ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

Contents lists available at ScienceDirect

ISPRS Journal of Photogrammetry and Remote Sensing

journal homepage: www.elsevier.com/locate/isprsjprs

The seasonal carbon and water balances of the Cerrado environment of

Brazil: Past, present, and future influences of land cover and land use
Arielle Elias Arantes a,, Laerte G. Ferreira a, Michael T. Coe b
a
b

Federal University of Gois, Image Processing and GIS Lab, Campus Samambaia, 74001-970 Goinia, Gois, Brazil
The Woods Hole Research Center, 149 Woods Hole Rd, Falmouth, MA 02540, United States

a r t i c l e

i n f o

Article history:
Received 9 July 2015
Received in revised form 3 February 2016
Accepted 9 February 2016

Keywords:
Cerrado
Carbon
Evapotranspiration
Phenology
Land cover and land use

a b s t r a c t
The Brazilian savanna (known as Cerrado) is an upland biome made up of various vegetation types from
herbaceous to arboreal. In this paper, MODIS remote sensing vegetation greenness from the Enhanced
Vegetation Index (EVI) and evapotranspiration (ET) data for the 20002012 period were analyzed to
understand the differences in the net primary productivity (NPP-proxy), carbon, and the evaporative flux
of the major Cerrado natural and anthropic landscapes. The understanding of the carbon and evaporative
fluxes of the main natural and anthropic vegetation types is of fundamental importance in studies regarding the impacts of land cover and land use changes in the regional and global climate. The seasonal
dynamics of EVI and ET of the main natural and anthropic vegetation types of the Cerrado biome were
analyzed using a total of 35 satellite-based samples distributed over representative Cerrado landscapes.
Carbon and water fluxes were estimated for different scenarios, such as, a hypothetical unconverted
Cerrado, 2002 and 2050 scenarios based on values derived from literature and on the PROBIO land cover
and land use map for the Cerrado. The total growing season biomass for 2002 in the Cerrado region was
estimated to be 28 gigatons of carbon and the evapotranspiration was 1336 gigatons of water. The mean
estimated growing season evapotranspiration and biomass for 2002 was 576 Gt of water and 12 Gt of carbon for pasture and croplands compared to 760 Gt of water and 15 Gt of carbon for the Cerrado natural
vegetation. In a modeled future scenario for the year 2050, the ET flux from natural Cerrado vegetation
was 394 Gt less than in 2002 and 991 Gt less than in an unconverted scenario, with only natural vegetation, while the carbon was 8 Gt less than in 2002 and 21 Gt less than in this hypothetical pre-conversion
Cerrado. On the other hand, the sum of the pasture and cropland ET flux increased by 405 Gt in 2050 relative to 2002 and the carbon by 11 Gt of carbon. Given the increasing global demand for agricultural
products and the insufficient protected areas in the Cerrado (with a significant area of remaining native
vegetation in privately owned lands that may be legally deforested), our analyses suggest that potential
future changes to the water and carbon balances are likely to be highly significant in the severely threatened Cerrado biome. On the other hand, our results also suggest that the recovery of degraded pastures
can have a positive impact on climate, due to the higher rates of carbon sequestration and water transfer
to the atmosphere.
2016 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier
B.V. All rights reserved.

1. Introduction
The Brazilian savanna (known as Cerrado in Brazil) is an upland
biome located in the central part of Brazil, covering about 25% of
the countrys land area (i.e. 2 million km2) (Eiten, 1972). The
rapid deforestation occurring in the Cerrado, with a mean
deforestation rate of 1.6% (Silva et al., 2009; Rocha et al., 2011),
Corresponding author. Tel.: +55 62 92143443.
E-mail addresses: aearantes@gmail.com
(L.G. Ferreira), mtcoe@whrc.org (M.T. Coe).

(A.E.

Arantes),

laerte@ufg.br

is potentially important for the energy, water, and carbon cycles,

as the replacement of natural vegetation by pastures and cropland
affects the land surface - atmosphere feedbacks. The soils of the
Cerrado are mainly deep with low fertility, high iron and
aluminum content, and excellent internal drainage (Buol, 2009).
The average annual temperature varies from 2026 C, and total
rainfall and evapotranspiration are 1481 mm and 895 mm, respectively (Marcuzzo et al., 2012; Eiten, 1972). The climate has two
defined seasons, a wet season from October to April and a dry
season from May to September (Camargo, 1963). The vegetation
of the Cerrado has many different structural forms (height, density,

http://dx.doi.org/10.1016/j.isprsjprs.2016.02.008
0924-2716/ 2016 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier B.V. All rights reserved.

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

and layers), varying from herbaceous, grassy, and shrubby vegetation to woodland (Goodland, 1971). The Cerrado trees and shrubs
generally have thick bark, twisted branches and trunks, glabrous
or soft and hairy leaves, relatively low leaf density, and crown
wider for its height than forest trees (Eiten, 1972). The ground
layer is more or less xeromorphic with grasses and sedges with
hard siliceous leaves. Most of the Cerrado species are perennial
with some annual species in the northeastern region.
The combination of different structural forms determines the
various Cerrado vegetation types, and these can be divided into
five main types: campo limpo (Cerrado grassland), campo sujo
(Cerrado shrubland), campo Cerrado (low tree and shrub Cerrado), Cerrado (Cerrado stricto sensu) and Cerradao (Cerrado
woodland). The most abundant vegetation type is the Cerrado
stricto sensu (22%), and the least abundant are the Cerrado
woodland (5%) and Cerrado grassland (4%) (Sano et al., 2010).
The forested vegetation types in the Cerrado biome differ from
the Cerrado woodland, in their structural components, such as
higher height and larger canopy size, and compositional components, like the occurrence of different tree species (Eiten, 1972).
These forests, when located away from river courses, are called
seasonal forests (floresta estacional). The forested vegetation
types are also classified based on the fraction of leaf loss during
the dry season, as evergreen (less than 20%), semi-deciduous
(20 to 50% leaf loss), or deciduous (more than 50%) (Pereira
et al., 2011). About 50% of the natural Cerrado vegetation has
been converted to pasture and agriculture, and less than 3%
are within protected areas (Couto et al., 2010; Garcia et al.,
2011).
The occupation of the Cerrado biome started in the 18th century with cattle ranching activities over natural pastures and
small subsistence farming (Silva et al., 2013). In the 70s, the
Brazilian government conducted studies that showed the agricultural potential of the Cerrado and the required technical implementations in order to increase its productivity (Silva et al.,
2013). Since the 1970s, the herbaceous and woody vegetation,
formerly used as natural pastures and sources of food for cattle,
have been replaced by exotic cultivated pastures of African origin
in the genre Brachiaria (more than 80% of the cultivated pastures
in the central part of Brazil), Panicum, and Andropogon (Brossard
and Barcellos, 2005). Cultivated pastures occupied 29% of the Cerrado biome in 2002, with 40% of that concentrated in the southern portion of the biome, particularly in eastern Mato Grosso do
Sul and western Gois (Sano et al., 2000; Sano et al., 2010). The
Cerrado biome supports 40% of the Brazilian cattle herd, over
50 million hectares of cultivated pastures, and contributes to
about 55% of the national meat production (Vendrame et al.,
2010; Brossard and Barcellos, 2005).
The transformation of the naturally poor and acid soils of the
Cerrado biome into productive soils in the 1970s by the introduction of correctives and fertilizers, and improved infrastructure, also
allowed for a boom in the expansion of soy, corn, and bean crops
(Cunha et al., 1994; Jepson, 2005; Klink and Machado, 2005). Such
technological advances and increased land profitability were
instrumental for transforming the Cerrado into the most prominent agricultural frontier of the country (Rezende, 2002). Together,
the commodity crops occupy about 10% of the total Cerrado area.
With the global demand for food rapidly increasing, future expansion of crops into degraded pastures and intensification of cattle
ranching (through the use of partial confinement and fodder) are
likely to occur (Mueller, 2003; Klink and Machado, 2005;
Brando et al., 2006)
Compared to forests, pasturelands and croplands have lower
above-ground and below-ground biomass, higher albedo,
decreased evapotranspiration, lower canopy interception of rainfall and less atmospheric turbulence (Arago et al. 2007;

