Extreme Precipitation Nowcasting using Transformer-based Generative Models

TL;DR

This paper introduces NowcastingGPT, a Transformer-based generative model with Extreme Value Loss regularization, for accurate short-term extreme precipitation nowcasting, demonstrating superior performance on KNMI data.

Contribution

The paper proposes a novel Transformer-based generative model with EVL regularization for improved extreme weather event prediction, addressing limitations of existing models.

Findings

NowcastingGPT-EVL outperforms baseline models in extreme precipitation forecasting.

The EVL regularization effectively captures extreme weather events.

Qualitative and quantitative analyses confirm the model's superior accuracy.

Abstract

This paper presents an innovative approach to extreme precipitation nowcasting by employing Transformer-based generative models, namely NowcastingGPT with Extreme Value Loss (EVL) regularization. Leveraging a comprehensive dataset from the Royal Netherlands Meteorological Institute (KNMI), our study focuses on predicting short-term precipitation with high accuracy. We introduce a novel method for computing EVL without assuming fixed extreme representations, addressing the limitations of current models in capturing extreme weather events. We present both qualitative and quantitative analyses, demonstrating the superior performance of the proposed NowcastingGPT-EVL in generating accurate precipitation forecasts, especially when dealing with extreme precipitation events. The code is available at \url{https://github.com/Cmeo97/NowcastingGPT}.