Semi-arid regions present unique challenges for maintaining aquatic biological integrity due to their complex evolutionary mechanisms. Uncovering the spatial patterns of aquatic biological integrity in these areas is a challenging research task, especially under the compound environmental stress. Our goal is to address this issue with a scientifically rigorous approach. This study aims to explore the spatial analysis and diagnosis method of aquatic biological based on the combination of machine learning and statistical analysis, so as to reveal the spatial differentiation patterns and causes of changes of aquatic biological integrity in semi-arid regions. To this end, we have introduced an innovative approach that combines XGBoost-SHAP and Fuzzy C-means clustering (FCM), we successfully identified and diagnosed the spatial variations of aquatic biological integrity in the Wei River Basin (WRB). The study reveals significant spatial variations in species number, diversity, and aquatic biological integrity of phytoplankton, serving as a testament to the multifaceted responses of biological communities under the intricate tapestry of environmental gradients. Delving into the depths of the XGBoost-SHAP algorithm, we discerned that Annual average Temperature (AT) stands as the pivotal driver steering the spatial divergence of the Phytoplankton Integrity Index (P-IBI), casting a positive influence on P-IBI when AT is below 11.8 °C. The intricate interactions between hydrological variables (VF and RW) and AT, as well as between water quality parameters (WT, NO–N, TP, COD) and AT, collectively sculpt the spatial distribution of P-IBI. The fusion of XGBoost-SHAP with FCM unveils pronounced north-south gradient disparities in aquatic biological integrity across the watershed, segmenting the region into four distinct zones. This establishes scientific boundary conditions for the conservation strategies and management practices of aquatic ecosystems in the region, and its flexibility is applicable to the analysis of spatial heterogeneity in other complex environmental contexts.

中文翻译：

---

集成机器学习揭示半干旱流域水生生物完整性模式


由于其复杂的进化机制，半干旱地区对维持水生生物完整性提出了独特的挑战。揭示这些地区水生生物完整性的空间模式是一项具有挑战性的研究任务，特别是在复合环境压力下。我们的目标是用科学严谨的方法解决这个问题。本研究旨在探索基于机器学习与统计分析相结合的水生生物空间分析与诊断方法，以揭示半干旱地区水生生物完整性的空间分异格局及变化原因。为此，我们引入了一种结合XGBoost-SHAP和模糊C均值聚类（FCM）的创新方法，成功识别和诊断了渭河流域（WRB）水生生物完整性的空间变异。该研究揭示了浮游植物的物种数量、多样性和水生生物完整性的显着空间变化，证明了生物群落在错综复杂的环境梯度下的多方面反应。深入研究 XGBoost-SHAP 算法，我们发现年平均温度 (AT) 是控制浮游植物完整性指数 (P-IBI) 空间散度的关键驱动因素，当 AT 时，对 P-IBI 产生积极影响低于 11.8 °C。水文变量（VF 和 RW）与 AT 之间以及水质参数（WT、NO-N、TP、COD）与 AT 之间错综复杂的相互作用，共同塑造了 P-IBI 的空间分布。 XGBoost-SHAP 与 FCM 的融合揭示了整个流域水生生物完整性的明显南北梯度差异，将该区域划分为四个不同的区域。这为该地区水生生态系统的保护策略和管理实践建立了科学的边界条件，其灵活性适用于其他复杂环境背景下的空间异质性分析。