67

DAlmeida 2007; Bonan, 2008; Coe et al., 2009, 2013; Loarie

et al., 2011; Lathuillire et al., 2012; Spracklen et al., 2012). In fact,
during the 18502000 period, land use change accounted for the
release of about 156 PgC globally (60% from the tropics)
(Houghton, 2003), while the net carbon flux from land use and land
cover change alone accounted for approximately 13% of the carbon
emissions from 1990 to 2010 (Houghton, 2012). This substantial
increase in atmospheric carbon hinders the absorption of carbon
by the oceans, changing the carbon-cycle feedbacks, which accelerates climate change.
Thus, land use transitions from natural vegetation to pastures
and crops decrease carbon stocks, increase greenhouse gas emissions (Soares-Filho et al., 2014), reduce evapotranspiration (Costa
and Pires 2009; Lathuilliere et al. 2012), and increase sensible
heat flux (Ferreira et al., 2011; Giambelluca et al., 2009), all of
which have significant environmental implications. Bustamante
et al. (2012) showed that GHG emissions from cattle ranching
in the Cerrado biome (229231 Mt CO2 eq) accounted for
approximately 2030% of all GHG emissions from cattle ranching
in Brazil (8131,090 Mt CO2 eq). The decrease in ET associated
with land use change leads to a soil moisture increase, with
excess water being exported via increased runoff and river discharge (Costa et al. 2003; Coe et al. 2011; Hayhoe et al. 2011),
which can, ultimately, reduce regional rainfall (Costa and Pires
2009).
Regarding the Cerrado landscapes, some studies have investigated the impact of land use change on soil carbon stocks
(Batte-Bayer et al., 2010; Silva et al., 2004; Pinto et al.; 2014;
Braz et al.,2013; Pimentel et al., 2012), as well as on GHG emissions from land use and land cover changes (Fearnside et al.,
2009; Bustamante et al., 2012; Cerri et al., 2009), and on the carbon and water fluxes of the different Cerrado vegetation types
(Rocha et al.; 2002; Paiva and Faria, 2007; Miranda et al., 1997;
Santos et al., 2003). A literature review by Miranda et al. (2014)
showed that the biomass of the Cerrado vegetation types varies
from a mean value of 24 Mg ha 1 for the Cerrado grassland,
58 Mg ha 1 for the Cerrado shrubland, and 98 Mg ha 1 for the
Cerrado forestlands (Miranda et al., 2014). Rocha et al. (2009)
estimated the water and heat fluxes for a gradient of natural vegetation from the Amazon forest to the Cerrado savanna biomes,
indicating evapotranspiration rates during the dry season of
2.5 mm d 1 in the forest and 1.0 mm d 1 in the Cerrado. Using
MODIS products (Moderate Resolution Imaging Spectroradiometer), Loarie et al. (2011) found that the conversion of native Cerrado to pasture or non-sugar cane crops resulted, on average, in a
0.6 mm d 1 decrease in evapotranspiration, showing that the Cerrado natural vegetation plays a key role in maintaining the water
balance.
Our study estimates both ET and carbon fluxes based on a regional scale, taking into consideration both the different Cerrado vegetation types, as well as the changes in ET and carbon fluxes from
the conversion of these natural vegetation types to anthropic vegetation types (pastures and crops). Differently from previous studies, focused on small, restricted areas, this is the first study that
estimates carbon and water fluxes for the entire Cerrado biome,
based on freely available and ready to use satellite products and
on simple and replicable approaches. Specifically, we used monthly
MODIS data on vegetation greenness from the Enhanced Vegetation Index (EVI) and evapotranspiration (ET) from 2000 to 2012,
to quantify the differences in the net primary productivity (NPPproxy) and water vapor flux (ET) of the major Cerrado natural
and anthropic landscapes. Additionally, we investigated the historical changes to biomass and water vapor flux that have resulted
from deforestation in the Cerrado and the potential changes that
may occur in the future if deforestation continues at its current
rates.

68

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

Fig. 1. Distribution of the 35 satellite-based samples over the major Cerrado vegetation types (the accompanying photographs depict the typical appearance of each
vegetation type during the growing season).

2. Methods and Data Analysis

The seasonal dynamics of EVI and ET of the main natural and
anthropic vegetation types of the Cerrado biome were analyzed
using a total of 35 satellite-based reference samples, with each
sample distributed over representative Cerrado landscapes
(Fig. 1). Each reference sample came from a large area of the same
vegetation type on the ground (>1 km2), in order to avoid mixture
of different vegetation types as detected by the MODIS sensor. The
35 samples, comprising five samples from each of the seven dominant land-cover classes in the biome, were selected based on
visual analysis of Landsat scenes and field information: a) natural
Cerrado classes (i.e. Cerrado grassland, Cerrado shrubland, Cerrado
stricto sensu, and Cerrado woodland vegetation types), b) deciduous forest, c) cultivated pastures and d) crops (soy).
The Cerrado grassland, Cerrado shrubland, and Cerrado stricto
sensu samples were chosen in national parks, based on sites from
a previous study by Ferreira and Huete (2004). The Cerrado woodland is rare because it occurs in highly fertile soils, which have
been intensively deforested for agricultural activities. These samples were chosen from Solrzano et al. (2012), who identified some
remaining Cerrado woodland fragments (Fig. 1). For the seasonally

deciduous forests, distributed in the central part of Brazil, mainly

over limestone terrains (Felfili et al., 2007), samples were chosen
from Pereira et al. (2011).
The soy samples were chosen by temporal analysis of Landsat 5
TM images from 2000 to 2012, in the municipalities of Rio Verde
and Jata (Gois State), known for their large-scale soy production
(Fig. 1). The cultivated pasture samples were based on field data
from the Rio Vermelho watershed (three samples of Brachiaria
brizantha) and from areas of cattle ranching in the municipality
of Nova Crixs (two samples of Brachiaria decumbens) (Gois State)
(Fig. 1).
The MOD13Q1 EVI (Enhanced Vegetation Index) images1 from
April 2000 to December 2012 were processed and organized for
the entire Cerrado biome, in order to analyze the seasonal photosynthetic responses. The MOD13Q1 vegetation index product is a 16day composite with 250 m spatial resolution, where each pixel,
selected based on the MVC-CV approach (Maximum Value Composite Constrained View), corresponds to the best possible observation, i.e. with minimum residual cloud and aerosol contamination

Source of MOD13Q1 images: http://reverb.echo.nasa.gov/reverb/.

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

69

Fig. 2. Flowchart depicting the main steps and approaches utilized for the assessment of carbon and water fluxes in the Brazilian Cerrado.

Table 1
Average (all samples) start of season (SOS), end of season (EOS) and length for the 12
growing seasons considered in this study.
Seasons

SOS

EOS

Length

1
2
3
4
5
6
7
8
9
10
11
12

29-Sep-00
30-Sep-01
30-Sep-02
16-Oct-03
15-Oct-04
30-Sep-05
14-Sep-06
16-Oct-07
15-Oct-08
14-Sep-09
16-Oct-10
30-Sep-11

10-Jun-01
25-May-02
10-Jun-03
9-Jun-04
10-Jun-05
25-May-06
10-Jun-07
9-Jun-08
10-Jun-09
10-Jun-10
10-Jun-11
25-Jun-12

8 months
7 months
8 months
7 months
7 months
7 months
8 months
7 months
7 months
8 months
7 months
8 months

11 days
25 days
10 days
23 days
25 days
25 days
25 days
26 days
25 days
26 days
25 days
25 days

and closer to nadir viewing geometry (Huete et al., 2002; Solano

et al., 2010).
The monthly mean 1-km resolution MOD16A2 evapotranspiration (ET) and potential evapotranspiration (PET) products2 were
processed and organized for the period from 2000 to 2012 to analyze
the energy and water balance. The land surface evapotranspiration
product is described by Mu et al. (2011) and is based on the Penman
Monteith equation. Its calculation uses the following input data:
land cover from MODIS Collection 4 (MOD12Q1), albedo
(MOD43C1), leaf area index (MOD15A2), enhanced vegetation index
(MOD13Q1) and daily meteorological data from the NASA Global
Modeling and Assimilation Office (GMAO).
Mu et al. (2011) reported good agreement of the annual global
ET estimated by the MOD16 product compared to literature values
(62.8  103 km3 MOD16 and 65.5  103 km3 from literature). However, they found that the satellite-based ET values showed discrepancies of about 24% when compared to the ET measured in the
eddy flux towers. At two sites in the Amazon and Cerrado biomes,
Ruhoff et al. (2011) found a correlation of 0.50 and a RMSE of
0.97 mm day 1 between the ET data derived from eddy flux towers
and the mean daily MODIS ET (8-day composite). At longer time
intervals (i.e. monthly ET composites), the correlation was higher
(r = 0.7) and the RMSE was 18% lower.
2

Source of MOD16 ET and PET images: http://www.ntsg.umt.edu/project/mod16.

