Formulaire de Réponses Devoir 2

Evaluation environnementale des systèmes agricoles
urbains en Afrique de l’Ouest : Implications de la

diversité des pratiques et de la variabilité des émissions
d’azote dans l’Analyse du Cycle de Vie de la tomate au
Bénin
Aurélie Perrin
To cite this version:

Aurélie Perrin. Evaluation environnementale des systèmes agricoles urbains en Afrique de l’Ouest :
Implications de la diversité des pratiques et de la variabilité des émissions d’azote dans l’Analyse du
Cycle de Vie de la tomate au Bénin. Sciences agricoles. AgroParisTech, 2013. Français. �NNT :
2013AGPT0080�. �tel-01168946�
HAL Id: tel-01168946

https://pastel.hal.science/tel-01168946
Submitted on 26 Jun 2015
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Doctorat ParisTech
THÈSE
pour obtenir le grade de docteur délivré par
L’Institut des Sciences et Industries

du Vivant et de l’Environnement
(AgroParisTech)
Spécialité : Sciences agronomiques et écologiques
présentée et soutenue publiquement par
Aurélie PERRIN
le 20 décembre 2013
Evaluation environnementale des systèmes agricoles urbains en Afrique de

l'Ouest: Implications de la diversité des pratiques et de la variabilité des
émissions d'azote dans l'Analyse du Cycle de Vie de la tomate au Bénin
Directeur de thèse : Benoit GABRIELLE
Co-encadrement de la thèse : Claudine BASSET-MENS
Jury
M. Ralph ROSENBAUM, Associate professor, Technical University of Denmark Rapporteur
M. Sylvain PERRET, Associate professor, Asian Institute of Technology, Bangkok and Cirad Rapporteur
Mme Véronique BELLON-MAUREL, Ingénieur en Chef des Ponts, des Eaux et
des Chaussées, HDR, IRSTEA Examinatrice
M. Christophe KOUAME, Directeur de recherche, Centre International de Recherche
en Agroforesterie, Cote d’Ivoire Examinateur
M. Benoit GABRIELLE, Professeur AgroParisTech, UMR EGC, Grignon Examinateur
Mme Claudine BASSET-MENS, Docteur, UMR Hortsys, Cirad Examinatrice
Mme Virginie PARNAUDEAU, Docteur, UMR SAS, INRA Invitée
UR Fonctionnement agro ecologique et performances des systèmes de culture horticoles

TA B-103/PS4 – Boulevard de la Lironde – 34 389 Montpellier Cedex 5
Doctorat ParisTech
THESIS
for the degree of doctor delivered by
Paris Institute of Technology for Life,

Food and Environmental Sciences
(AgroParisTech)
Major: Agronomic and ecological sciences
présentée et soutenue publiquement par
Aurélie PERRIN
le 20 décembre 2013
Environmental assessment of urban agricultural systems in West Africa:

Implications of the diversity of practices and the variability of nitrogen
emissions for the Life Cycle Assessment of tomato from Benin
Supervisor : Benoit GABRIELLE

Co-supervisor : Claudine BASSET-MENS
Jury
M. Ralph ROSENBAUM, Associate professor, Technical University of Denmark Rapporteur
M. Sylvain PERRET, Associate professor, Asian Institute of Technology, Bangkok and Cirad Rapporteur
Mme Véronique BELLON-MAUREL, Ingénieur en Chef des Ponts, des Eaux et
des Chaussées, HDR, IRSTEA Examinatrice
M. Christophe KOUAME, Directeur de recherche, Centre International de Recherche
en Agroforesterie, Cote d’Ivoire Examinateur
M. Benoit GABRIELLE, Professeur AgroParisTech, UMR EGC, Grignon Examinateur
Mme Claudine BASSET-MENS, Docteur, UMR Hortsys, Cirad Examinatrice
Mme Virginie PARNAUDEAU, Docteur, UMR SAS, INRA Invitée
UR Fonctionnement agro ecologique et performances des systèmes de culture horticoles

TA B-103/PS4 – Boulevard de la Lironde – 34 389 Montpellier Cedex 5
Remerciements
A l’occasion de cette fin de thèse je souhaiterais remercier tous ceux qui m’ont entourée avant et
pendant ces trois années mémorables. Je remercie d’abord mon équipe encadrante de choc et mes
collègues expatriés au Bénin pour leur aide incommensurable. Et c’est peu dire pour quelqu’un qui
s’est attachée à tout mesurer pendant 3 ans. Il y a aussi eu des tonnes de discussions intéressantes avec
des collègues de tous horizons et énormément de soutien de la part des amis et de la famille. Alors
merci à tous !
Table des matières
Liste des figures ................................................................................................ 1
Liste des tableaux ............................................................................................. 3
Liste des abréviations ....................................................................................... 5
Introduction générale ....................................................................................... 7

Nourrir la planète : un enjeu global et des solutions locales .................................................7
Les méthodes d’évaluation de l’agriculture .........................................................................8
ACV et qualité des données d’inventaire.............................................................................9
Evaluation environnementale de la tomate produite en agriculture urbaine au Bénin ......... 12
Contexte et objectifs de la thèse ........................................................................................ 14
Plan de la thèse et démarche scientifique ..........................................................................15
Références bibliographiques ............................................................................................. 17
Chapitres ........................................................................................................ 21
Chapitre 1. Analyse du cycle de vie des systèmes de culture maraichers: une revue centrée
sur la diversité des systèmes de culture et l’estimation des émissions au champ .................... 21
Avant-propos............................................................................................................................... 22
Résumé ........................................................................................................................................ 23
Life cycle assessment of vegetable products: a review focusing on cropping systems diversity
and the estimation of field emissions .......................................................................................... 25
Abstract ....................................................................................................................................... 26
1.2. Materials and methods ..................................................................................................... 28
1.2.2. Typology of cropping systems and aggregation of results .............................................. 28
1.3. Presentation of LCA studies selected ............................................................................... 29
1.3.1. Goal and scope .............................................................................................................. 29
1.3.2. Data and methods for LCI ............................................................................................. 31
1.3.3. Input flows and yields ................................................................................................... 34
1.3.4. LCIA results ................................................................................................................. 37
1.4. Relevance of the typology to compare LCIA results from a diversity of cropping systems
and vegetable products .............................................................................................................. 41
1.4.1. Inherent variability of cropping systems and representativeness issues ........................... 42
1.4.2. Discrepancies in methodological choices ....................................................................... 42
1.4.3. Weaknesses in reporting, a limitation to our typology approach ..................................... 44
1.4.4. Relevance and validity of the methods used for the estimation of reactive nitrogen
emissions .................................................................................................................................. 45
1.5. Conclusions and recommandations .................................................................................. 48
1.6. References ....................................................................................................................... 52
Contribution du chapitre 1 et perspectives dans le cadre de la thèse ........................................ 55
Chapitre 2. Méthodes et données pour une analyse du cycle de vie fiable et représentative
dans un contexte de faible disponibilité en données .............................................................. 57
Avant-propos............................................................................................................................... 58
Résumé ........................................................................................................................................ 59
Life Cycle Inventory for tomato urban gardens in Benin: low performances and high
environmental risks systems ....................................................................................................... 61
Abstract ....................................................................................................................................... 61
2.1. Introduction ..................................................................................................................... 62
2.2.1. Study sites and characterization of cropping systems ..................................................... 63
2.2.2. Sampling cropping systems for monitoring .................................................................... 64
2.2.3. Cropping systems data for environmental inventories .................................................... 65
2.2.4. Comparison with scenarios based on the literature ......................................................... 66
2.2.5. Assessment of the quality of LCI data ........................................................................... 67
2.3. Results and discussion ..................................................................................................... 68
2.3.1. Diversity of cropping systems sampled .......................................................................... 68
2.3.2. Crop protection: frequent applications of pesticides at excessive rates ........................... 71
2.3.3. Intensive fertilization management: important N and P surplus ...................................... 73
2.3.4. Variable and low tomato yields ..................................................................................... 77
2.3.5. Prospects for the Life cycle assessment of out-of season tomato in Benin ...................... 79
2.4. Conclusions ..................................................................................................................... 80
2.5. References ....................................................................................................................... 81
Contribution du chapitre 2 et perspectives dans le cadre de la thèse ........................................ 83
Chapitre 3. Boucler le bilan d’azote à l’échelle du système de culture : perspectives pour

l’estimation des émissions d’azote réactif pour la tomate en maraîchage urbain au Bénin ..... 85
Avant-Propos .............................................................................................................................. 86
Résumé ........................................................................................................................................ 87
Closing the nitrogen budget for tomato cropping systems in African urban gardens .............. 89
3.1. Introduction ..................................................................................................................... 89
3.2.1. An approach combining a crop model and a nitrogen budget ......................................... 91
3.2.2. Calculation of Nr emissions in STICS ........................................................................... 93
3.2.3. Parameterization of the model for the case study ........................................................... 96
3.2.4. Statistical analysis ......................................................................................................... 99
3.3. Results and discussion ................................................................................................... 100
3.3.1. Quality of prediction and model soundness for yields .................................................. 100
3.3.2. Analysis of drivers for nitrogen losses as simulated by STICS ..................................... 102
3.3.3. Evaluation of the combined approach .......................................................................... 106
3.4. Conclusion .................................................................................................................... 112
3.5. Reference ...................................................................................................................... 113
Contribution du chapitre 3 et perspectives dans le cadre de la thèse ...................................... 117
Chapitre 4. Exploration de la variabilité des impacts environnementaux calculés par
l’analyse du cycle de vie de la tomate en maraîchage urbain au Bénin ................................ 119
Avant-Propos ............................................................................................................................ 120
Résumé ...................................................................................................................................... 121
Investigation of the variability of life-cycle assessment of tomato production from Beninese
urban garden ............................................................................................................................. 123
4.1. Introduction ................................................................................................................... 123
4.2. Material and method ...................................................................................................... 124
4.2.1. Goal and scope definition ............................................................................................ 124
4.2.2. Inventory data ............................................................................................................. 127
4.2.3. Design of and comparison with literature LCIA reference scenarios ............................ 132
4.2.4. Design and analysis of simulated cropping system scenarios with the crop model STICS
. 133
4.3. Results .......................................................................................................................... 135
4.3.1. LCIA results across the modalities and the weighted mean .......................................... 135
4.4. Discussion ..................................................................................................................... 148
4.4.1. Impact of the irrigation management ........................................................................... 148
4.4.2. Impact of fertilization management ............................................................................. 148
4.4.3. Impact of crop protection management ........................................................................ 149
4.5. Conclusions ................................................................................................................... 150
4.6. References ..................................................................................................................... 151
Chapitre 5. Discussion générale........................................................................................ 155

5.1. Introduction ................................................................................................................... 156
5.2. Une approche de type méta-analyse pour capitaliser sur le matériau bibliographique ...... 159
5.3. Contribution à la réalisation d’inventaires environnementaux ......................................... 161
5.3.1. Proposition d’un protocole de collecte des données pour réaliser des inventaires
représentatifs de la diversité des systèmes de production ......................................................... 161
5.3.2. Proposition d’une approche intégrée pour l’analyse de la variabilité des émissions ...... 163
5.4. Contribution à l’évaluation environnementale de l’agriculture urbaine par la méthode ACV
. 165
5.4.1. Impact environnemental de la tomate produite en maraîchage urbain au Bénin ............ 165
5.4.2. Résultats de l’analyse de scénario : vers un raffinement des méthodes d’estimation au
champ 166
5.4.3. Identification des “hot-spots” à l’échelle de la population ............................................ 167
5.5. Recommandations et perspectives pour améliorer la robustesse des résultats d’ACV en
contexte tropical ..................................................................................................................... 169
5.5.1. Intégrer les aspects socio-économiques ........................................................................ 169
5.5.2. Collecter des données spécifiques sur les flux à la parcelle .......................................... 170
5.5.3. Mettre à disposition des méthodes et des outils adaptés................................................ 170
5.5.4. Améliorer la prise en compte des pesticides, des engrais organiques et de la
consommation en eau .............................................................................................................. 171
5.6. Conclusions ................................................................................................................... 173
5.7. Références bibliographiques .......................................................................................... 174
ANNEXES .................................................................................................... 177
Annex 1. Global warming potential (GWP), acidification potential (AP) and eutrophication
potential (EP) expressed per square meter for a vegetable crop cycle from cradle-to-harvest
for the 72 systems aggregated per cropping system type, product group and reviewed paper.
....................................................................................................................................... 179
Annexe 2. Liste des producteurs de tomate de contre saison pour la saison 2011-2012 sur
les communes de Cotonou, Ouidah et Grand-Popo.......................................................... 180
Annexe 3. Protocole d’enquête pour la collecte des données qualitatives par les conseillers
auprès des 12 agriculteurs sélectionnés ........................................................................... 184
Annexe 4. Liste des producteurs de tomate de contre saison pour la saison 2011-2012 sur
les communes de Cotonou, Ouidah et Grand-Popo.......................................................... 188
Annexe 5. Exemple des données complètes collectées pour l’itinéraire technique pour la
parcelle 6 ........................................................................................................................ 194
Annexe 6. Résultats des analyses de sol pour les 12 parcelles échantillonnées ................ 196
Annexe 7. Résumé des données climatiques produites par la station météo du centre de
recherche AfricaRice localisé à Cotonou......................................................................... 197
Annex 8. LCIA results from cradle-to-harvest per hectare of out-of-season tomato in Benin
for the 10 modalities accounting for 4 criteria, and the weighted mean............................ 199
Portefeuille de Compétences ..................................................................................................... 199
Liste des figures
Introduction
Figure 1 Relations entre les différentes étapes du processus d'analyse du cycle de vie (ISO, 2006) .... 10
Figure 2 Plan de la thèse et démarche scientifique ............................................................................ 15
Chapitre 1
Figure 1.1 Global warming potential (GWP), acidification potential (AP) and eutrophication potential
(EP) expressed per kg of fresh yield from cradle-to-harvest for the 72 systems aggregated per cropping
system type, product group and reviewed paper.. ............................................................................... 40
Figure 1.2 Résultats du chapitre 1 (boite verte) et contributions pour les chapitres suivants (rouge)... 55
Chapitre 2
Figure 2.1 Densité de population au Bénin (Center for International Earth Science Information
Network) .......................................................................................................................................... 58
Figure 2.2 The typology-based procedure to (1) sample 12 diverse cropping systems and (2) create a
representative inventory for out-of-season tomato cropping systems in studied urban areas in Benin
(2011-2012 - Cotonou, Ouidah, Grand Popo).. .................................................................................. 69
Figure 2.3 Treatment frequency index (TFI) per targeted pest: insect, disease and nematod for each
active substance used for crop protection, for cropping systems aggregated by modality for each
criterion: irrigation system, water resource, planting date and location and for the scenario
representative for the whole population (weighted mean). ................................................................. 72
Figure 2.4 Nitrogen and phosphorus inputs for main nutrient horizontal fluxes at field scale; for
cropping system aggregated by modality for each criterion: irrigation system, water resource, planting
date and location and for the scenario representative for the whole population (weighted mean) ........ 74
Figure 2.5 Nitrogen and phosphorus outputs for main nutrient fluxes at field scale for cropping
systems aggregated by modality for each criterion: irrigation system, water resource, planting date and
location and for the scenario representative for the whole population (weighted mean) ...................... 75
Figure 2.6 Nitrogen and phosphorus surplus at field scale for cropping systems aggregated by
modality for each criterion: irrigation system, water resource, planting date and location and for the
scenario representative for the whole population (weighted mean)..................................................... 77
Figure 2.7 Résultats du chapitre 2 (boite violette) et contribution pour les chapitres suivants (rouge) 83
Chapitre 3
Figure 3.1 Conceptual model for output variables of interest (circles) as simulated in STICS: main
influencing processes (white rectangle) and driving parameters (grey rectangles) .............................. 92
Figure 3.2 Effect of soil pH on volatilization from mineral fertilizers ............................................... 94
Figure 3.3 Effect of dry matter content of fertilizer and soil tillage on the volatilizable fraction ........ 94
Figure 3.4 Effects of pH (FPHN), Soil relative water content (FHN) and soil temperature (FTN) on
nitrification ....................................................................................................................................... 95
Figure 3.5 Effects of soil temperature (FDENT), soil nitrate content (FDENNO3) and soil properties
(FDENW) on denitrification ............................................................................................................. 96
Figure 3.6 Summary of climate variables for the 2011/2012 dry season at Cotonou, Benin, West
Africa ............................................................................................................................................... 97
Figure 3.7 Simulated versus observed yields .................................................................................. 100
Figure 3.8 Yield residues versus observed yields ............................................................................ 100
Figure 3.9 Simulated and observed yields versus cumulative global radiation over the crop cycle ... 101
Figure 3.10 Simulated and observed yields versus cumulative temperature over the crop cycle ....... 101
Figure 3.11 Simulated and observed yields versus total water ......................................................... 101
Figure 3.12 Simulated and observed yields versus nitrogen inputs .................................................. 101
Figure 3.13 Simulated and observed yields versus plant density ..................................................... 101
1
Figure 3.14 Total Nr emissions versus nitrogen input from mineralization of soil organic matter,
organic and mineral fertilizers ......................................................................................................... 102
Figure 3.15 NH4 emissions factors versus soil pH ........................................................................... 103
Figure 3.16 NH4 emissions factors for mineral fertilizers versus total water .................................... 103
Figure 3.17 N2O emission factors versus soil moisture at maximal capacity for all cropping systems
....................................................................................................................................................... 104
Figure 3.19 N2O emission factors for nitrification versus soil pH .................................................... 104
Figure 3.18 N2O emission factors versus soil moisture at maximal capacity for sandy soils ............ 104
Figure 3.20 NO3 emissions versus the total nitrified nitrogen .......................................................... 105
Figure 3.21 NO3 emissions versus total volume of water ................................................................ 105
Figure 3.22 Nitrogen budget simulated with STICS for the 12 cropping systems ............................ 106
Figure 3.23 Nitrogen input from soil mineralization versus organic fertilizers................................. 107
Figure 3.24 Simulated harvested nitrogen versus simulated yields .................................................. 107
Figure 3.25 Résultats du chapitre 3 (boite violette) et contribution pour les chapitres suivants (rouge)
....................................................................................................................................................... 117
Chapitre 4
Figure 4.1 Les contributions des chapitres précédents (rouge) pour répondre à l’objectif 4 .............. 120
Figure 4.2 LCIA results from cradle-to-harvest per kilogram of out-of-season tomato in Benin of the
10 modalities accounting for 4 criteria, and the weighted mean for 5 impacts categories: water
deprivation potential (WDP), climate change potential (CCP), human toxicity potential (HTP),
photochemical ozone formation potential (POFP) and marine ecotoxicity potential (METP)............ 137
10 modalities accounting for 4 criteria, and the weighted mean; for 3 impacts categories: terrestrial
acidification potential (TAP), freshwater and marine eutrophication potential (FEP and MEP) ........ 139
10 modalities accounting for 4 criteria, and the weighted mean; for 2 impacts categories: terrestrial and
freshwater ecotoxicity potential (TETP and FETP) .......................................................................... 141
Figure 4.5 Scenario analysis for climate change potential (CCP), terrestrial acidification potential
(TAP) and marine eutrophication potential (MEP) to minimum (Min) and maximum (Max) scenarios
for 4 drivers influencing Nr emissions: nitrogen rate, water volume, soil pH and maximum soil
moisture.......................................................................................................................................... 147
Chapitre 5
Figure 5.1 Principaux résultats de la thèse et progression de la discussion générale ......................... 158
2
Liste des tableaux
Chapitre 1
Table 1.1 LCI methods to estimate reactive nitrogen field emissions: guidelines and references,
calculation method using emission factors and parameters, validity domain and uses in selected studies
......................................................................................................................................................... 33
Table 1.2 Inputs flows for the field production stage expressed per ton of fresh yield and fresh yield
for the 72 systems aggregated by cropping system type, product group and reference. ....................... 36
Table 1.3 Summary of recommendations related to key parameters to be included in LCI and methods
to be used for the estimation of reactive nitrogen (Nr) emissions to account for specificities of
vegetable cropping systems ............................................................................................................... 49
Chapitre 2
Table 2.1 Distribution of the 10 types at the whole population scale (typology based on the irrigation
system, the water resource and the planting date) in terms of area (m²) and number of farms for each
district .............................................................................................................................................. 70
Table 2.2 Average number of applications and associated coefficient of variation per targeted pest:
insect, disease and nematod for cropping systems aggregated by modality for each criterion: irrigation
system, water resource and planting date and for the scenario representative for the whole population
(WM). .............................................................................................................................................. 71
Table 2.3 Average and range of variation for emission factors for the scenario representative
(weighted mean) where emissions were estimated based on the literature and for the Rotation scenario
where emissions were measured on field (18-month period) .............................................................. 76
Table 2.4 Cropping system data over one tomato crop cycle for cropping systems aggregated by
modality for each criterion: irrigation system, water resource, planting date and location and for the
scenario representative for the whole population (WM). .................................................................... 78
Table 2.5 Data quality indicators (Weidema and Wesnaes, 1996) for our LCI data set for out-of-
season tomato in Benin. The best score is 1, the worst is 5................................................................. 80
Chapitre 3
Tableau 3.1 Facteurs influençant la variation des émissions azotées au champ pour les cultures
maraichères....................................................................................................................................... 86
Table 3.2 Site-specific soil data used to feed the STICS model from 12 tomato cropping systems in
Benin ................................................................................................................................................ 97
Table 3.3 Management data used to feed the STICS model from 12 tomato cropping systems in Benin
......................................................................................................................................................... 98
Table 3.4 General parameters used to adapt the STICS crop model to tropical conditions ................. 99
Table 3.5 Performances of the structured estimation methods for Nr emissions ............................... 111
Chapitre 4
Table 4.1 Weighting factors based on area distribution used to calculate the weighted mean and
subsequent contribution of modalities within each criterion ............................................................. 126
Table 4.2 Parameters used for the inventory modeling of production processes per modality of each
criterion and for the weighted mean (WM) ...................................................................................... 129
Table 4. 3 Agricultural input data per modality of each criterion and for the weighted mean (WM). 131
Table 4.4 Characteristics for literature vegetable scenarios and Beninese out-of-season tomato ....... 132
Table 4.5 Key influencing drivers tested and corresponding favorable and unfavorable scenarios to
emissions for the scenario analysis of LCIA results ......................................................................... 134
Table 4.6 LCIA results per kg of product for WDP, global warming, acidification and eutrophication
potentials for literature vegetable scenarios and Beninese out-of-season tomato (this study) ............ 143
Table 4.7 Nr emission factors obtained with the Default Emission Factors (DEF) approach (calculated
for the weighted mean only) and estimated with the crop model method (CM) for the 8 agronomic
scenarios, minimum (Min) and maximum (Max) ............................................................................. 144
3
4
Liste des abréviations
ACV Analyse du Cycle de Vie / Life cycle Assessment (LCA)
AfricaRice International agricultural research center for research on rice in Africa (Cotonou, Bénin)
AP Acidification potential / Potentiel d’acidification
CEC Capacité d’Echange Cationique (propriété du sol)
Cirad Centre de coopération internationale en recherche agronomique pour le développement
CM Crop Model (see method for Nr emission estimation) / Modèle biophysique
CML Instititute of Environmental Sciences (Leiden, The Netherlands)
CS Cropping System / Système de culture (SdC)
CV Coefficient of Variation
DEF Default Method (see method for Nr emission estimation)
ECETOC European Center for Ecotoxicology and Technology of Chemicals
EF Emission Factor / Facteur d’émissions
EP Eutrophication Potential / Potentiel d’eutrophisation
ETP Evapotranspiration
FAO Food and Agriculture Organization of the United Nations
FC Field Capacity / Capacité au champ (propriété du sol)
GHG Greenhouse Gases / Gaz à effet de serre
GWP Global Warming Potential/ Potentiel de réchauffement climatique
ICV Inventaire du cycle de vie / Life cycle Inventory (LCI)
IFT Indice de fréquence de Traitement / Treatment Frequency Index (TFI)
ILCD International reference Life Cycle Data system (JRC, European Comission)
INRA Institut National de Recherche Agronomique
INRAB Institut National de Recherche Agronomique au Bénin
IPCC International Panel for Climate Change
ISO International Organization for Standardization
LCIA Life Cycle Impact Assessment
NP Engrais multiple à base d’azote et de phosphore
NPK Engrais multiple à base d’azote, de phosphore et de potassium
Nr Reactive nitrogen / Azote réactif
OECD Organisation for Economic Co-operation and Development
SETAC Society of Environmental Toxicology and Chemistry
STICS Simulateur mulTIdisciplinaire pour les Cultures Standard (modèle biophysique)
UK United Kingdom
WDP Water Deprivation Potential / Potentiel de privation en eau
WFPS Water-Filled Pore Space
WM Weighted mean / Moyenne pondérée
WSI Water Stress Index / Indice de stress hydrique
5
6
- Introduction générale -
Introduction générale
© Césaire Dakpogan
Circulation urbaine et pollution à Cotonou, Bénin
Nourrir la planète : un enjeu global et des solutions locales

D’ici 2050, la population du globe devrait dépasser 9 milliards de personnes, soit environ deux
milliards de plus qu'en 2010 (United Nations, 2012). Une augmentation de la production agricole est
cruciale pour le maintien de la stabilité économique et sociale à l’échelle mondiale (Tilman et al.,
2002). Dans les années 60-90, la révolution verte à grand renfort d’intrants a permis d’augmenter les
rendements, mais l’agriculture d’aujourd’hui est dépendante de nombreuses ressources non
renouvelables telles que les carburants fossiles, les minerais et l’eau et son empreinte écologique
(dégradation des sols, pollution de l’eau, réduction de la biodiversité) a atteint un niveau sans
précédent. Pourtant, les écosystèmes fournissent des services indispensables à la survie des
populations les plus vulnérables et leur développement doit aller de pair avec la préservation des
ressources naturelles et de l’environnement (Millennium Ecosystem Assessment, 2005).
L’amélioration de la sécurité alimentaire doit donc passer par le développement durable de
l’agriculture mondiale en général et dans les pays en développement en particulier. Malgré des progrès
récents prometteurs (Pretty et al., 2003), la transition vers des systèmes de production plus durables
dans les pays en développement reste un champ de recherche peu exploré, qui doit couvrir les produits
de base, les produits d’origine animale et aussi les fruits et les légumes.
7
Dans les pays en développement, le maraichage est présent sous trois formes : (1) les cultures de rente
destinées à l’exportation vers l’Europe comme le haricot du Kenya ou la tomate du Sénégal, (2) les
cultures extensives en zone rurale pendant la saison des pluies et (3) les cultures intensives en zone
urbaine et périurbaine tout au long de l’année. D’ici 2030, 60% de la population mondiale vivra en
zone urbaine (United Nations, 2012). Dans ce contexte, le développement de l’agriculture urbaine
représente un levier important pour le développement durable des villes dans les pays en
développement en général, et en Afrique en particulier (FAO, 2012; World Bank, 2013). En effet,
l’agriculture urbaine apporte des réponses pour réduire la pauvreté, améliorer la sécurité alimentaire,
réduire la malnutrition et pallier les contraintes environnementales. Cependant, face à la demande
croissante des urbains pour le foncier, les systèmes de production en zone urbaine se sont
particulièrement intensifiés, mettant en évidence des risques environnementaux associés à ces modes
de production (De Bon et al., 2010).
Une évaluation fiable de l’ensemble des impacts environnementaux associés aux systèmes de
production maraichers en Afrique est aujourd’hui nécessaire.
Les méthodes d’évaluation de l’agriculture

Une grande diversité de méthodes a été développée pour l’évaluation de l’agriculture (van der Werf &
Petit, 2002; Payraudeau & van der Werf, 2005; Bockstaller et al., 2009). L’objectif général de ces
méthodes peut consister en l’évaluation environnementale stricte ou l’évaluation de la durabilité
(performances économiques, environnementales et sociales). Ces méthodes permettent d’évaluer une
variété d’impacts depuis l’échelle locale (odeur, destruction d’une zone protégée) jusqu’à l’échelle
globale (changement climatique). Les critères ou indicateurs d’impacts peuvent être quantitatifs ou
qualitatifs, simples ou multiples. Le type d’impact évalué peut être réel ou potentiel et être exprimé par
une unité de référence unique ou multiple. Chaque méthode adresse plus particulièrement un niveau
d’intégration spécifique de l’agriculture depuis la région jusqu’au système de culture en passant par la
filière et l’exploitation agricole. Enfin, la pertinence d’une méthode d’évaluation repose sur la
pertinence de ses indicateurs en lien avec les objectifs qu’elle s’est assignée. Il est ainsi préférable de
choisir des indicateurs basés sur l’effet des pratiques (résultats) plutôt que sur la description des
pratiques (moyens) permettant ainsi aux gestionnaires d’atteindre librement leurs objectifs en fonction
de leur contexte de production (Payraudeau & van der Werf, 2005).
8
L’analyse des travaux sur les méthodes d’évaluation (Payraudeau & van der Werf, 2005) permet
d’aboutir à un certain nombre de recommandations pour l’évaluation environnementale de
l’agriculture que nous pouvons résumer ainsi :
 La prise en compte de l’ensemble des impacts environnementaux pertinents de type global ou

local pour les systèmes de production étudiés, ainsi que de l’ensemble des étapes de
production (site principal de production, production des intrants, commercialisation et
consommation) permet de limiter le risque de transfert d’impact,
 La possibilité d’exprimer les impacts en fonction de plusieurs unités de référence permet la
prise en compte de la multifonctionnalité des systèmes agricoles,
 L’évaluation quantifiée de l’effet des pratiques facilite la validation des résultats et permet une
agrégation transparente qui préserve l’information concernant les points faibles des systèmes,
 Les concepts et indicateurs impliqués dans les méthodes doivent être validés scientifiquement
et l’analyse de l’incertitude doit être intégrée dans le processus d’évaluation pour permettre
une interprétation scientifiquement correcte des résultats.
En suivant cette grille de recommandations, la méthode Analyse du cycle de vie a été identifiée
comme particulièrement pertinente pour l’évaluation environnementale des productions agricoles (van
der Werf & Petit, 2002).
ACV et qualité des données d’inventaire

L’analyse du cycle de vie (ACV) est une méthode d’évaluation standardisée (ISO 14044, 2006) qui
permet l’évaluation des impacts potentiels associés à un système de production. Ce système
correspond au cycle de vie du produit, et inclut l’ensemble des processus depuis l’extraction des
matières premières jusqu’à son utilisation ou sa destruction. On dit aussi « du berceau à la tombe ». La
méthode ACV définit 3 aires de protection prioritaires : les ressources naturelles, la santé humaine et
la santé des écosystèmes. A chaque zone de protection correspondent plusieurs catégories d’impact de
type global comme le changement climatique ou de type régional comme l’eutrophisation.
La mise en œuvre de l’ACV comprend 4 étapes définies par la norme ISO 14044 (Figure 1). La
première étape consiste à définir les objectifs et le champ de l’étude (ou périmètre). En pratique elle
consiste à identifier le ou les systèmes étudiés par rapport à une fonction et à sélectionner une ou
plusieurs unités de référence ou unités fonctionnelles grâce à laquelle l’ensemble des flux d’inventaire
(étape 2) et les impacts (étape 3) seront exprimés. Cette étape permet également de délimiter les
frontières du système et d’identifier les données et les méthodes à mettre en œuvre pour atteindre les
9
1. Objectif et périmètre objectifs de l’étude. L’étape suivante d’inventaire

de l’étude consiste à quantifier l’ensemble des flux entrants et
des flux sortants du système. Ces flux seront
2. Inventaire et analyse 4. Interprétation ensuite convertis en impacts (étape 3) par un
de l’inventaire des résultats
système d’équivalence entre les substances
inventoriées et les substances de référence
3. Evaluation de l’impact caractérisant chaque catégorie d’impact (le CO2
Figure 1 Relations entre les différentes étapes du pour le changement climatique). Enfin, la phase
processus d'analyse du cycle de vie (ISO, 2006)
d’interprétation consiste à évaluer la valeur et la
robustesse de tous les résultats, choix et hypothèses en relation avec les objectifs initiaux. La
méthodologie ACV prévoit une boucle itérative pour permettre le raffinement des étapes 1, 2 ou 3 si
les objectifs ne sont pas atteints.
Dans l’analyse du cycle de vie comme dans toute approche de modélisation, la qualité des données est
l’une des principales limites de l’application de la méthodologie. La variabilité des mesures entre les
systèmes de production, les divergences entre la bibliographie et les données de terrain et la sensibilité
des résultats aux choix méthodologiques constituent les principaux obstacles à la fiabilité et à la
comparabilité des résultats d’ACV (Coulon et al., 1997). Ces préoccupations sont au cœur des enjeux
scientifiques liés à l’application de l’ACV aux produits agricoles.
Empruntées au champ disciplinaire de l’agronomie, deux thématiques ont donné lieu à des
développements méthodologiques récents en ACV : la prise en compte de la diversité des systèmes de
production et de la variabilité des émissions.
Diversité des systèmes de production

La diversité des systèmes de production agricoles (ou fermes) résulte de décisions techniques et
stratégiques prises par les agriculteurs pour combiner leurs facteurs de production (moyens
disponibles) et leurs modes de production (objectifs de production). Chaque situation étant unique,
cette diversité est quasiment infinie. Or, la diversité de ses fermes est considérée comme l’une des
richesses d’une agriculture, comme un gage de sa capacité d’adaptation, et le maintien de cette
diversité est de plus en plus souvent évoqué comme un objectif en soi (Perrot & Landais, 1993).
La typologie a largement été utilisée pour décrire et structurer des populations de fermes pour
répondre à des besoins d’analyse et d’évaluation de l’agriculture. Historiquement, les méthodes de
typologie ont été développées et mises en œuvre (notamment en France) dans un objectif de
développement rural pour évaluer le fonctionnement des fermes, trouver des solutions aux problèmes
rencontrés et élaborer des recommandations adaptées (Capillon, 1993; Landais, 1998). Plus largement,
la typologie permet l’analyse d’une situation donnée définie d’un point de vue géographique (Köbrich
10
et al., 2003; Kristensen & Kristensen, 2004) et/ou technique (Tittonell et al., 2005; Maton et al., 2005;
Poussin et al., 2008). Les méthodes de typologie ont également été mises en œuvre pour concevoir des
systèmes de production innovants (Blazy et al., 2009) et analyser les déterminants de leur adoption
(Daskalopoulou & Petrou, 2002; Bidogeza et al., 2009). Enfin, les méthodes de typologie peuvent être
à la base d’approche ex-ante, pour analyser l’impact d’un changement de l’environnement biophysique
ou politique (Smit & Skinner, 2002; Andersen et al., 2007).
De manière générale, les méthodes de typologie consistent à élaborer des types pour analyser une
réalité complexe et ordonner des objets relativement similaires. Cette structuration plus ou moins
artificielle est basée sur des critères de distinction, qui découlent de l’identification des principales
causes explicatives de la diversité (Andersen et al., 2007). En pratique, l’objectif est de maximiser
l’homogénéité à l’intérieur d’un type particulier tout en maximisant l’hétérogénéité entre les types
(Köbrich et al., 2003). Une large gamme de méthodes a été développée : les approches les plus
qualitatives et supervisées (a priori) facilitent une analyse cognitive et fonctionnelle des fermes
d’autant plus qu’elles intègrent des notions de stratégie et de trajectoire temporelle. Ces approches
nécessitant des enquêtes approfondies et une analyse fastidieuse, sont souvent limitées à de petites
régions dans lesquelles l’agriculture est organisée. Les approches les plus quantitatives et automatisées
faisant appel aux méthodes statistiques (analyse en composante principale et classification
hiérarchique) permettent l’analyse de populations plus larges, pourvu qu’un grand nombre de données
soit disponible. Cependant ces approches sont très dépendantes de la représentativité de l’échantillon
et des variables sélectionnées. En pratique, les méthodes sont très souvent semi-supervisées de
manière plus ou moins générique, et incluent à la fois des variables quantitative et des indicateurs
qualitatifs.
L’approche par scénarisation développée par Basset-Mens et al. (2007) et l’approche de typologie de
fermes développée par Dalgaard et al. (2006) sont deux exemples de développement récents pour
améliorer la prise en compte de la diversité des systèmes de production dans le cadre de l’analyse du
cycle de vie. Les deux approches se basent sur des données mixtes issues de bases de données et
d’expertise. L’approche de Basset-Mens et al. (2007) distingue les systèmes sur la base du mode de
production (quels objectifs de production ?) alors que l’approche de Dalgaard et al. (2006) distingue
les systèmes sur la base des facteurs de production (quels moyens mis en œuvre ?).
Variabilité des émissions au champ

Les études relatives aux émissions gazeuses (NH 3, N2O, NOx, N2) s’accordent sur la grande variabilité
des émissions en fonction des pratiques et du contexte pédoclimatique, surtout à très petites échelles
(parcelle). Par exemple, Bussink et Oenema (1998) ont montré que le coefficient de variation associé
aux émissions de NH3 associées à l’application de fumier peut atteindre 100%. Une analyse
bibliographique portant sur des données mesurées et la modélisation des émissions de N 2O annuelles à
11
l’échelle globale (Stehfest & Bouwman, 2006) a conclu qu’aucune amélioration n’avait été réalisée
récemment concernant l'évaluation de N 2O à l’échelle globale. Un grand nombre de facteurs influant
sur les émissions d’azote réactif (NH 3, N2O, NOx, NO3) a été mis en évidence dans la littérature
spécifique à ces flux (Del Grosso et al., 2000; Sommer & Hutchings, 2001; Milroy et al., 2008). Ces
facteurs d’influence résultent de la combinaison entre les caractéristiques de sol, de climat et de
pratiques agricoles des systèmes de production et des processus biophysiques, ayant un effet direct sur
les émissions.
Contrairement aux systèmes de production clos comme les usines pour lesquels l’ACV a initialement
été développée, il est difficile de mesurer les émissions à la sortie des processus de production en
agriculture. D’une part parce que ces émissions sont majoritairement de nature diffuse dans le temps et
l’espace et d’autre part parce qu’elles résultent de processus biologiques complexes. Dans le champ de
l’ACV, le recours aux modèles stochastiques (Payraudeau et al., 2007) ou biophysiques (Gabrielle &
Gagnaire, 2008; Langevin, 2010; Bessou et al., 2013; Dufossé et al., 2013) sont deux exemples de
développement récents pour améliorer la prise en compte de la variabilité des émissions d’azote
associées à l’usage d’engrais dans les études ACV.
Evaluation environnementale de la tomate produite en agriculture urbaine au Bénin

L’horticulture urbaine est une des activités pratiquées par les agriculteurs urbains. Au même titre que
l’agriculture urbaine elle se caractérise par une recherche de valorisation forte des facteurs de
production dans un contexte de forte pression foncière, de besoin en main d’œuvre, d’accès limité aux
intrants (De Bon et al., 2010). Les cultures horticoles en général et la tomate en particulier sont très
sensibles aux aléas climatiques et à la pression phytosanitaire. Elles nécessitent des investissements
importants pour les infrastructures d’irrigation et les intrants. A cause du faible niveau d’éducation des
producteurs et de leur accès limité au soutien technique, les producteurs ont tendance à apporter des
intrants de manière excessive par rapport à leurs niveaux de rendement. Les perspectives de
développement de l’horticulture urbaine soulèvent l’existence de risques environnementaux associés à
un usage mal maitrisé des intrants (Orsini et al., 2013). Plusieurs études ont notamment relevé des
pratiques de fertilisation intensive entrainant des surplus azotés importants (Diogo et al., 2010;
Predotova et al., 2010; Predotova et al., 2011). D’autres études ont décrit des pratiques de protection
des cultures intensives, avec un usage excessif de pesticides (Cissé et al., 2003; Ahouangninou et al.,
2012).
Le cas de la tomate en jardin urbain au Bénin est particulièrement illustratif de cette problématique. Le
Bénin présente une densité de population deux fois plus élevée par rapport à la moyenne africaine (87
vs. 37 hab.km-²). La population croit deux fois plus vite en zone urbaine qu’en zone rurale (4 vs.2%)
et 6% de la population est en situation de malnutrition. La tomate est la première production horticole
du pays (FAO, 2013). Malgré tout, les villes sont obligées d’importer des tomates en saison sèche et
12
cela malgré les 400 ha additionnels alloués récemment à l’agriculture urbaine dans les communes de
Cotonou et Sèmè-Kpodji (FAO, 2012). La principale raison évoquée par la littérature grise pour
expliquer ce défaut de production est le manque de maitrise technique pour faire face à des conditions
climatiques chaudes et humides et à des attaques de ravageurs en tout genre. Ainsi, le développement
de l’horticulture urbaine en général et de la production de tomate en jardin urbain au Bénin en
particulier soulève de nouvelles questions agronomiques et environnementales. Des études et outils
d’évaluation environnementale fiables de ces systèmes sont nécessaires pour y répondre et
accompagner un développement durable de ces systèmes.
L’application de méthodologies de type ACV, très exigeantes en données, aux systèmes maraîchers
urbains, représente cependant un défi méthodologique important au regard des très faibles
disponibilités en données et en connaissances sur ces systèmes.
13
Contexte et objectifs de la thèse

Dans un contexte d'urbanisation croissante, les décideurs politiques, les scientifiques et les acteurs
locaux sont interpelés pour protéger et promouvoir l'agriculture urbaine (et péri-urbaine), parce qu'elle
contribue à la fois à satisfaire les besoins d'alimentation, d'emploi, et de traitement des déchets (De
Bon et al., 2010; Orsini et al., 2013). Pour concevoir des systèmes plus productifs, durables, économes
en intrants et moins nocifs pour l’environnement, il est nécessaire d’identifier des outils sur lesquels
s’appuyer pour faire des choix techniques pertinents. Cette thèse s’intègre dans un dispositif mis en
place par le Cirad depuis 2009 sur le thème de l’analyse de cycle de vie des produits tropicaux, qui
comprend une équipe de chercheurs spécialistes de ce nouveau champ de recherche, des outils de
calcul et des bases de données mutualisées.
La question scientifique posée dans cette thèse peut être formulée ainsi : comment prendre en compte
la variabilité pédoclimatique et technique dans la réalisation des inventaires environnementaux et ainsi
affiner les résultats d’analyse du cycle de vie des produits maraîchers en contexte tropical? Sur la base
de cette question, plusieurs objectifs sont formulés :
 Objectif 1 : réaliser l’état de l’art de l’ACV sur produits maraîchers afin de concevoir un
protocole d’étude adapté
 Objectif 2 : explorer les approches possibles pour mieux caractériser la diversité des systèmes
de culture au regard de leurs impacts environnementaux, et créer des inventaires
représentatifs,
 Objectif 3 : identifier les déterminants de la variabilité des émissions au champ et explorer les
approches possibles, telles que la modélisation, pour raffiner les méthodes d’estimation des
émissions azotées,
 Objectif 4 : évaluer l’impact environnemental de la tomate produite en jardins urbains au
Bénin et
 Objectif 5 : produire des recommandations pour améliorer la qualité des résultats d’ACV pour
les produits maraichers tropicaux.
14
Plan de la thèse et démarche scientifique

Ces objectifs nous ont amené à construire une démarche de travail cohérente, reprise dans la
structuration de la thèse (Figure 2), qui s’articule autour des 4 étapes de l’analyse du cycle de vie
présentées en figure 1. En préalable à notre propre contribution, une revue de la littérature doit nous
permettre d’identifier les grands enjeux méthodologiques de l’application de l’ACV aux produits
maraîchers (chapitre 1). L’étape de l’inventaire fera l’objet de développements méthodologiques
spécifiques pour répondre aux objectifs 2 (chapitre 2) et 3 (chapitre 3) explorant respectivement les
apports de la typologie puis de la modélisation biophysique. L’étape d’évaluation des impacts (objectif
4) sera construite autour de scénarios et présentée au chapitre 4. L’étape d’interprétation des résultats
sera approfondie d’abord sur la base des résultats d’ACV présentés dans les études publiées de notre
revue (chapitre 1) puis sur la base des résultats d’ACV pour le cas d’étude tomate Bénin (chapitre 5).
Enfin, des recommandations seront formulées (objectif 5) suite à l’analyse de l’ensemble de ces
tentatives exploratoires dans la discussion générale du document de thèse.
Objectif et périmètre de l’étude

 Tomates produites en jardin urbain au Bénin
OBJECTIF 2 OBJECTIF 3
Prise en compte de Identifier les

