Event-based 3D tracking of multi-type drought propagation in Xinjiang’s Mountain-Oasis-Desert system

Study region Xinjiang (XJ), located in northwestern China, is a prototypical arid region characterized by a coupled Mountain-Oasis-Desert (MOD) system. Study focus Conventional 2D analyses cannot fully capture the 3D spatiotemporal dynamics or complete propagation chains of multi-type droughts in arid MOD systems. To address this limitation, we used SPEI, SRI, SSMI, and SEDI to characterize meteorological (MD), hydrological (HD), agricultural (AD), and ecological droughts (ED), respectively, and developed an event-based 3D framework integrating spatiotemporal connectivity clustering with directed acyclic graph (DAG) network analysis. The framework tracks the growth, migration, and dissipation of drought events while reconstructing synchronous and lagged cross-type propagation linkages across XJ during 1960–2024. New hydrological insights for the region A total of 231 3D drought events were identified, including 55 MD, 61 HD, 83 AD, and 32 ED events; collectively, they exhibited functional-zone clustering and centroid-based directional migration, with trajectories broadly aligned with the terrain-river-oasis landscape configuration. DAG analysis revealed 43 propagation event groups, dominated by the linear cascading mode (56%), with parallel and complex modes accounting for 21% and 23%, respectively. Propagation exhibited path-dependent lag structures and asymmetric cross-system coupling, whereas network topology became more complex after 2000. Furthermore, drought propagation sensitivity showed a distinct tripartite spatial pattern: northern XJ was dominated by atmospheric forcing ( P Atmo =24.1%), southern XJ by hydro-agricultural feedback ( P Nexus ≈20%), and eastern XJ by the highest ecological response sensitivity ( P Eco =14%). These findings establish an event-based 3D network perspective for regional drought analysis and provide new insights into nonlinear propagation mechanisms in arid MOD systems.