Science‐based planning can support law enforcement actions to curb deforestation in the Brazilian Amazon

Abstract

While Brazil publicly committed to reduce deforestation in Amazonia at the 26th Conference of the Parties (COP26), the Brazilian parliament is moving toward weakening environmental laws. Deforestation rates continue ascending, reaching in 2021 the highest value since 2006 (13,235 km²). To overcome this paradox, strategies to curb deforestation are mandatory. The current strategy, “Plano Amazônia 21/22,” prioritizes law enforcement actions to curb illegal deforestation in only 11 Amazonian municipalities. Here, we show that this prioritization is likely to be insufficient since these municipalities account for just 37% of the current deforestation rate. This strategy may also be undermined by the leakage of deforestation actions to unmonitored municipalities. Using a set of spatially explicit datasets integrated into a deforestation‐prediction modeling approach, we propose a science‐based alternative method for ranking deforestation hotspots to be prioritized by law enforcement actions. Our prioritization method accounts for more than 60% of the deforestation, detecting larger deforested areas in both private and public lands, while covering 27% less territory than “Plano Amazônia 21/22.” Optimizing the detection of priority areas for curbing deforestation, as proposed here, is the first step to reducing deforestation rates and comply with the Brazilian legal commitment of 3925 km² year⁻¹.

INTRODUCTION

Stopping the persistent Amazon rainforest loss has a paramount global relevance. The continuous release of its 86 Pg of carbon stored in the biomass (Saatchi et al., 2007), totalizing around 8 years of global fossil fuel emissions, may have irreversible climatic effects from regional to global scales. As part of global agreements to fight climate change, Brazil publicly declared during the COP26 in November 2021 that the country would (i) reduce greenhouse gas emissions (GHG) by 50% in comparison to 2005 levels by 2030, (ii) reach climatic neutrality by 2050, and (iii) achieve zero illegal deforestation by 2028 (MMA, 2021). The recent increasing trend in Amazonia deforestation rates (INPE, 2022), however, contrasts with these goals. The path the world's largest tropical forest nation follows ahead is of international concern as it may put at risk all global efforts toward a sustainable planet, compromising the conservation of tropical biodiversity, and the well-being of Amazonian people.

After a historical deforestation reduction of 84% achieved in 2012 in relation to 2004, Brazilian Legal Amazon (BLA) deforestation rates relapsed and reached in 2021 the highest value since 2006 (13,235 km²) (INPE, 2022). The observed increase in deforestation rates since 2012 results from several controversial actions, which indirectly incentivizes land grabbing. Changes in the Brazilian Forest Code (BFC, Law 1,2651/2012) and the intensification of discussions in the Parliament about old and new bills possibly leading to the decriminalization of illegal occupation of public lands (PL-490/2007, PL-2,633/2020, and PL-510/2021), combined with the easing of environmental licensing (PL-3,729/2004, in progress in the Parliament), can directly motivate illegal activities. This scenario is aggravated by interrupting successful initiatives for curbing BLA deforestation in the past, such as the Action Plan for Prevention and Control of Deforestation (PPCDAm) (West & Fearnside, 2021).

In 2012, Brazil was close to the legal commitment reaffirmed by a Federal Decree in 2018 of reducing Amazonia's deforestation rate to 3925 km² year⁻¹ by 2020, a reduction of 80% in relation to the average rate between 1996 and 2005 (Decree 9,578/2018). The year 2021, however, exceeded this commitment by 237% (INPE, 2022). The consequence is an increased pressure on Brazil to control illegal deforestation. The Brazilian inability to curb deforestation can directly influence its export of commodities, as a consequence of buyer's embargos. Agribusiness generates over USD100 billion of income to the Brazilian economy, accounting for nearly half of the exports in 2020 (MAPA, 2021). Maintaining or even increasing these exports without deforesting, which could be achieved by increasing land productivity in already cleared BLA areas, can generate critical income to support the Brazilian postpandemic economic recovery.