OBJECTIF 1 la diversité et déterminants de la
inventaires variabilité des OBJECTIFS
Etat de l’art représentatifs émissions azotées 1, 2, 3, 4 & 5
ACV des
produits Chapitre 2 Chapitre 3 Discussion et
maraîchers  Approche de  Modélisation perspectives
typologie biophysique
Chapitre 1 Chapitre 5
 Méta-analyse
OBJECTIF 4
Evaluer l’impact environnemental de la

tomate produite en jardins urbains au Bénin
Chapitre 4
 Analyse de scénarios
Figure 2 Plan de la thèse et démarche scientifique
Le chapitre 1 présente une revue de la littérature sur l’analyse du cycle de vie appliquée aux produits
maraîchers. Grâce à une approche de type méta-analyse, nous avons analysé la qualité des données et
des méthodes d’inventaire et leurs effets sur la variation des résultats d’ACV. Nous avons ainsi
15
formulé des recommandations pour la réalisation d’ACV plus fiables et plus comparables pour
l’évaluation environnementale des produits maraichers.
Fort de ce constat général, nous avons identifié un cas d’étude pertinent par rapport aux enjeux
identifiés pour notre travail et sur lequel mettre en œuvre nos recommandations : la tomate produite en
jardin urbain au Bénin. La méthodologie ACV est particulièrement pertinente pour aborder les
problématiques environnementales associées à l’intensification des systèmes de production en zone
urbaine et péri-urbaine et plus particulièrement à leur usage excessif de pesticides et d’engrais. De
plus, le manque de données et de connaissances sur ces systèmes accentue les défis scientifiques
associés à la qualité des inventaires.
Au chapitre 2 nous présentons une procédure de création d’inventaires représentatifs basée sur la
typologie. Cette approche permet de caractériser la diversité des systèmes de production d’un point de
vue agronomique et de quantifier les flux d’inventaires pour la réalisation d’ACV.
Le chapitre 3 présente notre démarche d’exploration de la variabilité des émissions d’azote au champ
reposant sur l’usage d’un modèle biophysique combiné à une approche de bilan. A partir des données
de terrain présentées dans le chapitre précédent, nous mettons en évidence les principaux facteurs
influents des émissions d’azote dans notre cas d’étude et discutons les performances et les limites de
ce type d’approche pour raffiner l’estimation des émissions.
Au chapitre 4, nous intégrons les principaux résultats des deux chapitres précédents pour réaliser la
première ACV détaillée de la tomate en maraichage urbain au Bénin. Sont présentés dans ce chapitre
une gamme de profils environnementaux représentatifs de la diversité des systèmes de culture étudiés.
Une analyse de sensibilité des résultats d’ACV à la variabilité des émissions d’azote est également
proposée sur la base de 8 scénarios agronomiques.
Enfin, le chapitre 5 est une discussion générale de l’ensemble des résultats de la thèse. Il s’articule
autour de la discussion des résultats et méthodes testées pour atteindre les quatre premiers objectifs de
la thèse et permet ultimement d’atteindre l’objectif 5 : en formulant des recommandations pour
améliorer l’évaluation environnementale des systèmes de production de tomate en maraîchage urbain
et plus généralement des produits horticoles tropicaux.
16
Références bibliographiques
Ahouangninou C, Martin T, Edorh P, Bio-Bangana S, Samuel O, St-Laurent L, Dion S, Fayomi B (2012)
Characterization of health and environmental risks of pesticide use in market-gardening in the rural city of
Tori-Bossito in Benin, West Africa. Journal of Environmental Protection 3:241–248. doi:
10.4236/jep.2012.33030
Andersen E, Elbersen B, Godeschalk F, Verhoog D (2007) Farm management indicators and farm typologies as
a basis for assessments in a changing policy environment. Journal of environmental management 82:353–
362. doi: 10.1016/j.jenvman.2006.04.021
Basset-Mens C, van der Werf HMG, Robin P, Morvan T, Hassouna M, Paillat JM, Vertes F (2007) Methods and
data for the environmental inventory of contrasting pig production systems. Journal of Cleaner Production
15:1395–1405. doi: 10.1016/j.jclepro.2006.03.009
Bessou C, Lehuger S, Gabrielle B, Mary B (2013) Using a crop model to account for the effects of local factors
on the LCA of sugar beet ethanol in Picardy region, France. The International Journal of Life Cycle
Assessment 18:24–36. doi: 10.1007/s11367-012-0457-0
Bidogeza JC, Berentsen PBM, Graaff J, Oude Lansink AGJM (2009) A typology of farm households for the
Umutara Province in Rwanda. Food Security 1:321–335. doi: 10.1007/s12571-009-0029-8
Björklund AE (2002) Survey of approaches to improve reliability in LCA. The International Journal of Life
Cycle Assessment 7:64–72. doi: 10.1007/BF02978849
Blazy J-MM, Ozier-Lafontaine H, Doré T, Thomas A, Wery J (2009) A methodological framework that accounts
for farm diversity in the prototyping of crop management systems. Application to banana-based systems in
Guadeloupe. Agricultural systems 101:30–41. doi: 10.1016/j.agsy.2009.02.004
Bockstaller C, Guichard L, Keichinger O, Girardin P, Galan M-B, Gaillard G (2009) Comparison of Methods to
Assess the Sustainability of Agricultural Systems: A Review. Agronomy for Sustainable Development 769–
784. doi: 10.1007/978-90-481-2666-8_47
De Bon H, Parrot L, Moustier P (2010) Sustainable urban agriculture in developing countries. A review.
Agronomy for Sustainable Development 30:21–32. doi: 10.1051/agro:2008062
Bussink DW, Oenema O (1998) Ammonia volatilization from dairy farming systems in temperate areas: a
review. Nutrient Cycling in Agroecosystems 51:19–33. doi: 10.1023/a:1009747109538
Capillon A (1993) Guide pour la réalisation d’une typologie régionale d'exploitations à l'usage des agronomes.
Thèse doctorat, Institut national agronomique Paris-Grignon
Cissé I, Tandia AA, Fall ST (2003) Usage incontrôlé des pesticides en agriculture périurbaine: cas de la zone des
Niayes au Sénégal. Cahiers Agricultures 12:181–186.
Coulon R, Camobreco V, Teulon H, Besnainou J (1997) Data quality and uncertainty in LCI. The International
Journal of Life Cycle Assessment 2:178–182. doi: 10.1007/BF02978816
Dalgaard R, Halberg N, Kristensen IS, Larsen I (2006) Modelling representative and coherent Danish farm types
based on farm accountancy data for use in environmental assessments. Agriculture, Ecosystems &
Environment 117:223–237. doi: 10.1016/j.agee.2006.04.002
Daskalopoulou I, Petrou A (2002) Utilising a farm typology to identify potential adopters of alternative farming
activities in Greek agriculture. Journal of Rural Studies 18:95–103. doi: 10.1016/s0743-0167(01)00027-4
Diogo RC, Buerkert A, Schlecht E (2010) Horizontal nutrient fluxes and food safety in urban and peri-urban
vegetable and millet cultivation of Niamey, Niger. Nutrient Cycling in Agroecosystems 87:81–102. doi:
10.1007/s10705-009-9315-2
Dufossé K, Gabrielle B, Drouet J-L, Bessou C (2013) Using Agroecosystem Modeling to Improve the Estimates
of N2O Emissions in the Life-Cycle Assessment of Biofuels. Waste and Biomass Valorization 4:593–606.
doi: 10.1007/s12649-012-9171-1
FAO (2012) Growing greener cities in Africa. Food and Agriculture Organization of the United Nations, Rome
FAO (2013) FAOSTAT: Crop yield per country for the year 2000. http://faostat3.fao.org.
Gabrielle B, Gagnaire N (2008) Life-cycle assessment of straw use in bio-ethanol production: A case study
based on biophysical modelling. Biomass and Bioenergy 32:431–441. doi: 10.1016/j.biombioe.2007.10.017
Del Grosso SJ, Parton WJ, Mosier AR, Ojima DS, Kulmala AE, Phongpan S (2000) General model for N2O and
N2 gas emissions from soils due to dentrification. Global Biogeochemical Cycles 14:1045–1060. doi:
10.1029/1999GB001225
Huijbregts MAJ (1998) Application of uncertainty and variability in LCA. Part II: Dealing with parameter
uncertainty and uncertainty due to choices in life cycle assessment. The International Journal of Life Cycle
Assessment 3:343–351. doi: 10.1007/bf02979345
ISO 14044 (2006) Environmental management-life cycle assessment-requirements and guidelines. International
Standards Organization, 14044:2006 (E)
Köbrich C, Rehman T, Khan M (2003) Typification of farming systems for constructing representative farm
models: two illustrations of the application of multi-variate analyses in Chile and Pakistan. Agricultural
Systems 76:141–157. doi: 10.1016/S0308-521X(02)00013-6
17
Kristensen IT, Kristensen IS (2004) Farm types as an alternative to detailed models in evaluation of agricultural
practise in a certain area. Management Information Systems 241–250.
Landais E (1998) Modelling farm diversity. Agricultural Systems 58:505–527. doi: 10.1016/S0308-
521X(98)00065-1
Langevin B (2010) Prise en compte de la variabilité des émissions au champ dans l’Analyse de Cycle de Vie des
systèmes agricoles. Application à l'épandage de lisier. Thèse doctorat, Arts et Métiers ParisTech
Maton L, Leenhardt D, Goulard M, Bergez J-EE (2005) Assessing the irrigation strategies over a wide
geographical area from structural data about farming systems. Agricultural systems 86:293–311. doi:
10.1016/j.agsy.2004.09.010
Millennium Ecosystem Assessment (2005) Ecosystems and human well-being. n. 5, World Health Organization,
Geneva, Switzerland
Milroy SP, Asseng S, Poole ML (2008) Systems analysis of wheat production on low water-holding soils in a
Mediterranean-type environment: II. Drainage and nitrate leaching. Field Crops Research 107:211–220. doi:
10.1016/j.fcr.2008.02.008
Orsini F, Kahane R, Nono-Womdim R, Gianquinto G (2013) Urban agriculture in the developing world: a
review. Agronomy for Sustainable Development 33:695–720. doi: 10.1007/s13593-013-0143-z
Payraudeau S, van der Werf HMG (2005) Environmental impact assessment for a farming region: a review of
methods. Agriculture, Ecosystems & Environment 107:1–19. doi: 10.1016/j.agee.2004.12.012
Payraudeau S, van der Werf HMG, Vertes F (2007) Analysis of the uncertainty associated with the estimation of
nitrogen losses from farming systems. Agricultural systems 94:416–430. doi: 10.1016/j.agsy.2006.11.014
Perrot C, Landais E (1993) Exploitations agricoles: pourquoi poursuivre la recherche sur les méthodes
typologiques? Cahiers de la Recherche Developpement 33:13–23.
Poussin JC, Imache A, Beji R, Le Grusse P, Benmihoub A (2008) Exploring regional irrigation water demand
using typologies of farms and production units: An example from Tunisia. Agricultural Water Management
95:973–983. doi: 10.1016/j.agwat.2008.04.001
Predotova M, Bischoff W-A, Buerkert A (2011) Mineral-nitrogen and phosphorus leaching from vegetable
gardens in Niamey, Niger. Journal of Plant Nutrition and Soil Science 174:47–55. doi:
10.1002/jpln.200900255
Predotova M, Gebauer J, Diogo RVC, Schlecht E, Buerkert A (2010) Emissions of ammonia, nitrous oxide and
carbon dioxide from urban gardens in Niamey, Niger. Field Crops Research 115:1–8. doi:
10.1016/j.fcr.2009.09.010
Pretty JN, Morison JIL, Hine RE (2003) Reducing food poverty by increasing agricultural sustainability in
developing countries. Agriculture, Ecosystems & Environment 95:217–234. doi: 10.1016/S0167-
8809(02)00087-7
Smit B, Skinner MW (2002) Adaptation options in agriculture to climate change: a typology. Mitigation and
Adaptation Strategies for Global Change 7:85–114. doi: 10.1023/A:1015862228270
Sommer SG, Hutchings NJ (2001) Ammonia emission from field applied manure and its reduction—invited
paper. European Journal of Agronomy 15:1–15. doi: 10.1016/S1161-0301(01)00112-5
Stehfest E, Bouwman L (2006) N2O and NO emission from agricultural fields and soils under natural
vegetation: summarizing available measurement data and modeling of global annual emissions. Nutrient
Cycling in Agroecosystems 74:207–228. doi: 10.1007/s10705-006-9000-7
Tilman D, Cassman KG, Matson PA, Naylor R, Polasky S (2002) Agricultural sustainability and intensive
production practices. Nature 418:671–677. doi: 10.1038/nature01014
Tittonell P, Vanlauwe B, Leffelaar PA, Shepherd KD, Giller KE, Rowe EC (2005) Exploring diversity in soil
fertility management of smallholder farms in western Kenya: I. Heterogeneity at region and farm scale.
Agriculture, Ecosystems & Environment 110:149–165. doi: 10.1016/j.agee.2005.04.001
United Nations (2012) World Population Prospects: The 2012 Revision. http://esa.un.org/unpd/wpp/index.htm.
Weidema BP, Wesnæs MS (1996) Data quality management for life cycle inventories—an example of using data
quality indicators. Journal of Cleaner Production 4:167–174. doi: 10.1016/s0959-6526(96)00043-1
Van der Werf HMG, Petit J (2002) Evaluation of the environmental impact of agriculture at the farm level: a
comparison and analysis of 12 indicator-based methods. Agriculture, Ecosystems & Environment 93:131–
145. doi: 10.1016/S0167-8809(01)00354-1
World Bank (2013) Urban Agriculture - Findings from Four City Case Studies. n. 18, The World Bank Group,
Washington, USA
18
19
20
Chapitres
Chapitre 1. Analyse du cycle de vie des systèmes de culture maraichers: une

revue centrée sur la diversité des systèmes de culture et l’estimation des
émissions au champ
© Aurélie Perrin
Parcelle de choux dans un jardin urbain à Cotonou, Bénin
21
- Chapitre 1. Analyse du cycle de vie des systèmes de culture maraichers -
Avant-propos
Avant de se lancer dans la réalisation de l’ACV pour notre cas d’étude nous avons cherché à identifier
des précédents dans la littérature. Or au début de ce travail de thèse, la littérature ACV appliquée à des
produits maraîchers tropicaux se limitait à quelques cas de produits importés vers l’Europe et contenait
très peu de considérations méthodologiques. Parallèlement, la littérature ACV appliquée aux produits
maraîchers abordait une large gamme de systèmes de production mais les pratiques d’application de
l’ACV à ces produits étaient très variables et ne bénéficiaient pas de règles de bonnes pratiques
couvrant l’ensemble des particularités associées à ces systèmes de production.
A partir de ce constat, il était nécessaire de réaliser l’état de l’art des pratiques actuelles en ACV des
produits maraîchers pour identifier et évaluer les pratiques actuelles, les verrous méthodologiques, les
sources de variabilité et d’incertitude sur lesquels se baser pour construire notre propre protocole dans
notre cas d’étude mais aussi pour formuler des recommandations pour améliorer la qualité et la
comparabilité des études ACV des produits maraîchers. C’est l’objet du papier de revue de la
littérature qui suit et qui a été publié dans International Journal of LCA.
22
Résumé
Les récentes études d’Analyse de Cycle de Vie (ACV) des fruits et légumes mettent en évidence que la
phase de production agricole concentre une part prépondérante des impacts environnementaux de ces
produits. D’autre part, les systèmes de culture maraichers sont très divers selon les zones
géographiques et les systèmes de production. Ainsi, cette revue bibliographique s’attache à analyser
les méthodes et les données d’inventaires utilisées dans les différentes études ACV et à comparer les
impacts environnementaux potentiels de différentes productions maraîchères au stade de la ferme.
Dix études publiées dans des revues à comité de lecture ou dans des rapports répondant aux normes
ISO ont été sélectionnées. Nous avons d’abord analysé qualitativement les objectifs et le cadre des
différentes études. Ensuite, les résultats d’inventaires et les résultats d’impact pour trois catégories
(changement climatique, eutrophisation, acidification) ont été analysés au moyen d’une typologie des
systèmes de production, mais aussi par groupe de produit et par étude. Enfin, une attention particulière
a été portée sur les méthodes d’estimation des émissions d’azote réactif.
Parmi les différents systèmes de cultures identifiés, les systèmes sous serres chauffées présentent le
potentiel de changement climatique le plus élevé en raison de leur consommation d’énergie pour le
chauffage. Entre les différents groupes de produits, les haricots présentent les potentiels d’acidification
et d’eutrophisation les plus forts en raison de fortes émissions dues aux importantes quantités
d’engrais utilisées par tonne de produit. En analysant les résultats par étude, il ressort que les
principales sources de variation d’impact sont les rendements, les pratiques de fertilisation et les choix
méthodologiques relatifs aux règles d’allocation. Globalement, la comparabilité des études ACV pour
les productions maraîchères est restreinte par (1) le manque de transparence sur les objectifs et
périmètres des analyses, (2) le manque de données représentatives pour décrire les systèmes de
production et (3) l’hétérogénéité et le manque de spécificité des méthodes pour l’estimation des
émissions aux champs dans les conditions de production maraîchères. En particulier, il apparait que les
méthodes d’estimation des émissions d’azote sont souvent appliquées en dehors de leur domaine de
validité.
Cette première tentative de comparaison des impacts potentiels des productions maraichères a mis en
évidence plusieurs manques en termes de données et de méthodes dans la perspective de produire des
résultats représentatifs pour la phase de production agricole des produits maraîchers. Ainsi, une série
de recommandations a été formulée afin de mieux prendre en compte les spécificités des productions
maraichères, et ainsi d’améliorer la qualité des études ACV sur ces produits.
Mots clés : Stade ferme, maraîchage, typologie, méthodes d’estimation des émissions, azote réactif.
23
24
Life cycle assessment of vegetable products: a review focusing on cropping

systems diversity and the estimation of field emissions
Aurélie Perrin, Claudine Basset-Mens and Benoît Gabrielle
Aurélie Perrin () and Claudine Basset-Mens

CIRAD- UPR Hortsys, TA B-103/PS4, Boulevard de la Lironde
F-34398 Montpellier Cedex 5- France
 Tel : +33 4 67 61 71 49; Fax : +33 4 67 61 56 88; e-mail : aurelie.perrin@cirad.fr
Benoît Gabrielle
AgroParisTech, INRA, UMR 1091 Environnement et Grandes Cultures
F-78850 Grignon – France
Articled published in International Journal of Life Cycle Assessment

DOI 10.1007/s11367-014-0724-3
25
Abstract
Purpose Recent life cycle assessment studies for vegetable products have identified the agricultural
stage as one of the most important contributors to the environmental impacts for these products while
vegetable production systems are characterized by specific but also widely diverse production
conditions. In this context, a review aiming at comparing the potential impacts of vegetable products
and analysing the relevance of the methods and data used for the inventory of the farm stage appeared
necessary.
Method Ten papers published in peer-reviewed scientific journals or ISO-compliant reports were
selected. First, a presentation of the selected papers was done to compare the goal and scope and the
life cycle inventory data to the related sections in the ILCD Handbook. Second, a quantitative review
of input flows and Life Cycle Impact Assessment (LCIA) results (global warming, eutrophication and
acidification) was based on a cropping system typology and on a classification per product group.
Third, an in-depth analysis of the methods used to estimate field emissions of reactive nitrogen was
proposed.
Results The heated greenhouse system types showed the greatest global warming potential. The giant
bean group showed the greatest acidification and eutrophication potentials per kilogram of product
while, the tomato group showed the greatest acidification and eutrophication potentials per unit of
area. Main sources of variations for impacts across systems were yields and inputs variations and,
system expansion rules. Overall, the ability to compare the environmental impact for these diverse
vegetable products from cradle-to-harvest was hampered by (1) weaknesses regarding transparency of
goal and scope, (2) a lack of representativeness and completeness of data used for the field stage and
(3) heterogeneous and inadequate methods for estimating field emissions. In particular, methods to
estimate reactive nitrogen emissions were applied beyond their validity domain.
Conclusions and recommendations This first attempt at comparing the potential impacts of vegetable
products pinpointed several gaps in terms of data and methods to reach representative LCIA results for
the field production stage. To better account for the specificities of vegetable cropping systems and
improve the overall quality of their LCA studies, our key recommendations were: (i) to include
systematically phosphorus, water and pesticide fluxes and characterise associated impacts:
eutrophication, toxicity, water deprivation, (ii) to better address space and time representativeness for
field stage inventory data through better sampling procedures and reporting transparency and, (iii) to
use best available methods and when possible more mechanistic tools for estimating Nr emissions.
Keywords: Vegetable production; typology of cropping systems; emissions; Life Cycle

Assessment; reactive nitrogen; greenhouse; open-field systems
26
1.1. Introduction
The year-round demand for a diversity of vegetables and fruits from Western markets has led to the
development of complex supply chains (farming, grading, packing, transport, storage, refrigeration,
retailing…) often spread across distant countries. Particularly for off-season supply, this demand is
also responsible for the development of artificial cropping systems under temperate climate (soilless
and heated greenhouses) and of a range of cropping systems (from soilless and sheltered to open-field
productions) in Mediterranean and tropical countries. The growing awareness regarding the
environmental issues due to global food supply chains has now reached the fruits and vegetables
sector and the need for evaluating the different product origins and associated technologies has clearly
been expressed.
The Life Cycle Assessment (LCA) methodology was commonly used to assess the environmental
performance of agricultural products over the last fifteen years, revealing several scientific challenges
due in particular to the use of land by farming systems, their interactions with their environment, their
large variation in terms of soil, climate and practices and the importance of local and regional impacts
(Brentrup et al., 2004). Although the relevance of LCA to assess agricultural products was validated
for a large number of agricultural products (Brentrup et al., 2001; Mila i Canals et al., 2006), its
application to horticultural (i.e. fruits and vegetables) production systems was more recent and came
with renewed and specific scientific challenges.
Since horticultural products and production systems are characterized by specific conditions of
production and of interaction with their environment, but also by a large diversity of technologies and
management practices, a review of the state of the art of LCA studies for horticultural products, as a
recent object of study, appears necessary and timely. A recent review focusing on perennial crops
(Bessou et al., 2013) already provided important insights regarding specificities of LCA applied to
orchards (including fruit production). No relevant analysis has been made on LCA applied to annual
and even shorter crop length products, e.g. vegetables. In available LCA studies of vegetable products,
contribution analyses from cradle-to-grave and from-cradle-to-retail studies revealed a significant
contribution of post-harvest stages (Andersson et al., 1998; Sim et al., 2007). However, authors widely
agreed upon the field production stage being one of the greatest contributors especially to global
warming, acidification, eutrophication and toxicity potentials (Andersson et al., 1998; Antón et al.,
2005a; Mila i Canals et al., 2008; Hospido et al., 2009; Cellura et al., 2012b). The importance and
complexity of the field production stage in the LCA of vegetable products requires putting a particular
emphasis on the analysis of the methods and data used for this stage.
Vegetable cropping systems show a huge diversity of farming and environmental conditions of
production but also important specificities that largely impact the inventory, the associated on-field
27
estimation methods and the resulting LCA outcomes. Although this was already pinpointed in the
literature, the methods and data used for the description of vegetable products and the calculation of
their environmental inventory have never been reviewed nor have been their implications for LCA
results. We focus in this paper on the cradle-to-harvest stages, including the production of agricultural
inputs and the associated emissions. The objectives of this review were three-fold:
1. To assess the state of the art of published LCA studies on vegetables including an original
typology-based method, with a focus on data and methods used for the field production stage
2. To carry out a critical analysis of the methods used for the estimation of reactive nitrogen (Nr)
emissions
3. To make recommendations in terms of practice and scientific challenges for applying LCA to
vegetable cropping systems
1.2. Materials and methods

1.2.1. Selection of LCA studies and review process
A thorough literature search resulted in the identification of seventeen papers published in peer-
reviewed scientific journals or in ISO-compliant reports, which presented full LCA studies for
vegetable products. Studies which did not involve multi-criteria assessment (mainly those studies
focused on energy and global warming potential) were excluded from our review, as well as reports
which were not accessible on the Internet. A subsample of 10 studies featuring consistent and well-
described data for the modeling of cropping systems was found suitable for a qualitative and
quantitative analysis.
Our qualitative review of goals and scopes (section 1.3.1) covered the main aspects described in the
recent ILCD handbook related sections (European Commission, 2010). Our quantitative review of LCI
(sections 1.3.2 and 1.3.3) and LCIA results (section 1.3.4) were based on an original typology
described in the next section (section 1.2.2). Results were also presented according to their product
group and per reference. Across selected studies, only Global Warming Potential, Eutrophication
potential and Acidification potential were homogeneously calculated following the CML methodology
(Guinée et al., 2001) and could be selected for the quantitative analysis. The critical analysis (sections
1.4.1 to 1.4.3) followed the main recommendations from the ILCD handbook. An in-depth analysis
was also made on methods used for the estimation of nitrogen reactive emissions (section 1.4.4).
1.2.2. Typology of cropping systems and aggregation of results

The cropping system generally results from the combination of a product’s choice (targeted market),
certain objectives of production (yield and quality) and a location. For vegetable products, this leads to
diverse degree of artificialization in terms of protection status, heating and growing media to
overcome local natural constraints and reach the production objectives. To explore the cropping
28
systems diversity from our panel of papers we developed a typology of cropping systems, based on
three main criteria: protection and heating status (heated or cold greenhouse or open-field), growing
media (substrate or soil) and climate (temperate, Mediterranean or tropical context). This typology was
relevant to LCA as energy consumption is closely related to the greenhouse type and management
(heated or not), while the growing media and the climate characteristics affect inputs rates, yields and
field emissions. This typology resulted in 7 types named according to their characteristics. Greenhouse
systems are either heated (GH-heat) or cold (GH-cold). Both systems can be led soilless on substrate
(eg. GH-heat-sub) or directly on soil (eg. GH-heat-soil). Open-field cropping systems were classified
according to their respective climate: temperate (OF-Temp), Mediterranean (OF-Med) or tropical (OF-
Trop). Overall, 72 cropping systems were analysed and grouped into 7 cropping system types.
Results were also aggregated by product group. Indeed, crop species have specific management
requirements and potential yields, which affect their environmental performances. Across the 10
studies, 7 product groups were established namely, giant bean (Abeliotis et al., 2013), green bean
(Mila i Canals et al., 2008; Romero-Gámez et al., 2012), broccoli (Mila i Canals et al., 2008),
cauliflower (Martínez-Blanco et al., 2010), leek (de Backer et al., 2009), salad (Mila i Canals et al.,
2008) and tomato (Anton et al., 2004; Martínez-Blanco et al., 2009; Boulard et al., 2011; Martínez-
Blanco et al., 2011; Torrellas et al., 2012). Finally, we presented the results per reference to analyze
possible variations due to particular methodological choices by authors.
1.3. Presentation of LCA studies selected

1.3.1. Goal and scope
Intended applications
We can distinguish 3 objectives of the LCA studies reviewed:
 Comparing different farming practices (Romero-Gámez et al. 2012; Torrellas et al. 2012;
Martínez-Blanco et al. 2011; Martínez-Blanco et al. 2010; Martínez-Blanco et al. 2009; de
Backer et al. 2009, Antón et al., 2005a)
 Comparing different product‘s origins such locally-grown products versus imported equivalent
(Mila i Canals et al. 2008)
 Producing a reference for the environmental performance of a product in a given context
(Abeliotis et al. 2013; Boulard et al. 2011).
29
Function and functional unit

Most papers studied the function of production expressed in mass of fresh or dry product (Abeliotis et
al. 2013; Torrellas et al. 2012; Boulard et al. 2011; de Backer et al. 2009; Antón et al. 2005a). The
production function could be associated to quality aspects either from a commercial standard
(Romero-Gámez et al. 2012; Martínez-Blanco et al. 2011; Martínez-Blanco et al. 2010; Martínez-
Blanco et al. 2009) or from a nutritional value perspective (Martínez-Blanco et al. 2010). The
functions studied across those papers were often multiple including land occupation functions
(Abeliotis et al. 2013; de Backer et al. 2009) and could refer to functions beyond the production, such
as the consumption expressed in kilograms of vegetable consumed (Mila i Canals et al. 2008).
Comparison of systems
Across the 10 reviewed papers, contrasted vegetable systems were studied. All studies had in common
to present original inventories related to field production stages of vegetable products. Indeed, each
cropping system was unique in terms of management and yield. These management and yield
variations resulted from different modalities in terms of cropping system characteristics:
 The protection status: greenhouse versus open-field (Romero-Gámez et al. 2012; Martínez-
Blanco et al. 2011; Mila i Canals et al. 2008), greenhouse type (Torrellas et al. 2012; Boulard
et al. 2011) and greenhouse management (Torrellas et al. 2012; Romero-Gámez et al. 2012;
Antón et al. 2005a),
 The growing media: substrate versus soil (Boulard et al. 2011; Antón et al. 2005a),
 The location: temperate, Mediterranean or tropical contexts (Mila i Canals et al. 2008),
 The fertilization management, involving different types of fertilizer (mineral fertilizer versus
compost) and application rates (Martínez-Blanco et al. 2011; Martínez-Blanco et al. 2010;
Martínez-Blanco et al. 2009).
 The production orientation: conventional, integrated or organic (de Backer et al. 2009;
Abeliotis et al. 2013),
 The product cultivar (Abeliotis et al. 2013).
Finally, management and yield variations also resulted from between-farms and within-year variability
(Mila i Canals et al. 2008).
System boundaries and cut-off criteria

Reflecting those various functions but also the diversity of vegetable cropping systems studied, the
system limits were diverse. As a rule, the system boundaries for cradle-to-farm-gate LCA included
agricultural operations (including greenhouse management) and on-field emissions due to fertilizers
and pesticides within the foreground system (“foreground” refers here to the farm stage) and the
30
production, transport and use of buildings, machineries and agricultural inputs within the background
system (“background” refers to processes before the farm stage). Several additional stages were
recorded across studies. Eight studies explicitly included the waste treatment phase (Torrellas et al.
2012; Romero-Gámez et al. 2012; Boulard et al. 2011; Martínez-Blanco et al. 2011; Martínez-Blanco
et al. 2010; Martínez-Blanco et al. 2009; Antón et al. 2005a). Six studies explicitly included the
nursery phase (Torrellas et al. 2012; Boulard et al. 2011; Martínez-Blanco et al. 2011; de Backer et al.
2009; Antón et al. 2005a). Boulard et al. (2011) included the packaging phase occurring on the farm
and Mila i Canals et al. (2008) included fuel consumption related to the transport of workers. Some
studies included the biogenic fixation of atmospheric carbon dioxide in plants (Mila i Canals et al.
2008) or in soil after compost application (Martínez-Blanco et al. 2010; Martínez-Blanco et al. 2011).
We also noticed some punctual omissions, such as transport and packaging of agricultural inputs
(Romero-Gámez et al. 2012), building and machinery production and transport (Abeliotis et al. 2013;
de Backer et al. 2009), pesticides production and associated field-emissions (Torrellas et al. 2012;
Martínez-Blanco et al. 2011; Martínez-Blanco et al. 2010) or quite often simply pesticides field-
emissions (Abeliotis et al. 2013; Romero-Gámez et al. 2012; Martínez-Blanco et al. 2009).
Several procedures were described regarding multifunctional processes and flows crossing the
cropping system limits. Martínez-Blanco et al. (2009) and following papers (Martínez-Blanco et al.
2010; Martínez-Blanco et al. 2011) assigned avoided burdens from dumping organic wastes to their
system using compost. Torrellas et al. (2012) presented two scenarios regarding the expansion of their
heated greenhouse tomato system to co-generation of heat and electricity: (1) avoided burden from
producing the electricity used for the system using co-generation, (2) allocation of gas consumption
for greenhouse heating based on energy criteria. In Martínez-Blanco et al. (2011) and in Boulard et al.
(2011), the allocations of compost and amendments were based on plant uptake. The allocations of
infrastructure over annual rotations were based either on economic criteria (Boulard et al. 2011) or on
occupation time criteria (Mila i Canals et al. 2008).
1.3.2. Data and methods for LCI

Type, origin and quality of data
Across selected studies, type and origin of foreground data was of a wide diversity. Agricultural
practices data were always presented as site-specific, i.e. representative for one vegetable product at
the regional or the country scale. However, those data sets varied in time and space scales from
national production data recorded for one or a few years, to individual field data for a given crop cycle
defined according to its location, its duration and its position during the year. Data sources were also
very variable: Boulard et al. (2011) and de Backer et al. (2009) used data from large-scale surveys
carried out by agricultural development and extension services, while Abeliotis et al. (2013) and Mila i
Canals et al. (2008) used data based on a restricted number of farms surveyed specifically for their
31
study. Martínez-Blanco et al. (2009; 2010; 2011) and Romero-Gámez et al. (2012) used data from a 1
or 2-year trial, whose management followed regional recommendations. Torrellas et al. (2012) used a
mix of data collected by local partners, producers and sales cooperatives, as well as region-specific
literature and data from experimental stations. Antón et al. (2005a) used a mix of data from a farmers’
association for the soil cultivation scenario and from trials for the soilless scenarios. Background data
was quite homogeneous among the panel of 10 studies. Data regarding buildings, machinery and
inputs production were mainly based on literature or European and international databases. One
exception was Martínez-Blanco et al. (2009) who provided site-specific data regarding compost
production, based on data collected from a local compost plant.
Methods for field emissions

Table 1.1 shows the guidelines or references (column 1 and 2) used by authors (last column) to
estimate field emissions for various forms of reactive nitrogen (N r). Most studies referred to
recommendations from two general guidelines (Audsley et al., 1997; Nemecek & Kägi, 2007) and one
original paper (Brentrup et al., 2000), although transparency issues arose. Abeliotis et al. (2013) did
not explicitly state whether they included the emissions from mineral fertilizers or not. Torrellas et al.
(2012), Martínez-Blanco et al. (2011) and Martínez-Blanco et al. (2010) referred to several guidelines
without specifying which one was used for each type of Nr emission.
Some adaptations were explicitly described such as the extension to organic fertilizers of the
recommendation from Audsley et al. (1997) regarding nitrous oxide (N2O) emissions (Mila i Canals et
al., 2007; Mila i Canals et al., 2008). A few methods more specific to studied systems were also
mentioned: Boulard et al. (2011) used publications specific to greenhouse production systems to
estimate nitrate (NO3) emissions distinctly from open and closed-loop systems, while Torrellas et al.
(2012) deliberately omitted NO3 emissions from systems involving a recirculation of drainage water.
The calculation of phosphorus (P) emissions was only detailed in 4 studies. Abeliotis et al. (2013)
referred to recommendations from Nemecek and Kägi (2007) and de Backer et al. (2009) referred to
Audsley et al. (1997). Boulard et al. (2011) and Mila i Canals et al. (2007; 2008) used data from
specific publications, being respectively Sedilot et al.(2002) and Cowell and Clift (1998). Finally,
Antón et al. (2005a) deliberately omitted PO4 emissions on the basis that phosphorus was immobilized
in the topsoil due to its high pH.
Regarding pesticides emissions, only 3 studies explicitly mentioned their inclusion in the inventory.
Boulard et al. (2011) and Antón et al. (2005a) referred to the model from Hauschild (2000) adapted by
Antón et al.(2004). De Backer et al. (2009) referred to the model from Birkved and Hauschild (2006).
Finally, only Mila i Canals et al. (2008) explicitly included CH 4 emissions from ammonium fertilizer
application in the inventory.
32
Table 1.1 LCI methods to estimate reactive nitrogen field emissions: guidelines and references, calculation method using
emission factors and parameters, validity domain and uses in selected studies
Validity domain of the
Value, emission factors (EF) Uses in
Guideline References reference
- parameters papers a
Cropping system Location
NH3 emissions from mineral fertilizers
Audsley et al., 1997 Asman, 1992 EF % of N fertilizer Laboratory Europe 2, 4, 7, 9,
Nemecek and Kägi, - Fertilizer type experiment 10b
2007 (1, 3, 5, 6)
Brentrup et al., ECETOC, 1994 EF % of N fertilizer Mineral fertilizer European 8,
2000 - Fertilizer type Countries (3, 5, 6)
- Soil properties (pH, CaCO3)
NH3 emissions from organic fertilizers
Audsley et al., 1997 Menzi et al., EF= 50% of N-NH4 applied Cattle manure, Switzerland 7,
1998b - N-NH4 manure Grassland (5, 6)
Brentrup et al., Horlacher and - N-NH4 manure Cattle slurry Germany 8, 9,
2000 Marschner, 1990b - Operation & incorporation date (5, 6)
- Infiltration rate
- Temperature, Precipitation
Nemecek and Kägi, Menzi et al., - N-NH4 manure Cattle manure, Switzerland 1
2007 1998b - Fraction area spread Grassland
N2O emissions
EPA, 1995 Literature review EF % of N fertilizer Fluid fertilizer US (11)
- Fertilizer type
Audsley et al., 1997 Amstrong-Brown, EF % of N fertilizer Granular UK 2, 4, 9 e, 7,
1994 c - Fertilizer type, Temperature fertilizer, wet soil (3, 5, 6)
Brentrup et al., IPCC, 1996 EF= 1.25% of N fertilizer minus Mineral and Global 8, 10,
2000 NH3 and NOx losses organic fertilizer (3, 5, 6)
and manure
Nemecek et al., IPCC, 1996b, EF= 1.25 % of N soil , Mineral and Global (1)
2007 Schmid et al., EF= 1.25 % of N-NH3 losses, organic fertilizer
2000c EF= 2.5 % of N-NO3 losses, and manure
- Navailable fertilizers, Ncrop residues, N fixed
NOx emissions
Audsley et al., 1997 No reference EF= 10% of N- N2O losses na na 2, 3, 4, 5, 6,
7, 9, 10
Nemecek and Kägi, No reference EF= 21% of N2O losses na na (1)
2007
N2 emissions
Brentrup et al., Rheinbaben, EF= 9% of N applied Arable crop & Germany 2, 8
2000 1990c Grassland
NO3 emissions
Audsley et al., 1997 Simmelsgaard, - Reference leaching for Arable crops & Denmark 2, 7, 10 b
1998c recommended N fertilizer pastures, sandy & (3 f , 5, 6)
- N fertilizer loamy soil
Brentrup et al., Own method - Soil texture, NO3-N soil Crop production Germany d 8, (3, 5 ,6)
2000 - Precipitation
Nemecek and Kägi, Richner et al., - N uptake , N fertilizer Arable crops & Switzerland (1)
2007 2006c - Fraction N soil leachable meadow
- N mineralized , Soil depth
Cowell and Clift, 15 kg N-NO3.ha-1 na na 9
1998
Sedilot et al., EF= 20% (soil) and 48% Greenhouse France 4
2002 (substrate) of N applied
a. Between brackets are uncertain uses (not specific information in papers), Corresponding papers: 1. Abeliotis et al. (2013), 2.
Romero-Gámez et al. (2012), 3. Torrellas et al. (2012), 4. Boulard et al. (2011), 5. Martínez-Blanco et al. (2011), 6. Martínez-
Blanco et al. (2010), 7. Martínez-Blanco et al. (2009), 8. de Backer et al. (2009), 9. Mila i Canals et al. (2007 ; 2008), 10. Antón et
al. (2005a); b. Adaptation from the original method; c. Reference not published in English; d. Suggested value for parameters
adapted to German climate, e. Extended to organic nitrogen with 20% of N as ammonium and 30% as urea; f. no nitrate leaching
for recirculation irrigation systems; na = information no data available.
33
1.3.3. Input flows and yields

Table 1.2 presents foreground data related to the field production stage for 72 cropping systems
aggregated into 7 cropping system types presenting 7 product groups. For studies comparing systems
according to different allocations rules, each system associated with a given allocation ratio was
considered original and entered as an additional system.
Data completeness and reference flow

The completeness of field production data varied across studies. Data regarding energy consumption
were not provided in 4 studies (Romero-Gámez et al. 2012; Martínez-Blanco et al. 2011; Martínez-
Blanco et al. 2010; Martínez-Blanco et al. 2009). Torrellas et al. (2012) and Boulard et al. (2011) did
not provide data on fuel consumption for field operations. Martínez-Blanco et al. (2009) did not
provide specific data for water consumption.
Table 1.2 presents total fresh matter yields including commercialized and downgraded products. Those
yields ranged from 2 t.ha-1 for giant bean grown open-field in a Mediterranean context (Abeliotis et al.
2013) to 523 t.ha-1 for tomato grown on substrate under heated greenhouse (Torrellas et al. 2012).
Tomato showed yields above 100 t.ha -1 while the other crops showed yield below 60 t.ha -1, regardless
of the cropping system type.
Energy and input flows per ton of fresh matter yield

All cropping systems consumed diesel for field operations, except lettuce grown open-field in a
tropical context (Mila i Canals et al. 2008). Diesel consumption ranged from 28 MJ.t -1 for tomato
grown under cold greenhouse on substrate (Antón et al. 2005a) to 1,237 MJ.t-1 for giant bean grown
open-field in a Mediterranean context (Abeliotis et al. 2013). Some systems were equipped with
electrical pumps for irrigation leading to an electricity consumption from 5 MJ.t -1 for tomato grown on
soil under cold greenhouse (Antón et al. 2005a) to 301 MJ.t-1 for giant bean grown open-field in a
Mediterranean context (Abeliotis et al. 2013). This latter figure was due to its particularly low yield.
Regarding greenhouse heating, the sources of energy were electricity, natural gas or fuel, and in some
cases included a co-generation of electricity from fossil fuel leading to negative electricity
consumption (Torrellas et al. 2012). The global energy consumption for heating varied from 23,970
MJ.t-1 for tomato grown on substrate (Boulard et al. 2011) to 34,170 MJ.t-1 for a similar system
presented by Torrellas et al. (2012).
All cropping systems were irrigated, except leek (de Backer et al. 2009) and broccoli grown open-field
in a temperate context (Mila i Canals et al. 2008). The water volumes applied ranged from 1 m3.t-1 for
lettuce grown open-field in a tropical context (Mila i Canals et al. 2008) to 1,570 m3.t-1 for giant bean
34
grown open-field in a Mediterranean context (Abeliotis et al. 2013). However 80% of systems showed
water volumes below 100 m3.t-1. Systems showing the highest water consumption per ton were giant
bean, green bean and broccoli grown open-field in Mediterranean context with very low crop yields.
All cropping systems received synthetic fertilizers, except for the lettuce in open-field cropping
systems in Uganda (Mila i Canals et al. 2008). Variations of fertilizer inputs across systems were
lower than variations for other inputs. Nitrogen rates ranged from 1 kg.t -1 for lettuce grown open-field
in a tropical context (Mila i Canals et al. 2008) to 19 kg.t -1 for broccoli grown open-field in a
Mediterranean context (Mila i Canals et al. 2008). Cauliflower and tomato grown open field in Spain
(Martínez-Blanco et al. 2010; Martínez-Blanco et al. 2009; Martínez-Blanco et al. 2011), green bean
and leek grown open field in UK and Belgium (Mila i Canals et al. 2008; de Backer et al. 2009) did
not receive any phosphate-based synthetic fertilizers. Phosphate rates reached 29 kg.t -1 for green bean
grown open-field in a Mediterranean context (Romero-Gámez et al. 2012). Applications of organic
fertilizers were not systematic and never occurred for heated greenhouse types and types on substrate.
The types and input rates of organic fertilizers were systematically described. However the nutrient
contents of the organic fertilizers were not homogeneously reported in the studies. Organic fertilizers
rates reached 4 t.t-1 for giant bean grown open-field in a Mediterranean context (Abeliotis et al. 2013).
35
Table 1.2 Inputs flows for the field production stage expressed per ton of fresh yield and fresh yield for the 72 systems aggregated by cropping system type, product group
and reference.
Fresh Energy Synthetic fertilizers Organic

Cropping Water

Product group Reviewed paper CS a yield b Electricity Gaz/Fuel Diesel Nitrogen Phosphate fertilizers
system type
t.ha-1 MJ.t-1 MJ.t-1 MJ.t-1 m3.t-1 kg N.t-1 kg P2O5.t-1 t.t-1
Torrellas et al., 2012 4 523 -2835 37026 na 14 4 2 0
GH-Heat-Sub Tomato
Boulard et al., 2011 8 440 0 23970 na 29 6 3 0
GH-Heat-soil Lettuce Mila i Canals et al., 2008 2 58 1153 27498 46 7 1 1 0
Torrellas et al., 2012 1 165 0 0 na 29 5 3 0
GH-Cold-Sub Tomato
Anton et al., 2005 2 150 5 0 28 24 5 6 0
Green bean Romero-Gamez et al., 2012 4 55 na na na 88 1 7 0
Boulard et al., 2011 1 150 0 0 na 33 3 2 0
GH-Cold-Soil
Tomato Martinez-Blanco et al., 2011 3 161 na na na 33 2 0 0
Anton et al., 2005 1 120 5 0 35 29 6 3 0
36
Broccoli Mila i Canals et al., 2008 4 12 0 0 486 0 14 3 0

Leek de Backer et al., 2009 2 33 0 0 156 0 1 0 1
Lettuce Mila i Canals et al., 2008 6 22 0 0 742 37 6 1 0
Giant Bean Abeliotis et al., 2013 7 2 301 0 1237 1570 10 20 4
Green bean Romero-Gamez et al., 2012 2 19 na na na 265 4 29 0
Broccoli Mila i Canals et al., 2008 4 16 0 0 521 140 19 5 0
OF-Med Cauliflower Martinez-Blanco et al., 2010 3 57 na na na 41 3 0 0
Martinez-Blanco et al., 2009 6 124 na na na na 3 0 0
Tomato
Martinez-Blanco et al., 2011 3 127 na na na 45 3 0 0
Green bean Mila i Canals et al., 2008 2 24 0 0 42 36 1 1 1
OF-Trop
GH = Greenhouse; OF = Open-field; Temp = Temperate climate; Med = Mediterranean Climate; Trop = Tropical climate, na = data were not available in the published
documents a. Number of cropping systems aggregated within the line b. Fresh yield expressed in t.ha-1 for one crop cycle, including downgraded products
1.3.4. LCIA results

Figure 1.1 presents LCIA results expressed per kilogram of fresh matter yield for the three impact
categories selected (global warming, acidification and eutrophication potentials) for the 72 cropping
systems of the 10 reviewed studies. The 72 cropping systems were successively grouped by cropping
system type, product group and reviewed paper. When LCIA results were expressed using functional
units involving quality aspects, necessary conversions were used to homogenize the functional unit to
yields presented in table 1.2. LCIA results expressed per square meter and crop cycle are given in
Annex 1.
Global Warming Potential

The cropping system types under heated greenhouse: GH-Heat-Sub and GH-Heat-Soil had a greater
global warming potential (GWP) than other types, at 2.03 and 2.12 kg CO2-eq. per kg of fresh yield,
respectively (Figure 1.1). GWP variations across GH-Heat-sub systems were mostly due to differences
in the origin of the energy. Indeed, GWP from GH-Heat-Sub systems presented by Torrellas et al.
(2012) showed important variations (± 58%) mainly due to energy sources and associated allocations
and substitution rules (gas=5.00 kg CO 2-eq.kg-1 fresh yield; co-generation + allocation= 2.00 kg CO2-
eq.kg-1 fresh yield; co-generation + avoided burdens= 0.78 kg CO2-eq.kg-1 fresh yield; geothermal=
0.44 kg CO2-eq.kg-1 fresh yield). GWP variations across the GH-Heat-Soil systems were due to
different energy consumptions for greenhouse heating. According to authors presenting the two lettuce
cropping systems (Mila i Canals et al. 2008 ), one farm grew only one cycle of lettuce indoors, very
likely during the coldest period, while the other farm grew lettuce indoors all year-round, with on
average 5 cycles per year. The variation in energy consumption between systems of the GH-Heat-Soil
type could result from the crop cycle position within the year. GWP per kg fresh yield for all other
types ranged from 0.05 (OF-Trop) to 0.45 kg CO2-eq. (GH-Cold-Soil) without showing clear
differences due to large intra-type variations. For example, the higher impact of green bean under cold
greenhouse results from the misting system used to cool down the greenhouse temperature.
Products groups showed undifferentiated GWP values (Figure 1.1), ranging from -0.36 (Broccoli) to
0.89 kg CO2-eq.kg-1 fresh yield (Tomato). Negative GWP was observed in the articles by Martínez-
Blanco et al. (2009, 2010 and 2011) due to the avoided dumping of organic wastes in systems using
compost as a fertilizer. For example, the tomato system with compost only (Martínez-Blanco et al.
2009) showed a GWP of 0.15 kg CO2-eq. kg-1 fresh yield without considering avoided burdens and -
0.90 kg CO2-eq. kg-1 fresh yield with avoided burdens.
While being similar per kg fresh yield, the GWP for GH-Heat-Soil and GH-Heat-Sub systems were
significantly different per unit of area (Annex 1), the lowest being for the GH-Heat-Soil type. Such
discrepancy between impacts expressed by the two functional units resulted from the combined effect
of yield per day and crop duration, the lettuce being grown for two to three months only while tomato
37
were grown for up to 11 months. Moreover, while showing undifferentiated values between product
groups per kg fresh yield, the GWP per unit of area for the tomato group clearly exceeded the GWP of
other product groups. The sensitivity of results to the FU choice highlighted here the high intensity of
management (quantity of infrastructure and energy used per area) for tomato system grown under
heated greenhouse.
Acidification
Regarding the acidification potential (AP), expressed in g of SO 2 equivalent per kilogram of fresh
yield, cropping system types did not present clearly different means due to large variations within
types and products groups. However, we could relate the ranking of impact between product groups to
the fertilizer efficiency, i.e. the ratio of fertilizers inputs per kilogram of fresh yield (as shown in table
1.2). The giant bean group (Abeliotis et al. 2013), presenting the lowest manure efficiency (4t. t -1 fresh
yield) as well as a low mineral nitrogen efficiency (10 kg N.t -1 fresh yield), had the largest impact with
an average of 11.26 g SO2-eq. per kg of fresh yield. The authors of this study identified field emissions
from sheep manure as the major contributor for AP. The second greater impact (3.26 g SO 2-eq.kg-1
fresh yield) was shown by the broccoli group, which showed the lowest mineral nitrogen efficiency
(19 kg N.t-1 fresh yield). The particularly low yields shown by these two groups, suggested that AP
variations were greatly affected by yields. The lowest impact (0.42 g SO 2-eq.kg-1 fresh yield) was
shown by the leek group, presenting one of the highest mineral nitrogen efficiency (1 kg N.t -1 fresh
yield).
The field emissions and low yields of giant bean and broccoli contributed largely to the high AP of the
Mediterranean open-field systems type, averaging 4.28 g SO2-eq. kg-1 fresh yield. However, this mean
was associated with a particularly large coefficient of variation (± 72 %) stemming from broad
variations of input efficiency across systems within this type.
Variations of AP per unit of area (Annex 1) regardless of aggregation mode were greater than
variations for AP per kg of vegetable ( 156% vs.  71%). Such differences between results expressed
by the two functional units highlighted the yield effect, as already suggested previously. The variations
of AP between cropping system types and product groups were reduced per kg fresh yield by
integrating an aspect of input efficiency, as yields are supposed to be correlated to inputs of systems.
Similarly to GWP, the ranking of impact per unit of area for product groups and cropping system types
differed from the ranking using the functional unit per kilogram of fresh yield. Indeed, AP expressed
per unit of area for tomato systems clearly exceeded the AP of other products groups. Similarly AP
expressed per unit of area for GH-Heat-Soil systems clearly exceeded the AP of other cropping
systems types. Again, the sensitivity of results to the FU choice highlighted the high intensity of
fertilization management (quantity of fertilizer per unit of area) for tomato systems grown under
heated greenhouse.
38
Eutrophication
Regarding the eutrophication potential (EP) expressed in g of PO 43- equivalent, giant bean systems
showed a greater impact (1.75 g PO43--eq.kg-1 fresh yield) than other products groups. Broccoli and
green bean systems also showed a high impact (1.01 and 0.95 g PO 43--eq.kg-1 fresh yield respectively).
For broccoli systems, such high impact mainly resulted from low mineral nitrogen efficiency, while
for green bean it resulted from low mineral phosphorus efficiency (29 kg P 2O5. t-1 fresh yield). For
giant bean the high impact resulted from both low phosphorus and nitrogen efficiencies (10 kg N and
20 kg P2O5. t-1 fresh yield, respectively).We can notice that despite being a leguminous crop (i.e.
fixing nitrogen from the atmosphere), the giant bean group presented relatively high nitrogen inputs
contributing to its high EP. The Cauliflower group showed less impact than other groups (-1.53 g
PO43--eq.kg-1 fresh yield) due to the absence of phosphorus inputs, the high mineral nitrogen efficiency
(3 kg N. t-1 fresh yield) and the system expansion rules used by Martinez-Blanco et al. As already
explained for GWP, the latter obtained a negative EP thanks to negative nitrogen and phosphorous
emissions from avoided organic waste dumping in the system using compost as a fertilizer. Despite
low fertilizer rates, the leek group showed the third greater impact (0.68 g PO 43--eq.kg-1 fresh yield)
due to important phosphorus emissions.
System expansion related to compost fertilization contributed largely to artificially reduce the average
EP of Mediterranean open field cropping systems (only 0.45 g PO 43--eq.kg-1 fresh yield despite the
high impact of giant beans) which presented the highest standard error (211%) across all types,
products and impact categories. Across all types, GH-Heat-Sub showed the greatest EP (1.08 g PO43--
eq.kg-1 fresh yield). This greatest impact occurred despite the avoided electricity production thanks to
the use of natural gas for co-generation assumed by Torrellas et al. (2012). Over the 12 GH-Heat-Sub
systems, one system presented in Torrelas et al. (2012) and 20% of systems from Boulard et al. (2012)
were closed-loop.
Similarly to AP, the ranking of impact per unit of area for product groups and cropping system types
differed from the ranking of impacts per kilogram of fresh yield: the EP per unit of area of GH-Heat-
Soil systems exceeded the EP of other cropping systems types, including GH-Heat-Sub and the EP per
unit of area of tomato systems exceeded the EP of other products groups. This latter result highlighted
once again the high intensity of fertilization management per area for tomato system grown under
heated greenhouse.
39
Figure 1.1 Global warming potential (GWP), acidification potential (AP) and eutrophication potential (EP)
expressed per kg of fresh yield from cradle-to-harvest for the 72 systems aggregated per cropping system type,
product group and reviewed paper. Figures between brackets correspond to the number of systems in the group.
Grey boxes correspond to error bars centered on the mean of the group (red dash).
40
1.4. Relevance of the typology to compare LCIA results from a diversity of cropping
systems and vegetable products
While it is difficult to thoroughly analyze LCIA results from several individual studies, using a
typology allowed us to provide general conclusions on the environmental impacts of vegetable
products and to identify the main cropping systems components responsible for these impacts.
For GWP, we can conclude that the aggregation per cropping system type was the most relevant with
heated greenhouse types showing the greatest impact. In agreement with several studies dealing with
fruit and vegetable products (Hospido et al., 2009; Stoessel et al., 2012), we identified the type of
energy consumed for greenhouse heating as an important source of impact variations. However, GWP
for cold greenhouse and open-field types could not be distinguished due to low contribution of
infrastructures compared to the impact associated to material and input uses. Our analysis also showed
that the crop position within the year led to important GWP variations particularly due to the
requirement or not for heating. For AP our analysis suggested that the aggregation per product group
was the most comprehensible with the giant bean group showing the greatest impact and the leek
group showing the lowest impact. For AP and EP, we identified phosphorus and nitrogen emissions
from both mineral and organic fertilizers as the major contributors to the impacts. However, our
analysis identified system expansion rules as another important source of variations resulting in the
Cauliflower group showing the lowest impact. Finally, comparing two distinct functional units, we
showed that LCIA results per kilogram of fresh yield was largely affected by input efficiency per mass
of product, while LCIA results per area were largely affected by management intensity per area. This
discrepancy led to a change on the ranking of cropping system types and product groups according to
the functional unit.
Despite those important insights, the benefits of the typology were hampered by the occurrence of (1)
low variations between cropping system types and products and (2) high variations within types and
products.
We explored several hypotheses to explain these variations:
1. Inherent variability of cropping systems and products

2. Discrepancies and relevance of methodological choices
3. Limitations in our typology approach
4. We further elaborate on these three potential sources of limitations in sections 1.4.1, 1.4.2 and
1.4.3 with a special focus on the methods used for the estimation of Nr emissions in section
1.4.4.
41
1.4.1. Inherent variability of cropping systems and representativeness issues

Some variations within cropping system types can be explained by the actual variability attached to
cropping systems and products in general. Indeed, the variability of farmer’s practices, soil and
climate conditions affects cropping systems performances. However, across studies, the poor or
unspecified data representativeness reduced the possibilities to identify and interpret this inherent
variability.
Across studies, technological, geographical and time-related representativeness of data regarding goal
and scope were difficult to assess. Most studies provided detailed information on techniques involved
in the field production stage as they were often an object of comparison between systems. However,
the time and geographical scales were less detailed. We could locate the data used in time and space
but their representativeness regarding the studied region and period was poorly addressed. Only
Boulard et al. (2011) provided figures on national production and discussed the representativeness of
the data used for the LCA. Abeliotis et al. (2013) tried to address the intra-farms variability proposing
a weighted mean resulting from an area-based aggregation of different farms, but did not provide
standard deviations. Several LCA studies dealing with vegetable crops (Cellura et al., 2012a) and
other crops (Mouron et al., 2006) showed the influence of the management on the environmental
performances of systems which pleads for a better inclusion of farms variability in LCA studies.
1.4.2. Discrepancies in methodological choices

