Scholarlink

<p>With its increasingly serious and continuous need, effective spatiotemporal water quality prediction has become key to effective pollution control and decision-making. Current research primarily focuses on utilizing continuous time monitoring data to predict trends in time series within specific sections. However, the lack of spatially continuous and reliable observations limits the ability to achieve full spatial coverage prediction. To address this limitation, this study proposes an integrated framework, named SELC, which utilizes the Soil and Water Assessment Tool (SWAT), Environmental Fluid Dynamics Code (EFDC), Convolutional Neural Network (CNN), and Long Short-term Memory (LSTM), to predict the continuous spatiotemporal water quality of the Xiaoqing River Basin (China) using discrete cross-section monitoring data and mechanism model simulation. The SELC model framework integration is as follows: The CNN training uses on-site monitoring data and high-resolution spatial simulations from the coupled SWAT-EFDC models. LTSM is used to generate future temporal forcing data for SELC at monitoring sections. The verification results showed that CNN successfully replicated the spatially continuous distribution of pollutants, and the prediction results were highly consistent with the trend, peak position, and minimum value EFDC simulation results. In the verification, the average coefficients of determination (R2) of the model were 0.62 (NH₃-N) and 0.65 (chemical oxygen demand, COD), confirming its reliability. This study achieved high-resolution spatiotemporal water quality prediction by using only segmented monitoring input and future scenario prediction, thus overcoming the limitation of sparse spatial data. This framework provides a practical tool for identifying high-risk pollution areas and periods and supports targeted aquatic environmental 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

Continuous spatial prediction of river water quality based on a novel hybrid physical-data framework

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月25日

查看全文
  • Yinglan A,
  • Yan Cheng,
  • Puze Wang,
  • Guoqiang Wang,
  • Libo Wang,
  • Baolin Xue,
  • Yuntao Wang,
  • Jin Wu

摘要

With its increasingly serious and continuous need, effective spatiotemporal water quality prediction has become key to effective pollution control and decision-making. Current research primarily focuses on utilizing continuous time monitoring data to predict trends in time series within specific sections. However, the lack of spatially continuous and reliable observations limits the ability to achieve full spatial coverage prediction. To address this limitation, this study proposes an integrated framework, named SELC, which utilizes the Soil and Water Assessment Tool (SWAT), Environmental Fluid Dynamics Code (EFDC), Convolutional Neural Network (CNN), and Long Short-term Memory (LSTM), to predict the continuous spatiotemporal water quality of the Xiaoqing River Basin (China) using discrete cross-section monitoring data and mechanism model simulation. The SELC model framework integration is as follows: The CNN training uses on-site monitoring data and high-resolution spatial simulations from the coupled SWAT-EFDC models. LTSM is used to generate future temporal forcing data for SELC at monitoring sections. The verification results showed that CNN successfully replicated the spatially continuous distribution of pollutants, and the prediction results were highly consistent with the trend, peak position, and minimum value EFDC simulation results. In the verification, the average coefficients of determination (R2) of the model were 0.62 (NH₃-N) and 0.65 (chemical oxygen demand, COD), confirming its reliability. This study achieved high-resolution spatiotemporal water quality prediction by using only segmented monitoring input and future scenario prediction, thus overcoming the limitation of sparse spatial data. This framework provides a practical tool for identifying high-risk pollution areas and periods and supports targeted aquatic environmental management.