Long-term ecological monitoring is essential for sustainable management in fragile regions. This study assessed four decades (1986-2024) of ecological evolution in the Ji River Basin-a 1276.64 km(2) transitional loess-gully ecosystem in China's Yellow River Basin-using the Remote Sensing Ecological Index (RSEI). We integrated multi-temporal Landsat images via Google Earth Engine to construct a 40-year RSEI time series. The index couples greenness (NDVI), wetness (WET), heat (LST), and dryness (NDBSI) through principal component analysis, with PC1 explaining > 82% of the variance. Three evolutionary phases were identified: initial degradation (1986-1996), driven by slope cropland expansion; stabilization (1996-2006), coinciding with early 'Grain for Green' policies; and sustained recovery (2006-2024), characterized by the expansion of high-quality zones. We developed a novel resilience zoning framework integrating local spatial consistency, terrain constraints, and functional state (mean RSEI 2016-2024), which delineated three zones: high-resilience refugia (19.37%), moderate-resilience matrix (75.54%), and low-resilience corridors (5.09%). Mid-slope positions (TPI: 1.220-1.510) within moderate-resilience zones demonstrated optimal restoration efficiency, challenging conventional uniform approaches. The findings advocate spatially differentiated strategies-investing in transitional zones, retrofitting degraded corridors, and monitoring stable refugia-to advance the implementation of Sustainable Development Goal 15 in semi-arid regions globally.