System boundaries and completeness of studies
We highlighted a lack of homogeneity regarding system boundaries and cut-off rules. Authors listed
and justified most of excluded processes following recommendations from the ILCD handbook
(European Commission, 2010). However, the constraint of reporting an LCA through a publication
often led to a lack of transparency regarding included processes. In addition several exclusions were
not relevant with regard to their potential contribution to the environmental impact of vegetable
products.
First, some important stages were not systematically included. The waste treatment stage was
excluded from 3 studies over the 10 reviewed although Antón et al. (2005b) showed that it contributed
for 20 to 94% of GWP and 5 to 50% of EP for tomato production, depending on the type of
management considered. Our review did not allow identifying neither the diversity of waste
management techniques nor associated impact variations across systems. The nursery stage was not
explicitly included in 6 studies. To our knowledge there is no literature providing LCIA results for the
nursery stage of vegetable. However, for tomato production under cold greenhouse Boulard et al.
(2011) identified heating during the nursery stage as the major contributor for GWP of the whole
system.
42
Second, phosphorus emissions were not included in 5 studies with systems fertilized with mineral or
organic phosphorus. One of these studies explained that phosphorus was immobilized due to high soil
pH resulting in no losses by leaching. However, phosphorus losses occur as soluble phosphate through
leaching and run-off but also through the erosion of soil particles. Boulard et al. (2011) showed that
phosphate in soil and water contributed from 45% for heated greenhouse on substrate to 63% for cold
greenhouse on soil to the eutrophication potential at farm-gate. Due to the heterogeneous description
of drainage water management in reviewed papers, we could not analyze eutrophication potential
variations related to this aspect. However, Boulard et al. (2011) also stated that the phosphorus
emission could probably be reduced by 40% when switching from an open-loop system to a closed-
loop system.
Finally, the impacts of pesticide use could not be compared across studies since six of them did not
assess human or ecosystem toxicities. According to the ILCD handbook (European Commission,
2010), all impact categories that are relevant for the LCIA study shall be included. Several studies
showed that pesticide use dominated the life-cycle toxicity impact of vegetable production (Margni et
al., 2002; Anton et al., 2004). Among those papers which included human and ecosystem toxicity
impact categories, two did not explicitly include pesticides emissions in the field. Mila i Canals et al.
(2008) criticized current estimation methods for being either too simplistic (Audsley et al., 1997) or
too impractical (Birkved and Hauschild, 2006), and proposed an additional indicator based on
pesticide use.
Transparency and relevance of methods for field emissions

Readers should have the capacity to interpret the contribution of field emissions to the environmental
impacts. However, the references used to estimate these fluxes were not sufficiently detailed, neither
were the parameters values used for calculation. Four papers did not report the methodology used to
estimate each emission from fertilizer and pesticide use. In addition, authors did not detail the
parameter values used for the calculations, preventing the reader from recalculating emissions when
they were not included in the article. The acidification potential and the eutrophication potential
should be largely affected by reactive nitrogen-based and phosphorus-based field emissions. Unlike
Boulard et al. (2011) and Antón et al. (2005a), our review did not identify a reduction of the
eutrophication potential between systems on substrate and systems on soil, regardless of the functional
unit used. In addition we could not properly identify the respective contributions of phosphorus and
nitrogen emissions to the eutrophication potential. Considering the particularly weak description of
methods to estimate phosphorus emissions and the lack of consensus, we can assume that important
EP variations were related to methodological choices.
43
Detailed information on the estimation methods would have permitted to explain why those systems so
diverse a priori showed similar impacts for AP and EP and to assess the relevance of the methods
used. This is especially likely for water-borne emissions (the main contributors to EP), since
differences in climate, irrigation, fertilizer managements and growing media result in particular water
and nutrients dynamics. Since N r emissions were systematically included in inventories and often
responsible for significant acidification and eutrophication potential impacts, a special focus on their
estimation methods is proposed in section 1.4.4.
1.4.3. Weaknesses in reporting, a limitation to our typology approach

Our typology was an a priori typology based on a restricted number of characteristics of vegetable
cropping systems. A quantitative typology based on a statistical approach (principal component
analysis and hierarchical clustering) would have been preferable. However, it was not feasible in our
situation given the lack of transparency in the reporting of key cropping systems characteristics across
studies. Below, we identify often missing although critical parameters to be included and reported in
LCA studies for vegetable products. These parameters could allow improving the relevance of further
attempts at characterizing the variability of LCIA results across vegetable products. Such
characterization based on a meta-analysis should therefore provide information about important
influencing factors for the environmental performances of vegetable products for the attention of
policy makers, food industries, consumers and producers.
Our analysis showed that the crop position within the year led to important GWP variations. For some
vegetable crops, several cropping cycles can occur on the same field during one year. Cellura et al.
(2012a) showed that crop rotations can lead to a clear reduction of all environmental impacts for both
Tomato and Melon, especially with respect to waste production and allocation rules.
Our analysis also showed the important contribution of organic fertilizers to AP and EP. However,
information regarding the types and composition of organic fertilizers was not homogeneous across
studies. The type and the composition of the organic fertilizers inform not only on the nutrient inputs
to cropping systems, but also on the environmental risks associated to their application. Indeed, in
addition to nitrogen and phosphorus contents, compost (Shiralipour et al., 1992; Weber et al., 2007)
and manures are known to be sources of a diversity of trace contaminants such as heavy metals and
persistent organic pollutants. Finally, the method of application plays an important role regarding the
acidification and eutrophication potentials with some application techniques of manure especially
designed to abate ammonia emissions (Langevin, 2010).
Finally, irrigation practices appeared in recent studies in Mediterranean areas as an important
contributor for GWP and water consumption (Abeliotis et al. 2013, Cellura et al., 2012a&b, Romero-
Gámez et al., 2012). Our analysis did not allow identifying the contribution of irrigation practices to
44
the impact of vegetable production due to heterogeneous description of energy and water consumed.
However, the huge diversity of irrigation practices from recirculation systems to sprinkler irrigation
systems should have affected the environmental performances of vegetable production systems.
1.4.4. Relevance and validity of the methods used for the estimation of reactive nitrogen
emissions
Field emissions are substances emitted during the field production stage. Emissions can be either
directly caused by an input application (fertilizers or pesticides) or by processes occurring in the soil
and in the atmosphere (nitrogen deposition, drainage, soil mineralization). Field emissions do not
include emissions from fuel burned in tractors. In this section, we focused on reactive nitrogen (N r)
emissions because they were systematically included in inventories and often responsible for
significant acidification and eutrophication potential impacts. In table 1.1, the references detailing
calculation methods were analyzed regarding the parameters included and the validity domain of the
specific method, be it an emission factor or a more complex equation. In the next sub-sections, an in-
depth analysis for each emission process is proposed regarding the adequacy of each reference to the
specific conditions of the diverse vegetable cropping system types. Recommendations are made in the
last section of this paper.
Volatilization of applied fertilizers and associated ammonia emissions

Volatilization is a physico-chemical process leading to ammonia (NH3) emissions, which have high
acidifying properties. In the three guidelines used, NH 3 emissions from mineral and organic fertilizers
were dealt with separately.
For mineral fertilizers, the guidelines of Audsley et al. (1997) and Nemecek and Kägi (2007)
recommended emission factors from Asman (1992). This reference proposed emission factors
measured under controlled conditions in the laboratory for each type of fertilizer used in Europe.
These emission factors were used as references for the whole Europe, until the publication of the
ECETOC report (1994) recommended by Brentrup et al. (2000). ECETOC (1994)recommended a new
set of emission factors based on an extensive literature review, depending on the country considered
and the occurrence of soils with a pH greater than 7 (indicative of calcareous soils).
For NH3 emissions from organic fertilizers, the guidelines from Audsley et al. (1997) and Nemecek
and Kägi (2007) recommended a calculation method developed by Menzi et al.(1998). This method
was based on experiments in Switzerland, where liquid manure was applied to grassland. Audsley et
al. (1997) suggested simplified emission factors depending on the ammonia content of the organic
fertilizer. By contrast, Nemecek and Kägi (2007) slightly adapted the empirical equation from Menzi
et al. (1998) with parameters related to techniques of application. Finally, Brentrup et al. (2000)
recommended an empirical equation developed by Horlacher and Marschner (1990) in a temperate
45
context for cattle slurry spread on arable crops. Some parameters of this empirical equation may be
adjusted to account for different soils, crops and times of application with regard to incorporation or
precipitation.
None of those references are actually suitable to estimate emissions from open-field systems under
Mediterranean or tropical climates since high temperatures may entail large variations in the
emissions, especially if the fertilizer application occurred in the hottest season. For greenhouse
systems, the use of those calculation methods was very remote from their validity domain as wind
speed and temperature effects were not included, not to mention the effect of plastic mulch. Ammonia
emissions are largely variable according to crop type, local context and application techniques
(Bussink & Oenema, 1998; Langevin, 2010).
Denitrification of soil nitrogen and associated gaseous emissions

Nitrification and denitrification of soil nitrogen are responsible for the emission of nitrous oxide
(N2O), nitrogen oxides (NOx) and nitrogen gas (N2) emissions. N2O is a potent greenhouse gas, while
NOx contributes mainly to photochemical oxidant formation and terrestrial acidification. In the two
guidelines (Audsley et al., 1997; Nemecek and Kägi, 2007) and the original paper from Brentrup et al.
(2000), each emission is dealt with separately.
For N2O emissions, Audsley et al. (1997) associated N2O emissions to nitrogen from fertilizer
application and emission factors were based on an unpublished reference. These emission factors
depended on the fertilizer type and the soil temperature and belonged to the range [0.4-3%] (as a
fraction of applied N). Later on, the guidelines from Brentrup et al. (2000) and Nemecek and Kägi
(2007) recommended emission factors from IPCC (1996), based on an extended review of published
field experiments mainly from temperate regions. Brentrup et al. (2000) recommended using the direct
emission factor from the IPCC methodology. The direct emission factor corresponds to the proportion
of soil nitrogen turnover emitted as N2O, including N mineralized from mineral and organic fertilizer
N inputs, N fixation from legumes, and N in plant residues. In addition to the emission factor for direct
emissions, Nemecek and Kägi (2007) also included the two indirect emission factors from IPCC
(1996) via the equation developed by Schmid et al. (2000). The indirect emission factors correspond
respectively to the fraction of N inputs volatilized as ammonia, which is further deposited to soils and
emitted as N2O and, the fraction of N inputs leached as nitrates and further emitted as N 2O.
Only two guidelines included the estimation of NO x emissions. Brentrup et al. (2000) neglected this
emission, probably because there was no consensus on its emission factor. The guidelines from
Audsley et al. (1997) and Nemecek and Kägi (2007) agreed on NO x emissions equal to 21% N 2O
46
emissions. None of them quoted any published reference to justify such figure, which is actually
contradicted by recent literature (Stehfest & Bouwman, 2006). Finally, although not of environmental
relevance, N2 emissions were included by Brentrup et al. (2000) in their nitrogen balance approach to
infer nitrate leaching. They proposed a constant emission factor of 9% of the nitrogen applied
according to von Rheinbaben (1990).
Providing a unique emission factor, the IPCC guidelines (Tier 1) do not allow for discriminating
between systems but for calculating the soil nitrogen input. Nevertheless, the broad validity domain of
this reference and its wide use make it the most appropriate method currently available for open-field
systems, even under tropical conditions as a few field experiments used to calculate the emission
factors were led in tropical areas. However, its use for greenhouse systems was far beyond the scope
of the IPCC report. It does not appear that significant improvements can be possible using a unique
emission factor, since many parameters influencing N2O and NOx emissions were recently evidenced
and their interactions are not fully understood (Stehfest and Bouwman 2006). The effect of
temperature and fertilization on soil nitrogen content and soil pH and the effect of the irrigation on soil
humidity are particularly relevant to consider for vegetable cropping systems.
Leaching and associated nitrate emissions

Nitrate leaching occurs when there is an accumulation of NO 3 in the soil profile that coincides with or
is followed by a period of drainage. Therefore, excessive nitrogen fertilizer or N applications before
drainage events, ploughing, or long fallow without cover, can all potentially lead to NO 3 leaching
losses. In each guideline analyzed, a specific approach for nitrate leaching estimate is recommended in
relation to the specific context where the method was developed.
Audsley et al. (1997) recommended the use of a Danish approach adapted from a statistical equation
from Simmelsgaard and Djurhuus (1998). In this approach, a reference leaching loss for a
recommended fertilizer rate was used as a parameter to reflect the soil capacity for leaching. The
validity domain of this statistical model is restricted to Danish conditions where references are
available regarding the reference leaching. Brentrup et al. (2000) suggested the use of a conceptual
model based on a nitrogen balance coupled with an empirical equation reflecting drainage events.
Although following a more conceptual approach, this equation was calibrated under German
conditions for arable crops only. Finally, Nemecek and Kägi (2007) recommended the use of a
conceptual approach including a set of emission factors. These emission factors were calculated using
a mechanistic model developed by Richner et al.(2006), but were further calibrated under Swiss
conditions for arable crops. This conceptual approach included parameters related to soil, plant uptake,
soil coverage and risk of nitrogen leaching depending on the crop types and the months in which
47
fertilizers were applied. This range of parameters allowed the user to adapt the calculation within the
range of situations defined in the method (arable crops under temperate climate).
The current methods used to estimate nitrate emissions were more complete in terms of driving
parameters compared to other emissions, but remained mostly valid for arable crops in temperate
contexts. Overall, the only reviewed study using a suitable method for the estimation of leaching was
de Backer et al. (2009) who assessed leek cropping systems (without irrigation) under temperate
climate using the Brentrup’s method (Brentrup et al. 2000). Otherwise, none of these methods were
really suitable for estimating nitrate leaching for any other vegetable cropping systems.
1.5. Conclusions and recommandations

This paper reported on the current practice for the LCA of vegetable products with a focus on the field
production stage. It was a first attempt at comparing potential impacts of a large range of cropping
systems representative of the current European supply mix of vegetable products. It presented
aggregated LCIA results: i) according to the cropping system type through an original typology, ii)
according to the product group and iii) according to the reference. Setting a typology of systems
appeared as a suitable approach that should be generalized to characterize systems diversity and obtain
representative Life-Cycle Inventory (LCI) data, in relation to the function and scale defined in the goal
and scope. We can conclude that the heated greenhouse system types showed the greatest GWP impact
due to the crop position within the year in relation to energy requirement for heating. When GWP was
expressed per area, the tomato group showed the highest impact due to high management intensity per
unit of area. Regarding AP and EP the giant bean group showed the greatest impact due to low yields,
responsible for low input efficiency. When AP and EP were expressed per unit of area, the tomato
group showed the greatest impact due to high fertilizer management intensity per area.
We highlighted important sources of impact variations across systems. Important GWP and EP
variations were caused by system expansion rules related to energy and compost used. Important GWP
variations were due to the crop position within the year and the crop duration, in relation to energy
requirement for heating. Important AP and EP variations were caused by yield variations and diverse
fertilization managements.
However, the ability to compare the environmental impact for these diverse field production stages
was hampered by (1) weaknesses regarding transparency of goal and scope, (2) a lack of
representativeness and completeness of data used for the field stage and (3) heterogeneous and
inadequate methods for estimating on field emissions across reviewed studies.
48
Table 1.3 Summary of recommendations related to key parameters to be included in LCI and methods to be used for the estimation of reactive nitrogen (Nr) emissions to
account for specificities of vegetable cropping systems
Methods for the estimation of Nr emissions
Specificities of
Key parameters NH3 mineral NH4 organic
cropping systems N2O and NOx NO3

fertilizers fertilizers
 Fresh matter yield and commercial yield if suitable
 Nitrogen and phosphorus inputs along with
fertilizer types
Vegetable cropping  Water inputs and origin
systems  Irrigation system specificities    
 Crop duration and position within the year
 Previous and following crop
 Crop residue management
 Greenhouse materials and associated waste Brentrup et al. (2000) adapted
1. Greenhouse
management ø ø ø to soil type and irrigation
systems
 Allocation rules for infrastructures management
 Energy inputs and origin
49
1.1. Heated
 Allocation and substitution rules associated to
greenhouse systems
co-generation
1.2. Systems with  Substrate type and associated waste management
substrate  Drainage water management ø ø Bres (2009)
 Soil type
 Rainfall ECETOC
Brentrup et al.
IPCC (2006) & IPCC (2006) or Brentrup et
2. Open-field systems (2000)
 Origin and composition of organic fertilizers (1994)… Nemecek et al. (2007) al. (2000) adapted to…
 Allocation rules associated to organic fertilizers parameterized for…
2.1. Temperate … soil type and … soil type and irrigation

context … Gp. I 
plant cover management
2.2. Mediterranean … soil type and … soil type and irrigation
… Gp. III 
context plant cover management
Bouwman and Van Der Hoek (1997) … soil type, irrigation
2.3. Tropical context 
& ECETOC (1994) Gp. II management and rainfalls
 Level of specificity not homogenous enough to consider only one method, ø level of specificity which requires a specific methods but for which no method is available to
date,  level of specificity for which a specific methods could allow reducing the uncertainty.
Generally speaking, several aspects would deserve a systematic inclusion and a dedicated reporting in
the LCA of vegetable products. First, human and ecosystem toxicity impacts should be more
systematically assessed to consider pesticide practices, as well as the impact of freshwater use to
account for the frequent use of irrigation and the location of systems often in contexts of water stress.
Their absence can be partly explained by the lack of a consensual method at the time of their
publication regarding both methods for estimating emissions (for pesticides) and for characterization.
For these categories, there is an urgent need for further investigation. Second, geographical and time-
scale representativeness should be better addressed with regard to inventory data for the field
production stage of vegetable products. In agreement with the ILCD handbook (European
Commission, 2010), we strongly recommend providing data describing the sample of farms surveyed
with regard to its representativeness in connection with the scope of the LCA. Variability of the
management between farms and years should be better accounted for through valid sampling
procedures and discussed in the interpretation step, based on consensual quality criteria such as the
Data Quality Criteria proposed by Weidema and Wesnæs (1996). Finally, the waste treatment and the
nursery stage should be systematically included in inventories with detailed information about
techniques involved, inputs and energy consumption and emissions at least until their insignificant
contribution would be clearly demonstrated.
More specifically, an effort should be made to collect and report accurate, precise, and complete
inventory data regarding the main discriminating characteristics of cropping systems representative for
the studied function. Authors should more often take opportunities offered by many journals to
provide full inventories in supplementary materials. A quality reporting of key cropping systems’ data
would allow better quality meta-analyses including statistical analysis as an extension the analysis
proposed in this paper. Table 1.3 summarized those key parameters to be included in inventories
according to cropping system specificities. In agreement with recommendations from several authors
of LCA for vegetable products (de Backer et al. 2009; Andersson et al. 1998), land occupation
characteristics (the crop duration, its position within the year, the previous and following crops, crop
residues management) along with detailed allocation rules should be described to properly assess the
impact of vegetable crops on a yearly-basis. As proposed by several authors, another solution could be
to expand the system to the complete rotation. The organic fertilizer management should be
thoroughly taken into account for open-field cropping systems, in particular regarding the allocation
rules related to its use. There is also a pressing need to investigate impacts related to the contaminants
contained in organic fertilizers. Being an important source of nutrients especially regarding integrated
and organic production orientations, more detailed data on organic fertilization is deemed necessary.
Finally, regarding irrigation practices, the infrastructure and energy consumption should be better
documented with regard to its potential contribution to global warming. There is also a pressing need
to specify volumes and origin of the water used especially in a context where water is a scarce
50
resource. The irrigation infrastructure (drip or sprinkler), the quantities and nature of energy used
(diesel or electricity), the quantities and origin of water used (river, groundwater) should be
systematically detailed in LCA dealing with vegetable products.
Regarding field emissions, we recommend to systematically estimate emissions of phosphorus and

pesticides in line with the specificities of the cropping systems evaluated and to report on the methods
used for their estimation. Regarding nitrogen reactive emissions, we examined the relevance of the
methods used to the diversity of cropping systems studied. Table 1.3 summarizes identified gaps in
terms of methods availability and provides recommendations on the best available methods according
to cropping system specificities. We believe that available methods could not allow a proper
estimation of Nr emissions for all cropping systems, due to the diversity of vegetable systems. For
gaseous emissions from systems under greenhouse, there is currently no available method accounting
for such confined atmospheres (radiation, humidity and wind), plastic mulching or substrate
conditions. Conversely, for open-field systems, we can recommend the use of ammonia emission
factors from ECETOC (1994) for mineral fertilizers. For ammonia emissions from organic fertilizers,
the method proposed by Brentrup et al. (2000) has the advantage to allow the parameterization of soil
type and plant cover. For ammonia emissions under tropical climate, the emission factors proposed by
Bouwman and Van Der Hoek (1997) seem the most appropriate, and ECETOC (1994) group III can
be used as a complement for fertilizers not dealt with in the article. For nitrous oxide emissions, direct
and indirect emission factors proposed by IPCC (2006) seem the most appropriate. For both ammonia
and nitrous oxide emissions, further developments are required to calibrate more mechanistic tools
easily transposable in Mediterranean and tropical contexts. These tools should at least include
parameters to account for temperature, soil type and application techniques. For the estimation of
nitrate leaching, we can recommend the use of the nitrogen balance proposed by Brentrup et al.
(2000). However, a proper water balance should be done, accounting for the soil and climate
specificities in Mediterranean or tropical contexts, especially considering the common intensive
practices of irrigation in vegetable systems. For on-substrate systems, the methodology developed by
Bres (2009) seems the most appropriate provided that specificities of waste water management are
included (closed and open-loop systems). For each type of N r field emissions, there is a potential for
improvement from using measured data in different contexts to adapting empirical equations, or using
mechanistic and dynamic models to develop generic tools. The soil nitrogen balance should be
expanded to include the dynamics of organic nitrogen from soil, fertilizers and amendments. Indeed,
soil of vegetable gardens often shows high organic matter content due to frequent and important
organic matter inputs. However, difficulties will arise for short crop cycles and particularly in tropical
contexts given the quick turn-over of nitrogen fluxes.
We identified important potentials for improvement in capturing the variability of cropping systems
when assessing the environmental impact of vegetable products through the LCA methodology.
51
However, there will be a trade-off between the additional work required to follow these
recommendations and the benefits in terms of results quality, to be balanced according to users' own
interest. One important limitation remains the lack of quantitative data in certain contexts, such as
developing countries. Data availability has to be included as a constraint for the development of
methods, and different but consistent tiers of estimation tools should be proposed to users, similarly to
the IPCC guidelines for GHG inventories.
1.6. References
Abeliotis K, Detsis V, Pappia C (2013) Life cycle assessment of bean production in the Prespa National Park,
Greece. Journal of Cleaner Production 41:89–96. doi: 10.1016/j.jclepro.2012.09.032
Andersson K, Ohlsson T, Olsson P (1998) Screening life cycle assessment (LCA) of tomato ketchup: a case
study. Journal of Cleaner Production 6:277–288. doi: 10.1016/s0959-6526(98)00027-4
Anton A, Castells F, Montero JI, Huijbregts M (2004) Comparison of toxicological impacts of integrated and
chemical pest management in Mediterranean greenhouses. Chemosphere 54:1225–1235. doi:
10.1016/j.chemosphere.2003.10.018
Antón A, Montero JI, Muñoz P, Castells F (2005a) LCA and tomato production in Mediterranean greenhouses.
International Journal of Agricultural Resources, Governance and Ecology 4:102–112.
Antón MA, Muñoz P, Castells F, Montero JI, Soliva M (2005b) Improving waste management in protected
horticulture. Agronomy for Sustainable Development 25:447–453. doi: 10.1051/agro:2005045
Asman WHA (1992) Ammonia emission in Europe: updated emissions and emission variations. Report n.
2284700, National Institute of Public Health and Environment Protection, Bilthoven, Germany
Audsley E, Alber S, Clift R, Cowell SJ, Crettaz P, Gaillard G, Hausheer J, Jolliet O, Kleijn R, Mortensen B,
Pearce D, Teulon H, Weidema B, van Zeijts H (1997) Harmonisation of environmental life cycle assessment
for agriculture. Final repo, European Comm., DG VI Agriculture, Silsoe, UK
De Backer E, Aertsens J, Vergucht S, Steurbaut W (2009) Assessing the ecological soundness of organic and
conventional agriculture by means of life cycle assessment (LCA) A case study of leek production. British
Food Journal 111:1028–1061. doi: 10.1108/00070700910992916
Bessou C, Basset-Mens C, Tran T, Benoist A (2013) LCA applied to perennial cropping systems: a review
focused on the farm stage. The International Journal of Life Cycle Assessment 18:340–361. doi:
10.1007/s11367-012-0502-z
Birkved M, Hauschild MZ (2006) PestLCI - A model for estimating field emissions of pesticides in agricultural
LCA. Ecological Modelling 198:433–451. doi: 10.1016/j.ecolmodel.2006.05.035
Boulard T, Raeppel C, Brun R, Lecompte F, Hayer F, Carmassi G, Gaillard G (2011) Environmental impact of
greenhouse tomato production in France. Agronomy for Sustainable Development 31:757–777. doi:
10.1007/s13593-011-0031-3
Bouwman AF, Van Der Hoek KW (1997) Scenarios of animal waste production and fertilizer use and associated
ammonia emission for the developing countries. Atmospheric Environment 31:4095–4102. doi:
10.1016/s1352-2310(97)00288-4
Brentrup F, Küsters J, Kuhlmann H, Lammel J (2001) Application of the life cycle assessment methodology to
agricultural production: an example of sugar beet production with different forms of nitrogen fertilisers.
European Journal of Agronomy 14:221–233. doi: 10.1016/S1161-0301(00)00098-8
Brentrup F, Küsters J, Kuhlmann H, Lammel J, Barraclough P (2004) Environmental impact assessment of
agricultural production systems using the life cycle assessment methodology: I. Theoretical concept of a
LCA method tailored to crop production. European Journal of Agronomy 20:247–264. doi: 10.1016/s1161-
0301(03)00039-x
Brentrup F, Küsters J, Lammel J, Kuhlmann H (2000) Methods to estimate on-field nitrogen emissions from
crop production as an input to LCA studies in the agricultural sector. The International Journal of Life Cycle
Assessment 5:349–357. doi: 10.1007/BF02978670
Bres W (2009) Estimation of nutrient losses from open fertigation systems to soil during horticultural plant
cultivation. Polish Journal of Environmental Studies 18:341–345.
Cellura M, Ardente F, Longo S (2012a) From the LCA of food products to the environmental assessment of
protected crops districts: A case-study in the south of Italy. Journal of environmental management 93:194–
208. doi: 10.1016/j.jenvman.2011.08.019
52
Cellura M, Longo S, Mistretta M (2012b) Life Cycle Assessment (LCA) of protected crops: an Italian case
study. Journal of Cleaner Production 28:56–62. doi: 10.1016/j.jclepro.2011.10.021
Cowell SJ, Clift R (1998) Site-Dependency in LCAs Involving Agricultural Production. SETAC-Europe Annual
Conference
ECETOC (1994) Ammonia emissions to air in Western Europe. Report n. 62, European Centre for
Ecotoxicology and Toxicology of Chemicals, Brussels, Belgium
European Commission (2010) International Reference Life Cycle Data System (ILCD) Handbook—general
guide for life cycle assessment—detailed guidance. Joint Research Centre-Institute for Environment and
Sustainability, Luxembourg
Guinée JB, Gorrée M, Heijungs R, Huppes G, Kleijn R, De Koning A (2001) Handbook on life cycle
assessment. Institute of Environmental Sciences (CML), Leiden, The Netherlands
Hauschild M (2000) Estimating pesticide emissions for LCA of agricultural products. In: Weidema B, Meeusen
M (eds) Agricultural data for life cycle assessments. Agricultural Economics Research Institute, The Hague,
pp 64–79
Horlacher D, Marschner H (1990) Schätzrahmen zur Beurteilung von Ammoniakverlusten nach Ausbringung
von Rinderflüssigmist. Zeitschrift für Pflanzenernährung und Bodenkunde 153:107–115. doi:
10.1002/jpln.19901530209
Hospido A, Milà i Canals L, McLaren S, Truninger M, Edwards-Jones G, Clift R (2009) The role of seasonality
in lettuce consumption: a case study of environmental and social aspects. International Journal of Life Cycle
Assessment 14:381–391. doi: 10.1007/s11367-009-0091-7
IPCC (1996) Chapter 4: Agriculture: nitrous oxide from agricultural soils and manure management. In:
Houghton JT (ed) Revised IPCC guidelines for national greenhouse gas inventories. OECD, Institute for
Global Environmental Strategies, Paris, France, pp 4.1–4.140
IPCC (2006) Chapter 11: N2O emissions from managed soils, and CO2 emissions from lime and urea
application. In: Eggleston S, Buendia L, Miwa K, Ngara T, Tanabe K (eds) IPCC guidelines for national
greenhouse gas inventories. Institute for Global Environmental Strategies, Hayama, Japan, pp 11.1–11.54
Margni M, Rossier D, Crettaz P, Jolliet O (2002) Life cycle impact assessment of pesticides on human health
and ecosystems. Agriculture, Ecosystems & Environment 93:379–392. doi: 10.1016/S0167-8809(01)00336-
X
Martínez-Blanco J, Antón A, Rieradevall J, Castellari M, Muñoz P (2010) Comparing nutritional value and yield
as functional units in the environmental assessment of horticultural production with organic or mineral
fertilization. International Journal of Life Cycle Assessment 1–15. doi: 10.1007/s11367-010-0238-6
Martínez-Blanco J, Muñoz P, Antón A, Rieradevall J (2009) Life cycle assessment of the use of compost from
municipal organic waste for fertilization of tomato crops. Resources Conservation and Recycling 53:340–
351. doi: 10.1016/j.resconrec.2009.02.003
Martínez-Blanco J, Muñoz P, Antón A, Rieradevall J (2011) Assessment of tomato Mediterranean production in
open-field and standard multi-tunnel greenhouse, with compost or mineral fertilizers, from an agricultural
and environmental standpoint. Journal of Cleaner Production 19:985–997. doi: 10.1016/j.jclepro.2010.11.018
Menzi H, Katz PE, Fahrni M, Neftel A, Frick R (1998) A simple empirical model based on regression analysis to
estimate ammonia emissions after manure application. Atmospheric Environment 32:301–307. doi:
10.1016/s1352-2310(97)00239-2
Mila i Canals L, Burnip GM, Cowell SJ (2006) Evaluation of the environmental impacts of apple production
using Life Cycle Assessment (LCA): Case study in New Zealand. Agriculture, Ecosystems & Environment
114:226–238. doi: 10.1016/j.agee.2005.10.023
Mila i Canals L, Muñoz I, Hospido A, Plassmann K, McLaren SJ, Edwards-Jones G, Hounsome B (2008) Life
Cycle Assessment of domesctic vs. imported vegetables. Case studies on broccoli, salad crops and grean
beans. Working Paper n. 01/08, Center for environmental strategy University of Surrey, Guilford, UK
Mila i Canals L, Muñoz I, McLaren S, Brandão M (2007) LCA methodology and modelling considerations for
vegetable production and consumption. CES Working Paper n. 02/07, Centre for Environmental Strategy,
University of Surrey, Guilford, UK
Mouron P, Nemecek T, Scholz RW, Weber O (2006) Management influence on environmental impacts in an
apple production system on Swiss fruit farms: Combining life cycle assessment with statistical risk
assessment. Agriculture, Ecosystems & Environment 114:311–322. doi: 10.1016/j.agee.2005.11.020
Nemecek T, Kägi T (2007) Life Cycle Inventories of Agricultural Production Systems. n. 15, Swiss Centre for
Life Cycle Inventories Ecoinvent - ART, Zürich and Dübendorf, Switzerland
Von Rheinbaben W (1990) Nitrogen losses from agricultural soils through denitrification - a critical evaluation.
Zeitschrift für Pflanzenernährung und Bodenkunde 153:157–166. doi: 10.1002/jpln.19901530305
53
Richner W, Oberholzer HR, Freiermuth Knuchel R, Walther U (2006) Modell zur Beurteilung des
Nitratauswaschungspotenzials in Ökobilanzen-SALCA-NO3.
Romero-Gámez M, Suárez-Rey EM, Antón A, Castilla N, Soriano T (2012) Environmental impact of
screenhouse and open-field cultivation using a life cycle analysis: the case study of green bean production.
Journal of Cleaner Production 28:63–69. doi: 10.1016/j.jclepro.2011.07.006
Schmid M, Neftel A, Fuhrer J (2000) Lachgasemissionen aus der Schweizer Landwirtschaft. Schriftenreihe der
FAL 35, 131 p.
Sedilot C, Degas L, Le Quillec S, Brajeul E, Grasselly D (2002) Management of greenhouse effluents: a status
report for France. Infos Ctifl 181:50–52.
Shiralipour A, McConnell DB, Smith WH (1992) Physical and chemical properties of soils as affected by
municipal solid waste compost application. Biomass and Bioenergy 3:261–266. doi: 10.1016/0961-
9534(92)90030-T
Sim S, Barry M, Clift R, Cowell SJ (2007) The relative importance of transport in determining an appropriate
sustainability strategy for food sourcing. International Journal of Life Cycle Assessment 12:422–431. doi:
10.1065/lca2006.07.259
Simmelsgaard SE, Djurhuus J (1998) An empirical model for estimating nitrate leaching as affected by crop type
and the long-term N fertilizer rate. Soil Use and Management 14:37–43. doi: 10.1111/j.1475-
2743.1998.tb00608.x
Stoessel F, Juraske R, Pfister S, Hellweg S (2012) Life cycle inventory and carbon and water FoodPrint of fruits
and vegetables: application to a Swiss retailer. Environmental science & technology 46:3253–62. doi:
10.1021/es2030577
Torrellas M, Anton A, Ruijs M, Victoria NG, Stanghellini C, Montero JI (2012) Environmental and economic
assessment of protected crops in four European scenarios. Journal of Cleaner Production 28:45–55. doi:
10.1016/j.jclepro.2011.11.012
Weber J, Karczewska A, Drozd J, Licznar M, Licznar S, Jamroz E, Kocowicz A (2007) Agricultural and
ecological aspects of a sandy soil as affected by the application of municipal solid waste composts. Soil
Biology and Biochemistry 39:1294–1302. doi: 10.1016/j.soilbio.2006.12.005
54
Contribution du chapitre 1 et perspectives dans le cadre de la thèse

Les principaux résultats du chapitre sont résumé en figure 1.2. Dans le cadre de la thèse, ce chapitre
nous a permis de mieux définir les objectifs et le cadre de notre étude. Nous avons ainsi identifié 2
unités fonctionnelles pertinentes et complémentaires : un kilogramme de produit et un hectare de
production. Nous avons également identifié le périmètre des champs cultivés comme un cadre
pertinent pour une analyse approfondie de la diversité des systèmes de production et de la variabilité
des émissions azotées.
La quantité et la qualité des données nécessaires pour la réalisation des inventaires apparaissent
comme un verrou majeur pour l’application de l’ACV aux produits maraichers, soulignant la nécessité
d’une collecte de données de qualité pour notre cas d’étude que nous mettrons en œuvre dans le cadre
du chapitre 2. Nous avons identifié la typologie comme un moyen de caractériser la diversité des
systèmes de production avec des perspectives intéressantes dans le cadre de l’ACV.
Nous avons également identifié les limites des méthodes d’estimation des émissions azotées
couramment utilisées en ACV. La modélisation biophysique est une piste prometteuse pour raffiner
l’estimation des émissions que nous explorerons dans le chapitre 3.
Enfin, nous avons mis en évidence l’influence des choix méthodologiques sur les résultats d’ACV
(limites du système, allocations), et l’importance de l’aspect multicritères et de l’analyse de sensibilité
pour produire des références fiables et transparentes.

 Unités fonctionnelles
OBJECTIF 1: Etat de l’art ACV des  Limites du système
produits maraîchers
- Faible comparabilité des résultats OBJECTIF 2

d’impacts Chapitre 2 OBJECTIF 3
- Manque de transparence des études Chapitre 3
- Manque de représentativité des données  Données primaires à
- Inadéquation des méthodes vis-à-vis de collecter  Modélisation
la diversité et des spécificités des  Approche de biophysique
systèmes étudiés typologie
- Pas de prise en compte de la variabilité
des émissions
- Standardisation des inventaires par type
de systèmes de culture
OBJECTIF 4
Chapitre 4
 Analyse de sensibilité
 Sélection des catégories d’impact
Figure 1.2 Résultats du chapitre 1 (boite verte) et contributions pour les chapitres suivants (rouge)
55
56
Chapitre 2. Méthodes et données pour une analyse du cycle de vie fiable et
représentative dans un contexte de faible disponibilité en données
Le cas de la tomate en maraîchage urbain au Bénin
© Aurélie Perrin
Traitement phytosanitaire dans un jardin urbain à Cotonou, Bénin
57
- Chapitre 2. Méthodes et données pour l’ACV de la tomate en maraîchage urbain au Bénin -
Avant-propos
Comme souligné en introduction, le premier défi à relever dans le cadre de cette thèse est la création
de données représentatives pour l’évaluation environnementale de la tomate en maraîchage urbain au
Bénin. En effet, contrairement aux grandes cultures dans les pays développés, le maraichage urbain est
une production encore mal connue et peu décrite. Dans le chapitre 2 nous avons identifié l’influence
de la diversité des systèmes de production sur les résultats d’ACV. Au préalable de l’évaluation
environnementale de notre système, il nous a donc semblé nécessaire de mieux comprendre les
systèmes de culture de tomate en jardin urbain au Bénin, leurs spécificités et leurs performances
agronomiques. C’est la démarche que nous proposons dans ce chapitre. Ce passage sur le terrain est
loin d’être systématique dans la pratique de l’ACV. Nous montrerons qu’il est pourtant nécessaire
surtout dans des contextes d’agriculture tropicale où pratiquement aucune donnée n’est disponible
dans des bases de données ou des rapports.
Cette étude s’est déroulée sur la zone côtière du

Bénin. Dans le sud du pays, le climat est de type
équatorial avec 2 saisons sèches et deux saisons
humides. La tomate que nous avons étudiée est
produite pendant la longue saison sèche qui s’étend
de Septembre à Avril. Nous avons étudié une
saison de production à cheval sur les années 2011
et 2012.
Nous nous sommes plus particulièrement intéressés Grand-Popo
aux jardins urbains localisés aux alentours et au
Ouidah
cœur de 3 grandes villes côtières: Cotonou, la
capitale économique, Ouidah et Grand-Popo Cotonou
(Figure 2.1). Nous avons choisi ces zones de
production car elles concentrent les plus hauts
100
niveaux de densité de population.
km
Dans ces jardins urbains on trouve des sols
principalement sableux sauf le long des rivières et

Figure 2.1 Densité de population au Bénin (Center for
lagons ou les teneurs en limon et argiles peuvent International Earth Science Information Network)
être importantes. Ces systèmes sont caractérisés par un usage systématique d’engrais et de pesticides.
Nous avons suivi l’ensemble des opérations culturales pendant environ 100 jours pour chaque parcelle
du stade pépinière à la dernière récolte. Au total cela représente 6 mois de travail de terrain réalisé
avec l’aide précieuse d’un intermédiaire sur place (chercheur Cirad expatrié) et d’un stagiaire
Béninois.
58
Résumé
En Afrique sub-saharienne, l’agriculture urbaine présente un fort potentiel de réduction de la pauvreté
urbaine et d’amélioration de la sécurité alimentaire. Pour faire face à la demande croissante des villes
en produits frais et à la pression foncière, les maraîchers ont intensifié leurs systèmes de production.
Ainsi l’augmentation de l’utilisation d’engrais et de pesticides induit des risques croissants sur
l’environnement et la santé humaine. Afin d’encourager le développement de systèmes maraîchers
urbain et périurbains durables en Afrique, il est capital d’avoir des données fiables sur ces systèmes
(ex. rendement, engrais, pesticides, eau). Ces données doivent être représentatives de la diversité des
pratiques afin de permettre une évaluation environnementale pertinente. En se basant sur l’étude de la
tomate de contre-saison dans les jardins urbains au Bénin (Afrique de l’Ouest), l’objectif de cet article
est d’une part de fournir une méthode permettant de réaliser des Inventaires de Cycle de Vie de bonne
qualité, dans un contexte où les données sont rares, et d’autre part de fournir de nouvelles données
agronomiques pour les systèmes maraîchers urbains d’Afrique.
Une procédure originale basée sur la typologie a été développée et a permis la production de données
d’inventaire de qualité, selon l’index de qualité des données recommandé en ACV. Les ordres de
grandeur obtenus pour les paramètres clés tels que le rendement, l’eau, l’utilisation des engrais et des
pesticides et les émissions aux champs ont été validés par comparaison avec des données mesurées
disponibles dans la littérature. Notre étude démontre pour la période et l’échantillon étudiés que les
pratiques agronomiques ne sont pas optimisées, avec une forte variabilité des rendements et une
utilisation fluctuante et souvent excessive d’engrais et de pesticides. Au regard de ces données
d’inventaire, l’évaluation environnementale de ces produits par l’ACV devrait inclure au minimum :
l’écotoxicité pour les milieux aquatiques et terrestres, l’eutrophisation, l’acidification, le changement
climatique et la consommation d’eau. Enfin, ce travail montre qu’un important effort de recherche est
encore nécessaire afin de mieux connaitre le devenir des nutriments et des pesticides dans les systèmes
de maraîchage urbain Africains.
Mots-clefs : Agriculture urbaine ·Analyse du Cycle de Vie · Typologie · Création de données

·Diversité · Scénario
59
60
Life Cycle Inventory for tomato urban gardens in Benin: low performances
and high environmental risks systems
Aurélie Perrin 1,*, Claudine Basset-Mens 1, Joël Huat 1,2 , Wilfried Yehouessi 3
1
CIRAD, UPR Hortsys, F-34398 Montpellier Cedex 5, France
2
Africa Rice Center, 01 BP 2031, Cotonou, Benin
3
Université d’Abomey-Calavi, Faculté des Sciences Agronomiques, Cotonou, Bénin

Corresponding author e-mail: aurelie.perrin@cirad.fr
Part of this chapter published in Agronomy for Sustainable Development

DOI 10.1007/s13593-014-0241-6
Abstract
To face the increasing demand for fresh products from sub-Saharan African cities and the scarcity of
land, urban vegetable growers intensified their practices leading to potential risks for the environment
and for human health. Comprehensive environmental assessments, representative for a diversity of
agricultural practices are required and should be based on consensual methodology such as Life Cycle
Assessment (LCA). To feed the life cycle inventory (LCI) on which LCA is based, data on production
systems (pesticide use, fertilizer use and field emissions, water use and yield) is needed. Based on the
case study of out-of-season tomato from Beninese urban gardens, the objectives of this paper were (1)
to evaluate the performances and the environmental risks associated to tomato production in Benin and
(2) to assess the quality of LCI data using an LCA-dedicated procedure: the Data Quality Index.
An expert-based typology was used to select among 40 farmers, 12 cropping systems to be monitored
for six months and build representative LCI data to be compared with the literature. Treatment
Frequency Index (TFI) nutrient budgets and field emissions, were calculated based on best available
methods.
We presented LCI data for the main cropping system types and their weighted mean. Pesticide and
fertilizer rates were variable and excessive. The weighted mean TFI for insecticide reached 8.9.
Beyond the large field emissions to the environment, the weighted mean nutrient budget showed large
nutrient surplus: 120 kg N and 84 kg P. ha -1. Yields were low (11,501 t.ha-1 for weighted mean) and
also variable. Overall, practices were not optimized and revealed low performance and high risks for
the environment. According to the data quality index, the LCI dataset was of high quality, enabling a
future environmental assessment with the LCA methodology, which should at least include
ecotoxicity, eutrophication, acidification, global warming and water deprivation potentials.
Keywords Urban agriculture · Life-cycle assessment · Typology · Cropping system data · Treatment
frequency index · Nutrient budget
61
2.1. Introduction
In sub-Saharan Africa, the urban population is projected to increase from 400 million to 750 million
between 2010 and 2030 (UN-HABITAT, 2013). Urban agriculture holds the most potential to feed this
increasing population, curb urban poverty and ensure food security (Abdulkadir et al., 2012). To face
the increasing demand of fresh products from the cities and the scarcity of land available for
agriculture, urban farmers and especially vegetable gardeners intensified their practices (multiple crop
cycles in one year, continued cropping for many years) leading to improper uses of fertilizers and
pesticides (Drechsel & Zimmermann, 2005; De Bon et al., 2010). Such intensive practices started to
lead to deleterious environmental consequences (De Bon et al., 2010; Abdulkadir et al., 2012). In this
concerning context, a comprehensive environmental assessment, representative for a diversity of
agricultural practices for urban horticulture in Africa is required.
Comprehensive environmental assessments for agricultural products have been successfully based on
Life Cycle Assessment (LCA) which is a multicriteria and global approach. LCA is a normalized
methodology (ISO 14040, 2006; ISO 14044, 2006) based on a comprehensive environmental
inventory, or Life Cycle Inventory (LCI), which compiles all resources that are needed for and all
emissions that are released by the specific system under investigation and relates them to the function
of the system. However, this powerful decision helping tool has almost never been used to assess the
environmental impacts of vegetables in tropical contexts (Sim et al., 2007; Mila i Canals et al., 2008).
Applying the LCA framework to tropical vegetables constitutes the most challenging task for
practitioners because of the considerable variations in the resource use and performances between
farms of the same main activity (Haas et al., 2000) and also because of the extreme scarcity of data in
such contexts.
By quantifying all inputs and outputs including emissions to the environment, the LCI step of the LCA
methodology offers a consistent framework to study the performance and environmental risks of urban
horticulture systems. Indeed, emissions to the environment constitute risks, or hazards, for receiving
ecosystems. However to date, the literature on African urban horticulture does not offer sufficient data
including quantified inputs and outputs as required for a proper life cycle inventory. Although several
studies addressed environmental issues of urban horticulture in Africa, they focused either on
fertilizers or pesticides. Regarding pesticides, studies suggested frequent treatments with large rates of
active substances (Rosendahl et al., 2009; Ahouangninou et al., 2012). However to our knowledge, no
paper provided quantified and comprehensive data on crop protection practices for urban cropping
systems in Africa. Regarding fertilizers, large quantities of nutrients in the form of chemical fertilizers
and livestock manure are applied to urban gardens (Diogo et al., 2010; Sangare et al., 2012). Due to
low yields, an important part of nutrient input is lost to the environment (Predotova et al., 2010;
Predotova et al., 2011; Lompo et al., 2012). Nutrient emissions (as NH3, N2O, NO3 and phosphorus)
62
often contribute considerably to the environmental risk associated to urban cropping systems.
However, it is often difficult to estimate rates of nitrogen and phosphorus released to the air and to the
water, as measuring emission rates directly on field is time and cost-demanding, especially when
considering the variability of emissions at small scale (Bussink & Oenema, 1998; Stehfest &
Bouwman, 2006).
Furthermore, recent LCA guidelines have emphasized the importance of evaluating the quality of LCI
data in terms of reliability, completeness and representativeness of the studied function (ILCD
handbook, European Commission, JRC, 2010). An LCA-dedicated procedure is based on the so-called
pedigree matrix including the five quality criteria proposed by Weidema and Wesnæs (1996):
Reliability, Completeness, Temporal correlation, Geographical correlation and Technological
correlation. Although recent papers showed interesting attempts to account for the diversity between
horticulture farms (Cellura et al., 2012; Ingwersen, 2012), most LCA studies still lack a
comprehensive discussion of the representativeness and quality of LCI data and this recommended
procedure appears insufficiently applied. This is particularly detrimental considering the wide
diversity and variability of those systems and the tremendous lack of data.
Based on a comprehensive Life Cycle Inventory procedure applied to a typical African urban garden
case study showing extreme data scarcity, the objectives of this paper were (1) to evaluate the
performances and the environmental risks associated to tomato production in Benin including
cropping systems diversity and one representative scenario, and (2) to evaluate and discuss the quality
of LCI data according to recommended LCA procedure.

