SUPPORTING INFORMATION Future environmental and agricultural impacts of Brazil’s Forest Code Aline C. Soterroni1,2 , Aline Mosnier1 , Alexandre X. Y. Carvalho3 , Gilberto Cˆ amara2 , Michael Obersteiner1 , Pedro R. Andrade2 , Ricardo C. Souza2 , Rebecca Brock4 , Johannes Pirker1 , Florian Kraxner1 , Petr Havl´ık1 , Valerie Kapos4 , Erasmus K. H. J. zu Ermgassen5 , Hugo Valin1 & Fernando M. Ramos2? 1
International Institute for Applied System Analysis, Laxenburg, Austria National Institute for Space Research, S˜ao Jos´e dos Campos, Brazil 3 Institute of Applied Economic Research, Bras´ılia, Brazil 4 United Nations Environment Program, World Conservation Monitoring Centre, Cambridge, United Kingdom 5 University of Cambridge, Cambridge, United Kingdom 2
E-mail:
[email protected] (corresponding author)
Model framework Our model, called GLOBIOM-Brazil, combines the general framework of GLOBIOM [1] with a series of refinements that reflect Brazil’s specificities. GLOBIOM is based on the ASM-GHG model [2] and is a bottom-up economic partial equilibrium model focusing on major global landbased sectors, i.e., agriculture, forestry and bioenergy. Mathematically, the model simulates competition for land at the pixel level by solving a constrained linear programming problem: the maximization of welfare (i.e., the sum of producer and consumer surplus) subject to resources, technology and policy restrictions. The initial demand for each product in each region is computed using population and GDP growth from the SSP2 scenario [3], income elasticities from the USDA and food preferences from the FAO [4, 5]. The equilibrium production, consumption, and trade quantities and prices, are computed for each product and region and are the result of the optimization problem. The model relies on the homogeneous goods assumption, where the price difference between two regions is explained by trade costs only, allowing the model to represent bilateral trade flows. Thus, the model includes both tariffs and transportation costs for each product and trading partner. Trade tariffs come from the MAcMap database [6], and transportation costs, for which data are lacking, are computed by using the coefficients between the freight rates and distance and the
A C Soterroni et al
2
weight-over-value ratios of goods that have been estimated by Hummels [7]. The trade calibration method proposed by Jansson and Heckelei [8] is applied to reconcile observed bilateral trade flows, regional net trade, prices, and trading costs for the base year of 2000. Tariff adjustments for China have been introduced into the model for the year 2010. China joined the World Trade Organization (WTO) in December 2001, and since then, the country has undergone significant liberalization of its trade (e.g., Chinese soybean imports have increased from 14 million tons in 2000 to 45 million tons in 2010). In the scenario SSP2, the population in Brazil increases by 30% by 2030, and by 35% by 2050 compared to the population in the year 2000. For per capita GDP projections, Brazil is slightly above the world average, with an increase of 120% in 2030 and of 250% in 2050 compared to the level in 2000. The 2010 projections by World Energy Outlook (WEO) [9] are used to exogenously determine the bioenergy demand (biodiesel, bioethanol, charcoal, and heat and electricity), the biomass demand (heat and power generation) and the direct biomass consumption (e.g., charcoal for steel industry) per region. The biofuel demand is setup exogenously. The bioethanol demand for Brazil is from the Ministry of Mines and Energy and the Energy Research Enterprise (MME/EPE) [10]. From these projections, bioethanol use in Brazil will continue to strongly increase until 2030. Biodiesel use is also expected to increase in Brazil, but the overall level is still lower than that of bioethanol in 2030. A grid of 0.5◦ by 0.5◦ degrees (approximately 50 km x 50 km at the equator) for Brazil and a grid of 2◦ by 2◦ degrees (approximately 200 km x 200 km at the equator) for the rest of the world is used in GLOBIOM-Brazil to explicitly represent crop, livestock and wood production. There are 18 crops represented in the model, amounting to 86% of the cultivated area in Brazil in the year 2000. The biophysical model EPIC is used to estimate potential crop yields [11] for each crop and management system (subsistence, low-input rain-fed, high-input rain-fed, and high-input irrigated). Production costs are also defined at the grid level per crop type and management system based on the spatial heterogeneity from the EPIC model and the SPAM dataset [12]. Productivity increases endogenously from the reallocation of production to more suitable areas or through shifts between management systems, from low to high input, at the pixel level. This allows for endogenous yield changes in response to market signals. Exogenous yield growth is also possible due to technological changes over time, following economic growth projections [13]. The livestock sector of GLOBIOM covers 8 animal types that are spatially distributed and 7 animal products (bovine dairy and meat, sheep and goat dairy and meat, pig meat, and poultry meat and eggs). Ruminants (bovines, sheep and goats) can be produced in 8 production systems, and monogastrics (pigs and poultry) can be produced in 2 production systems. The management systems for ruminants are defined according to agro-ecological zones (arid, humid, and temperate/highlands) and feed requirements (grass-based, mixed crop-grass and other) based on ILRI/FAO [14, 15]. Monogastrics are differentiated across smallholders and industrial, and its management systems are based on the literature review. Production costs for livestock systems are based on FAOSTAT producer prices for product output and for grain input. The RUMINANT
A C Soterroni et al
3
