Upwind terrestrial influences on soil moisture variability across South America

Soil moisture across South America exhibits coherent spatial patterns shaped by upwind terrestrial conditions, yet their downwind propagation remains poorly quantified. We develop an observation-driven framework integrating atmospheric transport trajectories with deep learning to estimate vegetation's contribution to downwind soil moisture. The transport-informed model outperforms a local-only baseline, with upwind information contributing 67% of predictive power, of which vegetation explains 15%. This vegetation-mediated influence is enhanced over agricultural zones, forming hotspots aligned with land-cover change and drought exposure. The cross-regional influence is organized along major transport corridors and varies non-linearly with climate. In humid forests, downwind moisture supply is highly sensitive to vegetation change, with canopy degradation associated with reduced soil moisture; in semi-arid regions, initial greening is associated with increases, but further greening weakens or reverses the effect. These results quantify vegetation-soil moisture teleconnections, with implications for land-atmosphere coupling in Earth system models and hydrological consequences of land-use and climate change.