Investigating ecological network resilience to cascading failures in drylands under multi-hazard attack simulations: A case study of the northern slope area of the Tianshan Mountains, Xinjiang, China

Ecological networks in arid regions face dual pressures from climate-driven hazards and inherent fragility. Traditional purely topological resilience assessments are inadequate under complex disturbances. Hence, we developed a “topology-hazard” coupled simulation framework, within which we used circuit theory to construct the ecological network and Python NetworkX to conduct node-removal attack simulations, and then employed random forest to identify the dominant drivers of resilience decay. Results show that: (1) the network comprises 141 ecological sources and 344 corridors, with high-importance nodes clustered in the southern Tianshan and core ecological source areas to the west and northeast, forming dual vulnerable zones highly susceptible to cascading failures; (2) topology-targeted attacks have a much lower collapse threshold (40.9%) than hazard-topology coupled attacks (74.5%), indicating that purely topological approaches underestimate resilience under simulated scenarios; (3) low-temperature stress yields the lowest collapse threshold (84.7%), whereas compound hazard stress yields the highest (98.5%), indicating that hazard action modes modulate resilience responses; (4) network density decay contributes 53.2% of resilience decline and transmissibility decay 26.3%, together accounting for 79.5% and confirming density loss as the dominant driver. Based on these findings, we propose a three-tier protection strategy that prioritizes “dual vulnerable units” to enable precise resource allocation for multi-hazard collaborative prevention and network optimization. This study innovatively constructs a “topology-hazard” coupled simulation framework, which advances quantitative resilience assessment for arid regions and offers a practical pathway for multi-hazard collaborative prevention and ecological network optimization.