The TRMM (Tropical Rainfall Measuring Mission, Kummerow

et al., 1998)3 precipitation data (in millimeters), at 25 km spatial
resolution, was also processed and organized for the 2000 to 2012
period. The monthly average precipitation data was used to characterize the Cerrado precipitation regime, and compare the seasonality
of rainfall to EVI and ET responses.
The MOD13Q1 EVI images were used to determine the start
(SOS) and end (EOS) of the growing seasons and to derive secondary phenological parameters, such as the left and right derivative (green-up and senescence rates, respectively) and small
integral (EVI-integral used as NPP-proxy) (Fig. 2). The MODIS EVI,
with optimized vegetation signal in regions with high biomass
and low sensitivity to background and atmospheric contaminations, is a robust tool for retrieving information such as vegetation
type, net primary productivity, leaf area index, foliage cover, evapotranspiration, and phenology (Zhang et al., 2003; Tan et al., 2008;
Huete and Saleska, 2010; Mu et al., 2011; Zhang et al., 2012; Rosa
and Sano, 2013). Indeed, several studies have used MODIS EVI
time-series to estimate NPP and GPP (e.g. Rosa and Sano, 2013;
Potter et al., 2009; Bandaru et al.,2013; Wu et al., 2014), others
used EVI in models to estimate carbon in above-ground biomass
(e.g. Baccini et al., 2012; Li et al., 2015), while a few made use of
EVI to characterize the phenology of the vegetation (e.g. Goward
et al., 1985; Reed et al., 1994; Zhang et al., 2014; Wagle et al.,
2015).
Specifically for the Cerrado biome, field spectral measurements
tend to be well correlated to ground biophysical parameters (e.g.
green cover and LAI) and to satellite-based vegetation indices, corroborating the potential of satellite vegetation index data to discriminate among land cover types and to assess, in a more cost
effective manner, the seasonality and phenology of the main Cerrado vegetation types (e.g. Ferreira et al., 2003; Ratana and
Huete, 2005). Residual atmospheric noise in the EVI images were
smoothed using the Savitzky-Golay filter implemented in the
TIMESAT software (Jnsson and Eklundh, 2002; Jnsson and
Eklundh, 2004). TIMESAT was also used to identify the timing of
phenological phases from the smoothed data for all samples (35
in total) for the 13 years of analysis. In addition, the 13-year mean
for the start, end, and length of the growing seasons, on a per-pixel

Source of TRMM precipitation images: http://mirador.gsfc.nasa.gov/.

70

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

Fig. 3. (a) Start of the growing season (SOS), (b) end of the growing season (EOS), and (c) length of the growing season for the entire Cerrado biome.

basis, were estimated for the entire Cerrado biome, in order to

understand the differences in its phenological behavior due to
the latitudinal variability.
As proposed by Jnsson and Eklundh (2002), the start of the
growing season (SOS) and end of the growing season (EOS) were
defined as the calendar dates in each year of the EVI time-series,
when the EVI value crossed specific thresholds. In the case of
SOS, it is the first calendar day following the minimum seasonal
value for that year, when the EVI has increased to greater than
20% of the minimum value. For EOS, it is defined as the first calendar day following that years EVI peak, when EVI has decreased to
less than 80% of the peak value. Thus, SOS and EOS are the temporal bounds of the growing season. The left and right derivatives
were calculated as the total increase of the curve from the SOS to
the maximum value and the total decrease from the maximum
value to the EOS, respectively.
The accumulated EVI from the SOS to the EOS, also known as
the small integral in time-domain metrics, was used as a proxy
of net primary productivity (NPP-proxy) (Fig. 2). Goward et al.
(1985) showed that NPP, which corresponds to the carbon stored
by plants from photosynthesis minus what is used for respiration,
is well represented by vegetation indices, because the photosynthetic capacity of each vegetation type, measured by the vegetation
indices, is closely related to the ability of the plant to use water,
light, and nutrients. Indeed, satellite based vegetation indices have
been widely used as proxies for NPP (Running et al., 1988; Ruimy
et al., 1994; Potter et al., 2007; Rosa and Sano, 2013).
In this study we used the ratio of the MODIS-derived ET and PET
products as a dimensionless indicator of plant water use (referred
here as ETn). ETn describes that fraction of the total potential

atmospheric demand (PET) that was provided by the plants at

any location and time (Fig. 2). Thus, any differences in the ETn ratio
in space are more a function of the plant water availability and use
differences (e.g. total amount of soil moisture and vegetation) than
the regional differences in net radiation (PET).
The NPP-proxy and ETn were averaged for the 12 growing seasons for each of the 35 samples. In addition to the sample-based
analyses, we estimated the contribution of each land cover type
to the Cerrado-wide carbon and water budgets (Fig. 2). The mean
sample-based NPP-proxy and ETn over the 12 seasons (2000
2012) were multiplied by the area of each land cover type throughout the Cerrado, determined from the 2002 PROBIO land cover map
(Conservation and Sustainable Use of Brazilian Biological Diversity
Project) (Sano et al., 2010). From this, we assessed the relative state
of the carbon and water balances, as a function of the major land
cover types in the entire Cerrado. We also estimated the total biomass and evaporative flux (in Gt of C and H2O) from each land
cover type throughout the Cerrado.
Finally, in order to better understand the potential impact of
historical and future deforestation on the carbon and water fluxes
of the Cerrado, two hypothetical land use scenarios were considered: a scenario with no pasture and cropland, i.e. only natural vegetation throughout the Cerrado (hypothetical unconverted Cerrado),
and a scenario based on the Ferreira et al. (2012) modeled Cerrado
deforestation for the year 2050 (2050). For the hypothetical unconverted Cerrado scenario, we replaced the pasture and cropland
areas with natural vegetation based on the respective proportion
of each Cerrado land-cover type in 2002. For the 2050 scenario,
we assigned the modeled converted area in 2050 to be either pasture or crop according to the proximity to the respective (2002)

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

71

Fig. 4. Average EVI smoothed temporal profiles (and respective standard deviations) for the major Cerrado land cover classes (time interval encompasses 12 growing seasons,
from 2000 to 2012).

72

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

PROBIO land-use classes. To these hypothetical land-cover distributions, we applied the respective carbon (kg of above-ground biomass) from Baccini et al. (2012) pan-tropical biomass map and
median ET values from MOD16A2 (for all 12 growing seasons).
The main steps of this study, concerning data organization, processing and analysis are depicted in Fig. 2.
3. Results and Discussion
3.1. Phenological dates
A total of 12 growing seasons were identified in the time series,
with the respective lengths comprising the time interval between
the SOS and EOS (Table 1). For example, for the first year of observations the growing season was defined to begin (SOS) on 29
September and end (EOS) on 10 June, for a total growing season
length of 8 months and 11 days. All 12 growing seasons had similar
SOS (September to October) and EOS (May to June) dates, with
growing season lengths varying from 7 months and 25 days to
8 months and 26 days.
A marked spatial variability in mean SOS, EOS, and length of the
growing season is observed in the Cerrado biome (Fig. 3). While, for
most of the Cerrado area, the SOS occurs around SeptemberOctober and OctoberNovember, an earlier SOS (July and August) is
observed in its southern part (south of 20S) and in its extreme

Fig. 5. Distribution of the mean monthly precipitation and normalized ET and EVI
profiles for the entire Cerrado biome.

northern and western portions (as earlier as April and May). The
length of the growing season, in spite of this great variation, lasted,
in general, from seven to nine months. In the extreme western portion of the Cerrado, the length of the growing season was much
shorter, in part a result of errors in the detection of the SOS and
EOS, due to poor quality pixels (related to residual cloud and aerosol contamination) that hinder the detection of the SOS and EOS.
The end of the season showed little variation, with most of the Cerrado having the end of the growing season in June-July and July
August.
3.2. EVI and ETn seasonal profiles
The seasonal EVI profiles of all vegetation types showed one
peak per growing season (Fig. 4), with the exception of the soy,
which showed two EVI peaks. In years of favorable precipitation,
a secondary crop (e.g. corn) is planted late in the wet season after
the soy harvest. The soy samples had the highest EVI amplitude
(0.2 to 0.9), followed by the deciduous forest (0.2 to 0.7), pasture
(0.2 to 0.55), and the Cerrado vegetation types (0.2 to 0.5)
(Fig. 4). These ranges are in agreement with a study done by
Galford et al. (2008) that showed higher EVI values for the Cerrado
woodland (a mean of 0.6), compared to pastures (mean of 0.5) and
EVI values exceeding 0.8 for cropland.
As expected, the average seasonal behavior of the EVI and ETn
of the dominant Cerrado land-cover classes were consistent with
the precipitation dynamics of the Cerrado biome (Fig. 5), with both
EVI and ETn mean values peaking at the maximum of the wet season (January-March), and decreasing to values of about their wet
season peak by late in the dry season (AugustSeptember).
The green-up for all Cerrado vegetation types started in midSeptember to the end of October, at the onset of the rainy season,
with EVI reaching peak values from December through January.
ETn lagged EVI by about one month, peaking from January through
March (Fig. 6a, b). The green-up was particularly abrupt for the soy
and deciduous forest samples, due to human management and
plant physiology that respond rapidly to precipitation. The seasonal behavior of natural and anthropic vegetation types differed
in their response to senescence. All Cerrado vegetation types
showed a gradual senescence, starting as early as March or April.
The gradual senescence indicated by the low slope of the right
derivative of EVI can be related to the different responses of the
Cerrado species to water stress, some maintaining transpiration
in the beginning of the dry season due to access to deep soil water

Fig. 6. Sample-based mean 16-day EVI (a) and mean monthly normalized ET (b) seasonal profiles for the major Cerrado land-cover classes.

