Climate-policy pathways reshape cropland-expansion-driven deforestation and agricultural-carbon trade-off imbalance in the Amazon

Urbanization, population growth and agricultural commodity demand increasingly shift land-use pressure toward distant rural frontiers, intensifying cropland expansion into tropical forests and creating substantial trade-offs between agricultural production and carbon-sink capacity. Yet where regional climate-policy land-demand pathways translate into forest-conversion risk, and how these risks overlap with agricultural-value and carbon-sink outcomes at fine spatial resolution, remains insufficiently quantified. We couple the Global Change Assessment Model (GCAM) with the Future Land Use Simulation (FLUS) model to assess cropland-expansion-driven deforestation in South America from 1992 to 2050, with the Amazon basin as the focal analytical region, under reference and 1.5 °C -consistent (1p5) pathways. We estimate agricultural value using a data-driven approach and quantify carbon impacts as changes in vegetation carbon-sink capacity. From 1992–2020, Amazon cropland increased by 2.12 × 105 km2 while forests declined by 3.51 × 105 km2, accounting for more than 60% of South America forest loss. Under the reference pathway, forest loss reaches 8.23 × 105 km2, including 1.10 × 105 km2 converted to cropland, causing an annual loss of 1.04 × 1010 kg of forest carbon sequestration while generating 4.79 × 108 GK$ in agricultural revenue. Under the 1p5 pathway, Amazon forest-to-cropland conversion declines to 7.00 × 103 km2, a 93.6% reduction relative to the reference pathway, and the net land-conversion effect shifts toward a carbon-sink gain. These results indicate an asymmetric agriculture-carbon trade-off, in which cropland expansion produces relatively limited agricultural-value gains while imposing large carbon-sink costs. The GCAM-FLUS framework identifies where such trade-offs are spatially concentrated and where mitigation-consistent land-demand pathways may reduce conversion risk. These policy implications should be interpreted together with local governance, tenure and supply-chain conditions, which are not directly simulated in the present modelling framework.