Agricultural Systems 207 (2023) 103628

Contents lists available at ScienceDirect

Agricultural Systems
journal homepage: www.elsevier.com/locate/agsy

Optimal nitrogen fertilizer decisions for rice farming in a cascaded tank

system in Sri Lanka: An analysis using an integrated crop, hydro-nutrient
and economic model
H. Kanthilanka a, T. Ramilan a, b, R.J. Farquharson a, *, J. Weerahewa c
a
School of Agriculture and Food, The University of Melbourne, Australia
b
School of Agriculture and Environment, Massey University, New Zealand
c
Department of Agricultural Economics and Business Management, Faculty of Agriculture, University of Peradeniya, Sri Lanka.

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Subsidies for fertilizer will encourage

fertilizer use, but can cause health and
environmental problems.
• Fertilizer decisions of rice farmers in Sri
Lanka were analyzed via integrated
models with a social welfare
perspective.
• Social and private optimal Nitrogen
fertilizer rates varied over seasons, soil
types, weather and price levels.
• Application of socially optimal Nitrogen
rates slightly reduces crop yields but
helps in reducing water pollution.
• Social cost of fertilizer application is to
be internalized in recommending
optimal fertilizer rates.

A R T I C L E I N F O A B S T R A C T

Guest editor: Dr. Jagadish Timsina CONTEXT: The ancient irrigation systems in Sri Lanka, known as village tank cascade systems, were developed to
ensure an adequate and sustainable supply of good quality water to communities. However, there is growing
Keywords: concern about health and environmental issues related to the degradation of water quality caused by excessive
Rice nitrogen (N) levels from the overuse of chemical fertilizer. Subsidies for chemical fertilizer have encouraged
N fertilizer
fertilizer use for rice production in Sri Lanka.
Nitrate
OBJECTIVES: The objective was to evaluate the use of N fertilizers for rice production in the Thirappane cascaded
Water pollution
Cascade tank system and its impact on nitrate water quality. An optimal rate of N use was determined based on private
Integrated models (farm-level) decisions on fertilizer use. However, the private optimal fertilizer rate is not adequate for overall
social welfare due to market failures such as incomplete information and the lack of a market to account for the
negative impact of fertilizer use on tank water quality. The hypothesis is that the social optimal fertilizer rate is
lower than the private optimal rate due to this discrepancy. The study aims to identify the sources of inefficiency
in the sub-optimal use of fertilizers from a social perspective.
METHOD: We developed an integrated crop, hydro-nutrient and economic model to analyze fertilizer decisions in
the rice production process. The method involved conducting a marginal economic analysis based on simulated

* Corresponding author.
E-mail address: bob.farquharson@unimelb.edu.au (R.J. Farquharson).

https://doi.org/10.1016/j.agsy.2023.103628
Received 25 July 2022; Received in revised form 23 February 2023; Accepted 25 February 2023
Available online 14 March 2023
0308-521X/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

yield responses to N fertilizer and prices for inputs and outputs. The analysis was performed for three soil types
across the Maha (rainy) and Yala (dry) seasons and for three different weather scenarios within each season.
RESULTS AND CONCLUSIONS: When the negative impact of nitrate contamination on water quality is taken into
account, the optimal N fertilizer rate from a social perspective is always lower than the optimal rate determined
solely by private economic considerations. These optimal rates varied based on factors such as soil type, season,
weather conditions during the growing season, and fertilizer prices. At unregulated, higher, fertilizer prices, the
crop yields achieved at the social optimum were only slightly lower than those achieved under the private
economic optimum. However, under regulated, lower, fertilizer prices, achieving the social optimum would
require a larger reduction in N fertilizer use and result in a greater decrease in crop yields.
SIGNIFICANCE: A systematic analysis that takes into account the social costs can serve as a guide for creating
effective policies aimed at enhancing fertilizer decision making.

1. Introduction has been adjusted periodically in response to fluctuations in global

fertilizer prices and pressures on Government expenditure. In 2018 and
The village tank cascade system, a sustainable agricultural produc­ 2019 chemical fertilizer for rice farming was subsidized by 85% (the
tion system in the Dry Zone of Sri Lanka in ancient times (Abeywardana regulated price of urea fertilizer was 3.09 US$ per 50 kg bag, and the
et al., 2019; Leach, 1959), has been recognized as a Globally Important non-regulated price was 21.30 US$ per 50 kg bag in Yala3 and Maha4
Agricultural Heritage System (GIAHS) by the Food and Agriculture Or­ seasons, the Government subsidy was 18.21 US$ per 50 kg bag in 2018).
ganization (FAO, 2018). Historically, village tank cascade systems have In Yala in 2020, chemical fertilizer was provided free of charge. The
ensured the quantity, quality, and sustainability of water supply for importation of chemical fertilizer was abolished entirely on May 06,
rural populations (Aheeyar, 2013; Hoogesteger et al., 2023; Oyonarte 2021, and this decision was rescinded on November 30, 2021 (Ministry
et al., 2022; Ratnayake et al., 2021). They comprise low, mid and upper of Finance, 2019, 2020, 2021). Overall, the fertilizer subsidy policy has
landscape regimes designed to maximize available water resources with successfully increased agricultural productivity in Sri Lanka. However, it
varying cropping intensity, matched with the capacity of natural re­ is important to consider the potential negative impacts of the policy,
sources (soils and climate). The systems also helped to reduce the risk of such as environmental damage and unequal distribution of benefits.
soil erosion and nutrient losses. Traditionally, the systems were main­ Nitrate is a common pollutant in both surface and groundwater that
tained by strong social organizations (led by wel vidange1 under the comes from a variety of sources, including agricultural runoff, sewage,
rajakari2 system practice) (Panabokke et al., 2002; Sakthivadivel, 1997). and industrial waste, and excessive levels can lead to exceedance of
However, the village tank cascade systems have been degraded in recent permissible limits for water quality. Nitrate can readily move down the
years due to changes in land use, inappropriate crop management root zone in agricultural soils and reach the groundwater (Pretty et al.,
practices and abandonment of ancient management and rehabilitation 2000; Tilman, 1999) and is often associated with eutrophication (Cec­
practices (Ratnayake et al., 2021; Sakthivadivel et al., 1997; Sirimanna chin et al., 2021; Le et al., 2010; Tilman et al., 2002) and health issues
et al., 2022). The modernization of agricultural practices by use of (Ramalingam et al., 2022; Ward et al., 2018). Assessment of agricultural
chemical fertilizers and pesticides, and ignoring soil conservation practices regarding N fertilizer use is essential in considering water
practices, has caused a deterioration of water quality and human health quality-related issues in agricultural systems (Agouridis et al., 2005;
in farming communities (Abeysingha et al., 2021; Young et al., 2010). Gibbons et al., 2005). Many studies have demonstrated the variability in
This deterioration, reportedly caused by agricultural activities, has fertilizer rates across soil types (Link et al., 2006; Morris et al., 2018).
affected the ecology of the tanks and human health in the farming However, weather variability during the cropping season has been
community (Abeysingha et al., 2021; Dharma-Wardana et al., 2015). neglected (Bert et al., 2006; Lehmann et al., 2013; Monjardino et al.,
Thus an interest in the rehabilitation and reconstruction of the system 2013). This variability affects crop yield, crop N demand and nitrate
has been growing among government and non-government agencies losses to the environment. Growing season rainfall influences soil water
(UNDP, 2017). A recent study found that immediate actions are needed and nutrient cycling, which contributes to N supply to the crop (Hoch­
concerning proper land use planning, farmer awareness and integrated man and Waldner, 2020) and N losses to the environment (Hyytiäinen
nutrient management in the cascades to minimize external pressures et al., 2011; Miller et al., 2020). Further, Ranasinghe et al. (2023)
(Wickramasinghe et al., 2023). Kulasinghe and Dharmakeerthi (2022) emphasize that farmers must take adaptive measures against climate
recommended that land-use policies and input management need to be change to gain from the abundant natural resources in the cascades.
changed to ensure the sustainability of village tank cascade systems. However, apart from these influencing technical factors of farming,
A Government fertilizer subsidy policy has been the main instrument other compelling economic factors, including input and output prices
of agricultural support in Sri Lanka. The policy was initiated in 1962 and government policies, are also important in fertilizer decisions and
following introducion of high-yielding crop varieties (HYV) associated environmental outcomes (Kuhn et al., 2020; Sihvonen et al., 2021; Yu
with the Green Revolution. Higher fertilizer use and irrigation were et al., 2022).
needed for HYVs to achieve their yield potential. The introduction of the Studies addressing the application of economic theory for non-point
subsidy policy aimed to enhance crop productivity by promoting fer­ pollution control have used combined biophysical and economic models
tilizer application at the levels recommended by the Department of to address the complexity of relationships between agriculture and the
Agriculture at a low cost (Weerahewa et al., 2010). The fertilizer policy environment (Doole et al., 2013; Knowling et al., 2020; Kruseman et al.,
2020; Ramilan et al., 2011). Few studies have included the spatial
dimension or used social costs in decision-making (Lesschen et al., 2005;
Zhao et al., 2006). We include a factor to account for the social cost of N
1
Wel vidange is the person appointed by the farming community as the offical (SCN) in the decision framework. Systematic analysis can overcome the
in-charge of the management of the village tank cascade system
2
Rajakari system was the traditional system of land tenure in Sri Lanka until
the early 19th century where land was granted in exchange for services
3
rendered. The services expected were of two kinds: (1) public works, such as Maha (wet) season which is observed during October to March receives
road and bridge building or, in earlier days, the construction of irrigation rainfall from the North-East monsoon
4
works, and (2) special services elicited on the basis of a person's caste-related Yala (dry) season which is observed during April to September receives
occupation. rainfall from the South-West monsoon.