model [16] is used to estimate productivity and feed requirements. Switching between production systems allows for feed substitution and for intensification or extensification. All these features permit a detailed representation of the production of livestock, their feed requirements, and the link to land through grazing needs [1]. The forestry sector is represented in the model by five categories of primary products (sawn logs, pulp logs, biomass for energy, traditional fuel woods, and other industrial logs), which are consumed by industrial energy, cooking fuel demand, or processed and sold on the market as final products (i.e., wood pulp and sawn wood). These products are supplied from managed forests and short-rotation plantations. The forestry model G4 M [17] is used to compute harvesting costs and annual mean increments. Crops, livestock and forestry production are Leontief-type, i.e., fixed input and output ratios. Changes in the technological characteristics of primary product production are considered in the model to allow for multiple production types (ranging from subsistence to intensive agriculture). Land conversion cost is represented by a non-linear function. The cost per converted hectare increases with the total converted area. If production is no longer profitable, land can also be abandoned. Based on field data, land-use conversion inside protected areas in Brazil is not allowed by the model. In 2009, only 1.47% of these areas had been deforested in the Amazon [18]. GLOBIOM-Brazil is repeatedly run for 10-year time steps, from 2000 to 2050, and provides detailed trajectories for supply, demand, prices and land-use change variables. The originality of GLOBIOM comes from representing drivers of land-use change at two different geographical scales. While prices, final demands, processing quantities and trade are computed at the regional level, all land-related variables (i.e., land-use change, crop cultivation, timber production and livestock number) are modeled at the pixel (or local) level. This means that regional factors influence how land use is allocated at the local level by the model, and local constraints influence the outcome of the variables defined at the regional level. This feature ensures full consistency across multiple scales within short solution times, and allows the model to properly evaluate internal policies concerning land-use competition between the major land-based sectors. Land-cover/use map for Brazil GLOBIOM-Brazil land-use/cover classes are defined in Table S1. Given the total area of Brazil, we do not consider transitions in ‘Wetlands’, ‘Not related lands’, ‘Other agricultural land’ and ‘Protected Areas’. The model optimizes over ‘Cropland’, ‘Pasture’, ‘Unmanaged forest’, ‘Managed forest’, ‘Planted forest’ (or short-rotation tree plantation) and ‘Non-productive land’ that have a combined area of ≈585 Mha. The land conversion possibilities for these six land-use classes are restricted through biophysical land suitability and production potential, and through a matrix of endogenous land-use change (Fig. S2). Since the model does not properly include water constraints, we limited agricultural expansion in the Caatinga biome to a 10% increase per time step based on historical expansion.
A C Soterroni et al
4
A new land-use class named ‘Forest regrowth’ was created in GLOBIOM-Brazil to simulate the obligatory native vegetation restoration of Brazil’s Forest Code. Transitions from ‘Cropland’, ‘Pasture’ and ‘Non-productive land’ to ‘Forest regrowth’ are allowed in order to compensate for eventual deficits of legal reserves, but no transitions are allowed from ‘Forest regrowth’ to any other land-use class. The cost of active forest restoration depends on several factors (e.g., biome, topography, occupation history, scale), which vary in space and evolve in time as the recovery process progresses. In the absence of reliable statistics on this matter for all Brazilian biomes, ‘Forest regrowth’ areas are set aside only for passive regrowth. Data from various national and international sources (IBGE vegetation map [19], MODIS land-cover map [20], IBGE 2006 Agricultural Census and yearly crop (IBGE/PAM) and cattle (IBGE/PPM) surveys, SOS Mata Atlˆantica forest remnants, and the network of protected areas, including conservation units, public forests and indigenous lands) have been combined to produce a consistent land-cover/land-use map. The methodology is described in [21]. Internal transportation costs The 2012 National Plan for Logistics and Transportation (PNLT) data were used to collect information on the federal roads and the transportation costs within them. The costs vary from USD 0.35 to USD 1.1665 freight per ton-km. The internal transportation costs are computed at the spatial grid resolution of the model. We use an algorithm based on the generalized proximity matrix (GPM) proposed by Aguiar et al. [22]. We take the centroid of each grid cell as the starting point to compute the costs. In this algorithm, the path from the starting point to the ending point enters the road network only once. The path leaves the road only on the closest location or destination. When there is no road at the start or end points, we estimate an additional cost to enter or leave the roads. The algorithm requires that all roads must be connected. Roads in the Amazon that are not connected to the national network are connected to the nearest road, and they have their associated transportation cost doubled. The shortest paths inside the network are computed using the Dijkstra algorithm [23]. The proximity matrix was computed for the state capitals and for the export ports. Transportation costs range from USD 4.12 per ton to USD 1,000.84 for capitals, and from USD 9.93 per ton to USD 2,238.19 per ton for sea ports. Costs depend on the location of the production area, its connectivity to the road network, and where the goods are consumed. The internal transportation costs in USD per product type (i.e., solid, liquid and grain) and destination (nearest state capital or nearest seaport) were calculated for all products modeled (see Fig. S7). We derived the final transportation costs using the proportions of the internal consumption and the export per product. For example, if Brazil exports 44% of produced soybeans, then the transportation cost for this commodity for each grid cell is 0.44 times the cost of the nearest port plus 0.56 times the cost to the nearest capital.
