KAIST XAI Tutorial 2025 | Precipitation Forecasting | Seongyeop Jeong (KAIST, INEEJI)

This presentation introduces an AI-driven approach for ultra-short-term precipitation forecasting. We explore the capability of VQ-GAN and Transformer-based generative models to produce future precipitation fields and examine how expanding the input domain beyond the Korean Peninsula can improve prediction performance. In addition, we discuss a methodology for predicting and correcting the intrinsic forecast errors of the Korea Meteorological Administration’s numerical weather prediction (NWP) models using artificial intelligence. 본 발표에서는 인공지능(AI) 기술을 활용한 초단기 강수 예측 방법을 소개합니다. VQ-GAN 및 Transfo…