D 7.4 - Ex-ante multi-criteria analysis of supply chain governance initiatives targeting biodiversity protection considering interdependencies

Name of the Deliverable Ex-ante multi-criteria analysis of supply chain governance initiatives targeting biodiversity protection considering interdependencies Deliverable 7.4 2 Summary Work Package 7 Deliverable No 7.4 Dissemination Level Public Type R-document, report Lead Partner BASQUE CENTRE FOR CLIMATE CHANGE (BC3) Due Date 31st August 2025 Submission Date 8th October 2025 Status Final Authors Neus Escobar, Javier Ribal, Clara Frezal, David Leclère Other contributions Sibylle Rouet-Pollakis, Nelson Kevin Sinisterra-Solís, Neus Sanjuán About CLEVER Project Number 101060765 Project Title CLEVER: Creating leverage to enhance biodiversity outcomes of global biomass trade Topic HORIZON-CL6-2021-BIODIV-01-15 Start date 1st September 2022 End date 31st August 2025 Coordinator RHEINISCHE FRIEDRICH-WILHELMS-UNIVERSITAT BONN This document has been prepared in the framework of the project CLEVER. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. 3 This project has received funding from the European Research Executive Agency under HORIZON Research and Innovation Actions, grant agreement no101060765; and from the UK Research and Innovation (UKRI) under the UK government's Horizon Europe funding guarantee [grant number 10038491] 4 Contents 0. EXECUTIVE SUMMARY ................................................................................................... 5 1. BACKGROUND AND SCIENTIFIC CONTRIBUTION ................................................................. 9 2. OBJECTIVES ...................................................................................................................... 11 3. METHODS ........................................................................................................................ 12 3.1. Policy selection and scenario definition .................................................................... 12 3.2. Biodiversity loss indicators ....................................................................................... 17 3.3. SDG related indicators .............................................................................................. 18 3.3.1. Environmental impacts ........................................................................................................... 18 3.3.2. Socioeconomic impacts .......................................................................................................... 19 3.4. Trade-off analysis and interdependencies between SDG indicators ........................... 22 4. RESULTS AND DISCUSSION ............................................................................................... 24 4.1. Sustainability outcomes ............................................................................................ 24 4.2. Eco-efficiency ratio ................................................................................................... 29 4.3 Quantification of interdependencies .......................................................................... 32 4.3.1. Spearman rank correlation results .......................................................................................... 32 4.3.2. Results from the Principal Component Analysis (PCA) ........................................................... 35 CONCLUSIONS ..................................................................................................................... 38 PROJECT OUTPUTS ACHIEVED .............................................................................................. 39 REFERENCES ........................................................................................................................ 40 5 0. EXECUTIVE SUMMARY In response to the alarming decline in biodiversity linked to human socioeconomic development, CLEVER investigates the drivers, mechanisms, and scale of biodiversity loss across ecosystems. A central objective is to evaluate how different policies and supply chain initiatives influence global biodiversity, particularly in relation to existing land conservation schemes and climate change mitigation policies. Recognizing the role of international trade in accelerating land conversion and shifting impacts from consumers to producers, the project focuses on three key agricultural commodities with growing international demand: fishmeal, timber, and soy. Within WP6, the GLOBIOM modelling framework was enhanced to, on the one hand, refine the representation of these sectors in the model to better capture market dynamics for derivative products and intervention options across actors; on the other hand, to incorporate highresolution impact indicators that link biodiversity loss to specific drivers, building on advances from WP2. In WP7, the extended GLOBIOM framework has been applied to simulate combined policy scenarios, linked to empirical work in WP4 and WP5, and informed by expert stakeholder input. Deliverable D7.4 focuses on the soy supply chains and assesses the specific policy scenarios through a multi-criteria lens to identify trade-offs between biodiversity protection and other dimensions of sustainability, as well as key interdependencies. This deliverable combines the results from D7.2 and D7.3 to identify trade-offs and rank alternative interventions affecting the Brazilian soy sector according to their projected sustainability performance in 2050. The results provide comprehensive insights into the risks and benefits of different interventions along international soy supply chains, with a particular focus on the EU and Brazil, while also considering global spillover effects based on the conceptual framework of D5.2. Overall, these findings highlight the most effective approaches for promoting sustainability in Brazil and worldwide. The scenario analysis explores a range of possible futures for global soy markets and biodiversity, including conservation efforts in Brazil, due diligence policies, border adjustment tariffs and trade agreements/disputes, as well as global or regional dietary shifts and other demand-side measures. The “business-as-usual” (BAU) scenario represents a baseline under Shared Socioeconomic Pathway 2 (SSP2) or “middle of the road”, assuming no major policy changes, to assess impacts from current socioeconomic trends. The Integrated Action Portfolio (IAP) scenario envisions strong conservation and restoration measures aligned with the KunmingMontreal Global Biodiversity Framework (KMGBF) goal of reversing global biodiversity loss by 2050, combined with sustainable yield increases, shifts towards plant-based diets and reduced food waste. The Tr.Dis. scenario investigates the long-term consequences of the US–China trade war by capping China’s imports of US soy-based products. The ZNLBra scenario assumes zero conversion of natural lands to agriculture in Brazil after 2020, exploring the effects of strict land conservation on soy production and trade. Several scenarios focus on the impacts of EU policies, such as the EUDR, which reduces Brazil’s soy exports to the EU by 3% compared to BAU; and EUM+EUDR, which instead assumes a 15% increase due to combined effects of the EUMERCOSUR agreement and the EUDR. Variants of this include EUM+EUDR_ZNLBra, where stronger EU measures trigger enhanced conservation in Brazil, and EUM+EUDR_WeakLUR, where they coincide with weakened land-use regulation and reduced enforcement. Additional policy experiments test alternative EU approaches, such as the EUM+EUBBAM scenario, replacing the EUDR with a biodiversity border adjustment mechanism that taxes high-risk imports based on biodiversity impact; and EUM+EUDemSide, where demand-side measures— 6 like dietary shifts and reduced food waste—substitute the EUDR to mitigate environmental impacts from EU consumption. Trade-offs among selected environmental and economic indicators are firstly assessed for the different scenarios in relation to the BAU. Different sets of indicators are analysed together, combining a supply-oriented perspective, i.e., to assess trade-offs among impacts generated in soy sourcing regions, and a demand-oriented perspective, i.e., to assess trade-offs among impacts at the global level and in key soy consumer regions. All scenarios result in similar soybean production in Brazil in 2050, ranging between 210 and 230 Mt, with little differential impact on global food security (SDG2) or economic growth