Compound climate extremes shape landslide hotspot heterogeneity in Myanmar and Vietnam

Landslide hotspots refer to spatial clusters of slope failures where event densities significantly exceed surrounding background levels, providing key insight into how climatic, environmental, and anthropogenic factors interact to amplify landslide risk. In Southeast Asia, strengthening monsoon extremes, complex terrain, and rapid human disturbance jointly influence landslide activity, yet the spatial heterogeneity of hotspot patterns remains poorly understood.This study develops an integrated and interpretable framework combining hotspot mapping, multi-scale correlation analysis, Optimal Parameter GeoDetector (OPGD), and machine-learning interpretation to quantify the climatic, geomorphic, and anthropogenic drivers of landslide hotspot heterogeneity in Myanmar and Vietnam during 2015–2018. Optimal spatial resolutions were identified by linking landslide density with Getis–Ord Gi* Z scores across 1–5 km grids. CatBoost achieved the best predictive performance and was interpreted using Shapley Additive Explanations to reveal nonlinear responses and interactions. Results reveal contrasting hotspot regimes: in Myanmar, clustering is reinforced by persistent warm–humid conditions and compound hydroclimatic extremes, whereas in Vietnam hotspot emergence reflects combined effects of climate-driven soil moisture variability, terrain heterogeneity, vegetation patterns, and intensified human activity. By distinguishing hotspot intensity from conventional susceptibility analysis, this framework provides a novel approach for diagnosing cross-regional drivers of landslide clustering and supports climate-informed hazard management.