Assessing the effects of agricultural intensification on natural habitats and biodiversity in Southern Amazonia

RESEARCH ARTICLE Assessing the effects of agricultural intensification on natural habitats and biodiversity in Southern Amazonia Jan Go¨ pelID*☯, Jan Schu¨ ngel‡, Benjamin Stuch‡, Ru¨ diger Schaldach☯ Center for Environmental Systems Research (CESR), University of Kassel, Kassel, Germany ☯ These authors contributed equally to this work. ‡ These authors also contributed equally to this work. * jan.goepel@usf.uni-kassel.de Abstract The ongoing trend toward agricultural intensification in Southern Amazonia makes it essen- tial to explore the future impacts of this development on the extent of natural habitats and biodiversity. This type of analysis requires information on future pathways of land-use and land-cover change (LULCC) under different socio-economic conditions and policy settings. For this purpose, the spatially explicit land-use change model LandSHIFT was applied to calculate a set of high-resolution land-use change scenarios for the Brazilian states Para and Mato Grosso. The period of the analysis were the years 2010–2030. The resulting land- use maps were combined with maps depicting vertebrate species diversity in order to exam- ine the impact of natural habitat loss on species ranges as well as the overall LULCC- induced effect on vertebrate diversity as expressed by the Biodiversity Intactness Index (BII). The results of this study indicate a general decrease in biodiversity intactness in all investigated scenarios. However, agricultural intensification combined with diversified envi-

ronmental protection policies show least impact of LULCC on vertebrate species richness and conservation of natural habitats compared to scenarios with low agricultural intensifica- tion or scenarios with less effective conservation policies. 1. Introduction Human induced changes to the biosphere have caused severe losses of biodiversity [1, 2]. Important factors leading to a loss of natural habitats are land-use and land-cover changes (LULCC) in particular due to the expansion and intensification of agriculture [3]. Martinelli et al. [4] show that, the growth of Brazil’s agricultural sector from the 1970s until the end of the first decade of this century was an important driver of massive deforestation. During that period. an area of 18.8% of the original Brazilian Amazon biome, that is distin- guished as a biodiversity hotspot [5], has been cut down [6] and converted to a large extend to cropland and pasture posing a major threat to terrestrial biodiversity [7, 8]. Agriculture plays an important role in regard to Brazils GDP (6.1%) [9]. Even more important, a share of 39% of Brazils exported goods are agricultural commodities, and products [10]. This strong PLOS ONE PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 1 / 21 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Go¨pel J, Schu¨ngel J, Stuch B, Schaldach R (2020) Assessing the effects of agricultural intensification on natural habitats and biodiversity in Southern Amazonia. PLoS ONE 15(11): e0225914. https://doi.org/10.1371/journal. pone.0225914

Editor: Mirko Di Febbraro, University of Molise, Isernia, ITALY Received: November 15, 2019 Accepted: July 2, 2020 Published: November 25, 2020 Peer Review History: PLOS recognizes the benefits of transparency in the peer review process; therefore, we enable the publication of all of the content of peer review and author responses alongside final, published articles. The editorial history of this article is available here: https://doi.org/10.1371/journal.pone.0225914 Copyright: © 2020 Go¨pel et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: The data is available at https://doi.org/10.6084/m9.figshare.12014880. v1.

contribution of the agricultural sector to Brazil’s overall economic performance has had posi- tive impacts on social prosperity in the country. According to the Vieira et al. [11], the income of 29 million people has been considerably increased, lifting them out of poverty. Despite these positive numbers, the global demand for agricultural products is projected to continuously rise over the coming decades [12] driven by global population growth and increasing per capita demand for food, fodder, energy crops, and timber [13, 14]. Moreover, changes in food con- sumption patterns likely further increase food demands per capita [9]. These developments will most likely lead to further expansion and intensification of agricultural area in Brazil at the expense of natural habitats and biodiversity [15]. Among scholars, there is a controversial discussion regarding the impact of agricultural intensification on habitat and biodiversity loss. Some authors argue that intensification is key for a further increase in productivity [16], whereby the future destruction of natural habitats can be avoided by slowing down the expansion of agricultural land e.g. [17–19]. In contrast, there is the argument that agricultural intensification might even foster area expansion due to the so-called “rebound effect” [20] or increasing competiveness of agriculture and, thus higher attainable revenues [21]. The latter argument may only be applicable to situations where com- modities with high demand elasticity are involved [22].

Simulation models and scenarios are effective tools to explore current and future land-use changes and to enhance the scientific understanding of their dynamics and drivers. A number of studies examine land-use changes in the Amazon region by employing different models e.g. [23–30]. Moreover, several studies assess the impacts of land-use changes on biodiversity in the tropics e.g. [31–33]. However, Chaplin-Kramer et al. [34] focus on scenarios of total forest conversion without taking into account specific socioeconomic developments. Laurance et al. [15] highlight the general causes of a loss of tropical nature and their implications without identifying specific regions in which conservation efforts might be needed to increase. Others focus their work on a single specific species as an indicator for biodiversity [2, 32, 33]. This approach has its eligibility but is limited to the perspective of that single species while, for instance, Ritter et al. [35] advise against the use of one taxonomic group as an indicator for species richness across different taxonomic groups or, in other words, biodiversity [36]. New- bold et al. [2] adopt a rather broad perspective in terms of methodology as well as spatial extent, thus they cannot identify specific drivers of biodiversity reduction or offer specific rec- ommendations on how to confront possible implications. Widely used indicators for quantify- ing the effect of land-use change on biodiversity are the Mean Species Abundance (MSA) [37] and the Biodiversity Intactness Index (BII) [38]. Both indicators determine the impacts of

