Unraveling nonlinear landscape pattern effects on habitat quality: A Yangtze river delta case study for data-driven...

Since the onset of the Anthropocene, urbanization has increasingly altered natural landscape patterns, exerting profound impacts on ecosystems critical for human health. However, existing studies on ecosystem changes driven by landscape pattern dynamics often fail to fully account for their complex nonlinear effects, constrained by the linearity assumptions and black-box nature of conventional analytical models. This study aims to develop a transparent analytical framework to elucidate nonlinear landscape-ecosystem interactions and provide actionable management strategies for sustaining habitat quality (HQ) in the Yangtze River Delta (YRD). We apply interpretable machine learning, using the Light Gradient Boosting Machine (LightGBM) with SHapley Additive exPlanations (SHAP) to quantify the relative importance and nonlinear interactions of landscape metrics, enhancing model transparency and interpretability. Gaussian Mixture Model (GMM) clustering identifies two primary landscape modes by analyzing spatial patterns, enabling targeted ecological assessments. Our findings enhance understanding of the Intermediate Disturbance Hypothesis (IDH), demonstrating that quantifying the dynamic ranges and critical thresholds of key drivers helps maintain moderate landscape dynamics, thereby more effectively supporting ecosystem services. By integrating these landscape modes with environmental thresholds, we delineate key protection zones in the YRD aligned with high-quality habitat clusters. Multiscale modeling informed by spatial heterogeneity reveals that the high-aggregation, low-fragmentation landscape mode-though spatially limited-has a strong positive effect on regional ecology. This underscores the need for landscape-mode-informed multi-scale management. Therefore, this study transforms previously advocated directional recommendations-such as fully considering landscape interactions and setting thresholds-into actionable management pathways. Moreover, analyzing nonlinear mechanisms enables better coordination of current management strategies under specific ecological objectives. This provides policymakers with concrete guidance to advance contemporary environmental management.