A C Soterroni et al
5
Probability of enforcement The probability of enforcement of the illegal deforestation control, the most important measure of the Brazil’s Forest Code, is modeled by using the Becker’s [24] standard model of enforcement and the IBAMA’s decision problem as defined in B¨orner et al. [25] in a grid cell of 20 km x 20 km. In our IBAMA’s decision problem, or IBAMA’s enforcement strategy, we are taking into account the Amazon and the Cerrado biomes as the target area; the historical IBAMA’s embargoes occurrence in a grid cell during the period 2000-2014; the cost of visiting a grid cell given by the transportation costs equal to R$8,780 [26] multiplied by the transport time to visit that cell; the IBAMA’s administration cost for each embargo equal to R$2,165 [26]; the IBAMA’s annual operation budget between 2003 and 2008 with 80% (R$40M) allocated to the Amazon biome, 9.34% (R$4.67M) to the Cerrado biome, and 10.66% (R$5.33M) to the rest of Brazil; and the official historical deforestation in a grid cell for the period 2005-2013 from PRODES/INPE. The IBAMA’s enforcement strategy “seeks to minimize illegal deforestation by maximizing area of inspected illegal deforestation” [25]. For the Becker’s standard model of enforcement, the cost of punishment is given by the official deforestation fine equal to R$5,000 per ha. This approach assumes that farmers are aware of the approximate likelihood of getting punished for illegal deforestation, that their expectation is based on historical enforcement. In areas with low deforestation rates, the perceived probability of enforcement (p) is low, but that if deforestation rates spike, this will increase the farmer’s perceived p. Figure S14a shows the spatial distribution of the probability of enforcement as predicted by the IBAMA’s optimal enforcement strategy calculated for the Amazon and the Cerrado biomes, excluding the protected areas, and upscaled to a grid cell of 50 km x 50 km. Areas with a history of high deforestation rates have a higher enforcement probability of being inspected. Scenarios of imperfect illegal deforestation control (IDCImperfect) The scenarios of an imperfect illegal deforestation control use the probability of enforcement (p) as a proxy to allow or not illegal deforestation. Since this probability varies between 0 and 1 for each grid cell, it is directly used in the constraint that controls the illegal deforestation per cell within the model. If p is equal to 1 in a given cell, there is full compliance with the Forest Code and only LR surpluses can be converted (legal deforestation) in that cell. Similarly, if p is equal to zero in a cell, there is no compliance with the Forest Code in that cell and, consequently, no ban on the conversion of forests or native vegetation to any use (legal or illegal deforestation). Finally, if p is a value between 0 and 1 in cell, only a fraction of the available native vegetation (in addition to the LR surplus), namely 1 − p, can be converted in that given cell. For the scenario IDCImperfect1 we use the probability of enforcement p to control the illegal deforestation in the Amazon and the Cerrado biomes at the pixel level. The spatial distribution of p is shown in Fig. S14a. For the scenarios IDCImperfect2 and IDCImperfect3, the probability p
A C Soterroni et al
6
is increased by 25% and 50%, respectively, until the limiting value of 1 is attained (see Fig. S14b and S14c). Emissions calculation Carbon content in the equilibrium state of land-cover classes is used to estimate GHG emissions from land-use changes. Positive and negative emissions are determined by the difference between the carbon content of the original class and that of the new class. Positive emissions are generated by deforestation or native vegetation loss and other land-use changes (e.g., transitions from pasture to cropland, from non-productive land to either cropland or pasture). Afforestation from planted forests and passive restoration by forest regrowth cause negative emissions by removing CO2 from the atmosphere. In this study, positive emissions from deforestation, or native vegetation loss, and negative emissions from forest regrowth are estimated based on the carbon content from the Brazil’s Third Emissions Inventory, used in official communications to the UNFCCC in 2016. The carbon stocks of this map were estimated per vegetation type in each Brazilian biome (national coverage) taking into account values of living above- and belowground biomass; different land use data for the period 1994 to 2010 from projects such as RADAMBRASIL and PROBIO, as well as from the literature; different allometric equations to estimate biomass; and different biomass to carbon conversion factors [27]. For carbon uptake from short-rotation tree plantations, the carbon content is taken from Havl´ık et al. [28]. GLOBIOM-Brazil uses the biomass map of Ruesch and Gibbs [29] for the carbon content of pasture and non-productive land use classes used to estimate positive emissions from other land-use changes. Table S3 summarizes the emissions related to land-use change transitions modeled by GLOBIOM-Brazil. The release of carbon from the terrestrial biosphere to the atmosphere as CO2 occurs in one simulation period (10-year time step) for deforestation and other land-use changes (i.e., other LUCs). In contrast, CO2 removal from the atmosphere by forest regrowth takes several decades. The model accounts for carbon uptake from forest regrowth according to each biome. In the Amazon and the Atlantic Forest biomes, forest regeneration takes 25 years to recover 70% of the original biomass [30, 31]. In