nature sustainability

Article https://doi.org/10.1038/s41893-024-01475-8

Intensification of climate change impacts

on agriculture in the Cerrado due to
deforestation

Received: 26 February 2024 Argemiro Teixeira Leite-Filho 1 , Britaldo Silveira Soares-Filho1,

Ubirajara Oliveira 1 & Michael Coe2
Accepted: 29 October 2024

Published online: xx xx xxxx

Deforestation not only contributes to global warming but also induces
Check for updates changes in the regional climates that impact agriculture. Here, we analyse
the effects of deforestation-induced climate change on the soy–maize
double cropping of the Cerrado biome. Since the 1980s, there has been
an average delay of 36 days in the start of the agricultural rainy season,
a 36.7% reduction in total rainfall over the period and a 1.5 °C rise in the
temperature in the biome. Of the 8.1 million ha of soy–maize double
cropping, 99% experienced delays in the agricultural rainy season and 61%
faced reduced rainfall. Those changes have contributed to more frequent
and severe soybean and maize crop shortfalls. Delays in the onset of the
agricultural rainy season, reductions in rainfall and increases in maximum
air temperatures occur more frequently in areas with extensive native
vegetation loss. To sustain agricultural productivity in the Cerrado, it is
crucial to conserve and restore its native vegetation.

Despite their ecological relevance, tropical savannas are often under- to deforestation is the reduction in evapotranspiration 6,9, which
valued and inadequately protected worldwide1. This issue is particu- potentially consumes five times as much energy as that reflected by
larly critical in the Brazilian Cerrado (Supplementary Information an increased albedo in converted areas6, making it the dominant con-
Section 1.1), where the legal framework has proved unsuccessful for tributor to increased surface temperatures. This shift modifies the
conserving native vegetation2,3. Only 12.4% of this biome is protected local energy balance, reducing the latent heat flux and increasing the
with conservation units and indigenous lands and 50% of its native sensible heat flux, resulting in a rise in air temperature and surface tem-
vegetation has been converted into pastures and croplands, mainly peratures—which are important factors for the surface energy budget.
soybeans and maize (Fig. 1). Agribusiness has substantially transformed Previous research estimated the impacts of deforestation on the
the Cerrado biome into a major producer of grains. As of 2019, the Cerrado climate using field observations11,12, remote sensed data6,7,10,13
Cerrado accounted for 63% and 12% of Brazilian and global soybean and numerical modelling14,15 (Supplementary Table 1). Although these
production, respectively4. Yet this has come at a high environmental studies agree on the impacts of deforestation on the regional climate,
cost as agriculture continues to expand relentlessly at the expense of they have not yet separated the effects due to native vegetation loss
native vegetation. In 2023, even with the return of environmental law from those of anthropogenic greenhouse gas (GHG) climate change for
enforcement in Brazil, deforestation in the Cerrado biome was still the two crop seasons of the soy–maize double-cropping (DC) system
on the rise5. of the Cerrado. The empirical determination of impacts from native
Expansion of agriculture over native vegetation alters the regional vegetation loss on the climate and hence grain yields in the Brazilian
climate6–9, thereby impacting agriculture10. Studies indicate that the Cerrado thus remains a central challenge. First, deforestation trends
primary factor driving regional climatic change in the Cerrado due coincide with global climate change, making it difficult to separate

Center for Remote Sensing, Federal University of Minas Gerais, Belo Horizonte, Brazil. 2Woodwell Climate Research Center, Falmouth, MA, USA.
1

e-mail: [email protected]

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

a b
Changes in land use and occupation (1985–2022) State boundaries
Native vegetation
Non-vegetated area 5° S
Other natural formations MA
Other crops and forest plantation
Natural formation
Pasture
(forest or savanna)
Soy PI
10° S
Water
TO
55° W
BA
Other natural
formations
MT
DF
GO

Farming MG

20° S
MS
Non-vegetated area
Water SP
0 200 400 km
1985 2022

Fig. 1 | Land use and cover changes in the Cerrado over the past four decades. a, Sankey diagram of transitions between 1985 and 2022. b, Land use and cover by
2022. BA, Bahia; DF, Distrito Federal; GO, Goiás; MA, Maranhão; MT, Mato Grosso; MS, Mato Grosso do Sul; MG, Minas Gerais; PI, Piauí; SP, São Paulo; TO, Tocantins.
Map data in b from MapBiomas19.

one effect from the other. Second, the climate of the Cerrado exhibits temperature were assessed for each crop season. Together with the
pronounced spatial variation as well as large interannual variability onset of the agricultural rainy season, these variables are the ones for
and interdecadal fluctuations influenced by changes in sea surface which there are statistically significant shifts over the last four decades,
temperatures across the tropical Pacific and the Atlantic dipole mode. as determined by the Mann–Kendall test (P < 0.05) (Methods and Sup-
Here, we fill this gap by empirically assessing how deforestation has plementary Information).
intensified climate change across the Cerrado and, as a result, the risks A consistent trend in the delay of the onset of the agricultural
those changes pose to DC agriculture. rainy season resulted in a cumulative temporal shift of 36 days since
Our investigation involves an analysis on the long-term climate 1980, on average, for the Cerrado as a whole. In 1980, soybeans were
trends by using a dataset covering four decades from the Brazilian Daily usually sown in the first week of September and harvested in the first
Weather Gridded Database (BR-DWGD16), which has been extensively week of January. Yet, as of 2020, soybeans were sown around the third
validated17,18. Additionally, we explore the geographical modifications week of October. Roughly 90% of soybean and maize cultivation in
in the onset of the agricultural season, rainfall amount and maximum the Cerrado is rainfed22. If the rainy season begins later, it results in a
air temperature over the two cropping seasons by integrating 20 years delayed optimal cultivation window for these crops, as producers wait
of BR-DWGD climate data with annual land-use maps between 1999 for a period of reduced climate risk—that is, for the rainy season to be
and 2019 from MapBiomas collection 7 (ref. 19, Methods and Sup- firmly established. This shift has also entailed a tardy maize sowing
plementary Information). Our results pinpoint areas that already fall by early March, with harvests now occurring in the first week of May.
outside optimal climatic boundaries for DC as well as indicate the The average annual rainfall amount that was around 1,580 mm in the
probability of occurrence in terms of return period for surpassing 1980s dropped to 1,000 mm by 2020. These results agree with those
critical climate thresholds that may induce crop shortfalls. To isolate of previous studies that show a 9.2% decrease per decade in annual
the climate signals attributed to native vegetation loss, we applied a rainfall23,24. In parallel, there has been a significant increase in maximum
detrending procedure on the climate data series9,20. We then estimated air temperatures (P < 0.05). The average maximum air temperature has
the percentage of agricultural areas that have crossed the critical cli- reached 32 °C, a rise of 1.5 °C compared to that of the 1980s.
matic thresholds over the last decade (2009–2019). To identify crop
shortfalls induced by climate change, we used a generalized additive Geographic distribution of climate changes
model that considers non-climatic factors (technological develop- The recent climate changes in the Cerrado underscore the potential
ment) to estimate yield residues from crop production data. We also risks to agriculture, making farmers adapt to new temporal boundaries
derived cumulative probability distribution functions (CPDFs) to assess and climatic conditions. Yet, these changes are not geographically
the role of deforestation in augmenting the frequency of years with uniform and as such their impact on crops, particularly on DC, varies
suboptimal climatic conditions for DC (Methods and Supplementary across the biome (Fig. 2). Over the last decade (2010–2019) in compari-
Information). We concluded our study by suggesting conservation son with the previous decade (1999–2009), delays in the onset of the
strategies for the Cerrado biome to maintain the ecosystem services agricultural rainy season ranged from 10 to 30 days, depending on the
of the region, specifically in relation to climate regulation, which is key region under consideration (Fig. 2a). The southern and southeastern
to the success of agribusiness in Brazil. Cerrado, where DC is widely practiced, have experienced the most
significant delays in the rainy season, which have exceeded 21 days in
Mean long-term climate trend the states of Minas Gerais (MG), Mato Grosso do Sul (MS), Goiás (GO)
We estimated the onset of the agricultural rainy season for each grid and the Distrito Federal (DF).
cell of 28 × 28 km2 using the anomalous accumulation method21 on In addition to changes in the seasonality of the rainy season, the
BR-DWGD rainfall data16. On the basis of the onset of the rainy sea- eastern and southeastern Cerrado has experienced significant reduc-
son and the number of days necessary for cultivating soybeans and tions in rainfall amounts during both the first and second crop seasons,
subsequently maize in the Cerrado, we define the first crop season decreasing by as much as 169 mm in the first season and 358 mm in
from the first day of the onset to the 140th day and the second season the second season over the recent decade when compared with the
from 141st to 260th day. Trends in rainfall amount and maximum air previous one (Fig. 2b,c). The warming trend, on the other hand, is

