Scholarlink

<p>Reliable water quality assessment is essential for sustainable watershed management. However, many conventional water quality index models remain limited by instability or bias in parameter weighting and by the masking of severe pollution when aggregation functions dilute extreme values. To address these issues, a revised WQI framework was developed and applied to the Tanjiang River Basin, a major tributary of the Pearl River in southern China. The framework included four components and used eight representative physicochemical indicators as core variables, including COD, NH₃-N, NO₃⁻-N, TP, DO, CODMn, BOD₅, and pH. Subindices were derived using both linear and nonlinear scaling functions. Indicator weights were estimated using the analytic hierarchy process, the entropy weight method, and LightGBM, and were then integrated through a game theoretic procedure to obtain balanced composite weights. Three aggregation schemes, namely the weighted quadratic mean, the log weighted quadratic mean, and the sinusoidal weighted mean, were further compared to examine their behavior under extreme pollution scenarios. The integrated weighting scheme reduced reliance on any single method and yielded a more balanced representation of indicators related to nitrogen, phosphorus, and dissolved oxygen. The aggregation analysis showed that responses to extreme pollution were scheme dependent and varied across pollutants, leading to different WQI distributions and spatial classification patterns. In the Tanjiang River Basin, the framework highlighted hotspots of nutrient and organic pollution that were less apparent under conventional WQI approaches, providing a more informative basis for basin scale assessment and management.</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

A LightGBM–AHP Integrated Framework for Robust Water Quality Index Evaluation

  • 期刊论文

  • 发表日期：

    2026年5月2日

查看全文
  • Paike Ma,
  • Wenle Chen,
  • Juncai Huang,
  • Wei Liu,
  • Chongchong Zhu

摘要

Reliable water quality assessment is essential for sustainable watershed management. However, many conventional water quality index models remain limited by instability or bias in parameter weighting and by the masking of severe pollution when aggregation functions dilute extreme values. To address these issues, a revised WQI framework was developed and applied to the Tanjiang River Basin, a major tributary of the Pearl River in southern China. The framework included four components and used eight representative physicochemical indicators as core variables, including COD, NH₃-N, NO₃⁻-N, TP, DO, CODMn, BOD₅, and pH. Subindices were derived using both linear and nonlinear scaling functions. Indicator weights were estimated using the analytic hierarchy process, the entropy weight method, and LightGBM, and were then integrated through a game theoretic procedure to obtain balanced composite weights. Three aggregation schemes, namely the weighted quadratic mean, the log weighted quadratic mean, and the sinusoidal weighted mean, were further compared to examine their behavior under extreme pollution scenarios. The integrated weighting scheme reduced reliance on any single method and yielded a more balanced representation of indicators related to nitrogen, phosphorus, and dissolved oxygen. The aggregation analysis showed that responses to extreme pollution were scheme dependent and varied across pollutants, leading to different WQI distributions and spatial classification patterns. In the Tanjiang River Basin, the framework highlighted hotspots of nutrient and organic pollution that were less apparent under conventional WQI approaches, providing a more informative basis for basin scale assessment and management.