In this scenario, successful law enforcement actions against deforestation are required to protect BLA remaining forests (Trancoso, 2021). Clearly, these actions are not producing the desired effect, especially in 2019, 2020, and 2021 when deforestation rates surpassed 10,000 km². The continuous rising of deforestation rates highlights the urgent need for alternative solutions to tackle this problem. With budget limitations, new plans for curbing illegal deforestation must maximize operational success by focusing on the smallest geographic area as possible, contributing to the highest fraction of total deforestation. A new plan to guide law enforcement actions in the BLA (“Plano Amazônia 21/22” or PA21/22) started in 2021. This plan establishes 11 Amazonian municipalities to carry out law enforcement actions and sets a less ambitious deforestation reduction goal than that established in the Decree 9,578/2018, as described in the next section. This may be insufficient to reduce deforestation in the BLA to the required rates. Here, we use a set of spatially explicit datasets and a deforestation-prediction modeling approach to propose a science-based alternative method for prioritizing deforestation hotspots to support law enforcement. We argue that our method can (i) reduce the total area currently monitored by authorities, (ii) highlight new deforestation frontiers, and (iii) cover a larger proportion of deforestation hotspots contributing to total deforestation independently of political boundaries, which would reduce cross-municipality leakage of illegal activities. This is the first critical step to tackle deforestation and to redirect Brazil toward maintaining its environmental assets, achieving its international commitments, and protecting its commodity-based economy.

“PLANO AMAZÔNIA 21/22”

PA21/22 replaced “Operação Verde Brasil 2” (OVB-2) in May 2021 (DOU, 2021). This new plan establishes guidelines for law enforcement actions against deforestation in the BLA at the federal level. PA21/22 aims to reduce deforestation rates to 8719 km² by the end of 2022, which corresponds to the mean deforestation rate from 2016 to 2020 (DOU, 2021). Any action to reduce deforestation is valid; however, as PA21/22 lacks legal status, such as the Decree 9,578/2018, there is no legal obligation for guaranteeing that this timid deforestation reduction goal will be achieved. The PA21/22 has defined two strategies leading to major changes in how law enforcement actions were conducted in the BLA in 2019 and 2020: (i) enforcement actions are focusing on 11 priority municipalities (Figure 1) that, according to the plan, account for 70% of the BLA deforestation or environmental crimes (Supporting Information S1), and (ii) leadership of environmental inspections returned to environmental agencies instead of military forces (DOU, 2021). The method adopted for the first strategy, which defined the prioritization routine and estimated that 70% of the BLA deforestation or environmental crimes occur in these municipalities (Supporting Information S1), although unclear, permits achieving the proposed reduction. This is because 37% of the BLA deforestation increment in 2021 (4048 km²) was detected in these municipalities (INPE, 2022). It does not allow, however, to reach the national legal commitment established in the Decree 9,578/2018. Even if zero deforestation was achieved in these priority areas in 2021, the remaining Amazon-wide deforestation rate would still be 134% higher than the national legal commitment. Deforestation in the PA21/22 priority municipalities was significant over the last years (Figure 2a) and they indeed need to be monitored, but this is not enough. The prioritization of certain municipalities, in addition to not being sufficiently ambitious, may also induce the leakage of deforestation actions to unmonitored municipalities, causing the emergence of new deforestation hotspots. From February 18, 2021 to April 30, 2021, actions from the OVB-2 were already based on the PA21/22 strategy (DOU, 2021). Surprisingly, Brazil's official deforestation alert program (DETER) registered 524.89 km² of new deforestation alerts in the PA21/22 priority municipalities during this period. This was the highest value reported between February and April by DETER and is 105% above the average for this time period from 2017 to 2021. Alerts for the whole BLA during this period also set the highest value since the beginning of DETER (992.09 km²), 41% above the average between 2016 and 2021.

FIGURE 1. Brazilian Amazon and the location of the 11 priority municipalities prioritized by “Plano Amazônia 21/22.” ALT, Altamira; APU, Apuí; COL, Colniza; ITA, Itaituba; LAB, Lábrea; NPR, Novo Progresso; PAC, Pacajá; POR, Portel; POV, Porto Velho; RUR, Rurópolis; SFX, São Félix do Xingu. Base map is the Land Use and Land Cover classification of the Brazilian Amazon made available by the MapBiomas Project for the year 2019

FIGURE 2. (a) Accumulated deforestation estimated by PRODES from 2018 to 2021 within each 25 × 25 km grid cell. (b) Required avoided deforestation to meet the Brazilian legal commitment and the “Plano Amazônia 21/22” goal. (c) Priority classes for combating deforestation in the Brazilian Amazon in 2022 considering the regular grid of 25 × 25 km. (d) Contribution of the “Plano Amazônia 21/22” priority municipalities and the 2022 “High” priority areas to the Brazilian Amazon deforestation rates over time

The second strategy, correctly, brings back the legal attributions of law enforcement to environmental agencies. The PA21/22, however, does not define budgetary sources for the effective implementation of these actions. The environmental agencies responsible for the law enforcement, according to the PA21/22, are the Brazilian Institute of the Environment and Renewable Natural Resources (Ibama) and the Chico Mendes Institute for the Biodiversity Conservation (ICMBio). These agencies have experienced massive budget cuts, which jeopardize operations to curb deforestation (Supporting Information S2).

