Unpacking the Fragmentation–Ecology Nexus: Spatial Heterogeneity, Nonlinear Responses, and Associated Ecological Factors Across Urban Agglomerations in the Yellow River Basin

Ecological quality and landscape fragmentation are closely linked in rapidly urbanizing regions, yet their relationship across different urbanization contexts remains insufficiently understood. This study examined the spatiotemporal dynamics and coupling between the Remote Sensing Ecological Index (RSEI) and Landscape Fragmentation Index (LFI) across the Yellow River Basin Urban Agglomerations from 2000 to 2020. Geographically weighted regression, quantile-based mean-response analysis, and XGBoost–SHAP were applied to characterize spatial heterogeneity, nonlinear responses, and ecological quality drivers. Mean RSEI increased from 0.401 to 0.463, whereas mean LFI increased slightly from 0.474 to 0.486. Long-term trajectories differed across urbanization classes: NUr zones generally showed ecological improvement, NU zones exhibited declining RSEI and a significant reversal in LFI around 2013, and EU zones showed declines in both indices. Core–periphery profiles further revealed regionally heterogeneous and asynchronous RSEI–LFI responses. RSEI–LFI relationships exhibited strong spatial non-stationarity and nonlinear responses, with positive coupling in the slope-based model more prevalent in NUr zones and negative coupling dominant in urbanized zones. Precipitation was the dominant contributor to RSEI spatial variation, whereas topographic conditions and climate trends were more influential in the slope-based model. These findings highlight that fragmentation–ecology relationships are context-dependent rather than universally negative.