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

a b c
Days mm mm

5° S –169 138 5° S 5° S
5 13 –169 138
MA MA MA

PI PI PI
10° S 10° S 10° S
TO TO TO
55° W 55° W 55° W
BA BA BA

MT MT MT
DF DF DF
GO GO GO

MG MG MG

20° S MS 20° S MS 20° S MS

45° W 45° W 45° W
SP SP SP

d e f
°C °C Native vegetation loss (%)

5° S 5° S 80–100 5° S
–3.7 2.9 MA –3.7 2.9 MA 60–80 MA
40–60
20–40
PI PI PI
10° S 10° S 0–20 10° S
TO TO TO
TO
55° W 55° W 55° W
BA BA BA

MT MT MT
DF DF DF
GO GO GO

MG MG MG

20° S MS 20° S MS 20° S MS

45° W 45° W 45° W
SP SP SP
0 250 500 km
State boundaries

Fig. 2 | Climate and land-use changes across the Cerrado from 1999 to 2019. a–f, Changes in the onset of the agricultural rainy season (a), rainfall amount during the
first crop (b) and second crop (c), maximum air temperature during the first crop (d) and second crop (e) and accumulated native vegetation loss by 2019 (f).
Data from ref. 16.

more pronounced during the first crop season, with some regions climatic thresholds in the most recent decade (2009–2019). Critical
experiencing maximum air temperatures up to 2.8 °C higher than climatic thresholds refer to shifts in the onset of the agricultural rainy
that of the previous decade (1999–2019) (Fig. 2d). For example, in MS, season, maximum air temperature and rainfall that surpass climate
MG and GO, warming of up to 2.9 °C from the previous decade has boundaries crucial for the optimal growth of current soybean (S100)
been recorded during the first crop season. Additionally, MS, MG and and maize (M120) cultivars. Deviations from these critical climatic
MATOPIBA (acronym for the states of Maranhão, Tocantins, Piaui and thresholds can substantially impact crop yields and hence agricultural
Bahia) have recorded warming of up to 2 °C during the second crop production. For the onset of the agricultural rainy season, a delay of
season (Fig. 2e). These regional climate changes broadly coincide with 10 days or more in planting soybeans may jeopardize the maintenance
gradients of native vegetation loss (Fig. 2f). of high yields throughout the growth cycle of S100 and M120 cultivars25.
Among the current 8.1 Mha of DC in the Cerrado, nearly all of them Minimum thresholds for rainfall amount during the growing season
(99%) have experienced delays in the onset of the agricultural rainy are 450 mm for the soybean cultivar S100 and 400 mm for the maize
season. During the first crop season, ~61% of DC areas have experienced cultivar M120 (refs. 26,27). In turn, an ideal critical air temperature
a decrease in rainfall amount over the most recent decade (2009–2019), threshold is 30 °C since higher air temperature has a detrimental impact
while over half (53%) have recorded higher maximum air temperatures on S100 and M120 yields28.
compared with those of the previous decade (1999–2009). On the basis of the geographical distribution of average precipita-
The situation is even worse for the second crop season, with 84% tion from 2009 to 2019 and the current DC plantations, only 3% of DC
of croplands facing higher maximum air temperatures and 61% of them areas fall below the critical threshold of 450 mm during the first crop
experiencing a reduction in rainfall amount. season (Fig. 3a) and virtually none falls below the threshold of 400 mm
in the second crop season (Fig. 3b). Even though reduction in rainfall
Critical climate thresholds has not yet affected most cultivation areas, the Cerrado is inherently
To explore the impact of climate change on grain production in the a hot region, and even slight increases in temperature can push many
Cerrado, we assessed the percentage of DC areas exceeding critical areas beyond optimal conditions for crop cultivation. Thus, high air

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

a b

5° S 5° S

MA MA

PI PI
10° S 10° S
TO TO

55° W 55° W
BA BA

MT MT

GO DF GO DF

MG MG

20° S 20° S
MS MS
SP SP
45° W 45° W

c State boundaries
d
Double cropping

Soy single cropping 5° S 5° S

Regions outside MA MA
climate thresholds

PI PI
10° S
TO TO

55° W 55° W
BA BA

MT MT

GO DF GO DF

MG MG

20° S 20° S
MS MS
SP SP
45° W 45° W

0 250 500 km

Fig. 3 | Regions that have crossed the critical climate thresholds over the recent period (2009–2019). a, Rainfall <450 mm in the first crop season.
b, Rainfall <400 mm in the second crop season. c, Maximum air temperature >30 °C in the first crop season. d, Maximum air temperature >30 °C in the second crop
season. Climate data from ref. 16, double cropping from IBGE45 and ref. 46.

temperatures emerge as the primary factor pushing areas beyond technological developments, such as biotechnology, genetic improve-
ideal cultivation climate boundaries. We found that 84% and 57% of cur- ment, mechanization, soil correction and fertilization. To do so, we
rent DC areas, for the first (Fig. 3c) and second crop seasons (Fig. 3d), applied a generalized additive model on historical yield data to arrive
respectively, surpass the critical threshold of 30 °C for maximum air at residues29 (Methods). Yield data for DC come from the Brazilian
temperature. Institute of Geography and Statistics (IBGE). We observed that, despite
notable advancements in agricultural practices, which have led to a
Soy–maize crop shortfalls 20% increase in soybean yields and a 37% increase in maize yields, on
Because ~90% of croplands in the Cerrado are rainfed, they are highly average, from 2006 to 2019, there has been a rising trend in negative
susceptible to changes in rainfall and air temperature25. To verify the yield residues, indicating more frequent and acute crop shortfalls.
impact of climate change on soy and maize, we eliminated trends These shortfalls are also affecting a larger portion of soybean and maize
in yields associated with factors beyond climate change, that is croplands in recent years (Mann–Kendall test, P < 10−5) (Fig. 4). This

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

a b

2006

2007

2008

2009

2010

2011

2012
Year

2013

2014

2015

2016

2017

2018

2019

–3,000 –2,000 –1,000 0 1,000 2,000 3,000 –3,000 –2,000 –1,000 0 1,000 2,000 3,000
Yield residuals (ton ha–1)