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

(Garcia-Monteil et al., 2008), others reducing transpiration, and a

few transpiring during night (Fig. 6a,b).
The lowest and highest EVI and ETn values for the Cerrado vegetation types followed the vegetation gradient from predominant
ground cover to tree cover (Fig. 6). In the wet season, lowest EVI
and ETn values were found over the Cerrado grassland and Cerrado
shrubland (with predominance of grassy and herbaceous species),
followed by the Cerrado stricto sensu (with scattered low trees and
shrubs) and higher EVI and ETn estimations associated with the
higher tree density Cerrado woodland.
The Cerrado grassland consists of a mix of species of natural
grass with no tall woody plants, while the Cerrado shrubland is
composed mainly of herbaceous vegetation and shrubs (with few
low trees that have more ground cover biomass). The EVI (0.15
0.20 to 0.350.40) and ETn (0.20 to 0.40) of these vegetation types
had similar responses and the least variation from the dry to the
wet season (Fig. 6). From May to September, in the peak of the
dry season, most herbaceous vegetation types greatly reduce photosynthesis and transpiration. For the herbaceous species, with
average root depth of only 10 to 15 cm, the leaves die out, only
re-sprouting with the start of rainfall, when the air and soil
becomes moist again.
As the percentage of tree cover increases, there is a decrease in
water stress by the more woody vegetation types. The Cerrado
stricto sensu is a woody type of vegetation with total woody plant
cover of about 1030%, with closed or semi-opened low arboreal
forms (< 7 m) scattered with shrubs (Goodland, 1971). The seasonality response of this vegetation was similar to the herbaceous vegetation types (grassland and shrubland), where the EVI and ETn
start their ascending curves in October, with the first rainfall,
reaching its maximum value around January. Nevertheless, the
EVI and ETn responses, compared to those from the herbaceous
forms, were systematically higher year-around. In December, for
example, the ETn was 20% higher than the ETn of the herbaceous
vegetation types. Similar to the herbaceous vegetation, the Cerrado
stricto sensu had a gradual senescence, starting in March and reaching its lowest ETn and EVI values around June.
The Cerrado woodland, with medium-tall arboreal forms
(height > 9 m) and about 3060% tree cover (Goodland, 1971),
showed, relative to the other vegetation types, consistently higher
EVI and ETn values and very gradual senescence and green-up
rates (Fig. 6). The trees of this physiognomy have a greater proportion of leaves functioning in the dry season due to their roots that
reach up to 8 m of depth, enabling them to access deep soil water
(Palhares et al., 2010; Garcia-Montiel et al., 2008; Franco et al.,
2005). In the beginning of the dry season, the EVI and ETn
responses were slightly reduced, because of decreased soil water
availability. The reduction of the normalized ET in August can be
attributed to the long period without rainfall (more than 3 months)
(Fig. 6).
The deciduous forest was marked by a highly seasonal behavior,
with EVI varying from 0.2 and ETn 0.2 in the dry season to EVI 0.5
and ETn 0.9 in the wet season. Due to their preferred locations in
regions of fractured limestone (Felfili et al., 2007; Pereira et al.,
2011), the trees in deciduous forests have access to water from
the first rains, but the soils quickly become dry at the end of the
rainy season. Thus, green-up and senescence rates, as indicated
by the EVI and ETn responses, are abrupt due to the fast drainage
of water. During the wet season, the canopy cover is 50% to 70%
(Nascimento et al., 2007) and EVI and ETn values were about 1.4
times higher than those of the Cerrado woodland.
The soy and cultivated pasture seasonal responses are not only
a result of plant physiognomy, but also human management. In
particular, the anthropic land covers differ from the natural vegetation types in their faster senescence, which can be attributed to
the lack of biodiversity associated with monocultures and rapid

73

harvest after maturity (Fig. 6). The green-up of the soy was distinct
from the other vegetation types, starting in November after seeding, which generally occurs in October.
The cultivated pastures, although closely following the trends of
the herbaceous vegetation types (Cerrado grassland and Cerrado
shrubland), showed higher rates of photosynthesis and evapotranspiration and higher productivity during the wet season (Fig. 6).
Their higher productivity results from good management practices,
such as fertilization and rotational grazing, which increases its productivity relative to native grasses (Lascano, 1991). The higher
evapotranspiration rate is, in part, a consequence of the deep root
system of the Brachiaria brizantha pasture, which allows it to access
water at greater depths, contributing to higher ET compared to
native grasses (Santos et al., 2004).
The faster green-up rate of pastures is consistent with the study
of Meirelles et al. (2011) that showed that the Brachiaria brizantha
pastures quickly respond to the onset of the rainy season. It is
interesting to observe that the senescence of pastures, although
more abrupt, compared to the natural vegetation, tends to occur
later in May, which suggests, in agreement with the study of
Guenni et al. (2002), their potential for greater resistance to
drought (Fig. 6). Nevertheless, the fast decrease in the EVI and
ETn values, once senescence starts, can be attributed to the strategy of herbs and grasses to react to water stress by rapidly closing
the stomata and drying out their above-ground biomass (Rachid,
1947; Sack and Frole, 2006).
3.3. Phenological Parameters
The soy and deciduous forest are highly seasonal with consistently the highest green-up (left derivative) and senescence (right
derivative) rates, followed by pastures and the Cerrado vegetation
types (Fig. 7a and b). The mean green-up rate of soy (0.163) was
about three times greater than for cultivated pastures (0.053)
and natural Cerrado vegetation (0.040), and two times greater than
the deciduous forest (0.080). The senescence of soy varied from
year to year, with some years having up to two times lower senescence rate than the previous year (Fig. 7b). The high variability is a
direct result of the double-cropping system, in which, in some
years, two crops are not planted and therefore the calculated
senescence rate is very low (it starts early). In double cropping
years the rate is much higher because the corn is green late in
the season and EVI rapidly drops at harvest time (Fig. 7).
The green-up of the Cerrado vegetation types varied according
to the gradient of tree density to ground cover with lower to higher
values for the Cerrado woodland, Cerrado stricto sensu, Cerrado
grassland and shrubland (Fig. 7a). This was consistent with the
EVI and ETn temporal profiles, where the Cerrado grassland and
shrubland had higher green-up rates. The senescence rates were
very similar, with the Cerrado woodland having the lowest senescence rate and lowest temporal variation of the vegetation types
analyzed (Fig. 7b).
The NPP-proxy (small integral of the EVI over the growing season) followed the green-up patterns. The crop NPP-proxy was
highest, with significant variation from year to year (accumulated
EVI 5 to 7) (Fig. 7c). The deciduous forest had the least variation
(accumulated EVI 4 to 5) (Fig. 7c). Pasture was about two times less
productive than soy and showed small interannual variations
(accumulated EVI 3 to 4) (Fig. 6c). Pasture was consistently more
productive (about 1.3 times higher) than the Cerrado vegetation
types, except for two seasons, where the Cerrado shrubland had
higher NPP-proxy (Fig. 7c).
Among the Cerrado vegetation types, the low NPP-proxy of the
Cerrado woodland confirms less and slower accumulation of biomass during one growing season and higher total productivity
(accumulated EVI from past and current growing seasons). This

74

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

Fig. 7. Sample-based green-up (left derivative) (a), senescence (right derivative) (b) and net primary productivity (NPP small integral) for the major Cerrado land-cover
types along 12 growing seasons.

physiognomy, as well as the deciduous forest, had the least variation from year to year, in agreement with the study of Ma et al.
(2013) that showed increased temporal variability in phenology
in areas with herbaceous biomass. These results suggest that the
savanna response to climate variability is a function of the ratio
between woody and herbaceous biomass, where woody vegetation
promotes more stability and resilience to possible future climate
change.
3.4. Regional Analyzes
The mean normalized NPP-proxy values were lowest in the
extreme western, eastern, and northern portions of the Cerrado
(in the states of Mato Grosso, Bahia, and Maranho) (Fig. 8a). Remnants of Cerrado stricto sensu, Cerrado woodland, and forested vegetation vegetation types, associated to higher total NPP, but lower
seasonal NPP, predominate in these areas. By contrast, the high values of NPP-proxy were mostly located over cropland, with the
exception of a fragment composed mainly of deciduous forest in
the eastern portion of the state of Gois (Fig. 8a). These areas with
high NPP-proxy also have high ETn values, indicating that agriculture accumulates more carbon in the growing season and also
transfers more water to the atmosphere via evapotranspiration
(Fig. 8b). Intermediate normalized NPP-proxy and ETn values corresponded to pasture areas, shown in shades of yellow and green,
respectively.
The mean NPP-proxy and normalized ET values were associated
with the area of each land cover and land use class and summed to
understand the role of each vegetation type in the biome. In 2002,
the natural Cerrado vegetation covered about 48% of the Cerrado
biome and had the greatest contribution to NPP (76%) and ETn
(61%) (Fig. 9a, b). The pasture and cropland areas, covering 49%
of the biome, contributed to 24% of NPP-proxy and to 39% of ETn
(Fig. 9a, b).