2
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

deficiencies of decision analysis and provide information about the

inplications of including social costs associated with those decisions
(Mallawaarachchi et al., 2002; Mallawaarachchi and Quiggin, 2001).
This paper aims to explore the sources and magnitude of inefficiency
associated with sub-optimal fertilizer use within the concept of market
failure. This causes a divergence between N fertilizer rates for an indi­
vidual deciding on actions under incomplete information (private
benefit) and that of society (social benefit). The latter fertilizer rates are
often lower as a society must bear the burden of inefficiency.
This study uses the simulation of crop yield responses and nitrate
losses from rice fields in a village tank cascade in Sri Lanka to determine
the spatially targeted optimal N fertilizer rates. This approach aims to
understand the factors (both technical: soil type, season, weather, and
economic: the price of grain, price of fertilizer, the social cost of pollu­
tion, and government policies) influencing N fertilizer decisions for rice
production and associated trade-offs between agricultural production
and the water quality in the cascade tanks.
We hypothesize that operating at a ‘socially optimal N rate’ leads to
substantial yield and profit losses compared to a ‘privately optimal N
rate’. Thus the analysis is pursued by accounting for factors influencing
the nature of trade-offs by developing integrated crop, hydro-nutrient
and economic models for the Thirappane tank cascade of Sri Lanka.
We follow Gourevitch et al. (2018), who defined the socially optimum N
rate as the rate at which net benefits of N to society are maximized. The
private optimum N rate was defined as the rate of N that maximizes the
private net benefits only to the farmer.

2. Materials and methods

2.1. The study area

The study is conducted in the Thirappane tank cascade, located in the

North Central province of Sri Lanka. The cascade comprises a series of
six minor tanks each with less than 50 ha command area and a total
water holding capacity of 1,988,373 m3 (Fig. 1). The distance between
Fig. 1. Map of the Thirappane cascade illustrating the structure and the dis­
the uppermost (8.15 N, 80.52 E) and the lowermost tank (8.21 N, 80.51 tribution of the soil types.
E) is eight km, and the cascade is two km wide. The total cultivation area Note: The water flows from the upper stream tank (Vendarankulama) to the
is 207 ha. The command area of the tanks is relatively flat (maximum lower stream tank (Thirappne).
slope of 3%) (Jayatilaka, 2001). These minor tanks are mostly seasonal
in nature, and maintenance is conducted with the support of farmers. Nutrient use in crop production within the cascade has increased the
The tanks and their characteristics are listed in the Supplementary concentration of nutrients in tank water, mainly nitrate and phosphate.
Table S1. In accordance with government regulation,5 only rice (Oryza The downstream tank (Thirappane) has an elevated concentration of
sativa) can be cultivated in the lowlands of Sri Lanka when water is nitrate compared to other tanks in the cascade, illustrating the spatial
available in sufficient quantities. accumulation of nitrate down the tank cascade (Wijesundara et al.,
Three major soil types are present in the cascade: namely, Low 2012). This tank shows algal blooms in some months of the year (Zoysa
Humic Gley (LHG) poorly drained soil, Reddish Brown Earth (RBE) and Weerasinghe, 2016). A study by Abeysingha et al. (2021) showed
imperfectly drained soil, and RBE well-drained soil (Fig. 1). RBE is the that the nutrients in water within the cascade had reached the eutrophic
predominant soil type covering 75% of the total land area; however, level with possible environmental and health impacts.
LHG soil is predominant in rice cultivation areas. Tank water is used for
the irrigation of rice, as well as for domestic activities, such as bathing,
washing, and drinking. Occasionally, tank water is also used for live­ 2.2. Modelling approach
stock, such as cattle, buffaloes, and goats. Tank water is also sometimes
used for fish farming, which is an important source of income for many An integrated modelling framework was developed and applied
farmers. Tank water is used for recreational activities, such as swimming (Fig. 2) to derive optimum N fertilizer rates, which maximize the social
and boating. Rainfall is received in a bi-modal pattern, where 70% of benefits and private profits of rice farming constrained by nitrate levels
annual rainfall (on average 1490 mm) is received in the Maha season, in tank water. Socially optimal N rates were calculated using crop pro­
and nutrient levels in the tanks follow the rainfall pattern (Wijesundara duction and nitrate transport functions, including the social cost of N
et al., 2013). (SCN) and prices of rice and fertilizer. Privately optimal N rates were
calculated using crop production functions and prices for rice and
fertilizer.
5 A systematic analysis was undertaken by developing and integrating
The Agrarian Development Act (2000) restricts cultivation of paddy land,
from which the maximum production can be obtained, to only rice, with powers three models: a crop simulation model for yield responses, nitrate
to prescribe paddy lands vested in the Land Commissioner-General (LCG) under leaching, and runoff from the rice fields as a function of N fertilizer
the Paddy Lands Act (1958), which was later replaced by the Agricultural Lands applied; a hydro-nutrient model to predict the nitrate transport into
Law (1973) and also extended to uplands. Several deviations were allowed with tanks from the rice cultivation areas with flow paths of nitrate leaching,
written permission of the LCG. runoff and lateral flow; and an economic model to estimate socially and

3
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

Fig. 2. Schematic description of the integrated modelling framework.

privately optimal N levels by maximizing social benefits and private Census and Statistics of Sri Lanka for the Yala and Maha seasons from
profits respectively after accounting for soil heterogeneity, weather 1997 to 2018. Average fertilizer usage data was obtained from the Na­
variability over and within seasons, effects of fertilizer subsidies and tional Fertilizer Secretariat of Sri Lanka.
other policies, and the effects on society of reduced tank water quality.
The framework can be used to inform policy decisions on N fertilizer 2.2.2. Crop simulation model
subsidies and other policies related to N fertilizer use in rice farming. Rice yield and soil N dynamics were simulated using the Agricultural
Production Systems sIMulator (APSIM) (www.apsim.info) version 7.10.
2.2.1. Data APSIM was used to simulate grain yields, nitrate leaching, and runoff of
Socio-demographic, economic and physical data were sourced from a a typical rice cropping system under a 32-year series of historical climate
farm household survey conducted in the study area in 2019 (Kant­ data to account for yield variability induced by inter-annual climate
hilanka, 2022) using a structured questionnaire which included ques­ variations.
tions on farm household characteristics, agricultural production
activities, agricultural lands, farm assets, nutrient management (N, P 2.2.2.1. Model calibation. The model was calibrated using management
and K), weed control, irrigation, labour use, crop harvest, other income data. The calibration process involved adjusting the parameters of the
sources of farmers, and farmer perceptions of water quality and func­ model to match the observed crop growth and yield at the study loca­
tioning within the Thirappane tank cascade. A historical series of tion. The management data was collected through a household survey
climate data were available from the closest meteorological station for the main crop production seasons in 2018 and used to calibrate the
located at Mahaillupallama, representing the current climate of Thir­ crop simulation model. The most common rice varieties cultivated in
appane. Secondary data, including sowing dates, area of cultivation and Maha were “BG 359” and in Yala “BG 300”. Both are HYVs. The BG 359
fertilizer quantity provided under subsidy, were collected from the variety has a 3.5-month growing period and BG 300 has a three-month
Department of Agrarian Service Centre records at Thirappane. The re­ growing period. The phenological parameters for commonly grown rice
ported district average rice yields were obtained from the Department of varieties were obtained from Amarasingha et al. (2015), as shown in

4
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

Supplementary Table S2. Soil layer characteristics of the study area were had already been calibrated for the study area. The cascade water bal­
also obtained from Mapa and Pathmarajah (1995) and Mapa (2020). ance model was formulated based on a simple structure, incorporating
The soil characteristics are summarized in Supplementary Tables S3-S5. the dynamic hydrologic processes associated with a set of tanks in the
A historical series of climate data for yield simulations was available Thirappane tank cascade. The model considered all inflows and outflows
from the meteorological station located at Mahaillupallama. Climatic of water in a tank in the cascade. Inflows and outflows varied with the
datasets cover 1976–2019 and include daily values of rainfall, temper­ position of the tank in the node-link representation. After defining the
atures, relative humidity, wind speed and sunshine duration (used to inflow and outflow for each tank, the water balance was estimated daily
estimate solar radiation). Daily incoming radiation (MJm− 2d− 1) was (according to equations given in Supplementary Eqs. S1 to S11). Each
calculated using sunshine hours, latitude and longitude and angstrom tank's irrigation water supply potential was estimated based on crop
coefficient (a = 0.29, and b = 0.39) (Samuel, 1991). The model was water requirements after accounting for conveyance losses. Based on
configured for soil being puddled and levelled. Direct seeding was this, the maximum possible areas of cultivation for the Yala and Maha
implemented at 90 plants/m2 density. A seven-day sowing window was seasons were determined and used in the modelling. For agricultural
used in the simulations. Depending on the rainfall distribution, crop non-point source pollution, the process can be divided into the “source”
water requirements were supplemented through irrigation during the link of field pollution generation and the “sink” link of water trans­
growing period. The irrigation water used was considered later in the portation through drainage channels (Wan et al., 2021; Wriedt and
hydro-nutrient modelling. N fertilizer applications were simulated ac­ Rode, 2006). We also used the source-sink linkage in modelling nitrate
cording to farmer practice in the cascade. transport.