BOOSTING LAW ENFORCEMENT USING SCIENCE-BASED PLANNING

To achieve the national legal commitment (maximum of 3925 km² year⁻¹), 9310 km² of deforestation should have been avoided in 2021. Regarding the more flexible PA21/22 goal (maximum of 8719 km²), the required avoided deforestation in 2021 was 4516 km². While the national legal commitment was never achieved, the PA21/22 goal would have been achieved in most of the years since 2009 (Figure 2b), except, 2019, 2020, and 2021.

The total area to be monitored in the PA21/22 (the sum of all priority municipality areas) is 574,724 km². Here, we demonstrate that our grid cell-based prioritization strategy based on the machine-learning Random Forest algorithm (Breiman, 2001), which automatically builds sets of multivariate regressions to predict deforestation hotspots in the following year (Supporting Information S3), has advantages over the current PA21/22 strategy for guiding law enforcement actions. Our method identifies a larger fraction of areas at risk of deforestation in terms of total area and in public lands, which indicates illegal deforestation. The total area prioritized yearly by our method is on average 27% smaller than the combined area prioritized by PA21/22 (Table 1). Our model considered five deforestation predictors including deforestation in previous years, distance to grid cells with high cumulative deforestation in previous years, distance to infrastructures (highways, waterways, and highways or waterways), total area protected in the grid cell, and the number of active fires. Three classes of priority were defined based on the estimates of predicted deforestation: (i) “Low” (values below the 70th percentile), (ii) “Average” (values between 70th and 90th percentiles), and (iii) “High” (values above the 90th percentile). We have considered all grid cells classified as “High” to map the 2022 priority areas. The Root Mean Squared Error (RMSE) obtained after cross-validating the model was 2.78 ± 0.1 km². Predicted values of deforestation usually underestimated high values of deforestation (>10 km²), which are potential outliers, but most importantly the spatial distribution of predicted deforestation agreed with the overall spatial distribution of official deforestation increments (Supporting Information S4 and S5).

TABLE 1 Area of the priority classes defined by our method and by the “Plano Amazônia 21/22” from 2019 to 2022, and total and percentage of annual deforestation increment estimated by PRODES in the prioritization defined by each method

		Priority classes
		High	Average	Low	PA21/22
2019	Priority class area (Km2)	422,727	847,315	2,991,068	574,724
	PRODES deforestation increment (km2)	7261	2929	707	3443
	Annual deforestation increment (%)	67	27	6	32
2020	Priority class area (Km2)	423,150	847,118	2,990,841	574,724
	PRODES deforestation increment (km2)	6882	2540	1079	3908
	Annual deforestation increment (%)	66	24	10	37
2021	Priority class area (Km2)	414,603	850,154	2,996,351	574,724
	PRODES deforestation increment (km2)	7152	3047	654	4048
	Annual deforestation increment (%)	66	28	6	37
2022	Priority class area (Km2)	414,603	848,896	2,997,610	574,724
	PRODES deforestation increment (km2)	–	–	–	–
	Annual deforestation increment (%)	–	–	–	–

In total, we classified 414,603 km² as “High” priority in 2022 (Figure 2c; Table 1). The mapped “High” priority areas in 2022 concentrated 74% of the total deforestation increment in 2021 and, as observed for the priority municipalities, their contribution to the BLA deforestation is increasing over time (Figure 2d). Only 38% of the 2022 “High” priority grid cells were located within the PA21/22 priority municipalities. Moreover, 74% of total alerts emitted by DETER from August 1, 2021 to January 7, 2022 in the BLA occurred in the 2022 “High” priority areas (2362 km²), while PA21/22 priority areas concentrated 38% of total alerts (1213 km²).