land-use and its intensity level on species diversity and species abundance (as important aspects of biodiversity) and therefore are suitable tools for capturing the effects of agricultural intensification. Originally, the MSA was developed for global-scale analyses while the first application of BII was for a large-scale case study in Southern Africa (e.g. [39]). More recent articles propose the BII as a suitable indicator to measure the loss of species diversity in large- scale assessments [34, 40], in particular related to the Planetary Boundary concept [41]. The BII provides information to what extent vertebrate species abundance associated with each single grid-cell (900m x 900m) is influenced by LULCC by relating the average abun- dance of groups of organisms in current times (influenced by LULCC) to their average abun- dance in pre-industrial times [38]. This has the advantage of giving a broad overview of the effect of LULCC on vertebrate diversity on large to global scales. Lamb et al. [42] suggest that especially species abundance based indices perform very well as indicators for biodiversity change [42]. However, this broad view goes hand in hand with some downsides. One is the assumed (expert-based) and relatively coarse depiction of the degradation of natural habitats PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 2 / 21 Funding: This study has been conducted as part of the Carbiocial project (funding reference number

01LL0902A-01LL0902N) commissioned by the German Federal Ministry of Education and Research (https://www.bmbf.de/). J.G., J.S. received respective funding. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.

expressed by the population impact, a factor that summarizes the total impact of the anthropo- genic interference in natural ecosystems [43]. Main objective of the paper is to assess the potential effects of future agricultural intensifica- tion on habitat and biodiversity loss in the federal states Mato Grosso (MT) and Para´ (PA) in Brazil. For this purpose, we analyze a set of spatially explicit LULCC scenarios with a time horizon until 2030, characterized by different assumptions regarding agricultural and socio- economic development in the region as well as by different environmental policies. The first part of our study determines the effects of the loss of natural habitat due to the expansion of agricultural area on the distribution ranges of vertebrate species. In the second part of our analysis we use the BII as an indicator to evaluate the combined effects of expansion and inten- sification of agriculture. 2. Material and methods 2.1. Study area This study focusses on the two Brazilian federal states MT and PA (see Fig 1). These states dif- fer greatly in respect to their recent agricultural developments and their level of exploitation of natural habitats due to the Brazilian agricultural development frontier running through this region [44, 45]. PA has an area of 1.25 million km2 and a population of 8 million people [46]. Only 11,969 km2 of the land is used for soybean cultivation [46]. In 2015 1,881 km2 were deforested which is about the same amount of deforestation as in 2014 [6]. The dominant land use sector is cattle

ranching with a total herd size of 19.2 million animals [46]. A hot spot of LULCC is along the Cuiaba´-Santarem highway (BR-163), the most recent of the “development highways” which are used to acquire the agriculturally rather underdeveloped northern parts of Brazil for crop cultivation and cattle ranching [47]. The natural vegetation is dominated by dense rainforest [11] covering about 77.6% of the state’s area according to MODIS land cover data [48]. More than 40,000 vascular plant species can be found here, of which 30,000 are endemic [11]. Over 1,000 bird species are harbored in the Amazon biome [11] as well as a high concentration of mammals, of which many are endemic, especially along the courses of the rivers crossing this biome [5]. Of the 875 amphibian species in the country, approximately 50% are concentrated in the Amazon biome [5]. Especially here the potential for a loss of vertebrate diversity is high due to a high density of endemic, threatened, and small ranged species [5] as well as ongoing and expected future agricultural expansion [15]. MT has an area of 907,000 km2 and a population of 3.2 million inhabitants [46]. 69,807 km2 of land is used for soybean cultivation [46] and 1,508 km2 were deforested in 2015 which con- stitutes an increase of 16% in comparison to 2014 [6]. Another dominant land use sector is cat- tle ranching with a total herd size of 28.4 million animals [46]. Here the expansion of area used for soybean cultivation and cattle ranching could be identified as the primary cause of conver-

sion of natural ecosystems to agricultural land [49, 50]. In comparison to PA, MT is more con- solidated in terms of agricultural expansion. In recent years, the steep decline of availability of highly productive farmland, policies to curb deforestation and rising land prices have led to a development toward agricultural intensification and away from agricultural expansion [18, 51, 52]. MT is covered by two Brazilian biomes, the Brazilian Cerrado and the Amazon rainforest [5, 45]. Here, 7,000 plant species, of which 44% are endemic to the Cerrado, can be found. The Cerrado biome is especially rich in bird diversity with 837 species, which resembles 49% of all bird species found in Brazil. Also 150 reptile species (50% of all Brazilian reptile species) and 180 amphibian species (28% endemic to the Cerrado) are found here. The Cerrado biome and its waterways are home to 1,400 fish species, 40% of all fish species occurring in Brazil [53]. PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 3 / 21

Fig 1. Overview map of the study region (Para´ and Mato Grosso in Brazil). https://doi.org/10.1371/journal.pone.0225914.g001 PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 4 / 21

