Land-use input choice and policy-relevant environmental and sustainability indicators: a sensitivity assessment for Brazil

Global assessments of climate, biodiversity, and ecosystem services rely on projections of land-use change (LUC), typically from integrated assessment models (IAMs) at coarse resolution (25 to 50 km) and downscaled to the 30 to 300 m required by biophysical models. The choice of LUC projection introduces large variation in results, but this variation is rarely assessed. For Brazil, we systematically compare two LUC projections, a regionally calibrated model (Land Use and Cover Change Modelling Environment, LuccME) and a global product (Land-Use Harmonization 2, LUH2), downscaling both to 300 m with the Spatial Economic Allocation Landscape Simulator (SEALS) and computing five sustainability indicators (biomass carbon, biodiversity intactness, water yield, Forest Code compliance, and a diffuse-nitrogen index). The differences are large, from −37 to +45 % for biomass carbon, −20 to +23 % for water yield, −76 to +12 % for the diffuse-nitrogen index, and −0.27 to +0.17 in mean species abundance. The choice between the two projections also flips the Forest Code compliance call on 9 to 26 % of Brazil's land grid cells, and several biome-level shifts reverse sign under an alternative class correspondence, a sensitivity we quantify. Differences of this size could materially affect sustainability analyses, including REDD+ accounting, where a shifted carbon baseline could change results-based payments, as well as biodiversity assessment and nature-related financial-risk analysis.