The habitat quality (HQ) in the fragile ecosystem of the middle reaches of the Yellow River Basin (MRYRB) shows spatiotemporal divergence among sub-basins, yet remains unexplored. Based on a series of remote sensing data, an integrated approach combining InVEST-PLUS spatiotemporal modeling with XGBoost-SHAP machine learning method was used to quantify past-to-future HQ divergence across sub-basins in the MRYRB, and decipher nonlinear driving mechanisms of HQ changes under intensifying anthropogenic and climatic pressures. The results showed that HQ steadily increased by 6.3% from 1991 to 2020, with slower growth rate after 2011. Spatially, HQ displayed a northwest-high to southeast-low gradient, and the divergence among sub-basins was quite significant. HQ in the Hekou–Longmen Right Bank improved most rapidly, while HQ in Fen River Basin declined fastest. Future projections indicated continuous decline in HQ in the Longmen–Huayuankou section and the Wei River Downstream. HQ changes were primarily influenced by NDVI, elevation, and nightlight over the past three decades. When NDVI exceeded the threshold of 0.69 and elevation continued to rise without surpassing the threshold of 1220 m, their positive impacts on HQ changes were observed. When nightlight exceeded 1.65 W/m 2 ·sr, HQ declined sharply, highlighting the high sensitivity of HQ changes to human activities in the MRYRB. The framework provides actionable insights for spatially targeted ecosystem restoration and differentiated management strategies in vulnerable watersheds.
Xie et al. (Sun,) studied this question.