The Cerrado stricto sensu had the greatest ET contribution (32%)

of any single vegetation type. Pastures, because of the very large
area they cover, played a major role in the water cycle (Fig. 9b)
accounting for 28% of total ETn. The greatest contributions to
NPP came from the Cerrado stricto sensu (36%) and the shrub Cerrado (26%) (Fig. 9a), with pastures accounting for about 15%.
The mass of water and carbon represented in the ETn and NPP,
according to each land-cover type, were estimated from the
respective median MOD16A2 ET values (in millimeters of water)
and median carbon values (in tons per hectare) from Baccini
et al. (2012) pan-tropical biomass map. The median growing season ET varied from a minimum of 713 mm/yr for pasture to a maximum of 755 mm/yr for cropland. The biomass varied from a
minimum of 145 tons per hectare for deciduous forest to a maximum of 169 tons per hectare for the Cerrado woodland. The pasture biomass value was 168 tons per hectare and the soy 161
tons per hectare.
The role of each vegetation type in determining the total biome
carbon and water is a function of the total area occupied and the
differing stocks and flux rates. Anthropic environments provided
44% of the ET from 39% of the land area and the Cerrado vegetation
types 56% of the ET from 48% of the land area (Table 2). Pasture and
cropland contributed 576 Gt of water via evapotranspiration during the growing season, 420 Gt from pasture and 156 Gt from crops
(Table 2). The growing season evapotranspiration for all the native
Cerrado vegetation types was 760 Gt of water with greatest contributions from the Cerrado stricto sensu (332 Gt) and shrub Cerrado
(278 Gt) (Table 2).
There were 12 Gt of carbon in pasture and crop, 9 Gt for pasture
and 3 Gt for crops, which was 43% of the total Cerrado biomass (for
39% of the area) (Table 2). The areas of natural vegetation had 15 Gt
of carbon with the greatest contributions from the Cerrado stricto
sensu and shrub Cerrado, with 7 and 5 Gt of carbon, respectively,
accounting for 57% of the total biomass in the Cerrado (Table 2).

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

75

Fig. 8. Spatial distribution (per-pixel basis) of mean (12 seasons) normalized NPP-proxy (a) and normalized ETn (b) values (according to the major Cerrado land-cover and
land-use classes).

Fig. 9. The relative contributions of each land use class to total NPP-proxy (a) and ETn (b).

3.5. Scenarios
A scenario representing the Cerrado without anthropic environments (hypothetical unconverted Cerrado) was developed in order to
investigate the scale of the impact of historical deforestation on ET
and above ground biomass. There was 29 Gt of carbon from the
above-ground biomass and 1357 Gt of water transpired during
the growing season from this idealized landscape of only natural
Cerrado vegetation types. Those values were 21 Gt of water and
1 Gt of carbon greater than the values calculated for the mixed

landscape in 2002 (Tables 3 and 4). The lowest contributions of

the ET fluxes and above-ground biomass were for the deciduous
forest with 39 Gt of water and 0.7 Gt of carbon and the Cerrado
grassland with 102 Gt of water and 2 Gt of carbon (Tables 3 and
4). The greatest contributions were for the Cerrado stricto sensu
with 593 Gt of water and 13 Gt of carbon or 44% of the total ET
fluxes and 46% of the total above-ground biomass in the hypothetical unconverted Cerrado scenario, and shrub Cerrado with 497 Gt of
water and 10 Gt of carbon or 36% of the total ET fluxes and 34% of
the total biomass.

76

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

Table 2
ET fluxes and biomass in gigatons of water and carbon for each land cover type
according to the 2002 PROBIO land cover map.
Land cover and land use

ET 2002 (Gt)

Biomass 2002 (Gt)

Pasture
Cropland
Cerrado grassland
Cerrado shrubland
Cerrado Stricto Sensu
Cerrado woodland
Deciduous forest
Cerrado vegetation
Pasture + cropland
Total

420
156
57
278
332
71
22
760
576
1336

9
3
1
5
7
2
0.4
15
12
28

Table 3
Normalized ET fluxes in gigatons of water according to the 2002 landscapes (PROBIO),
a hypothetical unconverted Cerrado (pre-occupation) scenario (i.e. only natural
vegetation ET fluxes) and modeled conversions by 2050.
Land cover and land use

Hypothetical unconverted Cerrado

2002

2050

Pasture
Cropland
Cerrado grassland
Cerrado shrubland
Cerrado Stricto Sensu
Cerrado woodland
Deciduous forest
Cerrado vegetation
Pasture + cropland
Total ET fluxes

102
497
593
126
39
1357

1357

420
156
57
278
332
71
22
760
576
1336

787
194
23
145
151
25
22
366
981
1346

pasture and cropland areas occupy 54% and 13% (total 67%) of
the Cerrado biome, the pasture and cropland growing season ET
would be 981 Gt of water and the biomass would be 23 Gt of carbon. This is equivalent to 73% of the total ET flux from the Cerrado
and 75% of the total carbon biomass in this scenario. The native
Cerrado vegetation types in this scenario occupy 26% of the biome
area, and contribute 366 Gt of water and 8 Gt of carbon or 27% and
25% of the total ET and biomass (Tables 3 and 4).
In the 2050 scenario the biome total ET flux increased by 10 Gt
and biomass by 3 Gt of carbon compared to the 2002 estimate
(Table 3 and 4). The higher contribution of pasture and cropland
to ET fluxes and carbon in the 2050 scenario slightly increased
the total ET fluxes and carbon to 1346 Gt of water and 31 Gt of carbon (compared to 1336 Gt of water and 28 Gt of carbon in 2002),
which can be attributed to the higher ET and carbon median values
of pastures compared to the natural vegetation (for which the
greatest contribution was from the Cerrado stricto sensu). The ET
flux from natural Cerrado vegetation was 394 Gt less than in
2002 and 991 Gt less than in the hypothetical unconverted Cerrado
scenario, while the carbon was 8 Gt less than in 2002 and 21 Gt
less than in the hypothetical unconverted Cerrado scenario. The
sum of the pasture and cropland ET flux increased by 405 Gt relative to 2002 and the carbon by 11 Gt of carbon. This suggests that if
deforestation continues at or near current rates it will likely result
in further significant alteration of the carbon and water balance
with potentially negative implications for the climate system and
ecosystem services (Oliveira et al., 2013; Stickler et al., 2013; Coe
et al., 2013).

4. Conclusion
Table 4
EVI-based NPP according to the 2002 landscapes (PROBIO), a hypothetical unconverted Cerrado (pre-occupation) scenario (i.e. only natural vegetation C fluxes) and
modeled conversions by 2050.
Land cover and land use

Hypothetical unconverted Cerrado

2002

2050

Pasture
Cropland
Cerrado grassland
Cerrado shrubland
Cerrado Stricto Sensu
Cerrado woodland
Deciduous forest
Cerrado vegetation
Pasture + cropland
Total carbon

2
10
13
3
0.7
29

29

9
3
1
5
7
2
0.4
15
12
28

19
4
0.4
3
3
1
0.4
8
23
31

It is plausible that the Cerrado woodland would have occupied

most of the biome in the past 10,000 years before human settlements (Rizzini and Heringer, 1962; Eyre, 1963; Richards, 1969). If
so, our reconstructed landscape, which is dominated by Cerrado
stricto sensu, probably underestimates the pre-conversion evapotranspiration flux and carbon in biomass. If Cerrado woodland
occupied 50% of the Cerrado biome, the ET and carbon fluxes would
be 395 Gt of water and 8 Gt of carbon, about 269 Gt of water and
6 Gt of carbon greater than the hypothetical unconverted Cerrado
scenario.
Deforestation continues in the Cerrado at a rapid rate (Rocha
et al., 2011) and it has been identified as an important contributor
to Brazils greenhouse gas emission budget (Bustamante et al.,
2012) and, through ET reductions, a potential contributor to
altered rainfall and river discharge regionally (Costa and Pires,
2009; Coe et al., 2009, 2011, 2013). Based on a modeled deforestation scenario for 2050 (Ferreira et al., 2012), according to which

The MODIS EVI and normalized ET temporal profiles were

instrumental for understanding the seasonality of the Cerrado
biome and the diverse photosynthetic functioning of natural and
anthropic landscapes.
The EVI analysis identified one strong growing season per year
for all Cerrado vegetation types, with the exception of soy associated with a secondary crop. All natural Cerrado vegetation types
had earlier and more gradual senescence when compared to
anthropic pasture and soy/corn vegetation. The soy and deciduous
forest were highly seasonal with consistently the highest green-up
rate, senescence rate, and net primary productivity, followed by
pastures and the Cerrado vegetation types.
Using as reference a set of 35 samples, a regional analysis considering the area occupied by each vegetation type showed that
anthropic landscapes had 22% less ET and 52% less NPP-proxy than
natural landscapes. Our calculations of the total water and carbon
mass suggest that pasture and cropland together contributed 12 Gt
of carbon to the above-ground biomass pool and a 576 Gt of ET flux
during the growing season. Native Cerrado vegetation types combined contributed 15 Gt of carbon to the above-ground biomass
pool and a 760 Gt flux of water to the atmosphere.
Our analyses of the impacts of historical and potential future
changes suggest that human alteration of the landscape affects
the carbon and water balance. We found that deforestation up to
the year 2002 has contributed to a decline in water vapor flux to
the atmosphere from the Cerrado biome of 21 Gt (from 1357 Gt
prior to deforestation) and above-ground biomass of 1 Gt (from
29 Gt prior to deforestation). Analysis of the impacts of a land
use scenario for the year 2050 (based on modern deforestation
rates) suggests that conditions would be reversed, i.e. a 10 Gt
increase in ET and a 2 Gt increase in aboveground biomass compared to 2002. Pasture and cropland growing season ET in 2050
increases by 405 Gt and carbon by 11 Gt compared to 2002. The