in Brazil (SDG8). Only the IAP scenario, which combines global land conservation and restoration measures with shifts toward plant-based diets, results in a marked reduction in production of about 30% (to 152 Mt), lowering the value of soybean output in Brazil and globally, though with minimal changes in calorie availability. Differences in soybean production influence total natural land conversion, as land-use policies determine how agricultural expansion interacts with forest and other natural areas. The Zero Natural Land Loss policy in Brazil (ZNLBra) has the strongest effect, driving substantial forest and natural land expansion —about 67 Mha under ZNL and EUM+EUDR_ZNLBra, and around 40 Mha under IAP— compared with the BAU. IAP additionally results in about 120 Mha of restored land in Brazil, while other scenarios maintain restoration levels similar to BAU (around 18.5 Mha). Overall, these results highlight how different policy mixes influence the balance between agricultural production, land-use dynamics, and environmental sustainability, including biodiversity loss (SDG15). Differences in environmental impacts between scenarios are more pronounced than differences in socioeconomic outcomes. IAP reduces total emissions in Brazil by nearly 5 Gt CO₂eq through decreased consumption of animal products, reduced soy demand, and agricultural area savings. There is net carbon sequestration of 4.5 Gt CO₂eq from land use change (LUC), particularly in the Cerrado, Mata Atlantica, and Amazon. EUM+EUDR+ZNLBra and ZNL reduce LUC emissions by 319 Mt CO2eq (-97%), while the remaining scenarios have only minor impact on GHG emissions (SDG13). IAP also reduces freshwater biodiversity loss (-127%) and terrestrial biodiversity loss (-47%), with ZNLBra and EUM+EUDR+ZNLBra showing moderate reductions (16% and 32%, respectively). Trade disruption (Tr.Dis) slightly increases freshwater biodiversity loss (+1.7%) due to higher soy production and related agricultural inputs. Total irrigation water demand remains at 43 km3 in all scenarios except in IAP (45 km3 or 4.6% higher than BAU) due to an increase in the production of irrigated crops. The EUDR scenarios produce marginal changes (<1%) in output variables relative to the BAU, with small decreases in GHG emissions and biodiversity impacts under EUDR alone, but slightly increased impacts under EUM+EUDR due to higher soy production, primarily in the Cerrado. Overall, the Amazon biome exhibits more differences across scenarios than other regions, driven by land reallocation and soy expansion. Outcomes from the different scenarios have been combined into a set of climate and biodiversity Eco-efficiency Ratios, which measure the environmental impacts from soy production to the gross economic value of soy output. By including an economic metric (revenue) in the denominator, the eco-efficiency ratio assesses how efficiently economic value is created while minimizing environmental impacts, thereby linking environmental performance directly to economic productivity. In this way, higher values indicate greater environmental impact per USD generated, meaning that the economic activity is less environmentally efficient. Lower values suggest higher environmental efficiency, as less impact is generated per USD of 7 revenue. Comparing these metrics across scenarios over a given time horizon helps assess how effectively different policy mixes promote the decoupling of agricultural economic activities from environmental impacts. This involves achieving environmental protection at the lowest possible economic cost by balancing financial performance with sustainability goals (SDG8, SDG12). It is calculated as the ratio between the absolute environmental impacts (either AFOLU GHG emissions or total biodiversity impacts from farming) in a given scenario and time horizon, relative the value of soy production (AbsEIIR), for Brazil, the rest of the world (ROW), and globally. The IAP scenario demonstrates the most efficient impact intensity performance in Brazil, reducing GHG emissions per USD2000 generated to -0.75 kgCO₂e/USD2000. The ZNLBra scenarios perform moderately well (~1.3 kgCO₂e/USD2000), while the other scenarios are close to the BAU (2.65 kgCO₂e/USD2000). The latter is only surpassed by EUDR, EUM+EUDemSide and EUM+EUBBAM, which decrease soy gross output in Brazil. In terms of biodiversity loss, IAP again performs best in Brazil, with a negative total species richness loss (-5.4E-9 PDF.y/MillionUSD2000) driven by improved terrestrial biodiversity through reduced LUC. However, IAP shows slightly higher biodiversity impact intensity in ROW and globally (5.6-5.8E8 PDF.y/MillionUSD2000) due to water and land stress in regions such as North America or the rest of South America. Conversely, IAP exhibits the highest water stress intensity in Brazil (0.22 m³/USD2000) compared to other scenarios (0.14–0.15 m³/USD2000), with global and ROW values at 0.36 m³/USD2000, highlighting trade-offs between national efficiency gains and externalized environmental pressures. A composite Eco-Efficiency ratio (CsEIIR) has also been proposed to rank scenarios, considering two different sets of weights for the total biodiversity loss, GHG emissions and water stress. This enables the integration of multiple dimensions into a single score for easier scenario ranking, by scaling individual indicators to a common basis and applying weights that reflect policy priorities, in this case for biodiversity conservation and environmental protection. While this approach supports decision, its results are sensitive to normalization choices and weighting schemes, which introduces a degree of subjectivity and simplification. When using the same weights for the three impacts, IAP stands out as the most eco-efficient scenario for Brazil (0.3) but the worst for the ROW (0.7) and globally (0.7). EUM+EUDR+ZNLBra and ZNLBra have similar scores across regions (0.4), indicating a good performance and limited spillover of environmental impacts. In contrast, the remaining scenarios show the highest score (0.7) or lowest eco-efficiency in Brazil, with lower scores (around 0.3-0.5) for the ROW. When weighting only by biodiversity and GHG emissions, IAP is the most eco-efficient according to the Brazilian CsEIIR (0), followed by EUM+EUDR+ZNLBra (0.7) and ZNL (0.7). The remaining scenarios generate the highest scores for Brazil (1), while EUM+EUDemSide shows the highest score for the ROW (0.7) and globally (0.9). These figures facilitate comparison and enable a clearer classification of scenarios, although they omit the water stress dimension, which proved less decisive for evaluating soy-related policies. It should be noted that for scenarios involving broad demand-side interventions (EUM+EUDemSide and IAP), these metrics should not be interpreted as spillover effects from Brazil to the ROW, since their primary objective is not the mitigation of deforestation-related impacts in Brazil. Finally, correlations between selected indicators have been assessed through Spearman rank correlation and Principal Component Analysis (PCA). The Spearman rank coefficient shows positive correlations between soybean production and environmental impacts in Brazil, i.e., water stress, terrestrial biodiversity, and total GHG emissions. Freshwater biodiversity loss and 8 terrestrial biodiversity loss are strongly correlated with GHG emissions, since LUC emissions account for a large share of total GHG emissions, which also affect freshwater biodiversity through climate change. From the demand side, the calories consumed in Brazil and the soy production in the country are positively correlated with the total world GHG emissions and the global terrestrial biodiversity loss. The results also show a strong positive correlation between the calories consumed as food in the European Union (EUE) and Southern Asia (SAS) (incl. China), while negative correlations are observed between these and the value of the soy market in Brazil. This indicates that imports in these two regions respond to increases in Brazilian soy prices, although the results show that EUE relies on Brazilian soy imports to a lesser extent than SAS to meet their food calorie demand. This deliverable compares sustainability indicators and eco-efficiency of many policies and interventions affecting soy markets and supply chains, while proposing quantitative evidence to rank policy mixes according to their potential to mitigate impacts and trade-offs between biodiversity loss (SDG15), climate change mitigation (SDG13), water stress (SDG6), economic returns (SDG8), and food availability (SDG2). 