Critical phase transitions and early-warning frameworks for ecological networks in typical arid regions

With accelerating global environmental change, arid ecosystems are entering a destabilized state, exhibiting greater fragmentation, impaired ecological functions, and diminished resilience. However, the systemic thresholds and collapse dynamics underlying these transitions remain poorly understood. This study addresses this gap by proposing a multilayer network framework that integrates landscape ecological principles with complex network theory to diagnose structural vulnerabilities and simulate resilience pathways in Xinjiang, a representative inland dryland region. Ecological Security Patterns (ESPs) were reconstructed by integrating three ecosystem services vital to dryland integrity, including water and soil conservation, habitat quality, and carbon sequestration. Ecological sources were identified using Morphological Spatial Pattern Analysis (MSPA), and resistance surfaces were parameterized based on empirical ecological factors. In addition, percolation-based disruption modeling and cascading failure simulations were employed to assess both structural robustness and functional resilience under two disturbance scenarios: random failures and targeted attacks based on node centrality. Three key findings emerge from this study are: (1) Despite a 20.08 % increase in ecological source patches, accompanied by a 0.54 % total area contraction, network cohesion progressively declined due to intensified fragmentation (+21.7 % patch count) and corridor instability (+8.8 % length fluctuation); (2) Analysis of critical transitions reveals that structural vulnerabilities surpass functional vulnerabilities, with deliberate attacks degrading cohesion faster than random failures, as evidenced by hierarchical imbalance (<14 % hub nodes) and north-south disjunction, where annual corridor expansion (1.2 % yr⁻¹) exceeding the system's self-organization capacity, leading to efficiency decay (Δ = −0.004); (3) Pronounced north–south asymmetries in robustness, coupled with an emerging “two cores, one belt” spatial risk configuration, highlight the uneven distribution of ecological security and adaptive capacity across the region. This study integrates structural diagnostics with dynamic failure modeling to develop a scalable method for assessing resilience in dryland ecological networks. It moves beyond conventional static mapping approaches by explicitly capturing the dynamic interplay between network structure, functional response, and external disturbance pressures. This framework enables the identification of pre-collapse behavioral signals and critical vulnerability points, thereby enhancing the precision and adaptability of ecological restoration and conservation strategies.