2.2. Modeling and assessment protocol LULCC scenarios (section 2.3) were generated with the spatially explicit LandSHIFT model. The model is fully described in [54] and has been tested and validated in different case studies for Brazil [26, 55, 56]. It is based on the concept of land-use systems [57] and couples compo- nents that represent the respective anthropogenic and environmental sub-systems. In our study, LULCC was simulated on a raster with the spatial resolution of 900m x 900m that covers the territories of the federal states of MT and PA. A Table elaborately describing the input data used to initialize and run the model can be found in [58]. The LandSHIFT model generates digital maps for 2010 until 2030 in 5-year time steps that depict the resulting LULCC. For further analysis, we aggregated the land-use types used by the model (Table 1): The 12 crop types [54] were aggregated into the land-use class cropland, the 5 forest types [48] into the class rainforest, and the 2 savannah types [48] into the class Cerrado. Changes in location and area of the respective land-use types were determined by comparing the maps for 2010 and 2030 using GIS software. As a second step, we merged the simulated land-use maps of each calculated scenario with maps of vertebrate species distribution ranges [5] regarding three taxa and three categories by overlaying the land-use and land-cover maps with spatially explicit maps of vertebrate species distribution ranges, again using GIS software. Through this, we could correlate the species dis-

tribution ranges to natural habitats. Further, this enables quantifying the impact of simulated LULCC on natural habitat area (conversion of natural habitats) known to domicile vertebrate species. Finally, we calculated the Biodiversity Intactness Index (BII) for the reference year 2010 and 2030 according to eq 1. This was accomplished by assigning each land-use type a specific population impact and multiplying this population impact by the number of grid cells covered by the respective land-use type and the vertebrate species abundance (per taxon and category) associated with these grid cells. 2.3. Land-use scenarios In order to explore agricultural intensification and expansion in respect to different socio-eco- nomic and policy assumptions, 4 scenarios have been employed for modeling land use change. These scenarios have been developed during an interdisciplinary research project (CarBioCial; www.carbiocial.de) thematically covering the study area (MT, PA). They describe plausible future development pathways of Southern Amazonia until the year 2030. Each scenario con- sists of a qualitative part (storyline) that provides a short narrative of the future world and a set of quantitative information that describe the respective main drivers of LULCC [59, 60]. The storylines are elaborated by [61]. Table 1. Aggregation of LandSHIFT land-use types. LandSHIFT land-use types aggregated land-use types evergreen needle forest, evergreen broad-leafed forest, deciduous needle forest, deciduous broad-leafed forest, mixed forest rainforest

closed shrub land, open shrub land shrub land woody savannah, savannah savannah (Cerrado) tea, cocoa, coffee, maize, annual oil crops, pulses, rice, tropical roots and tubers, soybean, sugarcane, cassava, wheat cropland rangeland, pasture pasture https://doi.org/10.1371/journal.pone.0225914.t001 PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 5 / 21

The following paragraphs briefly describe the central assumptions of the scenarios. For a comprehensive overview of the quantitative scenario assumptions (crop production, crop yield, population, and livestock) see [56]. An elaborate description and discussion of the trans- lation process from qualitative to quantitative information and the respective results for all sce- nario assumptions is described in Scho¨nenberg et al. [62]. The Trend scenario describes a growing demand for agrarian products based on an extrapo- lation of growth trends from 1973 to 2000 specific for each modelled crop. Furthermore, it is assumed that environmental policies like the Brazilian Forest Code or the Soy- and Cattle Moratorium will not be implemented. Only the illegal conversion of natural habitats (pro- tected areas) is prohibited due to good law enforcement. The technological development of agricultural practices in the study area includes an intensification of agricultural production through increasing crop yields. The possibility to intensify pasture management is not consid- ered in this scenario. Two intensification scenarios (Legal Intensification and Illegal Intensification) assume a growing demand for agrarian products (see Trend Scenario) further reinforced by population and GDP growth generated in Asian countries (as export markets). The technological develop- ments of agricultural practices in the study area include a high degree of agricultural intensifi- cation including the intensification of pasture management. Both scenarios presume the

intensification of cattle ranching. In PA, we assume an intensification rate of 4.5% per time step up to a maximum of 30%. That means that the biomass productivity of any pasture grid cell is increased by 4.5% until biomass productivity is 30% higher than in the base year. As agriculture in MT is presumed to be more mechanized, large scale, and world market oriented [63, 64], we assume an intensification rate of 9% up to a maximum value of 50%. These assumptions are based on observed pasture intensification rates in Brazil. According to Wint et al. [65] and Lapola et al. [17], the stocking density of pastures in Brazil rose continuously from 1990 to 2010, with a total increase of 45% during that period. The two scenarios vary in terms of environmental law enforcement. While the Legal Intensification scenario assumes compliance with environmental policies (environmental protected areas, Brazilian Forest Code), the Illegal Intensification scenario presumes noncompliance with environmental law expressed as the defiance of environmental protected areas concerning agricultural expansion and the noncompliance with the Brazilian Forest Code. This scenario assumes the possibility to convert land that is under conservation (e.g. nature reserves), thus opening up spaces that are not allowed for conversion in all other scenarios. The Sustainable Development scenario describes a new social model. This new model includes citizenship, an inclusive economic system, clear land tenure rights, and strong law