9 1. BACKGROUND AND SCIENTIFIC CONTRIBUTION Socioeconomic development has led to the progressive increase of greenhouse gas (GHG) emissions, environmental degradation, and biodiversity loss, especially since the industrial revolution (Infante-Amate et al., 2025; Marques et al., 2019). The current rate of species extinction is estimated to be between 100 and 1000 times faster than it would be without human influence, which indicates that anthropogenic activities could be triggering the Sixth Mass Extinction (Ceballos et al., 2015). Land use change (LUC) is the main cause of terrestrial biodiversity impacts through habitat loss and degradation, which significantly impact both species and ecosystems (IPBES, 2019). This is particularly true in the tropics (Barlow et al., 2018; Socolar et al., 2025), the most biodiverse region on the planet, where deforestation is largely driven by agricultural commodity production (Maxwell et al., 2016; Pendrill et al., 2022). At the same time, climate change increasingly threatens ecosystems functioning (Cardinale et al., 2012), and can accelerate the reaching of critical tipping points (Dakos et al., 2019). Land use and LUC are not only drivers of biodiversity loss, but also play a central role for many Sustainable Development Goals (SDGs), including those related to food security, health, clean energy and climate action. Thus, consistent policy-making requires capturing these interlinkages and feedback effects. Integrated Assessment Models (IAMs) have proven particularly useful to identify and acknowledge trade-offs and leverage synergies in the pursuit of the SDG Agenda and the Paris Agreement (Popp et al., 2017; Riahi et al., 2017; Rogelj et al., 2018). These models are a simplified representation of the complex interactions between human and environmental systems, particularly in the context of global environmental change and sustainable development. Beyond simulating long-term dynamics, IAMs allow the analysis of exogenous interventions, including land protection and biodiversity conservation (Leclère et al., 2020; Veerkamp et al., 2020). Previous analyses have shown that alternative Paris-compliant policy measures, such as afforestation, biofuel and bioenergy targets, and dietary shifts, can have different effects on the GHG mitigation potential from Agriculture, Forestry and Other Land Uses (AFOLU) (Frank et al., 2019; Humpenöder et al., 2024; Rouhette et al., 2024). Most of them emphasize potential trade-offs through market-mediated effects, including indirect LUC and GHG leakage, although only a few quantify biodiversity impacts (Kozicka et al., 2023; Leclère et al., 2020; Read et al., 2022). In general, the above-mentioned studies conclude that a combination of supplyand demand-side initiatives and policies are needed to effectively and simultaneously progress towards conservation and broader sustainability goals. CLEVER employs the partial equilibrium, recursive-dynamic model GLOBIOM, which is the land use component of the MESSAGE-GLOBIOM IAM and has a detailed sub-national resolution to simulate LUC across agricultural and forestry sectors, including multiple environmental pressures. The challenge of modelling long-term biodiversity impacts in CLEVER is twofold: on the one hand, models need consistent quantitative indicators that represent the state of biodiversity through time, and how it changes in response to local, regional and global pressures, such as LUC or climate change. These indicators should be compatible with the temporal and spatial resolution of IAMs, most of which are global in scope. On the other hand, modellers need to identify and be able to represent, in a stylized way, the diversity of policies, strategies, and socioeconomic developments that potentially affect the distribution of land uses and associated biodiversity impacts at the global level. 16 7. EU Deforestation Regulation + EU-Mercosur RTA + Zero natural land loss in Brazil (EUM+EUDR_ZNLBra) scenario: this scenario explores the potential impacts of new EU trade-related policies (i.e., EU-Mercosur and EUDR) on Brazil-EU soy trade and associated biodiversity impacts, under the assumption that they would foster increased land conservation efforts in Brazil. This is done by simultaneously implementing the EUDR+EUM scenario and the ZNLBra scenario (developed in D7.2), which assumes zero absolute conversion from forest and non-forest natural land to agricultural land in Brazil after 2020 (see D7.2). This scenario assumes that international supply chains and the Brazilian society react to the new EU policies by ambitious supply chain traceability and land conservation efforts (well beyond that of current policies, or even pursued by the EUDR) in Brazil. This represents the manifestation in Brazil of a maximalist version of a ‘Brussels effect’, by which EU actions lead to the adoption of new, ambitious standards. 8. EU Deforestation Regulation + EU-Mercosur RTA + Weak Land Use Regulation (EUM+EUDR_WeakLUR) scenario: in contrast to the EUM+EUDR_ZNLBra scenario, this scenario explores the potential impact of the assumption that the new EU trade-related policies (i.e., EU-Mercosur and EUDR) lead to weaker supply chain and land conservation efforts in Brazil. This is done by simultaneously implementing the EUDR+EUM scenario and weakening key policies included in SSP2 and all other scenarios (see details on the Forest Code representation in D6.5). From 2030 onwards, this includes the removal of the ASM and a relaxation of the Forest Code. From 2030 onwards, an increase in land area dedicated to soy production is allowed in the Amazon biome, and the enforcement probabilities for illegal deforestation control as part of the Forest Code are reduced by a factor of 3 in both the Amazon and Cerrado biomes. However, as in all EUDR-based scenarios, it is assumed that a sufficient share of Brazil’s soy production can comply with the EUDR requirements, and that any soy produced from newly deforested land in Brazil is destined to other markets. This scenario provides insights into the implications of weakening both public and private land conservation efforts in Brazil as a potential adverse consequence of unilateral interventions from the EU, like the EUDR. 9. EU-Mercosur RTA + EU biodiversity border adjustment mechanism (EUM+EUBBAM) scenario: this scenario explores the impacts of substituting the EUDR by a biodiversity border adjustment mechanism applied to selected high-risk commodities. This is done by implementing, from 2020 onwards, a tax on the volume of EU imports of selected commodities (soya bean, soya oil, soya cake, palm oil, beef). The tax level is proportional to the difference between the exporting region and the EU in terms of the marginal biodiversity impact of land occupation embedded in the supply of a commodity. This measure accounts not only for local direct impacts but also for remote indirect impacts, such as those arising from feed production in other regions for livestock. The tax rate increases over time, reaching by 2050 a value aligned with medium-range estimates of the global economic value of nature (Costanza et al., 2014). For soya beans exported from Brazil to EU, this implies tax values of about 2, 113 and 221 USD2000 per t by 2030, 2040 and 2050, respectively. This scenario provides insights into the implications of EU mobilizing an alternative instrument to the EUDR for reducing the footprint of its imports, based on a biodiversity-focused pricebased approach. 10. EU-Mercosur RTA + EU demand-side measures (EUM+EUDemSide) scenario: this scenario explores the impacts of substituting the EUDR by demand-side sustainability efforts similar to those assumed applied globally in the IAP scenario (waste reduction, shift from animal- 17 based to plant-based products). This is done by implementing in the EU27 a progressive transition from 2030 onwards, to reaching by 2050 a 50% reduction in food waste and a substitution of 50% of the demand in animal-based products (meat and dairy) by plantbased products. This scenario provides insights into the implications of EU focusing on demand-side efforts as an alternative instrument to the EUDR for reducing the footprint of its imports. As part of an effort to improve the representation of land use dynamics in Brazil, key policy and private sector interventions including the Forest Code and the ASM were parameterized in the model. As further detailed in appendix and in D6.5, the parameterization of the Forest Code assumes restrictions on illegal deforestation in the Amazon and Cerrado biomes from 2020 onwards, adjusted by enforcement probabilities, together with restoration efforts from 2030 onwards modelled as a conversion of cropland and pasture into protected forest. For the ASM, no expansion in the land area dedicated to soy in the Amazon biome is assumed. These features apply both to the BAU scenario and all the intervention scenarios (except for the EUM+EUDR+WeakLURscenario where the ASM is removed from 2030 onwards). 