2.2.2.2. Model validation. The model was validated at temporal and 2.2.3.2. Nutrient transport model. The nutrient transport model con­
regional scales. The reported district average rice yields were obtained sisted of export coefficients for nitrates from the field on the assumption
from the Department of Census and Statistics of Sri Lanka for the Yala that land-use changes, including fertilizer decisions, are a major deter­
and Maha seasons from 1997 to 2018 and were compared with the minant of nitrate occurrence in the catchment (Johnes et al., 2007;
APSIM simulated yield in LHG soil for average N fertilizer use in Worrall and Burt, 1999; Zhang et al., 2019). Some components of this
respective years. Average N fertilizer usage data was obtained from the model have been developed with inputs from Neitsch et al. (2011) in
National Fertilizer Secretariat of Sri Lanka. The average N application SWAT as adapted by Lam et al. (2010). The nitrate transported in each
rates ranged from 60 to 150 N kg/ha from 1997 to 2018 in the Anu­ path was estimated via a set of equations adapted from the SWAT
radhapura district. (Neitsch et al., 2011) (see Supplementary eqs. S6 to S10).
As the regional scale model validation, the simulated rice yields for The nutrient transport model was coupled with the water balance
farmers' N fertilizer use at the cascade for 2018 were compared with model to evaluate nitrate movements within the cascade for different
observed farm yields collected from the household survey conducted at rates of N applications in the cultivation area in the source tank. The
the cascade in 2019. The model performance was evaluated using the amount of nitrate transported into the sink tank via each flow path,
root mean square error (RMSE) (Pham, 2019). including leaching, runoff and lateral flow, was estimated as the product
of the volume of water and the average concentration of nitrate in each
2.2.2.3. Model simulation. The validated model was used to estimate soil layer (Lam et al., 2010). The export coefficients used were those in
long-term rice-yield responses, nitrate leaching and nitrate runoff for the water balance model (see Supplementary Table S6).
different levels of N fertilizer applications using weather data from 1976
to 2019. Simulations were conducted for all three soil types in the Maha 2.2.3.3. Model validation and simulations. Daily tank water height data
season. Rice cultivation was restricted to LHG poorly drained soil and collected from April 2013 to 2015 by Oka et al. (2019) were used to
RBE imperfectly drained soil in the Yala season. There was always a calibrate and validate the cascade water balance model. The validated
restriction on the area of rice cultivated in Yala due to limited irrigation model was used to simulate tank water balance and nitrate transport
water availability in the tanks (Warnakulasooriya and Shantha, 2021). from 1997 to 2019 for five N rates (for Maha season 0, 50, 100, 150 and
Initial N application rates varied from 0 to 300 kg per ha per season 200 kg N/ha and for Yala season 0, 30, 60, 90 and 120 kg N/ha). All
(100% increase of current mean application at the cascade in 2018) per simulation analyses were modelled on a daily-time-step basis, and a
ha per season to identify the general trend of yield and nitrate-discharge seasonal total of variables was obtained where needed. Finally, nitrate
responses. transport functions were estimated for each soil type, season, and tank
combination.
2.2.3. Hydro-nutrient modelling
The hydrological model was a node-link network in which nodes 2.2.4. Economic modelling: Production function approach with profit
represent physical units (tanks) and links represent the connection be­ maximization
tween these units. The hydro-nutrient model comprised two sub-models: Rice yields and nitrate discharges from the soil responding to fer­
a water balance model and a nutrient transport model. Total daily ni­ tilizer applications were developed from crop simulations and are pre­
trate discharged (including leaching, runoff and lateral flow) from the sented in the results section. Conceptually, as the N fertilizer rate
cultivation area in the source tank and transported into the sink tank was increases, crop yields increase at a decreasing rate and nitrate discharges
simulated from 1997 to 2019 and summed annually for Maha and Yala from the soil increase at an increasing rate.
seasons across all soil types. The water balance model simulated the Diminishing return responses are common in biology, ecology and
water flow between tanks, and the nutrient transport model simulated animal production (Thornley and France, 2007). A modified
the transport of nitrate from the source tank to the sink tank. The model Mitscherlich-Baule (Brorsen and Richter, 2012) yield response function
was calibrated using observed data from the tanks and the surrounding was fitted for the relationship between N rate (x) and grain yield (Y) for
area. The model was used to simulate the hydrological and nutrient each scenario,
transport processes in the tank cascade system. ( )
Y = a + b 1 − e− kx . (1)
2.2.3.1. Water Balance model. Many previous studies have developed In eq. (1), Y is rice yield (t/ha), a is the yield at zero fertilizer
and used a particular water balance model for the Thirappane tank application (t/ha), b is the parameter above a where yield increases to
cascade (Itakura, 1995; Itakura and Abernethy, 1993; Jayatilaka et al., the asymptote, k is a coefficient of gain, and x is the rate of N application
2003; Oka et al., 2019; Shinogi, 1998). We adopted that model, which (kg/ha). The level of asymptotic yield is given by a + b.

5
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

Gourevitch et al. (2018) determined the socially optimum rates of N fertilizer prices for various subsidy scenarios and an assumed social cost
fertilizer application by evaluating the private and social costs and associated with reduced tank water quality.
benefits of N and identified the rate to maximize the net benefits of N to These scenarios address the market failure issues of distorted input
society. Keeler et al. (2016) investigated the SCN by quantifying the prices and the inclusion of community property rights associated with
damage costs of N to air, water and climate. acceptable water quality for human consumption. The framework does
A framework, including an estimate of the SCN for water quality, was not address other effects of market failure associated with using N fer­
adopted to develop a socially optimum rate of N. The concept of a so­ tilizer for rice production in Sri Lanka (e.g., nitrous oxide emissions
cially optimum fertilizer rate depends on the availability of complete contributing to global warming potential).
information about underlying natural processes, the effects of govern­ In Fig. 3, Ns is the social optimum N rate and NP1, NP2, NP3, and NP4
ment policies, and whether there are distorted or missing markets. Un­ are defined as sub-optimal N rates with different N prices. Ns was
certainties about underlying natural processes include crop yields and determined at the marginal social cost of nitrogen (M_SCN), which
nitrate losses arising from differing N fertilizer rates (depending on included the damage cost of nitrate-N into the water (SCN) and the
seasonal factors) and knowledge about nitrate transport from fields to world market price of N fertilizer (PN_W). The world market price of N is
tanks for differing N fertilizer rates. Policy distortions include subsidized shown at NP1. At NP2, the regulated market price of N (PN_RM) in Sri
fertilizer inputs, the fact that fertilizer is a regulated import, and land use Lanka is considered. The NP3 sub-optimal N rate is a ‘farmer's price
restrictions for where and when rice is grown. No clearly defined com­ (PN_av), which is included because paddy farmers can purchase N fer­
munity property rights are associated with acceptable water quality for tilizer from the market or from other farmers at a rate lower than the
human consumption. regulated market price and higher than the subsidized price. In practice,
The decision framework is presented (Fig. 3) to conceptualize N this happens. The NP4 sub-optimal N rate is in the absence of a quantity
fertilizer use from social and private viewpoints. This accounts for the restriction with a subsidized price (PN_S). This is a hypothetical scenario,
missing information, a lack of property rights, and distortions associated as only a quota of fertilizer is provided at a subsidized rate (see Table 1).
with the N fertilizer policy in Sri Lanka. The framework explicitly ad­ Damage cost estimates for nitrate-N in water are not available spe­
dresses input price uncertainties arising from the fertilizer subsidy pol­ cifically for Sri Lanka. Hence, an estimate from China (Yin et al., 2019)
icy and the social cost of adverse tank water quality. The framework was adapted for this study for illustrative purposes. The analysis esti­
provides a decision-making tool for farmers and policymakers to eval­ mated the nitrate transported; then nitrate was converted into nitrate-N
uate the optimal N fertilizer rate for a given crop and subsidy scenario. by multiplying by 0.226 (N accounts for 22.6% of the nitrate ion). The
The framework also provides a basis for evaluating the effects of within- damage cost of nitrate-N used in the study was US$ 1.32 per kg of
season variability in crop yields and nitrate transport on optimal fertil­ nitrate-N in the water. This includes health and environmental costs.
izer decisions. The framework can be used to evaluate the effects of The damage cost was converted into Sri Lankan rupees using the ex­
alternative fertilizer subsidy policies on optimal N fertilizer rates and the change rate in 2018 (1 US$ = LKR.161.81). The social and private
associated economic and environmental impacts. economic N rates were sensitive to input and output prices. We fixed the
Arising from the diminishing returns, crop yield responses encap­ rice grain price at 0.25 US$/kg (in 2018) in the analysis.
sulated in the Mitscherlich-Baule functional form and including crop The high, low and medium profit potential years were identified via
input and output prices, the analysis proceeded by developing a mar­ cumulative probabilities of yield (Farquharson, 2006) and net return
ginal value product (MVP) schedule, which typically declines as the N (Kandulu et al., 2018; Kandulu et al., 2012). We used three simulated
fertilizer input increases (Dillon and Anderson, 1990; Anderson et al., yield-outcome categories for growing season weather variability, low
1977)). Input prices, expressed as marginal costs (MC), are included to (10th percentile of yield), medium (50th percentile of yield) and high
develop optimal N fertilizer rates where the marginal benefit (MVP) (90th percentile of yield).
equals the MFC. In this analysis, the input prices include alternative