2.2.1. Study sites and characterization of cropping systems
Our data collection took place in Southern Benin from September 2011 to April 2012 which
corresponds to the long dry season (tropical climate). In this region, urban gardens are essentially
located within or in the surrounding of three coastal cities, Cotonou, Ouidah, and Grand-Popo, thanks
to the accessibility to water for irrigation and the proximity of urban markets. Those urban gardens
present sandy soils except for those located along the rivers or lagoons where the soil can content high
rate of silt and clay. Among vegetable productions, the out-of-season tomato is one of the highest
value crops grown by urban farmers, and aims at providing cash flow for urban farmers. Indeed the
tomato being the base for many traditional dishes, the important demand leads to high prices on the
market during this unfavorable period for tomato production (low rainfall and high pest pressure).
Unlike the seasonal tomato production mostly located in rural areas, the out-of-season production
requires irrigation. In addition, large amount of fertilizers and pesticides are applied to increase yields
63
and cope with the high land pressure in urban area. Those cropping system specificities are of major
importance for the environmental performance of tomato production in urban gardens in Benin.
With the support of the communal extension services of the three cities, we recorded 40 urban farmers
growing out-of-season tomatoes during the 2011-2012 season (see Annex 2). We implemented a
typology of the out-of-season tomato cropping systems― a cropping system being defined as a crop
grown with a specific soil, climate, and management. This typology addressed a dual-purpose: (1) to
allow the sampling of representative cropping systems for the population to which applying a detailed
monitoring (section 2.2.2), (2) to quantify the representativeness of the data and allow the production
of a unique representative scenario (section 2.2.3). To this aim, we identified three criteria easily
observable and quantifiable on farms―type of irrigation system; type of water resource; and planting
date―with the help of a panel of Tropical horticultural experts from CIRAD (the French research
center for agriculture in developing countries), from INRAB (the Beninese national research institute
of agriculture) and from the governmental extension service (CeRPA) which was identified as the best
mean to access farmers. For each criteria, three modalities were also identified: (1) type of irrigation
system: manual, hose, and sprinkler (being the most advanced system); (2) type of water resource:
depression (the most restricted resource), drilling, and river; (3) planting date: early (September),
median (November), and late (January).
2.2.2. Sampling cropping systems for monitoring

To comply with the need for a good representation of the tomato cropping systems diversity identified
through our typology and the time required for a quality monitoring, the sampling consisted in the
selection of 12 cropping systems. Indeed, to capture the main changes in daily cultural operations and
to record precisely occasional operations, we had to interview farmers every two weeks. We could
visit two farmers in one day and, we needed two days every week to check the data and to fill the
database. Moreover, to ensure the best quality of data, three feasibility criteria were added to the
typology: (1) selecting farmers able to provide detailed information on their practices, (2) selecting
tomato fields at the beginning of the crop cycle and, (3) selecting fields with healthy crops to ensure as
much as possible to reach the harvest stage. The resulting sample of 12 fields was the only possible
sample complying with the need for studying the broader diversity of cropping systems and the
feasibility criteria (see Annex 2).
For these 12 cropping systems, we collected data in two steps. First, farmers were asked to describe
systematically each field operation: date, plot concerned, container, measure and machine used. To
guarantee the accuracy of the data (no language or cultural misunderstanding issues) farmers’
declarations were collected by advisors from the communal extensions services involved in the
64
partnership. Then, a second step consisted in quantifying input flows. Another operator was in charge
of expressing the farmer’s declarations into quantified flows expressed in standard units (using timer,
scales and measuring cups) per hectare (measuring the area of concerned plots).
2.2.3. Cropping systems data for environmental inventories

Inventory data were presented for each modality of each criterion represented in our sample and for
their weighted mean. Data collected from the 12 cropping systems in Benin were aggregated into
modalities with horizontal averaging, i.e. each parameter was considered independent. To obtain the
most representative scenario for the whole population we first averaged data from cropping systems
belonging to the same type obtaining 6 virtual scenarios. Then, the weighted mean was calculated as
the weighted average of these averages for the 6 virtual scenarios according to the relative weight of
their respective corresponding type at population scale on an area-basis. For each virtual scenario, its
relative weight was calculated as the ratio between the cultivated area of the specific type and the sum
of the cultivated areas of corresponding types within the population. To analyze the dispersion of
cropping systems around this weighted mean we calculated the coefficient of variation (CV).
As recommended by Audsley et al. (1997) for an LCA of agriculture, nutrient balances were
calculated using the steady state hypothesis of organic matter stability (nitrogen from soil organic
matter was neglected). Basically, an element balance equation for any system can be stated as
(Oenema et al., 2003): PE IE -OE Eq. 1
where ΔPE, IE and OE stand for the change in the pool, the input and the output of element E,
respectively. Applied to our systems for nitrogen and phosphorus, Eq. (1) can be modified to Eq. (2)
with inputs at the left hand and output at the right hand:
MinFertN P OrgFertN P
RainN P IrrN P
ExportN P
VolatNH3 DenitN2O LeachNO3 P DenitN2 SoilN P Eq. 2
Nitrogen and phosphorus inputs from mineral (MinFert N&P) and organic fertilizers (OrgFert N&P) were
calculated based on farmer’s declaration referring to quantity of fertilizers applied and the nutrient
content of these fertilizers. For our systems we set nutrient inputs from rain (Rain N&P) and irrigation
water (Irr N&P) to 0. This amount could be significant if waste water was used for irrigation (Diogo et
al., 2010). Fresh yields were recorded according to farmer’s declaration. Due to the lack of measured
data, the exportation of nutrients in harvested tomatoes (Export N&P) were assumed to content 0.15% N
and 0.05% P of fresh matter (Diogo et al., 2010). Fertilizer emissions were also estimated, using the
best available methods used in LCA. For ammonia emissions from volatilization of mineral fertilizers
and poultry manure (Volat NH3), emission factors from Bouwman and Van Der Hoek (1997) were used
65
since they correspond to tropical conditions (4% for NPK, 20% for poultry manure and 25% for urea).
We used emission factors from IPCC (2006) to estimate direct (1% of nitrogen inputs) and indirect
(1% of NH3 emitted and 0.75% of NO3 emitted) nitrous oxide emissions from denitrification (Denit
N20) and to estimate nitrate leaching (LeachNO3 being 30% of nitrogen inputs). Given the lack of
specific data on phosphorus emissions in tropical conditions, we followed the recommendations from
Nemecek and Kägi (2007) to estimate phosphorus emissions from leaching and run-off (Leach P). A
part of the nitrogen inputs is emitted as N2 through denitrification (Denit N2) with no potential impact
on the environment. To estimate this pool, we applied the emission factor of 9% of nitrogen input as
suggested in Brentrup et al.(2000). The remaining nutrient pool in the soil (Soil N&P) was further
called nutrient surplus.
For pesticides, applied commercial products were composed by either a unique active substance or by
a mix of 2 active substances (ex. dimethoate / cypermethrin). Amounts of products applied per crop
cycle were recorded according to farmer’s declaration. We calculated the treatment frequency index
(TFI) per crop cycle for each pesticide corresponding to the sum of official European approved rates
applied. In practice, we divided the amount of active substance applied by its official rate. This
indicator allowed us to compare pesticide uses regardless of the nature of the product or the number of
actual applications. For each product, a TFI above 1 either meant that more than one application was
done or that the product was applied at an excessive rate. Generally speaking, the total TFI per target
pest informs about the intensiveness of practices, in terms of number of equivalent official rates.
Values for official European approved rates (OR) of each pesticide are provided to allow potential
LCA practitioners to calculate the pesticide inputs per hectare.
2.2.4. Comparison with scenarios based on the literature

The comparison and the validation of inventory data are essential for producing reliable LCA(Basset-
Mens et al., 2007). Inventory data for our six cropping system types and their weighted mean were
compared with agronomic data and measured field emissions from the literature. Seven published
articles were identified as suitable for comparison, from which we built three scenarios.
We called the first one “Eritrea scenario”, based on agronomic data collected by Asgedom et al.
(2011) from a survey of 146 tomato producers located in five regions in Eritrea. Those five regional
tomato systems differed from our case study as they were not exclusively located in urban gardens and
some were partly rain fed. However, it is to our knowledge the only dataset providing crop lengths
(nursery + plot) and crop densities associated to yields. We aggregated (simple average) data related to
these five tomato systems under the “Eritrea scenario”.
66
The second comparative scenario was the “Tomato scenario” based on two studies (Diogo et al., 2010;
Lompo, 2012). Diogo et al. (2010) collected agronomic data from a 24-month monitoring of ten urban
gardens at Niamey (Niger) between 2006 and 2009. Lompo (2012) collected agronomic data from a
24-month monitoring of two urban gardens (Kodéni and Kuinima) at Bobo Dioulasso (Burkina Faso)
between 2008 and 2010. These systems were very similar to our case study, even though tomato was
preferably grown during the rainy season (March – July) and was irrigated using cans. We extracted
from these two studies data related to six tomato crop cycles (Niger: three high input cycles and one
low input cycle; Burkina Faso: two cycles from Kodéni) and aggregated them under the “Tomato
scenario”.
The last scenario was the “Rotation scenario”, based on three studies dedicated to emission
assessments (Predotova et al., 2010; Predotova et al., 2011; Lompo et al., 2012). These authors
measured gaseous or leached emissions for an annual succession of crops from a selection of
monitored urban gardens from the previous studies, namely Goudel and Yantala in Niger, and Kodeni
and Kuinima in Burkina Faso. For comparison with our data we assumed that a 3-months tomato
cropping cycle (as in our Beninese case study) corresponded to 25% of the length of an annual rotation
(4 crop cycles on average). We extracted and aggregated data related to these four urban gardens and
divided them by a factor 4 under the “Rotation scenario”.
2.2.5. Assessment of the quality of LCI data

We applied to our dataset for urban tomato from Benin the LCA-dedicated procedure recommended
by most recent guidelines to assess the quality of the LCI data (ILCD handbook, European
Commission, JRC, 2010). This procedure is based on the so-called pedigree matrix including five
quality criteria proposed by Weidema and Wesnæs (1996): Reliability, Completeness, Temporal
correlation, Geographical correlation and Technological correlation.
67
2.3. Results and discussion

2.3.1. Diversity of cropping systems sampled
Using the three criteria of the typology, 10 cropping system types representative for the diversity of
out-of-season tomato cropping systems were identified over the 40 farmers of the population (Figure
2.2). The distribution of cropping system types among the population was uneven in terms of
cultivated area. As a result, most modalities were represented by several cropping systems sampled
allowing the calculation of within-group variations except for the Sprinkler modality, which was
excluded from the aggregation procedure. The 12 cropping systems were sampled within six cropping
system types: three belonged to T1, one to T3, one to T5, four to T6, two to T8 and one to T10 (Figure
2.2). Virtual scenarios were calculated for those 6 cropping system types (S1, S3, S5, S6, S8 and S10)
and were aggregated into a weighted mean according to their weight at population scale (Figure 2.2).
68
① ②
Aggregation of cropping Representative
Typology Sampling
systems by modality scenario
IS WR SD IS WR SD
Manual Depression Early T1 1, 2, 3 S1 S1
Surface
1-3
Median T2
Drilling Early 1-4 S3
T3 4 S3
Median T4
Hose Drilling Early
1-5,
12 S5
T5 5 S5
4-9,
12
Median
S6
T6 6, 7, 8, 9 S6
POPULATION = 13 ha
Late
T7
5-11 6-11 S8
River Median
T8 10, 11 S8 S10
Late
10,11
T9
x
Sprinkler Drilling Early

T10 12 12 S10
Figure 2.2 The typology-based procedure to (1) sample 12 diverse cropping systems and (2) create a
representative inventory for out-of-season tomato cropping systems in studied urban areas in Benin (2011-2012 -
Cotonou, Ouidah, Grand Popo). Our typology was based on three criteria easily observable and quantifiable on
farms: irrigation system (IS), water resource (WR), and planting date (PD) of the tomato plots. We ended up
with 10 cropping system types (T1 to T10). Twelve cropping systems (1 to 12) were sampled within 6 types, for
further analysis. Based on sampled types, 6 virtual scenarios were built (S1, S3, S5, S6, S8 and S10) and used to
calculate the representative scenario (weighted mean on an area-basis).
69
Four types were excluded from the sample. T2 and T4, with manual irrigation and crop cycles
occurring during the median period, had high pest attacks at the nursery stage and crops were not
transplanted. Moreover, late crop cycle types (T7 and T9) were not advanced enough at the time of the
data collection process and were therefore discarded. Regarding geographical diversity, T1, T3, T5
and T10 were mainly represented by farmers at Cotonou or Pahou while T6 and T8 were mainly
represented by farmers at Grand-Popo (Table 2.1). At sample scale we observed a gradient of farm
size across cities, resulting from high land pressure at Cotonou to low land pressure at Grand-Popo. To
further investigate the effect of land pressure, being one of the characteristics of urban agriculture, we
will add this criterion for cropping system aggregation and inventory analysis.
Table 2.1 Distribution of the 10 types at the whole population scale (typology based on the irrigation system, the
water resource and the planting date) in terms of area (m²) and number of farms for each district
Cotonou Ouidah Grand Popo Population
Irrigation Water Planting
Area Area Area Area Type
system resource date Farms Farms Farms Farms
(m²) (m²) (m²) (m²)
Manual Depression Early 1278 5 320 1 4478 6 1
Median 702 1 0 702 1 2
Drilling Early 400
1650 2 0 2 5650 4 3
Median 1575 4 800 1 2375 5 4
Hose Drilling Early 288 7 1800 3 30600 10 5
Median 90 1 00
161 3 400 1 16610 5 6
Late 20 11144 3 11144 3 7
River Median 440 1 20000 1 24400 2 8
Late 0 30000 1 30000 1 9
Sprinkler Drilling Median 2064 2 4000 1 6064 3 10
Total 7359 15 573 15 67344 10 13202 40
20 3
We present in sections 2.3.2. to 2.3.4 our field data aggregated by modality for each criterion and the
weighted mean compared to the data of the “Eritrean scenario”, the “Tomato scenario” and the
“Rotation scenario” based on the literature (Diogo et al., 2010; Predotova et al., 2010; Asgedom et al.,
2011; Predotova et al., 2011; Lompo, 2012; Lompo et al., 2012). The specificity of our studied tomato
cropping systems was to be out-of-season, i.e. occurring during the dry season, while tomato cropping
systems from the literature scenarios occurred mostly during the rainy season. Our data were
representative for only one crop cycle of tomato, while the “Rotation scenario” was representative for
one climatic year during which several vegetable cropping systems occurred.
70
2.3.2. Crop protection: frequent applications of pesticides at excessive rates

Crop protection practices revealed high and variable treatment frequency (TFI) indexes especially for
insect control. High TFI values highlighted both high frequencies of application (Table 2.2) and the
application of excessive rates compared to official approved rates. Farmers systematically applied
insecticides (CV=73%accross cropping systems, Table 2.2) while the use of pesticides for disease and
nematode control was more occasional (CV=78% and 127% respectively).
Table 2.2 Average number of applications and associated coefficient of variation per targeted pest: insect,
disease and nematod for cropping systems aggregated by modality for each criterion: irrigation system, water
resource and planting date and for the scenario representative for the whole population (WM).
Insect Disease Nematod
Criterion Modality
Average CV Average CV Average CV
Manual 5.3 75% 2.5 124% -
Irrigation system
Hose 2.4 62% 1.4 114% 0.4 124%
Depression 3.7 79% 3.3 96% -
Water resource Drilling 3.2 103% 1.4 114% 0.3 170%
River 4.0 0% 1.5 141% 1.0 0%
Early 4.5 72% 2.8 87% 0.2 245%
Planting date
Median 2.3 70% 1.0 126% 0.5 109%
Cotonou 5.5 67% 2.8 113% 0.3 200%
Location Grand-Popo 2.3 89% 2.4 89% 0.3 173%
Ouidah 2.4 47% 1.0 100% 0.4 137%
Representative scenario WM 3.7 73% 2.7 78% 0.4 127%
For insects’ control (Figure 2.3), the weighted mean presented a treatment frequency index of 8.2
(CV= 85%). Cropping systems with manual irrigation showed higher TFI than cropping systems with
hose irrigation (12.0 vs. 5.7) systems. Similarly, early cropping systems showed higher TFI than
median ones (9.9 vs. 5.1) and cropping systems from Cotonou showed higher TFI than that from other
locations (11.6 vs. 4.9 and 6.3 respectively). The highest TFI (manual irrigation systems) resulted
from frequent applications at excessive rates per application (5.3 applications at 2.3 fold the official
rate on average, Table 2.2). High TFI resulted mainly from the use of pyrethroids, i.e. cypermethrin,
lambda cyalothrin and deltamethrin (red colors vs. purple colors).
For diseases control (Figure 2.3), the weighted mean presented a TFI of 1.8 (CV=92%). Similarly to
insect control, the modalities manual irrigation systems, water from depression and Cotonou showed
higher TFI than other modalities within the same criterion. The highest TFI (manual irrigation with
TFI = 3.4) resulted more from frequent applications than excessive rates per application (3.4
applications at 1.4 fold the official rate on average, Table 2.2).High TFI resulted from several uses of
maneb, thiophanate methyl or mancozeb (blue colors vs. green colors).
71
72
Nematicides
Insecticides
Fungicides
Figure 2.3 Treatment frequency index (TFI) per targeted pest: insect, disease and nematod for each active substance used for crop protection, for cropping systems aggregated
by modality for each criterion: irrigation system, water resource, planting date and location and for the scenario representative for the whole population (weighted mean).
*Official European approved rates expressed in kg of active substance per hectare were used for calculation of TFIs.
For nematodes control (Figure 2.3), the weighted mean presented a lower TFI (TFI =1.8) compared to
insect control but was highly variable across cropping systems (CV= 156%) partly due to the absence
of treatments for cropping systems with manual irrigation. Cropping systems using water from river
showed higher TFI than cropping systems with drilling (4.3 vs. 1.6). Similarly, median cropping
systems showed higher TFI than early ones (2.5 vs. 0.8). The highest TFI (water from river) resulted
from an excessive rate per application (1 application at 4.3 fold the official rate, Table 2.2). The
highest TFI were associated to the use of carbofuran and terbufos. It seems that the choice of the active
substance to control nematodes depends on the location.
The orders of magnitude of treatment frequency index found in our study are in agreement with the
high number of pesticide applications recorded from studies based on a larger sample of urban
gardeners (superior to 100) in South-Benin (Ahouangninou et al., 2011) and Senegal (Cissé et al.,
2003). To our knowledge this is the first detailed dataset on crop protection practices for tomato
cropping systems in Africa. As suggested by (Ahouangninou et al., 2011) and (Cissé et al., 2003), pest
control practices were likely to result from a lack of knowledge and technical support. Farmers
struggle to identify the pests and to master instructions for application. We noticed that high TFI were
often associated to active substance with low official rates (pyrethroids, thiophanate methyl, maneb,
mancozeb, terbufos and carbofuran). It suggests that farmers had difficulties to apply pesticides at low
rate, most probably due to inaccurate measuring tools and small plots. This finding was in agreement
with observations made on the excessive use of Endosulfan (insecticide) for a large sample of urban
gardeners in South-Benin(Ahouangninou et al., 2012). According to Ahouangninou et al.(2012), the
chlorpyrifos ethyl, the pyrethroids and the mancozeb present particularly high toxicological risks (as a
result of high environment and health risk index). Owing to the wide use of pesticides on out-of-
season tomato in Benin, our inventory study should be completed with a proper environmental
assessment including impacts on freshwater and terrestrial ecotoxicity.
2.3.3. Intensive fertilization management: important N and P surplus

N and P inputs
Overall, N and P inputs as organic and mineral fertilizers were large and in the same range as that for
the literature scenarios (Figure 2.4). More specifically, the weighted mean presented similar total
nitrogen inputs to the “Tomato scenario” (309 kg N.ha -1) and a lower total nitrogen inputs than the
“Rotation scenario” (449 kg N.ha -1), due to lower amounts of mineral nitrogen. Unlike for the
nitrogen, the weighted mean presented a higher amount of total phosphorus inputs than the “Tomato
scenario” (104 vs. 68 kg P.ha-1), due to higher amounts of organic phosphorus. The total phosphorus
input for the weighted mean was slightly lower than the total P input for the “Rotation scenario” (102
kg P.ha-1). It seems that the relatively high phosphorus content of poultry manure was responsible for
high phosphorus inputs in our Beninese systems.
73
The analysis of nutrient input variations between modalities within each criterion was hampered by
large within-modalities variations (up to for 101% for the Depression modality). However, it appeared
that cropping systems from Cotonou showed higher amounts of nitrogen and phosphorus inputs
compared to other locations (711 vs. 447 and 122 kg N.ha -1respectively). This finding pleads for a
correlation between the location of cropping systems, as an indicator of land pressure, and the level of
intensification for fertilization practices. The contribution of mineral fertilizer to the total nitrogen
input varied from 3% (Grand-Popo) to 86% (Ouidah) and from 8% (Grand-Popo) to 83% (water from
river) for the total phosphorus input. For mineral fertilization, the three same types of fertilizer were
recorded across our types and across literature scenarios: NPK, NP and urea. The uneven contribution
pattern of the different nitrogen origins across cropping system types, suggests that the access to
fertilizers was most likely another factor driving management practices.
Figure 2.4 Nitrogen and phosphorus inputs for main nutrient horizontal fluxes at field scale; for cropping system
aggregated by modality for each criterion: irrigation system, water resource, planting date and location and for
the scenario representative for the whole population (weighted mean). We compared them to two scenarios from
the literature: the “Tomato scenario” corresponds to the average of 6 tomato cropping systems from urban
gardens in West Africa (Diogo et al., 2010; Lompo, 2012) and the “Rotation scenario” corresponds to the
average of 4 annual rotations from urban gardens in West Africa (Predotova et al., 2010; Predotova et al., 2011;
Lompo et al., 2012).
74
N and P outputs
Generally speaking, nutrients exportations through harvest were very low compared to inputs but were
larger than that for the “Tomato scenario” (Fig 2.5). More specifically, the weighted mean presented
slightly higher nitrogen exportation than the “Tomato scenario” (14 vs. 10 kgN.ha -1). However, the
weighted mean for nitrogen exported with harvest was lower than that for the “Rotation scenario” (124
kg N.ha-1). Similar pattern was observed for phosphorus exportations. These differences resulted from
the lower nutrient content of tomato compared with other vegetables as included in the “Rotation
scenario”.
Figure 2.5 Nitrogen and phosphorus outputs for main nutrient fluxes at field scale for cropping systems
aggregated by modality for each criterion: irrigation system, water resource, planting date and location and for
the scenario representative for the whole population (weighted mean). We compared them to two scenarios from
the literature: the “Tomato scenario” corresponds to the average of 6 tomato cropping systems from urban
gardens in West Africa (Diogo et al., 2010; Lompo, 2012) and the “Rotation scenario” corresponds to the
average of 4 annual rotations from urban gardens in West Africa (Predotova et al., 2010; Predotova et al., 2011;
Lompo et al., 2012).
75
As a rule, estimated nitrogen emissions were large especially through nitrate leaching and ammonia
volatilization (Figure 2.5) compared to the “Rotation scenario”. It is worth noticing that emissions
were not measured for the “Tomato scenario”. More specifically, cropping systems in Benin showed
greater nitrogen losses from leaching than from volatilization and denitrification. The weighted mean
presented a lower emission factor for NH 3 than the “Rotation scenario” (15 vs. 39% of the total
nitrogen inputs, Table 2.3), despite the high emission factors applied to estimate emissions from urea
and manure taken from a study dealing with tropical contexts (Bouwman and Van Der Hoek, 1997).
However, observed differences between our NH3 estimates and measured NH3 emissions from the
literature were probably not significant compared to the variability of the phenomenon at local scale
(Table 2.3). The weighted mean presented a lower emission factor for N 2O than the “Rotation
scenario” (1 vs. 15% of the total nitrogen inputs). By contrast, the weighted mean presented a higher
emission factor for NO3 than the “Rotation scenario” (30 vs. 1% of the total nitrogen inputs). Using
estimation methods in our case study most likely led to an underestimation of N 2O emissions and an
overestimation of NO3 leaching even considering variations. Finally, phosphorus emissions were very
low for our study as for the “Rotation scenario”. However, this does not anticipate the associated
potential impact on eutrophication, phosphorus being a limiting factor in natural ecosystems.
Owing to the amount of nitrogen emitted to air and water and potential risks associated to P losses, our
inventory study must be completed with a proper environmental assessment including impacts on both
acidification and eutrophication potentials.
Table 2.3 Average and range of variation for emission factors for the scenario representative (weighted mean)
where emissions were estimated based on the literature and for the Rotation scenario where emissions were
measured on field (18-month period)
Emission factors Weighted mean Rotation
kg N or P.100 kg-1 N or P A Variations Averag Variations
applied verage e
N-NH3 1 [4.5-25] 39.0% [7.2-75.4]
5.0%
N-N2O 1 - 14.9% [5.0-27.1]
.3%
N-NO3 3 - 1.0% [0.5-1.4]
0.0%
P-PO4 0 - 0.7% [0.2-1.1]
.9%
Nutrients surplus
Figure 2.6 showed that despite the inclusion of all emissions, the nitrogen and phosphorus surplus
remained large and similar to surplus observed in the literature (Diogo et al., 2010; Lompo, 2012).
This surplus resulted from high inputs compared to crop exportations. It suggests that emissions could
occur beyond the crop cycle duration in relation to the large pools of N and P stored in soils. The
76
commonly used assumption of stability of the soil organic matter appears questionable for tomato
urban gardens which are often recently settled and intensified cropping systems. This result raises the
question of the fate of these nutrients surplus after harvest. It also invalidates the hypothesis of organic
matter stability at cropping system scale. Our nitrogen budget also showed that neglecting N 2
emissions as was the case in the reviewed literature, can lead to large underestimations of nitrogen
outputs.
Figure 2.6 Nitrogen and phosphorus surplus at field scale for cropping systems aggregated by modality for each
criterion: irrigation system, water resource, planting date and location and for the scenario representative for the
whole population (weighted mean). We compared them to two scenarios from the literature: the “Tomato
scenario” corresponds to the average of 6 tomato cropping systems from urban gardens in West Africa (Diogo et
al., 2010; Lompo, 2012) and the “Rotation scenario” corresponds to the average of 4 annual rotations from urban
gardens in West Africa (Predotova et al., 2010; Predotova et al., 2011; Lompo et al., 2012).
2.3.4. Variable and low tomato yields

Overall, the crop yields for Beninese tomato were low and variable but in the same range as the yields
observed in literature scenarios (Table 2.4). The weighted mean presented a slightly lower yield than
the “Tomato scenario” (9,533 vs. 10,342 kg.ha-1) and a slightly greater yield than that the “Eritrean
77
scenario (7,300 kg.ha-1). However, any modality reached the yield of the weighted mean which results
from the high weighting factor associated to a cropping system presenting a yield twice as great as
other yields within the sample. The location and the planting date appear to be the most discriminant
criteria, with early cropping systems reaching higher yields than median cropping systems and
cropping systems from Ouidah showing the lowest yields.
The plot density weighted mean for our types was similar to that for the “Eritrean scenario” (4.9 vs.
5.4 plants.m-²). However, compared to similar tomato cropping systems from the literature (Asgedom
et al., 2011), the short crop lengths (78 vs. 105 days) indicated that farmers had to reduce their crop
cycles, limiting the potential yield. Cropping systems from Ouidah showed short crop lengths and the
lowest yields, due to random pest attacks. If no effective pesticide for these massive attacks was
available (Tetranychus evansi also called the red spider mites for example) the farmer may lose the
whole harvest.
Table 2.4 Cropping system data over one tomato crop cycle for cropping systems aggregated by modality for
each criterion: irrigation system, water resource, planting date and location and for the scenario representative
for the whole population (WM). We compared them to two scenarios from the literature: the “Eritrea scenario”
corresponds to the average of 5 tomato cropping systems in Eritrea (Asgedom et al., 2011), the “Tomato”
corresponds to the average of 6 tomato cropping systems from urban gardens in West Africa (Diogo et al., 2010;
Lompo, 2012). Some data were not available in reviewed papers (na).
Nursery
length Plot lenght Plot density Fresh yield
Criterion Modality n (days) (days) (plants.m-2) (kg.ha-1)
Average CV Average CV Average CV Average CV
Irrigation system Manual 4 43 33% 76 10% 3.4 44% 6309 83%
Hose 7 23 28% 74 28% 4.2 41% 5123 144%
Water resource Depression 3 36 15% 75 12% 3.7 48% 5787 108%
Drilling 7 29 58% 76 26% 4.1 43% 5219 147%
River 2 28 15% 68 23% 3.4 2% 3604 35%
Planting date Early 6 38 37% 77 13% 4.1 48% 7733 105%
Median 6 23 30% 71 29% 3.7 30% 2450 94%
Location Cotonou 4 45 28% 73 11% 3.9 27% 6307 83%
Ouidah 5 23 33% 69 32% 3.0 28% 1701 147%
Grand-Popo 3 25 18% 85 8% 5.5 37% 9121 115%
Scenarios WM 12 27 48% 78 20% 4.9 32% 9533 66%
Eritrea 5 31 13% 105 14% 5.4 30% 10342 26%
Tomato 4 na na na na na na 7300 64%
78
2.3.5. Prospects for the Life cycle assessment of out-of season tomato in Benin
Using the five quality criteria of the pedigree matrix from Weidema and Wesnæs (1996), we obtained
good scores for our data set (Table 2.5).
Reliability: The raw data were collected by local advisors directly from farmers’ declarations. Most
farmers could provide detailed technical data although they had no records. Due to the diversity of
measure units used by farmers, the difficulty relied in the conversion of farmer’s declarations into
quantified data expressed in standard units per hectare to obtain a homogeneous dataset. The
quantification of input flows (section 2.2.2) on the farm allowed a satisfactory level of data reliability
regarding the pedigree matrix criteria.
Completeness: We collected a complete dataset describing agricultural practices and resource use from
sowing to the last harvest. The cropping system selection allowed us to reach a fair representativeness
of the practices from the 40 identified farmers while maintaining a reasonable number of cropping
systems monitored with regard to time and labor constraints. With a capacity of monitoring 12
cropping systems, we could only cover 30% of existing cropping systems (on the basis of 40 farmers =
40 cropping systems). However, 70% of the farmers presented one of the six cropping system types
studied (Table 2.1) which represented 60% of the cultivated area of our farmers’ sample. A first
limitation of our dataset is the sample size compared to the high diversity of cropping systems within
the population. Some types being represented by only one cropping system, it excluded the possibility
to analyze the cropping systems’ variability within each type. We thus analyzed the effect of
modalities for each criterion independently. We highlighted interesting trends regarding practices
across modality and the weighted mean succeeded in providing an average profile regarding most of
the criteria. However, it appeared that the representative scenario (weighted mean) led to an
overestimation of the average yield for the population (Table 2.4). The second limitation of our dataset
refers to the lack of consideration for annual variations due to climate or socio-economic parameters.
However, compared to the scenarios based on the literature aggregating systems from several years
and gardens, the weighted mean showed comparable and often higher coefficients of variation. We
therefore most likely succeeded in capturing an important part of the inherent variability of cropping
systems through our typology.
Temporal, geographical, and technological correlations: The studied cropping systems presented the
three technical modalities of irrigation systems and sources of water. The distribution of modalities
within the population was respected in the sample, with a small over-representation of hose systems
and depression-sourced water systems. The main limitations were first our inability to study cropping
systems representative for the late planting date modality and second, the reduced number of cropping
systems with sprinkler irrigation system.
79
Following our typology, we succeeded at producing good quality data on cropping systems producing
out-of-season tomato in Benin to be used for life-cycle assessment (LCA). Such dataset could only be
achieved thanks to strong and multiple partnerships with locally established research organizations,
extension services, and farmers. The two first partners provided expertise to cope with the lack of
bibliography and selected criteria for the typology. They also provided a precious repertory of farmers
which guaranteed an unbiased selection of cropping systems, along with expertise and references to
ensure the quality of data. Finally, the farmers were able to answer to our questions although they had
no written records.
Table 2.5 Data quality indicators (Weidema and Wesnaes, 1996) for our LCI data set for out-of-season tomato in
Benin. The best score is 1, the worst is 5.
Quality
Score Data quality goal Performances and limitations
Indicator
Reliability 1/5 “Verified data based on Farmer’s declaration collected by local advisors
measurements” Straightaway quantification and checking
Completeness 3/5 “Representative data from an Representative for 70% of farmers’ practices
adequate number of sites but from The variability between cropping systems types and
shorter periods” between year could not be quantified
Temporal 1/5 “Less than three years of difference Data collected especially for the study
correlation to year of study” Homogenous period (out-of-season)
Geographical 1/5 “Data from the area under study” Representative for 60% of the production area
correlation
Technological 3/5 “Data from processes and materials Representative for 60% of types based on
correlation under study but from different technological criteria identified with an expert panel
technology” Sprinkler irrigation systems and late crop cycles
could not be analyzed
2.4. Conclusions
In this paper, we presented cropping system data for representative environmental inventories of one
hectare of out-of season tomato in Benin. We used a typology to sample 12 cropping systems within a
population of 40 farmers. We analyzed the different inventory data for the range of modalities for each
criterion of the typology and for their weighted mean and compared them to comparable data from
scenarios based on the literature. We then assessed the resulting inventory data according to
consensual quality criteria as recommended by the ILCD Handbook from the European Commission.
Pesticide and fertilizer rates were variable and excessive. The weighted mean TFI reached 8.2 for
insecticides and 1.8 for both fungicides and nematicides. TFI were highly variable between cropping
systems and within modalities. However, it appeared that cropping systems with either manual
irrigation, early crop cycle or being located at Cotonou presented more intensive pest and disease
control practices. Our results confirmed that pest control practices were likely to result from a lack of
80
knowledge and technical support, as most of high TFI resulted from excessive application rate and/or
frequent application of the same substance. It appeared that farmers encountered difficulties at
applying pesticides at low rate. Beyond the large field emissions to the environment, the weighted
mean nutrient budget showed large nutrient surplus: 120 kg N and 84 kg P. ha -1. These large surpluses
resulted from high nutrient inputs compared to crop exportations. Our results suggested a correlation
between the location of cropping systems, as an indicator of land pressure, and the level of
intensification for fertilization practices. Yields were low (11,501 t.ha -1 for the weighted mean) and
also variable, due to random pest attacks. Overall, practices were not optimized and revealed low
performance and high risks for the environment.
Our inventory data for out-of-season tomato in Benin were of good quality according to commonly
used criteria in LCA. Allowing the selection of 12 representative cropping systems over a population
of 40 farmers and their aggregation into a weighted mean, the typology allowed covering most of the
geographical and technical diversity characterizing the population. The environmental assessment of
those production systems should at least include freshwater and terrestrial ecotoxicity, eutrophication,
acidification, global warming and water deprivation potentials.
2.5. References
Abdulkadir A, Dossa LH, Lompo D-P, Abdu N, van Keulen H (2012) Characterization of urban and peri-urban
agroecosystems in three West African cities. International Journal of Agricultural Sustainability 10:289–
314. doi: 10.1080/14735903.2012.663559
Ahouangninou C, Fayomi BE, Martin T (2011) Évaluation des risques sanitaires et environnementaux des
pratiques phytosanitaires des producteurs maraîchers dans la commune rurale de Tori-Bossito (Sud-Bénin).
Cahiers Agricultures 20:216–222.
10.4236/jep.2012.33030
Asgedom S, Struik PC, Heuvelink E, Araia W (2011) Opportunities and constraints of tomato production in
Eritrea. African Journal of Agricultural Research 6:956–967.
15:1395–1405. doi: 10.1016/j.jclepro.2006.03.009
10.1016/s1352-2310(97)00288-4
81
Cellura M, Ardente F, Longo S (2012) From the LCA of food products to the environmental assessment of
208. doi: 10.1016/j.jenvman.2011.08.019
10.1007/s10705-009-9315-2
Drechsel P, Zimmermann U (2005) Factors influencing the intensification of farming systems and soil-nutrient
management in the rural-urban continuum of SW Ghana. Journal of Plant Nutrition and Soil Science
168:694–702. doi: 10.1002/jpln.200521775
Haas G, Wetterich F, Geier U (2000) Life cycle assessment framework in agriculture on the farm level. The
International Journal of Life Cycle Assessment 5:345–348. doi: 10.1007/BF02978669
Ingwersen WW (2012) Life cycle assessment of fresh pineapple from Costa Rica. Journal of Cleaner Production
35:152–163. doi: 10.1016/j.jclepro.2012.05.035
ISO 14040 (2006) Environmental management - Life cycle assessment - Principles and framework. International
Organization for Standardization, ISO 14040:2006
Lompo DJ-P (2012) Matter flows and balances in urban vegetable gardens of Bobo Dioulasso, Burkina Faso
(West Africa). PhD thesis, University of Kassel
Lompo DJ-P, Sangaré SAK, Compaoré E, Papoada Sedogo M, Predotova M, Schlecht E, Buerkert A (2012)
Gaseous emissions of nitrogen and carbon from urban vegetable gardens in Bobo-Dioulasso, Burkina Faso.
Journal of Plant Nutrition and Soil Science 175:846–853. doi: 10.1002/jpln.201200012
Oenema O, Kros H, de Vries W (2003) Approaches and uncertainties in nutrient budgets: implications for
nutrient management and environmental policies. European Journal of Agronomy 20:3–16. doi:
10.1016/S1161-0301(03)00067-4
10.1002/jpln.200900255
10.1016/j.fcr.2009.09.010
Rosendahl I, Laabs V, Atcha-Ahowe C, James B, Amelung W (2009) Insecticide dissipation from soil and plant
surfaces in tropical horticulture of southern Benin, West Africa. Journal of environmental monitoring
11:1157–1164. doi: 10.1039/b903470f
Sangare S, Compaore E, Buerkert A, Vanclooster M, Sedogo M, Bielders C (2012) Field-scale analysis of water
and nutrient use efficiency for vegetable production in a West African urban agricultural system. Nutrient
Sim S, Barry M, Clift R, Cowell SJ (2007) The relative importance of transport in determining an appropriate
sustainability strategy for food sourcing. International Journal of Life Cycle Assessment 12:422–431. doi:
10.1065/lca2006.07.259
UN-HABITAT (2013) UN-Habitat Global Activities Report 2013. Available from: http://www.unhabitat.org.
82

Ce chapitre présente le protocole de collecte des données qui seront utilisées dans la suite de la thèse
pour réaliser l’ACV de la tomate produite en jardin urbain au Bénin (Figure 2.7). Pour chaque parcelle
suivie, nous avons enregistré pour chaque opération culturale : la date, les produits utilisés et les doses
déclarés sous la forme des instruments de mesure utilisés par le producteur: une capsule, une poignée,
un sceau (les protocoles d’enquête sont visibles en annexe 3). L’ensemble des données collectées est
résumé en annexe 4 et un itinéraire technique complet est proposé en annexe 5. Enfin nous avons
également réuni des données pédoclimatiques: 1 analyse de sol par parcelle (Annexe 6) et les données
pour la station météo de Cotonou (Annexe 7). Ces données seront reprises pour simuler les rendements
et les émissions dans le chapitre 3, puis dans le chapitre 4 (Figure 2.7).
Le chapitre 2 met par ailleurs en application les recommandations du chapitre 1 pour l’estimation des
émissions d’azote réactif et aboutit à la définition d’une méthode de facteurs d’émission par défaut
(DEF) qui sera ensuite utilisée dans le cadre du chapitre 4 pour produire les résultats d’ACV dit ‘de
référence’ (Figure 2.7). Nous avons également mis en évidence un corpus bibliographique portant sur
les bilans de flux de nutriments à l’échelle de la parcelle. Ces références seront reprises dans le
chapitre 3 pour analyser la pertinence des émissions simulées par le modèle biophysique (Figure 2.7).
Enfin, ce chapitre permet d’aboutir à un niveau d’agrégation des parcelles basé sur 4 critères
discriminants (le système d’irrigation, l’origine de l’eau, la date de repiquage et la localisation
géographique) et à un scénario représentatif de la population basée sur la moyenne pondérée. Ces
scénarios seront utilisés dans le cadre du chapitre 4 pour produire les résultats d’ACV.

Tomate produite en jardin urbain au Bénin
OBJECTIF 2 : Prise en compte de la diversité et OBJECTIF 3

inventaires représentatifs Chapitre 3
- Protocole de collecte de données pour réaliser  Données complètes pour

des inventaires représentatifs modéliser les systèmes de
OBJECTIF 1
Chapitre 1
- Méthode d’estimation par défaut (DEF) culture (pratiques + sol)

- ICV des systèmes de cultures pour 10 modalités  Références bibliographiques
représentatives de 4 critères discriminants et 1 pour les émissions
moyenne pondérée
OBJECTIF 4
Chapitre 4
 Identification de critères discriminants pour les pratiques: système d’irrigation, origine de

l’eau, date de repiquage et localisation géographique
 Définition d’un scénario représentatif de la population
83
84
Chapitre 3. Boucler le bilan d’azote à l’échelle du système de culture :
perspectives pour l’estimation des émissions d’azote réactif pour la tomate en
maraîchage urbain au Bénin
© Aurélie Perrin
Carré de récolte sur une parcelle de tomate en jardin urbain à Cotonou, Bénin
85
- Chapitre 3. Boucler le bilan d’azote à l’échelle du système de culture -
Avant-Propos
L’analyse de la bibliographie dédiée aux flux d’azote réactif réalisée dans le chapitre 1 nous a permis
d’identifier l’espace de complexité qui existe entre les caractéristiques des systèmes de culture et les
émissions. Le tableau 3.1 présente la liste des facteurs susceptibles d’influencer les émissions d’azote
réactif dans le contexte d’étude, établie à partir de la bibliographie sur les émissions. Confronter cette
liste aux outils de modélisation disponibles, nous a permis d’identifier un modèle candidat permettant
d’intégrer un maximum de facteurs influents et pouvant être utilisé avec les données disponibles. Pour
notre application, nous avons choisi le modèle STICS car il permettait de simuler l’ensemble des
émissions d’azote réactif en plus des exportations d’azote par la plante, en prenant en compte de
manière robuste une grande diversité de conditions pédoclimatiques et de pratiques y compris celles de
nos systèmes. Nous présentons dans ce chapitre l’exploration de l’usage de ce modèle pour l’ACV de
notre cas d’étude.
Tableau 3.1 Facteurs influençant la variation des émissions azotées au champ pour les cultures maraichères
Facteurs influents NH3 N2O NO3
Sol Taux d’infiltration  
pH, CECa   
Teneur en calcaire et en argile 
Température du sol  
Taux de matière organique 
Norg b, [NH4]c, [NO3]c  
Potentiel de dénitrification 
Capacité au champ, Point de flétrissement permanant  
Densité apparente 
Climat Déficit de saturation /Pression de vapeur saturante  
ETPd 
Température de l’air 
Albédo  
Pratiques [NH4], [NO3] des engrais minéraux   
MO, Norg, [NH4], [NO3] des engrais organiques   
Date et fractionnement de la fertilisation   
Méthode d’incorporation engrais 
Date et fractionnement de la fertilisation l’irrigation   
Drainage et ruissellement induit par les pratiques 
LAIe  
Norg, [NO3] des résidus 
Azote absorbé par la plante  
a. Capacité d’échange cationique, b. azote organique, c. teneur en azote ammoniacal et nitrique à l’état initial, d.
évapotranspiration potentielle, e. Indice foliaire (Leaf Area Index)
86
Résumé
Dans le but d’évaluer l’impact environnemental des productions maraichères périurbaines en contexte
africain, il est nécessaire de connaitre les flux d’azote réactif et leurs interactions avec le climat, le sol
et les pratiques. Pour des systèmes sur lesquels les données sont rares et les mesures difficiles à mettre
en place, l’approche par modélisation est souvent incontournable. Une approche combinant un bilan
d’azote à l’échelle de la parcelle et l’estimation des flux d’azote par un modèle biophysique a été
appliquée à nos systèmes de production de tomate au Bénin. Les objectifs de cette approche sont (1)
d’identifier les facteurs explicatifs des flux d’azote (2) de comparer les bilans azotés de 12 systèmes de
production de tomate (3) d’identifier les incertitudes et de fournir des recommandations pour
l’estimation des flux d’azote réactif en contexte tropical.,
La modélisation des flux d’azote réactif a été réalisée à l’aide du modèle STICS, qui a l’avantage
d’être utilisable en contexte tropical et de fournir des données de flux détaillées réutilisables pour
calculer un bilan d’azote à l’échelle de la parcelle. Les variations entre les entrées et les sorties du
modèle ont été analysées à l’aide de régressions linéaires simples et multivariées. Les résultats obtenus
montrent la difficulté à modéliser les rendements, qui sont globalement surestimés par rapports aux
rendements observés. Ceci s’explique par une forte pression biotique (maladies et ravageurs) mal
contrôlée entrainant une faible corrélation entre les facteurs de production et la performance
agronomique des systèmes. Le modèle STICS ne permet pas d’estimer les pertes de rendement
associées aux stresses biotiques. Nous avons pu identifier 4 facteurs principaux ayant une influence
significative sur les émissions d’azote réactif : la dose d’azote organique, le volume d’eau d’irrigation,
le pH du sol et la capacité au champ. Nous avons également mis en évidence des incertitudes associées
à la modélisation des processus de nitrification et de dénitrification.
Nous pouvons conclure que l’intégration de ces 4 facteurs influents des émissions d’azote réactif
permet de prendre en compte une part significative de la variabilité des émissions au champ dans notre
cas d’étude. Notre analyse montre que la collecte d’un certain nombre de données spécifiques relatives
à la variété de tomate, au potentiel de dénitrification des sols et aux engrais organiques permettrait
d’améliorer la qualité des sorties du modèle. Néanmoins les incertitudes issues de l’absence de prise
en compte des pertes biotiques par le modèle STICS et issues du manque de jeux de données pour
valider les processus de nitrification, de dénitrification, et de devenir de l’azote minéral du sol après
récolte sont les principales limites pour l’utilisation de ce type d’approche pour estimer finement les
émissions au champ dans notre contexte d’étude.
Mots-clefs : Bilan azoté  Modèle biophysique  Azote réactif  Variabilité  Incertitude
87
88
Closing the nitrogen budget for tomato cropping systems in African urban
gardens
Prospects for Nr emission estimation in LCA
3.1. Introduction
Large quantities of nutrients in the form of chemical fertilizers and livestock manure are applied to
urban gardens (Khai et al., 2007; Diogo et al., 2010; Sangare et al., 2012). Due to low yields, an
important part of nitrogen input is lost to the environment (Predotova et al., 2010; Predotova et al.,
2011; Lompo et al., 2012). Reactive nitrogen emissions (Nr emissions being NH3, N2O, NOx and NO3)
often contribute considerably to the final results of the Life-Cycle Assessment (LCA) studies of
vegetable products (chapter 1). However, it is often difficult to derive exact rates of N released to the
air and to the water, because emission rates can greatly vary depending on soil type, climatic
conditions and agricultural management practices. Brentrup et al. (2000) recorded three ways of taking
these Nr emissions into account:
 To measure actual emission rates caused by the system under consideration,

 To use values derived from literature in a case-by-case procedure,
 To estimate potential emission rates using structured estimation methods.
The option of measuring emission rates directly on field is often discarded in LCA due to time and
cost constraints as well as issues regarding variability of emissions at small scale (Bussink & Oenema,
1998; Stehfest & Bouwman, 2006). Deriving values from the literature requires the identification of
representative references for the system. However, we found out that the best available references
resulted in the estimation of emissions much different from measurements in similar systems (chapter
2, Table 2.3). Without better knowledge on drivers for emissions in our contexts we cannot rely on
those measurements. Advantages of the last option (i.e. using structured estimation methods) are their
easy performance, implying less effort compared to do site –specific measurements and often their
better representativeness to the studied system compared to values derived from the literature
(Brentrup et al., 2000).
According to Audsley et al. (1997) structured estimation method should be based on the calculation of
mineral balance to ensure that the total input is equal to the total output and thus reduce the error on
the more uncertain component of the budget, i.e. Nr emissions. In line with the LCA framework,
spatial and time limits must be set to differentiate the system under analysis from its environment
(SETAC, 1998). Due to short crop cycles and several crop successions within a year, temporal limits
89
on a year-basis were found unsuitable for vegetable crops (chapter 1). To be in line with the literature
in agricultural sciences, the notion of “cropping system” (Sebillotte, 1990) can be applied as a set of
management procedures applied to a given, uniformly treated agricultural area, setting the atmosphere
and surrounding water pieces (groundwater and surface waters) as part of the environment. The
nitrogen budget at cropping system scale records all nitrogen inputs and outputs, including nitrogen
gains and losses within and from the crop-soil system.
Oenema et al. (2003) defined two main sources of uncertainty when calculating nutrient budgets:
fundamental uncertainties related to the completeness and the reliability of the method to calculate
budget components, and operational uncertainty, arising from the lack of data and knowledge, the
variability in space and time, and the changes in items and parameters. Conditions which influence the
nitrogen emissions are reflected by certain parameters (soil, climate, agricultural practice) which
should be available and used as an input for the estimation methods (Brentrup et al., 2000). We found
out that most up-to-date knowledge on Nr dynamics was embodied in modeling tools allowing the
simulation of Nr emissions (chapter 1, section 1.4.4). Several LCA studies already used dynamic crop
models to take into account local soil and climatic factors for the estimation of field-emissions,
especially regarding energy crops and manure management where Nr emissions are crucial for the life
cycle impact assessment (Adler et al., 2007; Gabrielle & Gagnaire, 2008; Langevin et al., 2010;
Bessou et al., 2013; Dufossé et al., 2013). However, none of them had to face large uncertainty issues
due to a context of extreme data scarcity associated with variable agricultural practices.
Based on the case study of tomato production in Beninese urban gardens, we applied the crop model
STICS to simulate nitrogen fluxes at cropping system scale and complete a nitrogen budget at
cropping system scale. Our objectives were: (1) identifying main influencing factors for nitrogen
fluxes and emission factors, (2) comparing the nitrogen budgets from 12 tomato cropping systems, (3)
identifying uncertainty issues and provide with recommendations for improving the reliability of Nr
emissions estimates.
90

3.2.1. An approach combining a crop model and a nitrogen budget
Selection of the crop model
Several crop models allows the simulation of Nr emissions such as NGAS-DAYCENT (Parton et al.,
1996; Parton et al., 1998; Del Grosso et al., 2000), DNDC (Li et al., 1992; Li, 2000) or, CERES-NOE
(Hénault et al., 2005; Gabrielle et al., 2006). The last one can be liken to the STICS model as it
includes similar components for the simulation of gaseous emissions. Advantages of the STICS model
are: (1) to be a generic and robust model based on mechanistic formalisms and able to simulate various
soil and climate conditions with reasonable accuracy (Brisson et al., 2002), (2) to allow the simulation
of tomato yields (Valdés-Gómez et al., 2009), (3) to be adaptable to tropical contexts (Sierra et al.,
2003; Scopel et al., 2004; Sierra et al., 2010). STICS shown a fair performance in simulating carbon
and nitrogen dynamics in arable soils (Gabrielle et al., 2002) and has already been used and to
simulate nitrate leaching (Schnebelen et al., 2004; Jégo et al., 2008) and gaseous emissions (Langevin,
2010). Data required to run the model are accessible and relate to climate, cropping system (crop type
and rotation, and agricultural practices) and soil properties (unchanging soil attributes and initial water
and nitrogen profiles).
STICS simulates both agronomic variables (leaf area index, biomass, yield and input consumption)
and environmental variables (soil profile water and contents, water drainage, nitrate leaching at the
base of the soil profile, ammonia volatilized and nitrous oxide emissions). Figure 3.1 presents the
conceptual model for the four output variables of interest for the case study. Nitrogen transformations
and transfers simulated in STICS are: volatilization, mineralization and immobilization, nitrification,
denitrification, leaching and crop exportation. Volatilization from mineral and organic fertilizers leads
to NH3 emissions. Nitrification (NH4  N2O  NO2  NO3) and denitrification (NO3  NO2 
NO  N2O  N2) processes lead to N2O, and eventually to N2 emissions. Leaching lead to NO3
emissions and yield formation leads to nitrogen exportations with harvested fruits. The formalisms
responsible for Nr emissions will be further described in section 3.2.2. Within STICS the
mineralization of soil organic matter is affected by soil relative water content (SrWC=soil water
content/field capacity), as described in Rodrigo et al.(1997), the soil temperature as described in Valé
(2006) and soil properties (bulk density, calcareous and clay content). STICS also simulates the
mineralization of organic residues (i.e. organic fertilizers in our case) as described in Nicolardot et al.
(2001) depending on organic fertilizers properties (see section 3.2.3). Both mineralization processes
lead to nitrogen inputs to the system, while immobilization leads to nitrogen outputs. Finally, the
nitrogen contained in exported fruits is calculated within the process of yield formation. This process
is based on the phenology development influenced by the temperature and the biomass production
influenced by the radiation, and is limited by stresses influenced by water and nitrogen availability.
91
Figure 3.1 Conceptual model for output variables of interest (circles) as simulated in STICS: main influencing
processes (white rectangle) and driving parameters (grey rectangles)
Calculation of the nitrogen Budget

Basically, an element balance equation for any system can be stated as (Oenema et al., 2003):
PE IE - OE Eq. 1
where ΔPE, IE and OE stand for the change in the pool, the input and the output of element E,
respectively. Khai et al. (2007) applied Eq. (1) for the field balance of nutrients in an agricultural
system. The input flows were derived from seedlings (S E), irrigation water (IWE), rainwater (RWE),
fertilizers (FE), atmospheric deposition (ADE), symbiotic nitrogen fixation (NE), the output flows from
harvested organs (HE), leaching (LE), runoff (RE), air emission (AEE), and ΔPN was the net change in
soil storage of mineral nitrogen (ΔSoilE). Thus, Eq. (1) can be modified to:
SoilE SE IWE RWE FE ADE SFE -(HE LE RE AEE ) Eq. 2
Soil nitrogen budgets were calculated over the cropping system period using Eq. 2 proposed by Khai
et al. (2007). Applying Eq. (2) using variables simulated by the crop model STICS (Figure 3.1), we
first had to include input from mineralization of soil organic matters (MH N), output from
immobilization into soil organic matter (IHN). We also detail nitrogen fertilizers as organic fertilizers
92
(FON) and mineral fertilizers (FMN). Then we refined atmospheric emissions (AEN, eq. 3) as ammonia
from volatilization of organic (VON) and mineral (VMN) fertilizers, nitrous oxide from nitrification
(NitN) and denitrification (DenitN) and, dinitrogen from denitrification (DenitN2). Even if N2 is not of
environmental relevance, N2 rates emitted to the air should be included in the nitrogen balance.
Indeed, it is the main product of denitrification and usually released in much higher rates than N2O.
NOx emitted through nitrification (NitNO) was the only Nr emission which was not simulated by the
crop model STICS. However STICS simulated the processes of nitrification and we can estimated on
average that 0.5% of nitrified nitrogen is emitted as NOx (Rolland et al., 2010).
AEN Denit N2O Denit N2 Nit N2O Nit NO VO NH3 VM NH3 Eq. 3
Finally, considering the high levels of nitrogen fertilizers rates and in agreement with several studies
dealing with nitrogen budgets of urban gardens in Africa (Lompo, 2012; Sangare et al., 2012) we set
nitrogen amounts from seedlings (SE), irrigation water (IWE), rainwater (RWE) and atmospheric
deposition (ADE) to zero. Thus Eq. (2) can be modified to Eq. (4) with inputs at the left hand and
output at the right hand:
FO N FM N MH N
IH N H N Denit N2O Den N2 Nit N2O Nit NO VO NH3 VM NH3 L NO3 Soil N Eq. 4
3.2.2. Calculation of Nr emissions in STICS