the Cerrado, Caatinga and Pantanal biomes, we are assuming that it takes 20 years for these biomes to recover their full biomass content, i.e., 70% in the first decade and 30% in the second decade. As the Pampa has grassland-based vegetation, we assume that regeneration takes 3 years and is completed in one time step (i.e., one decade). These regrowth periods in the Cerrado, Caatinga, Pantanal and Pampa biomes were estimated by the ratio between the global carbon estimates for woody savannahs and grasslands provided by Liu et al. [32] and the average mean annual increment per biome estimated by the G4M model [17]. For the Amazon forest and the Atlantic Forest, we use the growth curve defined in Ramankutty et al. [31]. Table S4 shows the carbon uptake schedule applied for each decade and Brazilian biome regarding passive forest regrowth. The CO2 removal from the atmosphere by short-rotation plantations within GLOBIOM-Brazil takes one simulation period (10 years) to be completed. Emissions estimates
A C Soterroni et al
7
from agriculture were not included in this study. When non-productive land is converted to agricultural use or to short-rotation plantation, we assume that all biomass is released into the atmosphere. Litter, dead wood, and soil organic carbon are not considered. This is the approach that Brazil has adopted to compute the forest reference emission level (FREL) submitted to the UNFCCC [33]. Table S2 summarizes emissions from the land-use change and forestry (LUCF) sector in PgCO2 e/yr per emissions type, as projected by the FC scenario. Validation GLOBIOM-Brazil projections are compared to the 2010 official statistics for validation purposes. These include the harvested areas from IBGE/PAM (Municipal Agriculture Census), livestock numbers from IBGE/PPM (Municipal Livestock Survey), and PRODES/INPE Amazon deforestation map for the period 2001-2010. Overall, a good agreement between census data and simulation results is observed. Differences range from +1% in total crop area for all modeled crops in Brazil to +10% in the Amazon biome, -2% in the Cerrado biome and -2% in the Atlantic Forest biome (see Fig. S8). Figures S9 and S10 show a comparison between area and production, respectively, of different crops. For example, in the 2010 projection, the soybean area is overestimated by 3% while soybean production is underestimated by 1%. Total livestock numbers differ from IBGE data by -3% (see Fig. S11). In 2010, beef production in Brazil differ from IBGE/PPM by -10% (see Fig. S12). The major discrepancies for cattle herd are found in the Caatinga (-2 MTLU) and the Amazon biomes (+1 MTLU) (Fig. S13). Accumulated deforestation from PRODES/INPE between 2001 and 2010 in the Amazon biome amounts to 16.53 Mha, whereas our model projects 16.45 Mha for the same period and region (Fig. 1). Differences are concentrated around the Xingu area and along the BR-163 road in the state of Par´a, and they are probably due to improvements in the local transportation network that have not yet been captured by the model. GLOBIOM-Brazil estimates of CO2 emissions from deforestation in the Amazon between 2001 and 2010 (793 MtCO2 e) are within 9% of those of Brazil’s FREL [33] (872 MtCO2 e) and 4.6% of Aguiar et al. [34] (831 MtCO2 e). GLOBIOM-Brazil estimates of CO2 emissions from the land-use change and forestry (LUCF) sector for all of Brazil (1,189 MtCO2 e) are within 16% of the estimates provided by SEEG [35] (1,421 MtCO2 e) for the period 2001-2010. Acknowledgments Sections “Model framework”, “Land-cover/use map for Brazil”, “Internal transportation costs” and “Emissions calculation” reproduce some parts of the methodology previously published by the authors in [21] under a Creative Commons Attribution-ShareAlike 4.0 International License. Similarly, Figures S1, S3, S4, S5 and Table S3 have been taken from [21]. Figure S2 has been
A C Soterroni et al
8
adapted from [1], published under Creative Commons Attribution-NonCommercialNoDerivatives International Public License (CC BY-NC-ND). 1. References [1] Havlik P, Valin H, Mosnier A, Obersteiner M et al. 2014 Climate change mitigation through livestock system transitions Proc. Natl. Acad. Sci. USA 111(10) 3709–14 [2] Schneider U, McCarl B and Schmid E 2007 Agricultural sector analysis on greenhouse gas mitigation in US agriculture and forestry Agricultural Systems 94(2) 128–40 [3] O’Neill B, Kriegler E, Riahi K, Ebi K and Hallegatte S 2014 A new scenario framework for climate change research: the concept of shared socioeconomic pathways Climate Change 122(3) 387–400 [4] Alexandratos N and Bruinsma J 2012 World agriculture towards 2030/2050: The 2012 revision Tech. rep. Food Agriculture Organization Rome [5] Valin H, Mosnier A, Herrero M, Schmid E and Obersteiner M 2013 Agricultural productivity and greenhouse gas emissions: Trade- offs or synergies between mitigation and food security? Environ. Res. Lett. 8(3) 035019 [6] Bouet A, Decreux Y and Fontagn´e L 2014 A consistent, ad-valorem equivalent measure of applied protection ´ across the world: The MAcMap-HS6 database Tech. rep. Centro d’Etudes Prospectives et d’Informations Internationales, Paris [7] Hummels D 1999 Toward a geography of trade costs Tech. rep. GTAP Working Papers [8] Jansson T and Heckelei T 2009 A new estimator for trade costs and its small sample properties Economic Modelling 26(2) 489–498 [9] OECD/IEA 2010 World energy outlook 2010 Tech. rep. International Energy Agency [10] MME/EPE Cen´ arios de oferta de etanol e demanda do ciclo otto: Vers˜ao estendida 2030. Tech. rep. Minist´erio de Minas e Energia, Empresa de Pesquisa Energ´etica [11] Williams J 