Fig. 3. Decision framework developed for optimal fertilizer rates for alternative N fertilizer prices includes subsidized price (PN_S), farmers' price(PN_AV), regulated
market price (PN_RM), and world market price (PN_W). The optimal (social) N rate is decided at the marginal social cost of N (M-SCN), including the damage cost and
world market price of N.

6
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

Table 1 discharge via leaching and runoff in all three soil types (for model
Price details associated with the decision framework cited in Fig. 3. validation, see Supplementary Figs. S1, S2, S3, and Tables S7 and S8).
Prices Details Price of N The present study identified the impact of the rate of N fertilizer on rice
yield and nitrate discharge via leaching, runoff, and lateral flow while
LKR/kg US$/kg
accounting for variability in weather and soils.
World Market World market equivalent price in
152 0.94
price (PN_W) 2018. Not paid by farmers.
Prior to 2021, a price subsidy was
3.1.1. Variability of rice yield over the season, soil and weather
Subsidized price Simulated rice yields for increased N rates showed general increases
provided up to the quota level of 86 kg 22 0.14
(PN_S)
of N/ha. at a decreasing rate up to a plateau in all soil types (Fig. 4). The rate of
The market price paid by farmers for N yield increase for N applications varied between soil types. The highest
Regulated Market over and above 86 kg of N/ha.
109 0.67 median yield response for applied N was observed in LHG poorly
price (PN_RM) Government funded the gap between
PN_W and PN_RM drained soil, while the lowest was in RBE well-drained soil. As expected,
The weighted average of subsidized 78 in 0.48 in the yield simulated at each level of N was stochastic due to climate
Farmers' price and regulated market prices. Specific Maha Maha variation over the years. In general, there was a lower variation in
(PN_av) for seasons according to average N 76 in 0.47 in simulated rice yield in LHG than in other soils over the years. The poor
usagea Yala Yala
Damage cost of
drainage condition (lower saturated hydraulic conductivity) of LHG soil
Based on Yin et al. (2019) 213.59 1.32 increases its ability to retain water and N, making more of them avail­
nitrate-N (SCN)
able for rice plant growth. The simulated yields over soil types in both
LKR is Sri Lankan Rupees, and the exchange rate in 2018 was 161.81 LKR/US$.
a Yala and Maha are shown in Supplementary Fig. S4.
The average N usage per ha per seasons in 2018 in the Thirappane Cascade,
data gathered from the survey.

3.2. Nitrate load simulations

3. Results and discussions
3.2.1. Simulations and validations of water flow
The results presented here focus on the LHG poorly drained soil (over
The results of this study suggest that the model could accurately
75% are distributed in the cultivation lands) and the main season used
simulate the water volume of the tanks. The observed and simulated
for rice production (Maha). The complete set of results is in the Sup­
water volume of the tanks showed good agreement (Fig. 5). The RMSE of
plementary materials.
each tank was Vendarankulama 23 (m3 ‘000), Bulankulama 17 (m3
‘000), Meegassagama 82 (m3 ‘000), and Allisthana 93(m3 ‘000). A close
3.1. Crop simulations match between the observed and predicted water volumes indicated that
the model could capture the hydrological dynamics of the tanks. Similar
The parametrized and validated APSIM Oryza model was used to model validation was observed in studies by Jayatilaka et al. (2003),
simulate long-term (1976 to 2019) seasonal rice yield and nitrate Fujihara et al. (2011) and Tan et al. (2018).

Fig. 4. Distribution of APSIM simulated rice yield from 1976 to 2019 for varying fertilizer application levels in the Maha season for all soil types at the Thirappane
tank cascade.

7
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

Fig. 5. Model validation with observed and simulated daily water volume in the tanks at the Thirappane tank cascade for the calibration period.

The average inflow and outflow contributions within the season were cultivation fields. Nitrate was carried from the return flows, including
variable (Supplementary Table S9). Runoff from the catchment was the seepage (leached nitrate) and irrigation water issue (nitrate runoff).
main inflow component in each tank in both seasons, from 60 to 75% in Around 2 to 5% of inflow in the Maha season consisted of water issue
Maha and 50 to 88% in Yala. Seepage was the main outflow component return, while in the Yala season, it was 4 to 10%. The seepage return
in any tank, accounting for 30 to 60% of outflow in Maha and 50 to 85% contribution was 2 to 8% and 20 to 32% in Maha and Yala, respectively
in Yala. Water issued for irrigating rice crops was the second-highest (see Supplementary Table S9).
outflow component from a tank at the cascade. As a sink of nitrate,
the return flows from the source tank are very important since the N 3.2.2. Nitrate discharge at fields and transport into tanks
fertilizer applications directly influence the flow path in the rice Simulated Nitrate discharges, including leaching, runoff and lateral

Fig. 6. Mean nitrate discharge from the cultivation area in tank (A) and mean nitrate transport into the sink tank (B) for different rates of N application in LHG poorly
drained soil for the Maha season at the Thirappane cascade. (error bar represents the standard deviation of nitrate discharge and transport at respective N level.

