Pathways for recent Cerrado soybean expansion: extending the soy moratorium and implementing integrated crop livestock systems with soybeans

We limit the scope of our study to the 870 municipalities that overlap with the Cerrado biome and that produced soybeans in 2014. This area spans 12 Brazilian states, including: Bahia (BA), Distrito Federal (DF), Goiás (GO), Maranhão (MA), Mato Grosso (MT), Rondônia (RO), Minas Gerais (MG), Mato Grosso do Sul (MS), Piauí (PI), Paraná (PR), São Paulo (SP), and Tocantins (TO) (figure 1, table 1). Together, these states accounted for 82% of Brazilian soy production and 84% of national deforestation across biomes in 2014 [2, 80].

Zoom In Zoom Out Reset image size

We investigate Soy Moratorium scenarios with benchmark years in 2008 and 2014. We chose 2008 rather than 2006 as a benchmark year as this is congruous with the current Soy Moratorium in the Amazon, and because this year had greater data availability. 2008 also represents a more compelling hypothetical because it was the year the Soy Moratorium was overhauled to become more consistent with the Forest Code, which was a missed opportunity to implement the strategy nation-wide. We chose the year 2014 as a proxy for a policy implemented today because this was the most recent year of data available in all relevant datasets.

We use three datasets to estimate the amount of cleared area before years 2008 and 2014 that do not currently grow soybeans, referred to hereafter as 2008 eligible area and 2014 eligible area respectively. This area should be viewed as land where Soy Moratorium compliant expansion could feasibly occur. Conversely, ineligible areas refer to land where compliant expansion may not occur, such as land under natural vegetation, or land already growing soybeans. We combine land use maps developed by TerraClass Cerrado with Agrosatélite soy maps for the 2013/2014 harvest year, and LAPIG's annual Cerrado deforestation data produced from MODIS images (MOD13Q1) and validated by LANDSAT and CBERS images for years 2008 through 2014 [25, 80].

We incorporate Soares-Filho's soy-suitability layer (2014) to determine area that was 2014 eligible and suitable, 2014 eligible and unsuitable, and suitable but under natural vegetation (figure 2) [35]. Finally, we remove annual deforestation polygons between 2008 and 2014 to obtain final estimates for 2008 (figure 2).

Zoom In Zoom Out Reset image size

Because agriculture is the main deforestation activity in the study area [45, 46] we consider the cleared land before 2014 to be equal to the total agricultural land use in 2014, which is the sum of all cropland, pastureland (natural and planted), and recently cleared area without designation in TerraClass. The latter category refers to land that has no vegetation, no agriculture, and no forestry uses present (0.18% of the total observed area) [80]. Second, we extract these land uses from Terraclass and mask the output with the Agrosatélite map of Cerrado soy area for the 2013/2014 crop year. We then exclude all area already growing soybeans to create 2014 eligible expansion estimates (figure 2).

We estimate the additional soybean production that ICLS systems could have contributed if implemented onto 2008 suitable pastures during the years between 2008 and 2014. Land was considered to be eligible for ICLS if they were pasture in 2008 and suitable for soybeans. ICLS producers typically rotate crops with pasture every 2–12 years [42, 47]. We assume a conservative rotation frequency where soybeans are planted onto suitable pasture once every 8 years (n = 8) and quantify municipal production increases using annual municipal average yield data from the Brazilian Institute of Geography and Statistics (IBGE) [2]. n = 8 is a conservative estimate intended only as a reference value. This assumption can be modified to other values of n, by multiplying our results by 8/n (equation (1)).

Of the 870 Cerrado municipalities that produced soy in 2014, 846 had suitable established 2008 pasture where ICLS could be implemented. Total pastureland area for each municipality was obtained from TerraClass Cerrado and bounded by Soares-Filho's suitability (2014) and LAPIG's deforestation (2008–2014) layers (figure 2). To determine ICLS quantity increases for each year between 2008 and 2014, we use the following calculation:

$$\begin{eqnarray}&&\begin{array}{l}{\rm{ICLS}}\,{{\rm{quantity}}}_{{m}}\\ \,=\,\displaystyle {\sum }_{{m}}y\left({\rm{soy}}\,{{\rm{yield}}}_{{m},{t}}\right.\\ \,\times \,\left.\displaystyle \frac{1}{8}2008\,{\rm{suitable}}\,{\rm{pasture}}\,{{\rm{area}}}_{{m}}\right),\end{array}\end{eqnarray} \tag{ 1 }$$

where ICLS quantity is the sum of the product of soybean yield in a year t for each municipality m and one eighth of 2008 suitable pasture area. Soy yield refers to annual IBGE yields, and pasture area uses values from the TerraClass database (2014).