enforcement including participatory monitoring of deforestation. Furthermore, it portrays a substantial change in terms of anthropogenic consumption pattern, away from a meat oriented diet toward a healthy and sustainable diet as proposed by the WHO [66, 67] including further intensification of crop production. Moreover, the conversion of areas classified as covered by rainforest into agricultural area is not allowed according to the assumptions of the scenario. 2.4. Maps of vertebrate diversity and Biodiversity Intactness Index (BII) We use maps of vertebrate diversity covering the whole area of Brazil [5] to illustrate the over- lapping of areas of vertebrate diversity and simulated LULCC in each investigated scenario. The species diversity maps were generated by deriving polygon range data concerning birds from BirdLife International and NatureServe [68] and polygon range data concerning mam- mals and amphibians from the International Union for the Conservation of Nature [69]. These polygon range datasets were rendered at a spatial resolution of 10×10 km in order to PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 6 / 21

produce species diversity maps considering these three groups of terrestrial vertebrates in Bra- zil [5]. These groups were further subdivided into the categories small-ranged species, threat- ened species, and endemic species. Small-ranged species were defined as those species that have a range smaller than the median for that taxon (2,250,813 km2 for birds, 1,230,901 km2 for mammals, 66,979 km2 for amphibians) in Brazil. For example, a bird species is considered to be small-ranged by occurring naturally in a range of less than 2,250,813 km2, which resem- bles the median distribution range for that taxon in Brazil. Threatened species were defined as vulnerable, endangered, or critically endangered according to the IUCN Red List [70]. Consid- ered threatened species are globally threatened as the use of taxanomies was inconsitent in the national RedList [5]. Endemic species were defined as having at least 90% of their range within Brazil and no part of their range extending more than 50 km beyond the Brazilian border. Overall, 1703 bird species, 637 mammal species and 875 amphibian species were considered in this study. We decided against the inclusion of the category total species richness in our assess- ment. Total species richness as an indicator for biodiversity can be misleading as it is mainly driven by wide-ranged species [5, 71] which might even be benefit from degraded habitats [72] while especially endemic and small-ranged species are dependent on the intactness of their respective ecosystems [73].

We calculated the BII accroding to [38] in order to assess the impact of LULCC on overall vertebrate diversity in the time from 2010 to 2030. The BII is defined as the population of a species group i under land-use activity k in ecosystem j, relative to a reference population on the same ecosystem type according to eq 1. BII ¼ SiSjSkRijAjkIijk SiSjSkRijAjk ð1Þ Eq 1: Biodiversity Intactness Index [38]. Iijk, the “population impact”, is the population of a certain species group i under land-use activity k in ecosystem j. Ajk is the area of land-use k in ecosystem j, Rij the number of species of taxon i in ecosystem j. Since the calculation is done on grid-cell level, each cell is associated with a specific land use type. The number of species is the sum of bird species, mammal species, and amphibian spe- cies assigned to one cell respectively. In order to formulate the population impact, a combina- tion of impact values from [74, 75] and [39] was employed. These values indicate the reduction of mean species diversity in respect to a certain type of land use. The values employed are shown in Table 2. A BII value of 1 indicates a species abundance on the pre-colo- nial level. An index of 0,5 indicates that the species abundance is reduced by half in reference to the pre-colonial level. A decreasing BII value is an expression for further reduction of biodiversity intactness due to LULCC affecting regions characterized by the occurrence of different species of different taxa. An increasing BII value expresses a recovery of biodiversity intactness mainly due to the

displacement of anthropogenic land-use out of these regions or by replacement of certain land-use types by “less harmful” land-use types (e.g. cropland to fallow land) within these regions. 3. Results & discussion The first section summarizes the main LULCC characteristics for each scenario. Thereafter, we assess the resultant effect on vertebrate species diversity, (1) by relating natural habitat area conversion and vertebrate species diversity as well as distribution ranges and (2) by calculating the BII. PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 7 / 21

3.1. Area conversion in the land-use change scenarios In the following, the main LULCC characteristics of each scenario are described in form of a land-change matrix (Table 3). The simulation of the Trend Scenario led to a strong conversion of rainforest to pasture area in PA while a smaller share of rainforest was converted into cropland. Only relatively small shares of other natural vegetation had to be converted into managed land. In MT, a compara- tively large area covered by natural vegetation had to be converted into pastures and a smaller area into cropland. Rainforest was only converted into pastures. An intensification of pasture management is not part of this scenario. In contrast, the Legal Intensification Scenario led to no conversion of rainforest into pas- tures in both Federal States. This is explained by the intensification of pasture management as described in section 2.3. Nonetheless, in PA the conversion of rainforest into cropland was increased by 221% in comparison with the Trend Scenario due to an increasing demand for, on the one hand, fodder crops (soybean) and, on the other hand, staple crops used for human consumption in Asian countries [61]. In addition to the rainforest that is converted to crop- land, 216,020 km2 of rainforest are classified as Legal Reserve in PA. This land is not converted from its original vegetation cover but is considered disturbed due to being fragmented by small-scale agriculture. Thus, it is considered with a population impact of 0.83 (see Table 3)