Fig. 4 | Soy and maize yield residues from 2006 to 2019. a, Soybean yield illustrates both positive residues and the more frequent and acute crop shortfalls
residues (first crop). b, Maize yield residues (second crop). The graph does towards the present for soybean and maize. Both trends are statistically
not cover the entire historical climate series because the systematic survey of significant (two-sided Mann–Kendall P < 0.05).
IBGE45 started distinguishing maize of second crop only after 2006. The graph

trend persists despite the expansion of cultivated areas, increased DC, we identified the climate signal resulting from native vegetation
application of fertilizers, adoption of irrigation and enhancements loss by removing factors unrelated to deforestation, such as El Niño/
in management practices and other adaptation measures. Moreover, Southern Oscillation (ENSO), anthropogenic GHG climate change,
while positive yield residues have become more pronounced, although geographic location and elevation9,20. Next, we developed CPDFs to
smaller than negative ones, those negative residues are becoming more evaluate the probabilities of crossing the critical climate thresholds
widely spread, indicating increased climate variability, especially in that could induce crop shortfalls. We used the same critical thresh-
some regions of the Cerrado (Fig. 2). olds, as defined above, for the delay in the onset of the agricultural
Native vegetation loss reinforces the crossing of climate thresh- rainy season, along with maximum air temperatures and reductions in
olds. Multiple large-scale factors, such as sea surface temperatures in rainfall amount within both crop seasons. We established categories
the North and South Atlantic and anthropogenic GHG climate change, using 20% incremental intervals of percentage of native vegetation
play a central role in climate variability across the Cerrado30. In addi- loss within grid cells, extracted the histograms for the variables and
tion, deforestation may also influence the regional climate10,31,32. To dif- then derived continuous CPDFs by running a Monte Carlo simulation
ferentiate the effects of native vegetation loss from the mean long-term with 10,000 iterations. We then transformed the probabilities of occur-
climate trends, we conducted a t-test comparing the onset of the agri- rence (Po) in terms of return period within the 20-year time period of
cultural rainy season, rainfall amount and maximum air temperature analysis (Methods).
between grid cells with limited native vegetation loss (≤20%) and those Results show that delays in the onset of the agricultural rainy
with extensive deforested areas (≥80%). This comparison was carried season (Extended Data Fig. 1), reductions in rainfall amount (Extended
out under similar geographical and interannual climatic conditions for Data Fig. 2) and increases in maximum air temperatures (Extended Data
two distinct periods: 1999–2009 and 2010–2019 (Methods). Fig. 3) occur more frequently in areas with extensive native vegetation
Grid cells with limited native vegetation loss (group 1) exhibit a loss. For the onset of the agricultural rainy season, a delay ≥10 days
reduction in rainfall amount of 29.1 ± 8.4 mm in the first crop season may occur every 2 years in regions with ≥80% of native vegetation
and 9.2 ± 2 mm in the second crop season between the two periods, loss (Po ≈ 0.5), whereas Po for regions with ≤20% of native vegetation
whereas grid cells with extensive deforested areas (group 2) show a loss is negligible. During the first crop season, regions with ≥80% of
decrease of 90.5 ± 13.5 and 109.1 ± 18.1 mm, respectively. Additionally, native vegetation loss may face a reduction in rainfall ≥200 mm in a
group 1 has experienced an increase in maximum air temperature of return period of every 5 years (Po ≈ 0.2), conversely regions with ≤20%
0.2 ± 0.04 mm in the first crop season and 0.4 ± 0.05 mm in the second of native vegetation loss may face a reduction in rainfall ≥200 mm
crop season. In turn, group 2 has had a more substantial increase of once every 10 years (Po ≈ 0.1) (Extended Data Fig. 2a). Considering the
1.1 ± 0.1 °C and 1.3 ± 0.2 °C, respectively. Both relationships are statisti- geographic distribution of precipitation across the Cerrado, a 200 mm
cally significant (P < 10−5). or more reduction in rainfall during the first crop season would render
To investigate the role of deforestation in increasing the occur- 65% of the current DC areas below the critical threshold of 450 mm
rence of years and regions with suboptimal climatic conditions for for growing soy (Fig. 5a). In the second crop season, areas with native

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

a b c

5° S 5° S 5° S
MA MA MA

PI PI PI
10° S 10° S 10° S
TO TO TO
55° W 55° W 55° W
BA BA BA

MT MT MT
DF DF DF
GO GO GO

MG MG MG

20° S 20° S 20° S

MS MS MS
45° W 45° W 45° W
SP SP SP

d e
State boundaries Deforestation (%)

Double cropping 5° S
Probability of occurrence due to native vegetation loss

50%
0–19.9 20–39.9 40–59.9 60–79.9 80–100
Soy single cropping MA

Regions outside
climate thresholds 40% First crop season Second crop season
PI
10° S
TO
30%
55° W
BA

MT 20%
DF
GO

MG 10%

20° S
MS
45° W 0%
SP Rainfall volume Maximum Rainfall volume Maximum Agricultural rainy
0 250 500 km decreasing temperature decreasing temperature season commencing
≥200 mm increase ≥1 °C ≥200 mm increase ≥1°C ≥10 days later

Fig. 5 | Additional effects of deforestation to the recent climate means. d, Maximum air temperature >30 °C in the second crop season. e, Cumulative
a, Rainfall <450 mm in the first crop season. b, Rainfall <400 mm in the second probability for the onset of the agricultural rainy season in each interval of native
crop season. c, Maximum air temperature >30 °C in the first crop season. vegetation loss. Climate data from ref. 16, double cropping from IBGE45 and ref. 46.

vegetation loss ≥80% may experience a reduction in rainfall ≥200 mm the current percentage of areas of DC in the Cerrado falling above the
every 5 years (Po ≈ 0.2). Conversely, regions with native vegetation air temperature limit of 30 °C. In the recent decade, this percentage for
loss ≤20% may experience this reduction once every decade (Po ≈ 0.1) the first crop situates at 80%, but with an additional 1 °C increase due to
(Extended Data Fig. 2b). Reductions ≥200 mm from the mean rainfall deforestation (Fig. 5e), it would expand to 92% (Fig. 5c). Likewise, for the
amount of the recent period could induce maize crop shortfalls in 20% second crop, the current percentage of 68% could rise to 94% (Fig. 5d).
of current DC areas (Fig. 5b). These results are consistent with recent
studies indicating that deforestation in the Cerrado has led to reduced Discussion
atmospheric water recycling through evapotranspiration7,10,13. Quantifying the impact of Cerrado deforestation on the regional cli-
Increase of at least 1 °C in maximum air temperature during the first mate is complex with uncertainties arising from the magnitude of the
crop season above the mean maximum air temperature of the recent deforestation effects, which vary as a function of the geographical scale
period may happen every 6 years (Po ≈ 0.16) in regions where native of observation20, climate–biosphere feedbacks33,34 and the diversity
vegetation loss is ≥80%, as opposed to once every 16 years in areas of native vegetation mosaics in the biome35. In addition to the direct
with ≤20% of native vegetation loss (Po ≈ 0.06) Extended Data Fig. 3a). effects of deforestation on the surface energy and moisture balance,
Similarly, during the second crop season, an increase of at least 1 °C in when deforestation involves or is followed by burning (a common case
maximum air temperature above the climate normal may occur every in the Cerrado), biomass burning alters both radiative and cloud prop-
5 years in regions with native vegetation loss ≥80% (Po ≈ 0.22) and once erties through aerosol emissions. While aerosols can serve as cloud
every decade in grid cells with ≤20% of native vegetation loss (Po ≈ 0.1) condensation nuclei, their presence can also reduce overall cloud cover,
(Extended Data Fig. 3b). Additional increase of 1 °C in maximum air allowing more solar radiation reaching the surface that contributes to
temperature resulting from the impact of deforestation would expand warming36, hence enhancing the deforestation effect on temperature.

