Graph Dual-stream Convolutional Attention Fusion for Precipitation Nowcasting

TL;DR

The paper addresses precipitation nowcasting across multiple regions by modeling disjoint areas as a spatiotemporal graph sequence. It introduces Graph Dual-stream Convolutional Attention Fusion (GD-CAF), which combines spatial and temporal attention within ST-Attention blocks and uses gated fusion and depthwise-separable convolutions to handle high-dimensional tensor node features directly. Evaluated on seven years of Copernicus CEMS ERA5 hourly precipitation maps over Europe and surrounding regions (16 nodes), GD-CAF outperforms Persistence, SmaAt-UNet, and RainNet, with attention visualizations providing interpretable insights into region-connection dynamics. The approach achieves strong predictive accuracy with fewer parameters and offers practical value for disaster management and weather nowcasting by exploiting higher-order correlations across regions and time steps.

Abstract

Accurate precipitation nowcasting is crucial for applications such as flood prediction, disaster management, agriculture optimization, and transportation management. While many studies have approached this task using sequence-to-sequence models, most focus on single regions, ignoring correlations between disjoint areas. We reformulate precipitation nowcasting as a spatiotemporal graph sequence problem. Specifically, we propose Graph Dual-stream Convolutional Attention Fusion, a novel extension of the graph attention network. Our model's dual-stream design employs distinct attention mechanisms for spatial and temporal interactions, capturing their unique dynamics. A gated fusion module integrates both streams, leveraging spatial and temporal information for improved predictive accuracy. Additionally, our framework enhances graph attention by directly processing three-dimensional tensors within graph nodes, removing the need for reshaping. This capability enables handling complex, high-dimensional data and exploiting higher-order correlations between data dimensions. Depthwise-separable convolutions are also incorporated to refine local feature extraction and efficiently manage high-dimensional inputs. We evaluate our model using seven years of precipitation data from Copernicus Climate Change Services, covering Europe and neighboring regions. Experimental results demonstrate superior performance of our approach compared to other models. Moreover, visualizations of seasonal spatial and temporal attention scores provide insights into the most significant connections between regions and time steps.

Figures (10)

• Figure 1: The input is a spatiotemporal graph sequence, and the output is a spatiotemporal graph.
• Figure 2: (a) GD-CAF architecture overview (b) Double convolution block with pooling (c) ST-Attention block in the $\ell$-th block. (d) Double convolution block with upsampling. The numbers above the double convolutional blocks indicate the number of input and output channels respectively.
• Figure 4: Precipitation data captured from the $4$-th region over a 6-hour time interval.
• Figure 8: MSE per region for GD-CAF, SmaAt-UNet and RainNet.
• Figure 11: True label, baseline models (Persistence, RainNet, and SmaAt-UNet), GD-CAF prediction from six selected regions.