when calculating the BII for PA. In MT, no rainforest was converted into managed land while the conversion of natural vegetation into managed land could be decreased by 3.1% compared to the Trend Scenario. In the case of the Illegal Intensification Scenario, in PA the conversion of rainforest to crop- land could be reduced by 99% and the conversion of rainforest to pastures could be decreased by 26% in comparison with the Trend Scenario. In PA, particularly pastures are relocated into protected areas covered by rainforest due to higher net primary productivity. In MT, a strong relocation of cropland into protected areas is discernible. Here, possible crop yields are higher than in other regions of the Federal State. In the case of the Sustainable Development Scenario, the conversion of rainforest into man- aged land was not allowed (see section 2.3). The shift of the consumption behavior towards a Table 2. Values used as population impact to calculate BII. land-use impact source Cropland 0.15 [74] weighted by proportion of high input agriculture to low input agriculture in Latin America Pasture extensive (PA) 0.6 [75] intensive (MT) 0.3 Mosaic Agricultural Area/rainforest (Legal Reserve in the transition matrix) 0.83� 20% cropland impact as calculated above and 80% undisturbed forest impact [74] Rainforest 1.0 [74] Grassland, Savannah, Shrubland, Wetland (natural vegetation in the transition matrix) 0.94 [39, 76] Fallow land 0.5 [74] Urban 0.05 [74] � A population impact value of 0.83 has been assumed for areas in the Amazon biome that are made up of 20%

cropland and 80% rainforest, the so called “Legal Reserve”, in which any kind of deforestation is prohibited according to the Brazilian Forest Code [77, 78]. This population impact value is considered as an expression for the fragmentation of rainforest. https://doi.org/10.1371/journal.pone.0225914.t002 PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 8 / 21

diet that is mainly based on crop products helped to avoid the necessity to convert rainforest or other natural vegetation into pastures. The reduction of meat demand led to a release of pas- ture area that could be utilized to expand cropland area that is now employed to produce addi- tional crops in order to substitute the share of calories that was originally consumed in the form of meat in PA. In order to realize this substitution of calorie sources in MT, compared with the Trend Scenario, 16% more natural vegetation had to be converted into cropland. Agricultural intensification has played an important role in regard to recent agricultural production growth in Brazil and is likely to further increase Brazil´s crop and beef production considerably [79]. The observed decoupling of production increases from deforestation in the latter half of the first decade of this century [29, 80] have shown that the intensification of agri- cultural systems not only supports food provisioning, it also limits the expansion of agricul- tural area; thus the destruction of natural habitats [81]. This trend could be confirmed. The negative effect of projected future agricultural production growth on the extent of natural hab- itats is considerably reduced through agricultural intensification and particularly through intensification of grazing intensities on pastures (compare with [19, 81]). Especially the Legal Intensification Scenario shows substantially less converted natural habitats compared to the

Trend Scenario based on constant crop yields and grazing intensities of the reference year 2010. This is also confirmed by Cohn et al. [18] who have shown that the encouragement of an intensification of pastures, either through subsidies of intensified systems or tax on extensive pastures, considerably limits the conversion of natural habitats until 2020. Nonetheless, the reducing effect of an intensification of pastures is much lower in the case of the Illegal Intensifi- cation Scenario where protected areas are made accessible for conversion into managed land. This leads to a relocation of managed land into these areas due to higher net primary produc- tivity on pastures or higher possible yields on cropland, causing higher conversion rates. This illustrates what has been confirmed by several other authors e.g. [82, 83], who found that the Table 3. Land-use and land-cover change matrix for the years 2010 to 2030 (103 km2). PA TREND LI ILI SUST CR PS ext. UR CR PS ext. UR LR CR PS ext. UR CR PS ext. UR RF 16.84 89.94 0.00 54.01 0.00 0.00 216.02 33.48 66.15 0.00 0.00 0.00 0.00 NV 5.82 6.38 0.01 8.95 5.00 0.01 0.00 8.84 6.29 0.01 4.80 0.00 0.01 FA 0.19 0.00 0.00 0.34 0.00 0.00 0.00 0.04 0.40 0.00 0.48 0.00 0.00 CR 139.33 0.00 0.03 139.33 0.00 0.03 0.00 128.62 4.47 0.03 125.26 0.00 0.02 PS 0.00 96.70 0.00 7.11 89.60 0.00 0.00 5.28 91.38 0.00 83.86 12.84 0.00 UR 0.00 0.00 0.56 0.00 0.00 0.56 0.00 0.00 0.00 0.56 0.00 0.00 0.56 MT TREND LI ILI SUST

CR PS int. UR CR PS int. UR LR CR PS int. UR CR PS int. UR RF 0.00 28.38 0.00 0.00 0.00 0.00 0.00 0.70 0.00 0.00 0.00 0.00 0.00 NV 6.62 27.82 0.00 18.51 17.05 0.00 0.00 51.26 3.33 0.00 40.10 0.00 0.00 FA 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 0.00 CR 184.63 23.47 0.00 208.52 0.00 0.00 0.00 181.51 5.90 0.00 208.51 0.00 0.00 PS 0.00 158.42 0.00 3.05 148.13 0.00 0.00 1.36 132.62 0.00 2.31 156.10 0.00 UR 0.00 0.00 0.85 0.00 0.00 0.85 0.00 0.00 0.00 0.85 0.00 0.00 0.85 (TREND = Trend Scenario, LI = Legal Intensification Scenario, ILI = Illegal Intensification Scenario, SUST = Sustainable Development Scenario) (CR = cropland, PS (extensive (ext.) in PA; intensive (int.) in MT) UR = urban area, RF = Rainforest, NV = natural vegetation, FA = fallow land). https://doi.org/10.1371/journal.pone.0225914.t003 PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 9 / 21