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

Notwithstanding uncertainties, our detrending procedure has at 28 × 28 km grid-cell resolution. The use of a shorter timeframe for
proved to be a sound methodology to assess the impact of defor- analysing the effects of deforestation on key climate variables for the
estation on the regional climate9,21 and consequently on DC yields, DC system is justified since both shorter 20-year periods (1999–2019)
thus providing a roadmap that policy-makers can use to implement and the extended 40-year period (1980–2019) have equivalent trends,
region-specific agricultural policies and programmes. Successful agri- Mann–Kendall test (P < 0.05) (Extended Data Fig. 4).
culture in the region is contingent upon a stable climate. The prevalence On the basis of the available information from the Municipal Agri-
of hotter and drier conditions therefore presents a challenge for grain cultural Production Survey provided by the IBGE45, we obtained the
production, resulting in more DC areas surpassing climatic boundaries average yield of soy–maize DC system (t ha−1) at the municipal level
for optimal crop growth. This trend coincides with a notable increase in between 2006 and 2019. The tabular data were also converted into
soybean and maize crop shortfalls, especially in regions largely defor- time-series maps with a grid-cell resolution of 28 × 28 km using the
ested, indicating that deforestation not only impacts large areas of inverse distance weighting interpolation.
current croplands, but also hinders the expansion of agriculture across We calculated the extent of native vegetation loss between 1999
the Cerrado. Mitigating climate risks requires investment in adaptation and 2019 using the land-use and land-cover classification dataset from
solutions, such as irrigation and the development of drought-tolerant the MapBiomas project, collection 7 (mapbiomas.org). To match the
cultivars. However, each strategy presents its own unique challenges. land-use and land-cover change data with the climate data, we trans-
While irrigation can mitigate yield losses under severe water stress and formed MapBiomas maps into a time-series of accumulated percent-
temperature increases of up to 2 °C, its effectiveness is greater in drier age of native vegetation loss per grid cell, at a spatial resolution of
regions, where water availability is limited and declining37 and irrigation 28 × 28 km.
systems require access to water, electricity and substantial investments
in technology and management38. Moreover, Brazil’s success in genetic Areas of soy–maize double cropping
improvement may face challenges due to climate change, including We performed the spatial allocation of soy–maize double cropping
the spread of pathogens, shifts in growing seasons and more frequent using the Otimizagro model20,46, a spatially explicit model that simu-
extreme events, such as floods, droughts and wildfires39. lates land use and land-use changes, including agriculture, forestry,
Given that deforestation is a pivotal factor amplifying the climate native vegetation loss and regeneration, under scenarios of agricultural
change impacts on DC, conserving the native vegetation of the biome land demand and Brazilian environmental policies (Supplementary
emerges as a vital measure to mitigate those impacts. Yet the legal Information Section 1.3).
protection framework for the Cerrado is insufficient to effectively
curb deforestation2,3, underscoring the need for additional initiatives Agricultural rainy season onset and crop seasons
such as the implementation of the Action Plan for the Prevention and On the basis of the Ri,j,t estimates from the Tropical Rainfall Measuring
Control of Deforestation in the Cerrado Biome (PPCerrado40). Brazil’s Mission product 3B42 v.7, we identified the onset of the agricultural
soy farmers feel justified in opening new areas whenever they have rainy season (Oi,j,t) for each 28 × 28 km grid cell (Supplementary Infor-
the economic means and motivation. Because these decisions are mation Section 1.4). We used the anomalous accumulation method
influenced by shifts in credit and policies41, conservation initiatives (AA; ref. 21 and equation (1)). We defined September as the start of the
should include prioritizing sustainable agriculture, excluding defor- hydrological year. The AA method is directly applicable to agriculture,
esters from agricultural supply chains as well as from bank loans via since the water demand of a soybean or maize plant is used as the refer-
public and transparent traceability platforms42, property level land-use ence value (Rref in mm d−1), so that:
monitoring43, territorial planning aimed at restricting legal deforesta-
day
tion in climate and ecological sensitive regions, payment for ecosystem AA
(day)
= ∑ (Ri,j,t − Rref ) (1)
services including economic subsidies to landowners that converse or n=1
restore native vegetation beyond the forest code requirements2 and
the promotion of a value-based bioeconomy44. Our research also holds where Ri,j,t is in mm d−1. We fixed the Rref as 2.5 mm d−1, which corresponds
potential significance for insurance companies in assessing crop fail- to the needs of a soybean seedling47. Onset of the agricultural rainy
ure or shortfall risks aimed at fixing insurance premiums. Companies season for each grid cell (Oi,j,t), where i, j and t refer to space and time
could charge lower premiums to farmers in regions with larger native dimensions, is defined as the day from which AA remains positive dur-
vegetation cover, because they have less frequent and lower variations ing the longest period recorded21.
in rainfall and air temperature. In turn, this strategy could also incentiv- The soy–maize DC system involves planting soybeans at the start
ize restoration in largely deforested regions. of the rainy season and maize after the soybean harvest. Our study
In sum, deforestation is a counterproductive agricultural strategy, focuses on analysing each crop season independently. To estimate the
therefore the conservation of Cerrado is potentially the most effec- crop seasons, we used the Oi,j,t values. We grouped Ri,j,t data (mm d−1)
tive policy to contribute to global food and biofuel production. The from the Tropical Rainfall Measuring Mission satellite and the daily
current trajectory of land use in the Brazilian Cerrado jeopardizes the maximum temperature (°C) from the BR-DWGD database from the day
sustainability of the country’s largely rainfed agricultural systems, as of onset of the agricultural rainy season (considered day 1) up to the
Brazil’s agribusiness and global partners push the limits of nature by 140th day. We assume this time interval for the cultivation of the first
expanding croplands at the expense of native vegetation. To reverse crop. The second crop season takes place from day 141 to 260, which
this course, we need to act now. Quantifying the impacts of deforesta- typically corresponds to January to April (Supplementary Information
tion on the ecosystem services of the region, specifically in relation to Section 1.5).
climate regulation for agriculture, is one important way to convince
policy-makers and stake-holders to act before it is too late. Climate anomalies due to native vegetation loss
To identify the deforestation signal on the onset of the agricultural
Methods rainy season (O′i,j,t), rainfall volume (R′i,j,t) and maximum temperature
Data (Tmax′i,j,t) in each crop season, we had to remove the influence of geo-
We used daily rainfall volume estimates (Ri,j,t in mm d−1) and daily maxi- graphic location, elevation and interannual variability (the last due
mum temperatures (Tmax i,j,t in °C), from the BR-DWGD at an original to large-scale climate mechanisms). We adopted a four-step process
resolution of ~1 × 1 km, which has a high correlation (r² ≈ 0.8–0.9) with based on ref. 20 (Supplementary Fig. 5) to determine anomalies and to
in situ data. The BR-DWGD data were aggregated into time-series maps assess the effectivness of our detrending method. Our methodology