8
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

flows, were positively related to N applications for all soil types (see outcomes are shown in Table 3.
Fig. 6A). Mean nitrate discharge at the cultivation fields in the source
tank within the season ranged from 22 to 50 kg/ha in LHG soil in the 3.3.1. The marginal social cost of nitrogen
Maha season. The higher discharge of nitrate at higher N rates can be In estimating the socially optimal N rate, the SCN includes the cost of
justified by the findings of Craswell (2021) that current crops use no fertilizer (world market price) and the damage cost of nitrate-N in tank
more than 50% of N fertilizer applied. The discharge of nitrate from water. The decision framework of Fig. 3 must be amended because the
cultivation fields is a soil-specific characteristic which did not vary much marginal SCN (M_SCN) is not constant for unit increases in N fertilizer
between tanks in the cascade (see Supplementary Figs. S5 & S6 for more since the nitrate transport functions increase exponentially (Fig. 6 and
results). Table 3). The revised framework is in Fig. 7.
Other studies have frequently reported the coupling of nitrate and
water movements (Chowdary et al., 2004; Mack et al., 2005; Sagehashi 3.3.2. Social and private economic N rates
et al., 2016; Zhang et al., 2020). We have followed the same procedure, The socially and privately economic optimum N rates (for the
with the export coefficients used in nitrate transport modelling being 0.1 framework in Fig. 7 and Table 1) were derived by combining the results
to 0.3 for the study area, which was within the range of reported values from yield response functions and nitrate transport functions. These
in other studies (Dash et al., 2015; Parashar et al., 1998). rates are shown in Table 4 and Supplementary Table S12. As expected,
As expected, nitrate transport was positively related to the rate of N the social optimum N rates were below the optimal private rates in all
applications (Fig. 6B). Mean nitrate transport into the immediate sink scenarios. The fertilizer price is a fundamental determinant of the dif­
tank within the season ranged from 3 to 5 kg/ha in LHG soil. However, ference between social and private economic N rates. As the fertilizer
as indicated in Supplementary Fig. S6, nitrate transport differences were price reduces under subsidy scenarios, the differences in N rates become
observed between tanks. Ten percent of nitrate discharges from culti­ larger.
vation areas in the source tanks of Allisthana tank and Meegassagama The reductions in N fertilizer rates, rice yields and nitrate transport
tank were transported into the sink tanks (into Thirappane tank and for the social optimum N rates compared to the optimum private rates
Allisthana tank, respectively). Twenty percent and 30% of nitrate are given in Supplementary Tables S13, S14 and S15. Results for the
discharge in cultivation areas of the Bulankulama tank and Vendar­ LHG soil type and Maha season are in Table 5.
ankulama tank were transported into the sink tank (Meegassagama The patterns of changes for the socially optimal N rates compared to
tank), respectively. The Vendarankulama and Bulankulama tanks the private (subsidized and distorted) N price scenarios in Table 5 show
showed comparably higher nitrate transported into the lower tank than that differences between social and private rates are small when the
other tanks. Both those tanks are sinks to the Meegassagama tank. The comparison uses higher (world market) prices for the private fertilizer
nitrate transport rate was higher in the Bulankulama and Vendar­ decisions. At highly subsidized (lower) fertilizer prices, the differences
ankulama tank than in the other tanks. This could be due to the higher are much larger.
slope of the cultivating fields in these tanks, compared to other tanks,
which increases the nitrate transport rate. 3.3.3. Sensitivity analysis
Using the simulated nitrate transport data, the nitrate transport Results presented in Table 5 were based on 2018 prices. Prices for N
functions have been estimated and the findings are presented in Table 2 fertilizer and rice have increased since 2018. A sensitivity analysis was
for LHG soil in the Maha season (see Supplementary Table S10 for co­ conducted for the socially optimal N rates with mid-2022 prices for rice,
efficients for other seasons). The estimated coefficients reveal that the N fertilizer and damage costs. The sensitivity analysis scenarios and
application rate significantly positively impacted the amount of nitrate changed prices are in Table 6. The grain price in 2022 increased three
transported into the sink tank from the cultivation area in the source times over the 2018 price, and the fertilizer price increased around
tank in LHG soil in Maha (see Table 2). Our findings are consistent with eleven times over the 2018 price. The nominal damage cost is more than
other studies, which demonstrated an exponential increase in N twice as high as in 2018. The results of the sensitivity analyses are in
discharge for small increases in N input based on field measurements Table 7 for LHG soil in the Maha season, and the complete set of results is
(Delin and Stenberg, 2014; Jiang et al., 2017; Tan et al., 2018). In in Supplementary Table S16.
conclusion, nitrate transport was positively related to the rate of N ap­ From the price changes in Table 6, we expect that in scenario 1 (as
plications and was also affected by the slope of the cultivating fields. the grain price is increased), the socially optimal N rates will increase
from the base. In scenario 2, owing to the much higher relative N fer­
tilizer price, we expect the optimal social N rates to decline from the
3.3. Optimal rates of N
base. These patterns are generally observed in Table 7 results.
From the decision framework in Fig. 7, we expect that the inclusion
The estimated coefficients of the fitted production function for the
of a damage cost for nitrate discharged into tank water will lead to a
season, soil types and yield outcomes revealed that the soil types have
lower optimal social N rate than the private decisions. These trends are
different yield responses to N fertilizer, indicating a difference in
apparent from the results in Table 4 for LHG soil in the Maha season.
inherent soil productivity.
Similar patterns for N fertilizer reductions, yield losses and nitrate
The detailed results are presented in Supplementary Table S11.
transport reduction are shown in Table 5. An important observation
Statistically significant relationships between the rate of N application
and rice yield for the LHG soil in the Maha season for all three yield
Table 3
Table 2 Estimated coefficients of the production function in LHG soil in the Maha season.
Estimated coefficients of the nitrate transport functions in LHG soil in the Maha Yield outcome category Y ¼ a þ b(1 ¡ e− kx
)
season.
A b k
Tank name NT = A * ek*N rate

Low 2655.1* 3099.7* − 0.011*

A k Medium 2954.1* 3238.4* − 0.014*
Vendarankulama 7.01*** 0.0035*** High 3959.4* 2701.4* − 0.015*
Bulankulama 4.69*** 0.0035*** Three simulated yield-outcome categories for growing season weather vari­
Meegassgama 2.31*** 0.0034***
ability - low (10th percentile of yield), medium (50th percentile of yield) and high
Allisthana 2.43*** 0.0038***
(90th percentile of yield).
*** *
Statistically significant at 0% of probability level. Statistically significant at 5% of probability level.

9
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

Fig. 7. Decision framework, including the non-linear marginal SCN and alternative N fertilizer prices, includes subsidized price (PN_S), farmer's price(PN_AV),
regulated market price (PN_RM), and world market price(PN_W).

Table 4 Table 5
Estimated optimal (social) and sub-optimal N rates in LHG poorly drained soil in Impacts on N fertilizer rates, crop yields and nitrate transport by reaching op­
the Maha season. timum N rates from sub-optimal N rates, LHG soil and Maha season under
Yield Optimum Sub-optimal N rates (kg/ha)
different N prices.
outcome (social) N Soil type Yield NP1 NP2 NP3 NP4
Category ratea outcome
(kg/ha) At world At At At
category
market regulated farmers' subsidized
NS NP1 NP2 NP3 NP4 price market price price
price
At world At At At
market regulated farmers' subsidized N fertilizer Low 3.00 33.00 63.00 178.00
price market price price reduction Medium 2.00 26.00 50.00 140.00
price (kg N/ha) High 2.00 24.00 46.00 131.00
Low 0.21 1.96 3.18 5.34
Low 197 200 230 260 375 Yield losses
Medium 0.13 1.42 2.32 3.90
Medium 175 177 201 225 315 (%)
High 0.12 1.21 1.99 3.38
High 156 158 180 202 287
Nitrate Low 0.08 0.90 1.81 6.38
a
Nitrate transport functions were tank-soil specific. First, tank-specific social transport Medium 0.05 0.65 1.30 4.31
optimal N rates were estimated. Then soil and season-specific social optimal N reduction
High 0.04 0.55 1.11 3.70
rates were obtained as weighted averages over tank-soil combinations. (kg/ha)

from these results is the impact of fertilizer prices as influenced by public

subsidization. Weerahewa and Dayananda (2023) also emphasize that a
policy. As fertilizer prices are more heavily subsidized, the difference
secure market for chemical fertilizers is needed to ensure financially and
between social and private optimal N rates is accentuated, leading to
environmentally sustainable cropping systems. Additionally, the pol­
higher nitrate transport into tanks. The sensitivity analysis results
icies related to fertilizer use and policy reforms in rice cultivation need
reinforce the importance of fertilizer prices on crop production decisions
to consider the structure and function of the cascade to ensure resource
and associated environmental impacts. The environmental impacts of
conservation and sustainable production (Sirimanna et al., 2022).
agricultural practices are the costs that are typically unmeasured and
often do not influence farmers or societal choices about production
4. Conclusions
methods when considering only the private benefits of agricultural
production (Tilman et al., 2002).
We developed and applied an integrated model to investigate the
Reducing N fertilizer use is indicated to mitigate the risk of envi­
implications of N fertilizer management on rice crop yields, optimum N
ronmental pollution and human health threats in Sri Lanka. This could
fertilizer rates, and nitrate discharges and transportation into nearby
be accomplished through stringent policies of fertilizer regulation, rec­
water bodies via leaching, runoff and lateral flow. The yield potential
ommendations for improved N management practices for farmers,
varied between soil types, and LHG soil was always the most productive.
training and education of farmers on nutrient management, and public
Soil, seasonal and weather variability cause heterogeneity in indicated
awareness of environmental protection. The N fertiizer-induced nega­
optimal social and private N fertilizer rates. Operating at socially
tive environmental impact can be mitigated if the socially optimal N
optimal N rates was predicted to reduce water quality depletion
rates are adopted, but this depends on the policy of fertilizer