This section is limited to a description of the STICS model (version 11) for specific variables of
interests within this study. The reader can refer to Brisson et al. (2008) for more information about
conceptual basis, formalizations and parameterization.
Volatilization
The volatilization of ammonia (NH 3) from nitrogen-based fertilizers is divided into two processes
calculated on a daily-step basis (d): volatilization from mineral fertilizers and from organic residues.
First, the volatilization rate from mineral fertilizers (VM N, eq.5) depends on the crop nitrogen uptake
rate (VABSMOY), the amount of mineral nitrogen applied (FM) and soil pH (FPH) according to the
following relationship:
VABS2P
VMNH3 (d) VOLENGP (f) VABS2P VABSMO (I)
FM (d) FPH Eq. 5
with VOLENGP a fertilizer specific parameter defining the maximal proportion of nitrogen losses by
volatilization (see section 3.2.3) and VABS2P a parameter defining the crop uptake rate at which
volatilization losses reach 50% of their maximum (2 kg.ha -1.d-1). Figure 3.2 presents the variations of
the pH factor according to soil pH.
93
Second, the volatilization rate from organic fertilizers (VON, eq. 6) depends on the potential NH3
volatilization rate (FSNH3, in μg N.m-2.s-2):
VONH3 (d) 0.036 FSNH3 (d) Eq. 6
The calculation of FSNH3 is based on equations developed by Génermont and Cellier (1997) for the
acido-basic and liquid-gas equilibria and varies according to the concentration of ammonium (NH 4) in
the top-soil, CEC, soil water content, pH, temperature and wind speed (Beutier & Renon, 1978;
Shuttleworth & Wallace, 1985). The volatilization from organic fertilizers (VO) cannot exceed the
amount of NH4 at soil surface nor the NH 4 volatilizable pool. On the day of application, the
volatilizable pool corresponds to the NH 4 content of the organic fertilizer minus the fraction which
infiltrates into the subsoil. This volatilizable fraction, so called PROPVOLAT, depends on the dry
matter content of the organic fertilizer and is assumed to be affected by soil tillage (Figure 3.3). The
volatilizable pool is then updated daily.
Figure 3.2 Effect of soil pH on volatilization from Figure 3.3 Effect of dry matter content of fertilizer
mineral fertilizers and soil tillage on the volatilizable fraction
Nitrification and denitrification

In STICS, the emissions of nitrous oxide (N2O) result from two main processes calculated on a daily-
step basis: nitrification and denitrification. N2O emitted through nitrification (NitN, eq. 7) depends first
on the fraction of NH4 transformed into nitrate (NO3) in each layer (TNITRIF, eq.8) from the surface
to the last active layer for mineralization (PROFHUM). It also depends on the amount of NH 4 in the
specific soil layer (Z). N2O emitted is assumed to be a constant proportion of the nitrified NH 4
(Garrido et al., 2002), so called RATIONITP (with a default value of 0.0016):
Nit N2O (d) RATIONITP ∑PROFHUM

z 1 TNITRIF ( ,d) NH4 ( ,d) Eq. 7
94
The fraction of NH4 transformed into NO3 (TNITRIF, eq. 10) cannot exceed the daily maximum
fraction of ammonium transformed into nitrate (FNXP = 0.5, Sierra et al. (2003)) and is affected by
soil pH (FPHN), soil relative water content (FHN) and soil temperature (FTN):
TNITRIF ( ,d) FN P FPHN FHN ( ,d) FTN ( ,d) Eq. 8
Figure 3.4 presents the variations of the pH factor, Soil relative water content factor (SrWC) and the
temperature factor.
Figure 3.4 Effects of pH (FPHN), Soil relative water content (FHN) and soil temperature (FTN) on nitrification
At this stage, we can also deduce quantities of NO x emitted through nitrification (NitNO, eq. 9) based
on measurements from Rolland et al. (2010):
NitNO (d) 0.005 ∑PROFHUM

z 1 TNITRIF ( ,d) NH4 ( ,d) Eq. 9
N2O emitted through denitrification (DenitN, eq. 10) is calculated according to the model proposed by
Hénault et al.(2005). It is a constant proportion of the denitrified NO3, so called RATIODENIT P (with
a default value of 0.1). For each denitrifying layer (PROFDENIT P, 0-20cm) N2O losses cannot exceed
the total denitrification potential rate (1kg N.ha -1.d-1) and are affected by soil temperature (FDENT),
the amount of nitrate in soil (FDENNO3) and soil properties (FDENW):
VPOTDENIT P
DenitN2O (d) RATIODENITP PROFDENIT P
∑PROFDENIT
z 1
S
FDENT ( ,I) FDENNO3 ( ,d) FDENW ( ,I) Eq. 10
Figure 3.5 presents the variations of the soil temperature factor, the soil nitrate content factor and the
soil properties factor referring to both water content at field capacity (FC) and bulk density (BD).
95
Figure 3.5 Effects of soil temperature (FDENT), soil nitrate content (FDENNO3) and soil properties (FDENW)
on denitrification
At this stage, we can also deduce quantities of N2 emitted through denitrification (DenitN2, eq. 11):
VPOTDENIT P
DenitN2 (d) (1-RATIODENITP ) PROFDENIT P
∑PROFDENIT
z 1
S
FDENT ( ,d) FDENNO3 ( ,d) FDENW ( ,d) Eq. 11
Nitrate leaching
Water transfers are calculated using a reservoir-type analogy. The excess water above field capacity is
drained downward and leads to leaching of nitrates (NO3) soluble in the drained soil water.
3.2.3. Parameterization of the model for the case study

Data for agricultural practices were collected for 12 cropping systems representative of the production
of out-of-season tomato in Beninese urban gardens. This section is limited to a brief description of the
data used to feed the model for this case study. The reader can refer to the previous chapter for more
detail on the data collection procedure and inventory data for these cropping systems. The data refers
to the 2011/2012 dry season (i.e. from September to April).
STICS input consists in four files: climate, soil, management and initialization. The climate file
contains daily data regarding rainfall (mm), wind speed (m.s -1), radiation (MJ.m-2), minimum and
maximum temperature (°C) and, relative humidity (%). The same file based on data from the
international research center located at Cotonou (AfricaRice) was used for the simulation of the 12
cropping systems. The evapotranspiration (ETP) was automatically calculated in STICS based on the
Shuttleworth and Wallace resistive approach (Shuttleworth & Wallace, 1985). Figure 3.6 shows
variations of minimum temperature (CV=7%), radiation (CV=21%) and, ETP (CV=34%) were low.
However, rainfalls were particularly low from December to March (CV=356%) and maximal
temperature slowly increased from September to March (CV=69%).
96
 Rainfall  ETP (mm)  Radiation (MJ/m²)

 Tmin  Tmax (°C)
45 25
40
35 20
30
15
25
20
10
15
10 5
5
0 0
September  October  November  December  January  February  March  April
Figure 3.6 Summary of climate variables for the 2011/2012 dry season at Cotonou, Benin, West Africa
For each cropping system, the soil files contain two types of data due to data availability constraints:
site specific data and estimations from the literature. Soil analyses were performed by the soil
laboratory at the University of Abomey-Calavi from soil samples collected in the 12 farms at the
beginning of each crop cycle. Table 3.2 presents site specific parameters, regarding clay, organic
nitrogen and calcareous (Ca) contents, pH, C:N (carbon to nitrogen ratio), field capacity at saturation
(FCS) and at wilting point (FCW) and bulk density (BD). Other parameters were estimated based on
references from the model documentation(Brisson et al., 2008).
Table 3.2 Site-specific soil data used to feed the STICS model from 12 tomato cropping systems in Benin
Clay Organic nitrogen Total CaCO3 pH C:N Moisture (%) Bulk density
% % % - - Maximal capacity Wilting point g.cm-3
1 7.8 0.05 0.18 6.8 15.2 13.4 2.8 1.4
2 1.8 0.04 0.16 6.0 20.9 13.4 2.7 1.2
3 4.5 0.03 0.34 6.6 16.8 10.8 2.0 1.5
4 13.0 0.03 0.46 5.7 30.4 13.6 5.2 1.2
5 9.8 0.05 0.34 6.0 21.8 16.1 4.0 1.3
6 15.0 0.05 0.18 6.8 19.0 10.6 3.1 1.4
7 0.8 0.02 0.57 6.4 23.5 9.6 1.7 1.5
8 3.3 0.03 0.34 6.9 19.1 11.4 2.2 1.2
9 2.0 0.06 0.34 6.6 17.9 18.6 3.4 1.5
10 2.0 0.03 0.57 4.8 21.8 13.0 3.5 1.5
11 7.8 0.05 0.23 5.0 22.2 24.7 4.2 1.4
12 65.0 0.13 0.11 4.6 20.6 61.6 33.3 1.3
For each cropping system, the management files contain data regarding transplanting (date, crop
density), irrigation (date and volumes for each irrigation event), fertilization (date and rates for each
97
fertilization event) and harvest (harvest frequency, last harvest date). Table 3.3 presents those
management data. For each type of mineral fertilizer (NPK vs. Urea), a set of parameters was defined
based on the STICS documentation:
 Ammonium proportion (100 % vs. 50 %),

 Maximal proportion of N losses by volatilization (35% vs. 12%),
 Maximal amount of microbial immobilized N (37.5 kg N.ha -1 vs. 30.0 kg N.ha-1)
For poultry manure the set of parameters was taken from measurements from Chabalier et al. (2006):
 Carbon content (22%)

 C:N ratio (6)
 Mineral nitrogen content (3%)
 Water content (54%)
Table 3.3 Management data used to feed the STICS model from 12 tomato cropping systems in Benin
Crop Poultry Last
Length Yield Transplant Water Urea NPK
density manure harvest
days kg.ha-1 date plt.m-2 m3.ha-1 kg N.ha-1 kg N.ha-1 kg N.ha-1 date
1 80 4902 9-Oct 3.9 7200 1045 33 107 28-Dec
2 64 12452 11-Nov 5.3 6360 64 146 104 14-Jan
3 80 8 5-Oct 1.8 586 117 0 19 24-Dec
4 48 7875 4-Nov 2.7 10400 473 43 25 22-Dec
5 92 21163 30-Oct 7.4 5318 0 0 22 30-Jan
6 82 127 21-Nov 3.8 4515 563 0 235 11-Feb
7 36 0 21-Nov 2.4 919 0 0 58 27-Dec
8 91 5662 14-Dec 3.3 10581 778 0 0 14-Mar
9 83 1703 5-Dec 5.7 7048 0 59 0 26-Feb
10 79 4498 5-Jan 3.3 771 0 202 72 24-Mar
11 57 2710 17-Nov 3.5 3759 450 0 0 13-Jan
12 67 0 23-Nov 3.7 2570 559 0 211 29-Jan
In the initial condition file, the nitrogen content was set to 0 and the water content was set to 100% of
the field capacity, as important water volumes were applied to tomato before transplanting, leading to
the leaching of most of the residual soil nitrogen.
STICS also requires a set of general parameters regarding the crop and the formalisms. Due to a lack
of data specific to our crop growth, we used the generic tomato crop file, with the Capitan cultivar as
the better approximation for our determinate tropical cultivar (Mongal). We did not change general
parameters except for climate dependent ones (Table 3.4).
98
Table 3.4 General parameters used to adapt the STICS crop model to tropical conditions
Process
Parameter description Temperate Tropical
involved
N transfer Threshold of soil nitrate concentration for leaching (kg N.ha -1) 1.01 (1) 0.20 (1)
Mineralization Minimal water content for mineralization (% of field capacity) 0.30 (2) 0.22 (1,3)
Optimal water content for mineralization (% of field capacity) 1.0 (2) 0.67 (1,3)
Nitrification Minimal temperature below which nitrification stops 5 (1) 20 (1)

Optimal temperature for nitrification 20 (1) 35 (1)
Maximal temperature above which nitrification stops 45 (1) 50 (1)
Denitrification Parameters for the effect of temperature 11 (1) 20 (4)

20 (1) 29 (4)
1. Brisson et al. (2008) 2. Rodrigo et al. (1997) 3. Sierra et al. (2003) 4. Sierra et al.(2010)
3.2.4. Statistical analysis

Variations of observed and simulated variables were analyzed using simple and multivariate linear
regressions computed with the R software (V 2.15.2). We thus assumed that the dependent variable (Y)
was a linear combination of the parameters (a and b) and the explanatory variables (X, i.e. Y=aX+b).
The coefficient of correlation (r²) corresponds to the fraction of variance explained and the slope (a)
informs about the effect of the predictor on the dependent variable. The significance of the slope (a)
was tested using a Student’s t-test with H0: a=0 and H1:a≠0. Significance codes were assigned (`for
5%,* for 1%, ** for 0.1% and *** for less than 0.1%) corresponding to the risk in accepting H 1 when
H0 is true. If the slope was not significant, we could not conclude on an effect of the predictor. When
necessary, the independence of variables was tested with a Pearson's chi-squared test where the
contingency coefficient (CC) is a measure of the degree of relationship between the dependent
variables.
99
3.3. Results and discussion

3.3.1. Quality of prediction and model soundness for yields
On a crop cycle basis, STICS consistently overestimated yields except for one cropping system
(Figure 3.7). As a result we ended up with a poor quality of prediction with r²=0.25. The distribution
of the error seems to follow a linear function (Figure 3.8, R²=0.51) indicating particularly poor
predictions for low yields.
Figure 3.7 Simulated versus observed yields Figure 3.8 Yield residues versus observed yields
To analyze the discrepancy between simulated and observed yields, we tested the effect of identified
drivers for the process of yield formation within STICS (Figure 3.1) on simulated and observed yields.
Figures 3.9 and 3.10 show that an important part of simulated yield variations can be explained by the
cumulative global radiation (r²=0.65) and the cumulative temperature (r²=0.66). It means that
simulated yields were limited by climatic drivers (a=14** and a=67** for respectively radiation and
temperature). The cumulated temperature being significantly correlated to the cumulative radiation
(CC=0.97), we could not identify which one of these drivers was the most limiting for yields. The
positive effect of total water, i.e. irrigation water plus rainfall (Figure 3.11) indicated that simulated
yields were limited by water stresses (r²=0.39*). However, yield and nitrogen inputs were correlated
neither in the simulation nor from observations (Figure 3.12). The lack of correlation might result from
the absence of nitrogen stresses resulting from excessive inputs. It might also results from important
variations of nitrogen efficiency caused by important nitrogen losses.
100
Figure 3.9 Simulated and observed yields versus Figure 3.10 Simulated and observed yields versus
cumulative global radiation over the crop cycle cumulative temperature over the crop cycle
Figure 3.11 Simulated and observed yields versus Figure 3.12 Simulated and observed yields versus
total water nitrogen inputs
Figure 3.13 Simulated and observed yields versus

plant density
101
The weak correlation between observed yields and those drivers (radiation, temperature, nitrogen,
water) indicated that other drivers were more limiting for observed yields. Figure 3.13 shows that plant
density was the best explicative variable we found for observed yields (r²=0.53). The higher effect of
plant density for the observed yield (a=2976**) compared to simulated yields (a=1578`) suggested a
higher production per plant on field than the one simulated by STICS.
We could not properly assess the model efficiency to simulate yields because we could not compare
simulated yield to yield obtained in controlled conditions. The extreme biotic pressure on studied
cropping systems resulted in major yield variations which could not be reproduced with STICS
(maximum r² was 0.53). However we could test the model soundness. We can conclude that simulated
tomato yields were sensitive to climate conditions (temperature and/or radiation) and volumes of water
applied. Considering the dry and hot climate and the variable water volumes applied across fields,
STICS behavior can be considered as satisfactory. For these 3 drivers (radiation, temperature and
water), effects were similar for simulated and observed yields (same slope). It suggests that STICS
succeeded in simulating the response of yields to these drivers. However, the low correlation of
observed yield to these drivers indicated that observed yield were mainly affected by other drivers. In
agreement with Huat et al. (2013) who identified planting density and pest attacks as main limiting
drivers for yields of tomatoes fields grown by smallholders in tropical regions, a significant part of
yield variations could be explained by plant density whether in the simulations or in the observations.
The higher effect of plant density for observed yield suggested that parameters affecting the
production per plant were not adapted to our tropical cultivar.
3.3.2. Analysis of drivers for nitrogen losses as simulated by STICS

In Life-cycle assessment, Nr losses
(NH3, N2O, NOx and NO3) to the
environment are estimated using
emission factors, i.e. the ratio
between the amounts of nitrogen
emitted on the amount of nitrogen
input. Figure 3.14 shows a
correlation between the amounts of
nitrogen emitted on the amount of
Figure 3.14 Total Nr emissions versus nitrogen input from nitrogen input (r²=0.91). It appears
mineralization of soil organic matter, organic and mineral fertilizers that main drivers for nitrogen
emissions are organic fertilizers
102
inputs (r²=0.94) followed by nitrogen from soil organic matter mineralization (r²=0.82). In the
following section we will analyze the sensitivity of emissions factors for NH 3, N2O and NO3 to soil,
management and climate drivers.
Ammonia volatilization
The total ammonia emission factor (NH 3 emissions/total fertilizer inputs) varied from 0 to 37 % across
the 12 cropping systems, with an average of 10%. These results were lower than NH 3 emissions factors
(39% on average, see chapter 2 Table 2.3) calculated from ammonia measured on urban gardens in
Niger (Predotova et al., 2010) and Burkina Faso (Lompo et al., 2012). NH3 emission factors were
expressed for each type of fertilizers, i.e. organic and mineral. The emission factors from organic
fertilizers, corresponding to the main origin of emission, varied from 0.9 to 73% across the 12
cropping systems with an average of 31%. This result was higher to the emission factor of 20%
suggested by Bouwman and Van Der Hoek (1997). We tested the correlation between NH 3 emission
factors and the identified drivers for volatilization within STICS (Figure 3.1). We found out that NH3
emission factors were correlated to soil pH (Figure 3.15). The effect of pH was more important for
NH3 emission from organic fertilizers (a=0.31**) than from mineral fertilizers (a=0.02**). The
equations from Génermont and Cellier (1997) included an acido-basic equilibrium explaining the
increase of emissions with soil pH through its effect on the NH 4+:NH3 ratio in the soil (Figure 3.2).
However, we could not identify any other variable with a significant effect on NH 3 emissions from
organic fertilizers. Regarding NH 3 emissions from mineral fertilizers, the volatilization rate was
expected to increase with soil pH (equation 5). The second effect was due to the nature of the
fertilizers: NPK or urea (VOLENGP for urea> VOLENGP for NPK). None of these drivers showed
significant effect on NH3 emissions from mineral fertilizers. Unexpectedly, NH3 emissions from
mineral fertilizers were correlated to the total water inputs (Figure 3.16). It suggests that the plant
nitrogen uptake velocity (being the last effect variable in equation 5) was influenced by water inputs.
Figure 3.15 NH4 emissions factors versus soil pH Figure 3.16 NH4 emissions factors for mineral fertilizers
versus total water
103
N2O emissions from nitrification and denitrification

The total nitrous oxide emission factor (N2O emissions /total nitrogen inputs including soil
mineralization) varied from 0.08 to 1.96 % across the 12 cropping systems with an average of 0.28%.
This result was lower than N2O emissions factors (15% on average, see chapter 2, Table 2.3)
calculated from N2O measured on urban gardens in Niger (Predotova et al., 2010) and Burkina Faso
(Lompo et al., 2012). N2O emission factors were expressed for each process, i.e. nitrification and
denitrification. We tested the correlation between N 2O emission factors and the identified drivers for
nitrification and denitrification within STICS (Figure 3.1). The major part of N2O emission factor
from denitrification variations can be explained by soil moisture at maximal field capacity (r²=0.93,
Figure 3.17) even if we excluded the extreme clay soil (r²=0.63, Figure 3.18). An important part of
N2O emission factor from nitrification variations, can be explained by soil pH (r²=0.40, Figure 3.19).
While the effect of changing pH from 4.5 to 5.5 agreed with the expected trend (Figure 3.2), the
decrease of N2O emission factor for great pH probably resulted from another influencing driver.
Figure 3.17 N2O emission factors versus soil moisture Figure 3.18 N2O emission factors versus soil moisture at
at maximal capacity for all cropping systems maximal capacity for sandy soils
Figure 3.19 N2O emission factors for nitrification versus

soil pH
104
Nitrate leaching
The nitrate emission factor (NO 3 emissions/total nitrogen inputs including soil mineralization) varied
from 0 to 52% across the 12 cropping systems, with an average of 10%. This result was higher than
NO3 emissions factors (1% on average, see chapter 2, Table 2.3) calculated from nitrate measured on
urban gardens in Niger (Predotova et al., 2011) and Burkina Faso (Lompo, 2012), which is logical
considering the intensive irrigation in our systems. Figure 3.20 shows that NO3 emissions were
positively correlated to the nitrification rate of (r²=0.74, a**). This effect was in agreement with the
reservoir-type approach for leaching, being the result of drainage of water containing NO 3. However,
figure 3.21 shows a poor correlation between NO3 emissions and the total water inputs (r²=0.19). It
suggests that other drivers influenced drainage events (probably a combination of soil parameters).
Figure 3.20 NO3 emissions versus the total nitrified Figure 3.21 NO3 emissions versus total volume of water
nitrogen
105
3.3.3. Evaluation of the combined approach

Figure 3.22 presents the nitrogen budgets for the 12 cropping systems simulated with STICS. Nitrogen
budget for the 12 cropping systems differed in terms of importance (195 to 1814 kg N.ha-1) and
diversity of nitrogen fluxes.
The net change in soil storage of mineral nitrogen, also called nitrogen surplus ( SoilN ) varied from
145 to 876 kg N.ha -1 with an average of 463 kg N.ha-1. It represented on average 59% of the total
output. Such a large nitrogen surplus agrees with results from several studies dealing with nitrogen
budget of urban gardens (Diogo et al., 2010; Lompo, 2012). Nitrogen fertilizers constituted on average
57% of nitrogen inputs, with respectively 43% for organic fertilizers and 14% for mineral fertilizers.
Figure 3.22 Nitrogen budget simulated with STICS for the 12 cropping systems
106
Soil mineralization
One particularity of using a dynamic
crop model to calculate a nitrogen
budget referred to the inclusion of the
mineralization of soil organic matter
and organic residues. While
immobilized nitrogen into soil
organic matter did not contribute
much to nitrogen outputs (3% on
average), nitrogen mineralized
(MHN) contributed on average 51%
of nitrogen input, with variable
Figure 3.23 Nitrogen input from soil mineralization versus organic
fertilizers quantities across cropping systems
(CV=54%). Figure 3.23 shows that most of the variation of nitrogen from mineralization can be
explained by organic fertilizers inputs (r²=0.91). As expected, the nitrogen mineralized from residues
(i.e. from organic inputs) was perfectly correlated to organic fertilizers inputs (r²=1, a=0.33***). It
means that each organic fertilizer application was followed by an additional release of mineral
nitrogen equivalent to 33% of the mineral nitrogen contained in the fertilizers. To eventually refine the
estimation of nitrogen inputs from the mineralization of organic fertilizers, we could collect site-
specific parameters for poultry manure, instead of using generic data from literature. However, the
good performances of STICS to simulate mineralization in a similar context (Sierra et al., 2010) and
the satisfactory soundness of our simulations allowed us to conclude that no major uncertainty was
associated with the estimation of this component of the nitrogen budget.
Harvested nitrogen
Harvested nitrogen constituted on
average only 8% of the total output,
and presented low variations across
cropping systems (CV=26%).
Quantities of nitrogen exported per
hectare varied from 9 to 45, with an
average of 37 kg N.ha-1. These
quantities were higher than average
quantities (11 kg.ha-1 on average, see
chapter 2, Figure 2.5) calculated from
Figure 3.24 Simulated harvested nitrogen versus simulated yields

measurements on urban gardens in
107
Niger (Diogo et al., 2010) and Burkina Faso (Lompo, 2012). Figure 3.24 shows exported nitrogen was
correlated to simulated yields with an average nitrogen concentration of fruit estimated at 0.18%. This
nitrogen concentration was similar to the 0.15% estimated from the literature (see chapter 2 and dotted
line in Figure 3.24). We can extrapolate from the yields variations analysis (section 3.3.1) that
variations of harvested nitrogen across cropping systems were multifactorial and very likely not
representative for variations of observed data. To improve the simulation of harvested nitrogen we
could collect data to calibrate cultivar-specific parameters regarding the production per plant and the
nitrogen concentration of fruits. However to properly simulate tomato yields and subsequent nitrogen
harvested for Beninese tomato cropping system, the model should also include losses due to parasitism
stresses, as crop protection has often been shown inefficient in such context (Huat et al., 2013). To our
knowledge, no model including parasitism stresses and associated crop protection is available.
Nr emissions
Nr emissions (NH3, N2O, NOx and NO3) constituted on average 21% of the total output and presented
large variations across cropping systems (CV=122%). Ammonia emissions contributed significantly to
the total emissions (58%) and quantities varied across cropping systems (CV=160%). The main
influencing driver for the total NH3 emission was the soil pH (Figure 3.15). We also found out that the
total water input influenced NH3 emissions from mineral fertilizers (Figure 3.16). However, as
variations of NH3 emissions from mineral fertilizers was much lower compared to variations of NH 3
emissions from organic fertilizers, we can conclude that water input resulted in low variations of the
nitrogen budget when referring to its effect on NH3 emissions from mineral fertilizers. A part of NH 3
emission from organic fertilizers variation remained unexplained using linear regressions. Considering
the number of parameters involved to calculate the equilibrium between the different forms of nitrogen
(Génermont & Cellier, 1997), influencing drivers were very likely multiple and thus difficult to
identify. In STICS (Equation 6), NH3 volatilized from organic fertilizers also depend on the
volatilizable fraction of organic fertilizers, which was an estimated based on the water content of
organic fertilizer. To eventually refine the estimation of NH 3 emissions we could collect site-specific
parameters for poultry manure, instead of using generic data from literature. However, the likelihood
between simulated NH3 emissions and measurements from the literature (Predotova et al., 2010;
Lompo et al., 2012) and the satisfactory soundness of our simulations allowed us to conclude than no
major uncertainty was associated with the estimation of this component of the nitrogen budget.
Nr emissions from nitrification and denitrification processes contribute significantly to the total
emissions (25%) but quantities did not vary much across cropping systems compared to other
emissions (CV=75%). Nitrous oxide emissions presented contributed for 11%, however quantities
varied across cropping systems (CV=130%). Nitrogen oxide emissions contributed for 14%, and
quantities did not vary much across cropping systems (CV=60%).
108
At first sight, conditions seemed favorable for nitrification processes (high temperature, high water
and nitrogen inputs). This assumption was validated as we obtained an average emission factor of
0.4% for NOx (CV=7%), above the reference of 0.1% of the total N inputs, commonly used for NO x
(Nemecek and Kägi, 2007). We could identify soil pH as an influencing driver for nitrification.
However, we could question the validity of our assumption to estimate NOx emissions (0.05% of
nitrified nitrogen) which was based on measurements in a very different context compared to our case
study.
For N2O emissions, the measurements from the literature in similar contexts (Predotova et al., 2010;
Lompo et al., 2012), suggested that the emissions factor would be higher than the IPCC (IPCC, 2006)
reference of 1% of the total N inputs for N2O. However, we obtained an average emission factor of
0.3% for N2O (CV=175%). A significant part of the difference between the IPCC reference and the
simulated emission factors may results from the high contribution of soil organic nitrogen to the total
nitrogen inputs. When only considering fertilizers as nitrogen inputs we obtained an average emission
factor of 0.6% . We found out that for soils with moisture at maximal field capacity below 18%, the
model STICS did not simulate any denitrification process. It can be explained by the effect of the field
capacity in the calculation of the daily water-filled pore space (WFPS) as a function of soil water
content (maximum corresponding to field capacity) and bulk density. Indeed, the denitrification, as an
anaerobic process, does not occur until the WFPS exceeds 50–60%, then it increases exponentially as
WFPS increases and levels off as soils approach saturation (Parton et al., 1998). In STICS, this
threshold has been set to 62% based on the NOE equations (Hénault et al., 2005). Considering the low
soil moisture at field capacity and associated bulk density of our soils most cropping systems cannot
reach this threshold. This result was in favor of a satisfactory soundness of STICS results. However,
on dryland cropping systems (Del Grosso et al., 2002) showed that a major source of model error
(DNDC in this case) was the inability to correctly simulate WFPS. This can be explained by the
difficulty in estimating amounts of water from the macroporosity. To improve our confidence on N 2O
emissions we could collect data to validate the estimation of WFPS for our soils. To refine the
estimation of N2O emissions we could also provide site-specific data for VPOTDENIT P. Finally Chen
et al. (2008) underlined in a review of simulation approaches for N2O emissions that an important
challenge referred to the partitioning of gaseous N losses into N2O and N2. In STICS this ratio was
fixed for denitrification (RATIODENIT P) based on the literature (Garrido et al., 2002). However it
was found that this ratio varied (Del Grosso et al., 2000) especially in irrigated vegetable cropping
systems (He et al., 2009). This analysis suggested that important uncertainties were associated with the
estimation N2O emissions from denitrification. We can conclude that emissions from nitrification and
denitrification cannot be validated at this stage.
Nitrate emissions contributed significantly to the total emissions (21%) and quantities varied across
cropping systems (CV=179%). An important part of NO3 emission variations could be explained by
nitrification rates. To eventually improve our confidence on this variable we could collect data to
109
validate soil NO3 concentration at the beginning and at the end of the crop. However, the
performances of STICS to simulate nitrate leaching in contrasted contexts (Schnebelen et al., 2004;
Jégo et al., 2008) and the satisfactory soundness of our simulations allowed us to conclude that no
major uncertainty was associated with the estimation of this component of the nitrogen budget.
Qualitative assessment of the overall approach

Table 3.5 summarizes the performances of the suggested estimation methods for Nr emissions
combining a crop model and a nutrient budget. The budget approach allowed us to identify the main
fluxes of nitrogen, while the modeling approach allowed us to identify influencing driver for each
important nitrogen flux and associated uncertainties. As a result the nitrogen budget of out-of-season
tomato production in Beninese urban gardens was mainly influenced by 4 drivers:
1. The quantity of organic fertilizers because of its contribution to the total inputs and its effect
on mineralized and volatilized nitrogen,
2. The volumes of water applied because of its effects on volatilized and leached nitrogen,
3. The soil pH because of its effect on volatilized and nitrified nitrogen,
4. The soil moisture at maximal field capacity because of its effect on denitrified nitrogen,
Regarding uncertainties, we can express reservations on the reliability of harvested nitrogen, N 2O and
NOx emissions. Regarding harvested nitrogen, a first lead for improvement consists in collecting
cultivar-specific parameters. However the fundamental uncertainty associated to yield losses due to
pest pressure cannot be undertaken with the current tool. Regarding N2O, a first lead for improvement
consists in collecting specific parameters for the potential of denitrification. However, considering the
lack of references for N2O and NOx emissions in our context, further data to validate water and
nitrogen dynamics would be needed to improve the quality of model outputs. For other emissions and
fluxes, although we still have room for improvement in reducing operational uncertainties, the
reliability of estimates could be considered as satisfactory. Back to the Life-Cycle Assessment
framework, there was a last source of uncertainty which refers to the budget approach: what is the fate
of the nitrogen surplus? To resolve this issue we could run multiple simulations using the nitrogen
surplus as initial conditions until we reach a steady state. However this procedure would require
additional data on agricultural practices for several full climatic years. This question regarding the
limit of the system will be further discussed in chapter 5.
110
Table 3.5 Performances of the structured estimation methods for Nr emissions

Influencing drivers Operational uncertainty Fundamental uncertainty
N mineralization Organic fertilizer inputs Site-specific parameters for
poultry manure
N harvest Crop density Cultivar-specific Yield losses due to biotic

Temperature and/or parameters stresses
radiations
Volumes of water inputs
NH3 emissions pH Site-specific parameters for

Volumes of water inputs poultry manure
NO3 emissions Rate of nitrified nitrogen Validation of soil nitrogen

 organic fertilizers dynamic
N2O & NOx Soil moisture at maximal Validation of on soil water RATIODENITP
emissions field capacity dynamic RATIONITP
Soil pH Field-specific value for NOx=0.05% of nitrified
VPOTDENITP nitrogen
Combined 1. Organic fertilizer inputs 1. Cultivar parameters 1. Yield losses due to biotic
approach 2. Volumes of water inputs 2. VPOTDENITP stresses
3. Soil pH 3. Site specific parameters 2. Validation of soil water
4. Soil moisture at maximal for poultry manure content dynamic
field capacity 3. Validation of soil nitrogen
content dynamic
4. Exploration of the fate for 
SoilN?
111
3.4. Conclusion
In this paper we proposed a structured method for identifying main drivers for Nr emissions based on
the association of a nitrogen budget and a dynamic crop model. The advantage of the crop model
approach was to include a diversity of parameters influencing Nr emissions, according to up-to-date
knowledge on biological processes involved. The budget approach allowed us to rank influencing
drivers and uncertainties associated to the estimation of the different components of the budget
according to their contribution to input and output fluxes. We underlined one additional uncertainty
related to the fate of the nitrogen surplus.
We compared simulated yield to observed yield and underlined the error involved by the absence of
biotic stresses within the crop model. Regarding emissions, we assessed the model soundness with
regard to measured literature data and variation of influencing parameters. We found out that despite
the many parameters identified for each process there were only a few drivers affecting heavily the
components of the nitrogen budget across the 12 cropping systems. We calculated the nitrogen budget
for the twelve cropping systems and qualitatively assessed the uncertainty linked to the estimation of
major contributing components. For those components we provided recommendations to refine or
improve the quality of the estimation based on our expertise on the model and the literature. Finally
we identified and ranked four influencing drivers and five origins of uncertainty associated to the
calculation of the nitrogen budget.
The sensitivity of Nr emissions to soil type and agricultural practices suggested potential variations of
the associated environmental impacts. In the next chapter we will use the simulated emissions with the
crop model STICS in the modeling of 8 agronomic scenarios representative for the variability of the
environmental impact of one hectare of out-of-season tomato.
112
3.5. Reference
Adler PR, Grosso SJ Del, Parton WJ (2007) LIFE-CYCLE ASSESSMENT OF NET GREENHOUSE-GAS
FLUX FOR BIOENERGY CROPPING SYSTEMS. Ecological Applications 17:675–691. doi: 10.1890/05-
2018
Assessment 18:24–36. doi: 10.1007/s11367-012-0457-0
Beutier D, Renon H (1978) Representation of NH3-H2S-H2O, NH3-CO2-H2O, and NH3-SO2-H2O vapour-
liquid equilibria. Industrial & Engineering Chemistry Product Research and Development 17:220–230. doi:
10.1021/i260067a002
10.1016/s1352-2310(97)00288-4
Brisson N, Launay M, Mary B, Beaudoin N (2008) Conceptual basis, formalisations and parameterization of the
STICS crop model. Quae, Nancy, France
Brisson N, Ruget F, Gate P, Lorgeou J, Nicoullaud B, Tayot X, Plenet D, Jeuffroy MH, Bouthier A, Ripoche D
(2002) STICS: a generic model for simulating crops and their water and nitrogen balances. II. Model
validation for wheat and maize. Agronomie 22:69–92. doi: 10.1051/agro:2001005
Chen D, Li Y, Grace P, Mosier AR (2008) N2O emissions from agricultural lands: a synthesis of simulation
approaches. Plant and Soil 309:169–189. doi: 10.1007/s11104-008-9634-0
10.1007/s10705-009-9315-2
doi: 10.1007/s12649-012-9171-1
Gabrielle B, Gagnaire N (2008) Life-cycle assessment of straw use in bio-ethanol production: A case study
based on biophysical modelling. Biomass and Bioenergy 32:431–441. doi: 10.1016/j.biombioe.2007.10.017
Gabrielle B, Laville P, Hénault C, Nicoullaud B, Germon JC (2006) Simulation of Nitrous Oxide Emissions
from Wheat-cropped Soils using CERES. Nutrient Cycling in Agroecosystems 74:133–146. doi:
10.1007/s10705-005-5771-5
Gabrielle B, Mary B, Roche R, Smith P, Gosse G (2002) Simulation of carbon and nitrogen dynamics in arable
soils: a comparison of approaches. European Journal of Agronomy 18:107–120. doi: 10.1016/s1161-
0301(02)00099-0
Garrido F, Henault C, Gaillard H, Perez S, Germon J. C, Hénault C, Pérez S (2002) N2O and NO emissions by
agricultural soils with low hydraulic potentials. Soil Biology and Biochemistry 34:559–575. doi:
10.1016/S0038-0717(01)00172-9
Génermont S, Cellier P (1997) A mechanistic model for estimating ammonia volatilization from slurry applied to
bare soil. Agricultural and Forest Meteorology 88:145–167. doi: 10.1016/S0168-1923(97)00044-0
Del Grosso S, Ojima D, Parton W, Mosier A, Peterson G, Schimel D (2002) Simulated effects of dryland
cropping intensification on soil organic matter and greenhouse gas exchanges using the DAYCENT
ecosystem model. Environmental Pollution 116:S75–S83.
Del Grosso SJ, Parton WJ, Mosier AR, Ojima DS, Kulmala AE, Phongpan S (2000) General model for N2O and
N2 gas emissions from soils due to dentrification. Global Biogeochemical Cycles 14:1045–1060. doi:
10.1029/1999GB001225
He F, Jiang R, Chen Q, Zhang F, Su F (2009) Nitrous oxide emissions from an intensively managed greenhouse
vegetable cropping system in Northern China. Environmental Pollution 157:1666–1672. doi:
10.1016/j.envpol.2008.12.017
Hénault C, Bizouard F, Laville P, Gabrielle B, Nicoullaud B, Germon JC, Cellier P (2005) Predicting in situ soil
N2O emission using NOE algorithm and soil database. Global Change Biology 11:115–127. doi:
10.1111/j.1365-2486.2004.00879.x
113
Huat J, Dore T, Aubry C (2013) Limiting factors for yields of field tomatoes grown by smallholders in tropical
regions. Crop Protection 44:120–127. doi: 10.1016/j.cropro.2012.11.003
Jégo G, Martínez M, Antigüedad I, Launay M, Sanchez-Pérez JM, Justes E (2008) Evaluation of the impact of
various agricultural practices on nitrate leaching under the root zone of potato and sugar beet using the
STICS soil-crop model. Science of The Total Environment 394:207–221. doi:
10.1016/j.scitotenv.2008.01.021
Khai NM, Ha PQ, Öborn I (2007) Nutrient flows in small-scale peri-urban vegetable farming systems in
Southeast Asia—a case study in Hanoi. Agriculture, Ecosystems & Environment 122:192–202. doi:
10.1016/j.agee.2007.01.003
Langevin B, Basset-Mens C, Lardon L (2010) Inclusion of the variability of diffuse pollutions in LCA for
agriculture: the case of slurry application techniques. Journal of Cleaner Production 18:747–755. doi:
10.1016/j.jclepro.2009.12.015
Li C, Frolking S, Frolking TA (1992) A model of nitrous oxide evolution from soil driven by rainfall events: 1.
Model structure and sensitivity. Journal of Geophysical Research: Atmospheres 97:9759–9776. doi:
10.1029/92jd00509
Li CS (2000) Modeling trace gas emissions from agricultural ecosystems. Nutrient Cycling in Agroecosystems
58:259–276. doi: 10.1023/A:1009859006242
Lompo DJ-P (2012) Matter flows and balances in urban vegetable gardens of Bobo Dioulasso, Burkina Faso
(West Africa). PhD thesis, University of Kassel
Lompo DJ-P, Sangaré SAK, Compaoré E, Papoada Sedogo M, Predotova M, Schlecht E, Buerkert A (2012)
Gaseous emissions of nitrogen and carbon from urban vegetable gardens in Bobo-Dioulasso, Burkina Faso.
Journal of Plant Nutrition and Soil Science 175:846–853. doi: 10.1002/jpln.201200012
Nicolardot B, Recous S, Mary B (2001) Simulation of C and N mineralisation during crop residue
decomposition: A simple dynamic model based on the C:N ratio of the residues. Plant and Soil 228:83–103.
doi: 10.1023/a:1004813801728
Oenema O, Kros H, de Vries W (2003) Approaches and uncertainties in nutrient budgets: implications for
nutrient management and environmental policies. European Journal of Agronomy 20:3–16. doi:
10.1016/S1161-0301(03)00067-4
Parton WJ, Hartman M, Ojima D, Schimel D (1998) DAYCENT and its land surface submodel: description and
testing. Global and planetary Change 19:35–48. doi: 10.1016/S0921-8181(98)00040-X
Parton WJ, Mosier AR, Ojima DS, Valentine DW, Schimel DS, Weier K, Kulmala AE (1996) Generalized
model for N2 and N2O production from nitrification and denitrification. Global Biogeochemical Cycles
10:401–412. doi: 10.1029/96GB01455
10.1002/jpln.200900255
10.1016/j.fcr.2009.09.010
Rodrigo A, Recous S, Neel C, Mary B (1997) Modelling temperature and moisture effects on C–N
transformations in soils: comparison of nine models. Ecological Modelling 102:325–339. doi:
10.1016/S0304-3800(97)00067-7
Rolland MN, Gabrielle B, Laville P, Cellier P, Beekmann M, Gilliot JM, Michelin J, Hadjar D, Curci G (2010)
High-resolution inventory of NO emissions from agricultural soils over the Ile-de-France region.
Environmental Pollution 158:711–722. doi: 10.1016/j.envpol.2009.10.017
Sangare S, Compaore E, Buerkert A, Vanclooster M, Sedogo M, Bielders C (2012) Field-scale analysis of water
and nutrient use efficiency for vegetable production in a West African urban agricultural system. Nutrient
Schnebelen N, Nicoullaud B, Bourennane H, Couturier A, Verbeque B, Revalier C, Bruand A, Ledoux E (2004)
The STICS model to predict nitrate leaching following agricultural practices. Agronomie 24:423–435. doi:
10.1051/agro:2004039
Sebillotte M (1990) Système de culture, un concept opératoire pour les agronomes. Un point sur les systèmes de
culture, INRA éditions, Paris 165–196.
SETAC (1998) Life-cycle Impact Assessment: The State-of-the-art. Society of Environmental Toxicology and
Chemistry, Brussels and Pensacola (Florida)
114
Shuttleworth WJ, Wallace JS (1985) EVAPORATION FROM SPARSE CROPS - AN ENERGY

COMBINATION THEORY. Quarterly Journal of the Royal Meteorological Society 111:839–855. doi:
10.1256/smsqj.46909
Sierra J, Brisson N, Ripoche D, Déqué M (2010) Modelling the impact of thermal adaptation of soil
microorganisms and crop system on the dynamics of organic matter in a tropical soil under a climate change
scenario. Ecological Modelling 221:2850–2858. doi: 10.1016/j.ecolmodel.2010.08.031
Sierra J, Brisson N, Ripoche D, Noël C (2003) Application of the STICS crop model to predict nitrogen
availability and nitrate transport in a tropical acid soil cropped with maize. Plant and Soil 256:333–345. doi:
10.1023/a:1026106208320
Valdés-Gómez H, Ortega-Farías S, Argote M (2009) Evaluation of the Water Requirements for a Greenhouse
Tomato Crop using the Priestley-Taylor Method . Chilean Journal of Agricultural Research 69:3–11.
Valé M (2006) Quantification et prédiction de la minéralisation nette de l’azote du sol in situ, sous divers
pédoclimats et systèmes de culture français. Thèse doctorat, INP, Toulouse
115
116

Ce chapitre présente une approche basée sur un bilan d’azote et l’usage d’un modèle biophysique pour
analyser la variabilité des émissions d’azote réactif au champ. Les résultats de cette expérimentation
avec le modèle candidat STICS ne nous ont pas permis de conclure que ce modèle permettrait de
raffiner les estimations des émissions dans le cadre de l’ACV appliquée à notre cas d’étude. Ce
résultat en demi-teinte résulte notamment de l’incapacité du modèle à prendre en compte les pertes de
rendements liées à la pression parasitaire, principale source de variabilité des rendements dans notre
cas d’étude. D’autre part la grande quantité de données nécessaires pour paramétrer le modèle
(culture) et valider les sorties (émissions) limite le potentiel de ce type de modèle pour un usage en
ACV et plus particulièrement en contexte tropicale ou ces données sont rarement disponibles.
Néanmoins, la robustesse de ce modèle et l’analyse fine des variations d’émissions réalisée dans ce
chapitre nous a permis d’identifier 4 principaux paramètres influençant les émissions dans notre
gamme de situations pédoclimatiques. Ainsi nous pourront propager ces variations aux résultats
d’ACV sur la base de scenarios virtuels dans le chapitre 4 (Figure 3.25).

 Tomate produite en jardin urbain au Bénin
OBJECTIF 3 : Identifier les déterminants de la

variabilité des émissions azotées
- Approche pour analyser la variabilité des

OBJECTIF 2 émissions d’azote réactif au champ
Chapitre 2 - Modèle candidat pour simuler la variation des
OBJECTIF 1 émissions en fonction des pratiques et du sol
Chapitre 1 (méthode CM)
- 4 paramètres de variation des émissions
d’azote réactif
OBJECTIF 4
Chapitre 4
 4 paramètres pour construire des scénarios pour l’analyse de la variabilité des

émissions
117
118
Chapitre 4. Exploration de la variabilité des impacts environnementaux calculés
par l’analyse du cycle de vie de la tomate en maraîchage urbain au Bénin
© Aurélie Perrin
Jardin urbain équipé d’un système d’irrigation par aspersion à Cotonou, Bénin
119
- Chapitre 4. Exploration de la variabilité de l’ACV de la tomate au Bénin -
Avant-Propos
Nous avons montré dans les chapitres 2 et 3 que les systèmes de production de tomate en jardin urbain
au Bénin étaient très diverses et avaient des flux très variables (intrants, rendements et émissions).
Dans ce cadre, il est difficile d’en évaluer les performances environnementales uniquement sur la base
d’un inventaire moyen. Nous proposons dans ce chapitre un grand nombre de résultats d’ACV pour
plusieurs catégories d’impact, plusieurs scénarios et plusieurs choix méthodologiques. La figure 4.1
résume les données et les méthodes issues des chapitres précédents qui sont mises en œuvre dans ce
chapitre. Au travers de cette diversité d’approches et de scénarios, ce chapitre vise à produire une
connaissance robuste sur les impacts environnementaux de ces systèmes en prenant en compte la
complexité mise en évidence précédemment.

