Abstract Soil temperature is a fundamental variable for the Earth system, yet it remains highly sensitive to climate change, particularly in tropical regions such as Brazil. Despite its importance, no bias-corrected soil temperature dataset based on the latest CMIP6 projections is currently available for the country. Climate projections are model-dependent and often exhibit systematic biases, which makes bias correction an essential step for subsequent applications. To address this gap, we present STEM-BR: Soil Temperature Under Climate Change for Brazil, a new gridded dataset of historical (1950–2014) and future (2015–2100) soil temperature derived from 15 CMIP6 General Circulation Models (GCMs) under the SSP2-4.5, SSP3-7.0, and SSP5-8.5 scenarios. The dataset was generated through systematic regridding, bias correction based on ERA5 dataset, and statistical refinement, and provides a monthly series at 0.25° × 0.25° resolution, including both raw and bias-corrected outputs. We applied the Quantile Delta Mapping (QDM) approach to correct biases in monthly time series of soil temperature at three depths (0.07, 0.28, and 1.00 m). The corrected dataset is available for both historical and future simulations and shows significant improvements in representing observed climatology. Bias correction reduces residual errors in long-term means and distribution tails to within ±10%, while preserving latitudinal, vertical, and climatic gradients. Seasonal cycles are more accurately reproduced, with reduced warm biases and narrower inter-model spread, particularly in dry climate classes where raw ensemble errors are largest. Future projections indicate consistent warming across Brazil, with increases of about +2 °C (mean) to +5 °C (95th percentile) under SSP2-4.5, and up to +5 °C (mean) to +10 °C (95th percentile) under SSP5-8.5 by 2100. Warming is strongest in the Northeast, North, and Central-West regions, while subsurface layers exhibit smoother but still significant changes. By reducing systematic biases and providing multi-model projections across depths and scenarios, SoilFutures-BR fills a critical data gap and offers a valuable resource to support climate impact assessments and adaptation strategies for agriculture, food security, and water resource management in Brazil.
Schwamback et al. (Thu,) studied this question.