10
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

Table 6 Acknowledgement
Scenarios to identify the impacts on optimal N rate by changed grain price,
fertilizer price and exchange rate from 2018 to 2022. Jeevika Weerahewa acknowledges the technical support provided by
Scenario Grain Damage Np1 Damage Np1 AHEAD/DOR/52 project administered by the University of Peradeniya,
price cost (2018) cost (2022) Sri Lanka, to conduct this study.
(LKR/ (2018)a (LKR/ (2022)b (LKR/
kg) (LKR/kg) kg) (LKR/kg) kg)
Appendix A. Supplementary data
Base (current result) 40 214 152
Increase
Scenario Supplementary data to this article can be found online at https://doi.
grain price 120 214 152
1
only org/10.1016/j.agsy.2023.103628.
Increase
grain and References
Scenario fertilizer
120 476 1739
2 prices, and Abeysingha, N., Dissanayake, S., Sumanaweera, S., De Silva, S., 2021. Cascade tank
damage water quality management: a case study in Thirappane tank cascade system, Sri
cost Lanka. J. Environ. Earth Sci. 3 (1).
a Abeywardana, N., Schütt, B., Wagalawatta, T., Bebermeier, W., 2019. Indigenous
The damage cost was converted into Sri Lankan rupees using the exchange agricultural systems in the dry zone of Sri Lanka: management transformation
rate in 2018. (1 US$ = LKR.161.81) assessment and sustainability. Sustainability 11 (3), 910.
b
The damage cost was converted into Sri Lankan rupees using the exchange Agouridis, C.T., Workman, S.R., Warner, R.C., Jennings, G.D., 2005. Livestock grazing
rate in 2022. (1 US$ = LKR. 360.76). management impacts on stream water quality: a review 1. JAWRA J. Am. Water Res.
Assoc. 41 (3), 591–606.
Aheeyar, M., 2013. Alternative Approaches to Small Tank/Cascade Rehabilitation: Socio-
Economic and Institutional Perspective: Hector Kobbekaduwa Agrarian Research
Table 7 and Training Institute Colombo.
Impacts on optimal N rate by changes in grain price, fertilizer price, damage cost Amarasingha, R.P.R.K., Suriyagoda, L.D.B., Marambe, B., Galagedara, L.W., Silva, G.L.L.
and exchange rate from 2018 to 2022. P., Punyawardena, R., Wijekoon, R., Nidumolu, U., Howden, M., 2015. Modelling
the impact of changes in rainfall distribution on the irrigation water requirement and
Season Yield outcome Optimum N rate yield of short and medium duration rice varieties using APSIM during Maha season
Category (kg/ha) in the dry zone of Sri Lanka. Trop. Agricult. Res. 26 (2).
Anderson, J.R., Dillon, J.L., Hardaker, J.B., 1977. Agricultural Decision Analysis. The
Base Scenario 1 Scenario 2 Iowa State University Press, Ames Iowa.
Low 197 175 78 Bert, F.E., Satorre, E.H., Toranzo, F.R., Podestá, G.P., 2006. Climatic information and
decision-making in maize crop production systems of the Argentinean pampas.
Maha Medium 175 253 82
Agric. Syst. 88 (2), 180–204. https://doi.org/10.1016/j.agsy.2005.03.007.
High 156 229 69
Brorsen, B.W., Richter, F.G.C., 2012. Experimental designs for estimating plateau-type
Low 128 187 55
production functions and economically optimal input levels. J. Prod. Anal. 38 (1),
Yala Medium 195 306 53 45–52. Retrieved from HTTP://unimelb.hosted.exlibrisgroup.com/sfxlcl41?
High 128 186 54 sid=google&auinit=BW&aulast=Brorsen&atitle=Experimental%20designs%20for
%20estimating%20plateau-type%20production%20functions%20and%
20economically%20optimal%20input%20levels&id=doi%3A10.1007%2Fs11123-
compared to the private economic N rates when accounting for the ef­ 010-0204-0&title=Journal%20of%20Productivity%
20Analysis&volume=38&issue=1&date=2012&spage=45&issn=0895-562X.
fects of nitrate transport into the tank water. The null hypothesis was
Cecchin, A., Pourhashem, G., Gesch, R.W., Lenssen, A.W., Mohammed, Y.A., Patel, S.,
that fertilizer rates would need to be substantially lowered to reduce Berti, M.T., 2021. Environmental trade-offs of relay-cropping winter cover crops
nitrate's adverse environmental effects on tank water quality. However, with soybean in a maize-soybean cropping system. Agric. Syst. 189, 103062 https://
there is very little difference between the social and private optimal N doi.org/10.1016/j.agsy.2021.103062.
Chowdary, V., Rao, N., Sarma, P., 2004. A coupled soil water and nitrogen balance model
fertilizer rates under unregulated prices. The socially optimal N rates for flooded rice fields in India. Agric. Ecosyst. Environ. 103 (3), 425–441.
were slightly lower, and the yield differences were quite small. Hence, Craswell, E., 2021. Fertilizers and nitrate pollution of surface and ground water: an
we reject the null hypothesis when unregulated (world market) prices increasingly pervasive global problem. SN Appl. Sci. 3 (4), 1–24.
Dash, J., Sarangi, A., Singh, D., Singh, A., Adhikary, P.P., 2015. Prediction of root zone
for N fertilizer were used for comparison. The fertilizer subsidy policy water and nitrogen balance in an irrigated rice field using a simulation model. Paddy
provides private economic benefits to smallholder farmers, yet at the Water Environ. 13 (3), 281–290.
risk of excessive nitrate transport into cascade tanks with consequent Delin, S., Stenberg, M., 2014. Effect of nitrogen fertilization on nitrate leaching in
relation to grain yield response on loamy sand in Sweden. Eur. J. Agron. 52,
effects on water quality for human use if farmers have access to un­ 291–296.
limited quantities of fertilizer at a subsidized rate. Dharma-Wardana, M.C., Amarasiri, S.L., Dharmawardene, N., Panabokke, C.R., 2015.
This modelling framework can be applied to simulate the effect of Chronic kidney disease of unknown aetiology and groundwater ionicity: study based
on Sri Lanka. Environ. Geochem. Health 37 (2), 221–231.
alternative fertilizer technologies on crop production decisions, for
Dillon, J.L., Anderson, J.R., 1990. The analysis of response in crop and livestock
example, the use of organic N fertilizers and slow-release fertilizers. This production. In: In, 3rd ed. Pergamon Press, pp. 66–83.
model can be applied to evaluate policy questions such as changing the Doole, G.J., Marsh, D., Ramilan, T., 2013. Evaluation of Agri-environmental policies for
reducing nitrate pollution from New Zealand dairy farms accounting for firm
fertilizer subsidy policy and including environmental effects on farmer
heterogeneity. Land Use Policy 30 (1), 57–66.
decisions by other policy mechanisms. FAO, 2018. Sri Lanka among Globally Important Agricultural Heritage Systems.
Farquharson, R.J., 2006. Production response and input demand in decision making:
Declaration of Competing Interest nitrogen fertilizer and wheat growers. Aust. Agribusiness Rev. 14 (1673-
2016–136786).
Fujihara, Y., Oda, M., Horikawa, N., Ogura, C., 2011. Hydrologic analysis of rainfed rice
The authors declare that they have no known competing financial areas using a simple semi-distributed water balance model. Water Resour. Manag. 25
interests or personal relationships that could have appeared to influence (9), 2061–2080.
Gibbons, J.M., Sparkes, D.L., Wilson, P., Ramsden, S.J., 2005. Modelling optimal
the work reported in this paper. strategies for decreasing nitrate loss with variation in weather – a farm-level
approach. Agric. Syst. 83 (2), 113–134. https://doi.org/10.1016/j.
Data availability agsy.2004.02.010.
Gourevitch, J.D., Keeler, B.L., Ricketts, T.H., 2018. Determining socially optimal rates of
nitrogen fertilizer application. Agric. Ecosyst. Environ. 254, 292–299.
Data will be made available on request. Hochman, Z., Waldner, F., 2020. Simplicity on the far side of complexity: optimizing
nitrogen for wheat in increasingly variable rainfall environments. Environ. Res. Lett.
15 (11), 114060.