Tomate produite en jardin urbain au Bénin
OBJECTIF 1
Chapitre 1 OBJECTIF 3
OBJECTIF 2 Chapitre 3
Chapitre 2
 Méthode
d’estimation des  Modèle candidat pour simuler la
émissions basée sur  ICV pour des systèmes de variation des émissions en
la littérature culture agrégés par modalité pour fonction des pratiques et du sol
(méthode DEF) 4 critères discriminants (méthode CM)
 Références  ICV pour la moyenne pondérée  4 paramètres influençant les
d’impact pour représentative de la population émissions de Nr
comparer nos
résultats (GWP, AP,
EP)
OBJECTIF 4 : Evaluer l’impact environnementale de la tomate produite

en jardins urbains au Bénin
Figure 4.1 Les contributions des chapitres précédents (rouge) pour répondre à l’objectif 4
120
Résumé
A ce jour, aucune étude ne traite de l’évaluation environnementale des produits issus de l’agriculture
urbaine Africaine, qui est en fort développement. L’application de la méthodologie Analyse du Cycle
de Vie (ACV) à ces systèmes représente un défi scientifique par rapport à la prise en compte de la
diversité des systèmes de production et de la variabilité des émissions au champ. Les objectifs de ce
chapitre sont (1) de produire des références de résultats d’ACV pour la tomate en maraîchage urbain
au Bénin et (2) d’explorer la variabilité des résultats associée à 4 facteurs influençant les émissions
d’azote réactif au champ.
L’unité fonctionnelle choisie correspond à 1 kg de tomate. Les inventaires ont été construits à partir de
données issues de 12 parcelles, regroupées selon 4 critères : le système d’irrigation, la source d’eau, la
date de semis et la localisation géographique. Ces parcelles ont également été agrégées en une
moyenne pondérée représentative de la fonction étudiée. Les catégories d’impact analysées sont la
privation d’eau douce, le changement climatique, la toxicité humaine, la formation photochimique
d’ozone, le potentiel d’acidification terrestre, le potentiel d’eutrophisation (marine et eau douce), et le
potentiel d’écotoxicité (terrestre, marine et eau douce). L’analyse de sensibilité des émissions d’azote
réactif a été réalisée à partir de 8 scénarios agronomiques (minimum et maximum pour 4 drivers) pour
lesquels les émissions ont été estimées avec un modèle biophysique. Les drivers analysés sont le
volume d’irrigation, la fertilisation azotée, le pH du sol et l’humidité maximal à capacité au champ.
Il apparait que les impacts environnementaux de ces systèmes de production de tomate en jardin
urbain au Bénin sont supérieurs aux impacts de systèmes maraichers Européens. Les avantages du
climat favorable au sud du Bénin et du faible besoin en infrastructure sont annulés par de faibles
performances des systèmes d’irrigation, d’importantes émissions azotées et l’utilisation excessive
d’insecticides. L’analyse de sensibilité montre que les résultats d’impact sont fortement sensibles aux
variations d’émissions obtenues avec le modèle. L’analyse met notamment en évidence les risques de
transferts de pollutions (ex : la réduction de la volatilisation induit une augmentation du lessivage) et
la complexité des mécanismes biophysiques impliqués dans les émissions.
Il apparait que l’intégration des paramètres relatifs à l’irrigation et aux propriétés du sol (pH, capacité
au champ) dans l’estimation des émissions au champ permettrait de produire des résultats d’ACV plus
spécifiques et des recommandations plus fines pour réduire l’impact des systèmes étudiés.
Mots-clefs: ACV  Agriculture urbaine  Diversité des systèmes de production  Variabilité des
émissions azotées
121
122
Investigation of the variability of life-cycle assessment of tomato production

from Beninese urban garden
4.1. Introduction
At the beginning of the last century, only 14% of the world population lived in cities. Since 2007, the
share of the world population living in cities reached 50% and is projected to reach 70% by 2030
(Kulikowski, 2007). This increase of population is correlated to an increase of the level of poverty in
cities.
Urban agriculture in general, and urban gardening in particular host the most potential to increase
citizen livelihood providing balanced food, reliable incomes and labor opportunities for urban citizens
(Orsini et al., 2013). The rapid expansion of those production systems led to the intensification of
agricultural practices to cope with the important land pressure on urban areas. Several studies
underlined the environmental risk associated to this intensification leading to excessive use of
fertilizers and pesticides (De Bon et al., 2010; Orsini et al., 2013). However, to date, a comprehensive
environmental assessment, representative for a diversity of agricultural practices for urban horticulture
in Africa is still lacking.
Life Cycle Assessment (LCA) is a normalized methodology (ISO 14040 and 14044) for the
environmental assessment of products and services consisting of four steps: the goal and scope
definition, the inventory, the environmental assessment and the interpretation. The relevance of LCA
to study the environmental impacts of agricultural systems has been validated for a large number of
products. In Europe, the use of detailed databases to characterize production system diversity (Mouron
et al., 2006; Boulard et al., 2011) or the use of modeling tools to accurately estimate emissions
(Bessou et al., 2013; Dufossé et al., 2013) led to significant improvement of the quality of LCA
studies applied to agricultural systems.
By contrast, the application of LCA to vegetables in tropical contexts is almost inexistent. The
achievement of life-cycle impact assessment results (LCIA) for vegetable production from urban
gardens in Benin represents a vast scientific challenge. The lack of technical and environmental data
and the absence of specific methods to estimate field emissions for tropical vegetable cropping
systems reduce the reliability and the comparability of LCIA results for their products (chapter 1). The
comparison of contrasted production systems is limited by uneven practices reasoning and
agronomical performances (chapter 2). The data scarcity limits the possibility to accurately estimate Nr
emissions with modeling tools (chapter 3). Finally, the lack of LCA studies applied to similar systems
hinders the ability to conclude on the meaning of potential impacts. There is an urgent need for
123
transparent LCA studies addressing diversity and variability issues specific to vegetable systems from
tropical contexts.
Based on a case study for out-of-season tomato in Benin showing extreme data scarcity, the objectives
of this paper were (1) to produce novel LCIA results for vegetable production from urban gardens in
Africa, and (2) to investigate the potential sources of LCIA variability associated to the estimation of
Nr emissions.
4.2. Material and method

4.2.1. Goal and scope definition
Here we present the methodology of the study which corresponds to the first step of the LCA
methodology as described initially in ISO 14044 (2006) and more recently in the ILCD Handbook
(European Commission, 2010).
Goals of the study

The aims of this LCA study were to determine the environmental performances of out-of-season
tomato production in Benin. To investigate the variability arising from both soil/climate drivers and
management practices across cultivation fields associated to LCIA results, two different aspects were
considered:
 The variability of the environmental impacts of cropping systems according to 4 key criteria:
irrigation system, water resource, planting date and location,
 The sensitivity of results to 4 drivers affecting Nr emissions: water inputs, nitrogen inputs, soil
pH and field capacity at saturation.
Functional unit
One kilogram of fresh yield and one hectare of tomato production were chosen as functional units as a
reference to express input, output flows and subsequent LCIA results. Out-of-season tomato
production in Benin shows a multiplicity of functions, i.e. producing fresh food all-year round,
utilizing in the most effective manner one hectare of agricultural urban land, providing cash flow to
farmers and employment for urban citizens. However, given the great variability of yield (chapter 2)
and profit for those systems, choosing 1 kg of tomato or one unit of profit as unique functional unit
would have introduced a bias in our LCIA results. Thus, we identified the land use as an interesting
complementary function allowing to account for each individual farm whatever its yield and profit.
124
Data origin and quality

As presented in chapter 2, a typology-based procedure allowed exploring the actual diversity of
cropping systems for out-of-season tomato production in Benin. Ten types were defined as the
existing combination of modalities for the three selected technical criteria (irrigation system, water
resource and planting date) within the population. Based on this typology, 12 cropping systems were
selected and site-specific primary data were collected during the 2011-2012 dry season. The dataset
obtained covered all cropping stages from sowing to harvest and included irrigation, fertilization and
crop protection aspects. The quality of this dataset was assessed using the matrix of pedigree and
scored (1, 3, 3, 1, 1) over the five criteria, namely reliability, completeness, technological,
geographical and temporal correlation. From this dataset we built virtual systems in two steps (see
chapter 2 for more details).
A first step consisted in building one synthetic and representative system called weighted mean. Our
sample of 12 cropping systems allowed building 6 virtual scenarios on the basis of the typology of
cropping systems recorded at population scale. Those 6 virtual scenarios resulted from the arithmetic
average of data from monitored cropping system belonging to each of the 6 types for which cropping
systems were sampled (1 virtual scenario = 1 type = 1 to 4 fields). The weighted mean was then
calculated based on the weighted average of our 6 virtual scenarios according to their relative weight
on an area-basis for the population (Table 4.1).
A second step consisted in building a range of systems representative for the diversity at population
scale. Beyond the typology results as types and virtual scenarios, the 12 cropping systems may also be
gathered by modality for the 3 technical criteria and their location representative of the diversity of
cropping systems at population scale:
 The irrigation can be manual (14% of the population on area-basis) or done by hose requiring
a pump (79% of the population),
 The water originated mainly from drilling (65%), but can also originate from depression (6%)
or river (30%),
 The crop cycle can be set early (from September to October - 66%) or median (from
November to December - 34%),
 The cropping systems can be located at Cotonou (19%), Ouidah (20%) or Grand-Popo (61%).
 This aggregation led to 10 modality-specific systems: 2 irrigation systems, 3 water resources,
2 planting dates and 3 locations.
125
Table 4.1 Weighting factors based on area distribution used to calculate the weighted mean and subsequent
contribution of modalities within each criterion
Cropping systems 1 2 3 4 5 6 7 8 9 10 11 12
Virtual scenarios S1 S3 S5 S6 S8 S10
Weighting factors 2% 2% 2% 8% 45% 1% 1% 1% 1% 15% 15% 7%
Irrigation system Manual (14%) Hose (79%)
Water resource Depression (6%) Drilling (65%) River (30%) Dr.
Planting date Early (66%) Median (34%) E.

Location* C C O C G O O O G G O C
*C=Cotonou (19%), O=Ouidah (20%), G=Grand-Popo (61%)
Processes inventoried in the Ecoinvent database v2.2 were used as background data for energy
production (Dones et al., 2007), material production (Hischier, 2007), fertilizer production (Nemecek
& Kägi, 2007) and pesticide production (Sutter, 2010). Due to the complex logistic regarding the
importation of energy material and inputs to Africa we did not include their transports in our
inventories.
Methods for the estimation of field emissions

Gaseous emissions from energy combustion were calculated according to recommendations from
Nemecek and Kägi (2007) for petrol. Given the lack of specific data on phosphorus and pesticide
emissions in tropical conditions, we followed the recommendations from Nemecek and Kägi (2007) to
estimate these emissions which corresponded to the more up-to-date guidelines for agriculture. We
also based our estimation of NOx emissions on these guidelines (21% of N2O emissions). We used
emission factors from IPCC (2006) to estimate direct (1% of nitrogen inputs) and indirect (1% of NH3
emitted and 0.75% of NO3 emitted) nitrous oxide emissions (N2O) and to estimate nitrate (NO3)
leaching (as 30% of nitrogen inputs). Despite the lack of specificity of its emission factors, the IPCC
report remains the most consensual method to estimate emissions in our context. For ammonia (NH 3)
emissions from mineral fertilizers and poultry manure, emission factors from Bouwman and Van Der
Hoek (1997) were used since they correspond to tropical conditions (4% for NPK, 20% for poultry
manure and 25% for urea). Overall, to estimate fertilizer emissions we used the most recent or
appropriate methods available, that we will further refer to as the default emission factor method
(DEF).
Category of impact and LCIA methods

To assess the impact on resources, the water deprivation potential (WDP, expressed in m 3 equivalent)
was calculated using the ReCiPe method (Goedkoop et al., 2009) associated to water stress index
(WSI) from Pfister et al. (2011). The water deprivation potential results from the quantification of
126
water consumptions during the life-cycle, related to the water scarcity index (WSI) associated to the
area where the water was withdrawn.
The following 9 impact categories selected as relevant to assess the environmental impact of studied
systems were homogeneously calculated using the ReCiPe method.
To assess the impact on human health we selected three impact categories showing different
contribution profiles. First, the climate change potential (CCP, expressed in kg CO 2 equivalent) was
calculated using the equivalency factors from the 2007 IPCC report as recommended in the ReCiPe
method. It results from the quantification of greenhouse gases emitted to air, related to their specific
climate change potential over 100 years. Second, the human toxicity potential (HTP, expressed in kg
1,4-DB equivalents) results from the quantification of substances emitted to air, soil and water,
relative to their specific human exposition route (exposure) and their toxicity (effect). Third, the
photochemical ozone formation potential (OFP, expressed in kg NMVOCs 1 equivalent) results from
the quantification of NOx and NMVOCs emitted to urban air, related to their ability to increase the
ozone concentration in the atmosphere.
To assess the impact on ecosystems we selected seven impact categories showing different
contribution profiles. The terrestrial acidification potential (TAP, expressed in kg SO2 equivalents)
results from the quantification of substances emitted to the atmosphere which cause changes in levels
of acidity when deposited on soils, related to their fate and effect on ecosystems. The freshwater
eutrophication potential (FEP, expressed in kg P equivalent) and the marine eutrophication potential
(MEP, expressed in kg N equivalent) result from the quantification of all emissions of P (FEP) and N
(MEP) to air, water and soil and of organic matter to water, related to their fate in ecosystems and the
sensitivity of receiving compartment. Terrestrial and freshwater ecotoxicity potentials (TETP and
FETP expressed in kg 1,4-DB equivalents) result from the quantification of chemical substances
emitted to soil (TETP) and water (FETP), related to their specific environmental persistence (fate), and
toxicity (effect).
4.2.2. Inventory data

One inventory was built for each individual cropping system monitored as the assembly of:
 One hectare of nursery, weighted by the nursery area actually transplanted,

 One hectare of plot weighted by its duration,
 Irrigation infrastructure according to their duration of use.
1
Non Methane Volatile Organic Compounds
127
The resulting 12 inventories were assembled into the weighted mean using the weighting factors
presented in table 4.1. In parallel, the 12 inventories were assembled into 10 modality-specific
inventories. In this section we present calculated inventory data in consistence with the LCIA results,
that is aggregated according to the same rules. Table 4.2 presents the parameters used to build
inventories. The nursery area used to transplant 100 square meters of tomatoes did not vary much
across cropping systems. Main variations resulted from variations of plant density. Fuel consumption
did not vary much except within modalities where cropping systems with pump were mixed with
cropping systems without pump (manual irrigation due to one system with intermediary tank, early
planting date and systems at Cotonou). Irrigation material was broken down into four elements, the
pump, the pipe connecting the pump to the plot, the hose used to irrigate and the sprinklers. For each
element, their specific weight and life-span were estimated based on farmer’s declarations.
128
Table 4.2 Parameters used for the inventory modeling of production processes per modality of each criterion and for the weighted mean (WM)
Irrigation System Water Resource Planting Date Location
WM
Manual Hose Dep. Drilling River Early Median Cotonou Ouidah GPP
Transplant on plots
Mean 0.005 0.009 0.005 0.009 0.005 0.008 0.007 0.007 0.004 0.01 0.007
Nursery area (m².100 m-2)
CV 47% 80% 57% 71% 110% 77% 73% 16% 105% 43% 71%
Nursery length Mean 43 23 36 29 28 38 23 45 23 25 27
(days) CV 33% 28% 15% 58% 15% 37% 30% 28% 33% 18% 48%
Plot density Mean 3.4 4.2 3.7 4.1 3.4 4.1 3.7 3.9 3.0 5.5 3.6
(plant.m-2) CV 44% 41% 48% 43% 2% 48% 30% 27% 28% 37% 39%
Plot length Mean 76 74 75 76 68 77 71 73 69 85 78
(days) CV 10% 28% 12% 26% 23% 13% 29% 11% 32% 8% 20%
Irrigation energy
Mean 0.3 1.5 0 1.4 1.3 0.8 1.3 0.5 1.1 1.7 0.9
Fuel consumption (l.h -1)
CV 200% 51% - 55% 28% 140% 33% 115% 67% 69% 74%
Mean 450 2496 0 2067 3300 933 2579 900 1860 2724 1591
Pump efficiency (m3.h-1)
CV 200% 30% - 25% 13% 110% 31% 115% 67% 30% 50%
Irrigation material
129
Mean 7 27 0 27 27 14 27 14 22 27 17
Pump (kg)
CV 200% 0% - 0% 0% 110% 0% 115% 56% 0% 35%
Pump Mean 828 5585 0 4093 7527 2416 4868 1152 3412 7346 3438
Life-span (h) CV 200% 58% - 75% 35% 161% 59% 136% 104% 30% 75%
Mean 0 14 0 10 13 1 15 0 11 15 7
Hose (kg)
CV - 34% - 79% 47% 245% 31% - 70% 35% 60%
Hose Mean 0 4219 0 2677 5396 618 4304 0 3740 3612 2260
Life-span (h) CV - 27% - 72% 23% 245% 28% - 62% 27% 51%
Mean 9 136 0 106 162 32 146 29 87 172 81
Pipe (kg)
CV 200% 70% - 89% 48% 119% 68% 129% 119% 48% 83%
Pipe Mean 552 5220 0 3947 5611 1210 5266 579 4990 3862 2873
Life-span (h) CV 200% 34% - 59% 56% 168% 37% 188% 62% 24% 64%
Mean 0 0 0 0.6 0 0.7 0 1.1 0 0 0.2
Sprinkler (kg)
CV - - - 265% - 245% - 200% - - 278%
Sprinkler Mean 0 0 0 289 0 338 0 506 0 0 104
Life-span (h) CV - - - 265% - 245% - 200% - - 278%
Table 4.3 presents a summary of inventory data for 10 modality-specific production processes and the
weighted mean. Further details can be found in chapter 2 and appendix 4-5. Cropping systems with
manual irrigation, early crop cycle or located at Cotonou presents particularly high insecticide
treatment frequency index (TFI). Frequent insecticide treatments were mainly based on Cypermethrin
(chapter 2, Figure 2.3). Fungicide and nematicide treatments were variable across cropping systems
(CV = 126 and 169% respectively). While fungicide treatments appear unevenly distributed across
cropping systems, the absence of nematicide treatments seems to be specific to cropping systems with
manual irrigation and water from depression. While most of cropping systems mainly applied products
based on terbufos, cropping systems with early crop cycles and located at Cotonou only applied
products based on carbofuran (chapter 2, Figure 2.3). All cropping systems groups present organic and
mineral fertilizer input. The location appears determinant regarding fertilization practices with high
fertilizer use at Cotonou and low fertilizer use at Grand-Popo. This pattern suggests that land-pressure
affects fertilizations practices (see chapter 2, section 2.3.3). Irrigation water volumes did not vary
much across cropping systems (CV=77%). Cropping systems with manual irrigation and located at
Cotonou high water use mainly due to one cropping system with intermediary tank used to stock
pumped water before manual irrigation, which led to water losses through evaporation. By contrast
cropping systems using water from river present low water use, mainly due to the soil type around
river (clay and silt) which reduces water losses through drainage.
130
Table 4.3 Agricultural input data per modality of each criterion and for the weighted mean (WM).
Irrigation system Water resource Planting date Location
WM
Manual Hose Dep. Drilling River Early Median Cotonou Ouidah GPP
Crop protection (TFI*)

Mean 12.0 5.7 7.6 7.4 7.8 9.9 5.1 11.6 4.9 6.5 8.2
Insect
CV 85% 92% 82% 124% 47% 87% 107% 91% 99% 87% 85%
Mean 2.5 1.1 3.4 1.1 1.8 2.7 0.8 3.3 0.4 2.0 1.9
Disease
CV 136% 131% 109% 134% 141% 99% 165% 95% 115% 90% 119%
Mean 0.0 2.1 0.0 1.6 4.3 0.8 2.5 1.2 2.8 0.2 1.8
Nematod
CV - 162% - 172% 119% 245% 146% 200% 138% 173% 132%
Organic fertilizers (kg .ha-1)
Mean 433 259 419 342 226 384 301 543 385 3 221
N
CV 104% 129% 131% 96% 141% 103% 115% 73% 83% 173% 157%
Mean 151 91 147 120 79 134 105 190 134 1 105
P2O5
CV
131
104% 129% 131% 96% 141% 103% 115% 73% 83% 173% 115%
Mineral fertilizers (kg .ha-1)
Mean 64 55 77 79 36 82 61 112 62 31 52
N-NPK
CV 76% 153% 65% 128% 141% 93% 150% 68% 159% 118% 140%
Mean 30 24 37 43 16 47 27 64 29 14 30
P2O5-NPK
CV 63% 153% 47% 140% 141% 115% 150% 95% 147% 118% 125%
Mean 56 37 61 15 101 38 43 56 0 87 37
N-urea
CV 111% 204% 125% 174% 141% 150% 187% 111% - 120% 187%
Irrigation water Mean 6175 4791 4763 5992 2290 5472 4664 6671 4099 4546 4768
(m3.ha-1) CV 66% 73% 76% 62% 93% 64% 82% 48% 99% 74% 73%
Yields Mean 6309 5123 5787 5219 3604 7733 2450 6307 1701 9121 9533
(kg.ha-1) CV 83% 144% 108% 147% 35% 105% 94% 83% 147% 115% 60%
*Treatment frequency index, see chapter 2, section 2.2.3 for more details
4.2.3. Design of and comparison with literature LCIA reference scenarios

For water deprivation potential (WDP), we can compare our LCIA results with literature scenarios
based on water volumes weighted by the WSI (Water Stress Index) from Pfister et al. (2011) for
tomato produced in Benin (“Benin x tomato”) and tomato produced in Spain (“Spain x tomato”). We
used yields from FAOSTAT for the year 2000 (FAO, 2013) to calculate water volume per hectare,
consistently with data on water consumption used in the original paper (Table 4.4). It is important to
notice that those literature scenarios only referred to water used for irrigation and not to water used in
background processes. However, volumes of water used in the background processes within studied
systems were low compared to volumes for irrigation water.
We can compare our LCIA results for global warming (GWP), acidification (AP) and eutrophication
(EP) potentials with literature scenarios based on results reported in a review of LCA of vegetable
products (chapter 1). The comparison with these average scenarios allowed us to get rid of the extreme
variability of LCA results on a case-to-case basis (chapter 1, Figure 3.1) while highlighting major
tendencies. The “Cold-greenhouse vegetable” scenario corresponds to the average impact from 12 cold
greenhouse soil tomato systems extracted from 5 papers (Antón et al., 2005; Boulard et al., 2011;
Martínez-Blanco et al., 2011; Romero-Gámez et al., 2012; Torrellas et al., 2012). The “Open-field
tomato” scenario corresponds to the average impact of 46 open-field tomato systems extracted from 7
(Mila i Canals et al., 2008; de Backer et al., 2009; Martínez-Blanco et al., 2009; Martínez-Blanco et
al., 2010; Martínez-Blanco et al., 2011; Romero-Gámez et al., 2012; Abeliotis et al., 2013). Main
characteristics are summarized in table 4.4.
Table 4.4 Characteristics for literature vegetable scenarios and Beninese out-of-season tomato (this study)
Benin x Spain x Cold-greenhouse Open-field Our study

Scenarios
tomato tomato vegetable vegetable (WM)
Yield (kg.ha-1) 5 399 60 469 119 808 41 582 9533
Irrigation water (m3.kg-1) 0.38 0.02 0.05 0.32 0.50
WSI 0.018 0.715 Ø Ø 0.023
-1 1
Energy consumption (MJ.kg ) Ø Ø 0.01 0.46 0.01
-1
Mineral fertilizers (kg N.kg ) Ø Ø 0.003 0.007 0.009
-1 2
Organic fertilizers (kg N eq.ha ) Ø Ø 0.001 0.025 0.023
1. Quantities of energy were converted to megajoul based on specific energy density for fuel (36 MJ.l -1), 2.
Quantities of nitrogen equivalent for organic fertilizers were calculated using an average nitrogen content of 3%
132
4.2.4. Design and analysis of simulated cropping system scenarios with the crop model
STICS
We used the crop model STICS (Brisson et al., 2002) as an alternative method to the DEF method to
estimate Nr emissions for the scenario analysis. This method will be further called the crop model
method (CM). This method aimed at including a diversity of specific parameters which influenced N r
emissions at the LCIA result level. Relevant impact categories for this scenario analysis were: Climate
change Potential, Eutrophication potential and Acidification potential. STICS was selected for its
generic aspect and its robustness to simulate crop yields (including open-field tomato), water and
nitrogen dynamics for a range of crops and environmental contexts. Input data consists in four files:
climate, soil, management and initialization. Climate data for Benin were provided by the local
international research center named AfricaRice. Soil data were collected for the 12 cropping systems.
Further detail on the model and input data can be found in the chapter 3 (sections 3.2.2 and 3.2.3) of
this thesis.
Regarding simulated scenarios, we built first a reference scenario based on the environmental
inventory of the weighted mean. To this end, we used the same aggregation procedure based on types
and weighting factors to calculate soil parameters used as inputs for the crop model (see chapter 2,
section 2.2.3). To build the daily management data used as inputs for STICS we considered:
 55 irrigation events occurring every day from the transplanting date up to the harvest with
same volume of water per event
 2 applications of organic fertilizer occurring respectively 5 and 30 days after transplanting
with the same nitrogen rate
 2 applications of mineral fertilizer occurring respectively 10 and 25 days after transplanting
with the same nitrogen rate.
Second, we selected 4 drivers responsible for important variations of N r emissions (chapter 3, Table
3.5). The variation of nitrogen inputs affects all emissions through the variation of nitrogen quantities
sensitive to emission processes and through the link between hexogen nitrogen and soil mineralization.
The variation of water use affects a large range of variables: yield, volatilized nitrogen and to a lesser
extent leached nitrogen. The variation of soil pH mainly affects the volatilization processes and the
variation of soil moisture at maximal field capacity affects the denitrification process.
Finally, we built 8 scenarios based on minimum (Min) and maximum (Max) values regarding each
driver independently (Table 4.5). Irrigation water use varied from 586 to 10,581 m3.ha-1 across the 12
cropping systems sampled. Both the number of irrigation events and the daily rates were correlated to
this variable. We adapted the number of irrigation events to 26 for the minimum scenario and 77 for
the maximum scenario. Nitrogen input rates varied from 21 to 1,184 kg N.ha -1. The type of fertilizer
133
used, the number of fertilization events and the application rate were correlated to this variable. We
used the overall distribution of fertilizer types shown by the weighted mean system (i.e. 71% as
poultry manure, 17% as NPK and 12% as urea), and varied the number of fertilizer applications from
1 to 3 for organic products and from 1 to 4 for mineral products. Soil pH varied from 4.6 to 6.9. This
parameter did not seem to be correlated to any other soil parameters, thus we could change it
independently of the other soil properties. Soil moisture at maximal capacity (shortened as maximal
soil moisture) varied from 9.6 to 61.6%. Unlike soil pH, this parameter was correlated to the clay
content, the quantity of organic nitrogen and the field capacity at wilting point. We therefore used the
actual values of maximal soil moisture for the two different soils of our sample.
Table 4.5 Key influencing drivers tested and corresponding favorable and unfavorable scenarios to emissions for
the scenario analysis of LCIA results
Influencing Scenarios
Influence on emissions Correlated variables
drivers Min Max
Water inputs The more water was supplied, Irrigation events: 26 to 77
586 10 581
(m3.ha-1) the higher was the leaching rate Daily rate
Types of fertilizers used

Nitrogen inputs The more nitrogen was supplied, Fertilization events: 1 to 3
21 1 184
(kg N.ha-1) the more Nr was emitted (organic) or 4 ( mineral)
Application rate
The higher was the soil pH,

Soil pH 4.6 6.9
the higher was the volatilization rate
Maximal soil The higher was the maximal soil Clay content
moisture moisture, the higher was the 9.6 61.6 Soil organic nitrogen content
(%) denitrification rate Minimal soil moisture
Finally STICS did not simulate indirect N2O emissions. We thus applied for those emissions the same
emissions factors as for the DEF method (i.e. N2Oi= 1% of NH4 as simulated with STICS + 0.75% of
NO3 emissions as simulated with STICS).
134
4.3. Results
4.3.1. LCIA results across the modalities and the weighted mean
Figures 4.2 to 4.4 present the environmental impacts of the 10 modalities and the weighted mean
calculated with the DEF method, expressed per kilogram of tomatoes. Each vertical panel presents one
criterion to which modalities were associated. Each horizontal panel presents one impact category
expressed in its reference substance. The figures also present the contribution analysis for each
modality and for the weighted mean. We presented the contribution of irrigation practices through the
energy production and use, the irrigation material production and the water consumption; the
contribution of fertilization practices through the fertilizer production, N r emissions and phosphorus
emissions during the crop stage and emissions during the nursery stage; and the contribution of crop
protection practices through the production and use of nematicides, fungicides and insecticides.
Categories mainly influenced by irrigation practices

Figure 4.2 presents LCIA results for five impact categories showing similar pattern of variation across
modalities: water deprivation potential (WDP), climate change potential (CCP), human toxicity
potential (HTP), photochemical ozone formation potential (POFP) and marine ecotoxicity potential
(METP). Main contributors for these impact categories were associated to irrigation practices.
The weighted mean had a WDP of 0.014 m3 eq. per kilogram of tomatoes. The major contributor was
the water used for irrigation (71%). Despite low water volumes involved in the inventory of energy
production and use and fertilizer production, their quite high contributions to WDP : 15% and 13%,
respectively can be explained by the use of the World average WSI for background processes
(WSI=0.669) which is higher than the local WSI used for irrigation water on farm (WSI=0.023 on
average). The WDP varied from 0.009 (modality Grand-Popo) to 0.097 (modality Ouidah) m3 eq. per
kilogram across modalities.
The weighted mean showed a CCP of 3.08 kg CO2 eq. per kilogram of tomatoes. The major
contributor was the energy production and use for water pumping (84%). The second most important
contributor was the nitrous oxide (N2O) emissions (direct N2O emissions with an 8% contribution and
indirect N2O emissions with 3%). The third contributor was the production of fertilizers (3%). The
CCP varied from 1.33 (Depression) to 11.55 (Ouidah) kg CO2 eq. per kilogram across modalities.
The weighted mean showed a HTP of 1,553.7 kg 1,4-DB eq. per kilogram of tomatoes. The only
contributor was the energy production and use for water pumping. The HTP varied from 0.1
(Depression) to 4,465.3 (Ouidah) kg 1,4-DB eq. per kilogram across modalities.
The weighted mean showed a POFP of 0.006 kg NMVOC per kilogram of tomatoes. The major
contributor was the energy production and use for water pumping (88%). The second most important
135
contributors were the nitrogen oxide (NOx) emissions and fertilizer production with 4% each. The
POFP varied from 0.002 (Depression) to 0.019 (Ouidah) kg NMVOC per kilogram across modalities.
The weighted mean showed a METP of 0.16 kg 1,4-DB eq. per kilogram of tomatoes. The only
contributor was the energy production and use for water pumping. The HTP varied from 0.01
(Depression) to 0.46 (Ouidah) kg 1,4-DB eq. per kilogram across modalities.
For the criteria: irrigation system, planting date and location, LCIA results variations resulted mainly
from yield variations. Nevertheless, some discrepancies can be observed for WDP. The influence of
irrigation volume led to similar WDP for the modalities of irrigation. In addition, there was also an
effect of the water stress index associated to the location (WSI= 0.0129 at Grand-Popo vs. 0.0325 at
Ouidah and Cotonou) which more clearly appeared when WDP was expressed per hectare of tomato
production (Annex 8). Regarding the water resource criterion, cropping systems using water from
depression presented the lowest CCP, HTP, POFP and METP thanks to the absence of energy
consumption for irrigation. Despite higher yields cropping systems using water from drilling presented
higher CCP, HTP, POFP and METP than cropping systems using water from river. These differences
resulted from higher irrigation water volumes for cropping systems using water from drilling (5219 vs.
2290 m3.ha-1), because the clay and silt soils around rivers reduce water losses through drainage. These
differences also resulted from lower fuel consumption, which can be attributed to lower energy needs
to pump surface water than ground table water.
136
Figure 4.2 LCIA results from cradle-to-harvest per kilogram of out-of-season tomato in Benin of the 10
modalities accounting for 4 criteria, and the weighted mean for 5 impacts categories: water deprivation potential
(WDP), climate change potential (CCP), human toxicity potential (HTP), photochemical ozone formation
potential (POFP) and marine ecotoxicity potential (METP)
137
Impacts mainly influenced by fertilization practices

Figure 4.3 presents LCIA results for three impact categories showing similar pattern of variation
across modalities: terrestrial acidification potential (TAP), freshwater eutrophication potential (FEP)
and marine eutrophication potential (MEP). Main contributors for these impact categories were
associated to fertilization practices.
The weighted mean showed a TAP of 0.02 kg SO2 eq. per kilogram. The main contributor was the
ammonia (NH3) emitted from organic fertilizers (71%) and from mineral fertilizers applications
(14%). Energy production and use and mineral fertilizers production also contributed to the TAP
impact of the weighted mean (8% and 4% respectively). The TAP varied from 0.01 (Grand-Popo) to
0.16 (Ouidah) kg SO2 eq. per kilogram across modalities.
The weighted mean showed a FEP of 0.0006 kg P eq. per kilogram. The main contributor was
fertilizer production (71%). Phosphorus emission and energy production and use also contributed to
the impact of the weighted mean (19% and 7% respectively). The FEP varied from 0.0004 (Grand-
Popo) to 0.0033 (Ouidah) kg P eq. per kilogram across modalities.
The weighted mean showed a MEP of 0.009 kg N eq. per kilogram. The main contributor was nitrate
emission (92%). The ammonia (NH 3) emitted from organic fertilizers also contributed to the impact of
the weighted mean (5%). The MEP varied from 0.004 (Grand-Popo) to 0.086 (Ouidah) kg N eq. per
kilogram across modalities.
For the planting date and location criteria, LCIA results variations resulted mainly from yield
variations. Consequently systems located at Cotonou showed lower impact than systems located at
Ouidah despite higher fertilizer uses. Regarding irrigation systems and water resources criteria there
was no clear impact difference between modalities. However, when LCIA were expressed per hectare
(Annex 8), we could identify the influence of mineral phosphorus use on FEP variations across
modalities of water resources.
138
modalities accounting for 4 criteria, and the weighted mean; for 3 impacts categories: terrestrial acidification
potential (TAP), freshwater and marine eutrophication potential (FEP and MEP)
Impacts mainly influenced by crop protection practices

Figure 4.4 presents LCIA results for two impact categories (TETP and FETP) each category showing a
different pattern of variation across modalities.
The weighted mean showed a terrestrial ecotoxicity potential (TETP) of 0.21 kg 1,4-DB eq. per
kilogram of tomato. The major contributor was the insecticide production and use (90%). Nematicide
139
and energy productions and uses contributed respectively for 7 and 3% of the total impact. The TETP
varied from 0.16 (Grand-Popo) to 0.38 (Cotonou) kg 1,4-DB eq. per kilogram across modalities.
Unlike other categories, TETP variations were not much driven by yield variations. Despite higher
yield, cropping systems with manual irrigation systems showed higher TETP than systems with hose
irrigation, due to greater use of insecticides. For the same reasons there was not much difference of
TETP between modalities of water resource and planting date and cropping system at Cotonou
presented higher TETP than cropping system at Ouidah. Despite lower yields, similar insecticide use
and relatively high contribution of nematicides, cropping systems using water from river presented
lower TETP. This can be explained by a lower share of cypermethrin which presents the highest
toxicity potential over all active substances.
The weighted mean showed freshwater ecotoxicity potential (FETP) of 0.05 kg 1,4-DB eq. per
kilogram of tomato. Similarly to the TETP, an important contributor was the insecticide production
and use. However the contribution for the latter only reached 50%. This lower relative contribution of
insecticide production and use resulted from the higher contribution of phosphorus emissions (35%),
nematicide production and use (8%), and energy production and use (4%). The FETP varied from 0.04
(Grand-Popo) to 0.14 (Ouidah) kg 1,4-DB eq. per kilogram across modalities. FETP variations were
very similar to TETP variations, i.e. mainly led by insecticide use in general and cypermthin use in
particular. However, the use of nematicide in general and Terbufos in particular contributed to reduce
the FETP difference between water resource modalities.
140
modalities accounting for 4 criteria, and the weighted mean; for 2 impacts categories: terrestrial and freshwater
ecotoxicity potential (TETP and FETP)
141
4.3.2. Comparison with scenarios from the literature

Table 4.6 presents LCIA results expressed per kilogram of product for the 3 literature scenario and for
the weighted meaning representative for the out-of-season tomato production in urban garden in
Benin.
The WDP of one kilogram of out-of-season tomato in Benin (our study, weighted mean) was higher
than the WDP of one kilogram of tomato in Benin (“Benin x tomato” scenario). This can be explained
by the intensive irrigation requirement linked to the out-of-season production. Indeed, in Benin most
of the tomato production is produced during the rainy season. Then, we also compared our results with
a literature scenario accounting for tomato from Spain (Pfister et al., 2011) to assess the effect of
changes in the value of WSI. The WDP of one hectare of out-of-season tomato in Benin (our study,
weighted mean) was similar to the WDP of one hectare of tomato in Spain (“Spain x tomato”
scenario). Indeed the higher water consumption for irrigation per kilogram of off-season tomatoes in
Benin was compensated the low WSI associated to water extracted in Benin. Indeed, the coastal area
of Benin presenting a sub-tropical climate, the impact of water use is lower than in drier areas such as
Spain.
The GWP of one kilogram of out-of-season tomato in Benin was higher than the GWP of one
kilogram of both European open-field vegetables (“Open-field x vegetable” scenario) and European
greenhouse vegetables (“Cold-greenhouse vegetable” scenario). The yield difference is likely to be the
main driver for such impact difference. However while the yield of European open-field vegetables
was four times the yield of out-of-season tomato in Benin, the GWP increased by a factor 30. This
difference can be explained first by higher irrigation water inputs for Benin tomato (0.50 for the WM
vs. 0 0.32 m3.kg-1 for the “Open-field vegetable” scenario). Indeed, our systems being out-of-season
they required more water than seasonal systems. The difference can also be explained by the source of
energy used, which was an important driver for variations of the global warming potential across
vegetable production systems (chapter 1). Unlike European producers Beninese producers had no
access to electricity in their urban gardens. One unit of energy from fuel shows a higher GWP than
one unit of the European electricity mix which often included electricity from nuclear power.
In agreement with the results for GWP, the AP for one kg of out-of-season tomato in Benin was higher
than the AP for one kg of European vegetable whether under cold-greenhouse or open-field. As the
AP increased by a factor 5 between the open-field vegetable scenario and our result, the yield
difference is likely to be the main driver of such an impact difference.
In agreement with the results for GWP and AP, the EP for one kg of out-of-season tomato in Benin
was higher than the EP for one kg of European vegetable whether under cold-greenhouse or open-
field. The EP increased by a factor 15 between the “Open-field vegetable” scenario and our result
despite the low contribution of energy production and use. This difference suggests that the high
142
impact of one kg of out-of-season tomato in Benin was due to high emission factors associated to
volatilized and leached nitrogen emissions.
Table 4.6 LCIA results per kg of product for WDP, global warming, acidification and eutrophication potentials
for literature vegetable scenarios and Beninese out-of-season tomato (this study)
Cold-greenhouse Open-field Benin x Spain x Our study

Scenarios
vegetable vegetable tomato tomato (WM)
WDP (m3 eq.) Ø Ø 0.007 0.015 0.014
*
Global warming (kg CO2 eq.) 0.37 0.11 Ø Ø 3.08
Acidification (g SO2 eq.)* 1.11 3.38 Ø Ø 16.74
*
Eutrophication (g PO4 eq.) 0.37 0.58 Ø Ø 8.15
* For comparison purposes we calculated our LCIA results using the CML methodology.
Regarding ecotoxicity impact categories, there was few relevant LCIA result from the literature to
compare our results with. Indeed, in the few papers dealing with ecotoxicity impacts of vegetable
production systems (Anton et al., 2004; Boulard et al., 2011; Abeliotis et al., 2013) the methodology
applied for LCIA calculation, and the estimation of emissions were not consistent with our case study.
However, we can notice a very different contribution pattern between our result compared to Boulard
et al. (2011) which presented a low contribution of pesticides to terrestrial ecotoxicity potentials (0.4%
and 11% of the total impact for respectively the multi-span greenhouse and the tunnel systems). If we
compare the terrestrial ecotoxicity potential for Beninese tomato to those presented by Boulard et al.,
(2011) for Tunnel system (excluding pesticide emissions), we end up with a higher impact for one
kilogram of out-of-season tomato in Benin (0.0047 vs. 0.21 kg 1,4-DB eq. per kilogram for the
weighted mean). This suggests that our results were particularly high which can be explained by the
frequent use of pesticides, and especially of some insecticides with very high terrestrial ecotoxicity
potential such as cypermethrin. We can notice the change in the scale of the y axis between terrestrial
and freshwater ecotoxicities (Figure 4.4). It suggests a higher terrestrial ecotoxicity potential of crop
protection practices compared with their freshwater ecotoxicity potential. The assumption of a 100%
emission to soil used in this study might lead to an overestimation of the terrestrial ecotoxicity
potential of some active substances. Indeed, simulations with recent estimation methods such as the
Pest-LCI model (Dijkman et al., 2012) suggest that an important part of pesticides do not reach the
soil, either because of crop interception or drift during spreading. Part of it is also biodegraded.
4.3.3. Scenario analysis to 4 drivers influencing N r emissions

Variability of Nr emissions
Choices we made regarding the estimation of Nr emissions contributed largely to the high impact
shown by one kilogram of out-of-season tomato in Benin regarding CCP, AP and EP. In the following
section we will analyze the sensitivity of these three impact categories to N r emissions within the
143
range of variations for the 12 surveyed fields. To consider the variations of N r emissions, we selected 4
drivers to build 8 agronomical scenarios for which we simulated N r emissions using the crop model
STICS. Table 4.7 presents the Nr emission factors simulated with the crop model method for the
weighted mean and the minimum and maximum scenarios for the 4 drivers compared with N r emission
factors obtained with the default emission factors (DEF).
Table 4.7 shows that using the crop model STICS to estimate the emissions for the weighted mean led
to a decrease of the emission factors for N2O (from 1 to 0.3%), for NH3 from organic fertilizers (from
20% to 12%) and, for NH3 from mineral fertilizers (from 11 to 3%). The volatilization rate was limited
by the effect of soil pH, while the denitrification rate was limited by the effect of maximum soil
moisture (chapter 3). By contrast, the leaching emission factor was increased (from 30% to 45%) due
to large water applications on soils sensitive to drainage (if water is not stored in soil, it drains).
Table 4.7 Nr emission factors obtained with the Default Emission Factors (DEF) approach (calculated for the
weighted mean only) and estimated with the crop model method (CM) for the 8 agronomic scenarios, minimum
(Min) and maximum (Max)
Weighted Maximum soil
Nitrogen rate Water volume Soil pH
Mean capacity
DEF CM Min Max Min Max Min Max Min Max
N2O 1.0% 0.3% 2.0% 0.2% 0.2% 0.3% 0.3% 0.3% 0.2% 3.7%
NH3 from organic
20% 12% 14% 12% 17% 12% 1% 51% 23% 4%
fertilizers
NH3 from mineral
11% 3% 4% 2% 3% 3% 0% 6% 3% 3%
fertilizers
NO3 30% 45% 0% 53% 0% 64% 42% 29% 71% 0%
Simulated yields
17,824 17,794 17,848 14,693 18,149 17,812 17,824 17,870 17,706
(kg.ha-1)
Regarding agronomic scenarios (Table 4.7), the variation of nitrogen rate led to a variation of all
emission factors. As expected, the NO 3 emission factor was lower for the minimum scenario and
higher for the maximum scenario compared to the weighted mean (leaching rate depends on the soil
nitrate concentration during drainage events). By contrast, emission factors for N2O and NH3 were
higher for the minimum scenarios than for the weighted mean. Similarly, N 2O and NH3 emission
factors were lower for the maximum scenario than for the weighted mean, but not for NH 3 from
organic fertilizers which remained constant. This means that changes in fluxes of N 2O and NH3
emissions were not proportional to changes of N rate. This can be explained by the large contribution
of mineralized nitrogen to emissions.
As expected the variation of water use led to a variation of NO 3 emission factors from 0 to 64% of the
total nitrogen fertilizer input. The minimum scenario showed a higher NH 3 emission factor for organic
fertilizer, however N2O and NH3 (min) emission factors were not much influenced by the irrigation
volume.
144
Regarding the effect of soil pH as expected, the higher was the soil pH, the higher were the NH 3
emissions. NH3 emission factors for organic fertilizers varied from 1 to 51% and from 0 to 6% for
mineral fertilizers. However, we noticed an unexpected effect on the NO 3 emission factor which
decreased compared to the weighted mean to reach 29% for the maximum scenario.
For soil moisture at maximum field capacity only the maximum scenario led to an important variation
of the denitrification rate. Thus, N2O emission factor reach 3.7% for the maximum. However, we
noticed that the variation of maximal soil moisture led to an opposite variation of NO 3 emission factor
(from 71 to 0%) and an opposite variation of NH3 emission factor from organic fertilizer (from 23 to
4%). The increase of leaching rate when maximal soil moisture decreased resulted from the higher
amount of drained water. The negative correlation of volatilization rate for organic fertilizers to
maximal soil moisture was more unexpected and probably resulted from the complex mechanism
involved in the acid-basic equilibrium governing the speciation of nitrogen in soil (chapter 3). Finally,
we can notice that the water volume was the only driver responsible for important yield variation
(from 14 693 to 18 149 kg.ha-1 vs. 17 824 kg.ha -1 for the weighted mean.
Variability of LCIA results

We then propagated this N r emission variability into LCIA results for the 3 impact categories impacted
by Nr emissions, namely climate change potential (CCP), terrestrial acidification potential (TAP) and
marine eutrophication potential (MEP). Figure 4.5 presents the relative LCIA results for the maximum
(Max) and minimum scenarios (Min) for each influencing driver independently. Results are expressed
as a percent of the impact for the weighted mean calculated with the crop model method for N r
emission estimation. Similarly to figures 4.2 to 4.4, figure 4.5 also presents the contribution analysis
for each scenario and each impact category.
For CCP, the most influencing driver was the nitrogen rate, because of higher indirect N2O emissions
resulting from the increase in NH 3 and NO3. The CCP of the minimum scenario only reached 83% of
the CCP of the weighted mean, while the maximum scenario presented a CCP 3 fold higher than the
weighted mean. The second most influencing drivers was the water volume, due to the high
contribution of energy consumptions for irrigation which was correlated to water inputs (Figure 4.2).
The CCP of the minimum scenario only reached 43% of the CCP of the weighted mean, while the
maximum scenario presented a CCP 2.6 fold higher than the weighted mean. The increase of CCP for
the maximum scenario can be explained by both the increase of energy consumption and the increase
of indirect N2O emissions due to the increase of NO3 emissions. As expected for the maximum soil
moisture driver, the maximum scenario showed a higher CCP (+38%) due to higher N 2O emissions.
However the minimum scenario showed an even higher difference to CCP than the favorable scenario
(+62%) due to higher indirect emissions. For the soil pH driver, both the minimum and the maximum
145
scenarios showed higher CCP compared with the weighted mean (+34% and +26% respectively) due
to indirect emissions either from NO3 emissions (minimum scenario) or from NH 3 emissions
(maximum scenario).
For TAP, the most influencing driver was the nitrogen rate, due to the high contribution of both NH 3
emissions and fertilizers production. The impact of the minimum scenario only reached 37% of the
weighted mean, while the maximum scenario was 3 fold higher. As expected, the TAP was also
largely influenced by the soil pH. The impact of the minimum scenario only reached 43% of the
weighted mean, while the maximum scenario was 2.9 fold higher. To a lesser extent, AP was
influenced by maximal soil moisture, with a higher TAP for minimum scenario (+49%) and a higher
TAP for the maximum scenario (-35%) compared to the weighted mean. Both the minimum and the
maximum scenario for water input exceeded the impact of the weighted mean (+62% and +30%
respectively) due to higher NH3 emissions.
For MEP, the most influencing driver was the nitrogen input, due to the high contribution of NO 3
emission. The MEP of the maximum scenario was 4.5 fold higher than the weighted mean, while the
minimum scenario reached only 3% of the MEP of the weighted. By contrast for the maximum soil
moisture driver, the MEP of the maximum scenario reached only 3% of the MEP of the weighted
mean, while the MEP of the minimum scenario was 1.6 fold higher. Similar MEP results were found
for the scenarios related to water volume and soil pH compared to the weighed mean. For the soil pH
driver it can be explained by a compensative variation in NH 3 and NO3 emissions associated to a
relatively low contribution of NH 3 emissions. For the water volume driver, this absence of effect
resulted from the variation of yields which softened the effect of the higher NO 3 emissions.
146
Figure 4.5 Scenario analysis for climate change potential (CCP), terrestrial acidification potential (TAP) and
marine eutrophication potential (MEP) to minimum (Min) and maximum (Max) scenarios for 4 drivers
influencing Nr emissions: nitrogen rate, water volume, soil pH and maximum soil moisture
147
4.4. Discussion
4.4.1. Impact of the irrigation management
In Benin, urban gardeners grow tomato out-of-season i.e. from September to April while tomato is
often grown during the season from April to August in other urban gardens in Africa (Diogo et al.,
2010; Sangare et al., 2012). The importance of out-of-season tomato production in urban gardens in
Benin results from the high competition with extensive seasonal production in rural area with no
irrigation system and low fertilization management, which reduces the profitability of tomato for
urban gardeners during the season. As an out-of season production, the advantage of the studied
system was its localization under a sub-tropical climate. Despite lower yields and higher water
volume, out-of-season tomato showed similar water deprivation potential compared to the Spanish
scenario due to the low WSI (Water Stress Index) calculated for this area (Pfister et al., 2011).
However, we could question the validity of an annual WSI, especially for a short-cycle crop in a
region where the water availability can vary from scarce to excessive within a climatic year. Monthly
WSI would allow a better capturing of water scarcity issues along the year (Pfister & Baumann, 2012)
and especially when the out-of-season tomato crop is grown.
Being under a sub-tropical climate prevents from the need for any specific infrastructure, which
potentially reduces climate change potential compared to average European out-of-season vegetables
grown under greenhouse. However, its low yield and high fuel consumption (high water inputs and
low equipment efficiency) led to higher climate change potential than the average European vegetable
whether under cold greenhouse and open-field. More generally, the major effect of irrigation
management on CCP was confirmed by the scenario analysis with STICS. It also demonstrated that
parameters referring to irrigation practices should be systematically included for the estimation of N r
emissions. Such inclusion may allow optimizing the irrigation management to reduce the water
deprivation potential, the climate change potential, as well as the acidification and eutrophication
potentials.
4.4.2. Impact of fertilization management

Similarly to other urban gardens in West Africa (Diogo et al., 2010; Sangare et al., 2012), Beninese
urban gardens showed intensive fertilization practices and relatively low yields. Organic and mineral
fertilizers inputs provide nutrient amounts largely above crop exportations (chapter 2). As a result,
average nutrients inputs per kilogram of tomato for our systems were higher than average nutrients
inputs for European open-field vegetables. Acidification and eutrophication potentials were higher
than that for European vegetable scenarios. This can be explained by the high emission factors
associated to Nr emissions which were selected to better represent our tropical context (Bouwman &
Van Der Hoek, 1997). On this respect, the emission factors we selected from Bouwman and Van Der
Hoek (1997) were above the emission factors calculated with STICS for the weighted mean. It
148
suggests that in most situations of our sample, the emission factor associated to volatilization would be
lower than the reference from the literature and that it would be desirable to include key parameters
associated to the volatilization process in the estimation method. The scenario analysis using STICS
for calculating Nr emissions showed that the main changes in acidification and eutrophication
potentials actually resulted from the variability of nitrogen input rates. The last point underlined by the
scenario analysis with STICS was the compensation between emission processes. Our results
suggested that reducing the volatilization rate may increase the leaching rate, due to higher soil
nitrogen content (compensative losses). Similarly, reducing the denitrification rate by reducing water
logging (low soil moisture capacity) led to higher leaching rate by increasing drainage and also led to
higher volatilization rate (this aspect remains unexplained though). These opposite behaviors for
emissions highlighted the complexity of bio-physical mechanisms involved in the processes of N r
emissions. More generally these results demonstrated that environment-friendly fertilization practices
could only be achieved by considering the irrigation management and soil properties.
4.4.3. Impact of crop protection management

