Long-term ecological quality dynamics and spatial associations in arid and semi-arid Northwest China using an adaptive PRSEI

Long-term ecological monitoring in arid and semi-arid regions requires indices that capture region-specific stressors while remaining comparable through time. We developed an adaptive particular remote-sensing ecological index (PRSEI) for Gansu Province, China, using growing-season kNDVI, wetness (WET), land surface temperature (LST), sandification index (SI), and a PM10-based particulate indicator (TI) from 2000 to 2024. A unified principal component analysis (PCA) with common scaling defined the primary time series, while annual PCA and fixed weights were used for sensitivity testing. Unified PC1 explained 79.12% of total variance, and mean PRSEI increased from 0.343 to 0.419 (slope = 0.00432 yr⁻¹; R² = 0.778). Unified PCA and fixed weights agreed on 99.90% of pixel-level trend directions, whereas annual and unified PCA agreed on only 39.73%, indicating strong methodological sensitivity. Adaptive PRSEI was highly correlated with original PRSEI ( r = 0.991) and traditional RSEI ( r = 0.975), showing the strongest consistency with mapped restoration transitions (81.47%) but weaker consistency with degradation transitions (34.22%). Precipitation, soil type, and land use/land cover were the most stable spatial explanatory factors. Overall, ecological quality improved broadly but heterogeneously, supporting adaptive PRSEI as a complementary multi-stressor index for drylands rather than a universal replacement for RSEI.