1995 The EPIC model. Computer models of watershed hydrology Tech. rep. Water Resources Publications Highlands Ranch, CO [12] Skalsk` y R, Tarasoviˇcov´ a Z, Balkoviˇc J et al. 2008 GEO-BENE global database for bio-physical modeling Tech. rep. GEOBENE project [13] Valin H, Havl´ık P, Forsell N, Frank S, Mosnier A, Pters D, Hamelinck C, Sp¨ottle M and Berg M 2013 Description of the globiom (iiasa) model and comparison with the mirage-biof (ifpri) model Tech. rep. Technical Report, E3tech, ECOFYS [14] Ser´e C, Steinfeld H and Groenewold J 1995 World livestock production systems: current status, issues and trends Consultation on Global Agenda for Livestock Research, Nairobi (Kenya), 18-20 Jan 1995 [15] Notenbaert A, Herrero M, Kruska R et al. 2009 Classifying livestock production systems for targeting agricultural research and development in a rapidly changing world Tech. rep. ILRI [16] Herrero M, Havlik P, Valin H, Notenbaert A, Rufino M, Thornton P et al. 2013 Biomass use, production, feed efficiencies, and greenhouse gas emissions from global livestock systems Proc. Natl. Acad. Sci. USA 110(52) 20888–93 [17] Kindermann G, McCallum I, Fritz S and Obersteiner M 2008 A global forest growing stock, biomass and carbon map based on FAO statistics Silva Fennica 42(3) 387–396 [18] Verissimo A, Rolla A, Vedoveto M and Futada S 2011 Protected areas in the Brazilian amazˆonia: challenges and opportunities Tech. rep. Imazon, Bel´em, Brazil [19] IBGE 2012 Manual t´ecnico da vegeta¸c˜ ao brasileira (technical manual of brazilian vegetation Tech. rep. IBGE, 2nd edition [20] Friedl M, Sulla-Menashe D, Tan B, Schneider A, Ramankutty N, Sibley A and Huang X 2010 MODIS collection 5 global land cover: algorithm refinements and characterization of new datasets Remote Sensing of Environment 114(1) 168–182
A C Soterroni et al
9
[21] Cˆ amara G, Soterroni A, Ramos F, Carvalho A, Andrade P, Souza R, Mosnier A, Mant R, Buurman M, Pena M, Havl´ık P, Pirker J, Kraxner F, Obersteiner O, Kapos V, Affonso A, Esp´ındola G and Bocqueho G 2012 Modelling land use change in Brazil: 2000–2050 Tech. rep. INPE/S˜ao Jos´e dos Campos, IPEA/Bras´ılia, IIASA/Laxenburg, UNEP-WCMC/Cambridge. This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. [22] Aguiar A, Camara G, Monteiro A and Souza R 2003 Modelling spatial relations by generalized proximity matrices Tech. rep. [23] Dijkstra E 1959 A note on two problems in connection with graphs Numerische Mathematik 1(1) 269–71 [24] G B 1968 Crime and punishment: an economic approach Journal of Political Economy 76 169–217 [25] B¨ orner J, Marinho E and Wunder S 2015 Mixing carrots and sticks to conserve forests in the Brazilian Amazon: a spatial probabilistic modeling approach PlosOne pp. 1–20 [26] B¨ orner J, Wunder S, Wertz-Kanounnikoff S, Hyman G and Nascimento N 2014 Forest law enforcement in the Brazilian Amazon: Costs and income effects Glob Environ Change 29 294–305 [27] MCTI 2014 Estimativas anuais de emiss˜ oes de gases de efeito estufa no Brasil Tech. rep. Brazilian Ministry of Science and Technology, Bras´ılia [28] Havlik P, Schneider U, Schmid E, Bottcher H, Fritz S, Skalsky R, Aoki K, De Cara S, Kindermann G, Kraxner F, Leduc S, McCallum I, Mosnier A, Sauer T and Obersteiner M 2011 Global land-use implications of first and second generation biofuel targets Energy Policy 39(10) 5690–5702 [29] Ruesch A and Gibbs H 2008 New IPCC tier-1 global biomass carbon map for the year 2000 [30] Houghton R, Skole D, Nobre C, Hackler J, Lawrence K and Chomentowski W 2011 Annual fluxes of carbon from deforestation and regrowth in the Brazilian Amazon Nature 403(6767) 301–304 [31] Ramankutty N, Gibss H, Achard F, Defries R, Foley J and Houghton R 2007 Challenges to estimating carbon emissions from tropical deforestation Glob Change Biol 13 51–66 [32] Liu Y, van Dijk A, de Jeu R, Canadell J, McCabe M, Evans J and Wang G 2015 Recent reversal in loss of global terrestrial biomass Nature Climate Change 5 470–474 [33] MMA 2013 Brazil’s submission of a forest reference emission level for deforestation in the Amazonia biome for results-based payment for REDD+ under the UNFCCC Tech. rep. Brazilian Ministry of Environment, Bras´ılia [34] Aguiar A, Ometto J, Nobre C, Lapola D, Almeida C, Vieira I, Soares J, Alvala R, Saatchi S, Valeriano D et al. 2012 Modeling the spatial and temporal heterogeneity of deforestation-driven carbon emissions: the INPE-EM framework applied to the Brazilian Amazon Glob. Chang. Biol. 18(11) 3346–3366 [35] Acessed: 01 Nov. 2016 Sistema de estimativas de emiss˜ao de gases de efeito estufa Available at: http://plataforma.seeg.eco.br/total emission [36] Soares-Filho B, Raj˜ ao R, Macedo M, Carneiro A, Costa W, Coe M, Rodrigues H and Alencar A 2014 Cracking Brazil’s Forest Code Science 344(6182) 363–364 [37] Guidotti V, Freitas F, Sparovek G, Pinto L, Hamamura C, Carvalho T and Cerignoni F 2017 N´ umeros detalhados do novo c´ odigo florestal e suas implica¸c˜oes para os pras Tech. rep. IMAFLORA
A C Soterroni et al
10
Figure S1. Brazil’s six biomes: Amazon (mainly tropical rain forest), Cerrado (tropical savanna), Caatinga (semi-arid deciduous shrubland and semi-deciduous dry forests), Atlantic Forest (tropical and subtropical forest, much depleted), Pantanal (extensive wetlands) and Pampa (mainly natural grassland). Protected areas are indicated in light green superposing the biomes and include federal, state and municipal conservation units and indigenous lands. The network of protected areas accounts for ≈243 million hectares, and the private properties cover ≈572 million hectares. Source: Reproduced from Cˆ amara et al. (2015) [21] CC BY-SA 3.0.