The Cerrado contains 67.9 Mha of 2008 eligible cleared area. Of this area, 31.1 Mha (46%) are considered unsuitable for soybeans, leaving 36.8 Mha of suitable eligible area where soybeans could expand and remain in compliance with a 2008 Cerrado Soy Moratorium. Of this suitable area, 29% (10.8 Mha) was found in Mato Grosso do Sul, 22% (8.2 Mha) in Goiás, 21% (7.7 Mha) in Minas Gerais, 15% (5.6 Mha) in Mato Grosso, and 8% (2.8 Mha) in São Paulo. The remaining states contained less than 4% each (table 1, figure 3).

Zoom In Zoom Out Reset image size

The same states that possess the most suitable eligible area also have the most suitable area under natural vegetation; Mato Grosso has 12.7 Mha (35%); Minas Gerais has 6.7 Mha (19%); Goiás has 5.2 Mha (15%); and Mato Grosso do Sul has 5.2 Mha (14%) (table 1, figure 3).

Approximately 19% of all 2008 eligible cleared area was already being used to grow soy in 2014. Of this, 47% of Piauí's pre-2008 cleared area is already used for soy, 41% of Paraná's, 40% of Mato Gross's, 34% of Distrito Federal's, and 31% of Bahia's, while the remaining states use less than 20% combined (figure 3(a)). If they were able to maximize the amount of suitable eligible area present, states Mato Grosso do Sul and São Paulo could both increase their 2014 soy acreage by over 8 times, while Distrito Federal, Minas Gerais, Goiás, and Mato Grosso could more than double their 2014 soy acreage. Matopiba states could increase less; soy area in Bahia could increase by 67%, in Tocantins by 43%, in Maranhão by 28%, and in Piauí by 15%. Piauí contains more soy acreage on land cleared after 2008 (0.2 Mha) than it has eligible cleared area (0.1 Mha), suggesting the state may have had a land debt if the policy had been adopted in 2008. While the Matopiba states together possess roughly 12.7 Mha of 2008 eligible area, 88% of this area is not considered suitable for soybeans. If land improvements were to occur in these areas, their potentials for Soy Moratorium compliance may increase as land becomes more suitable.

We find that between 2008 and 2014, 1.4 Mha of soy-suitable area was cleared. Roughly half (0.7 Mha) of this did not grow soybeans in 2014, making it eligible for compliant expansion given a 2008 Moratorium. The other half (0.7 Mha) did grow soy in 2014 and would be in violation of a 2008 Soy Moratorium. Of this, 82% of this acreage is located in Matopiba, and 96% in Matopiba and Mato Grosso states combined. These areas highlight where recent soy expansion has come at the cost of forest between 2008 and 2014. This violation area contributed 8.9 Mt of soybeans to the market between 2008 and 2014.

The Cerrado contains an additional 4.2 Mha of 2014 eligible cleared area compared to 2008 levels (72.1 Mha total). 3.5 Mha (83%) of this are unsuitable for soybeans, leaving 0.7 Mha of suitable 2014 eligible area for future compliant conversion (37.5 Mha total). States possessing the largest additions of the latter include Distrito Federal (20%), Mato Grosso (20%), Bahia (18%), Minas Gerais (13%), and Mato Grosso do Sul (10%) (figures 3 and 4).

Zoom In Zoom Out Reset image size

In addition to possessing the highest amounts of 2014 eligible area Mato Grosso, Minas Gerais, and Mato Grosso do Sul also possessed the highest concentrations of 2008 eligible area. Bahia's additional 2014 eligible expansion area could enable the expansion of its 2014 soy area by up to 9% compared to 2008 eligible area levels. Other Matopiba states would show little additional expansion capacities as most of their pre-2014 cleared area already grew soy in 2014; Tocantins and Piauí could expand their 2014 soy area up to an additional 3%, and Maranhão up to 2% compared to 2008 levels (figure 4).

Of the 3.8 Mha of land cleared between 2008 and 2014, 77% of deforestation occurred in Matopiba (62%) and Mato Grosso (15%). The vast majority of post-2008 deforestation occurred on unsuitable land for soybeans. States with the highest proportion of their suitable land under natural vegetation are found to be Tocantins (83%), Maranhão (75%), Mato Grosso (69%), Bahia, (68%), and Distrito Federal (64%). Mato Grosso do Sul (32%), and São Paulo (34%) have the lowest proportions of suitable area under natural vegetation.