11
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

Hoogesteger, J., Bolding, A., Sanchis-Ibor, C., Veldwisch, G.J., Venot, J.P., Vos, J., Mallawaarachchi, T., Quiggin, J., 2001. Modelling socially optimal land allocations for
Boelens, R., 2023. Communality in farmer managed irrigation systems: insights from sugar cane growing in North Queensland: a linked mathematical programming and
Spain, Ecuador, Cambodia and Mozambique. Agric. Syst. 204, 103552. choice modelling study. Aust. J. Agric. Resour. Econ. 45 (3), 383–409.
Hyytiäinen, K., Niemi, J.K., Koikkalainen, K., Palosuo, T., Salo, T., 2011. Adaptive Mallawaarachchi, T., Monypenny, R., Rayment, G.E., 2002. An integrated strategy to
optimization of crop production and nitrogen leaching abatement under yield enhance profitability and environmental compliance in the Australian sugar
uncertainty. Agric. Syst. 104 (8), 634–644. https://doi.org/10.1016/j. industry. In: Paper presented at the Proceedings of the 2002 Conference of the
agsy.2011.06.006. Australian Society of Sugar Cane Technologists held at Cairns, Queensland,
Itakura, J., 1995. Water Balance Model for Planning Rehabilitation of a Tank Cascade Australia, 29 April-2 May 2002.
Irrigation System in Sri Lanka. Retrieved from. http://sfx.unimelb.hosted.exlibrisgr Mapa, R.B., 2020. The Soils of Sri Lanka [electronic resource]:. Springer.
oup.com/sfxlcl41/?genre=article&isbn=&issn=edsagr&title=&volume=&issue Mapa, R.B., Pathmarajah, S., 1995. Contrasts in the physical properties of three soils of
=&date=19950101&atitle=Water%20balance%20model%20for%20planning% an Alfisol catena in Sri Lanka. Soil Use Manag. 11 (2), 90–93.
20rehabilitation%20of%20a%20tank%20cascade%20irrigation%20system%20in% Miller, C.M.F., Waterhouse, H., Harter, T., Fadel, J.G., Meyer, D., 2020. Quantifying the
20Sri%20Lanka&aulast=Itakura%2C%20J.&spage=&sid=EBSCO%3AAGRIS&pid. uncertainty in nitrogen application and groundwater nitrate leaching in manure
Itakura, J., Abernethy, C.L., 1993. Water Management in a Tank Cascade Irrigation based cropping systems. Agric. Syst. 184, 102877 https://doi.org/10.1016/j.
System in Sri Lanka: First Seasonal Report of TARC-IIMI Joint Project 1991/1992 agsy.2020.102877.
Maha Season, Vol. 24. IWMI. Ministry of Finance, 2019. Annualk Report 2019. Retrieved from. Colombo.
Jayatilaka, C.J., 2001. Predicting water availability in irrigation tank cascade systems: Ministry of Finance, 2020. Annual Report 2020. Retrieved from. Colombo.
the cascade water balance model. Retrieved from. http://sfx.unimelb.hosted.exlibris Ministry of Finance, 2021. Annual Report 2021. Retrieved from. Colombo.
group.com/sfxlcl41/?genre=article&isbn=&issn=edsagr&title=&volume=&issue Monjardino, M., McBeath, T.M., Brennan, L., Llewellyn, R.S., 2013. Are farmers in low-
=&date=20010101&atitle=Predicting%20water%20availability%20in%20irrigatio rainfall cropping regions under-fertilizing with nitrogen? A risk analysis. Agric. Syst.
n%20tank%20cascade%20systems%3A%20the%20cascade%20water%20balance% 116, 37–51. https://doi.org/10.1016/j.agsy.2012.12.007.
20model&aulast=Jayatilaka%2C%20C.%20J.&spage=&sid=EBSCO%3AAGRIS&pi Morris, T.F., Murrell, T.S., Beegle, D.B., Camberato, J.J., Ferguson, R.B., Grove, J.,
d=. McGrath, J.M., 2018. Strengths and limitations of nitrogen rate recommendations
Jayatilaka, C.J., Sakthivadivel, R., Shinogi, Y., Makin, I.W., Witharana, P., 2003. for corn and opportunities for improvement. Agron. J. 110 (1), 1–37.
A simple water balance modelling approach for determining water availability in an Neitsch, S.L., Arnold, J.G., WIlliams, J.R., 2011. Soil and Water Assessment Tool
irrigation tank cascade system. J. Hydrol. (Amsterdam) 273 (1–4), 81–102. Theroritical Documentation Version 2009. Retrieved from Texas A & M University
Retrieved from. http://sfx.unimelb.hosted.exlibrisgroup.com/sfxlcl41?sid=google System, Collage Station, Texas.
&auinit=CJ&aulast=Jayatilaka&atitle=A%20simple%20water%20balance% Oka, N., Yamaji, E., Witharana, P., 2019. Validation error calculation of water volume
20modelling%20approach%20for%20determining%20water%20availability%20in estimation a tank cascade system. IDJR J. 308 (87), 71–79.
%20an%20irrigation%20tank%20cascade%20system&id=doi%3A10.1016% Oyonarte, N.A., Gómez-Macpherson, H., Martos-Rosillo, S., González-Ramón, A.,
2FS0022-1694%2802%2900360-8&title=Journal%20of%20Hydrology&volume Mateos, L., 2022. Revisiting irrigation efficiency before restoring ancient irrigation
=273&issue=1-4&date=2003&spage=81&issn=0022-1694. canals in multi-functional, nature-based water systems. Agric. Syst. 203, 103513.
Jiang, Y., Deng, A., Bloszies, S., Huang, S., Zhang, W., 2017. Non-linear response of soil Panabokke, C.R., Sakthivadivel, R., Weerasinghe, A.D., 2002. Evolution, Present Status
ammonia emissions to fertilizer nitrogen. Biol. Fertil. Soils 53 (3), 269–274. and Issues Concerning Small Tank Systems in Sri Lanka. International Water
Johnes, P.J., Foy, R., Butterfield, D., Haygarth, P., 2007. Land use scenarios for England Managemnet Institute, Colombo, Sri Lanka.
and Wales: evaluation of management options to support ‘good ecological status’ in Parashar, D., Kulshrestha, U., Sharma, C., 1998. Anthropogenic emissions of NOx, NH3
surface freshwaters. Soil Use Manag. 23, 176–194. and N2O in India. Nutr. Cycl. Agroecosyst. 52 (2), 255–259.
Kandulu, J.M., Bryan, B.A., King, D., Connor, J.D., 2012. Mitigating economic risk from Pham, H., 2019. A new criterion for model selection. Mathematics 7 (12), 1215.
climate variability in rain-fed agriculture through enterprise mix diversification. Pretty, J.N., Brett, C., Gee, D., Hine, R., Mason, C., Morison, J., Raven, H., Rayment, M.,
Ecol. Econ. 79, 105–112. Van der Bijl, G., 2000. An assessment of the total external costs of UK agriculture.
Kandulu, J., Thorburn, P., Biggs, J., Verburg, K., 2018. Estimating economic and Agric. Syst. 65 (2), 113–136.
environmental trade-offs of managing nitrogen in Australian sugarcane systems Ramalingam, S., Panneerselvam, B., Kaliappan, S.P., 2022. Effect of high nitrate
taking agronomic risk into account. J. Environ. Manag. 223, 264–274. contamination of groundwater on human health and water quality index in semi-arid
Kanthilanka, H., 2022. Evaluation of Nitrogen Fertilizer Decisions by Rice Farmers: region, South India. Arab. J. Geosci. 15 (3), 1–14.
Accounting for the Effects of Nitrate on Water Quality for Communities in Cascaded Ramilan, T., Scrimgeour, F., Levy, G., Marsh, D., Romera, A.J., 2011. Simulation of
Village Tank Systems in Sri Lanka (Unpublished PhD thesis).. The University of alternative dairy farm pollution abatement policies. Environ. Model. Softw. 26 (1),
Melbourne, Melbourne. 2–7.
Keeler, B.K., Gourevitch, Jessie D., Polasky, Stephen, Isbell, Forest, Tessum, Chris W., Ranasinghe, R.D.A.K., Korale-Gedara, P.M., Weerasooriya, S.A., 2023. Climate change
Hill, Jason D., Marshall, Julian D., 2016. The social costs of nitrogen. Sci. Adv. 2 adaptation and adaptive capacities of dairy farmers: evidence from village tank
(e1600219). cascade systems in Sri Lanka. Agric. Syst. 206, 103609.
Knowling, M.J., White, J.T., McDonald, G.W., Kim, J.-H., Moore, C.R., Hemmings, B., Ratnayake, S.S., Kumar, L., Dharmasena, P.B., Kadupitiya, H.K., Kariyawasam, C.S.,
2020. Disentangling environmental and economic contributions to hydro-economic Hunter, D., 2021. Sustainability of village tank cascade systems of Sri Lanka:
model output uncertainty: An example in the context of land-use change impact exploring cascade anatomy and socio-ecological nexus for ecological restoration
assessment. Environ. Model Softw. 127, 104653. planning. Challenges 12 (2), 24.
Kruseman, G., Bairagi, S., Komarek, A.M., Molero Milan, A., Nedumaran, S., Petsakos, A., Sagehashi, M., Mori, H., Hareyama, Y., Sakuma, K., Akiba, M., Hosomi, M., 2016.
Yigezu, Y.A., 2020. CGIAR modeling approaches for resource-constrained scenarios: Integration of the rice paddy water management into a coupled surface-subsurface
II. Models for analyzing socioeconomic factors to improve policy recommendations. water flow model in the Sakuragawa River watershed (Japan). Hydrol. Res. 47 (1),
Crop Sci. 60 (2), 568–581. 137–156.
Kuhn, T., Enders, A., Gaiser, T., Schäfer, D., Srivastava, A.K., Britz, W., 2020. Coupling Sakthivadivel, R., 1997. Rehabilitation Planning for Small Tanks in Cascades: A
crop and bio-economic farm modelling to evaluate the revised fertilization Methodology Based on Rapid Assessment.
regulations in Germany. Agric. Syst. 177, 102687 https://doi.org/10.1016/j. Sakthivadivel, R., Fernando, N., Brewer, J.D., 1997. Rehabilitation Planning for Small
agsy.2019.102687. Tanks in Cascades: A Methodology Based on Rapid Assessment: International
Kulasinghe, T.N., Dharmakeerthi, R.S., 2022. Effects of land use type and tank Irrigation Management Institute (IIMI).
components on soil properties and sustainability of tank cascade system in the dry Samuel, T.D.M.A., 1991. Estimation of global radiation for Sri Lanka. Sol. Energy 47 (5),
zone of north Central Sri Lanka. Agric. Syst. 201, 103474. 333–337.
Lam, Q., Schmalz, B., Fohrer, N., 2010. Modelling point and diffuse source pollution of Shinogi, Y., 1998. Simulation of the Water Balance in a Dry Zone Tank Cascade. Paper 68
nitrate in a rural lowland catchment using the SWAT model. Agric. Water Manag. 97 of Session 2. Retrieved from. http://sfx.unimelb.hosted.exlibrisgroup.com/sfxlcl
(2), 317–325. 41/?genre=article&isbn=&issn=edsagr&title=&volume=&issue=&da
Le, C., Zha, Y., Li, Y., Sun, D., Lu, H., Yin, B., 2010. Eutrophication of Lake waters in te=19980101&atitle=Simulation%20of%20the%20water%20balance%20in%20a%
China: cost, causes, and control. Environ. Manag. 45 (4), 662–668. https://doi.org/ 20dry%20zone%20tank%20cascade.%20Paper%2068%20of%20Session%
10.1007/s00267-010-9440-3. 202&aulast=Shinogi%2C%20Y.&spage=&sid=EBSCO%3AAGRIS&pid=.
Leach, E.R., 1959. Hydraulic society in Ceylon. Past & Present 15, 2–26. Sihvonen, M., Pihlainen, S., Lai, T.-Y., Salo, T., Hyytiäinen, K., 2021. Crop production,
Lehmann, N., Finger, R., Klein, T., Calanca, P., Walter, A., 2013. Adapting crop water pollution, or climate change mitigation—which drives socially optimal
management practices to climate change: modeling optimal solutions at the field fertilization management most? Agric. Syst. 186, 102985 https://doi.org/10.1016/j.
scale. Agric. Syst. 117, 55–65. https://doi.org/10.1016/j.agsy.2012.12.011. agsy.2020.102985.
Lesschen, J.P., Verburg, P.H., Staal, S.J., 2005. Statistical Methods for Analyzing the Sirimanna, S., Kahathuduwa, K.K.P.N., Prasada, D.V.P., 2022. Are cascade reservoir
Spatial Dimension of Changes in Land Use and Farming Systems. International systems sustainable agroecosystems? A comparative assessment of efficiency,
Livestock Research Institute Kenya. effectiveness and resource footprint in a Sri Lankan micro-cascade. Agric. Syst. 203,
Link, J., Graeff, S., Batchelor, W.D., Claupein, W., 2006. Evaluating the economic and 103493.
environmental impact of environmental compensation payment policy under Tan, X., Shao, D., Gu, W., 2018. Improving water reuse in paddy field districts with
uniform and variable-rate nitrogen management. Agric. Syst. 91 (1), 135–153. cascaded on-farm ponds using hydrologic model simulations. Water Resour. Manag.
https://doi.org/10.1016/j.agsy.2006.02.003. 32 (5), 1849–1865.
Mack, U.D., Feger, K.H., Gong, Y., Stahr, K., 2005. Soil water balance and nitrate Thornley, J.H.M., France, J., 2007. Mathematical Models in Agriculture: Quantitative
leaching in winter wheat–summer maize double-cropping systems with different Methods for the Plant, Animal and Ecological Sciences, 2nd ed. CAB International,
irrigation and N fertilization in the North China plain. J. Plant Nutr. Soil Sci. 168 (4), Wallingford UK.
454–460.

