Ecological Degradation Diagnosis and Restoration Planning in Resource‐Depleted Cities: Integrating a Modified Remote Sensing Index and Machine Learning

ABSTRACT

Ecological degradation in resource‐depleted cities (RDCs) involves complex surface stripping and loss of soil function, yet current frameworks often lack sensitivity to mining‐specific features and objectivity in spatial planning. To bridge this gap, this study proposes an integrated diagnosis‐to‐optimization framework, demonstrated in Huangshi, China. First, precise ecological diagnosis was achieved by developing a modified remote sensing ecological index (MRSEI) using Sentinel‐2 and Landsat data. A rock index (RI) was specifically incorporated to capture mining‐induced surface exposure, enabling accurate ecological environment quality (EEQ) assessment. Second, the optimal parameter‐based geographical detector (OPGD) characterized key explanatory patterns. Third, biodiversity (BIO), water yield (WY), soil conservation (SC), and carbon storage (CS) were quantified using the InVEST model. And the ecological resistance surface was estimated by comparing eight machine‐learning algorithms, with XGBoost incorporated into the minimum cumulative resistance (MCR) framework for ecological security pattern (ESP) construction. Results indicate that EEQ followed a decline‐then‐recovery trajectory, exhibiting a pronounced south‐high, north‐low spatial gradient. Notably, the combined diagnostic and planning results suggest a restoration‐illusion pattern, wherein visible greening and BIO improvement are not accompanied by a proportional recovery of SC capacity. Land use and land cover change (LUCC) was identified as the primary explanatory factor. Furthermore, the combined diagnostic and planning results suggested a divergence in ecosystem recovery, in which BIO improved while SC remained weak, indicating asymmetric recovery across ecological functions. Consequently, a planning‐oriented restoration network comprising 20 primary and 24 secondary corridors was delineated. This study provides an integrated framework for ecological diagnosis and restoration prioritization in RDCs.