Our method surpasses territorial delimitations and helps to prioritize actions in boundary areas of the 11 priority municipalities, where deforestation activities are concentrated. Moreover, it captures other areas of increasing deforestation that are not monitored by PA21/22. The proposed method here overcomes the deforestation leakage problem to other unmonitored areas by the PA21/22, since the “High” priority areas are determined on a yearly basis and are not dependent on geopolitical frontiers (e.g., municipalities). Additionally, the method detects a larger portion of deforestation while covering much less territory. It also contemplates new deforestation hotspots, such as the ones in the state of Roraima (Supporting Information S1). Our evaluation indicates that our method could be readily assessed by authorities and law enforcement agencies, directing actions to the “High” priority areas that are not contemplated by PA21/22. If successful, it could be an operational planning tool for the entire BLA. States that have “High” priority areas but are not included in PA21/22, such as Acre and Roraima, could also benefit from this method. Overall, our method is potentially a critical tool for supporting actions aiming to meet the challenging commitments established during COP26, especially achieving zero illegal deforestation by 2028.

We demonstrate that the largest proportion of the official deforestation increment in 2021 in both PA21/22 and our method was located in private farms (1818 and 3045 km² deforested, corresponding, respectively, to 45% and 41% of the total deforestation increment in each method), followed by nondesignated lands and rural settlements defined by Freitas et al. (2017) (Figure 3; Table 2). Accordingly, Rajão et al. (2020) estimated that 62% of all potentially illegal deforestation in the Amazon and Cerrado biomes concentrates in 2% of private properties. A better strategy for enforcing the law in private farms is urgently required, since planning enforcement operations in private properties is more complex compared to public lands. A potential solution for curbing illegal deforestation in private farms might be blocking the access of landowners not complying with environmental laws to public and private credits. We must, however, emphasize that a fraction of the deforestation occurring in private farms may be legal, since the BFC allows 20% of private properties located in the BLA to be legally deforested.

FIGURE 3. Spatial location of the land tenure classes, the PRODES 2021 deforestation increment, the “Plano Amazônia 21/22” proposed prioritization, and the “High” priority class derived from our method for the year 2021 in the Brazilian Amazon. PAs in this figure is referring to Protected Areas

TABLE 2 Land tenure class associated with the PRODES 2021 deforestation increment located within the 2021 “High” priority areas defined by our method and the “Plano Amazônia 21/22” priority municipalities

Land tenure class	2021 deforestation PA21/22 priority municipalities (km2)	2021 deforestation “High” priority class (km2)	Area difference (km2)
No data	4.45 (0.11%)	12.21 (0.17%)	7.76
Military areas	22.48 (0.56%)	22.37 (0.30%)	–0.10
Indigenous lands	134.23 (3.32%)	142.48 (1.93%)	7.95
Full protection PAs	491.43 (12.14%)	833.24 (11.31%)	341.81
Sustainable use PAs	119.84 (2.96%)	155.28 (2.11%)	35.44
Nondesignated lands	859.53 (21.23%)	1506.47 (20.45%)	646.94
Private farms	1818.83 (44.93%)	3045.14 (41.35%)	1226.31
Rural settlements	582.45 (14.39%)	1611.04 (21.87%)	1028.59
Quilombolas	0.01 (0.00%)	0.16 (0.00%)	0.15
Urban, transportation, and water	14.75 (0.36%)	36.82 (0.50%)	22.07
Total	4048 (100.00%)	7365 (100.00%)	3316.93

Abbreviation: PAs, protected areas.

Despite the recent push to ease environmental regulations, protected areas and Indigenous lands avoid deforestation more effectively than unprotected areas. Combined, these land tenure classes accounted for 18% (745 km²) and 15% (1131 km²) of the 2021 deforestation increment in the priority areas of PA21/22 and our method, respectively. These classes cover 61% (351,946 km²) and 18% (78,792 km²) of the total area to be monitored by PA21/22 and our method, respectively. We detected 385 km² more deforestation in these land tenure classes than PA21/22 while covering 272,704 km² less territory. The effectiveness of protected areas for containing illegal deforestation depends on government enforcement (Nolte et al., 2013). If enforcement actions are undermined by political reasons or the protected areas are not properly monitored, conservation efforts can be lost (Carvalho et al., 2019). Enforcing the law in these areas is key for maintaining Amazonia's remaining forests.