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

2050 ET flux from native Cerrado vegetation types is reduced by

394 Gt and carbon biomass by 8 Gt in 2050 compared to 2002.
These scenarios indicate that important changes in biomass and
water have already occurred in the Cerrado biome. Further, our
analyses of potential future conditions shows that given the high
global demand for agricultural products, the very low density of
protected areas in the Cerrado, and the significant area of remaining native vegetation on private land that may be legally deforested, future changes to the water and carbon balances are likely
to be significant. An increase in the contribution of pastures and
cropland to total ET and NPP relative to natural Cerrado vegetation
could alter the degree of surface and atmosphere coupling, potentially impacting the climate. On the other hand, as our results also
suggest, the higher net primary productivity of pastures compared
to the natural Cerrado vegetation suggests that the recovery of
degraded pastures into well managed pastures can lead to significant amounts of carbon assimilation and water transfer to the
atmosphere with potential positive impacts to the local and global
climate.
Finally, and worth of mentioning, the methods and approaches
on which this study relied are simple and easily replicable to other
landscapes. To this end, most of the time-series analysis tools and
datasets (for the entire Brazil we used, are freely and easily accessible thorough our pasture data gateway (pastagem.org).
Acknowledgements
The authors thank the Gois State Foundation for Research Support (FAPEG/2012/00766130154) and the Gordon and Betty Moore
Foundation for funding this research. Laerte Guimares Ferreira
and Arielle Arantes thank CNPq for individual research grants.
Michael T. Coe acknowledges the support of the NASA Terrestrial
Ecology program (grant NNX12AK11G).
References
Arago, L.E., Malhi, Y., Roman-Cuesta, R.M., Saatchi, S., Anderson, L.O., Shimabukuro,
Y.E., 2007. Spatial patterns and fire response of recent Amazonian droughts.
Geophys. Res. Lett. 34 (7).
Baccini, A., Goetz, S.J., Walker, W.S., Laporte, N.T., Sun, M., Menashe, D.S., Hackler, J.,
Beck, P.S.A., Dubayah, R., Friedl, M.A., Samanta, S., Houghton, R.A., 2012.
Estimated carbon dioxide emissions from tropical deforestation improved by
carbon density maps. Nat. Clim. Change 2.
Bandaru, V., West, T.O., Ricciuto, D.M., Izaurralde, R.C., 2013. Estimating crop net
primary production using national inventory data and MODIS-derived
parameters. ISPRS J. Photogram. Remote Sens. 80, 6172.
Batlle-Bayer, L., Batjes, N.H., Bindraban, P.S., 2010. Changes in organic carbon stocks
upon land use conversion in the Brazilian Cerrado: a review. Agric. Ecosyst.
Environ. 137 (1), 4758.
Bonan, G.B., 2008. Forests and climate change: forcings, feedbacks, and the climate
benefits of forests. Science 320 (1444).
Brando, A.S.P., Rezende, G.C., Marques, R.W.C., 2006. Crescimento Agrcola no
Perodo 1999/2004: a Exploso da Soja e da Pecuria Bovina e Seu Impacto
Sobre o Meio Ambiente. Economia Aplicada 10 (2), 249266.
Braz, S.P., Urquiaga, S., Alves, B.J.R., Jantalia, C.P., Guimares, A.P., dos Santos, C.A.,
dos Santos, S.C., Pinheiro, E.F.M., Boddey, R.M., 2013. Soil carbon stocks under
productive and degraded Brachiaria Pastures in the Brazilian Cerrado. Soil Sci.
Soc. Am. J. 77 (2), 914928.
Brossard, M., Barcellos, A.O., 2005. Converso do Cerrado em Pastagens Cultivadas e
Funcionamento de Latossolos. Cadernos de Cincia & Tecnologia 22 (1), 153
169.
Buol, S.W., 2009. Soils and agriculture in central-west and north Brazil. Scientia
Agricola 66 (5), 697707.
Bustamante, M.M.C., Nobre, C.A., Smeraldi, R., Aguiar, A.P.D., Barioni, L.G., Ferreira, L.
G., Longo, K., May, P., Pinto, A.S., Ometto, J.P.H.B., 2012. Estimating greenhouse
gas emissions from cattle raising in Brazil. Clim. Change 115, 559577.
Camargo, A., 1963. In: Ferri, M.G. (Ed.), Clima do cerrado. Simpsio sobre o Cerrado.
EDUSP, So Paulo, pp. 7595.
Cerri, C.C., Maia, S.M.F., Galdos, M.V., Cerri, C.E.P., Feigl, B.J., Bernoux, M., 2009.
Brazilian greenhouse gas emissions: the importance of agriculture and
livestock. Scientia Agricola 66 (6), 831843.
Coe, M.T., Costa, M.H., Soares-Filho, B.S., 2009. The Influence of historical and
potential future deforestation on the stream flow of the Amazon River land
surface processes and atmospheric feedbacks. J. Hydrol. 369 (2), 165174.