3.2. Biodiversity loss indicators This study uses outcomes from D7.2 and D7.3 in terms of species richness loss as PDF·yr based on LC-IMPACT endpoint CFs. These express the potential damage to Areas of Protection (AoPs), in this case, ecosystem quality, by distinguishing several environmental mechanisms that affect both terrestrial and freshwater biodiversity. Specifically, in GLOBIOM, the following environmental pressures have been considered: land stress, water stress, climate change, and freshwater eutrophication (see Table 2). Other drivers available in LC-IMPACT (i.e., photochemical ozone formation, terrestrial acidification, freshwater and terrestrial ecotoxicity, marine eutrophication) could not be linked to GLOBIOM variables, because the related environmental flows are not covered in the model. Based on the work from D6.4, biodiversity loss associated with environmental pollution downstream the supply chain (farming, transport) has also been measured. Additionally, refined indicators from D2.4 have also been assessed, which only capture species richness loss associated with land stress, as shown in Table 2. The CFs for species richness loss associated with land and water stress and freshwater eutrophication were implemented at the highest level of resolution available for Brazil (ca. 50 x 50 km grid) in GLOBIOM, consistent with the definition of Land Unit Identification (LUID) units to model agricultural supply; and then aggregated at the relevant granularity (see Table 2); whereas climate change-driven CFs are global in scope. Other socioeconomic indicators cannot be reported at this level because of a lack of subnational modelling of key variables such as demand, trade, processing, and prices. 18 Table 2. Biodiversity loss indicators considered in GLOBIOM and underlying drivers. Impact assessment method Impact driver Spatial granularity of CF Harmonisation needed for GLOBIOM LC-IMPACT (Verones et al., 2020) Climate change World Not applicable Land stress Ecoregions based on share of ecoregion area in LUID Water stress 0.05° by 0.05° based on share of water use in LUID (CWATM data) Freshwater Eutrophication Freshwater ecoregions (FEOW) based on share of FEOW Fertiliser and manure application in LUID (Potter et al., 2012) D2.4 (Oliveira et al., 2019, 2024) Land stress Ecoregions based on share of ecoregion area in LUID 3.3. SDG related indicators The following indicators were measured through the ex-ante scenario assessment, which can be related to specific SDG targets. 3.3.1. Environmental impacts We focus on indicators reflecting changes in land and water use, GHG emissions and biodiversity impacts. This leads to four types of indicators: • Changes in land cover areas (SDG15): areas of land (in million ha) for four land use categories (cropland, pasture, forest, and other natural land). The projected values reflect the initial land use and land cover for the year 2000 and explicitly simulated conversions between individual land uses at the subnational level (e.g., ca. 50 by 50 km in Brazil, coarser in the rest of the world – see D6.2) at each decadal time step, accumulated throughout reported time horizon. Results for each time step arise from agricultural production changes in response to the simulated agricultural market dynamics and LUC, subnational land endowments, and technological changes (e.g., long-term yield improvements, adoption of alternative technologies for irrigation, etc.). ➔ Since agricultural land expansion is the most important driver of deforestation and natural cover loss globally (West et al., 2025), this indicator is related to SDG15 Life on Land, e.g., Target 15.2: By 2020, halt deforestation, restore degraded forests and substantially increase afforestation and reforestation globally. • Changes in water use (SDG6): blue water withdrawn for irrigation (in km3). The projected values reflect the initial distribution of irrigated harvested area for the year 2000 and agricultural production changes in relation to regional agricultural market dynamics, subnational surface water, and groundwater endowments, and technological changes (e.g., long-term yield improvements, adoption of alternative technologies for irrigation, etc.). 19 ➔ Globally, agricultural irrigation accounts for ∼70% of the total freshwater withdrawal and 80–90% of human water consumption (Hoekstra & Mekonnen, 2012). This indicator is related to SDG6 Clean Water and Sanitation, e.g., Target 6.4: By 2030, substantially increase water-use efficiency across all sectors and ensure sustainable withdrawals and supply of freshwater to address water scarcity. • Changes in GHG emissions (SDG13): GHG emissions from agricultural activities (primarily N2O emissions from cropland and pastures, and CH4 from enteric fermentation in ruminant animals, manure management and rice cultivation) and LUC (CO2 emissions through changes in above-ground carbon stocks). The projected values reflect the initial distribution for the year 2000 and subsequent dynamics in agricultural activities, as well as crop and livestock management systems (that differ in emission intensity). Emissions from crop production, livestock production, and LUC are reported as separate categories. ➔ This indicator is directly related to SDG13 Climate Action, e.g., Target 13.2: Integrate climate change measures into national policies, strategies and planning; Indicator 13.2.2: Total greenhouse gas emissions per year. • Changes in biodiversity impacts from local production (SDG15): species richness loss from agricultural activities in PDF·y. The projected values represent of biodiversity impacts on freshwater and terrestrial ecosystems at the endpoint level (Verones et al., 2020), associated with the simulated environmental pressures (climate change, land occupation, land transformation, water stress, and freshwater eutrophication – see section 3.2 and Methods in D6.3 for further details). ➔ This indicator is directly related to SDG15 Life on Land, e.g., Target 15.1: Ensure the conservation, restoration and sustainable use of terrestrial and inland freshwater ecosystems and their services; Target 15.5: Take urgent and significant action to reduce the degradation of natural habitats, halt the loss of biodiversity and, by 2020, protect and prevent the extinction of threatened species. 3.3.2. Socioeconomic impacts GLOBIOM quantifies different socioeconomic indicators that cover various supply chain steps, from production to consumption, as well as various groups of commodities: e.g., primary crop products with a distinction of soy vs other crops, secondary products – e.g., protein cakes and vegetable oils – and primary livestock products altogether. This deliverable specifically focuses on the following indicators: • Changes in food availability (SDG2). Food availability is a consumer-oriented metric and one of the components of food security (together with food access, utilization, and stability). It represents the quantities of food products available to the consumer in a given country, including through domestic supply and imports. It is estimated as per capita daily availability in terms of energy content (kilocalories per capita per day, kcal/c/d), with a split between plant-based and animal-based products to assess dietary shifts. ➔ This indicator is related to SDG2 Zero Hunger, e.g., Target 2.1: By 2030, end hunger and ensure access by all people, in particular the poor and people in vulnerable situations. • Changes in production and net trade physical volumes (SDG2, SDG17). Production and net trade (defined as the difference between exports and imports) are market balance indicators relevant to both production (e.g., agricultural supply and destination), trade (e.g., net trade balance), and consumption (e.g., domestic demand as the sum of production and 20 