Scholarlink

<p>Water resource governance in the Ningxia section of the Yellow River is of great significance to the economy and ecology along its banks. Existing water quality prediction models struggle to separate long-term trends from short-term fluctuations and provide inadequate representations of spatiotemporal dependencies and cross-regional higher-order associations among monitoring stations. Therefore, this paper proposes a model to predict water quality in the Ningxia reach of the Yellow River based on the Wavelet Dual Graph Spatiotemporal Network (WDGS-Net). First, the Multi-Level Cascaded Wavelet Processing (MCWP) Module is employed to decompose and reconstruct water quality data, thereby enabling the adaptive enhancement of multiscale features for both long-term trends and short-term fluctuations. Second, the Shared Attention Directed Graph Construction (SADGC) mechanism is introduced. Within the full historical window highlights the spatial correlation features of monitoring stations that are more influential. Subsequently, by employing Dual Graph Spatiotemporal Network (DGS-Net), the propagation of water quality characteristics between upstream and downstream river reaches and the temporal evolution of higher-order associations among monitoring stations are perceived. Finally, the Topology-Aware Dual Graph Gating (TADG) module is employed to fuse information from the dual graph and obtain the water quality prediction results. This paper performs predictions using pH and dissolved oxygen (DO) data from six monitoring stations. The results show that WDGS-Net outperforms the baseline models, reducing average MAE by 35% and the RMSE by 25%, while improving <InlineEquation ID="IEq1"><EquationSource Format="TEX">\(R^{2}\)</EquationSource><EquationSource Format="MATHML"><math><msup><mi>R</mi><mn>2</mn></msup></math></EquationSource></InlineEquation> by 0.05. The WDGS-Net performs well in multi-step predicting and provides decision support for collaborative watershed management and sustainable water environment governance.</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

Study on water quality prediction model for the Ningxia reach of the Yellow River based on a wavelet dual graph spatiotemporal network

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月4日

查看全文
  • Bingbing Lei,
  • Wenjing Xie,
  • Meng Han,
  • Xiaofeng Wang,
  • Xue Chai

摘要

Water resource governance in the Ningxia section of the Yellow River is of great significance to the economy and ecology along its banks. Existing water quality prediction models struggle to separate long-term trends from short-term fluctuations and provide inadequate representations of spatiotemporal dependencies and cross-regional higher-order associations among monitoring stations. Therefore, this paper proposes a model to predict water quality in the Ningxia reach of the Yellow River based on the Wavelet Dual Graph Spatiotemporal Network (WDGS-Net). First, the Multi-Level Cascaded Wavelet Processing (MCWP) Module is employed to decompose and reconstruct water quality data, thereby enabling the adaptive enhancement of multiscale features for both long-term trends and short-term fluctuations. Second, the Shared Attention Directed Graph Construction (SADGC) mechanism is introduced. Within the full historical window highlights the spatial correlation features of monitoring stations that are more influential. Subsequently, by employing Dual Graph Spatiotemporal Network (DGS-Net), the propagation of water quality characteristics between upstream and downstream river reaches and the temporal evolution of higher-order associations among monitoring stations are perceived. Finally, the Topology-Aware Dual Graph Gating (TADG) module is employed to fuse information from the dual graph and obtain the water quality prediction results. This paper performs predictions using pH and dissolved oxygen (DO) data from six monitoring stations. The results show that WDGS-Net outperforms the baseline models, reducing average MAE by 35% and the RMSE by 25%, while improving \(R^{2}\)R2 by 0.05. The WDGS-Net performs well in multi-step predicting and provides decision support for collaborative watershed management and sustainable water environment governance.