Similarly to other urban gardens in West Africa (Cissé et al., 2003; Ahouangninou et al., 2011),
Beninese urban gardens showed intensive crop protection practices in terms of application frequency
and applied rates. Our results showed that studied system had particularly high terrestrial and
freshwater ecotoxicity potentials compared to European cold-greenhouse production systems (Boulard
et al., 2011). This difference can be partly explained by the emission estimation method we chose
which considered 100% of pesticides were emitted to soil (Nemecek & Kägi, 2007). Recent
developments showed that a significant part of pesticides applied can be absorbed and biodegraded on
plant leafs or in soils (Dijkman et al., 2012). However, we could also explain these large impacts by
the use of certain active substances with high terrestrial and freshwater ecotoxicity potentials such as
the terbufos and the cypermethrin. Our results suggested that the terbufos was even more critical than
the carbofuran (also a nematicide) often quoted as the most dangerous product for the environment
(Ahouangninou et al., 2012). The frequent use of cypermethrin at high rates was of particular concern
regarding its persistence in aquatic ecosystems and its high toxicity (Feo et al., 2010). Finally, in our
study, we did not consider the effects of pesticide residues on human health (Margni et al., 2002).
Considering the important pesticide residues found on vegetables from Beninese garden by
Ahouangninou et al. (2012), this aspect appears as an important limitation of our results and would
warrant further research.
149
4.5. Conclusions
We assessed the impact of one kilogram of out-of-season tomato production in Benin from cradle-to-
harvest. To this aim we calculated life-cycle impact assessment (LCIA) results for 10 modalities and
the weighted mean with default emission factors for field emissions. The 10 modalities presented
different characterization profiles for the 10 impact categories considered. The analysis of the
representative weighted mean and the diverse modalities allowed identifying hot-spots for one
kilogram of out-of-season tomato production in Benin. The results for our weighted mean were
compared with reference results for vegetables from the literature, highlighting the high environmental
impact associated to tomato production in urban gardens in Benin. Then we analyzed the sensitivity of
3 impact category results (CCP, AP and EP) to 4 drivers influencing N r emissions. This analysis
allowed us to underline the variations associated to LCIA results in order to refine the interpretation of
these results and suggest improvements for emission estimation methods.
We can conclude that the LCA results of out-of-season tomato production in Benin were diverse and
highly variable. We highlighted the fuel consumption, organic fertilizer use and insecticides
applications as major hot-spots for the studied function. Some variations due to management practices
and local drivers could be identified and quantified. We highlighted the major effect of irrigation on all
Nr emissions, which stresses the need of including systematically this parameter into the emission
estimation methods dedicated to irrigated systems. We also demonstrated that soil parameters played a
significant role for the estimation of N r emissions. Finally, our results also revealed weaknesses within
the approach to estimate emissions based on emission factors from nitrogen fertilizers as it does not
involve nitrogen inputs from soil mineralization which was a major input for our systems.
Based on the agronomical analysis of systems (chapter 3) and our LCIA results (this chapter) we
propose to explore in the general discussion potential margins for improvement for producers to
reduce the impact of their systems. We underlined methodological limitations for the application of
LCA to tomato systems grown on urban gardens in Africa. Based on our expertise on the inventory
modeling (chapter 3), on the analysis of N r emissions (chapter 4) and on LCIA results, we will
formulate in the general discussion scientific prospects which could allow improving the robustness of
LCA applied to such systems.
150
4.6. References
Abeliotis K, Detsis V, Pappia C (2013) Life cycle assessment of bean production in the Prespa National Park,
Greece. Journal of Cleaner Production 41:89–96. doi: 10.1016/j.jclepro.2012.09.032
Ahouangninou C, Fayomi BE, Martin T (2011) Évaluation des risques sanitaires et environnementaux des
pratiques phytosanitaires des producteurs maraîchers dans la commune rurale de Tori-Bossito (Sud-Bénin).
Cahiers Agricultures 20:216–222.
10.4236/jep.2012.33030
Anton A, Castells F, Montero JI, Huijbregts M (2004) Comparison of toxicological impacts of integrated and
chemical pest management in Mediterranean greenhouses. Chemosphere 54:1225–1235. doi:
10.1016/j.chemosphere.2003.10.018
Antón A, Montero JI, Muñoz P, Castells F (2005) LCA and tomato production in Mediterranean greenhouses.
International Journal of Agricultural Resources, Governance and Ecology 4:102–112.
De Backer E, Aertsens J, Vergucht S, Steurbaut W (2009) Assessing the ecological soundness of organic and
conventional agriculture by means of life cycle assessment (LCA) A case study of leek production. British
Food Journal 111:1028–1061. doi: 10.1108/00070700910992916
Assessment 18:24–36. doi: 10.1007/s11367-012-0457-0
Boulard T, Raeppel C, Brun R, Lecompte F, Hayer F, Carmassi G, Gaillard G (2011) Environmental impact of
greenhouse tomato production in France. Agronomy for Sustainable Development 31:757–777. doi:
10.1007/s13593-011-0031-3
10.1016/s1352-2310(97)00288-4
Brisson N, Ruget F, Gate P, Lorgeou J, Nicoullaud B, Tayot X, Plenet D, Jeuffroy MH, Bouthier A, Ripoche D
(2002) STICS: a generic model for simulating crops and their water and nitrogen balances. II. Model
validation for wheat and maize. Agronomie 22:69–92. doi: 10.1051/agro:2001005
Dijkman TJ, Birkved M, Hauschild MZ (2012) PestLCI 2.0: a second generation model for estimating emissions
of pesticides from arable land in LCA. The International Journal of Life Cycle Assessment 17:973–986. doi:
10.1007/s11367-012-0439-2
Dones R, Bauer C, Bolliger R, Burger B, Faist Emmenegger M, Frischknecht R, Heck T, Jungbluth N, Röder A,
Tuchschmid M (2007) Life cycle inventories of energy systems: results for current systems in Switzerland
and other UCTE countries. 5, Swiss Centre for LCI, PSI, Dübendorf and Villigen, Switzerland
doi: 10.1007/s12649-012-9171-1
FAO (2013) FAOSTAT: Crop yield per country for the year 2000. http://faostat3.fao.org.
Feo ML, Ginebreda A, Eljarrat E, Barcelo D (2010) Presence of pyrethroid pesticides in water and sediments of
Ebro River Delta. Journal of Hydrology 393:156–162. doi: 10.1016/j.jhydrol.2010.08.012
Goedkoop M, Heijungs R, Huijbregts M, De Schryver A, Struijs J, van Zelm R (2009) ReCiPe 2008: A life cycle
impact assessment method which comprises harmonised category indicators at the midpoint and the endpoint
level. Report I:
Hischier R (2007) Life cycle inventories of packagings and graphical papers. 11, Swiss Centre for LCI, Empa -
TSL, Dübendorf, Switzerland
151
Kulikowski M (2007) Mayday 23: World Population Becomes More Urban Than Rural. North Carolina State
University Newsroom, online: http://news.ncsu.edu/releases/2007/may/104.html
Margni M, Rossier D, Crettaz P, Jolliet O (2002) Life cycle impact assessment of pesticides on human health
and ecosystems. Agriculture, Ecosystems & Environment 93:379–392. doi: 10.1016/S0167-8809(01)00336-X
Martínez-Blanco J, Antón A, Rieradevall J, Castellari M, Muñoz P (2010) Comparing nutritional value and yield
as functional units in the environmental assessment of horticultural production with organic or mineral
fertilization. International Journal of Life Cycle Assessment 1–15. doi: 10.1007/s11367-010-0238-6
Martínez-Blanco J, Muñoz P, Antón A, Rieradevall J (2009) Life cycle assessment of the use of compost from
municipal organic waste for fertilization of tomato crops. Resources Conservation and Recycling 53:340–
351. doi: 10.1016/j.resconrec.2009.02.003
Martínez-Blanco J, Muñoz P, Antón A, Rieradevall J (2011) Assessment of tomato Mediterranean production in
open-field and standard multi-tunnel greenhouse, with compost or mineral fertilizers, from an agricultural
and environmental standpoint. Journal of Cleaner Production 19:985–997. doi:
10.1016/j.jclepro.2010.11.018
Pfister S, Baumann J (2012) Monthly characterisation factors for water consumption and application to
temporally explicit cereals inventory. Proceedings of the 8th international conference on life cycle
assessment in the agri-food sector
Pfister S, Bayer P, Koehler A, Hellweg S (2011) Environmental Impacts of Water Use in Global Crop
Production: Hotspots and Trade-Offs with Land Use. Environmental Science & Technology 45:5761–5768.
doi: 10.1021/es1041755
Romero-Gámez M, Suárez-Rey EM, Antón A, Castilla N, Soriano T (2012) Environmental impact of
screenhouse and open-field cultivation using a life cycle analysis: the case study of green bean production.
Journal of Cleaner Production 28:63–69. doi: 10.1016/j.jclepro.2011.07.006
Sutter J (2010) Life cycle inventories of pesticides. Swiss Centre for Life Cycle Inventories , St Gallen,
Switzerland
Torrellas M, Anton A, Ruijs M, Victoria NG, Stanghellini C, Montero JI (2012) Environmental and economic
assessment of protected crops in four European scenarios. Journal of Cleaner Production 28:45–55. doi:
10.1016/j.jclepro.2011.11.012
152
153
154
Chapitre 5. Discussion générale
© Aurélie Perrin
Irrigation au Tuyau sur une parcelle de tomate en jardin urbain à Ouidah, Bénin
155
- Chapitre 5. Discussion générale -
5.1. Introduction
Le taux de croissance urbain a atteint 4.6% en Afrique sub-saharienne, soit trois fois le taux de
croissance urbain des pays développés (UN-HABITAT, 2013). Ce changement dans la répartition de
la démographie mondiale s’est accompagné d’une augmentation de la pauvreté urbaine.
L’agriculture urbaine en général et l’horticulture urbaine en particulier est l’une des stratégies
largement développées, notamment en Afrique, pour réduire la pauvreté et améliorer le bien-être des
habitants des villes (Orsini et al., 2013). En effet les fruits et les légumes représentent une composante
importante pour un régime alimentaire équilibré (minéraux, vitamines et autres éléments
antioxydants). En plus de leur valeur nutritive, les fruits et légumes sont des cultures à haute valeur
ajoutée qui peuvent constituer une source de revenus pour les petits agriculteurs, et particulièrement
pour les femmes.
Aujourd’hui, la promotion de l’agriculture urbaine est freinée par le manque de données complètes et
fiables sur ces systèmes de production (FAO, 2012; Orsini et al., 2013; World Bank, 2013). En effet,
ce constat sur l’importance de l’agriculture urbaine renouvelle les enjeux associés à l’usage (souvent
non maitrisé) des fertilisants minéraux et organiques et des produits phytosanitaires. Plusieurs études
mettent en évidence les risques potentiels de l’agriculture urbaine sur la santé humaine et sur
l’environnement (De Bon et al., 2010; Drechsel & Dongus, 2010; Ahouangninou et al., 2012).
Néanmoins, la mise en œuvre de l’évaluation environnementale de ces systèmes, notamment par
l’analyse du cycle de vie, soulève de grands défis scientifiques parmi lesquels la prise en compte de la
variabilité de ces systèmes est sans doute la plus importante. La question centrale dans cette thèse a
ainsi été : comment prendre en compte la variabilité du sol, du climat et des pratiques pour la
réalisation d’ACV fiables ?
Pour ce travail, nous avons adopté une démarche intégrative combinant l’analyse agronomique
(typologie, modélisation, indicateurs simples) et l’analyse de cycle de vie (objectifs et cadre de l’étude,
inventaires et évaluation de l’impact). Nous avons travaillé à l’échelle de la parcelle et du cycle
cultural. Nous étions conscients dès le départ des fortes interactions à l’échelle supérieure, à savoir
l’exploitation que ce soit spatialement avec les autres activités de l’exploitation agricole ou
temporellement avec les autres composantes de la rotation. Néanmoins, la parcelle est la plus petite
unité de raisonnement de l’itinéraire technique, ainsi que l’échelle de simulation avec le modèle
biophysique. A ce titre, elle représente l’unité adéquate pour explorer dans un premier temps
l’influence de la variabilité du sol, du climat et des pratiques dans le cadre de l’évaluation
environnementale par l’ACV. Les implications de l’échelle n+1 et leur possible prise en compte seront
discutées dans le cadre de nos perspectives de travail.
Les principaux résultats de cette thèse sont résumés en figure 5.1 qui présente aussi le plan de
progression de cette discussion.
156
Nous avons tout d’abord réalisé l’état de l’art de l’ACV des produits maraîchers pour nous construire
une expertise sur les grands enjeux méthodologiques de l’ACV appliquée à ces produits. Nous avons
aborder chacun de ces enjeux dans la réalisation d’une ACV pour le cas d’étude de la tomate en
maraichage urbain au Bénin. Nous avons commencé par mettre au point un protocole de collecte de
données d’inventaire en réalisant une typologie sur des critères agronomiques et proposé une première
méthode d’estimation des émissions au champ (méthode dite par défaut ou DEF) basée sur les
meilleures références disponibles dans la littérature. Nous avons ensuite appliqué une approche
combinant un bilan d’azote avec l’utilisation du modèle biophysique STICS pour analyser la
variabilité des émissions d’azote réactif au champ et mis en évidence 4 principaux paramètres
influençant le bilan d’azote à la parcelle. Les jeux de données nouveaux ainsi obtenus ont été mis en
œuvre dans une Analyse de Cycle de Vie exhaustive puisqu’elle inclut en plus des catégories d’impact
plus classiques telles que changement climatique, eutrophisation, acidification, les impacts liés à
l’usage d’eau douce et les différents impacts de toxicité. Au final, quatre grands types de résultats ont
été produits : (1) une expertise sur les principaux enjeux relatifs à l’application de l’ACV aux produits
maraichers, (2) des développements méthodologiques pour la réalisation d’ACV en contexte tropical
pauvre en données, (3) des connaissances agronomiques et environnementales sur les systèmes
étudiées et (4) des références relatives aux résultats d’impacts environnementaux qui mettent en
évidence les points critiques des systèmes de production. Ces résultats nous ont permis de formuler
des recommandations pour de futures ACV de produits maraichers tropicaux et d’identifier des
perspectives de recherche communes aux disciplines de l’ACV et de l’agronomie.
Dans cette discussion nous allons donc dans un premier temps revenir sur l’approche de type méta-
analyse appliquée à la bibliographie ACV des produits maraichers pour mettre en valeur son potentiel
en termes d’expertise et de capitalisation des travaux existants. Ensuite nous reviendrons dans la
seconde section sur les moyens mis en œuvre pour prendre en compte la diversité des systèmes de
production (section 5.3.1) et intégrer la variabilité en termes de climat, de sol, et de pratiques agricoles
(section 5.3.2) au niveau de l’inventaire environnemental de notre système. Dans la troisième section,
nous reviendrons sur les résultats d’impact pour les systèmes de culture de tomate en maraichage
urbain, en se focalisant sur la portée des résultats et sur les marges de progrès envisageables. Dans la
dernière section de cette discussion, seront formulés des recommandations et des axes de recherches
communs aux deux disciplines de l’agronomie et de l’évaluation environnementale par l’ACV.
157
Objectif et périmètre de l’étude OBJECTIF 1 A 5
Tomate produite en jardin urbain au Bénin Chapitre 5
Discussion
(1) les principaux enjeux relatifs à
OBJECTIF 2 OBJECTIF 3 l’application de l’ACV aux produits
OBJECTIF 1 Chapitre 2 Chapitre 3
Chapitre 1 maraichers,
- Protocole de collecte de - Approche pour analyser la (2) des développements méthodologiques

- Faible comparabilité des
données pour réaliser des variabilité des émissions d’azote pour l’ACV en contexte tropical pauvre en
résultats d’impacts
inventaires représentatifs réactif au champ données
- Manque de transparence

- Méthode d’estimation par - Modèle candidat pour
des études (3) des connaissances agronomiques et
défaut (méthode DEF) simuler la variation des
- Manque de
- Caractéristiques émissions en fonction des environnementales
représentativité des données
agronomiques et ICV pour 10 pratiques et du sol (méthode (4) des références relatives aux résultats
- Inadéquation des
modalités représentatives de 4 CM)
méthodes vis-à-vis de la d’ACV
critères discriminants et 1 - 4 paramètres de variation des
diversité des systèmes
158
moyenne pondérée émissions d’azote réactif (5) Recommandations et perspectives :

étudiés
- Pas de prise en compte de
la variabilité des émissions OBJECTIF 4 Recommandations
Chapitre 4 - Intégrer les aspects socio-économiques
- Standardisation des
- Mesurer les flux à la parcelle
inventaires par type de
systèmes de culture - 1ère référence de résultats d’impact pour 10 catégories (Moyenne - Mettre à disposition des outils flexibles et
pondérée, DEF) faciles d’utilisation pour les émissions au
- Identification des points critiques à l’échelle de la population (10 champ
modalités, DEF) - Améliorer l’évaluation de l’impact des
- Analyse des sources de variabilité des émissions azotées à l’échelle de pesticides, des engrais organiques et de la
la population (8 scenarios, CM) consommation en eau
Typologie des systèmes de culture, Modélisation à l’échelle parcelle, Calculs d’indicateurs simples
Figure 5.1 Principaux résultats de la thèse et progression de la discussion générale

5.2. Une approche de type méta-analyse pour capitaliser sur le matériau

bibliographique
Revenons à l’une des finalités affichées de l’ACV qui consiste à produire des références fiables et
comparables concernant les performances environnementales des systèmes de production.
L’engouement récent pour l’évaluation environnementale des produits alimentaires a entrainé la
multiplication des études ACV dans le domaine agricole. L’approche de type méta analyse offre un
moyen efficace de capitaliser sur les études réalisées pour produire des références robustes et
pertinentes. Néanmoins, le peu de méta-analyses réalisées dans le domaine agricole s’accordent sur le
besoin (1) d’homogénéisation des objets d’étude, (2) d’informations relatives à l’échelle temporelle et
spatiale de l’étude et (3) d’informations relatives à la taille de l’échantillon et à l’importance de
l’erreur (Bengtsson et al., 2005; Seufert et al., 2012). Il est donc important de remettre en question les
moyens à mettre en œuvre pour réaliser des ACV fiables et comparables pour trouver des solutions qui
concilient le besoin de standardisation des pratiques d’ACV et de spécificité des outils et des données.
En préalable au travail réalisé sur le cas d’étude Bénin nous avons donc effectué une analyse
bibliographique des articles et rapports présentant des ACV « complètes » de produits maraichers
(chapitre 1). Dans ce cadre nous avons appliqué une approche de type méta-analyse pour analyser de
manière systématique et organisée les pratiques actuelles mises en œuvre en ACV ainsi que leurs
résultats. La méta-analyse consiste en l’analyse statistique d’une grande quantité de résultats issus de
différentes études dans l’objectif de conforter et mutualiser les résultats existants et aussi de dégager
de nouveaux résultats (Glass, 1976). Cette approche nous a d’abord conduits à analyser la qualité des
données d’inventaire présentées dans les différents documents disponibles. L’hétérogénéité de ce
matériau nous a conduits à proposer plusieurs méthodes d’agrégation des résultats d’ACV basés sur
une approche de typologie (à défaut de pouvoir appliquer des méthodes statistiques propres à la méta-
analyse).
Dans un premier temps, nous avons mis en évidence l’importante diversité des systèmes de production
maraichers et de leurs impacts. Cette diversité résulte d’une augmentation de la demande de fruits et
légumes frais tout au long de l’année et se traduit par une augmentation du niveau d’artificialisation
des systèmes de production. Nous avons établi qu’une part non négligeable des variations d’impact (en
particulier pour le changement climatique) peut être expliquée par un nombre restreint de paramètres
clefs (tableau 1.3). Par exemple, nous avons identifié le rendement et les pratiques de fertilisation
comme les principales sources de variation des performances environnementales des systèmes de
production maraîchers. Nous avons aussi mis en évidence que des paramètres très spécifiques comme
l’origine de l’énergie dans le cas de serres chauffées, ou les règles d’allocation associées à l’usage de
compost pour des systèmes plein champ pouvaient entrainer des variations non négligeables des
résultats.
159
Nous avons également identifié l’estimation des émissions au champ comme un point critique
générant des variations entre les études. En effet, non seulement la mise en œuvre des méthodes est
souvent décrite de manière insuffisamment précise, mais en plus les méthodes choisies elles-mêmes
présentent des domaines de validité inappropriés et une marge d’adaptabilité insuffisante pour
l’évaluation des flux des systèmes maraîchers. Un approfondissement bibliographique sur ces
méthodes d’estimation des flux au champ nous a permis de constater qu’aucune des méthodes
d’estimation employées dans les études ne peut être appliquée de manière générique à tous les
systèmes maraichers (tableau 1.3). Les recommandations formulées dans cette thèse peuvent
contribuer à la mise en place de standards propres à l’évaluation environnementale des différents
systèmes de culture maraîchers en tant que catégorie de produits plus ou moins homogène.
Finalement, nous avons aussi dû constater une grande variabilité des résultats dont il n’est pas toujours
évident de déterminer l’origine. Que celle-ci soit liée à des différences de qualité de données
employées ou à des différences réelles entres les systèmes étudiés, le manque de transparence des
études rend difficile son analyse. Cette variabilité doit être mieux caractérisée dans l’étape de
définition des objectifs et du champ de l’étude. Peu d’études peuvent prétendre présenter un niveau
d’information compatible avec les recommandations du guide ILCD (European Commission, 2010)
notamment en terme de description de la qualité des données (représentativité et fiabilité). Des
informations quantifiées sur la représentativité de l’étude et sur la taille de l’échantillon permettrait
d’intégrer cette variabilité dans la méta-analyse et de quantifier leur effet. Par exemple la prise en
compte d’un critère « échelle de l’étude » dans la méta-analyse consacré à l’effet de l’agriculture
biologique sur la biodiversité (Bengtsson et al., 2005) a montré que la différence entre agriculture
biologique et agriculture conventionnelle était plus marquée dans les études à petite échelle.
Plus généralement, ces standards doivent définir les données nécessaires et suffisantes à renseigner
pour la réalisation et la publication d’ACV et leur inclusion future dans une approche de type méta-
analyse. Ils doivent aussi proposer des méthodes transparentes et flexibles pour intégrer la diversité
des systèmes de production et identifier finement les sources de variabilité principales des impacts
environnementaux de la catégorie des produits maraîchers. C’est à cette ambition que nos
recommandations apportent une première contribution. Nous n’avons pas traité l’ensemble des sources
de variation. Notamment la diversité des logiciels et des méthodologies de caractérisation des impacts
n’a pas été abordée et pourrait faire l’objet d’une revue bibliographique dédiée.
160
5.3. Contribution à la réalisation d’inventaires environnementaux

5.3.1. Proposition d’un protocole de collecte des données pour réaliser des inventaires
représentatifs de la diversité des systèmes de production
La création d’inventaires représentatifs de la diversité des systèmes de production est l’un des défis
majeurs de l’application de l’ACV dans le domaine de l’agriculture. Les méthodes d’agrégation
comme la moyenne pondérée (Coltro et al., 2009) ou la scénarisation (Basset-Mens et al., 2007) et les
méthodes de désagrégation comme la typologie (Dalgaard et al., 2006) ou la modélisation (Nemecek
& Erzinger, 2005) ont montré leur potentiel. Ces tentatives intéressantes ne permettent cependant pas
de créer des inventaires dans un contexte où les données sont rares et l’expertise est difficile à
mobiliser. Dans cette thèse, la création d’inventaires environnementaux qui prennent en compte la
diversité des systèmes de production a reposé sur l’usage de la typologie à dire d’expert pour
sélectionner l’échantillon de systèmes à étudier et agréger les données collectées.
Cette démarche avait pour objectif de pallier les fortes contraintes de disponibilités de données dans le
contexte d’étude. Elle peut s’appliquer plus généralement dans le cadre d’analyses de systèmes
innovants ou de systèmes « de niche » pour lesquels les données génériques ne sont pas valides et les
données spécifiques sont insuffisantes. Elle a abouti au développement et à l’application d’une
procédure de création d’inventaires présentée dans le chapitre 2 (Figure 2.2). Nous avons ainsi créé
des inventaires environnementaux pour 10 modalités issues de 4 critères discriminants pour les
systèmes de culture de tomate de contre-saison au Bénin, et pour une moyenne pondérée qui agrège
cette diversité sur la base de la répartition des types au sein de la population en fonction de leur surface
relative. Ces inventaires présentent une qualité satisfaisante au regard des indicateurs de qualité des
données employées en ACV. Sur la base de la matrice des pédigrées développée par Weidema et
Wesnæs (1996) et dont l’usage est recommandé en ACV par les guides de bonne pratique récents
(European Commission, 2010), on atteint un score maximal de 1/5 pour les critères de fiabilité et de
corrélation temporelle et géographique. Seuls les critères de complétude et de corrélation technique
présentent des lacunes relatives à la durée de la période de collecte de données (seulement un cycle de
culture) et à la répartition des systèmes échantillonnés dans les types (pas suffisamment de système
avec irrigation par aspersion et pas de systèmes à cycle tardif).
Nous avons pu relier certaines caractéristiques agronomiques aux systèmes groupés par modalités de
la typologie. Aux trois critères initialement sélectionnés : le type de système d’irrigation, l’origine de
l’eau et la date de plantation, nous avons ajouté le critère de pression foncière qui repose sur la
distribution des types dans chaque commune et le gradient décroissant de pression foncière entre
Cotonou et Grand-Popo (la troisième commune de l’étude, Ouidah, présentant un profil intermédiaire).
Le critère origine de l’eau d’irrigation a un effet indirect sur la consommation en eau au travers de
l’effet type de sol par la présence de sols argileux aux alentours des rivières (l’une des sources d’eau
d’irrigation).
161
L’usage d’insecticides (et plus particulière de la cypermethrine) et de fongicide est plus intense pour
les systèmes avec une irrigation manuelle ou ayant recours à l’eau puisée dans une mare, ou avec un
cycle de culture précoce ou localisée à Cotonou. La répartition des parcelles au sein de ces modalités
se recoupant nous n’avons pu identifier si l’un de ces facteurs avait un effet dominant. Nous avons
proposé des hypothèses explicatives qui reposent sur une expertise de terrain mais qui n’ont pas pu
être vérifiées par les données techniques collectées (niveau de maitrise technique, accessibilité des
intrants, pression parasitaire).
Seule la localisation géographique semble être corrélée avec les pratiques de fertilisation avec des
pratiques d’autant plus intenses que le gradient de pression est fort. Les variations de rendement sont
plus difficiles à caractériser.
De manière plus globale et pour notre échantillon, nous avons mis en évidence un niveau
d’optimisation des systèmes de production imparfait, souvent lié à des conditions de production très
contraignantes (pression parasitaire, température élevée, faible accessibilité aux moyens de
production) et associé à une faible maitrise technique (faible appui technique, matériel de mesure
inadapté, non maitrise des instructions relatives aux intrants). Ce constat explique notre difficulté à
caractériser des types contrastés présentant une cohérence globale.
Nous avions envisagé d’autres pistes pour aller plus loin dans l’usage de la typologie pour la
réalisation d’inventaires diversifiés et contrastés. L’une d’elle consistait à utiliser les méthodes de
typologie quantitative, et notamment celles basées sur l’analyse en composante principale et la
classification hiérarchique. Ces méthodes ont démontré leur robustesse dans des contextes où la
typologie doit être établie avec peu de connaissances préalables et dans des contextes de grande
variabilité (Köbrich et al., 2003). De plus, nous avons identifié le potentiel de cette procédure pour
identifier de manière objective des individus représentatifs de la population (appelés parangons) sans
avoir recourt à l’agrégation de données qui nécessite des hypothèses et le calcul de moyennes au
détriment de la qualité des inventaires. Considérant le nombre important de variables descriptives des
systèmes de production, l’application de ces méthodes quantitatives nécessite cependant
l’échantillonnage d’un grand nombre d’individus. La typologie présentée par Blazy et al. (2009)
réalisée sur 46 variables à partir d’un échantillon de 66 fermes et présentant une forte discrimination
des résultats donne un bon exemple de l’ordre de grandeur minimum du ratio entre le nombre
d’individus et le nombre de variables étudiées. La taille de notre échantillon n’était pas suffisante et les
résultats préliminaires ont montré une faible puissance de discrimination (valeurs propres faibles
distribuées sur un grand nombre de dimensions), ce qui nous a poussé à abandonner cette analyse dans
notre contexte d’étude.
162
5.3.2. Proposition d’une approche intégrée pour l’analyse de la variabilité des émissions
La prise en compte de la variabilité des systèmes de production est l’un des défis majeurs de
l’application de l’ACV dans le domaine de l’agriculture. La variabilité des pratiques agricoles a été
analysée dans plusieurs articles par le biais d’analyses de scénarios (Mouron et al., 2006; Cellura et al.,
2012). Ces tentatives intéressantes ne permettent cependant pas de relier l’influence de pratiques
agricoles clés comme l’irrigation avec les émissions au champ faute de modèles d’estimation des
émissions suffisamment raffinés. D’autre part, les articles traitant de l’analyse de scénarios des
impacts aux émissions soulèvent le problème de covariance entre les paramètres impliqués dans les
processus d’émission (Payraudeau et al., 2007). L’approche que nous proposons s’intègre dans la
continuité des travaux sur l’analyse de scénarios, mais se distingue par une intégration cohérente des
paramètres de pratiques et de sol.
Compte tenu de la liste importante de paramètres influents (Tableau 3.1), il était nécessaire de
caractériser la complexité dans la gamme de variation du cas d’étude pour identifier un nombre
restreint de paramètres sensibles à propager sur les résultats d’ACV. Nous avons donc exploré la
variabilité des émissions dans la gamme de parcelles représentatives de la fonction étudiée. Pour cela,
nous avons combiné l’approche bilan d’azote avec la modélisation biophysique des flux (chapitre 3,
section 3.2.1). Cette combinaison nous a permis d’utiliser de manière cohérente un modèle dynamique
à pas de temps journalier dans un cadre d’analyse à l’échelle du cycle de culture. Nous avons identifié
4 paramètres présentant un degré de corrélation élevé avec les variations des composantes du bilan
d’azote (Tableau 3.5). Il s’agissait de la dose d’azote, du volume d’irrigation, du pH du sol et de
l’humidité du sol à la capacité au champ.
Compte tenu de la corrélation entre les différentes caractéristiques des systèmes de production la mise
en œuvre des méthodes de propagation de l’incertitude de type Monte Carlo nécessite le calcul
d’indices de covariances entre ces caractéristiques (Payraudeau et al., 2007), pour lesquels nous ne
disposons pas d’information a priori dans notre cas d’étude. Nous avons donc réalisé dans un troisième
temps des scénarios agronomiques pour propager cette variabilité sur les résultats d’impact. Sur la
base de valeurs minimum et maximum de chacun des paramètres au sein de la population, nous avons
construit des scénarios représentatifs de conditions réalistes favorables ou défavorables aux émissions
dans le domaine de variation de notre échantillon. Pour les paramètres relatifs aux pratiques agricoles,
ce travail de scénarisation a nécessité la modélisation à l’échelle journalière (date et fractionnement)
de variations de quantités à l’échelle du cycle de culture. Ce changement d’échelle s’est basé sur
l’expertise acquise sur les 12 parcelles étudiées. Pour les facteurs relatifs au sol, nous avons dû
considérer les relations biophysiques entre les différents paramètres d’entrée du modèle afin de
proposer des jeux de paramètres cohérents.
163
Cette approche nous a permis d’explorer la variabilité des inventaires de manière systématique et
transparente. Elle nous a permis d’identifier les principales sources d’incertitude relatives au modèle.
Le manque de données pour paramétrer finement le modèle (variété, potentiel de dénitrification,
engrais organique), et le manque de données mesurées pour le valider (bilan hydrique et azoté)
limitent la portée de ce type de modèle complexe pour raffiner l’estimation des émissions car il réduit
sa qualité prédictive (Wallach & Genard, 1998). Par ailleurs, l’incapacité du modèle à reproduire les
variations de rendement représente une limite d’ordre conceptuel. Compte tenu du degré de maitrise
technique faible des producteurs, il semblerait plus pertinent d’aborder la question du rendement avec
des outils probabilistes plutôt que mécanistes. Nous pouvons suggérer comme principaux paramètres
(ou ensembles de paramètres) pour établir des relations statistiques plus robustes : la pression
parasitaire, la température à la floraison et la connaissance technique des agriculteurs.
Enfin, l’approche bilan nous a permis d’invalider l’hypothèse d’un état stable de la matière organique
du sol pour lequel les entrées d’azote seraient équivalentes aux sorties. Il devient nécessaire d’explorer
le devenir de l’azote au-delà des limites du système en simulant avec le modèle un grand nombre
d’années climatiques. Néanmoins, une année climatique étant composée de plusieurs cycles de
cultures d’espèces différentes et présentant des itinéraires techniques propres, cette exploration
nécessite des données ou de l’expertise sur l’ensemble de la rotation. Ces limites soulèvent de
nouvelles questions scientifiques relatives aux données et aux outils qui feront l’objet d’une analyse
complémentaire dans la section perspective de cette discussion.
164
5.4. Contribution à l’évaluation environnementale de l’agriculture urbaine par la

méthode ACV
5.4.1. Impact environnemental de la tomate produite en maraîchage urbain au Bénin
Nous avons réalisé l’analyse environnementale d’un kilogramme de tomate produite en maraîchage
urbain au Bénin sur la base de 11 catégories d’impact. L’analyse de la moyenne pondérée nous a
permis de formuler pour chaque pratique, quelles sont les composantes critiques des systèmes de
culture par rapport à des références de systèmes maraîchers. L’analyse comparative des systèmes
regroupés par modalité pour les 4 critères discriminants de notre typologie nous a permis d’identifier
les leviers d’action pour réduire l’impact sur l’environnement de la fonction étudiée.
Nos résultats d’ACV sont les premiers pour un système de production de jardin urbain africain d’où la
difficulté à comparer avec d’autres références équivalentes. Quand cela était possible (WDP, GWP,
AP, et EP), nous avons identifié des scénarios issus de la littérature pour comparer nos résultats
d’impact par kilogramme de tomate (chapitre 4, section 4.3.2).
Il ressort qu’un kilogramme de tomate produite en maraîchage urbain présente un potentiel de

privation en eau plus élevé que la moyenne des systèmes de production de tomate au Bénin. Il ressort
également qu’un hectare de tomate produite en maraîchage urbain présente des potentiels de
changement climatique, d’acidification et d’eutrophisation plus élevés que les systèmes maraîchers
européens. Nous avons identifié la nature de l’énergie et la quantité d’énergie consommée par volume
d’eau prélevé comme les points critiques pour l’impact changement climatique. Nous avons identifié
l’effet du climat plus chaud sur le processus de volatilisation et la dose importante d’engrais
(notamment d’origine organique) par rapport aux apports totaux d’azote comme les points critiques
pour les impacts d’acidification et d’eutrophisation. Nous remarquons qu’une part non négligeable de
ces forts impacts repose sur les choix méthodologiques adoptés pour l’estimation des émissions au
champ dans la méthode de facteurs d’émission par défaut (DEF).
Enfin, malgré l’absence de références strictement comparables d’un point de vue méthodologique
(calcul des émissions et caractérisation des impacts), il ressort qu’un kilogramme de tomate produite
en maraîchage urbain présenterait des potentiels d’écotoxicité terrestre et aquatique plus élevés que les
systèmes maraîchers européens. Nous avons identifié l’utilisation de molécules avec un potentiel de
toxicité particulièrement élevé (des molécules qui sont d’ailleurs non homologués en Europe), les
applications systématiques et le surdosage comme les points critiques relatifs au potentiel de toxicité
des écosystèmes.
165
5.4.2. Résultats de l’analyse de scénario : vers un raffinement des méthodes d’estimation

au champ
Les effets prépondérants associés au choix du facteur d’émission sur le potentiel d’acidification et
d’eutrophisation remettent en question l’usage de méthodes d’estimation des émissions basées sur des
facteurs d’émissions peu spécifiques (méthode DEF). Les facteurs d’émission relatifs et génériques
sont-ils valides dans nos contextes de production spécifiques ? L’ACV peut-elle apporter des éléments
de réponse plus locaux pour limiter l’impact de ces émissions ? Le modèle biophysique STICS a été
utilisé pour simuler les émissions pour un scénario de référence (moyenne pondérée) et pour les 8
scénarios agronomiques représentatifs de situations favorables et défavorables aux émissions (chapitre
4, section 4.2.4). L’analyse de ces 8 scénarios nous a permis d’identifier les variations d’impact
associées à la variabilité des paramètres influençant le plus les émissions azotées.
En moyenne, le modèle suggère un facteur d’émission de protoxyde d’azote proche de celui proposé
par IPCC (2006). Par contre, les taux de volatilisation estimés par le modèle pour nos systèmes sont
inférieurs au facteur d’émission trouvé dans la littérature (Bouwman et Van Der Hoek, 1997) ainsi
qu’aux mesures dans des contextes comparables (Predotova et al., 2010; Lompo et al., 2012). Le taux
de lessivage semblerait lui plus élevée que le facteur générique proposé par IPCC (2006a) ce qui peut
s’expliquer par la forte occurrence de sols sableux et les forts volumes d’eau d’irrigation dans notre
échantillon.
Nous avons montré que les facteurs d’émission n’étaient pas constants et qu’ils variaient en fonction
des volumes d’eau apportés, de la dose d’azote et des paramètres de sol tels que le pH et la capacité au
champ (chapitre 4, tableau 17.4) et que ces variations étaient complexes à cause de l’interdépendance
des processus biophysiques impliqués. Les variations d’émissions engendrées par les 4 paramètres
d’influence étudiés entrainent des variations d’impact non négligeables. La quantité d’eau maximum
appliquée à l’échelle de la population entraine un potentiel de changement climatique 3 fois supérieur
à la référence. La dose d’azote maximum appliquée à l’échelle de la population entraine un potentiel
d’acidification 2,5 fois supérieur à la référence et un potentiel d’eutrophisation 3.2 fois supérieur à la
référence. Ces résultats montrent que les milieux sont très sensibles à une augmentation de la dose
d’azote du fait que la culture ne valorise pas ce surplus de nutriments. Enfin, l’analyse de scénarios a
permis de mettre en évidence des risques de transfert d’impact en fonction des interactions milieu x
pratiques. Ainsi une faible capacité au champ permet de réduire la dénitrification et donc la
contribution des émissions de N2O direct au changement climatique. Cependant, une faible capacité au
champ augmente aussi le drainage et la volatilisation entrainant une augmentation du potentiel
d’acidification, d’eutrophisation et au final une augmentation du potentiel de changement climatique
associée à de plus fortes émissions indirectes de N 2O. Ce phénomène est d’autant plus marqué que les
cultures sont irriguées et donc que les rendements sont peu sensible à d’éventuels stresses hydriques
associés à une faible capacité au champs.
166
Ces résultats soulignent le besoin de prendre en compte l’irrigation, le pH du sol et la capacité au

champ dans l’estimation des émissions azotées. L’usage de modèle biophysiques est une piste
prometteuse pour atteindre ce niveau de spécification et des exemples réussis se multiplient dans la
littérature (Gabrielle et Gagnaire, 2008; Bessou et al., 2013b; Dufossé et al., 2013). Cependant, la
complexité des modèles de culture engendre un besoin important en données de paramétrage et de
validation (section 5.3.2 de cette discussion générale) qui limite leur usage dans des contextes de
carence en données et en expertise. La nécessité de mettre à disposition des méthodes et des outils
adaptés et faciles d’utilisation fait l’objet d’une analyse complémentaire dans la section perspective de
cette discussion.
5.4.3. Identification des “hot-spots” à l’échelle de la population

Au Bénin, les producteurs des zones urbaines et péri-urbaines cultivent principalement la tomate en
contre saison (de septembre à avril), ce qui implique des pratiques d’irrigation particulièrement
intensives (en termes de système d’irrigation et de volumes d’eau apportés) par rapport à d’autres
jardins urbains en Afrique de l’Ouest où la tomate est principalement cultivée à partir d’avril en saison
plus humide. Cette particularité du Bénin résulte de la production massive de tomate pendant les
saisons plus humides en zone rurale. Ces modes de production plus extensifs affectent largement les
prix à cette période et rendent la production intensive en jardins urbains beaucoup moins rentable. Les
pratiques d’irrigation ont un impact environnemental principalement en termes de potentiel de
privation en eau, de potentiel de changement climatique, de toxicité humain, de potentiel de formation
photochimique d’ozone et d’écotoxicité marine.
Les performances des pratiques d’irrigation peuvent être améliorées par la promotion des systèmes
d’irrigation plus économes qui permet de réduire le volume d’eau appliqué. Dans un second temps, le
renouvellement et l’entretien des pompes permettrait de limiter les quantités d’essence consommées
par volume d’eau appliqué. Enfin, il pourrait également être intéressant de promouvoir la production
au bord des rivières ou des lagunes où les sols présentent une plus grande capacité de rétention de
l’eau, ceci en s’assurant que cela ne conduise pas à un transfert de pollution sous la forme de plus
grandes émissions de protoxyde d’azote (N2O) par dénitrification.
Au même titre que les autres jardins urbains d’Afrique de l’Ouest, les systèmes de production de
tomate en maraîchage urbain au Bénin présentent des pratiques de fertilisation intensives basées à la
fois sur des engrais minéraux (ou synthétiques) et des fertilisants organiques. Ces pratiques de
fertilisation ont principalement des impacts en termes de potentiel d’acidification et de potentiel
d’eutrophisation.
167
Nous avons constaté grâce au bilan d’azote à l’échelle du cycle de culture que les flux d’intrants azotés
et phosphorés étaient considérablement supérieurs aux exportations par la plante. Compte tenu de la
faible concentration en azote des fruits.et du niveau des rendements nos systèmes de culture Béninois
exportent peu d’azote comparé à d’autres systèmes maraîchers (chapitre 1). Il est nécessaire
d’optimiser les pratiques de fertilisation par rapport aux rendements. Cette optimisation doit prendre
en compte la sensibilité particulière de l’urée et de la fiente de volaille à la volatilisation en contexte
tropical. La limitation de l’usage des engrais organiques et de l’urée sera d’autant plus utile sur les sols
à pH élevé. Cependant, une diminution significative des émissions en azote ne sera permise qu’en
alliant une diminution de la dose totale d’engrais à une optimisation des pratiques d’irrigation.
Au même titre que les autres systèmes maraîchers d’Afrique de l’Ouest, les systèmes de production de
tomate en maraîchage urbain au Bénin présentent des pratiques de protection des cultures intensives
résultant à la fois de l’usage fréquent et en quantités importantes des produits phytosanitaires. Ces
pratiques phytosanitaires ont principalement des impacts en termes de potentiels de toxicité des
écosystèmes terrestres et d’eau douce, responsables d’une réduction de la qualité des écosystèmes.
Une nette amélioration des performances environnementales peut être atteinte pour nos systèmes par la
limitation de l’usage de certaines matières actives comme la cyperméthrine et le terbufos. Nous avons
émis l’hypothèse que la pression parasitaire était plus importante pendant le cycle de production
précoce. Il serait intéressant d’étudier les variations de pression parasitaire au cours du temps pour
promouvoir la culture de tomate pendant les périodes de pression parasitaire moindre. Enfin, la
formation et le suivi technique des producteurs sont indispensables pour réduire l’usage de produits de
manière systématique et à dose trop élevée.
168
5.5. Recommandations et perspectives pour améliorer la robustesse des résultats

d’ACV en contexte tropical
5.5.1. Intégrer les aspects socio-économiques
Notre approche a consisté à analyser les performances d’un système de culture, c’est-à-dire sans
intégrer pleinement les logiques d’exploitation à l’échelle temporelle (rotation) et spatiale
(assolement). Ce choix est cohérent avec l’approche produit de l’ACV, d’ailleurs la majorité des
études analyse les systèmes à cette échelle (chapitre 1, section 1.3.1). En effet, l’analyse à l’échelle du
cycle de culture s’est révélée pertinente pour l’identification de leviers d’action liés aux conditions du
milieu, permettant un raisonnement en fonction des interactions milieu x pratiques. Néanmoins, une
récente revue bibliographique (Peacock et al., 2011) rapporte que ce choix est souvent une limite
majeure des résultats d’ACV et nous ne pouvons qu’être d’accord avec ce constat. Pour notre cas
d’étude, nous avons constaté qu’une approche à l’échelle de l’exploitation présente un potentiel
explicatif plus fort pour caractériser les systèmes (section 5.3.1 de cette discussion générale). Dans un
contexte où les données et l’expertise sont rares, une meilleure compréhension des déterminants des
systèmes de production nous aurait permis d’élargir le cadre de notre analyse et de produire des
préconisations directement à l’attention du centre de prise des décisions : l’agriculteur. Nous avons
également constaté un défaut du bouclage du bilan d’azote à l’échelle du cycle de culture et émis
l’hypothèse que ce défaut pouvait être résolu à l’échelle d’une année climatique en prenant en compte
les autres cultures de la rotation (section 5.3.2 de cette discussion générale). A ce stade nous ne
pouvons qu’encourager fortement les initiatives d’évaluation à l’échelle de l’exploitation.
Dans le cadre de la petite agriculture, notamment en Afrique la ferme est assimilée à un ménage qui
représente l’unité de base sur laquelle sont réalisées les activités de production, de consommation et
d’éducation. Plusieurs études montrent que ce niveau d’intégration est pertinent pour l’analyse des
performances des systèmes de production et de leur impact environnemental (Bidogeza et al., 2009;
Tittonell et al., 2010; Sakané et al., 2012). Cependant, la description d’un ménage nécessite la collecte
d’un grand nombre de variables caractérisant la structure familiale, les différentes activités
économiques, les facteurs de production (travail et capital), ce qui nécessite un grand nombre
d’individus enquêtés. Par exemple, la typologie des ménages au Rwanda présenté par Bidogeza et al.
(2009) est basée sur un échantillon de 90 fermes. Dans cette perspective, nous avions envisagé
d’inclure des paramètres socio-économiques à notre typologie dès le début de la thèse mais ceci n’a
pas pu être mis en place faute de moyens suffisants. L’expertise acquise dans le cadre de cette thèse
souligne que ce type d’approche nécessite à la fois des compétences spécifiques (notamment en
sciences économiques et sociales) et la mise en place d’un protocole de collecte de grande ampleur.
Utiliser des modèles à l’échelle de la ferme voire de la région pour analyser l’allocation des ressources
est une alternative prometteuse (Parsons et al., 2011; Affholder et al., 2013). Ce type d’approche
« top-down » permet à la fois de limiter le nombre de données à collecter, de valoriser l’expertise
169
existante à l’échelle de l’exploitation et d’intégrer des logiques de stratégie d’exploitation à plus long
terme.
Enfin, l’intégration de paramètres socio-économiques au niveau de la définition de l’objet d’étude

représente un premier pas vers des démarches intégratrices à l’échelle de l’évaluation des impacts.
L’ACV sociale est un domaine de recherche en pleine expansion (Benoît et al., 2010) et son
application aux productions horticoles tropicales est pertinente et prometteuse (Feschet et al., 2013).
5.5.2. Collecter des données spécifiques sur les flux à la parcelle

Nos résultats démontrent que l’amélioration de la robustesse des résultats d’ACV ne pourra être
possible que par la production de données et de connaissances nouvelles notamment par le biais de
mesures biophysiques associées à des outils d’exploration. Pour notre cas d’étude, deux questions
précises peuvent être formulées : (1) Quel mode de gestion des engrais azotés permettrait de limiter
l’ensemble des émissions d’azote réactif vers l’air et l’eau? (2) Quelles sont les pratiques qui
permettraient de réduire le transfert des insecticides en général et des pyrethrinoïdes en particulier vers
les eaux de surface et les eaux souterraines? De plus, notre démarche a soulevé des questions de
représentativité et de fiabilité qui doivent être intégrées à ces questions scientifiques pointues pour
permettre à la connaissance théorique spécifique d’être utilisée de manière pratique dans un cadre plus
large d’évaluation des système de production. Par exemple, la problématique des pesticides doit tenir
compte des différents niveaux de risque associés aux variations de pression parasitaire, du manque de
maitrise technique des producteurs par rapport à l’identification des nuisibles et au dosage des produits
et de la toxicité spécifique de chaque molécule. D’autre part, la problématique des engrais azotés est
associée à la question des limites du système (délai entre l’apport d’azote et l’émission) et d’allocation
des impacts (répartition de l’apport en nutriments au niveau de la rotation). Nous avons montré dans
cette thèse que l’analyse du cycle de vie était génératrice de questions de recherche pointues sur le
fonctionnement des systèmes de production. L’ACV permet de cibler et de hiérarchiser les lacunes
scientifiques relatives à la fonction étudiée. Nous pensons qu’à l’avenir l’ACV mériterait d’être
intégrée de manière plus systématique en amont des démarches de définition de projets scientifiques.
5.5.3. Mettre à disposition des méthodes et des outils adaptés

Nous avons exploré la problématique des méthodes et des outils tout au long de cette thèse. Nous
avons d’abord insisté sur la nécessité qu’ils soient transparents et accessibles pour optimiser leur
utilisation par les praticiens en ACV. Après une analyse plus détaillée et une application dans le cadre
du cas d’étude, nous avons montré que ces méthodes et outils gagneraient à être également plus
flexibles et plus spécifiques. Enfin, notre expérience relative à l’utilisation du modèle STICS démontre
que ces méthodes et outils ne doivent pas être trop gourmands en données et/ou trop complexes à
170
prendre en main. Reprenons les 3 grandes stratégies possibles mises en évidence par Affholder et al.
(2012) dans un article sur la modélisation en agronomie : (1) déduire un modèle conceptuel des
connaissances acquises, (2) dériver un modèle simplifié à partir de modèles complexes et (3) identifier
et hiérarchiser les causes de variation de la ou des variables d’intérêt et utiliser des méthodes multi
variées pour les prédire. La première stratégie n’est pas facilement applicable aux productions
maraîchères en contexte urbain compte tenu du manque d’expertise. La dernière stratégie est limitée
par la disponibilité en données, néanmoins nous avons montré qu’il est possible de mettre en évidence
un nombre restreint de paramètres influents. Cette approche n’est donc pas à mettre de côté dans des
domaines qui bénéficient d’une littérature abondante. L’usage d’un modèle complexe validé présente
donc le meilleur potentiel pour une utilisation en ACV car il permet de couvrir une grande diversité de
systèmes, à différentes échelles spatiales et temporelles.
La première étape consiste à valider un modèle (adéquate) en définissant explicitement son domaine
de validité dans le temps et dans l’espace. Pour la diffusion de cet outil deux options sont alors
envisageables. La première option consiste à proposer une interface simplifiée tout en gardant
l’intégrité du modèle. Cette approche a notamment été utilisée en Australie pour mettre à disposition
des agriculteurs le modèle APSIM pour suivre les variations de leur potentiel de rendement et ainsi
adapter leur pratique au cours du cycle de production (Hochman et al., 2009). Une base de données est
mise à disposition de l’utilisateur pour lui permettre de décrire son système (climat, type de sol,
culture), et un nombre restreint de paramètres est suffisant pour spécifier les conditions de production
et analyser la variabilité associée à ces paramètres (irrigation, fertilisation, voire même pH). La
deuxième option est de simplifier le modèle complexe et de produire ce qu’on appelle un méta-
modèle. Britz et Leip (2009) présentent un exemple de méta modélisation (basée sur le modèle
DNDC) pour déterminer des facteurs d’émission spécifiques. Cette procédure consiste à réduire le
modèle à une fonction de réponse déterminée par un nombre limité de paramètres. Cette approche
demande un travail conséquent en amont de la mise à disposition de l’outil car elle nécessite des
données supplémentaires pour calibrer et valider le méta modèle. Par contre, ce mode de simplification
est ensuite moins coûteux en temps pour l’utilisateur.
5.5.4. Améliorer la prise en compte des pesticides, des engrais organiques et de la

consommation en eau
Le besoin en méthodes plus spécifiques largement abordée dans la section précédente sur les
émissions azotées est aussi valable pour les émissions de substances phytosanitaires. L’usage de
méthodes d’estimation plus fines permettrait d’apporter un niveau supplémentaire d’analyse et de
préconisation. Les émissions d’insecticides comme les pyréthrinoïdes vers les eaux sont
particulièrement préoccupantes à cause de leur persistance et de leur toxicité (Feo et al., 2010). Des
outils tels que PestLCI (Dijkman et al., 2012) permettent de raffiner la distribution des émissions dans
171
les différents compartiments de l’environnement en fonction de paramètres spécifiques de sol et de

climat. Cependant, ce type d’approche est souvent limité par un accès restreint aux données
nécessaires à l’évaluation et à la validation du modèle, garantissant une utilisation optimum. Enfin,
nous n’avons pas pris en compte la présence de résidus de produits phytosanitaires sur les tomates
après récolte. De récents développements permettent de quantifier ces résidus (Fantke et al., 2011)
pour évaluer leur impact potentiel sur la santé humaine. Cet aspect serait intéressant à intégrer compte
tenu de l’absence récurrente dans la pratique de délai entre le dernier traitement et la récolte.
Nous avons montré que l’usage d’engrais organiques est un point critique des performances
environnementales de la tomate en maraîchage urbain. L’usage de déchets de ferme et de déchets
urbains compostés ou non est une problématique majeure de l’agriculture urbaine (Orsini et al., 2013).
Une récente revue bibliographique identifie de nombreux bénéfices du compost au-delà de l’apport en
nutriments et propose des moyens pour mieux en tenir compte en ACV (Martínez-Blanco et al., 2013).
D’autre part d’autres travaux mettent en évidence des impacts (négatifs) qui ne sont pas toujours pris
en compte en ACV comme les émissions avant application (fabrication du compost) (Basset-Mens et
al., 2007) ou la dégradation des sols causée par les métaux lourds contenus dans les fertilisants
organiques (Anikwe & Nwobodo, 2002; Shao et al., 2012). L’intégration de ces flux (émissions avant
application et métaux lourd) et l’impact sur la qualité du sol permettrait d’améliorer la pertinence des
résultats d’ACV.
Enfin, l’évaluation du potentiel de privation en eau souligne la nécessité de considérer la localisation

des systèmes dans l’analyse de l’impact environnemental des systèmes irrigués. Il renforce également
le besoin de raffiner les facteurs de caractérisation. D’une part, nous n’avons pas pu discriminer
l’impact des différentes origines de l’eau. Or, le contexte régionale (climat, démographie, topographie)
et le contexte local (sensibilité des milieux environnants) infllue largement sur la nature et la gravité
des impacts. Au sud Benin par exemple il semble que prélever de l’eau dans un cours d’eau est moins
impactant que de prélever de l’eau dans la nappe qui est également l’unique ressource en eau pour les
consommations domestiques. Un niveau de spécification plus élevé aurait renforcé l’intérêt de
privilégier les zones de production en bord de rivière ou de lagune. D’autre part le manque de prise en
compte de la saisonnalité dans la caractérisation de la rareté en eau est une limite importante dans nos
contextes où l’eau peut être rare en saison sèche puis en excès en saison des pluies.
172
5.6. Conclusions
Nous insisterons pour conclure sur le fait que l’ACV est un outil de connaissance et de
compréhension. Associé à une analyse agronomique, elle nous a permis de souligner 3 principes
importants à retenir pour la réalisation d’ACV agricoles :
1. la représentativité des inventaires dépend de la qualité de prise en compte de la diversité des

caractéristiques agronomiques et socio-économiques des systèmes de production et de leur
niveau d’optimisation,
2. des indicateurs agronomiques simples tels que les bilans d’azote et les indices de traitements
en pesticides permettent un éclairage complémentaire efficace aux résultats d’impacts
environnementaux,
3. la prise en compte des processus biophysiques dans l’estimation des émissions d’azote réactif
augmente les performances prédictives de l’ACV et sa capacité à identifier d’éventuels
transferts de pollution.
Nous avons commencé cette discussion en soulignant le besoin de standardisation. Nous la terminons
en proposant des pistes de recherche très pointues. La méthodologie ACV propose une procédure
normalisée, mais c’est l’utilisateur qui définit le niveau de résolution. A nous chercheurs de proposer
des moyens efficaces pour repousser les limites de la connaissance et pour gagner en performance de
prédiction. Il reste du ressort des utilisateurs de bien poser leurs questions. J’espère que ma thèse aura
contribué à établir un dialogue fructueux entre agronomes et « ACV’istes » en mettant en évidence les
enjeux partagés.
173
5.7. Références bibliographiques