net trade). They can be indicative of regions with export dependency for production (e.g., positive net trade representing a significant share of production), or import dependency for consumption (e.g., negative net trade representing a significant share of consumption). We report them in physical volumes (e.g., thousand tons of fresh matter), and we distinguish livestock final products from crop final products. Within crop final products, we further distinguish soy, other crops, soy protein meals, other cakes, soy oil, and other vegetable oils. ➔ This indicator can be used to measure progress towards SDG2 Zero Hunger, e.g., Target 2.3: By 2030, double the agricultural productivity and incomes of small-scale food producers; SDG17 Partnerships for the Goals, e.g., Target 17.11: Significantly increase the exports of developing countries, in particular with a view to doubling the least developed countries’ share of global exports by 2020. • Changes in the value of production (SDG2, SDG8). This metric is a production-oriented indicator, defined at the product level as the domestic production volume multiplied by the producer price or gross revenue. This figure represents the total potential income from selling soybeans at market prices, but it does not account for costs. This indictor complements the indicator above on physical production volumes by integrating the impact of changes in market prices that relate to the scarcity of a product as demand and supply co-evolve. It is reported as production value for various primary product aggregates: soy, other crops, and livestock products. ➔ This indicator can be used to measure progress towards SDG2 Zero Hunger, e.g., Target 2.3: By 2030, double the agricultural productivity and incomes of small-scale food producers; and SDG8 Decent Work and Economic Growth, e.g., Target 8.1: Sustain per capita economic growth in accordance with national circumstances; Target 8.2: Achieve higher levels of economic productivity through diversification, technological upgrading and innovation. • Eco-efficiency Ratio (SDG8, SDG12). This indicator was proposed in D6.4, representing the environmental productivity of a product or service in economic terms. It measures Environmental Impact Intensity per Revenue (EIIR) and can be derived from the abovementioned indicators. Specifically, it is calculated as the ratio of the environmental impacts from soy production to the gross economic value of soy output. By including an economic metric (revenue) in the denominator, the eco-efficiency ratio assesses how efficiently economic value is created while minimizing environmental impacts, thereby linking environmental performance directly to economic productivity. In this way, higher values indicate greater environmental impact per USD generated, meaning that the economic activity is less environmentally efficient. Lower values suggest higher environmental efficiency, as less impact is generated per USD of revenue. ➔ This indicator can be used to measure progress towards SDG8 Decent Work and Economic Growth, e.g., Target 8.2: Achieve higher levels of economic productivity through diversification, technological upgrading and innovation; Target 8.4: Improve progressively global resource efficiency in consumption and production and endeavour to decouple economic growth from environmental degradation; and SDG12 Responsible Consumption and Production, e.g., Target 12.2: By 2030, achieve the sustainable management and efficient use of natural resources. The EIIR is calculated in two different ways, as Absolute Eco-efficiency or AbsEIIR (a) or as Composite Eco-efficiency for all impacts or CsEIIR (b). Whereas a) directly gives the specific 21 environmental impact per unit of (gross) revenue (eq. 1), b) provides the normalized ecoefficiency impact score across all impacts (eq. 2). The specific definitions and interpretation of the two indicators are as follows: • AbsEIIR (Absolute Environmental Impact Intensity per Revenue) shows the environmental cost per unit of gross economic output — for example, how much greenhouse gas, water use, or biodiversity loss is caused for each unit of revenue. Lower values mean better ecoefficiency (less impact per dollar earned). • CsEIIR (Composite Environmental Impact Intensity per Revenue) combines several environmental impacts into one normalized score so that different types of impacts can be compared on the same scale. The closer the score is to 0, the more eco-efficient (better); the closer to 1, the less eco-efficient (worse). For CsEIIR, the AbsEIIR values for water scarcity, total GHG emissions, and total biodiversity loss (terrestrial plus freshwater species richness loss) are normalized according to eq. 3 (Min–Max normalization), expressed relative to the maximum value (as a fraction of the maximum impact), where a lower value (closer to 0) indicates better performance (higher eco-efficiency), and a higher value (closer to 1) indicates worse performance (lower eco-efficiency). In eco-efficiency analysis, Min–Max normalization provides a simple, interpretable, and bounded indicator that directly supports scenario comparison and policy communication. Finally, the normalized metrics are used to calculate a CsEIIR per scenario (eq. 2), considering two different sets of weights, i.e., assuming equal weight for each of the three impacts or neglecting water scarcity, since soy is mostly rainfed and the diverse scenario deliver minor differences (eq. 2). It should be noted that, in this case, impacts on biodiversity are also caused by water stress and GHG emissions through the chain of causal effects, but they capture different categories of impact, either at the midpoint or at the endpoint levels. In this way, scenarios with a negative AbsEIIR have lower CsEIIR (closer to 0). While AbsEIIR is useful to measure and compare eco-efficiency for a single indicator across scenarios, NorEIIR enables comparison of the overall sustainability performance of the policy scenarios. Both AbsEIIR and CsEIIR are calculated for Brazil, ROW, and globally to understand potential spillover effects. In this context, spillovers refer to cases where impacts —whether positive or negative— spread to other countries beyond Brazil as the target region. For the scenarios that involve broad demandside interventions (EUM+EUDemSide and IAP), these metrics cannot be interpreted as spillovers in the ROW vs Brazil, since their primary aim is not the mitigation of deforestation-related impacts in Brazil. 𝐴𝑏𝑠𝐸𝐼𝐼𝑅𝑖,𝑠 = 𝐼𝑖,𝑠 / 𝑉 𝑠 (eq. 1) 𝐶𝑠𝐸𝐼𝐼𝑅𝑠= ∑(𝑁𝑜𝑟𝐸𝐼𝐼𝑅𝑖,𝑠 × 𝑤𝑖) 𝑖 (eq. 2) 𝑁𝑜𝑟𝐸𝐼𝐼𝑅𝑖,𝑠 = 𝐴𝑏𝑠𝐸𝐼𝐼𝑅𝑖,𝑠 − min(𝐴𝑏𝑠𝐸𝐼𝐼𝑅𝑖,𝑠) max(𝐴𝑏𝑠𝐸𝐼𝐼𝑅𝑖,𝑠) − min(𝐴𝑏𝑠𝐸𝐼𝐼𝑅𝑖,𝑠) (eq. 3) Where i is the specific environmental indicator (GHG emissions, total biodiversity loss, water scarcity); s is the specific scenario; I is the impact value; V the gross soy production value; and w is the weight assigned to each indicator i. 22 3.4. Trade-off analysis and interdependencies between SDG indicators To assess interdependencies among indicators and quantify the magnitude of the observed correlations, the Spearman rank correlation coefficient was estimated (Myers & Sirois, 2005). It is a non-parametric measure of rank correlation, which assesses how well the relationship between two variables (in this case, pairs of indicators) can be described by a monotonic function (a relationship that consistently increases or decreases, but not necessarily linearly). The Spearman rank correlation coefficient is robust for small sample sizes, skewed distributions, or ordinal-type data. In this case, we use it to assess correlations between scenarios in 2050, the year when all interventions are fully implemented, resulting in a sample of 50 observations per indicator for impacts in Brazil (10 scenarios × 5 biomes, excl. Caatinga, without soy production) and 70 observations for global impacts (10 scenarios × 7 regions). Principal Component Analysis (Jolliffe & Cadima, 2016) is also applied, which is a statistical technique used to reduce the dimensionality of a dataset by transforming correlated variables into a smaller number of uncorrelated variables called principal components, which capture the maximum variance in the data with minimal information loss. Furthermore, trade-offs between environmental and economic performance indicators were systematically assessed by normalizing the impact and values quantifying a unique sustainability metric across scenarios to ultimately rank them according to their normalized