Spatial-Temporal Graph Fusion Transformer for Long-term Water Quality Prediction

Over the past decades of rapid development, the global water pollution problem has become prominent. Accurate water quality prediction can detect the trend and anomaly of water quality changes in advance, thereby taking timely measures to avoid the occurrence of water quality problems. Traditional statistical methods for water quality prediction make it difficult to capture the complex relationship between multiple variables and deep learning models make it difficult to capture temporal dependence and spatial correlation of the water quality simultaneously. To solve the above problems, this work proposes an adaptive and dynamic graph fusion water quality prediction model based on a spatiotemporal attention mechanism named Spatlal-Temporal Graph Fusion Transformer (STGFT). It in-tegrates a spatial attention encoder (SAE), a temporal attention encoder (TAE), an adaptive dynamic adjacency matrix generator (ADMG), and a multi-graph fusion layer. Among them, SAE and TAE are adopted to capture the spatial correlations and temporal characteristics among different water quality monitoring stations, respectively. ADMG generates adaptive and dynamic adjacency matrices to reflect potential spatial relationships in the river net-work. Experimental results with real-life water quality datasets demonstrate that STGFT outperforms current state-of-the-art models regarding prediction accuracy.