Affholder F, Poeydebat C, Corbeels M, Scopel E, Tittonell P (2013) The yield gap of major food crops in family
agriculture in the tropics: Assessment and analysis through field surveys and modelling. Field Crops
Research 143:106–118. doi: 10.1016/j.fcr.2012.10.021
Affholder F, Tittonell P, Corbeels M, Roux S, Motisi N, Tixier P, Wery J (2012) Ad Hoc Modeling in
Agronomy: What Have We Learned in the Last 15 Years? Agronomy Journal 104:735. doi:
10.2134/agronj2011.0376
10.4236/jep.2012.33030
Anikwe MAN, Nwobodo KCA (2002) Long term effect of municipal waste disposal on soil properties and
productivity of sites used for urban agriculture in Abakaliki, Nigeria. Bioresource Technology 83:241–250.
15:1395–1405. doi: 10.1016/j.jclepro.2006.03.009
Bengtsson J, Ahnström J, WEIBULL A (2005) The effects of organic agriculture on biodiversity and abundance:
a meta-analysis. Journal of applied ecology 42:261–269.
Benoît C, Norris GA, Valdivia S, Ciroth A, Moberg A, Bos U, Prakash S, Ugaya C, Beck T (2010) The
guidelines for social life cycle assessment of products: just in time! The International Journal of Life Cycle
Assessment 15:156–163. doi: 10.1007/s11367-009-0147-8
Bidogeza JC, Berentsen PBM, Graaff J, Oude Lansink AGJM (2009) A typology of farm households for the
Umutara Province in Rwanda. Food Security 1:321–335. doi: 10.1007/s12571-009-0029-8
Blazy J-MM, Ozier-Lafontaine H, Doré T, Thomas A, Wery J (2009) A methodological framework that accounts
for farm diversity in the prototyping of crop management systems. Application to banana-based systems in
Guadeloupe. Agricultural systems 101:30–41. doi: 10.1016/j.agsy.2009.02.004
Britz W, Leip A (2009) Development of marginal emission factors for N losses from agricultural soils with the
DNDC-CAPRI meta-model. Agriculture Ecosystems & Environment 133:267–279. doi:
10.1016/j.agee.2009.04.026
Cellura M, Ardente F, Longo S (2012) From the LCA of food products to the environmental assessment of
208. doi: 10.1016/j.jenvman.2011.08.019
Coltro L, Mourad AL, Kletecke RM, Mendonça TA, Germer SPM (2009) Assessing the environmental profile of
orange production in Brazil. The International Journal of Life Cycle Assessment 14:656–664. doi:
10.1007/s11367-009-0097-1
Dalgaard R, Halberg N, Kristensen IS, Larsen I (2006) Modelling representative and coherent Danish farm types
based on farm accountancy data for use in environmental assessments. Agriculture, Ecosystems &
Environment 117:223–237. doi: 10.1016/j.agee.2006.04.002
Dijkman TJ, Birkved M, Hauschild MZ (2012) PestLCI 2.0: a second generation model for estimating emissions
of pesticides from arable land in LCA. The International Journal of Life Cycle Assessment 17:973–986. doi:
10.1007/s11367-012-0439-2
Drechsel P, Dongus S (2010) Dynamics and sustainability of urban agriculture: examples from sub-Saharan
Africa. Sustainability Science 5:69–78. doi: 10.1007/s11625-009-0097-x
Fantke P, Juraske R, Antón A, Friedrich R, Jolliet O (2011) Dynamic multicrop model to characterize impacts of
pesticides in food. Environmental Science & Technology 45:8842–8849.
FAO (2012) Growing greener cities in Africa. Food and Agriculture Organization of the United Nations, Rome
Feo ML, Ginebreda A, Eljarrat E, Barcelo D (2010) Presence of pyrethroid pesticides in water and sediments of
Ebro River Delta. Journal of Hydrology 393:156–162. doi: 10.1016/j.jhydrol.2010.08.012
Feschet P, Macombe C, Garrabé M, Loeillet D, Saez AR, Benhmad F (2013) Social impact assessment in LCA
using the Preston pathway. The International Journal of Life Cycle Assessment 18:490–503. doi:
10.1007/s11367-012-0490-z
Glass G V (1976) Primary, secondary, and meta-analysis of research. Educational researcher 5:3–8.
Hochman Z, van Rees H, Carberry PS, Hunt JR, McCown RL, Gartmann A, Holzworth D, van Rees S, Dalgliesh
NP, Long W, Peake AS, Poulton PL, McClelland T (2009) Re-inventing model-based decision support with
174
Australian dryland farmers. 4. Yield Prophet (R) helps farmers monitor and manage crops in a variable
climate. Crop & Pasture Science 60:1057–1070. doi: 10.1071/cp09020
Köbrich C, Rehman T, Khan M (2003) Typification of farming systems for constructing representative farm
models: two illustrations of the application of multi-variate analyses in Chile and Pakistan. Agricultural
Systems 76:141–157. doi: 10.1016/S0308-521X(02)00013-6
Martínez-Blanco J, Lazcano C, Christensen TH, Muñoz P, Rieradevall J, Møller J, Antón A, Boldrin A (2013)
Compost benefits for agriculture evaluated by life cycle assessment. A review. Agronomy for Sustainable
Development 33:721–732. doi: 10.1007/s13593-013-0148-7
Nemecek T, Erzinger S (2005) Modelling Representative Life Cycle Inventories for Swiss Arable Crops. The
International Journal of Life Cycle Assessment 10:68–76. doi: 10.1065/lca2004.09.181.8
Parsons D, Nicholson CF, Blake RW, Ketterings QM, Ramirez-Aviles L, Fox DG, Tedeschi LO, Cherney JH
(2011) Development and evaluation of an integrated simulation model for assessing smallholder crop-
livestock production in Yucatan, Mexico. Agricultural systems 104:1–12. doi: 10.1016/j.agsy.2010.07.006
Payraudeau S, van der Werf HMG, Vertes F (2007) Analysis of the uncertainty associated with the estimation of
nitrogen losses from farming systems. Agricultural systems 94:416–430. doi: 10.1016/j.agsy.2006.11.014
Peacock N, De Camillis C, Pennington D, Aichinger H, Parenti A, Rennaud J-P, Raggi A, Brentrup F, Sára B,
Schenker U, Unger N, Ziegler F (2011) Towards a harmonised framework methodology for the
environmental assessment of food and drink products. International Journal of Life Cycle Assessment
16:189–197. doi: 10.1007/s11367-011-0250-5
Sakané N, Becker M, Langensiepen M, van Wijk MT (2012) Typology of Smallholder Production Systems in
Small East-African Wetlands. Wetlands 33:101–116. doi: 10.1007/s13157-012-0355-z
Shao XH, Hu XJ, Shan J, Liao LX, Tan JY, Kwizera C (2012) Environmental risk assessment of soil and
groundwater contamination with bio-organic fertilizer application. Journal of Food Agriculture &
Environment 10:1209–1212.
Tittonell P, Muriuki A, Shepherd KDD, Mugendi D, Kaizzi KCC, Okeyo J, Verchot L, Coe R, Vanlauwe B
(2010) The diversity of rural livelihoods and their influence on soil fertility in agricultural systems of East
Africa – A typology of smallholder farms. Agricultural systems 103:83–97. doi: 10.1016/j.agsy.2009.10.001
UN-HABITAT (2013) UN-Habitat Global Activities Report 2013. Available from: http://www.unhabitat.org.
Wallach D, Genard M (1998) Effect of uncertainty in input and parameter values on model prediction error.
Ecological Modelling 105:337–345. doi: 10.1016/S0304-3800(97)00180-4
World Bank (2013) Urban Agriculture - Findings from Four City Case Studies. n. 18, The World Bank Group,
Washington, USA
175
176
ANNEXES
© Césaire Dakpogan
Paniers utilisés pour la récolte des tomates, sur la route entre Grand-Popo et Cotonou, Bénin
177
178
ANNEXES
Annex 1. Global warming potential (GWP), acidification potential (AP) and

eutrophication potential (EP) expressed per square meter for a vegetable crop
cycle from cradle-to-harvest for the 72 systems aggregated per cropping system
type, product group and reviewed paper.
Figures in brackets correspond to the number of systems in the group. Grey boxes correspond to error
bars centered on the mean of the group (red dash)
179
Annexe 2. Liste des producteurs de tomate de contre saison pour la saison 2011-2012 sur les communes de Cotonou,
Ouidah et Grand-Popo
Cette liste a été établie en collaboration avec les Centres Communaux de Promotion Agricole (CeCPA) en charge de l’accompagnement des producteurs. Le
stade correspond aux observations et prévisions réalisées en Octobre 2012 (JAR/SAR/MAR= Jour/semaine/mois après repiquage). Dans la colonne
« Traçabilité » sont identifiés les problèmes de traçabilité dues à un stade trop avancé de la culture (A) et à l’incertitude associée aux systèmes les plus tardifs
(T). Les croix dans la colonne « Etat sanitaire » identifie les parcelles qui ont déjà subi des attaques parasitaires à la date de l’établissement de cette liste.
Surface Système Localisation du Etat Echan

Ville Nom du producteur Variété Précocité Source d’eau Stade * Tracabilité Type
m² d’arrosage site sanitaire -tillon
SOGLOHOUN
Cotonou 216 Mongal Précoce Mannuel Dépression PLM 4JAR 1 1
Léonard
ANNEXES
Cotonou HOUNTHO Faustin 234 Mongal Précoce Mannuel Dépression PLM Pépinière 1 2
180
Ouidah ALLIANCE Zodoté 3200 Akli. Précoce Mannuel Dépression Mahuclo 15JAR 1 3
Cotonou DANSOU Dominique 492 Mongal Précoce Mannuel Dépression PLM 1MAR A 1
AMOUSSOU
Cotonou 84 Mongal Précoce Mannuel Dépression Houéyiho 5JAR 1
Hamadou
Cotonou DANFADJI 252 Mongal Précoce Mannuel Dépression PLM Floraison A 1
Cotonou SEKANDE Bathélémy 702 Mongal Médian Mannuel Dépression PLM Pépinière 2
Forage +
Cotonou SANTOS Victor 1356 Caraibe Précoce Mannuel pompe + Fidjrossè Pépinière 3 4
reservoir
Forage +
Ouidah TONOUKOIN Arsène 3200 Akli. Précoce Mannuel pompe + Ahozon 10JAR x 3
reservoir
Cotonou ADIMONSSO Félix 294 Tropimech Précoce Mannuel Forage COMAES Maturité A 3
Ouidah BADJISSI Emile 800 Mongal Précoce Mannuel Forage Mahuclo Repiquage x 3
AMOUSSOU Jean Akpakpa
Cotonou 360 Mongal Médian Mannuel Forage Pépinière x 4
Claude Dodomè
GBESSEMEHLAN Akpakpa
Cotonou 405 Mongal Médian Mannuel Forage Pépinière x 4
Antoine Dodomè
Ville Nom du producteur Variété Précocité Source d’eau Stade Tracabilité Type
Akpakpa
Cotonou HOUNSSOU Vincent 405 Mongal Médian Mannuel Forage Pépinière x 4
Dodomè
Akpakpa
Cotonou KPATENON Sonia 405 Mongal Médian Mannuel Forage Pépinière x 4
Dodomè
Ouidah AGBO Inès 800 Mongal Médian Mannuel Forage Ouidah Colas Pépinière x 4
Grand Forage +
GOUFLE Théophile 1000 Mongal Précoce Tuyau Hounsoukouè Repiquage 5 5
Popo pompe
Grand Forage +
EWE Théophile 360 Mongal Précoce Tuyau Yodocondji 10JAR x 5
Popo pompe
Forage +
Ouidah YEHOUENOU René 5000 Mongal Précoce Tuyau Houdjava 10JAR x 5
pompe
Forage +
Ouidah DOUKPO Jean 5000 Mongal Précoce Tuyau Houdjava 15JAR x 5
pompe
Grand Forage +
DJIMEZO Franck 440 Mongal Précoce Tuyau EWE-CONDJI 21JAR A 5
Popo pompe
ANNEXES
Forage +
Ouidah FAGBOHOUN Janvier 7000 Mongal Précoce Tuyau Houdjava 2JAR x 5
pompe
181
Forage +
Ouidah DOUKPO Augustin 4400 Mongal Précoce Tuyau Houdjava Repiquage x 5
pompe
Forage +
Ouidah NOUWATIN Gervais 1600 Akli. Précoce Tuyau Ahozon Repiquage x 5
pompe
Forage +
Ouidah TONATO Sylvain 800 Mongal Précoce Tuyau Ouidah Centre Repiquage x 5
pompe
Forage +
Ouidah YEHOUENOU Paulin 5000 Mongal Précoce Tuyau Houdjava Repiquage x 5
pompe
Forage + Pépinère
Ouidah GUEZODJE Charles 4400 Mongal Médian Tuyau Akadjamè 6 6
pompe Novembre
Forage + Pépinière
Ouidah ADJANOHOUN Elie 6200 Mongal Médian Tuyau Houdjava 6 7
pompe Novembre
KPATENON Forage + Pépinière
Ouidah 5520 Mongal Médian Tuyau Houdjava 6 8
Barthélémy pompe Décembre
Grand Forage + Pépinière
EGBEMIKPO Michèle 400 Mongal Médian Tuyau EWE-CONDJI 6 9
Popo pompe Novembre
Forage +
Cotonou TCHIBOZO Maurice 90 Petomech Médian Tuyau Houéyiho Pépinère 6
pompe
Ville Nom du producteur Variété Précocité Source d’eau Stade * Tracabilité Type
Grand ADANGBEDOU Forage + Pépinière
2560 Mongal Tardif Tuyau Yodocondji T 7
Popo Daniel pompe Février
KPODO Régina 3584 Mongal Tardif Tuyau Yodocondji T 7
Popo pompe Janvier
KASSA Félix 5000 Mongal Tardif Tuyau Yodocondji T 7
Popo pompe Mars
Grand Rivière + Pépinère
TOSSOU Codjo 20000 Mongal Médian Tuyau Onkouihoué 8 10
Popo pompe Décembre
AGBOGANON Rivière +
Ouidah 4400 Mongal Médian Tuyau Akadjamè Pépinère 8 11
Ignace pompe
Grand Rivière + Pépinière
TOSSOU Dodji 30000 Mongal Tardif Tuyau Onkouihoué T 9
Popo pompe Janvier
Forage + Akpakpa
Cotonou GNAMBODE Francis 450 Mongal Précoce Asperseur Pépinière 10 12
pompe Dodomè
Grand Forage +
ZINGAN Symphorien 4000 Mongal Précoce Asperseur EWE-CONDJI 21JAR A 10
ANNEXES
Popo pompe
Tropimec Forage +
182
Cotonou KINHOUDE Léonard 1614 Précoce Asperseur COMAES 3SAR A 10

h pompe
ANNEXES
183
ANNEXES
Annexe 3. Protocole d’enquête pour la collecte des données qualitatives par les
conseillers auprès des 12 agriculteurs sélectionnés
FICHE COLLECTE PARCELLE – Stade pépinière et repiquage
Date visite : Nom enquêteurs : Localité :

……………………………… ……………………………… ………………………………
N° Parcelle : Nom& Prénom Producteur : Variété :
……………………………… ……………………………… ………………………………
STADE PEPINIERE
Date début
Date fin
Mise en place
Nom Quantité
Variété
Outils
Opérateur
Fertilisation
Fréquence
Nom Quantité
Produits
Outils
Opérateur
Irrigation
Fréquence
Nom Quantité
Mode d’apport
Opérateur
Protection phytosanitaire
Fréquence
Nom Quantité
Produits
Outils
Opérateur
184
ANNEXES
REPIQUAGE
Date début
Date fin
Repiquage
Nom Quantité
Variété
Outils
Opérateur
Si apport au repiquage :
Fertilisation
Nom Quantité
Produits
Outils
Opérateur
Irrigation
Nom Quantité
Mode d’apport
Opérateur
Protection phytosanitaire
Nom Quantité
Produits
Outils
Opérateur
185
FICHE COLLECTE PARCELLE- Stade culture
Date visite : Nom enquêteurs : …………………….…………………….……………………. Localité :
…………………. …………………….…………………….
N° Parcelle2 : Nom& Prénom Producteur : Variété :
……………………. …………………….…………………….……………………. …………………….…………………….
Fertilisation 1 Fertilisation 2
Date/Période
Objectif (Croissance,
fructification, carence…)
Fréquence sur la
période
Nom Quantité Nom Quantité
Produits
ANNEXES
Outils d’épandage
186
Opérateur
(producteur, MOF 3,
MOS, autre)
Traitement phyto 1 Traitement phyto 2

Date/Période
Cible
(maladie, insectes,
mauvaises herbes…)
Fréquence sur la
période
2
Faire précéder le numéro de parcelle de celui de l’exploitant (Cf liste)
3
MOF Main d’œuvre Familiale, MOS Main d’Œuvre Salariée, autre main-d’œuvre ponctuelle, entraide voisins…(à préciser)
Traitement phyto 1 Traitement phyto 2
Produits
Outils de
pulvérisation
Opérateur
(producteur, MOF, MOS,
autre)
Irrigation 1 Irrigation 2
Date/Période
Fréquence sur la
période
Nom Quantité/Durée Nom Quantité/Durée
Mode d’apport
ANNEXES
Opérateur
187
autre)
Récolte 1 Récolte 2
Date/Période
Fréquence sur la
période
Outils de collecte/
contenant
Opérateur
autre)
Tomate récolté
Annexe 4. Liste des producteurs de tomate de contre saison pour la saison 2011-2012 sur les communes de Cotonou,
Ouidah et Grand-Popo
Résumé des données - Stade pépinière
Parcelles ID 1 2 3 4 5 6 7 8 9 10 11 12
Date de semis 1/9/11 2/10/11 5/9/11 2/9/11 9/10/11 1/11/11 1/11/11 1/12/11 5/11/11 11/12/11 17/10/11 17/10/11
Durée pépinière (jours) 38 40 30 63 21 20 20 13 30 25 31 37
Surface (m²) 2 1 1.8 12 12 6 6 2 11 21.28 7 1.5
-
Densité (Graines.m ²) 714 1429 3175 1190 1190 4762 2905 3571 1299 671 1741 952
Apport fiente (nbre) 1 1 1 1 1 1 1 1 1

-1
N (kg.ha ) 1319 2639 1233 258 486 530 608 817 1808
-1
K2O (kg.ha ) 905 1811 847 177 333 364 417 560 1241
ANNEXES
-1
P2O5 (kg.ha ) 1056 2113 988 206 389 424 487 654 1448
188
Apport engrais minéral (nbre) 2
-1
N (kg.ha ) 839
-1
K2O (kg.ha ) 213
-1
P2O5 (kg.ha ) 213
Jour irrigués (nbr) 37 39 29 31 20 19 18 12 29 23 30 36

3 -1
Volume (m .ha ) 6383 11700 4833 1550 11111 950 750 27000 22594 1087 32143 1440
Traitement fongicide (nbr) 1 4 1 1 2 1

TFI 3.0 7.6 2.5 1.1 2.7 1.0
Traitement insecticide (nbr) 4 1 2 1 2 1 1

TFI 9.5 1.7 1.8 2.7 13.3 1.7 1.6
Traitement nématicide (nbr) 1 1 1

TFI 7.1 5.8 3.1
Résumé des données - Stade culture
Parcelles ID 1 2 3 4 5 6 7 8 9 10 11 12
Date de repiquage 9/10/11 11/11/11 5/10/11 4/10/11 30/10/11 21/11/11 21/11/11 14/12/11 5/12/11 5/1/12 17/11/11 23/11/11
Surface pépinière utilisée 100% 100% 100% 25% 100% 33% 100% 100% 100% 100% 16% 100%
Durée culture (jours) 80 64 80 79 92 82 36 91 83 79 57 67
Surface (m²) 306 126 921.5 546 576 168.72 1914.9 759.83 963 2200 895.2 196
Densité (Plants.m-²) 3.85 5.32 1.81 2.74 7.42 3.82 2.40 3.30 5.67 3.34 3.46 3.67
Apport fiente (nbre) 3 2 1 2 2 2 2 1
N (kg.ha-1) 1045 64 117 473 563 778 450 558.756
K2O (kg.ha-1) 717 44 81 325 387 534 309 383.46
-1
P2O5 (kg.ha ) 836 51 94 379 451 623 360 447.37
Apport engrais minéral (nbre) 5 5 1 7 1 3 2 1 1 5 2
N-Urea (kg.ha-1) 33 146 0 43 0 0 0 0 59 202 0
ANNEXES
N-NPK (kg.ha-1) 107 104 19 25 22 235 58 0 0 72 211
K2O (kg.ha-1)
189
107 104 38 25 22 235 58 0 0 72 346

-1
P2O5 (kg.ha ) 107 104 38 25 22 235 58 0 0 72 211
Jour irrigués (nbr) 45 48 45 78 69 38 23 64 66 27 50 65
Volume (m3.ha-1) 7200 6360 586 10400 5318 4515 919 10581 7048 771 3759 2570
Traitement fongicide (nbr) 7 1 2 4 2 1 3 3
TFI 7.5 2.0 0.6 2.3 1.2 0.4 3.5 3.8
Traitement insecticide (nbr) 7 2 2 10 3 3 2 1 4 4 3
TFI 14.7 3.7 4.3 25.1 8.8 12.8 1.5 0.6 10.5 5.2 2.8
Traitement nématicide (nbr) 1 1 1 1

TFI 6.1 0.7 7.9 4.9
Récolte (nbre) 1 6 1 3 4 1 1 4 4 5 1 1
Date dernière récolte 28/12/11 14/1/12 24/12/11 22/12/11 30/1/12 11/2/12 27/12/11 14/3/12 26/2/12 24/3/12 13/1/12 29/1/12
Rendements cumulés (kg.ha-1) 4902 12452 8 7875 21163 127 0 5662 1703 4498 2710 0
Détail des données - Protection phytosanitaire, stade culture
Parcelles ID 1 2 3 4 5 6 7 8 9 10 11 12
Fongicides
Cuivre
Nombre traitements 1
TFI 0.9
Mancozèbe
Nombre traitements 2 2 1 1 1
TFI 0.6 1.9 0.5 0.4 1.8
Manèbe
Nombre traitements 7 1 1
TFI 7.5 2.0 2.13
ANNEXES
Souffre
190
TFI 0.4
Thiophanate-méthyl
TFI 0.7 0.8 1.7
Insecticides
Chlorpyrifos-éthyl
Nombre traitements 2 1
TFI 4.3 2.3
Cypermethrine
TFI 5.2 7.9 4.5
Deltamethrine
TFI 4.9 1.9
Parcelles ID 1 2 3 4 5 6 7 8 9 10 11 12
Insecticides (Suite)
Dimethoate
TFI 1.5 2.2
Dimethoate / Cypermethrine
Nombre traitements 7 5 2 1 3
TFI 14.7 15.6 3.6 0.6 2.8
Lambda cyalothrine
TFI 3.7 3.6 1.1
Spinozad
ANNEXES
TFI 9.5
191
Nematicide
Carbofuran
TFI 4.9
Fenamiphos
TFI 0.7
Terbufos
TFI 6.1 7.9
Description des produits phytosanitaires utilisés
Producteurs Substance Active 1 [SA1] Substance [SA2] Dose homologuée

Nom commercial Cible Formulation
utilisateurs (nbr) (SA1) (g.kg-1 / .L-1) Active 2 (SA2) (g.kg-1 / .L-1) (kg / L.ha-1)
Cydim 23 Insectes Liquide Dimethoate 400 Cypermethrine 36 0.7
Cypercal 6 Insectes Liquide Cypermethrine 50 0.5
Lazer 5 Insectes Liquide Spinozad 480 0.15
Lambda Finer 2,5EC 4 Insectes Liquide Lambda cyalothrine 25 0.5
Dimex 3 Insectes Liquide Dimethoate 400 0.5
SUMITEX 40 EC 3 Insectes Liquide Dimethoate 400 0.5
SUNPYRIFOS 48%EC 2 Insectes Liquide Chlorpyrifos-éthyl 480 0.5
Decis 2 Insectes Liquide Deltamethrine 15 0.6
PACHA 25 EC 2 Insectes Liquide Lambda cyalothrine 15 Acetamiprid 10 1
WREKO 2.5EC 2 Insectes Liquide Lambda cyalothrine 25 0.5
ANNEXES
Dursban 4E 1 Insectes Liquide Chlorpyrifos-éthyl 480 0.5
Manate 9 Maladies Solide Manèbe 800 2
192
TOPSIN-M 70 WP 7 Maladies Solide Thiophanate-méthyl 700 1

MANCOZEB 80% WP 4 Maladies Solide Mancozèbe 800 2
Ridomil plus 3 Maladies Solide Cuivre 600 Métalaxyl 120 20
Sulfa 80WDG 3 Maladies Solide Souffre 800 10
FOKO 2 Maladies Solide Mancozèbe 800 2
CHAMPION 1 Maladies Solide Cuivre 770 10
Kocide 2000 1 Maladies Solide Cuivre 538 2.5
Conti-Zeb "5" 80%
WP 1 Maladies Solide Mancozèbe 800 1
Dithane-M 1 Maladies Solide Mancozèbe 800 2
Contitop 70% WP 1 Maladies Solide Thiophanate-méthyl 700 1
Nématue 10G 4 Nématodes Solide Terbufos 100 2.2
Carbofuran 5G 2 Nématodes Solide Carbofuran 50 10
Némacur 10G 1 Nématodes Solide Fenamiphos 100 100
Matériel d’irrigation
ID parcelles 4 5 6 7 8 9 10 11 12
-1
Débit au niveau de la pomme d'arrosage (L.h ) 2000 1800 1500 3000 2571 3600 3000
-1
Debit au niveau de la pompe (L.h ) 1800 1800
Profondeur source d'eau (m) 7 7 13 12 12 7 0 0 9

Consomation essence réelle (L.h-1) 1.0 3.0 1.0 1.0 2.0 1.0 1.0 1.5 1.0
Conduite d'eau (poids kg*) 37 78 39 31 259 222 217 107 77

Durée de vie des conduites (h) 2208 4944 4410 5570 7152 3240 3402 7820 108
Tuyau flexible (poids kg*) - 9 9 18 18 18 18 9 -
Durée de vie des tuyaux (h) - 3708 4410 4456 3576 2592 4536 6256 -
ANNEXES
Asperseurs (poids kg*) - - - - - - - - 4
Durée de vie des asperseurs (h) - - - - - - - - 2025
193
Pompe (poids kg*) 27 27 27 27 27 27 27 27 27

Durée de vie de la pompe (h) 3312 9888 2646 3240 1788 6480 5670 9384 1296
*Données utilisées pour estimer les poids:

Conduite d'eau: 5.6 kg pour 4m
Tuyau flexible =5.26 kg pour 30m
Asperseur = 0.16 kg par unité
ANNEXES
Annexe 5. Exemple des données complètes collectées pour l’itinéraire technique

pour la parcelle 6
ID parcelle: 1
Producteur: SOGLOHOUN
Localisation: Cotonou (PLM)
Conseiller: ODRADO Clémence
Système d'irrigation: Manuel Type: 1
Origine de l'eau: Depression
Position du cycle: Précoce
Surface pépinière: 2 m² La totalité utilisé pour le repiquage
Surface repiquée: 372 m² 51 planches de 6m²
Date Dose Unité Materiel
SEMIS 01/09/2011 714 grains.m-2 Houe, rateau
IRRIGATION Journalière 173 m3.ha-1.j-1 Arrosoir
FERTILISATION
Fiente 14/09/2011 43 t.ha-1 Epandage manuel
NPK 16 16 16 21/09/2011 1329 kg.ha-1 Arrosoir
Urée 21/09/2011 1363 kg.ha-1 Arrosoir
PROTECTION
Cydim 01/09/2011 0.7 TFI Pulvérisateur
REPIQUAGE 09/10/2011 3.9 plants.m-2 Houe, rateau, fil
Journalière >
IRRIGATION 80 m3.ha-1.j-1 Arrosoir
23/11/2011
FERTILISATION
NPK 16 16 16 16/10/2011 52.1 kg.ha-1 Arrosoir
Urée 16/10/2011 71.2 kg.ha-1 Arrosoir
NPK 16 16 16 03/11/2011 205.3 kg.ha-1 Capsule au pied des plantes
194
ANNEXES
Date Dose Unité Materiel

PROTECTION
Manate 30/10/2011 0.8 TFI Pulvérisateur
RECOLTE 28/12/2011 4902 kg.ha-1 Manuelle
BILAN Pepinière Culture

Fertlisation
organique
N 1319 1045 kg.ha-1
P2O5 905 717 kg.ha-1
K2O 1056 836 kg.ha-1
Fertlisation minérale
N 839 140 kg.ha-1
P2O5 213 107 kg.ha-1
K2O 213 107 kg.ha-1
Protection
Insecticides 9.5 14.7 TFI
Nematicides 0 7.5 TFI
Fongicides 0 TFI
Irrigation m3.ha-1
Rendements cumulés >28/12/2011 4902 kg.ha-1
Informations complémentaires
Fiente N = 3.06%, P2O5=2.45%,K2O=2.1%
Cydim Cypermethine = 36 g.L-1, Dimethoate = 400 g.L-1
Dose de reference = 0.7 L.ha-1
Manate Manèbe = 800 g.kg-1; Dose de reference = 2 kg.ha-1
195
Annexe 6. Résultats des analyses de sol pour les 12 parcelles échantillonnées
Laboratoire: Université d'Abomey-Calavi, Faculté des Sciences Agronomiques
Date: 03/01/2012
Unit 1 2 3 4 5 6 7 8 9 10 11 12 WM
Granulométrie
Limons % 0.75 1 1.75 1.5 1.75 1.25 5 0.75 1.25 12.75 2 0.75 3.26
Argile * % 7.75 1.75 13 15 4.5 2 9.75 7.75 0.75 65 2 3.25 13.41
Sables % 91.5 97.25 85.25 83.5 93.75 96.75 85.25 91.5 98 22.25 96 96 83.33
Densité apparente * g.cm-3 1.43 1.23 1.2 1.41 1.51 1.48 1.34 1.35 1.48 1.27 1.48 1.16 1.412
Humidités caractéristiques*
Capacité maximale de rétention % 13.4 13.4 13.6 10.6 10.8 13.0 16.1 24.7 9.6 61.6 18.6 11.4 20.1
Point de fletrissement permanent % 2.8 2.7 5.2 3.1 2.0 3.5 4.0 4.2 1.7 33.3 3.4 2.2 7.0
C/N * - 15.2 20.9 30.4 19.0 16.8 21.8 21.8 22.2 23.5 20.6 17.9 19.1 19.1
ANNEXES
Azote total % 0.064 0.050 0.039 0.059 0.042 0.048 0.064 0.064 0.036 0.148 0.070 0.045 0.064
NH4 ppm 4.5 3.8 4.4 2.3 2.2 7.3 1.5 4.6 2.9 3.1 3.1 2.2 2.9
196
NO3 ppm 13.5 9.1 8.8 6.1 13.3 8.8 9.0 10.7 15.8 13.9 10.8 10.3 11.8
Azote organique * % 0.046 0.037 0.026 0.051 0.026 0.032 0.054 0.049 0.017 0.131 0.056 0.032 0.050
Carbone organique % 0.98 1.05 1.19 1.12 0.7 1.04 1.4 1.43 0.86 3.06 1.25 0.86 1.25
Matière organique % 1.69 1.82 2.05 1.92 1.21 1.79 2.42 2.46 1.48 5.28 2.16 1.48 2.16
Capacité Echange Cationique meq.100g-1 5 5 7 7 3 6 6 5 4 16 4 4 6
Calcaire Total * % 0.18 0.16 0.34 0.46 0.34 0.18 0.57 0.34 0.34 0.57 0.23 0.11 0.35
pH * - 7 6 6 7 7 5 6 5 6 5 7 7 6
Paramètres calculés* (Brisson et al., 2008)
q0 mm 6.28 5.41 7.21 7.48 5.94 5.49 7.21 6.28 5.30 16.66 5.60 5.60 7.50
Albedo - 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.28 0.25 0.25 0.25
Inflitration
Horizon 1 - 50.0 50.0 12.7 12.7 50.0 50.0 12.7 50.0 50.0 0.9 50.0 50.0 38.4
Horizon 2 - 34.5 34.5 6.9 6.9 34.5 34.5 6.9 34.5 34.5 0.3 34.5 34.5 26.2
Horizon 3 - 18.3 18.3 3.5 3.5 18.3 18.3 3.5 18.3 18.3 0.3 18.3 18.3 13.8
* Données utilisées pour paramétrer le modèle STICS

Annexe 7. Résumé des données climatiques produites par la station météo du centre de recherche AfricaRice localisé à
Cotonou
ANNEXES
197
ANNEXES
198
ANNEXES
Annex 8. LCIA results from cradle-to-harvest per hectare of out-of-season

tomato in Benin for the 10 modalities accounting for 4 criteria, and the weighted
mean
Impact categories influenced by irrigation practices:

 The water deprivation potential (WDP expressed in m3 equivalent) calculated using water stress
index (WSI) from Pfister et al. (2011),
 The climate change potential (CCP expressed in kg CO 2 equivalent),
 The human toxicity (HTP expressed in kg 1,4-DB equivalents)
 The photochemical ozone formation potential (POFP expressed in kg NMVOCs equivalent)
 Marine ecotoxicity potentials (METP expressed in kg 1,4-DB equivalents),
199
ANNEXES
Impact categories influenced by fertilization practices:

 The terrestrial acidification potential (TAP, expressed in kg SO 2 equivalents)
 The freshwater eutrophication potential (FEP, expressed in kg P equivalent)
 The marine eutrophication potential (MEP, expressed in kg N equivalent)
200
ANNEXES
Impact categories influenced by crop protection practices:

 Terrestrial ecotoxicity potentials (TETP expressed in kg 1,4-DB equivalents)
 Freshwater ecotoxicity potentials (FETP expressed in kg 1,4-DB equivalents),
201
ANNEXES
202
ANNEXES
Portefeuille de Compétences
Madame Aurélie PERRIN
Soutenance le 20/12/2013
Doctorat : Sciences agronomiques et écologiques

Ecole doctorale : ABIES - Agriculture, Alimentation, Biologie, Environnements et Santé
Etablissement : AgroParisTech
Encadrement : Benoît GABRIELLE, Mme Armelle BAEZA-SQUIBAN, Mme Claudine
BASSET-MENS
Sujet de thèse : Quelle prise en compte de la variabilité sol, climat et pratiques dans l’analyse de
cycle de vie des produits maraîchers ? Mise en œuvre au cas de la tomate
Ouverture scientifique et culturelle

 Journées ABIES 29 et 30 Mars 2011 - 29 mars 2011 - 14 heures. Mis en place par : ABIES -
Agriculture, Alimentation, Biologie, Environnements et Santé
 Journées ABIES 28 et 29 février 2012 - 28 février 2012 - 14 heures. Mis en place par : ABIES -
Agriculture, Alimentation, Biologie, Environnements et Santé
Formations scientifiques
 Analyse environnementale du Cycle de Vie - - 21 heures
Formations à la recherche
 Communication scientifique en Anglais et Français - 28 heures
Formations Langues Étrangères et F.L.E

 Atelier d'écriture d'articles scientifiques en anglais - 28 heures
Formation aux logiciels

 EndNote - 14 heures
Cours scientifiques et techniques

 Analyse de sensibilité, propagation d'incertitudes et exploration numérique de modèles en
sciences de l'environnement - 27 heures
 ACV: approfondissement du cadre conceptuel (14 au 16 mai 2012) - 15 heures
203
ANNEXES
Conférences
 12th congress of the European Society of Agronomie - - 20 heures
 8th conference on Life Cycle Assessment in the Agri-Food Sector - - 22 heures
Connaissance du Milieu Professionnel

 Préparation à l'insertion professionnelle - Objectif premier emploi - Public : doctorants en 3è
année de thèse - Novembre 2012 - 24 heures. Mis en place par : ABIES - Agriculture, Alimentation,
Biologie, Environnements et Santé
Productions scientifiques
 Aurélie Perrin, Claudine Basset-Mens and Benoît Gabrielle. Life cycle assessment of
vegetable products: a review focusing on cropping systems diversity and the estimation of field
emissions The International Journal of Life Cycle Assessment
Autres valorisations techniques

 Farm and crop models:
 APSIM (Agricultural Production Systems sIMulator)
 APES (Agricultural Production and Externalities Simulator) STICS(Simulateur
mulTIdiscplinaire pour les Cultures Standard)
 Computer: Word, Excel, Power Point, Access and Internet & notions about SPSS, R,
SQL(PgAdmin) and Java(Eclipse)
 Agronomie, Analyse du cycle de vie et évaluation environnementale, Typologie, Flux d'azote à
l’échelle de la parcelle
Enseignements et encadrement de stages
 Cours à Montpellier Supagro -Introduction à l'analyse du cycle de vie appliquée aux systèmes
horticoles tropicaux- 2x 1.5h Ecole Chercheur ELSA- Introduction à la Typologie, usage en ACV-
1.5h
 Wilfried Yehouessi- Etude des performances agronomiques de la culture de tomate de contre
saison dans la zone côtière du Sud – Bénin (Cotonou – Pahou – Grand-Popo). Octobre 2011- 6 mois.
Université Abomey Calavi Benin, stage Master 1.
 Thais Juliane Do Prado : Prise en compte des techniques d’application et du l’état de la
couverture végétale dans l’estimation des émissions de pesticides vers l’environnement. Juin 2012.
2 mois. Université de l’Etat de Sao Polo, Bresil, stage Master 1.
204
ANNEXES
CV - Aurélie Perrin
53 avenue olande d’Aragon aurelie.perrin6@orange.fr
49100 ANGERS aurelie.perrin@cirad.fr²
06 66 94 01 76
Née le 31/12/1985 à Dijon
O B J E C T I F Mettre à contribution mes compétences pour le développement des systèmes de culture

dans un objectif de réduire la dépendance aux intrants et plus particulièrement aux fertilisants.
EXPERIENCE
Doctorante | U R H O R T S Y S - C I R A D - M o n t p e l l i er
20 Septembre 2010 – Décembre 2013
Gérer un projet de 3 ans, créer des données sur les systèmes de production de tomate au Bénin,
rédiger des publications scientifiques et des rapports techniques pour les partenaires, utiliser des
modèles de culture existants en collaboration avec les développeurs, superviser des stagiaires M1 et
M2, présenter des résultats en anglais dans des conférences internationales.
Chargé de mission | A I V B - L R ( i n t er p r o f es s i o n des vit icu lt eurs b io) -
M o n t p e l l i er
1er Avril 2010 – 27 Aout 2010
Collecter les informations sur les engrais et amendements organiques bio auprès des industries,
caractérisation ces produits et de leurs usages pour une meilleure gestion de la fertilisation.
Chargé de mission | U M R S Y S T E M - M o n t p e l l i e r S u p a g r o
1er Novembre 2009 – 05 Septembre 2010
Coordonner la communication au sein du comité scientifique pour la mise en place des programmes et
du contenue des conférences, créer et gérer la plate-forme web dédiée aux participants du congrès.
Stage de fin d’étude | P P S g r o u p - W a g e n i n g e n U n i v e r s i t y - P a y - B a s
15 Mars 2009 – 15 Septembre 2009
Evaluer de manière ex-ante les systèmes de cultures avec un focus sur les pratiques de fertilisation
azotées, calibrer et évaluer un outil de modélisation (Blé, Pays-Bas).
Stage en cours de formation | A P S R U - C S I R O - T o o w o o m b a , A u s t r a l i e
01 Juin 2008 – 15 Septembre 2008
Etudier la production fourragère en milieu sub-tropical à tendance aride, conduire des
expérimentations pour la collecte de données physiologiques et pédologiques.
DIPLOMES
M o n t p e l l i er S u p a g r o ( 3 4 )
2009 Diplôme d’ingénieur Agronome spécialité Production Végétale Durable
L E G T A E t i e n n e M u n i er ( 7 0 )
2005 BTSA Technologies végétales option Agronomie et système de culture
Lycée St e Bénigne (21)
2003 Baccalauréat scientifique option SVT
AUTRES COMPETENCES
Langues: Anglais courant parlé et écrit (niveau C1), Espagnol parlé et écrit (niveau A2)
Informatique : Pack Office, SQL (PgAdmin®), Java (Eclipse®), Html (eZpublish®), R
(statistique), Simapro® (ACV)
205
ANNEXES
PUBLICATIONS SCIENTIFIQUES
C o n f é r e n c e s i n t e rn a t i o n a l e s
Perrin A, Basset-Mens C, Yehouessi W, Huat J, Gabrielle B, and Malezieux E (2012)
Agronomical analysis of Life-Cycle Assessment (LCA) variability for different
horticultural cropping systems at regional scale. In the 12th Congress of the European
Society for Agronomy. August 20-24, Helsinki, Finland.
Perrin A, Basset-Mens C, Yehouessi W, Huat J, and Gabrielle B (2012) An LCIA-based

typology for more representative results and refined data collection of a horticultural
cropping system in the Tropics. In the 8th International Conference on LCA of the Agri-
Food sector (LCA Food). October 1-4, Saint-Malo, France.
Perrin A, Basset-Mens C, and Gabrielle B (2013) The use of a mechanistic model to better
estimate Nitrogen reactive (Nr) emission for Life-Cycle Assessment (LCA) of Tomato
production in Benin, West Africa. Under review for the 8th Australian Conference on
Life Cycle Assessment (ALCAS 2013). July 15-18, Sydney, Australia.
Burns M, Perrin A, Roux P, and Basset-Mens C (2013) Do the current life cycle inventory
(LCI) methods adequately account for the life of a pesticide in horticultural production
systems? A review. Under review for the 8th Australian Conference on Life Cycle
Assessment (ALCAS 2013). July 15-18, Sydney, Australia.
J o u rn a u x à c o m i t é d e l e c t u r e
Adam M, Belhouchette H, Corbeels M, Ewert F, Perrin A, Casellas E, Celette F, and Wery J
(2012) Protocol to support model selection and evaluation in a modular crop modelling
framework: An application for simulating crop response to nitrogen supply. Computers
and Electronics in Agriculture 86, 43-54.
Perrin A, Basset-Mens C, and Gabrielle B (2013) Life cycle assessment of horticultural

cropping systems: a review focusing on systems variability and the estimation of field
emissions. The Internationnal Journal of Life Cycle Assessment DOI 10.1007/s11367-
014-0724-3
Perrin A, Yehouessi W, Huat J, and Basset-Mens C (2013) Methods and data for producing
representative LCA results in a context of extreme data scarcity – the case of out-of-
season tomato from Benin, West Africa. Under review
CENTRES D’INTÉRÊTS
Voyages pour découvrir d’autres cultures et paysages, Sports et loisirs : équitation, escalade, randonnée,
ski de randonnée.
206
ANNEXES
207
RESUME : L’agriculture urbaine représente une opportunité de réduire la pauvreté et d’améliorer la
sécurité alimentaire des habitants des villes d’Afrique de l’Ouest. L’objectif général de cette thèse
est de produire des données d’inventaire représentatives ainsi qu’une évaluation
environnementale robuste de ces systèmes de production par la méthodologie Analyse du Cycle
de Vie (ACV). Notre cas d’étude a été la tomate des jardins urbains au Bénin. Notre état de l’art
nous a permis d’identifier que prendre en compte la diversité des systèmes de production et la
variabilité des émissions au champ étaient deux enjeux majeurs de l’ACV des produits
maraîchers. Nous avons donc développé un protocole de collecte de données basé sur la
typologie qui prend en compte la diversité des systèmes puis mis au point une méthode
d’estimation des flux d’azote au champ combinant un bilan d’azote et l’usage d’un modèle
biophysique. Nous avons ainsi créé des inventaires pour 6 types de systèmes et une moyenne
pondérée représentative de la population étudiée. L’analyse des performances agronomiques de
ces systèmes a montré une forte variabilité des rendements et une utilisation fluctuante et souvent
excessive d’engrais et de pesticides. L’exploration de la variabilité des flux d’azote à l’échelle de la
parcelle et du cycle de culture a permis l’identification de 4 facteurs majeurs: les volumes d’eau
d’irrigation, la dose d’azote, le pH du sol et la capacité au champ. A l’aide de scénarios favorables
et défavorables aux émissions pour chacun de ces 4 facteurs, nous avons montré que leurs effets
sur les résultats d’ACV étaient importants. L’intégration de ces données dans une ACV finalisée a
montré qu’un hectare de production de tomate au Bénin était plus impactant que les productions
maraichères Européennes. Les avantages du climat favorable du sud Bénin à la production de
tomate en contre saison sont annulés par les faibles performances des systèmes d’irrigation,
l’usage fréquent d’insecticides et d’importantes émissions azotées. Des données mesurées et des
connaissances nouvelles sont nécessaires sur ces systèmes pour valider et affiner nos
conclusions.
MOTS-CLEFS : Inventaire du cycle de vie  Typologie  Méthodes d’estimation des émissions 

Modèle biophysique  Analyse de sensibilité
ABSTRACT: Urban agriculture provides opportunities to reduce poverty and ensure food safety for
cities inhabitants in West Africa. The general objective of this thesis is producing representative
inventories and a robust environmental assessment for those production systems using the Life
Cycle Assessment (LCA) methodology. Our case study was the tomato production in urban
gardens in Benin. Our state of the art identified the integration of the diversity of systems and the
variability of field emissions as two major challenges for the LCA of vegetable products. We
therefore developed a typology-based protocol to collect cropping systems data that includes their
diversity and an approach combining a nitrogen budget and the use of a biophysical model to
estimate nitrogen field emissions. We created inventories for 6 cropping system types and one
weighted mean representative for the urban tomato growers in Benin. The analysis of the
agronomical performances of these systems highlighted the important yield variability and the
variable and often excessive use of pesticides and fertilizers. The investigation of nitrogen fluxes
variability at plot and crop cycle scales led to the identification of 4 major influencing factors: water
use, nitrogen input, soil pH and field capacity. Using favorable and unfavorable scenarios for
nitrogen emissions for each of these 4 factors, we demonstrated that the LCA results were
sensitive to their variations. The implementation of LCA using those contrasted data showed that
one hectare of tomato production in Benin was more impacting than European vegetable
productions. The benefits from the favorable climate for producing out-of-season tomatoes were
hampered by the low efficiency of irrigations systems, the frequent use of insecticides and large
nitrogen emissions. Measured data and new knowledge on these systems are needed to validate
and refine our conclusions.
KEYWORDS: Life-cycle inventory  Typology  Emissions estimation methods  Crop model 

Sensitivity analysis

Formulaire de Réponses Devoir 2

Transféré par

Droits d'auteur :

Formats disponibles

Formulaire de Réponses Devoir 2

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Formulaire de Réponses Devoir 2

Transféré par

Droits d'auteur :

Formats disponibles

Evaluation environnementale des systèmes agricoles

urbains en Afrique de l’Ouest : Implications de la

To cite this version:

HAL Id: tel-01168946

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

L’Institut des Sciences et Industries

présentée et soutenue publiquement par

Evaluation environnementale des systèmes agricoles urbains en Afrique de

UR Fonctionnement agro ecologique et performances des systèmes de culture horticoles

Paris Institute of Technology for Life,

présentée et soutenue publiquement par

Environmental assessment of urban agricultural systems in West Africa:

Supervisor : Benoit GABRIELLE

UR Fonctionnement agro ecologique et performances des systèmes de culture horticoles

Liste des figures ................................................................................................ 1

Liste des tableaux ............................................................................................. 3

Liste des abréviations ....................................................................................... 5

Introduction générale ....................................................................................... 7

Chapitre 3. Boucler le bilan d’azote à l’échelle du système de culture : perspectives pour

Chapitre 5. Discussion générale........................................................................................ 155

Circulation urbaine et pollution à Cotonou, Bénin

Nourrir la planète : un enjeu global et des solutions locales

Les méthodes d’évaluation de l’agriculture

 La prise en compte de l’ensemble des impacts environnementaux pertinents de type global ou

ACV et qualité des données d’inventaire

1. Objectif et périmètre objectifs de l’étude. L’étape suivante d’inventaire

Diversité des systèmes de production

Variabilité des émissions au champ

Evaluation environnementale de la tomate produite en agriculture urbaine au Bénin

Contexte et objectifs de la thèse

Plan de la thèse et démarche scientifique

Objectif et périmètre de l’étude

Prise en compte de Identifier les

Evaluer l’impact environnemental de la

Figure 2 Plan de la thèse et démarche scientifique

Chapitre 1. Analyse du cycle de vie des systèmes de culture maraichers: une

Life cycle assessment of vegetable products: a review focusing on cropping

Aurélie Perrin, Claudine Basset-Mens and Benoît Gabrielle

Aurélie Perrin () and Claudine Basset-Mens

Articled published in International Journal of Life Cycle Assessment

Keywords: Vegetable production; typology of cropping systems; emissions; Life Cycle

1.2. Materials and methods

1.2.2. Typology of cropping systems and aggregation of results

1.3. Presentation of LCA studies selected

Function and functional unit

System boundaries and cut-off criteria

1.3.2. Data and methods for LCI

Methods for field emissions

1.3.3. Input flows and yields

Data completeness and reference flow

Energy and input flows per ton of fresh matter yield

Fresh Energy Synthetic fertilizers Organic

- Chapitre 1. Analyse du cycle de vie des systèmes de culture maraichers -

Broccoli Mila i Canals et al., 2008 4 12 0 0 486 0 14 3 0

1.3.4. LCIA results

Global Warming Potential

1. Inherent variability of cropping systems and products

1.4.1. Inherent variability of cropping systems and representativeness issues

1.4.2. Discrepancies in methodological choices