We estimate that Cerrado pastureland for the year 2008 covered roughly 58.9 Mha, of which 54% (31.9 Mha) was suitable for soybeans and thus eligible for ICLS implementation. Of the 1375 Cerrado municipalities with established pasture in 2008, 846 municipalities had area suitable for soybeans. If ICLS management strategies were to have been adopted on all 2008 suitable pasture between 2008 and 2014, average annual municipal yields would allow an extra 63.4 Mt of soy to have been produced throughout the period. 31% (19.7 Mt) of this potential exists in Mato Grosso do Sul, 26% (16.8 Mt) in São Paulo, 19% (12.2 Mt) in Mato Grosso, and 17% (11.0 Mt) in Goiás (table 1, figure 5). The remaining states had less than 4% of the potential each, with Matopiba containing less than a combined 2% of the increased production potential in the Cerrado due to persisting low yields (figure 5) [2]. This increase would have grown annual national production by an average of 13%.

Zoom In Zoom Out Reset image size

An 8 year ICLS rotation would have led to a 4.0 Mha increase in soy area throughout the study period; 1.3 Mha of this land is found in Mato Grosso do Sul (33%), 0.9 Mha in Goiás (24%), 0.8 Mha in Minas Gerais (20%), 0.7 Mha in Mato Grosso (16%), and less than 5% in the remaining states. Matopiba only embodied a combined 16% of the new soy area potential due to lack of established pasture in the region (table 1, figure 5). This compares to roughly 5 Mha of soy expansion in the study area that actually occurred during the period. In order to completely accommodate this acreage on suitable pasture, it would have required a minimum rotation frequency of one planting every 6 years.

In the unlikely scenario that all suitable area were to adopt soybean production and rotate every year (n = 1 for equation (1)), we estimate the maximum upper boundary for additional soy area during the period to be 31.9 Mha, and the maximum quantity produced given municipal average yield to be 507 Mt. Likewise, if n = 100, and soybeans were rotated only once every hundred years, the additional soy land added to the Cerrado would be 0.3 Mha and the additional quantity produced would equal 5.1 Mt in the study period.

Production increases can generally be achieved through intensification or extensification. Previous work has shown that soy yields are already at 92.5% of their potential in the Cerrado region, suggesting that it may be very difficult to close the gap further [62]. While vast increases in compliant production are possible through expansion onto suitable eligible area, pathways are limited to areas currently occupied by other land uses. Leakage risk outlined by previous work should be a key consideration [11, 27, 63]. Cattle remains the single greatest driver of deforestation in the tropics, and established pasture in the Cerrado should not be allowed to deforest new area while ceding land for soybeans [11, 16, 24]. Policy interventions and management strategies that help to deescalate land use competition between soy and beef are urgently needed.

ICLS management practices present an alternative where new production area can occur without additional clearing, and without other land use displacement. Previous studies have shown that rotating soy with pasture may increase productivity in each commodity, improve crop resilience to drought and frost, enhance soil health, promote water conservation, and increase carbon storage [39, 42, 64]. We show that a conservative ICLS implementation between 2008 and 2014 could grow total national soy production by 13% in the same period, though Matopiba showed limited potential as the region has not engaged in large scale ranching in the past [65].

The examined scenario assumes that already established pasture would completely accommodate new soy expansion between 2008 and 2014. A 100% soy-onto-pasture expansion rate is not far from what occurred in reality for the period. From 2000 to 2014, 80% of Brazil's total cropland expansion has been onto pasture [66]. in the Cerrado specifically, previous work has shown that between 2007 and 2014, 74% of new soy area expanded onto pasture or non-soy cropland—non-soy crops comprise approximately 3% of all agricultural land in the Cerrado, so the vast majority of this expansion can be assumed to have been onto pasture [25]. Trends differed in Matopiba and Mato Grosso, where for the same time period 62% and 68% of soy expansion occurred over native vegetation, respectively [62]. For Matopiba, the low soy-onto-pasture expansion rates and limited quantities of established pasture may serve as insurmountable obstacles for producers implementing ICLS management at a meaningful scale. However, we show Mato Grosso as having one of the highest potentials for increased soy production through ICLS (figure 5), suggesting that pivoting the state towards ICLS management could serve a role in slowing high rates of expansion onto native vegetation. Further, if excluding these regions, the remaining states (DF, GO, MG, MS, PR, SP) had 89% of their soy expansion occur over pasture between 2007 and 2014 [62]. These high rates coupled with high potentials for production increases (figure 5) make these places exceptionally well-poised to emphasize ICLS implementation.