A C Soterroni et al
11
Figure S2. Matrix of endogenous land-use and land-cover changes allowed in GLOBIOM-Brazil and an illustration of possible model grid cell land use compositions. For Brazil, the model optimizes on a uniform grid of 0.5◦ by 0.5◦ amounting to 3,001 pixels with a spatial resolution of approximately 50 km x 50 km at the equator. The land-use class “Forest Regrowth” was created in GLOBIOMBrazil to model the obligatory native vegetation restoration of Brazil’s Forest Code. The dashed arrows in the matrix of endogenous land-use and land-cover changes represent land abandonment. Source: Adapted from Havl´ık et al. [1].
A C Soterroni et al
12
Figure S3. Overview of the methodology used to create a consistent land cover/land use map for Brazil. The complete description of this methodology can be found in the REDD-PAC Technical Report [21]. Source: Reproduced from Cˆamara et al. (2015) [21] CC BY-SA 3.0.
A C Soterroni et al
0
100
13
200
300
(a)
400
500
0
100
200
300
400
500
(b)
Figure S4. Spatial distribution of the transportation costs per pixel for (a) soybeans and (b) beef. Color bar values are expressed in USD per ton per cell. The exchange rate is USD 1.00 = BRL 1.954. Source: Reproduced from Cˆamara et al. (2015) [21] CC BY-SA 3.0.
Figure S5. Spatial distribution of the Legal Reserve (LR) requirements from Soares et al. [36] downscaled to 50 km x 50 km pixels. Source: Reproduced from Cˆamara et al. (2015) [21] CC BY-SA 3.0.
A C Soterroni et al
0
14
50
100
150
200
(a) Native vegetation
250
300
0
50
100
150
200
250
300
(b) LR surpluses
Figure S6. Spatial distribution of the (a) native vegetation stocks in 2010, and the (b) LR surpluses of native vegetation that could be legally deforested according to the Forest Code. In the Amazonas state, we consider only 20% of the forest surpluses.
Land use/cover classes Unmanaged Forest
Managed forest Forest regrowth Planted forest
Cropland
Pasture Non-productive Other agricultural Wetland Not relevant
Description This class covers all unmanaged forests and native vegetation. Both the evergreen rain forest of Amazon, the deciduous forest of the Caatinga, and woody savannas of the Cerrado are included in this class. These are forests that are exploited in a sustainable way. Transitions from mature to managed forest are not considered deforestation in the model. Passive forest and other native vegetation regrowth necessary to compensate legal reserve deficits. These are short-rotation plantations, with single or few species and uniform planting density, that are used by the wood and paper industries. Brazil has a significant number of planted forests with pinus and eucalyptus species. Areas planted with one of the 18 GLOBIOM crops: barley, dry beans, cassava, chick peas, corn, cotton, groundnut, millet, potatoes, rapeseed, rice, soybeans, sorghum, sugarcane, sunflower, sweet potatoes, wheat, and oil palm. GLOBIOM crops cover 86% of the total cultivated area in Brazil in 2000 according to IBGE/PAM. Areas used for livestock ranching. Anthropized natural vegetation mosaic areas not currently under production Areas planted with crops not modelled by GLOBIOM. In Brazil, these include coffee and fruit tress, for example. Areas with permanent water cover, or areas that are regularly flooded. Bare areas, water bodies, snow and ice.
Table S1. Description of GLOBIOM-Brazil land use/cover classes.
A C Soterroni et al
0
20
15
40
60
80
100
0
20
(a) APP debts
40
60
80
100
(b) LR debts
Figure S7. Environmental debts of (a) Areas of Permanent Preservation (APP debts) and (b) Legal Reserves (LR debts) based on the Rural Environmental Registry (Portuguese acronym: CAR) from December 2016 and consolidated by Guidotti et al. [37]. APP debts account to ≈ 8 million hectares, mainly in the Atlantic Forest biome, and LR debts account to ≈ 11 million hectares, mainly in the Cerrado and the Amazon biomes. The data from Guidotti et al. [37] is available at municipality level and it was downscaled to the model gridcell (about 50 km x 50 km at the equator). According to the Forest Code, the APP debts should be fully recovered and the LR debts could be compensated by the environmental reserve quotas (CRA). Color bar values are expressed in thousands of hectares per cell.
LUCF Emissions (PgCO2 e/yr)
FC 2010
2020
2030
2040
2050
0.972
0.250
0.197
0.125
0.067
0
0
-0.161
-0.124
-0.063
Afforestation
-0.012
-0.018
-0.019
-0.024
-0.025
Other LUC
0.228
0.351
0.143
0.071
0.083
Net LUCF
1.188
0.583
0.160
0.049
0.062
Deforestation Forest Regrowth
Table S2. Emissions from land-use change and forestry (LUCF) sector in PgCO2 e/yr. Emissions are broken down by emissions type: deforestation, forest regrowth, afforestation and other land-use change (Other LUC).
A C Soterroni et al
16 IBGE/PAM
GLOBIOM-‐Brazil
28
Crop area (Mha)
24 20 16 12 8 4 0 Amazon
Cerrado
Atlan4c Forest
Caa4nga
Pampa
Pantanal
Figure S8. Crop area comparison per biome between GLOBIOM-Brazil projections and IBGE/PAM for the year 2010. 1 Mha = 104 km2 .
