Quantifying nonlinear vegetation loss thresholds in response to cumulative and lagged drought effects in East Africa.

Abstract

Drought is a major environmental stressor affecting ecosystem stability in East Africa, where strong hydroclimatic variability and limited water availability increase vegetation vulnerability. However, common and indicator-specific vegetation responses to cumulative and lagged drought remain insufficiently understood. This study developed a multi-indicator framework to quantify probabilistic vegetation drought-loss thresholds across East Africa during 2001-2022. NDVI, LAI, SIF, and GPP, representing vegetation greenness, canopy structure, photosynthetic activity, and ecosystem carbon uptake, respectively, were integrated with multiscale SPEI. A Copula-Bayes conditional probability framework was used to estimate thresholds at different vegetation low-value levels, while random forest regression and partial dependence analysis characterized nonlinear associations with environmental variables. Vegetation responses were dominated by short-term processes. Dominant cumulative timescales were concentrated within 2-4 months, whereas lagged responses occurred mainly at a 1-month delay. GPP nevertheless showed a distinct 10-12-month lagged tail in some regions, suggesting more persistent drought-related influences on ecosystem carbon uptake. With increasing vegetation-loss severity, thresholds generally shifted toward more negative SPEI values. Temperature, surface soil moisture, and elevation showed the highest predictive importance and nonlinear associations with threshold variability. By comparing four complementary vegetation indicators within a unified framework, this study identifies both shared short-term responses and process-specific differences in persistence and threshold behavior. The results help distinguish apparently stable vegetation with constrained functioning from states in which structure and function decline together, providing a useful basis for drought monitoring, ecological risk assessment, and early warning in East Africa.