12
H. Kanthilanka et al. Agricultural Systems 207 (2023) 103628

Tilman, D., 1999. Global environmental impacts of agricultural expansion: the need for 20of%20Nitrate%20and%20Phosphate%20in%20Selected%20Six%20Small%
sustainable and efficient practices. Proc. Natl. Acad. Sci. 96 (11), 5995–6000. 20Tanks%20of%20Dry%20Zone%20in%20Sri%20Lanka&id=doi%3A10.4038%
Tilman, D., Cassman, K.G., Matson, P.A., Naylor, R., Polasky, S., 2002. Agricultural 2Ftar.v23i3.4665.
sustainability and intensive production practices. Nature 418 (6898), 671. Wijesundara, W., Nandasena, K., Jayakody, A., 2013. Spatial and temporal changes in
UNDP, 2017. Project Document - Strengthening the Resilience of Smallholder Farmers in nitrogen, phosphorus and potassium concentration in water in the Thirappane tank
the Dry Zone to Climate Variability and Extreme Events through an Integrated Cascade in dry zone of Sri Lanka. J. Environ. Professionals Sri Lanka 1 (1).
Approach to Water Management. United Nations Development Programme. Worrall, F., Burt, T., 1999. The impact of land-use change on water quality at the
Wan, W., Han, Y., Wu, H., Liu, F., Liu, Z., 2021. Application of the source–sink landscape catchment scale: the use of export coefficient and structural models. J. Hydrol. 221
method in the evaluation of agricultural non-point source pollution: first estimation (1–2), 75–90.
of an orchard-dominated area in China. Agric. Water Manag. 252, 106910 https:// Wriedt, G., Rode, M., 2006. Modelling nitrate transport and turnover in a lowland
doi.org/10.1016/j.agwat.2021.106910. catchment system. J. Hydrol. 328 (1–2), 157–176.
Ward, M.H., Jones, R.R., Brender, J.D., De Kok, T.M., Weyer, P.J., Nolan, B.T., Van Yin, Y., Ying, H., Xue, Y., Zheng, H., Zhang, Q., Cui, Z., 2019. Calculating socially optimal
Breda, S.G., 2018. Drinking water nitrate and human health: an updated review. Int. nitrogen (N) fertilization rates for sustainable N management in China. Sci. Total
J. Environ. Res. Public Health 15 (7), 1557. Environ. 688, 1162–1171.
Warnakulasooriya, W., Shantha, A., 2021. The impact of water management practices on Young, S.M., Pitawala, A., Gunatilake, J., 2010. Fate of phosphate and nitrate in waters
Paddy productivity in the dry zone of Sri Lanka. Aust. J. Basic Appl. Sci. 15 (5), 1–9. of an intensive agricultural area in the dry zone of Sri Lanka. Paddy Water Environ. 8
Weerahewa, J., Dayananda, D., 2023. Land use changes and economic effects of (1), 71–79. Retrieved from. https://doi.org/10.1007/s10333-009-0186-6.
alternative fertilizer policies: a simulation analysis with a bio-economic model for a Yu, W., Yue, Y., Wang, F., 2022. The spatial-temporal coupling pattern of grain yield and
Tank Village of Sri Lanka. Agric. Syst. 205, 103563. fertilization in the North China plain. Agric. Syst. 196, 103330 https://doi.org/
Weerahewa, J., Kodithuwakku, S.S., Ariyawardana, A., 2010. In: Pinstrup-Andersen, P., 10.1016/j.agsy.2021.103330.
Cheng, F. (Eds.), The fertilizer subsidy program in Sri Lanka. Food policy for Zhang, H., Yang, R., Wang, Y., Ye, R., 2019. The evaluation and prediction of agriculture-
developing countries: case studies. Cornell University, Ithaca. Retrieved August, 26, related nitrate contamination in groundwater in Chengdu Plain, southwestern China.
2014. Hydrogeol. J. 27 (2), 785–799.
Wickramasinghe, M.R.C.P., Dayawansa, N.D.K., Jayasiri, M.M.J.G.C.N., De Silva, R.P., Zhang, H., Yang, R., Guo, S., Li, Q., 2020. Modeling fertilization impacts on nitrate
2023. A study on external pressures of an ancient irrigation cascade system in leaching and groundwater contamination with HYDRUS-1D and MT3DMS. Paddy
SriLanka. Agric. Syst. 205 https://doi.org/10.1016/j.agsy.2022.103593. Water Environ. 18 (3), 481–498.
Wijesundara, W., Nandasena, K.A., Jayakody, A.N., 2012. Temporal variation of nitrate Zhao, S., Peng, C., Jiang, H., Tian, D., Lei, X., Zhou, X., 2006. Land use change in Asia
and phosphate in selected six small tanks of dry zone in Sri Lanka. Trop. Agricult. and the ecological consequences. Ecol. Res. 21 (6), 890–896.
Res. 23 (3), 277. Retrieved from. http://sfx.unimelb.hosted.exlibrisgroup.com/sfxlcl Zoysa, D., Weerasinghe, T., 2016. Study on the chemical and biological status of Malwatu
41?sid=google&auinit=W&aulast=Wijesundara&atitle=Temporal%20Variation% oya basin, Anuradhapura in Sri Lanka. Br. J. Environ. Sci. 4 (1), 1.

13