Impact of large-scale climate indices on total nitrogen concentrations in drinking water sources cannot be overlooked

文献类型： 外文期刊

第一作者： Li, Ning

作者： Li, Ning;Yong, Jiayi;Ning, Ziyu;Chen, Miao;Li, Ning;Zhang, Zheng;Chen, Yaxiong;Xiong, Shengwu;Li, Ning;Zhang, Zheng;Chen, Yaxiong;Xiong, Shengwu;Li, Ning;Yong, Jiayi;Ning, Ziyu;Chen, Miao;Xiao, Jingfeng;Bai, Rui;Xin, Shuli

作者机构：

关键词： Water quality; Machine learning algorithms; Harmonized Landsat and Sentinel-2 (HLS); Climate drivers; Spatiotemporal variability

期刊名称：JOURNAL OF HYDROLOGY （ 影响因子：6.3； 五年影响因子：6.9 ）

ISSN： 0022-1694

年卷期： 2025 年 661 卷

页码：

收录情况： SCI

摘要： Total nitrogen (TN) is a crucial indicator of contamination in drinking water sources. Its primary drivers include both natural and anthropogenic factors such as meteorological elements, atmospheric nitrogen deposition, land use and pollutant emissions. However, the impact of large-scale climate indices on TN concentrations has rarely been studied. Investigating this impact is hindered by several challenges, including the identification of an ideal study area, scarcity of TN data and low concentrations with narrow variation range in drinking water sources. This study focused on a representative tropical drinking water source situated within a watershed where has no industrial activities and minimal land use changes. Harmonized Landsat and Sentinel-2 (HLS) satellite imagery from November 2016 to October 2022, combination with manual sampling and laboratory analysis with 147 samples and five machine learning (ML) algorithms, were used to develop TN concentrations models. We then analyzed the spatial and temporal patterns of the resulting gridded TN concentrations estimates, examining 1008 potential climate drivers for TN concentrations changes by 26 meteorological elements and 110 climate indices data with a maximum lag time of six months. Average wind velocity on 6-month lag (summed normalized feature importance scores, SNFIS = 0.37) and the Western Hemisphere Warm Pool Index on 5-month lag (SNFIS = 0.31) were the main climate drivers for TN concentrations, indicating that the impact of climate indices on TN concentrations was similar to that of meteorological elements and should not be overlooked. Other climate indices, including the Eastern Pacific Subtropical High Northern Boundary Position Index, Kuroshio Current Sea Surface Temperature Anomaly Index and Pacific North American Index, also influence TN concentrations. The impact of climate drivers on TN concentrations exhibited a significant lag of approximately 5-6 months. The improved ML model (residual artificial neural networks, ResANN) was the optimal retrieval model for TN concentrations (R2 = 0.68, RMSE = 0.15 mg/L and MPE =-2.25 % in the test set). Higher TN concentrations were observed in nearshore areas and wet seasons. These findings provided a novel insight that climate indices could shaping TN concentrations in drinking water sources, and offered a broader perspective on the significance of climate indices in water environmental management.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序