Scholarlink

<p>Lake eutrophication poses a pressing global challenge, yet identifying its spatially heterogeneous drivers remains difficult in complex environments. This study presents an explainable automated machine learning (XAutoML) framework, integrating AutoGluon for robust modeling and KernelSHAP for interpretability, to quantify eutrophication drivers across China. Focusing on the substantial spatial heterogeneity characteristic of continental-scale environments, this study targets China, a region bridging diverse climatic zones and developmental stages, to validate the Hu Huanyong Line (Hu Line) not merely as a demographic divide but as a critical ecological boundary governing lake trophic state. AutoGluon achieved superior predictive performance (R² = 0.8407) compared to traditional models. The SHAP analysis revealed a stark spatial dichotomy in eutrophication drivers aligning with the Hu Line. In western China, natural factors, particularly soil organic matter (SOM) and mean normalized difference vegetation index, functioned as primary buffers, displaying a negative contribution to TSI (protective effect). However, this protection is non-linear and notably weakened by extreme precipitation. Conversely, in eastern China, industrial nitrogen emissions (TN_ID) and agricultural phosphorus inputs exhibited strong positive contributions to TSI, driving eutrophication through synergistic anthropogenic pressures. This study validates the XAutoML framework as a scalable prototype for managing global environmental gradients. By effectively disentangling the complex interactions between anthropogenic pressures and natural background conditions, the framework offers a transferable methodology for achieving precise, spatially differentiated ecosystem management in heterogeneous regions worldwide.</p> Graphical Abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Identification of spatially heterogeneous drivers of lake eutrophication in China with explainable automated machine learning

  • 期刊论文

  • 发表日期：

    2026年3月4日

查看全文
  • Wenjie Qin,
  • Yuyi Zhang,
  • Xianghan Zheng,
  • Jingping Hu,
  • Huijie Hou,
  • Jiakuan Yang

摘要

Lake eutrophication poses a pressing global challenge, yet identifying its spatially heterogeneous drivers remains difficult in complex environments. This study presents an explainable automated machine learning (XAutoML) framework, integrating AutoGluon for robust modeling and KernelSHAP for interpretability, to quantify eutrophication drivers across China. Focusing on the substantial spatial heterogeneity characteristic of continental-scale environments, this study targets China, a region bridging diverse climatic zones and developmental stages, to validate the Hu Huanyong Line (Hu Line) not merely as a demographic divide but as a critical ecological boundary governing lake trophic state. AutoGluon achieved superior predictive performance (R² = 0.8407) compared to traditional models. The SHAP analysis revealed a stark spatial dichotomy in eutrophication drivers aligning with the Hu Line. In western China, natural factors, particularly soil organic matter (SOM) and mean normalized difference vegetation index, functioned as primary buffers, displaying a negative contribution to TSI (protective effect). However, this protection is non-linear and notably weakened by extreme precipitation. Conversely, in eastern China, industrial nitrogen emissions (TN_ID) and agricultural phosphorus inputs exhibited strong positive contributions to TSI, driving eutrophication through synergistic anthropogenic pressures. This study validates the XAutoML framework as a scalable prototype for managing global environmental gradients. By effectively disentangling the complex interactions between anthropogenic pressures and natural background conditions, the framework offers a transferable methodology for achieving precise, spatially differentiated ecosystem management in heterogeneous regions worldwide.

Graphical Abstract