TOWARD A SUSTAINABLE FUTURE WITHIN THE NEW GLOBAL MARKET CONCEPTS

As an initial step, but critical for avoiding illegal deforestation (Trancoso, 2021), law enforcement alone is unable to significantly reduce deforestation rates. Long-term inhibition of illegal deforestation requires (i) identification and accountability of actors infringing environmental protection laws and profiting from illegal deforestation (Aragão et al., 2020) to avoid future infractions, (ii) regularization of public and Indigenous lands (Silva Junior et al., 2021), necessary for managing the environment adequately, ceasing social conflicts and violence episodes, and guaranteeing local populations’ rights, (iii) promotion of environmental education (Aragão et al., 2020) and intensification of environmental conservation awareness via public media (Caetano, 2021), since a large part of the population is unaware of the critical environmental services offered by forests (e.g., maintaining climate stability), undermining the importance of standing forests, (iv) enhancement of market-derived pressures on agricultural products from illegally deforested areas (Rajão et al., 2020; Silva Junior et al., 2021) such as the Soy Moratorium that reduced deforestation in the BLA (Heilmayr et al., 2020), and (v) addition of economic value to the standing forest by promoting and financing forest economy activities (e.g., investing on the production chain of Amazonian products such as cupuaçu) (Nobre et al., 2021). All these actions are complementary to the environmental laws and their enforcement. If successfully implemented, a drop in illegal deforestation would be very likely.

Updating and strengthening the PPCDAm, which promoted a drastic conservation reform in Brazil leading to the reduction in Amazonian deforestation specially from 2007 to 2012 (West & Fearnside, 2021), would be a logical step to curb deforestation in the following years. The four stages of the PPCDAm included (i) the development of the DETER program (Diniz et al., 2015) to support ground-level law enforcement actions, (ii) the creation of a “blacklist” of the BLA municipalities mostly contributing to deforestation and restricting public rural credit and legal deforestation permits in these municipalities, (iii) blocking loans from public banks to landowners in noncompliance with environmental and rural regulation, (iv) promoting the Rural Environmental Registry (CAR) system to support land tenure regularization, and (v) the expansion of protected areas (West et al., 2019; West & Fearnside, 2021).

The Brazilian paradoxical environmental policy cancels the efforts to curb illegal deforestation, causing preoccupation regarding the postpandemic economic recovery. It is a risk that international investors will not ignore, which may clearly lead to economic losses due to sanctions (Supporting Information S6). The growing global demand for deforestation-free supplies requires that producers urgently adapt to the new consumer's habits by controlling deforestation (Supporting Information S7).

About 11% of GHG emissions are caused by poor forestry and land-use management, including commodity-driven deforestation (Masson-Delmotte et al., 2019). Currently, two-thirds of the already-cleared BLA land is underused, degraded, or abandoned (Zero Deforestation Working Group, 2017). Increasing agricultural production in already-cleared areas by supporting technological development instead of land expansion will change the misconception that deforestation is necessary for Brazil's economic growth. Moreover, it is necessary to protect the secondary forests that have grown in abandoned lands, especially due to their potential as a nature-based solution to mitigate climate change (Heinrich et al., 2021). The BLA sustainable economic and social development is required, but strategic long-term territorial planning for the conservation of remaining forests is paramount. Science-based solutions, such as the method evaluated here, offer this possibility. There is absolutely no place for uncontrolled illegal deforestation to occur in Amazonia, especially if we take into consideration Brazil's current social, economic, and environmental aspects.

ACKNOWLEDGMENTS

All authors thank the Brazilian Space Agency (AEB) for paying the article publication charge. G.M., R.D., M.E.D.C., and L.E.O.C.A. thank the São Paulo Research Foundation (FAPESP) (grants 2019/25701-8, 2019/21662-8, 2021/07382-2, and 2016/02018-2, respectively). L.E.O.C.A. thanks CNPq (grant 314416/2020-0).

AUTHOR CONTRIBUTIONS

Guilherme Mataveli, Luiz E. O. C. Aragão, Gabriel de Oliveira, and Michel E. D. Chaves conceptualized the study. Guilherme Mataveli, Luiz E. O. C. Aragão, Ricardo Dalagnol, Fabien H. Wagner, and Alber H. S. Ipia designed the methodology. Guilherme Mataveli, Gabriel de Oliveira, Michel E. D. Chaves, and Celso H. L. Silva Jr. wrote the original draft. Luiz E. O. C. Aragão, Ricardo Dalagnol, Fabien H. Wagner, and Alber H. S. Ipia reviewed and edited the manuscript.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT

PRODES and DETER data are available at: http://terrabrasilis.dpi.inpe.br/en/home-page/. All data and code to define the priority classes are available at: https://github.com/albhasan/prioritizedeforestationhotspots.