EIIR (see eq. 1-3). Different combinations of variables have been considered to assess trade-offs and correlations based on GLOBIOM outcomes for Brazil and the globe, as indicated in Table 3. Table 3. Groups of indicators selected for assessing correlations and trade-offs across scenarios for the year 2050. Regional scope Correlations Indicators Description Brazil Between soy output and environmental impacts Soy production (1000 t) Soy output in Brazil GHG emissions (MtCO2eq/yr) GHG emissions from LUC, total crop and livestock production activities across biomes Water consumption (km3) Irrigation water consumed in crop production across biomes Terrestrial biodiversity loss with LC Impact (PDF.year) Impacts on terrestrial species richness based on LC Impact CFs, driven by both land occupation and land transformation across biomes Freshwater biodiversity loss with LC Impact (PDF.year) Impacts on freshwater species richness based on LC Impact CFs, driven by climate change, water stress and freshwater eutrophication Terrestrial biodiversity loss with CLEVER CFs from D6.3 (PDF.year) Impacts on terrestrial species richness based on CLEVER CFs, driven by both land occupation and land transformation across biomes 23 Brazil Between soy areas, land cover changes, LUC emissions and terrestrial biodiversity Soy area (1000 ha) Soy areas across biomes Pasture area (1000 ha) Pasture areas across biomes Forest area (1000 ha) Forestland areas across biomes Natural land area (1000 ha) Other natural land areas across biomes Restored area (1000 ha) Restored areas across biomes LUC emissions (MtCO2eq/yr) Net GHG emissions from LUC across biomes Terrestrial biodiversity loss with LC Impact (PDF.year) Impacts on terrestrial species richness based on LC Impact CFs, driven by both land occupation and land transformation across biomes World Between soy production, market value, calorie availability in Brazil, EUE and SAS, biodiversity loss and GHG emissions Soy production (1000 t) Output of soy in Brazil, in physical quantities Soy production value (million USD, constant 2000 prices) Output of soy in Brazil, in monetary value Food calorie availability in Brazil (kcal/cap/d) Calories associated with the consumption of plantand animalbased products in country’s food demand Food calorie availability in European Union (EUE) (kcal/cap/d) Calories associated with the consumption of plantand animalbased products in countries’ food demand Food calorie availability in Southern Asia (SAS), incl. China (kcal/cap/d) Calories associated with the consumption of plantand animalbased products in countries’ food demand Terrestrial biodiversity loss with LC Impact (PDF.year) Impacts on global terrestrial species richness based on LC Impact CFs, driven by both land occupation and land transformation across world regions, incl. Brazil Freshwater biodiversity loss with LC Impact (PDF.year) Impacts on global freshwater species richness based on LC Impact CFs, driven by climate change, water stress and freshwater eutrophication across world regions, incl. Brazil GHG emissions (MtCO2eq/yr) GHG emissions from LUC, total crop and livestock production activities across world regions, incl. Brazil 24 4. RESULTS AND DISCUSSION 4.1. Sustainability outcomes Socioeconomic and environmental sustainability outcomes from the scenarios in Table 1 are firstly represented as relative changes to the BAU in 2050, since this is the last year of the period when all policies and constraints are implemented. Error! Reference source not found. shows results for the selected indicators for Brazil as a whole and across underlying biomes, to understand trade-offs from a supply-oriented perspective, i.e., among impacts generated in soy sourcing regions. Figure 2 shows trade-offs among selected indicators for the globe and key specific regions, from a demand-oriented perspective. As can be seen in Error! Reference source not found., all the scenarios lead to similar soybean production output in Brazil, between 210 and 230 Mt in 2025 (Fig. 1a). Only IAP causes a significant reduction, by around 30% (152 Mt). EUM+EUDR and EUM+EUDR_WeakLUR slightly foster soybean production (up to 3%), while EUDR, EUM+EUBBAM, EUM+EUDR_ZNLBra and EUM+EUDemSide cause marginal decreases in production (up to 3.1%). This has implications for total natural land conversion (as the sum of forest and other natural land), since soy production interacts with other land uses, governed by the different policies. As expected, the ZNL policy in Brazil plays a big role across ZNLBra, EUM+EUDR+ZNLBra, and IAP scenarios, leading to expansion in forest and other natural land areas by 67 Mha in Brazil (40 Mha in IAP) relative to the BAU, up to a total of 248 Mha in 2050. The other scenarios have barely any effect on this indicator. IAP leads to additional +120 Mha of restored land in Brazil, while the other scenarios have same restored areas as BAU (18.5 Mha in 2050). As for AFOLU emissions resulting from the sum of GHG emissions from LUC, crop and livestock production, the IAP scenario causes a decrease in total GHG emissions relative to BAU of 3.5 Gt CO2eq globally and 5 Gt in Brazil, with net GHG savings of 4.2 Gt CO2eq in the latter. As indicated in D7.2, this is mainly through the reduction in food consumption of animal products, which leads to a sharp decrease in livestock production as well as in the demand for soy cake and associated soy production in IAP. This lowers GHG emissions from LUC and livestock across regions, especially in Brazil, where there is net carbon sequestration (4.5 Gt CO2eq) from LUC through reduction in pastureland areas and natural areas clearing, also as a result of the increased land conservation and restoration efforts. These GHG savings from LUC occur mainly in Cerrado (-1.4 Gt), Mata Atlantica (-1.3 Gt) and the Amazon (-1.1 Gt). In the BAU, LUC emissions reach 328 Mt CO2eq in 2050, mainly in the Amazon (271 Mt) (Fig. 1b). The ZNLBra and EUM+EUDR+ZNLBra scenarios cause a 97% emissions reduction in Brazil, while the other scenarios have only marginal effects in terms of AFOLU emissions in the country – ranging from -2.2% in EUM+EUDR+WeakLUR to +0.42% in EUM+EUBBAM. As a driver of freshwater biodiversity loss (FreshW.BD) according the LC Impact method, aforementioned decreases in total GHG emissions in IAP lead to a notable decrease in species loss relative to BAU (-127%). ZNLBra, and EUM+EUDR+ZNLBra decrease FreshW.BD loss by about 16%, due to limited natural land loss and associated GHG emissions. Whereas the impacts of EUDR alone are minor, trade disruption (Tr.Dis.) increases FreshW.BD loss relative to BAU by 1.7%, due to the overall net increase in soy production in Brazil (4.85% or +10.7 Mt), which entails greater agricultural inputs consumption, leading to higher eutrophication and water stress. EUM+EUDR_WeakLUR, EUM+EUDemSide and EUM+EUBBAM decrease FreshW.BD 25 impacts in Brazil by 3.3%, 1.7% and 0.7%, respectively, mainly due to lower freshwater ecosystems impacts in Cerrado and Mata Atlantica. Figure 1. Spider charts for sustainability outcomes in Brazil and the soy producing biomes. Results show relative changes (%) of selected indicators across scenarios in 2050, relative to the Business-as-Usual (BAU) scenario. 32 behaviour with IAP and Tr.Dis. showing slightly lower CsEIIRs (0.5). EUM+EUDemSide shows the highest score for the world (0.9), mainly due to freshwater biodiversity loss in ROWexEUE. These latest figures facilitate comparison and enable a clearer classification of scenarios, although at the cost of losing the dimension of water stress, less decisive in comparing scenarios of soyrelated policies. Table 6. Composite Environmental Impact Intensity ratio (CsEIIR) across scenarios (dimensionless), considering two weighting schemes: same weight for the three impacts, or same weight for total biodiversity loss and GHG emissions. CsEIIR_BRA w1 CsEIIR_ROW w1 CsEIIR_WORLD w1 CsEIIR_BRA w2 CsEIIR_ROW w2 CsEIIR_WORLD w2 IAP 0.3 0.7 0.7 0.0 0.5 0.5 EUM+EUDR+ ZNLBra 0.4 0.4 0.4 0.7 0.6 0.5 ZNLBra 0.4 0.4 0.3 0.7 0.6 0.5 EUM+EUDR+ WeakLUR 0.7 0.4 0.4 1.0 0.6 0.7 Tr.Dis. 0.7 0.3 0.3 1.0 0.5 0.5 EUM+EUDR 0.7 0.4 0.4 1.0 0.6 0.7 BAU 0.7 0.4 0.5 1.0 0.6 0.7 EUDR 0.7 0.4 0.5 1.0 0.6 0.7 EUM+EUDemSide 0.7 0.5 0.6 1.0 0.7 0.9 EUM+EUBBAM 0.7 0.4 0.5 1.0 0.6 0.7 4.3 Quantification of interdependencies 4.3.1. Spearman rank correlation results The Spearman rank coefficient captures how well one variable increases/decreases as the other increases, regardless of linearity. The results in Figure 2 confirm the positive correlation between the selected output variables across the eight scenarios (see Table 3). All impacts are positively correlated with the soybean production (in tonnes) in Brazil, especially water stress (0.77), terrestrial biodiversity with LC Impact CFs (Terr.BD, 0.55) and total GHG emissions (0.54). As expected, FreshW.BD is strongly correlated with GHG emissions (0.80), and the two Terr.BD indicators (CLEVER vs. LC Impact) are strongly correlated (0.81), as both are determined by the same land use and LUC effects. FreshW.BD is correlated with Terr.BD since the latter is partly caused by LUC, which represents a large share of GHG emissions that also cause FreshW.BD through climate change. 