77

Coe, M.T., Latrubesse, E.M., Ferreira, M.E., Amsler, M.L., 2011. The effects of
deforestation and climate variability on the streamflow of the Araguaia River,
Brazil. Biogeochemistry 105, 119131.
Coe, M.T., Marthews, T.R., Costa, M.H., Galbraith, D., Greenglass, N., Imbuzeiro, H.M.
A., Levine, N.M., Malhi, Y., Moorcroft, P., Muza, M.N., Powell, T.L., Saleska, S.,
Solorzano, L.A., Wang, J., 2013. Deforestation and climate feedbacks threaten
the ecological integrity of south-southeastern Amazonia. Philos. Trans. Remote
Sens. Soc. 368.
Costa, M.H., Botta, A., Cardille, J.A., 2003. Effects of large-scale changes in land cover
on the discharge of the Tocantins River, Southeastern Amazonia. J. Hydrol. 283,
206217.
Costa, M.H., Pires, G.F., 2009. Effects of Amazon and Central Brazil deforestation
scenarios on the duration of the dry season in the arc of deforestation. Int. J.
Climatol. 30 (13), 19701979.
Couto, M.S.D.S., Ferreira, L.G., Hall, B.R., Silva, G.J.P., Garcia, F.N., 2010. Identificao
de reas Prioritrias para Conservao da Biodiversidade e Paisagens no Estado
de Gois: Mtodos e Cenrios no Contexto da Bacia Hidrogrfica. Revista
Brasileira de Cartografia (62).
Cunha, A.S. (Ed.), 1994. Avaliao da sustentabilidade da agricultura nos cerrados.
Relatrios de pesquisa, Braslia. IPEA.
DAlmeida, C., Vrsmarty, C.J., Hurtt, G.C., Marengo, J.A., Dingman, S.L., Keim, B.D.,
2007. The effects of deforestation on the hydrological cycle in Amazonia: a
review on scale and resolution. Int. J. Climatol. 27 (5), 633647.
Eiten, G., 1972. The Cerrado Vegetation of Brazil. The Botanical Review, vol. 38.
Springer, pp. 201341, n. 2.
Eyre, S.R., 1963. Vegetation and Soils. Edward Arnold, London.
Fearnside, P.M., Righi, C.A., Graa, P.M.L.A., Keizer, E.W.H., Cerri, C.C., Nogueira, E.M.,
Barbosa, R.I., 2009. Biomassa and greenhouse-gas emissions from land-use
change in Brazils Amazonian arc of deforestation: the states of Mato Grosso
and Rondnia. For. Ecol. Manage. 258, 19681978.
Felfili, J.M., Nascimento, A.R.T., Fagg, C.W., Meirelles, E.M., 2007. Floristic
composition and community structure of a seasonally deciduous forest on
limestone outcrops in Central Brazil. Revista Brasileira Botanica 30 (4), 611621.
Ferreira, L.G., Yoshioka, H., Huete, A., Sano, E.E., 2003. Seasonal landscape and
spectral vegetation index dynamics in the Brazilian Cerrado: an analysis within
the Large-Scale-Biosphere-Atmosphere Experiment in Amaznia (LBA). Remote
Sens. Environ. 87, 534550.
Ferreira, L.G., Huete, A.R., 2004. Assessing the seasonal dynamics of the Brazilian
Cerrado vegetation through the use of spectral vegetation indices. Int. J. Remote
Sens. 25 (10), 18371860.
Ferreira, L.G., Asner, G.P., Knapp, D.E., Davidson, E.A., Coe, M.T., Bustamante, M.M.C.,
Oliveira, E.L., 2011. Equivalent water thickness in savanna ecosystems: MODIS
estimates based on ground EO-1 Hyperion data. Int. J. Remote Sens. 32 (22).
Ferreira, M.E., Ferreira, L.G., Miziara, F., Soares, B.S., 2012. Modeling landscape
dynamics in the central Brazilian savana biome: future scenarios and
perspectives for conservation. J. Land Use Sci. 8 (4), 403421.
Franco, A.C., Bustamante, M., Caldas, L.S., Goldstein, G., Meinzer, F.C., Kozovitz, A.R.,
Rundel, P., Coradin, V.T., 2005. Leaf functional trait of Neotropical savanna trees
in relation to seasonal water deficit. Trees 19, 326335.
Garcia, F.N., Ferreira, L.G., Leite, J.F., 2011. reas Protegidas no Bioma Cerrado:
fragmentos vegetacionais sob forte presso. In: Simpsio Brasileiro de
Sensoriamento Remoto, Curitiba, PR, 2011.
Galford, G.L., Mustard, J.F., Melillo, J., Gendrin, A., Cerri, C.C., Cerri, C.E.P., 2008.
Wavelet analysis of MODIS time series to detect expansion and intensification
of row-crop agriculture in Brazil. Remote Sens. Environ. 112, 576587.
Garcia-Montiel, D.C., Coe, M.T., Cruz, M.P., Ferreira, J.N., Silva, E.M., 2008. Estimating
seasonal changes in volumetric soil water content at landscape scales in a
savanna ecosystem using two-dimensional resistivity profiling. Earth Interact.
12 (2).
Giambelluca, T.W., Scholz, F.G., Bucci, S.J., Meinzer, F.C., Goldstein, G., Hoffmann, W.
A., Franco, A.C., Buchert, M.P., 2009. Evapotranspiration and energy balance of
Brazilian savannas with contrasting tree density. Agric. For. Meteorol. 149,
13651376.
Goodland, R.A., 1971. Physiognomic analysis of the Cerrado vegetation of Central
Brasil. J. Ecol. 59 (2), 411419.
Goward, S.N., Tucker, C.J., Dye, D.G., 1985. North American vegetation patterns
observed with the NOAA-7-advanced very high resolution radiometer.
Vegetation 64, 214.
Guenni, O., Marn, D., Baruch, Z., 2002. Responses to drought of five Brachiaria
species. I. Biomass production, leaf growth, root distribution, water use and
forage quality. Plant Soil 243, 229241.
Hayhoe, S., Neill, C., McHorney, R., Porder, S., Lefebvre, P., Coe, M.T., Elsenbeer, H.,
Krusche, A., 2011. Conversion to soy on the Amazonian agricultural frontier
increases streamflow without affecting stormflow dynamics. Glob. Change Biol.
17, 18211833.
Huete, A.R., Saleska, S.R., 2010. Remote sensing of tropical forest phenology: issues
and controversies. In: International Archives of the Photogrammetry, Remote
Sensing and Spatial Information Science, Kyoto, Japan.
Huete, A., Didan, K., Miura, T., Rodriguez, E.P., Gao, X., Ferreira, L.G., 2002. Overview
of the radiometric and biophysical performance of the MODIS vegetation
indices. Remote sens. Environ. 83, 195213.
Houghton, R.A., 2003. Revised estimates of the annual net flux of carbon to the
atmosphere from changes in land use and land management 18502000. Tellus
55, 378390.
Houghton, R.A., 2012. Estimated carbon dioxide emissions from tropical
deforestation improved by carbon density maps. Nature Clim. Change 2.

78

A.E. Arantes et al. / ISPRS Journal of Photogrammetry and Remote Sensing 117 (2016) 6678

Jepson, W., 2005. A dissapearing biome? Reconsidering land-cover change in the

Brazilian savanna. Geograph. J. 171 (2), 99111.
Jnsson, P., Eklundh, L., 2002. Seasonality extraction by function fitting to timeseries of satellite sensor data. IEEE Trans. Geosci. Remote Sens. 40 (8).
Jnsson, P., Eklundh, L., 2004. TIMESAT a program for analyzing time-series of
satellite sensor data. Comput. Geosci. 30, 833845.
Klink, C.A., Machado, R., 2005. Conversation of the Brazilian Cerrado. Conserv. Biol.
19 (3), 707713.
Kummerow, C., Barnes, W., Kozu, T., Shiue, J., Simpson, J., 1998. The Tropical Rainfall
Measuring Mission (TRMM) sensor package. J. Atmos. Ocean. Technol. 15.
Lascano, C.E., 1991. Managing the grazing resource for animal production in
savannas of tropical America. Tropic. Grassl. 25, 6672.
Lathuillire, M.J., Johnson, M.S., Donner, S.D., 2012. Water use by terrestrial
ecosystems: temporal variability in rainforest and agricultural contributions
to evapotranspiration in Mato Grosso, Brazil. Environ. Res. Lett.
Li, L., Guo, Q., Tao, S., Kelly, M., Xu, G., 2015. Lidar with multi-temporal MODIS
provide a means to upscale predictions of forest biomass. ISPRS J. Photogram.
Remote Sens. 102, 198208.
Loarie, S.R., Lobell, D.B., Asner, G.P., Mu, Q., Field, C.B., 2011. Direct impacts on local
climate of sugar-cane expansion in Brazil. Nat. Clim. Change 10 (67).
Ma, X., Huete, A., Yu, Q., Restrepo Coupe, N., Davies, K.P., Broich, M., Ratana, P.,
Beringer, J., Hutley, L.B., Cleverly, J.R., Boulain, N.P., Eamus, D., 2013. Spatial
patterns and temporal dynamics in savanna vegetation phenology across the
North Australian Tropical Transect. Remote Sens. Environ. 139 (1), 97115.
Marcuzzo, F.F.N., Cardoso, M.R.D., Faria, T.G., 2012. Chuvas no Cerrado da Regio
Centro-Oeste do Brasil: anlise histrica e tendncia futura. Atlie Geogrfico 6
(2), 112130.
Meirelles, M.L., Franco, A.C., Farias, S.E.M., Bracho, R., 2011. Evapotranspiration and
plantatmospheric coupling in a Brachiaria brizantha pasture in the Brazilian
savannah region. Grass Forage Sci. 66, 206213.
Miranda, A.C., Miranda, H.S., Lloyd, J., Grace, J., Francey, R.J., Mcintyre, J.A., Meir, P.,
Riggan, P., Lockwood, R., Brass, J., 1997. Fluxes of carbon, water and energy over
Brazilian cerrado: an analysis using eddy covariance and stable isotopes. Plant,
Cell Environ. 20, 315328.
Miranda, S.C., Bustamante, M., Palace, M., Hagen, S., Keller, M., Ferreira, L.G., 2014.
Regional variations in biomass distribution in Brazilian savanna woodland.
Biotropica 46 (2), 125138.
Mu, Q., Zhao, M., Running, S.W., 2011. Improvements to a MODIS global terrestrial
evapotranspiration algorithm. Remote Sens. Environ.
Mueller, C.C., 2003. Expansion and modernization of agriculture in the Cerrado the
case of soybeans in Brazils Center-West. Universidade de Braslia
Departamento de Economia.
Nascimento, A.R.T., Felfili, J.M., Fagg, C.W., 2007. Canopy opennes and LAI estimates
in two seasonally deciduous forests on limestone outcrops in Central Brazil
using hemispherical photographs. Revista rvore 31 (1), 151159.
Oliveira, L.J.C., Costa, M.H., Soares-Filho, B.S., Coe, M.T., 2013. Large-scale expansion
of agriculture in Amazonia may be a no-win scenario. Environ. Res. Lett. 8.
Paiva, A.O., Faria, G.E., 2007. Estoque de carbono do solo sob cerrado sensu stricto no
Distrito Federal, Brasil. Revista Trpica 1 (1), 59.
Palhares, D., Franco, A.C., Zaidan, L.B.P., 2010. Respostas fotossintticas de plantas
de cerrado nas estaes seca e chuvosa. Revista Brasileira de Biocincias 8 (2),
213220.
Pereira, B.A.S., Venturoli, F., Carvalho, F.A., 2011. Florestas Estacionais no Cerrado:
Uma Viso Geral. Pesquisa Agropecuria Trop. 41 (3), 446455.
Potter, C., Gross, P., Genovese, V., Smith, M., 2007. Net primary productivity of forest
stands in New Hampshire estimated from Landsat and MODIS satellite data.
Carbon Balance Manage. 2 (9).
Potter, C., Klooster, S., Huete, A., Genovese, V., Bustamante, M., Ferreira, L.G.,
Oliveira, R.C., Zepp, R., 2009. Terrestrial carbono sinks in the Brazilian Amazon
and Cerrado region predicted from MODIS satellite data and ecosystem
modeling. Biogeosciences 6, 937945.
Pimentel, R.M., 2012. Propriedades Fsicas, Carbono e Nitrognio do Solo em
Sistemas Agropecurios. Master in Animal Husbandry, Universidade Federal de
Lavras, Brazil.
Pinto, J.C., Pimentel, R.M., Zinn, Y.L., Chizzotti, F.H., 2014. Soil organic carbon stocks
in a Brazilian Oxisol under different pasture systems. Trop. Grasslands-Forrajes
Tropicales 2 (1), 121123.
Rachid, M., 1947. Transpirao e sistemas subterrneos da vegetao de vero dos
campos cerrado de Emas. Boletim da Faculdade de Filosofia 80 (5), 3769.
Ratana, P., Huete, A.R., 2005. Analysis of Cerrado Vegetation types and conversion in
the MODIS seasonal-temporal domain. Earth Interact. 9 (3).
Reed, B.C., Brown, J.F., Vanderzee, D., 1994. Measuring phenological variability from
satellite imagery. J. Veg. Sci. 5, 703714.
Rezende, G.C., 2002. Ocupao Agrcola e Estrutura Agrria no Cerrado: O Papel do
Preo da Terra, dos Recursos Naturais e da Tecnologia. Embrapa (access in:
14.04.14). <http://www22.sede.embrapa.br/unidades/uc/sge/ocupacao_agraria.
pdf>.
Richards, P.W., 1969. Unpublished Report to R.S./R.G.S Brazil Expedition Committee.
Rizzini, C.T., Heringer, E.P., 1962. Preliminares acerca das formaes vegetais e do
reflorestamento no brasil central. Servio de Informao Agrcola e Ministrio
da Agricultura, Rio de Janeiro.