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

includes the following steps: (1) using machine learning algorithms and Space Administration of United States (https://www.usgs.gov/
to model climate variability across the Brazilian Cerrado, (2) evaluat- centers/eros/science/usgs-eros-archive-digital-elevation-shuttle-
ing model accuracy, (3) applying a detrending procedure to remove radartopographymission-srtm-non?qt-science_center_objects=
trends related to other factors than deforestation and (4) validating 0#qt-science_center_objects). Yield of soy–maize DC system is based
the detrending process to ensure its reliability. Each step is explained on the available information from the Municipal Agricultural Pro-
in detail in Supplementary Information Section 1.6. duction Survey by the IBGE (https://sidra.ibge.gov.br/pesquisa/lspa/
To verify the veracity of the detrending procedure at isolating the tabelas). Climate anomalies due to Cerrado native vegetation loss are
climate signal resulting from native vegetation loss, we conducted available via Figshare at https://doi.org/10.6084/m9.figshare.27273177.
several assessments (Supplementary Information Section 1.7). First, v1 (ref. 50), accumulated deforestation in the Cerrado biome is avail-
we calculated the annual means of each climate variable for the entire able via Figshare at https://doi.org/10.6084/m9.figshare.27273294.v1
study region after applying the detrending procedure. The detrended (ref. 51) and soybean and yield residues are available via Figshare at
annual means converged to zero, indicating that the procedure success- https://doi.org/10.6084/m9.figshare.27276621.v1 (ref. 52).
fully removes long-term trends associated with geographical location,
elevation and interannual variability. In addition, we examined the Code availability
spatial variability of the mean climate anomalies to check whether All analytical and simulation models were developed using Dinam-
they exhibit any spatial gradients. This analysis was performed by ica EGO (dinamicaego.com) and R (r-project.org). Codes used in
using the Cramer’s V (ref. 48) and the Spearman rank order correlation the paper are available via Figshare at https://doi.org/10.6084/
coefficients49. These metrics allowed us to evaluate the spatial associa- m9.figshare.27277701.v1 (ref. 53).
tion between percentage of native vegetation loss in each grid cell and
the estimated climate anomalies. The results indicate that, after the References
detrending procedure, the climate anomalies exhibited spatial vari- 1. Pennington, R. T., Lehmann, C. E. R. & Rowland, L. M. Tropical
ability that was solely correlated with percentage of native vegetation savannas and dry forests. Curr. Biol. 28, 541–545 (2018).
loss per grid cell. 2. Soares-Filho, B. S. et al. Cracking Brazil’s forest code. Science
344, 363–364 (2014).
Soybean and maize yield residues 3. Rajão, R. et al. The rotten apples of Brazil’s agribusiness. Science
The calculation of the soybean and maize yield residues is needed 369, 246–248 (2020).
to eliminate the influence of other factors, such as technological 4. Statistical Yearbook 2022 (FAO, 2022); https://doi.org/10.4060/
advancements, which could bias our analyses. To achieve this, we cc2211en
calculated soybean and second-crop maize yield residues using a 5. Projeto Prodes—Monitoramento de Desmatamento na Amazônia
three-step procedure (Supplementary Information Section 1.8). Legal (INPE, 2023); http://terrabrasilis.dpi.inpe.br/downloads/
First, we applied generalized additive models29 to fit mathematical 6. Loarie, S. R. et al. Direct impacts on local climate of sugar-cane
equations to the historical series of each grid cell as a function of time. expansion in Brazil. Nat. Clim. Change 1, 105–109 (2011).
Soybean and maize yields (Yst and Ymt, in kg ha−1) were the dependent 7. Arantes, A. E., Ferreira, L. G. & Coe, M. T. The seasonal carbon
variables and time (t, in years) was the independent variable. Here, and water balances of the Cerrado environment of Brazil: past,
we assume that the yields follow a Poisson distribution. This is based present, and future influences of land cover and land use. ISPRS J.
on the consideration that the probability of a series of events occur- Photogram. Remote Sens. 117, 66–78 (2016).
ring over a specific period is calculated assuming that these events 8. Spera, S. A., Winter, J. M. & Partridge, T. F. Brazilian maize yields
are independent of the time of the last event. In the second step, the negatively affected by climate after land clearing. Nat. Sustain. 3,
angular coefficients of the fitted equations were used to estimate the 845–852 (2020).
yield values (^Yst and ^Ymt), representing the trend for Yst and Ymt. In 9. Leite-Filho, A. T. et al. Climate risks to soy–maize double-cropping
the third step, we subtracted the fitted yield values from the observed due to Amazon deforestation. Int. J. Climatol. 44, 1245–1261 (2024).
values (equations (2) and (3)). 10. Rodrigues, A. A. et al. Cerrado deforestation threatens regional
′ ∧
climate and water availability for agriculture and ecosystems.
Yst = Yst − Yst (2)
Glob. Change Biol. 28, 6807–6822 (2022).
11. Nóbrega, R. L. B. et al. Effects of conversion of native Cerrado
Ymt ′ = Ymt −∧ Ymt (3) vegetation to pasture on soil hydro-physical properties,
evapotranspiration and streamflow on the Amazonian agricultural
The resulting values are considered the yield residue, isolating frontier. PLoS ONE 12, e0179414 (2017).
the influence of other climate factors. Supplementary Information 12. Anache, J. A. A. et al. Hydrological trade-offs due to different land
Sections 1.9 and 1.10 provide further description of methods and covers and land uses in the Brazilian Cerrado. Hydrol. Earth Syst.
Supplementary Information Section 2 gives a detailed description of Sci. 23, 1263–1279 (2019).
uncertainties. 13. Spera, S. A. et al. Land-use change affects water recycling in Brazil’s
last agricultural frontier. Glob. Change Biol. 22, 3405–3413 (2016).
Reporting summary 14. Coe, M. T. et al. The effects of deforestation and climate variability
Further information on research design is available in the Nature on the streamflow of the Araguaia River, Brazil. Biogeochemistry
Portfolio Reporting Summary linked to this article. 105, 119–131 (2011).
15. Pires, G. & Costa, M. Deforestation causes different subregional
Data availability effects on the Amazon bioclimatic equilibrium. Geophys. Res.
All the data that support the findings of this study are obtained from Lett. 40, 3618–3623 (2013).
publicly available sources. Annual deforestation data are obtained from 16. Xavier, A. C. et al. New improved Brazilian daily weather gridded
the MapBiomas project, collection 6 (https://brasil.mapbiomas.org/). data (1961–2020). Int. J. Climatol. 42, 8390–8404 (2022).
Climate data use the Brazilian Daily Weather Gridded Database from 17. Ruane, A. C., Goldberg, R. & Chryssanthacopoulos, J. Climate
ref. 16 (https://sites.google.com/site/alexandrecandidoxavierufes/ forcing datasets for agricultural modeling: merged products for
brazilian-daily-weather-gridded-data). Elevation data come from the gap-filling and historical climate series estimation. Agric. For.
Shuttle Radar Topography Mission from the National Aeronautics Meteorol. 200, 233–248 (2015).

