Scholarlink

<p>Water quality and ecological health in Gaoyou Lake are at risk because of agricultural nonpoint source (NPS) pollution. The primary objective is to develop an IoT-based predictive framework that identifies and quantifies agricultural pollution risks in Gaoyou Lake, enabling accurate temporal modeling, early warnings, and actionable decision support for environmental agencies and agricultural stakeholders. A Contracted Fox Search-driven Temporal Convolutional Network-tuned with Self-Attention (CFS-TCN-SA) is proposed. Fox Search Optimization fine-tunes TCN parameters, while SA improves long-range dependency modeling, ensuring precise multi-step forecasting and robust generalization across diverse hydrological and pollution-loading conditions in the Gaoyou Lake watershed. High-frequency data obtained by Internet of Things (IoT) sensor network are nitrate, ammonium, total phosphorus, turbidity, pH, temperature, conductivity, dissolved oxygen, rainfall, and flow. The sensor streams are synchronized in terms of time and the Kalman filter (KF) is used to smooth the noisy measurements, and Min–max normalization is used to provide scale consistency between the variables. The optimized model has higher predictive performance as shown by higher RMSE (0.0068) scores, than baselines as implemented in Python. Risk heatmaps, dynamic IoT time-series comparisons and model-explanation plots are shown in an interactive visualization. The suggested model is effective in forecasting the risks of agricultural NPS pollution in Gaoyou Lake to support the pro-active control. Sustainable watershed monitoring and environmental decision-making can be offered through the integration of IoT sensing, advanced temporal modeling, and interactive visualization, which is scalable, replicable, and sustainable.</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 deep learning and IoT-based model and visualization study for measuring agricultural nonpoint source pollution risks in Chinese Gaoyou Lake

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月6日

查看全文
  • Zhirong Guo,
  • Dong Cao,
  • Qing Wang

摘要

Water quality and ecological health in Gaoyou Lake are at risk because of agricultural nonpoint source (NPS) pollution. The primary objective is to develop an IoT-based predictive framework that identifies and quantifies agricultural pollution risks in Gaoyou Lake, enabling accurate temporal modeling, early warnings, and actionable decision support for environmental agencies and agricultural stakeholders. A Contracted Fox Search-driven Temporal Convolutional Network-tuned with Self-Attention (CFS-TCN-SA) is proposed. Fox Search Optimization fine-tunes TCN parameters, while SA improves long-range dependency modeling, ensuring precise multi-step forecasting and robust generalization across diverse hydrological and pollution-loading conditions in the Gaoyou Lake watershed. High-frequency data obtained by Internet of Things (IoT) sensor network are nitrate, ammonium, total phosphorus, turbidity, pH, temperature, conductivity, dissolved oxygen, rainfall, and flow. The sensor streams are synchronized in terms of time and the Kalman filter (KF) is used to smooth the noisy measurements, and Min–max normalization is used to provide scale consistency between the variables. The optimized model has higher predictive performance as shown by higher RMSE (0.0068) scores, than baselines as implemented in Python. Risk heatmaps, dynamic IoT time-series comparisons and model-explanation plots are shown in an interactive visualization. The suggested model is effective in forecasting the risks of agricultural NPS pollution in Gaoyou Lake to support the pro-active control. Sustainable watershed monitoring and environmental decision-making can be offered through the integration of IoT sensing, advanced temporal modeling, and interactive visualization, which is scalable, replicable, and sustainable.