PostRainBench: A comprehensive benchmark and a new model for precipitation forecasting

TL;DR

This model is the first deep learning-based method to outperform NWP approaches in heavy rain conditions and highlights the potential impact of the model in reducing the severe consequences of extreme rainfall events.

Abstract

Accurate precipitation forecasting is a vital challenge of societal importance. Though data-driven approaches have emerged as a widely used solution, solely relying on data-driven approaches has limitations in modeling the underlying physics, making accurate predictions difficult. We focus on the Numerical Weather Prediction (NWP) post-processing based precipitation forecasting task to couple Machine Learning techniques with traditional NWP. This task remains challenging due to the imbalanced precipitation data and complex relationships between multiple meteorological variables. To address these limitations, we introduce the \textbf{PostRainBench}, a comprehensive multi-variable NWP post-processing benchmark, and \textbf{CAMT}, a simple yet effective Channel Attention Enhanced Multi-task Learning framework with a specially designed weighted loss function. Extensive experimental results on the proposed benchmark show that our method outperforms state-of-the-art methods by 6.3\%, 4.7\%, and 26.8\% in rain CSI and improvements of 15.6\%, 17.4\%, and 31.8\% over NWP predictions in heavy rain CSI on respective datasets. Most notably, our model is the first deep learning-based method to outperform NWP approaches in heavy rain conditions. These results highlight the potential impact of our model in reducing the severe consequences of extreme rainfall events. Our datasets and code are available at https://github.com/yyyujintang/PostRainBench.

Figures (6)

• Figure 1: An overview of the proposed PostRainBench and CAMT framework. (a) benchmark's attributes. (b) input composition. (c) distribution of the German dataset. The bottom section illustrates our CAMT workflow: (d) NWP inputs undergo processing by the Channel Attention Module, followed by a Swin-Unet backbone. (e) Multi-task learning with hybrid weighted loss.
• Figure 2: An illustrate of NWP post-processing task. NWP predictions $X_t$ with a time sequence length of $L$ is used as input, while rain observation ${y}_t$ is used as ground truth.
• Figure 3: The architecture of Swin-Unet, which is composed of encoder, bottleneck, decoder and skip connections. Encoder, bottleneck and decoder are all constructed based on swin transformer block.
• Figure 4: CSI scores of Korea Dataset for rain and heavy rain classification with lead times ranging from 6 to 87 hours.
• Figure 5: Valiation loss on Germany Dataset with Swin-Unet.