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

18. Bender, F. & Sentelhas, P. Solar radiation models and gridded 38. Lathuilliere, M. et al. Rain-fed and irrigated cropland-atmosphere
databases to fill gaps in weather series and to project climate water fluxes and their implications for agricultural production in
change in Brazil. Adv. Meteorol. https://doi.org/10.1155/ Southern Amazonia. Agric. For. Meteorol. 256–257, 407–419
2018/6204382 (2018). (2018).
19. Project—Collection 7 of Brazil’s Annual Coverage and Land Use 39. Oliveira, M. W. et al. Dry matter and nutrient cycling by soil cover
Map Series (MapBiomas, 2022). plants in an intensive corn silage production system. Res. Soc.
20. Leite-Filho, A. T. et al. Deforestation reduces rainfall and Dev. 11, e45611831008 (2022).
agricultural revenues in the Brazilian Amazon. Nat. Commun. 12, 40. Plano de Ação para Prevenção e Controle do Desmatamento e das
2591 (2021). Queimadas no Bioma Cerrado (PPCerrado): 4a. Fase (2023 a 2027)
21. Liebmann, B. et al. Onset and Offset the rainy season in South (MMA, 2023).
America in observations and the ECHAM 4.5 atmospheric general 41. Aragão, R. B. A. et al. To clear or not to clear: unpacking soy
circulation model. J. Clim. 20, 2037–2050 (2007). farmers’ decision-making on deforestation in Brazil’s Cerrado.
22. Atlas de Irrigação 2021 (Agência Nacional de Águas, 2021); Front. Sustain. Food Syst. 6, 942207 (2022).
https://portal1.snirh.gov.br/ana/apps/storymaps/stories/ 42. Nunes, F. et al. Environmental Compliance of Coffee, Soy, and
a874e62f27544c6a986da1702a911 Forest Plantations in the State of Minas Gerais, Brazil (CSR/UFMG,
23. Campos, J. O. Variabilidade da Precipitação no Cerrado e sua 2024); https://csr.ufmg.br/csr/wp-content/uploads/2024/03/
Correlação com a Mudança no uso da Terra. MSc thesis, Univ. policy_conformidade_cadeias_sv_mg_13_03_24_en.pdf
Brasília (2018). 43. Balanço do Código Florestal Vol. 1 (CSR, 2022); https://csr.ufmg.
24. Van Dijkhorst, H., Kuepper, B. & Piotrowksi, M. Cerrado br/csr/wp-content/uploads/2022/08/boletim_cf_vol.1.pdf
Deforestation Disrupts Water Systems and Poses Business Risks 44. Carvalho-Ribeiro, S. et al. Bioeconomic markets based on
for Soy Producers (Chain Reaction Research, 2018); https:// the use of native species (NS) in Brazil. Ecol. Econ. 218, 108124
chainreactionresearch.com/wp-content/uploads/2018/10/ (2024).
Cerrado-DeforestationDisrupts-Water-Systems- and-Poses- 45. Produção Agrícola Municipal 2022 (IBGE, 2022); www.sidra.ibge.
Business-Risks-for-Soy-Producers3.pdf gov.br/bda/pesquisas/pam
25. Brumatti, L. M., Pires, G. F. & Santos, A. B. Challenges to the 46. Rochedo, P. et al. The threat of political bargaining to climate
adaptation of double cropping agricultural systems in Brazil under mitigation in Brazil. Nat. Clim. Change 8, 695–698 (2018).
changes in climate and land cover. Atmosphere 11, 1310 (2020). 47. Abrahão, G. M. & Costa, M. H. Evolution of rain and photoperiod
26. Berlato, M. A., Matzenauer, R. & Bergamaschi, H. limitations on the soybean growing season in Brazil: the rise (and
Evapotranspiração máxima da soja e relações com a possible fall) of double cropping systems. Agric. For. Meteorol.
evapotranspiração calculada pela equação de Penman. 256, 32–45 (2018).
Evaporação do tanque “classe a” e radiação solar global. Agron. 48. Cramér, H. Mathematical Methods of Statistics Ch. 21 (Princeton
Sul. 22, 243–259 (1986). Univ. Press, 1946).
27. Manual of Safety and Quality for Soy Crop (EMBRAPA, 2005); 49. Spearman, C. The proof and measurement of association
https://ainfo.cnptia.embrapa.br/digital/bitstream/item/25249/1/ between two things. Am. J. Psychol. 15, 72–101 (1904).
manualsegurancaqualidadeparaaculturadesoja.pdf 50. Leite-Filho, A., Soares-Filho, B. & Oliveira, U. Climate anomalies
28. Andrade, C. L. T. et al. Simulação do Crescimento da Planta e da due to Cerrado native vegetation loss [Dataset]. Figshare
Dinâmica de Agua e Nitrogênio na Cultura Milho: 1. Fitomassa, Area https://doi.org/10.6084/m9.figshare.27273177.v1 (2024).
Foliar e Produtividade de Grãos (EMBRAPA, 2006); 51. Leite-Filho, A., Soares-Filho, B. & Oliveira, U. Accumulated
https://ainfo.cnptia.embrapa.br/digital/bitstream/item/29773/1/ deforestation in the Cerrado biome (1999 to 2019) [Dataset].
Simulacao-crescimento-3.pdf Figshare https://doi.org/10.6084/m9.figshare.27273294.v1
29. Hastie, T. J. & Tibshirani, R. J. Generalized Additive Models (2024).
(Chapman & Hall/CRC, 2017). 52. Leite-Filho, A. Soybean and maize yield residuals in the
30. Marengo, J. A. et al. Onset and end of the rainy season in the Cerrado biome [Dataset]. Figshare https://doi.org/10.6084/
Brazilian Amazon Basin. J. Clim. 14, 833–852 (2001). m9.figshare.27276621.v1 (2024).
31. Spracklen, D., Arnold, S. R. & Taylor, C. M. Observations of 53. Leite-Filho, A., Soares-Filho, B. & Oliveira, U. Intensification
increased tropical rainfall preceded by air passage over forests. of climate change impacts on the Cerrado agriculture due to
Nature 489, 282–285 (2012). deforestation (Leite-Filho et al., 2024, Nat. Commun.).
32. Keys, P. W., Wang-Erlandsson, L. & Gordon, L. J. Megacity Figshare https://doi.org/10.6084/m9.figshare.27277701.v1
precipitation sheds reveal teleconnected water security (2024).
challenges. PLoS ONE 13, e0194311 (2018).
33. Swann, A. L., Fung, I. Y. & Chiang, J. C. Mid-latitude afforestation Acknowledgements
shifts general circulation and tropical precipitation. Proc. Natl A.T.L.-F. is supported by the Brazilian National Council for
Acad. Sci. USA 109, 712–716 (2012). Scientific and Technological Development (CNPq 140147/2019-5),
34. Liu, L. et al. Impact of biomass burning aerosols on radiation, BiomasBR MCTI—Cerrado: Financiadora de Estudos e Projetos
clouds, and precipitation over the Amazon during the dry season: (FINEP 01.22.0254.00) and Brazilian Coordination for the Improvement
relative importance of aerosol–cloud and aerosol–radiation of Higher Education Personnel (CAPES 88887.933209/2024-00).
interactions. Atmos. Chem. Phys. 20, 13283–13301 (2020). B.S.S.-F. is supported by the Brazilian National Council for Scientific
35. Ribeiro, J. F. & Walter, B. M. T. in Cerrado: Ecologia e Flora (eds and Technological Development (CNPq 304736/2023-6), Alexander
Sano, S. M. et al.) 153–212 (EMBRAPA, 2008). von Humboldt Foundation (3.4-11611976-BRA-IP) and Climate and
36. Archer-Nicholls, S. E. et al. Aerosol–radiation–cloud interactions Land Use Alliance (G-2305-59018). U.O. is supported by BiomasBR
in a regional coupled model: the effects of convective MCTI—Cerrado: Financiadora de Estudos e Projetos (FINEP
parameterisation and resolution. Atmos. Chem. Phys. 16, 01.22.0254.00). M.C. is supported by the US National Science
5573–5594 (2016). Foundation (INFEWS programme 1802754). A.T.L.-F. and B.S.S.-F. are
37. IPCC Climate Change 2022: Impacts, Adaptation and Vulnerability also supported by Rainforest Foundation Norway. We also thank A.
(eds Pörtner, H.O. et al.) (Cambridge Univ. Press, 2022). Oliveira and B. Alves for cartographic support.