33 Figure 2. Spearman rank correlation index for environmental output indicators for Brazil. Figure 3. Spearman rank correlation index for land use areas and environmental indicators for Brazil. A more in-depth analysis of the correlations between land use areas, LUC-related GHG emissions, and terrestrial biodiversity loss in Figure 3 shows both negative and positive correlations. This could be expected since some uses expand at the cost of other in the different scenarios. For instance, it is observed that an increase in the forestland area in Brazil weakly correlates with both other natural land (0.32) and restored areas (0.29) in a positive fashion, but 34 shows a strong negative correlation between soy (-0.88) and other crop areas (-0.90), and a relatively weaker negative correlation with Terr.BD impacts (-0.85) and LUC emissions (-0.61). Other natural land areas only show a significant correlation with restored areas (-0.52). Since, by definition, restored areas in the model expand at the expense of agricultural and pasture lands, this outcome can be explained by two mechanisms: (a) in some cases, the agricultural land being restored might have otherwise been abandoned, and (b) in other cases, converting agricultural land back to forest may indirectly drive the conversion of other natural areas into new agricultural land. Finally, Figure 3 indicates that terrestrial biodiversity loss is strongly and positively correlated with the area of other crops (0.87) and soy (0.82), whereas LUC emissions show a higher correlation with the area of other crops (0.72) than with soy (0.59). The same analysis is done for variables related to global consumption of soy (Figure 4). As expected, the calories consumed in Brazil are strongly correlated with the soy production in the country (0.83), and both are positively correlated with the total world GHG emissions and the global Terr.BD loss. The results also show a strong positive correlation between the calories consumed as food in EUE and SAS (0.66), while these are only weakly correlated with the global Terr.BD loss. Negative correlations are observed between the calories consumed in SAS and EUE, and the value of the soy market in Brazil (-0.81 and -0.36, respectively), which indicates that imports in these two regions respond to increases in Brazilian soy prices. The negative correlation coefficient is however lower (higher negative correlation) between the calories consumed in EUE and the soy market value in the ROW (-0.88), which indicates that EUE relies less on Brazilian soy imports than SAS to meet their food calorie demand. Accordingly, calorie availability in SAS is negatively correlated with production of soy in Brazil (-0.53, vs. -0.27 for EUR). Other significant positive correlations exist, showing that increased soy production in Brazil is related to increased GHG emissions (0.37) and Terr.BD impacts (0.48) globally, while FreshW.BD is negatively correlated with Brazilian soy production (-0.32). Global GHG emissions are positively correlated with global Terr.BD (0.83). Figure 4. Spearman rank correlation index for selected environmental and economic indicators for key world regions and the world. LCFE is the freshwater biodiversity impact and LCTE is the terrestrial biodiversity impact with LC Impact method. ROW: rest of the world (except Brazil). 35 4.3.2. Results from the Principal Component Analysis (PCA) PCA reduces the dimensionality of the data by summarizing the main correlation patterns into a few principal components (PCs), identifying the axes that best explain the overall variability in the results. For the analysis of the impacts in Brazil, disaggregated per biomes, the scree plot reveals that two axes explain more than 71% of the variability. The PCA in Figure 5 displays the first principal component (PC1) in x-axis and second principal component (PC2) in the y-axis, where each point represents one of the 50 values (10 scenarios x 5 biomes). We observe strong positive correlations between GHG emissions (EMIS), Terr.BD with CLEVER CFs, and Terr.BD with LC IMPACT CFs. In the PCA, water stress, soy production, and FreshW.BD load positively on PC1 but negatively on PC2, grouping together in the bottom-right quadrant, which means they tend to increase together. In the top-right quadrant we find the biomes in which the scenarios lead to high GHG emissions and Terr.BD impacts, mainly the Amazon. In the bottom-right, we find scenarios-biome combinations that lead to high soy production and high-water stress and FreshW.BD, mainly in Cerrado and Mata Atlantica biomes. Figure 5 shows that the Amazon, Pantanal and Pampa are similarly affected in terms of total impacts and correlations, as are the Cerrado and Mata Atlantica. Figure 5. Principal Component Analysis (PCA) biplot showing the relationships among impacts across Brazilian biomes and their contributions to the first two principal components (PC1 and PC2). Arrows indicate the direction and strength of variable loadings, while points represent individual observations for the biomes projected into the reduced dimensional space. PCA results in Figure 6 show a complementary point of view, without assessing correlations by biome, only by scenario. At the Brazilian level, there is a strong correlation between GHG emissions, soy production, and FreshW.BD, while there is also a positive correlation between Terr.BD with CLEVER CFs and Terr.BD with LC Impact CFs. The scenarios behave very similarly, except for IAP and EUM+EUDR+ZNLBra. As indicated above, IAP increases water stress and 36 decreases GHG emissions, while the other scenarios show Terr.BD impacts similar to BAU (SSP2) with the two sets of CFs. From the demand side, Figure 7 shows strong correlations between the production of soy in Brazil, calories consumed in Brazil, the global Terr.BD impacts and the value of the soy market in ROW, while world GHG emissions are negatively correlated. Moreover, there are strong correlations between the calories consumed in EUE, SAS and the market value of Brazilian soy, as indicated above. Finally, the assessment of correlations between biomes and scenarios shows that the biome is a much more decisive factor in determining the different impact metrics for Brazil than the scenario itself. The observations for the scenarios appear rather grouped in the biplot, while the biomes appear away from each other, except for Amazon, Pampa, and Mata Atlantica (Figure 8). Only IAP one the one hand, and EUM+EUDR+ZNLBra and ZNLBra on the other, show a different behaviour, as highlighted through the results section. Figure 6. Principal Component Analysis (PCA) biplot showing the relationships among impacts in Brazil as a whole and their contributions to the first two principal components (PC1 and PC2). Arrows indicate the direction and strength of variable loadings, while points represent individual observations for the scenarios projected into the reduced dimensional space. SSP2 refers to BAU. 37 Figure 7. Principal Component Analysis (PCA) biplot showing the relationships among impacts in world regions and their contributions to the first two principal components (PC1 and PC2). Arrows indicate the direction and strength of variable loadings, while points represent individual observations for the scenarios projected into the reduced dimensional space. SSP2 refers to BAU. Figure 8. Principal Component Analysis (PCA) biplot showing the relationships among scenarios and biomes and their contributions to the first two principal components (PC1 and PC2). Points represent individual observations for the scenarios-biomes projected into the reduced dimensional space. SSP2 refers to BAU. 38 CONCLUSIONS This deliverable evaluates