intensification of agricultural production, if it goes hand in hand with the protection of natural habitats, have the highest impact in terms of limiting the conversion of natural habitats, and thus, promoting the conservation of vertebrate species diversity. 3.2. Effects of natural habitat loss on the distribution ranges of vertebrate species 3.2.1. Para´. Fig 2 illustrates that the highest effect in terms of a distrubance of natural hab- itats and distribution ranges of vertebrate species in PA can be expected in case of the Trend Scenario. Allmost all of the 119,185 km2 of converted natural habitats in the case of the Trend Sce- nario is identified as a distribution range of bird and mammal species. Also, close to all of this coverted area is a ditribution range of endemic amphibians. Only the distribution ranges of small-ranged and threatened aphibians are not affected strongly by the conversion of natural habitats into managed land. 67,914 km2 of natural habitats were converted in the case of the Legal Intensification Sce- nario excluding 216,022 km2 of Legal Reserve as it remains in its original land-cover type “Rainforest”. The calculation of the area-weighted change of affected distribution ranges, including the Legal Reserve as it is considered affected by fragmentation, could be reduced by 46% for all mammal and bird categories. The afflicted distribution ranges could be decreased by 47% for endemic amphibians, by 45% for small-ranged and by 44% for threatened amphibi- ans compared to the situation in the Trend Scenario.

In the case of the Illegal Intensification Scenario, 114,608 km2 of natural habitat area is con- verted into managed land. In this case, the calculation of the area-weighted change shows a reduction of the affected distribution ranges by 40% for endemic bird and mammal species and by 38% for all other bird and mammal categories in comparison to the Trend Scenario. Also, the afflicted distribution ranges of endemic, small-ranged and threatened aphibians could be decreased by 46%, 31% and 37% respectively. The least negative effect on vertebrate species diversity due to a conversion of natural habi- tats is discernible in the case of the Sustainable Development Scenario. Not only the area of con- verted natrual habitats is less as compared to the other scenarios, also the share of this area that is a known habitat to vertebrate species is relatively low. The affected area known as Fig 2. Natural habitat area of investigated vertebrate species affected by conversion over total converted area in Para´. https://doi.org/10.1371/journal.pone.0225914.g002 PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 10 / 21

distribution range of all investigated vertebrate species could be reduced by over 99% (area- weighted change) compared to the Trend Scenario. 3.2.2. Mato Grosso. As Fig 3 shows, the assumptions made for the Trend Scenario in MT result in a strong disturbance of vertebrate species diversity. An area of 62,824 km2 of natural habitats is converted in the case of the Trend Scenario. An especially strong impact can be seen for all categories of bird species, endemic and threatened mammals and endemic amphibians. The distribution range of small-ranged mammals is slightly less affected. Small-ranged amphibian distribution ranges are least afflicted. The situa- tion concerning threatened amphibians (for all scenarios) may not be consistent with the actual situation as threatened amphibians are especially data deficient in Brazil [5]. In the case of the Legal Intensification Scenario 35,555 km2 of natural habitats were con- verted into managed land. The calculation of the area-weighted change of affected distribution ranges illustrates a reduction of 53% for all amphibian, bird and mammal taxa and categories except small ranged amphibians (-30%) and small-ranged mammals (-46%) in comparison with the Trend Scenario. Concerning the Illegal Intensification Scenario, the negative effect of a conversion of natural habitats (55,301 km2) on the distribution ranges of all investigated taxa and categories per taxon could be reduced (area-weighted change) by 44%, except small-ranged amphibians (-73%) and small-ranged mammals (-35%) despite the fact that the total converted area was

only reduced by 12% in comparison with the Trend Scenario. The assumptions made in the case of the Sustainable Development Scenario lead to a reduc- tion of converted natural habitats to 40,166 km2. The positive effect expressed as a reduction of the afflicted distribution ranges (area-weighted change) amounts to 36% for endemic amphibi- ans and mammals as well as small-ranged and threatened birds. This positive effect can also be witnessed for small-ranged amphibians (-32%), endemic birds (-44%), small-ranged (-26%) as well as threatened mammals (-38%). 3.2.3. Factors influencing natural habitat loss and the distribution ranges of vertebrate species: A scenario comparison. In MT, the area of affected natural habitats in the case of the Illegal Intensification scenario could be reduced by 12% compared to the Trend Scenario. This reduction of transformed natural habitats due to agricultural intensification is surpassed in the Legal Intensification Scenario. Here, the conversion of natural habitats could be Fig 3. Natural habitat area of investigated vertebrate species affected by conversion over total converted area in Mato Grosso. https://doi.org/10.1371/journal.pone.0225914.g003 PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 11 / 21

decreased by 43% compared to the Trend Scenario. Also in both cases, the affected distribution ranges of vertebrate species could be reduced considerably. These results suggest that intensifi- cation measures are effective. This is confirmed by Strassburg and Latawiec who found in their study of possible future agricultural productivity scenarios that an increase of pasture produc- tivity to 49–52% of its potential productivity in Brazil would suffice for the projected increase of agricultural production without further appropriation of natural habitats [84]. Although, the effect of combined intensification and adequate conservation policies can contribute stron- ger to the preservation of the assessed vertebrate species as the results of the Legal Intensifica- tion Scenario (see Section 2.3) show. This can be confirmed by Sparovek et al. who indicate that in all of their simulated future land use scenarios, effective conservation policies played the most important role in conserving natural habitats [85]. Further, the effective protection and conservation of forested land will not impair Brazil´s ability to produce agricultural ressources for demanding markets in the near future as has been confirmed by de Souza Fer- reira Filho et al. [86]. This picture is confirmed in PA. Here, the amount of converted natural habitats could be limited by 43% (Legal Intensification Scenario) and by 3% in the case of the Illegal Intensification Scenario. Despite this relatively low reduction of converted natural habi-