Rocha, H.R., Freitas, H.C., Rosolem, R., Jurez, R.I.N., Tannus, R.N., Ligo, M.A., Cabral,
O.M.R., Dias, M.A.F.S., 2002. Measurements of CO2 exchange over a woodland
savanna (Cerrado sensu stricto) in southeast Brasil. Biota Neotropica 2 (1).
Rocha, H.R., Manzi, A.O., Cabral, O.M., Miller, S.D., Goulden, M.L., Saleska, S.R., Coupe,
N.R., Wofsy, S.C., Borma, L.S., Artaxo, P., Vourlitis, G., Nogueira, J.S., Cardoso, F.L.,
Nobre, A.D., Kruijt, B., Freitas, H.C., Randow, C., Aguiar, R.G., Maia, J.F., 2009.
Patterns of water and heat flux across a biome gradient from tropical forest to
savanna in Brazil. J. Geophys. Res. 114.
Rocha, G.F., Ferreira Jnior, L.G., Ferreira, N.C., Ferreira, M.E., 2011. Deteco de
desmatamentos no bioma Cerrado entre 2002 e 2009: padres, tendncias e
impactos. Revista Brasileira de Cartografia 63, 341349.
Rosa, R., Sano, E.E., 2013. Determinao da Produtividade Primria Lquida (NPP) de
Pastagens na Bacia do Rio Paranaba, Usando Imagens MODIS. GeoFocus 13,
367395.
Ruhoff, A.L., Arago, L.E., Collischonn, W., Rocha, H.R., Mu, Q., Running, S., 2011.
MOD16: Desafios e limitaes para a estimativa global de evapotranspirao.
In: XV Brazilian Remote Sensing Symposium, Curitiba, Brazil, pp. 5124.
Ruimy, A., Saugier, B., 1994. Methodology for the estimation of terrestrial net
primary production from remotely sensed data. J. Geophys. Res. 99 (3), 5263
5283.
Running, S.W., Nemani, R., 1988. Relating seasonal patterns of the AVHRR
vegetation index to simulated photosynthesis and transpiration of forests in
different climates. Remote Sens. Environ. 24, 347367.
Sack, L., Frole, K., 2006. Leaf structural diversity is related to hydraulic capacity in
tropical rain forest trees. Ecology 87, 483491.
Sano, E.E., Barcellos, A.O., Bezerra, H.S., 2000. Assessing the spatial distribution of
cultivated pastures in the Brazilian savanna. Pasturas Tropicales 22 (3).
Sano, E.E., Rosa, R., Brito, J.L.S., Ferreira, L.G., 2010. Land cover mapping of the
tropical savanna region in Brazil. Environ. Monit. Assess. 166, 113124.
Santos, A.J.B., Silva, G.T.D.A., Miranda, H.S., Miranda, A.C., Lloyd, J., 2003. Effects of
fire on surface carbono, energy and water vapour fluxes over campo sujo
savanna in central Brazil. Funct. Ecol. 17, 711719.
Santos, A.J.B., Quesada, C.A., Silva, G.T., Maia, J.F., Miranda, H.S., Miranda, A.C., Lloyd,
J., 2004. High rates of net ecosystem carbono assimilation by Brachiaria pasture
in the Brazilian Cerrado. Glob. Change Biol. 10, 877885.
Silva, J.E., Resck, D.V.S., Corazza, E.J., Vivaldi, L., 2004. Carbon storage in clayey
Oxisol cultivated pastures in the Cerrado region, Brazil. Agric. Ecosyst.
Environ. 103, 357363.
Silva, E.B., Ferreira, L.G., Rocha, G.F., Couto, M.S.D.S., 2009. Taxas de desmatamento
em Otto bacias do bioma Cerrado obtidas atravs de imagens ndice de
vegetao MODIS. In: XIV Brazilian Remote Sensing Symposium, Natal, Brazil,
pp. 62416248.
Silva, E.B., Ferreira, L.G., Anjos, A.F., Miziara, F., 2013. An spatial distribution of
cultivated pastures in the Brazilidas no bioma Cerrado entre 1970 e 2006.
Revista IDeAs 7 (1), 174209.
Soares-Filho, B., Rajo, R., Macedo, M.N., Carneiro, A., Costa, W., Coe, M.T., Rodrigues,
H., Alencar, A., 2014. Cracking Brazils forest code. Science 344 (6182), 363364.
Solano, R., Didan, K., Jacobson, A., Huete, A.R., 2010. MODIS vegetation indices
(MOD13) user s guide. <http://tbrs.arizona.edu/project/MODIS/MOD13.C5UsersGuide-HTML-v1.00> (accessed on 25.02.11).
Solrzano, A., Pinto, J.R., Felfili, J.M., Hay, J.D.V., 2012. Perfil florstico e estrutural do
componente lenhoso em seis reas de cerrado ao longo do bioma Cerrado. Acta
Botanica Brasilica 26 (2), 328341.
Spracklen, D.V., Arnold, S.R., Taylor, C.M., 2012. Observations of increased tropical
rainfall preceded by air passage over forests. Nature 489 (7415).
Stickler, C.M., Coe, M.T., Costa, M.H., Dias, L.C., Nepstad, D.C., McGrath, D.G.,
Rodrigues, H.O., Soares-Filho, B.S., 2013. The Dependence of hydropower energy
generation on forests in the Amazon Basin at local and regional scales. Proc. Nat.
Acad. Sci. 13.
Tan, B., Morisette, J.T., Wolfe, E., Gao, F., Ederer, G.A., Nightingale, J., Pedelty, J.A.,
2008. Vegetation phenology metrics derived from temporally smoothed and
gap-filled MODIS data. In: International Geoscience and Remote Sensing
Society, 2008.
Vendrame, P.R.S., Brito, O.R., Guimares, M.F., Martins, E.S., Becquer, T., 2010.
Fertility and Acidity Status of Latossols (oxisols) under pasture in the Brazilian
Cerrado. An Acadmica Brasileira de Cincia 82 (4).
Wagle, P., Xiao, X., Suyker, A.E., 2015. Estimation and analysis of gross primary
production of soybean under various management practices and drought
conditions. ISPRS J. Photogram. Remote Sens. 99, 7083.
Wu, C., Gonsamo, A., Zhang, F., Chen, J.M., 2014. The potential of the greenness and
radiation (GR) model to interpret 8-day gross primary production of vegetation.
ISPRS J. Photogram. Remote Sens. 88, 6979.
Zhang, J., Feng, L., Yao, F., 2014. Improved maize cultivated area estimation over a
large scale combining MODIS-EVI time series and crop phenological
information. ISPRS J. Photogram. Rem. Sens. 94, 102113.
Zhang, X., Friedl, M.A., Schaaf, C.B., Strahler, A.H., Hodges, J.C.F., Gao, F., Reed, B.C.,
Huete, A., 2003. Monitoring vegetation phenology using MODIS. Remote Sens.
Environ. 84, 471475.
Zhang, X., Friedl, M.A., Tan, B., Goldberg, M.D., Yu, Y., 2012. Long-term detection of
global vegetation phenology from satellite instruments. Phenol. Clim. Change.