TSWP demo

Intelligent Weather and Data Science Lab (IWDSL) at National Taiwan University brings together a team of meteorologists who blend deep expertise in severe weather with cutting-edge AI and machine learning. We focus on developing advanced deep learning methods to process radar and satellite observations more effectively, creating AI-based weather prediction models that deliver more accurate forecasts under a changing climate, and conducting scientific research that deepens our understanding of atmospheric processes based on real-world operational data.

Collaborating closely with operational weather centers, we aim to bridge the gap between innovative research and practical forecasting applications — turning scientific advancements into real-world impact.

IWDSL is led by Dr. Buo-Fu Chen, a scientist at the forefront of artificial intelligence and atmospheric science. His research centers on developing data-driven models for predicting severe weather phenomena, including tropical cyclones and mesoscale weather systems.

[LinkedIn] www.linkedin.com/in/chen-buofu
[Google Scholar] https://scholar.google.com/citations?user=BVA_6zRXl-oC&hl=zh-TW&oi=sra

Research Projects

🌪️ Deep Learning for Tropical Cyclones

We use AI to better understand and predict tropical cyclones. Our research combines typhoon forecasting, value-added data processing, and data assimilation to improve prediction accuracy. Recent works include intensity estimation, rapid intensification forecasting, and storm structure analysis. We also develop deep learning models to retrieve 2-D surface winds and analyze convective structures—bridging research and operational forecasting.

🌧️ Deep Learning for Rainfall Forecasting

We apply cutting-edge deep learning to enhance rainfall prediction and nowcasting. Our models improve radar-based QPF, short-term rainfall forecasts, and downscaling to finer spatial and temporal scales. These tools help forecasters better anticipate severe weather and understand rainfall predictability.

🌏 Data-driven Limited Area Mesoscale Prediction (DLAMP)

IWDSL is actively developing the DLAMP model to provide kilometer-scale, high spatiotemporal resolution forecasts. From a computer science perspective, DLAMP explores complex multi-scale interactions through large-output generative AI. From an atmospheric science perspective, it focuses on high-fidelity 3D convection to better represent mesoscale severe weather. The project is closely tied to local challenges, aiming to deliver high-resolution forecasts for Taiwan and the surrounding region.