Work

Diagnosing ecological quality dynamics in a karst restoration belt: Trajectory instability, response breakpoints, and adaptive restoration implications

Year: 2026

Type: article

Abstract: Conventional global attribution models tend to smooth over spatially conflicting ecological signals, thereby limiting the identification of trajectory-dependent association patterns, nonlinear responses, and latent instability in heterogeneous fragile landscapes. To address this limitation, this study developed a diagnostic adaptive assessment framework (DAAF) and applied it to the rocky desertification restoration belt of eastern and southeastern Yunnan, China. Using annual remote sensing ecological index (RSEI) data from 2000 to 2022 as the operational indicator of ecological environmental quality (EEQ), Sen's slope, the Mann–Kendall test, and the Hurst exponent were jointly employed to diagnose EEQ trends, characterize Hurst-based persistence tendencies and potential trajectory instability, and delineate eight dynamic trajectory types. Spatially stratified XGBoost–SHAP analysis, breakpoint detection, and network-assisted mediation analysis were then conducted to identify trajectory-specific environmental associations, response breakpoints, and indirect association patterns. The results showed that 88.29% of the study area exhibited improving EEQ, whereas 34.52% fell within trajectory-instability zones, indicating widespread improvement alongside anti-persistent or fluctuation-prone trajectories. Improving trajectories showed stronger associations with reduced external human pressure, whereas degrading trajectories were associated with weaker vegetation-cover conditions, hydroclimatic stress, and cumulative disturbance patterns. The fitted association between land-use intensity (LUCI) and annual RSEI showed an approximate response breakpoint around 2.09. In addition, FVC and precipitation exhibited separate approximate model-based response breakpoints around 0.63 and 970 mm, respectively, within localized degradation trajectories. Their co-occurring nonlinear patterns were consistent with a process-informed karst rainfall-paradox hypothesis under weakened vegetation protection. Mediation-assisted analysis further identified indirect association patterns linking land-use intensity, cumulative human pressure, and annual RSEI, whereas steep slopes showed terrain-related buffering associations with disturbance patterns. By linking trend diagnosis, stratified attribution, and breakpoint identification, the proposed framework provides a diagnostic workflow for EEQ assessment and spatially differentiated restoration management in fragile landscapes.

Source: Ecological Indicators

DOI: 10.1016/j.ecolind.2026.115526