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

Author contributions Peer review information Nature Sustainability thanks Francina

A.T.L.-F., B.S.S.-F. and U.O. conceived the study. A.T.L.-F. conducted Dominguez and the other, anonymous, reviewer(s) for their
the formal analysis. A.T.L.-F., B.S.S.-F. and U.O. developed the contribution to the peer review of this work.
methodology. A.T.L.-F. wrote the original draft. A.T.L.-F., B.S.S.-F., U.O.
and M.C. contributed to the writing—review and editing. Reprints and permissions information is available at
www.nature.com/reprints.
Competing interests
The authors declare no competing interests. Publisher’s note Springer Nature remains neutral with regard
to jurisdictional claims in published maps and institutional
Additional information affiliations.
Extended data is available for this paper at https://doi.org/10.1038/
s41893-024-01475-8. Springer Nature or its licensor (e.g. a society or other partner) holds
exclusive rights to this article under a publishing agreement with
Supplementary information The online version contains the author(s) or other rightsholder(s); author self-archiving of the
supplementary material available at https://doi.org/10.1038/s41893- accepted manuscript version of this article is solely governed by the
024-01475-8. terms of such publishing agreement and applicable law.

Correspondence and requests for materials should be addressed to © The Author(s), under exclusive licence to Springer Nature Limited
Argemiro Teixeira Leite-Filho. 2024

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

Extended Data Fig. 1 | Cumulative probability density functions for rainy season onset by native vegetation loss intervals. Cumulative probability density
functions (CPDF) for the onset of the agricultural rainy season in each interval of native vegetation loss. For the onset of the agricultural rainy season, we consider 10
days as the maximum delay.

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

Extended Data Fig. 2 | Cumulative probability density functions for rainfall anomalies by native vegetation loss intervals. Cumulative probability density
functions (CPDF) for anomalies of rainfall amount in different native vegetation loss intervals in the (a) first and (b) crop seasons in Cerrado. Critical rainfall reduction
of 200 mm for both crop seasons.

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

Extended Data Fig. 3 | Cumulative probability density functions for loss in the Cerrado for: (a) maximum air temperature in the first crop season;
maximum air temperature by native vegetation loss intervals. Cumulative (b) maximum air temperature in the second crop season. Critical maximum air
probability density functions (CPDF) for different intervals of native vegetation temperature increases of 1° Celsius.

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

Extended Data Fig. 4 | Climate trends in the Cerrado biome. Figure S4. Maximum air temperature trends for the periods (e) 1980–2019 and
Climate trends in the Cerrado biome over different periods. Average onset of (f) 1999–2019. All trends are statistically significant, as assessed by using the
the agricultural rainy season for the periods (a) 1980–2019 and (b) 1999–2019. Two-Sided Mann-Kendall test (p < 0.05). Data from Xavier et al.40.
Annual rainfall amount for the periods (c) 1980–2019 and (d) 1999–2019.

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

Extended Data Fig. 5 | Methodology for calculating climate anomalies. location and elevation; R *i, j,t is the difference between observed annual rainfall
Diagram with method steps to calculate the climate anomalies for each grid values and the estimated ones due to geographical location and elevation; R t
cell. i, j, t are the subscripts representing space (i,j) and time (t) dimensions and are the annual averages of rainfall calculated throughout the study region; Ri′, j,t
(C) represents the climate variables. In the example of application, Ri, j,t are the are the annual rainfall anomalies; φ, λ, ζ are the subscripts representing latitude,
annual rainfall values, R i, j are the estimated values of rainfall due to geographical longitude (in degrees), and elevation (in meters). Adapted from Leite Filho et al.10.

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

Extended Data Fig. 6 | Observed versus modeled average rainfall during Random Forest algorithm for the (d) first crop season and (e) second crop
crop seasons. Average observed rainfall during the (a) first crop season and season, and bias between observed and modeled data for the (c) first crop season
(b) second crop season from 1999 to 2019. Modeled average rainfall from the and (f) second crop season.

Nature Sustainability
Article https://doi.org/10.1038/s41893-024-01475-8

Extended Data Fig. 7 | Observed versus modeled maximum air temperature average maximum air temperature from the Random Forest algorithm for the
during crop seasons. Average observed maximum air temperature during the (d) first crop season and (e) second crop season, and bias between observed and
(a) first crop season and (b) second crop season from 1999 to 2019. Modeled modeled data for the (c) first crop season and (f) second crop season.

Nature Sustainability
 nature portfolio | reporting summary
Corresponding author(s): Argemiro Teixeira Leite FIlho
Last updated by author(s): Oct 26, 2024

Reporting Summary
Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.

Software and code

Policy information about availability of computer code
Data collection The Dinamica EGO platform (version 5) was used for environmental modeling. Data were obtained through direct downloads from publicly
accessible repositories. R Studio and version 4.4.2 of R were utilized for data analysis and processing.

Data analysis Data analysis for this study involved several key steps to investigate the impacts of deforestation on climate change and agricultural
productivity in the Cerrado biome. The analysis used a comprehensive dataset comprised of historical climate data and land-use maps:

Data Sources: The primary climate data utilized in this research was obtained from the Brazilian Daily Weather Gridded Database (BR-DWGD)
covering a span of four decades (1980-2020). Additionally, land-use maps from the MapBiomas collection (1999-2019) provided crucial
insights into vegetation changes and agricultural expansion.

Trend Analysis: Long-term climate trends were assessed using statistical methods, including the Mann-Kendall test, to identify significant
shifts in the onset of the agricultural rainy season, annual rainfall amounts, and maximum air temperature across different regions of the
Cerrado.

Spatial Analysis: A spatial analysis was conducted on 28×28 km grid-cells to assess the geographical distribution of changes in climate
variables. This involved mapping regional variations in the timing of the agricultural rainy season, rainfall distribution, and temperature
April 2023

increases to understand how these factors correlate with native vegetation loss.

Detrending Procedures: To isolate the climate signals attributed to native vegetation loss, a detrending procedure was applied to the climate
data series. This approach helped differentiate the effects of deforestation from those related to global anthropogenic climate change.

1
 Generalized Additive Models (GAM): A generalized additive model was employed to estimate crop yield residues by considering non-climatic
factors, such as technological advancements. This analysis facilitated the identification of yield trends and shortfalls in response to climatic

nature portfolio | reporting summary

variations.

Cumulative Probability Distribution Functions (CPDFs): CPDFs were developed to evaluate the probabilities of crossing critical climatic
thresholds that could induce crop shortfalls. This statistical tool provided insights into the frequency of extreme climate occurrences in
relation to native vegetation loss.

