Thresholds of drought and terrain complexity shape biomasspatterns in South America's Caatinga

Seasonally Tropical Dry Forests (STDFs) are widespread throughout the world and store significant amounts of carbon; however, they are often overlooked in spatial assessments compared to humid tropical forests. The Caatinga, which is the largest seasonally dry tropical forest in South America, covers approximately 862,000 km{superscript 2} in northeastern Brazil and supports millions of people. Unfortunately, its carbon dynamics has not been thoroughly quantified, especially after centuries of land-use transformation and wood extraction that have significantly diminished its biomass stocks. In this study, we model the potential aboveground biomass (AGB) that Caatinga could sustain under current climatic, atmospheric, and topographic conditions. We integrated data from 301 geo-referenced plots along with high-resolution environmental predictors. The principal Component Analysis revealed two main gradients: a hydro-thermal axis dominated by precipitation, temperature, and severity of drought, and a topographic axis reflecting slope, ruggedness, and terrain position. Random Forest models, validated through both random and spatial cross-validation, explained a significant amount of variation in AGB (R{superscript 2} = 0.81, RMSE = 21 Mg ha). Our findings indicated that AGB is highly sensitive to water availability. In particular, biomass increased sharply when the maximum cumulative water deficit (MCWD) was less than -500 mm and annual rainfall exceeded approximately 1,100 mm. In contrast, elevated vapour pressure deficit (VPD) and potential evapotranspiration (PET) were associated with reduced carbon storage. Topographic heterogeneity further influenced AGB, with rugged and concave terrains supporting potential biomass levels more than twice as high as those found in flat, convex areas. Our predictive map reveals a mosaic of low-biomass cores and localized high-biomass refugia, highlighting the dual influence of hydroclimatic and topographic factors. These findings reposition Caatinga as a heterogeneous and dynamic potential carbon reservoir - historically degraded by changes in land-use but still capable of storing substantial carbon. This research offers critical insights for the restoration, conservation and mitigation efforts for climate change in tropical drylands worldwide.