We estimate that it would require a minimum rotation frequency of one planting every six years in order to accommodate all of the Cerrado's soy expansion for the study period onto 2008 suitable pasture using ICLS. However, any new soy area on already established pasture may or may not lead to land sparing. Under specific policy contexts, intensification has been associated with positive conservation outcomes [67–70]. The Amazon is a particularly successful example of this; being subject to intensive land use zoning regulation, credit restrictions for bad actors, and private market mechanisms has synergistically worked to create conditions where intensification contributes to land sparing [21, 61, 67–70]. However, in regions that are absent of these conditions, a 'rebound effect' has been found to occur, where increased productivity leads to more incentive to clear new lands and thus exacerbates land conversion pressures in the region [71]. If proper conditions were present in the Cerrado, it is possible that this 4.0 Mha of new soy area could result in land spared. However, successfully leveraging ICLS for conservation outcomes depends heavily on the presence of regulatory conditions highlighted in previous work [21, 72], likely in the form of land-use-zoning policy that guides implementation, introduction programs targeted to specific areas, and credit incentives to help bridge initial costs of adoption [67, 73]. While it is possible that another Soy Moratorium could play a role in this regulation, it is likely that a more comprehensive zoning policy that accounts for indirect land use change and combines economic and ecological standards (e.g. Map of Ecological Zoning for Sugarcane) would be more effective [72]. It is important to note that if these ideal policy conditions were present, the improved pasture yield that results from ICLS implementation may contribute an additional land sparing effect, which should be studied in-depth in future work.

The Brazilian Agricultural Research Corporation (EMBRAPA) has started emphasizing integrated systems as a means to qualify for Brazil's ABC Plan, which rewards sustainable production practices with low-interest loans [64]. The conditions under which ranchers may decide to adopt ICLS are complex. Cattle ranches are most often large properties (>300 ha) located far from towns, and are associated with having limited access to credit and machinery [74]. Access to machinery, in particular, correlates with whether cattle ranches use crop-pasture systems versus cattle alone [74]. Some frontier regions face other socio-economic barriers to implementing ICLS systems, such as lack of technical and economic expertise, lack of quality labor, lack of credit access, high financial costs, being far from supply chain infrastructure, and the absence of business models that are applicable to small-scale farmers [42, 43, 73, 75]. These obstacles have resulted in low adoption rates historically [22, 76], yet surveys show integrated systems are becoming more widely recognized as financially beneficial in the long term and as a welcome opportunity for specialized farmers to diversify within intensified systems [42]. In the current landscape, where ranchers face increasingly low prices offered by the meatpacking industry along with high pasture recuperation costs, ranchers may be more eager to experiment with income augmentation through integrating soy production onto pasture [57, 77–78]. One previous study showed average ICLS stocking densities of 3.4 animals per hectare versus 0.98 animals per hectare in conventional systems, and the ICLS cattle reached slaughter weight a year earlier than normal [42]. While this result presents an optimistic picture for producers, more research is needed to quantify the long-term financial implications for adoption.

The land sparing effects of ICLS-based cattle intensification and potential adoption for soy producers should also be carefully examined in future work. To date, adoption efforts have focused primarily on soy producing municipalities, as adoption is both lower risk and lower cost than for cattle producers [64]. This analysis makes the case for a more concerted effort to extend programs into pasture producing municipalities to achieve maximum benefit. In general, strategies should be tailored to each audience. For ranchers, the rehabilitation of degraded pastures and eventual high stocking densities should be emphasized, and for soy producers, the low initial cost and quick returns.

Our estimates that use data from 2008 to 2014 assume that land uses in the Cerrado have remained relatively static over time and do not move around from year to year. Municipalities face additional expansion limitations from other policy, such as the Legal Reserve requirement in the Forest Code. Completely compliant expansion under both the Forest Code and a Cerrado Soy Moratorium dually requires land cleared before 2008 or 2014 and Forest Code surplus [35]. The model presented is not dynamic and is meant for simple bookkeeping. Leakage, laundering, and rebound effects have not been incorporated. Soy expansion is also subject to many variables outside of the scope of this study, including profitability, market networks, connectivity, and infrastructure [49–51]. While this work focuses specifically on land availability, future work that integrates these variables would be invaluable to ongoing discussion surrounding the viability of a Cerrado Soy Moratorium and ICLS adoption.

We consider an eight-year crop-pasture rotation to be conservative given common regimes highlighted in ICLS literature [40–42, 63, 75]. We chose this conservative rotation frequency because ICLS has never been explored at the scale we discuss here, and may have other barriers to adoption. The financial benefits of ICLS may be insufficient to drive widespread adoption. For example, the perception that cattle ranching requires less labor than other more complex crop systems may work to dissuade those considering adoption in areas where labor is scarce, meanwhile, uncertain markets increase the appeal of acquiring low-risk savings in the form of cattle while enjoying the elevated social status associated with the profession [44, 74, 79].

Finally, combining datasets from four different sources can result in compatibility uncertainties. Our methods integrate LANDSAT-based 30 × 30 resolution data from TerraClass Cerrado, Agrosatélite, and LAPIG, with 60 × 60 resolution data from Soares-Filho (2014). To reconcile the different resolutions of these datasets ArcGIS automatically resamples using the nearest neighbor assignment to the coarsest resolution of the input datasets, creating marginal losses of accuracy.