Statistical Comparisons: T-tests were utilized to compare climate variables between grid-cells with varying levels of native vegetation loss,
allowing for the examination of the relationship between deforestation and regional climate changes.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A description of any restrictions on data availability
- For clinical datasets or third party data, please ensure that the statement adheres to our policy

All the data that support the findings of this study are obtained from publicly available sources. Annual deforestation data are obtained from the MapBiomas
project, collection 6 (https://brasil.mapbiomas.org/). Climate data use the Brazilian Daily Weather Gridded Database from Xavier et al. [16] (https://
sites.google.com/site/alexandrecandidoxavierufes/brazilian-dailyweather-gridded-data?authuser=0). Elevation data come from the Shuttle Radar Topography
Mission (SRTM) from the National Aeronautics and Space Administration of U.S.
(https://www.usgs.gov/centers/eros/science/usgs-eros-archive-digital-elevation-shuttle-radartopographymission-srtm-non?qt-science_center_objects=0#qt-
science_center_objects). Yield of soy-maize double cropping system based on the available information from the Municipal Agricultural Production Survey by the
Brazilian Institute of Geography and Statistics (https://sidra.ibge.gov.br/pesquisa/lspa/tabelas).
Access to processed data is available at Figshare repository (Leite-Filho, Argemiro; Soares-Filho, Britaldo; Oliveira, Ubirajara (2024). Climate anomalies due to
Cerrado native vegetation loss. figshare. Dataset. https://doi.org/10.6084/m9.figshare.27273177.v1
Leite-Filho, Argemiro; Soares-Filho, Britaldo; Oliveira, Ubirajara (2024). Accumulated deforestation in the Cerrado Biome (1999 to 2019). figshare. Dataset. https://
doi.org/10.6084/m9.figshare.27273294.v1)
Leite-Filho, Argemiro (2024). Soybean and maize yield residuals in the Cerrado Biome. figshare. Dataset. https://doi.org/10.6084/m9.figshare.27276621.v1

Research involving human participants, their data, or biological material

Policy information about studies with human participants or human data. See also policy information about sex, gender (identity/presentation),
and sexual orientation and race, ethnicity and racism.
Reporting on sex and gender N/A

Reporting on race, ethnicity, or N/A

other socially relevant
groupings

Population characteristics N/A

Recruitment N/A

Ethics oversight N/A

Note that full information on the approval of the study protocol must also be provided in the manuscript.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Ecological, evolutionary & environmental sciences study design

April 2023

All studies must disclose on these points even when the disclosure is negative.

Study description This study examines the intensifying impacts of climate change on agriculture in the Cerrado biome, specifically focusing on soy-
maize double cropping, due to deforestation. Since the 1980s, deforestation has led to a 36-day delay in the agricultural rainy season,

2
 an 8.6% reduction in rainfall, and a 1.5°C increase in temperature, affecting 99% of the 8.1 million hectares of double-cropped land.
Regions with extensive native vegetation loss experience more drastic climate changes, resulting in increased frequency of crop

nature portfolio | reporting summary

shortfalls. The research underscores the urgent need to conserve and restore native vegetation to mitigate adverse climate effects
on agricultural productivity and ensure the sustainability of Brazil's agribusiness. Effective policies and conservation strategies are
essential to balance agricultural expansion with ecological preservation, ultimately enhancing climate regulation and food security in
the region.

Research sample The research sample for this study encompasses a comprehensive assessment of the Cerrado biome, focusing specifically on the
effects of deforestation on climate variables and agricultural productivity related to soy-maize double cropping. The sample includes:

Sampling strategy No sampling was involved; all data were used in their entirety.

Data collection All the data that support the findings of this study are obtained from publicly available sources. Annual deforestation data are
obtained from the MapBiomas project, collection 6 (https://brasil.mapbiomas.org/). Climate data use the Brazilian Daily Weather
Gridded Database from Xavier et al. [16] (https://sites.google.com/site/alexandrecandidoxavierufes/brazilian-dailyweather-gridded-
data?authuser=0). Elevation data come from the Shuttle Radar Topography Mission (SRTM) from the National Aeronautics and Space
Administration of U.S.
(https://www.usgs.gov/centers/eros/science/usgs-eros-archive-digital-elevation-shuttle-radartopographymission-srtm-non?qt-
science_center_objects=0#qt-science_center_objects). Yield of soy-maize double cropping system based on the available
information from the Municipal Agricultural Production Survey by the Brazilian Institute of Geography and Statistics (https://
sidra.ibge.gov.br/pesquisa/lspa/tabelas).
Access to processed data is available at Figshare repository (Leite-Filho, Argemiro; Soares-Filho, Britaldo; Oliveira, Ubirajara (2024).
Climate anomalies due to Cerrado native vegetation loss. figshare. Dataset. https://doi.org/10.6084/m9.figshare.27273177.v1
Leite-Filho, Argemiro; Soares-Filho, Britaldo; Oliveira, Ubirajara (2024). Accumulated deforestation in the Cerrado Biome (1999 to
2019). figshare. Dataset. https://doi.org/10.6084/m9.figshare.27273294.v1)
Leite-Filho, Argemiro (2024). Soybean and maize yield residuals in the Cerrado Biome. figshare. Dataset. https://doi.org/10.6084/
m9.figshare.27276621.v1

Timing and spatial scale This research covers a temporal span of four decades, analyzing climate trends and agricultural impacts from 1980 to 2020 in the
Cerrado biome. The study focuses on the changes in the agricultural rainy season, rainfall amounts, and maximum air temperatures,
specifically assessing their effects on soy-maize double cropping practices.

Spatially, the research centers on the entire Cerrado region, which encompasses approximately 2 million square kilometers in Brazil.
The analysis utilizes grid-cells of 28×28 km to examine variations in climate and agricultural practices across different geographical
areas within the biome.

Data exclusions N/A

Reproducibility All primary data used are open source and readily available, facilitating study replication, and the methods are clearly described.
Access to processed data is available at Figshare repository (Leite-Filho, Argemiro; Soares-Filho, Britaldo; Oliveira, Ubirajara (2024).
Climate anomalies due to Cerrado native vegetation loss. figshare. Dataset. https://doi.org/10.6084/m9.figshare.27273177.v1
Leite-Filho, Argemiro; Soares-Filho, Britaldo; Oliveira, Ubirajara (2024). Accumulated deforestation in the Cerrado Biome (1999 to
2019). figshare. Dataset. https://doi.org/10.6084/m9.figshare.27273294.v1)
Leite-Filho, Argemiro (2024). Soybean and maize yield residuals in the Cerrado Biome. figshare. Dataset. https://doi.org/10.6084/
m9.figshare.27276621.v1

All analytical and simulation models were developed using Dinamica EGO (dinamicaego.com) and R (r-project.org). Access to code is
available at Figshare (Leite-Filho, Argemiro; Soares-Filho, Britaldo; Oliveira, Ubirajara (2024). Codes used in the manuscript
'Intensification of climate change impacts on the Cerrado agriculture due to deforestation' (Leite-Filho et al., 2024, Nature
Communications).. figshare. Software. https://doi.org/10.6084/m9.figshare.27277701.v1).

Randomization In this study, the absence of randomization is justified due to the nature of the research design and the specific objectives aimed at
understanding the impacts of deforestation on climate and agriculture within the Cerrado biome.

Blinding In this study, blinding is not applicable as the research design primarily involves observational data analysis rather than experimental
intervention.

Did the study involve field work? Yes No

Reporting for specific materials, systems and methods

We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.
April 2023

3
Materials & experimental systems Methods

nature portfolio | reporting summary

n/a Involved in the study n/a Involved in the study
Antibodies ChIP-seq
Eukaryotic cell lines Flow cytometry
Palaeontology and archaeology MRI-based neuroimaging
Animals and other organisms
Clinical data
Dual use research of concern
Plants

Plants
Seed stocks N/A

Novel plant genotypes N/A

Authentication N/A

April 2023