tats in the latter case, this simulated decrease still led to a considerable reduction of afflicted distribution ranges of vertebrate species due to a spatial shift of habitat conversions away from the distribution ranges of the investigated vertebrate species. Concerning the Sustainable Development Scenario in MT, cropland area expands especially strong due to a shift of anthro- pogenic consumption patterns away from meat toward crop intake while there is only a slight release of pasture area. Therefore, the decrease of pasture area can only partially counteract the expansion of cropland, and thus the loss of natural habitat area in MT remains relatively high compared to the situation in PA. This is also expressed in the reduction of affected distribution ranges. The simulated relocation of cropland onto released pasture area in PA almost completely prevents the loss of natural habitats that are known distribution ranges of the inves- tigated vertebrate species. This development is confirmed by Alkimim et al. who found that 50 million hectares of pasture land in Brazil are suitable for crop prodcution and thus, can spare forest area from conversion into agricultural land [87]. Especially here, the shift of human con- sumption toward a plant based diet is the key factor for limiting the loss of natural habitats resembling distribution ranges of vertebrate species as the intensification of pastures is not part of the assumptions for the Sustainable Development Scenario. The optimal combination

of intensification and conservation measures as well as the interplay with changes of anthropo- genic consumption patterns with the goal of a maximized reduction of converted natural habi- tats depends on the present situation in the respective region. This is illustrated by the heterogeneity of conversions of natural habitats in PA and MT, under the respective scenario assumptions. 3.3. The combined impacts of expansion and intensification of agriculture on vertebrate diversity 3.3.1. Para´. The effects of a conversion of natural habitats on vertebrate species diversity are confirmed by our assessment of the BII in PA. Table 4 shows, the impact on species diver- sity, as expressed by changes of the BII, is strongest as calculated in the case of the Trend Sce- nario in PA. We found especially strong redcutions of BII for small-ranged amphibians, threatened mammals, followed by threatened amphibians and endemic mammal species. In the case of the Legal Intensification Scenario, we see BII value decreases for all taxa and categories. Espe- cially strong disturbances can be discerned in the case of small-ranged amphibian species, PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 12 / 21

endemic mammals as well as threatened mammals. In the case of the Illegal Intensifcation Sce- nario especially small-ranged amphibian species, endemic mammals, and threatened amphib- ian species were strongly affected. The lowest negative effect was simulated in the case of the Sustainable Development Scenario. The highest decrease was calculated for threatened amphib- ians while the BII and threatened mammals decreased. All other BII values remained constant or even increased as is the case for small-ranged amphibians and small-ranged bird species. This slight increase can be attributed to the conversion of unused cropland cells to pasture cells. 3.3.2. Mato Grosso. Table 5 shows, the impact on species diversity in MT is more moder- ate in the case of the Trend Scenario compared to the situation in PA. The highest reduction of BII in MT was simulated in the case of the Trend Scenario. Here, threatened bird species, threatened mammals, and endemic mammals are especially affected. In the case of the Legal Intensification Scenrio, we see decreasing BII values for all taxa and cat- egories with the exception of small-ranged birds which remains constant. We found especially strong decreases for threatened mammals and threatened birds. Concerning the Illegal Intensi- fication Scenario, especially small-ranged amphibians, threatened mammals and threatened bird species are impacted strongly. Interestingly, the Sustainable Development Scenario in MT results in a strong negative impact. It becomes obvious that all taxa and categories are affected

Table 4. Changes of BII in Para´ between 2010 and 2030. taxon category Trend 2010 Trend 2030 rel. Change [%] LI 2030 rel. Change [%] ILI 2030 rel. Change [%] SD 2030 rel. Change [%] Amphibians endemic 0.79 0.71 -10.1 0.71 -10.1 0.73 -7.6 0.79 0 small-ranged 0.65 0.52 -20 0.53 -18.5 0.57 -12.3 0.66 1.5 threatened 0.59 0.49 -16.9 0.52 -11.9 0.53 -10.2 0.58 -1.7 Birds endemic 0.8 0.71 -11.3 0.71 -11.3 0.73 -8.8 0.8 0 small-ranged 0.85 0.77 -9.4 0.78 -8.2 0.78 -8.2 0.86 1.2 threatened 0.79 0.72 -8.9 0.73 -7.6 0.75 -5.1 0.79 0 Mammals endemic 0.79 0.68 -13.9 0.69 -12.7 0.71 -10.1 0.79 0 small-ranged 0.86 0.77 -10.5 0.78 -9.3 0.79 -8.1 0.86 0 threatened 0.79 0.65 -17.7 0.69 -12.7 0.72 -8.9 0.78 -1.3 (Trend = Trend Scenario, LI = Legal Intensification Scenario, ILI = Illegal Intensification Scenario, SD = Sustainable Development Scenario). https://doi.org/10.1371/journal.pone.0225914.t004 Table 5. Changes of BII in Mato Grosso between 2010 and 2030. taxon category Trend 2010 Trend 2030 rel. Change [%] LI 2030 rel. Change [%] ILI 2030 rel. Change [%] SD 2030 rel. Change [%] Amphibians endemic 0.67 0.62 -7.6 0.66 -1.5 0.67 0 0.62 -7.5 small-ranged 0.56 0.54 -3.6 0.55 -1.8 0.36 -35.7 0.54 -3.6 threatened n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Birds endemic 0.66 0.6 -9.1 0.65 -1.5 0.67 1.5 0.61 -7.6 small-ranged 0.62 0.57 -8.1 0.62 0 0.6 -3.2 0.59 -4.8 threatened 0.59 0.51 -13.6 0.55 -6.8 0.5 -15.2 0.53 -10.2