IBGE/PAM
GLOBIOM-‐Brazil
30
Crop area (Mha)
25 20 15 10 5 0 Soybeans Maize Sugarcane
Dry Beans
Rice
Wheat
Cassava
Figure S9. Crop area comparison per crop type between GLOBIOM-Brazil projections and IBGE/PAM for the year 2010 at national level. 1 Mha = 104 km2 .
A C Soterroni et al
17 IBGE/PAM
GLOBIOM-‐Brazil
80 Crop produc)on (Mton)
70 60 50 40 30 20 10 0 Soybeans Maize
Cassava
Wheat
Pota
Rice
Dry Beans
Figure S10. Crop production comparison per crop type between GLOBIOM-Brazil projections and IBGE/PAM for the year 2010 at national level. 1 Mton = 106 ton.
IBGE/PPM
GLOBIOM-‐Brazil
156
Livestock (MTLU)
130 104 78 52 26 0 Bovines
Sheep & Goats
Pigs
Poultry
Figure S11. Livestock comparison between GLOBIOM-Brazil projections and IBGE/PPM for the year 2010. Values are expressed in million tropical livestock units (MTLU). Tropical livestock units (TLU) are used to provide a standardized measure of livestock biomass. 1 TLU = 0.7 cattle; 0.25 pig; 0.10 sheep or goat; 0.01 poultry.
A C Soterroni et al
18
9
Beef ca'le (Mton)
8 6 IBGE/PPM
5
GLOBIOM-‐Brazil
3 2 0 2010
(a)
(b)
Figure S12. Brazil’s aggregated numbers for (a) cattle herd in million tropical livestock unit (MTLU), and (b) beef cattle production in million tons (Mton) in 2010. 1 MTLU = 0.7 cattle head.
IBGE/PPM
GLOBIOM-‐Brazil
Ca#le heads (MTLU)
55 46 37 27 18 9 0 Amazon
Cerrado
Atlan8c Forest
Caa8nga
Pampa
Pantanal
Figure S13. Cattle herd comparison per biome between GLOBIOM-Brazil projections and IBGE/PPM for the year 2010. 1 TLU = 0.7 cattle heads.
A C Soterroni et al
0.2
0.4
0.6
0.8
1.0
0.0
0.2
(a) IDCImperfect1
0.4
0.6
0.8
1.0
0.0
(b) IDCImperfect2
0.2
0.4
0.6
0.8
1.0
(c) IDCImperfect3
Figure S14. Spatial distribution of probability of enforcement used in the scenarios (a) IDCImperfect1, (b) IDCImperfect2 and (c) IDCImperfect3 after 2010. Scenario abbreviations: IDCImperfect1 = partial illegal deforestation control in the Amazon and the Cerrado biomes, full control in the Atlantic Forest biome and no control in the rest of Brazil, and no forest restoration; IDCImperfect2 = partial illegal deforestation control increased by 25% in the Amazon and the Cerrado biomes, full control in the Atlantic Forest biome and no control in the rest of Brazil, and no forest restoration; IDCImperfect3 = partial illegal deforestation control increased by 50% in the Amazon and the Cerrado biomes, full control in the Atlantic Forest biome and no control in the rest of Brazil, and no forest restoration.
Amazon ● Cerrado
Caatinga
AtlanticForest
Brazil
2.0
Cattle productivity (TLU/ha)
0.0
19
1.5
1.0 ● ●
0.5
●
●
2010
2020
●
0.0 2030
2040
2050
Year
Figure S15. Evolution of cattle productivity (TLU/ha) in Brazil and its main biomes as projected by the FC scenario. Abbreviation: FC = Forest Code fully implemented scenario. 1 TLU = 0.7 cattle heads.
A C Soterroni et al
20 NoFC ● IDCAmazon
IDCBrazil
FCnoCRA
FC
Other non−forest land (Mha)
120
●
80
● ●
● ●
40
0 2010
2020
2030
2040
2050
Year
Figure S16. Evolution of other non-productive land not currently under production as projected by the NoFC, IDCAmazon, IDCBrazil, FC and FCnoCRA scenarios. Scenario abbreviations: NoFC = no illegal deforestation control except in the Atlantic Forest biome and no forest restoration, i.e., no enforcement of the Forest Code; IDCAmazon = illegal deforestation control in the Amazon and the Atlantic Forest biomes with no forest restoration; IDCBrazil = illegal deforestation control everywhere in Brazil with no forest restoration; FC = Forest Code fully implemented, i.e., illegal deforestation control, compensation of LR debts by the environmental reserve quotas (CRA) and forest restoration; FCnoCRA = FC scenario without the CRA. 1 Mha = 104 km2 . Managed
Non-‐managed 250
200
200
150
150
Pasture (Mha)
Pasture (Mha)
Non-‐managed 250
100
50
0 2020
Managed
100
50
2030
(a) NoFC
2040
2050
0 2020
2030
2040
2050
(b) FC
Figure S17. Distribution of managed and non-managed pastures between 2020 and 2050 as projected by the (a) NoFC and the (b) FC scenarios. Abbreviations: NoFC = no illegal deforestation control except in the Atlantic Forest biome and no forest restoration, i.e., no enforcement of the Forest Code; FC = Forest Code fully implemented, i.e., illegal deforestation control, compensation of LR debts by the environmental reserve quotas (CRA) and forest restoration. 1 million ha = 104 km2 .