sustainability and eco-efficiency of policy interventions in Brazil’s soy sector, using quantitative data to rank policy combinations – based on evidence and expert opinions – according to their potential to address biodiversity loss, climate change, water stress, economic returns, and food availability. D7.4 combines results from D7.2 and D7.3 to assess 2050 scenarios through a multi-criteria lens, highlighting trade-offs, interdependencies, and the most effective strategies for promoting sustainability in Brazil and beyond. The assessment of environmental and economic indicators across policy mix scenarios in relation to a BAU scenario shows trade-offs among impacts generated in soy sourcing regions, as well as among impacts at the global level and in key soy consumer regions. While most scenarios maintain soybean output in Brazil between 210–230 Mt in 2025, IAP reduces production by 30% (152 Mt) through conservation and restoration measures and dietary shifts, leading to major benefits in terms of GHG emissions and biodiversity (SDG13, SDG15). IAP reduces GHG emissions and generates net carbon sequestration of 4.5 Gt CO₂eq, mainly in the Cerrado, Mata Atlantica, and Amazon. It also delivers significant biodiversity gains, with freshwater and terrestrial species losses reduced by 127% and 47% respectively, but 4.6% higher water demand in agriculture than BAU. ZNLBra and EUM+EUDR+ZNLBra scenarios deliver more moderate improvements relative to BAU, while Tr.Dis. increases freshwater biodiversity loss slightly. From an eco-efficiency perspective, IAP again stands out, showing negative GHG emissions per USD2000 generated in Brazil and the strongest performance in biodiversity impact intensity, though at the cost of higher water stress. ZNL and EUM+EUDR+ZNLBra offer intermediate improvements with limited spillovers, while other scenarios perform close to BAU. Composite eco-efficiency ratios confirm IAP as the most efficient scenario for Brazil, while for the rest of the world, outcomes are more balanced, with EUM+EUDR+ZNLBra and ZNLBra also performing well. Importantly, the weighting of indicators for the composite indicator affects rankings: when biodiversity and GHG emissions are emphasized, all scenarios converge to similar scores, with IAP remaining the best option for Brazil. Interestingly, EUM+EUDemSide and EUM+EUBBAM have the worst eco-efficiency in the ROW due to rebound effects driven by price changes in soy and soy cake markets. Trade-offs between impacts and biomes must be carefully considered in designing effective soy-related policies, with the Amazon being the region most sensitive to the different policy combinations evaluated and underlying LUC. The Spearman rank coefficient shows positive correlations between soybean production and environmental impacts in Brazil. Freshwater and terrestrial biodiversity loss are strongly correlated with GHG emissions, given the large contribution that LUC makes to total GHG emissions, which also affect freshwater biodiversity through climate change. From the demand side, the calories consumed in Brazil and the soy production in the country are positively correlated with the total world GHG emissions and the global terrestrial biodiversity loss. The results also show a strong positive correlation between the calories consumed as food in EUE and SAS, while negative correlations are observed between these and the value of the soy market in Brazil. This indicates that imports in these two regions respond to increases in Brazilian soy prices, although the results show that EUE relies less on Brazilian soy than SAS to meet their food calorie demand. 39 Absolute eco-efficiency ratios, such as kilograms of CO₂ per unit of economic output, provide a transparent and benchmarkable measure of environmental intensity that can be compared across time, regions, and sectors. However, they treat each impact dimension separately, making it difficult to evaluate trade-offs among indicators like GHG emissions, biodiversity loss, and water stress. A normalized and weighted composite indicator addresses this by scaling individual indicators to a common basis and applying weights that reflect policy priorities, enabling the integration of multiple dimensions into a single score for easier scenario ranking. While this approach supports systematic decision-making and highlights trade-offs, its results are sensitive to normalization choices and weighting schemes, which introduces a degree of subjectivity. Comparing these metrics across scenarios over a given time horizon helps assess how effectively different policy mixes promote the decoupling of agricultural economic activities from environmental impacts. PROJECT OUTPUTS ACHIEVED • Quantification of synergies and trade-offs between biodiversity protection and climate mitigation for the governance scenarios outlined in D7.3. • Analysis of trade-offs and correlations between environmental and economic indicators defined for the soy sector in WP6 • Ranking of policy mix scenarios according to their normalized eco-efficiency, considering spillover effects and trade-offs 40 REFERENCES Alcamo, J. (2008). Chapter Six The SAS Approach: Combining Qualitative and Quantitative Knowledge in Environmental Scenarios. In Developments in Integrated Environmental Assessment (Vol. 2, pp. 123–150). Elsevier. https://doi.org/10.1016/S1574101X(08)00406-7 Barlow, J., França, F., Gardner, T. A., Hicks, C. C., Lennox, G. D., Berenguer, E., Castello, L., Economo, E. P., Ferreira, J., Guénard, B., Gontijo Leal, C., Isaac, V., Lees, A. C., Parr, C. L., Wilson, S. K., Young, P. J., & Graham, N. A. J. (2018). The future of hyperdiverse tropical ecosystems. Nature, 559(7715), 517–526. https://doi.org/10.1038/s41586018-0301-1 Börner, J., Schulz, D., Wunder, S., & Pfaff, A. (2020). The Effectiveness of Forest Conservation Policies and Programs. Annual Review of Resource Economics, 12(1), 45–64. https://doi.org/10.1146/annurev-resource-110119-025703 Cardinale, B. J., Duffy, J. E., Gonzalez, A., Hooper, D. U., Perrings, C., Venail, P., Narwani, A., Mace, G. M., Tilman, D., Wardle, D. A., Kinzig, A. P., Daily, G. C., Loreau, M., Grace, J. B., Larigauderie, A., Srivastava, D. S., & Naeem, S. (2012). Biodiversity loss and its impact on humanity. Nature, 486(7401), 59–67. https://doi.org/10.1038/nature11148 Ceballos, G., Ehrlich, P. R., Barnosky, A. D., García, A., Pringle, R. M., & Palmer, T. M. (2015). Accelerated modern human–induced species losses: Entering the sixth mass extinction. Science Advances, 1(5), e1400253. https://doi.org/10.1126/sciadv.1400253 Chaudhary, A., & Brooks, T. M. (2018). Land Use Intensity-Specific Global Characterization Factors to Assess Product Biodiversity Footprints. Environmental Science & Technology, 52(9), 5094–5104. https://doi.org/10.1021/acs.est.7b05570 41 Chaudhary, A., & Kastner, T. (2016). Land use biodiversity impacts embodied in international food trade. Global Environmental Change, 38, 195–204. https://doi.org/10.1016/j.gloenvcha.2016.03.013 Costanza, R., De Groot, R., Sutton, P., Van Der Ploeg, S., Anderson, S. J., Kubiszewski, I., Farber, S., & Turner, R. K. (2014). Changes in the global value of ecosystem services. Global Environmental Change, 26, 152–158. https://doi.org/10.1016/j.gloenvcha.2014.04.002 Dakos, V., Matthews, B., Hendry, A. P., Levine, J., Loeuille, N., Norberg, J., Nosil, P., Scheffer, M., & De Meester, L. (2019). Ecosystem tipping points in an evolving world. Nature Ecology & Evolution, 3(3), 355–362. https://doi.org/10.1038/s41559-019-0797-2 Dockendorff, C., Fuss, S., Agra, R., Guye, V., Herrera, D., & Kraxner, F. (2022). Committed to restoring tropical forests: An overview of Brazil’s and Indonesia’s restoration targets and policies. Environmental Research Letters, 17(9), 093002. https://doi.org/10.1088/1748-9326/ac8ab2 Escobar, N., Tizado, E. J., zu Ermgassen, E. K. H. J., Löfgren, P., Börner, J., & Godar, J. (2020). Spatially-explicit footprints of agricultural commodities: Mapping carbon emissions embodied in Brazil’s soy exports. Global Environmental Change, 62, 102067. https://doi.org/10.1016/j.gloenvcha.2020.102067 Frank, S., Havlík, P., Stehfest, E., Van Meijl, H., Witzke, P., Pérez-Domínguez, I., Van Dijk, M., Doelman, J. C., Fellmann, T., Koopman, J. F. L., Tabeau, A., & Valin, H. (2019). Agricultural non-CO2 emission reduction potential in the context of the 1.5 °C target. Nature Climate Change, 9(1), 66–72. https://doi.org/10.1038/s41558-018-0358-8