Mammals endemic 0.63 0.57 -9.5 0.62 -1.6 0.64 1.6 0.59 -6.4 small-ranged 0.66 0.6 -9.1 0.65 -1.5 0.64 -3.0 0.63 -4.6 threatened 0.68 0.61 -10.3 0.63 -7.4 0.6 -11.8 0.63 -7.3 (Trend = Trend Scenario, LI = Legal Intensification Scenario, ILI = Illegal Intensification Scenario, SD = Sustainable Development Scenario). https://doi.org/10.1371/journal.pone.0225914.t005 PLOS ONE Agricultural intensification effects on natural habitats and biodiversity PLOS ONE | https://doi.org/10.1371/journal.pone.0225914 November 25, 2020 13 / 21

negatively, especially endemic amphibians and endemic birds as well as threatened birds and threatened mammals. 3.3.3. The combined effects of agricultural expansion and intensification on vertebrate diversity: A scenario comparison. In general, when looking at sections 3.3.1. and 3.3.2. as well as the following discussion of the illustrated results, one has to keep in mind that the BII values in MT are on average 0.14 points below those calculated for PA as MT is more consoli- dated in agricultural terms [51, 52]. Newbold et al. [7] calculated BII values of around 70% for the Brazilian Cerrado (mainly located in MT) as well as 85% for the Amazon biome (mainly located in PA). This agrees well with our calculation for the year 2010 of 0.59–0.68 (MT) and 0.65–0.86 (PA) respectively (see Tables 4 and 5). The fact that the estimates found in our study are slightly lower than those estimated by Newbold et al. [7] is explained by taking into account that they focused on the whole Cerrado and Amzonia region. We assess a sub-region that is and was characterized by especially strong LULCC dynamics. Interestingly in PA, the negative effect in the case of the Illegal Intensification Scenario is lower compared to the Legal Intensification Scenario. This can be explanied by the consider- ation of the 216,022 km2 of rainforest share of the total Legal Reserve area (270,027 km2) in PA as affected area with a population impact of 0.83 (see Table 3). Here, the fragmentation of the total Legal Reserve area by small scale agriculture, and consequently the degradation of rain-

forest habitats e.g. [88, 89], has a negative impact that leads to a stronger decrease of the BII as compared to the Illegal Intensification Scenario. This is true despite the fact that in the Illegal Intensification Scenario a larger amount of natural habitat area is actualy converted into man- aged land with a higher impact of either 0.15 for cropland or 0.6 for extensive pasture. In the case of the Sustainable Development Scenario in PA, the impact of a conversion on natural hab- itats, distribution ranges of vertebrate species and, thus calculated BII values could be reduced substantially in comparison to all other scenarios. This effect is attributable to a substantial reduction of the global meat intake which leads to a significant reduction of pasture area which, in turn, can be utilized for an expansion of cropland area. A total of 94% of the crop- land expansion was realized on released pasture area (see Table 3), thus decreasing the amount of transformed natural habitat area considerably. Alkimim et al. come to a similar conclusion in their study [87]. In PA, strong decreases of the BII can be witnessed. In the case of the Illegal Intensification Scenario in MT strongly decreasing BII value for threatened vertebrate species and especially small-ranged amphibian species can be observed due to the opening up of protected areas for agricultural expansion. Especially the Pantanal, known for its species richness in regard to birds and amphibians [90], is affected by the dis- placement of managed land into formerly protected areas. Moreover, we see decreasing BII

conserved, which displaces LULCC from the Pantanal to other not conserved, less biodiverse areas. This prevents strongly decreasing BII value for threatened vertebrate species and espe- cially small-ranged amphibian species in the Legal Intensification scenario in contrast to the sit- uation in the Illegal Intensification Scenario. As has been discussed before, the expansion of cropland due to a anthropogenic consumption shift towards plant based consumption can only be marginally realized on released pasture area in the Sustainable Development Scenario in MT, resulting in BII decreases that are stronger compared to the situation in the Sustainable Development Scenario in PA. Concerning both Federal States, it has to be mentioned that the impact of a conversion of natural habiats on the distribution ranges of vertebrate species expressed by the BII seems rela- tively high concerning small-ranged and threatened amphibians in PA and small-ranged amphibians in MT. This can be explained by the small extent of the distribution ranges of these vertebrate species [5] (see section 2.4) compared to the other investigated vertebrate spe- cies. Only a marginal conversion of natural habitats can mean a severe impact, especially if expressed with the help of an indicator as the BII. 3.4. Limitations and uncertainties of the study First and foremost, a model-based scenario analysis with a focus on environmental impacts should not be misunderstood as a method to predict concrete future events. Rather, it provides