A C Soterroni et al
0
50
21
100
150
200
250
(a) FC
300
0
50
100
150
200
250
300
(b) FCnoCRA
Figure S18. Spatial distribution of the (a) LR debts offset by the quotas mechanism (CRA), and the (b) LR surpluses used to offset the LR debts. Only cells with deficits overlapping soybeans and sugarcane areas are taken into account to be compensated. The total amount of LR debts and surpluses traded by the CRA is ≈6 Mha. All the surpluses used in the CRA system are prevented from legal deforestation. The border of the MATOPIBA region is indicated in red. Color bar values are expressed in thousands of hectares per cell. Abbreviations: LR = Legal Reserve; FC = Forest Code fully implemented, i.e., illegal deforestation control, compensation of LR debts by the environmental reserve quotas (CRA) and forest restoration; FCnoCRA = FC scenario without the CRA; Matopiba = a region in the states of Maranh˜ao, Tocantins, Piau´ı and Bahia along the border between the Cerrado and the Caatinga biomes. 1 Mha = 104 km2 .
Emissions Sign Action Deforestation Positive Other LUC
Afforestation Negative Reforestation
Land Use Transitions From To Unmanaged Forest Cropland Unmanaged Forest Pasture Pasture Cropland Non-productive Cropland Non-productive Pasture Cropland Planted Forests Pasture Planted Forests Non-productive Planted Forests Cropland Forest Regrowth Pasture Forest Regrowth Non-productive Forest Regrowth
Table S3. Land-use change transitions and associated emissions taken into account in GLOBIOMBrazil model. Source: Reproduced from Cˆamara et al. (2015) [21] CC BY-SA 3.0.
A C Soterroni et al
0
20
22
40
60
(a) FC (12.9 Mha)
80
100
0
20
40
60
80
100
(b) FCnoCRA (18.7 Mha)
Figure S19. Spatial distribution of native vegetation restoration in 2050 as projected by (a) FC and (b) FCnoCRA scenarios. Color bar values are expressed in thousands of hectares per cell. Scenarios abbreviations: FC = Forest Code fully implemented, i.e., illegal deforestation control, compensation of LR debts by the environmental reserve quotas (CRA) and forest restoration; FCnoCRA = FC scenario without the CRA; 1 million ha = 104 km2 .
Brazilian biomes Cerrado Decades Amazon Caatinga Atlantic Forest Pantanal First 28% 70% Second 28% 30% Third 17% Fourth 6% Fifth 6% -
Pampa 100% -
Table S4. Carbon uptake schedule per biome applied for passive forest regrowth in GLOBIOMBrazil.
A C Soterroni et al
−300
−250
23
−200
−150
−100
−50
0
50
100
−300
−250
(a) NoFC
−300
−250
−200
−150
−100
−200
−150
−100
−50
0
50
100
0
50
100
(b) IDCImperfect1
−50
(c) IDCImperfect2
0
50
100
−300
−250
−200
−150
−100
−50
(d) IDCImperfect3
Figure S20. Spatial distribution of cumulative loss (orange) or gain (blue) of native vegetation for the scenarios (a) NoFC, (b) IDCImperfect1, (c) IDCImperfect2 and (d) IDCImperfect3 between 2010 and 2050. Color bar values are expressed in thousands of hectares per cell. Scenario abbreviations: NoFC = no implementation of the Forest Code; IDCImperfect1 = partial illegal deforestation control in the Amazon and the Cerrado biomes, full control in the Atlantic Forest biome and no control in the rest of Brazil, and no forest restoration; IDCImperfect2 = partial illegal deforestation control increased by 25% in the Amazon and the Cerrado biomes, full control in the Atlantic Forest biome and no control in the rest of Brazil, and no forest restoration; IDCImperfect3 = partial illegal deforestation control increased by 50% in the Amazon and the Cerrado biomes, full control in the Atlantic Forest biome and no control in the rest of Brazil, and no forest restoration; FC = Forest Code fully implemented, i.e., illegal deforestation control, compensation of LR debts by the environmental reserve quotas (CRA) and forest restoration.
A C Soterroni et al NoFC
24 IDCImperfect1
IDCImperfect2
IDCImperfect3
●
FC
NoFC
IDCImperfect1
IDCImperfect2
IDCImperfect3
●
FC
260 250
Pasture (Mha)
●
● ●
●
220
200
Cattle herd (MTLU)
● ●
240
●
200 ●
●
150
●
100 2010
2020
2030
Year
(a) Pasture
2040
2050
2010
2020
2030
2040
2050
Year
(b) Cattle herd
Figure S21. (a) Pasture area evolution and (b) Cattle herd in Brazil as projected by the FC, NoFC, IDCImperfect1, IDCImperfect2 and IDCImperfect3. Scenario abbreviations: NoFC = no implementation of the Forest Code; IDCImperfect1 = partial illegal deforestation control in the Amazon and the Cerrado biomes, full control in the Atlantic Forest biome and no control in the rest of Brazil, and no forest restoration; IDCImperfect2 = partial illegal deforestation control increased by 25% in the Amazon and the Cerrado biomes, full control in the Atlantic Forest biome and no control in the rest of Brazil, and no forest restoration; IDCImperfect3 = partial illegal deforestation control increased by 50% in the Amazon and the Cerrado biomes, full control in the Atlantic Forest biome and no control in the rest of Brazil, and no forest restoration; FC = Forest Code fully implemented, i.e., illegal deforestation control, compensation of LR debts by the environmental reserve quotas (CRA) and forest restoration.
