Graduate Exam Abstract

Abstract: Numerical weather prediction (NWP) models such as the Global Forecast System (GFS) play a key role in weather forecasting but have inherent systematic errors caused by numerical approximations, parameterization errors, and uncertainty in boundary conditions. These biases can lead to large forecasting errors that affect renewable energy management, disaster prevention, and resource planning. To address this issue, this study utilizes machine learning to correct the biases of forecasts from the GFS, with an emphasis on 2-m temperature (2m-T) and 10- and 100-m wind speed (10m-WS and 100m-WS) correction. In particular, two machine learning models based on XGBoost and U-Net are designed and implemented. This study demonstrates machine learning's ability to enhance NWP forecast accuracy through post-processing. It contributes to more precise, data-driven bias correction methods that can be applied to improve operational weather forecasting and inform decision-making in climate-sensitive areas.

Adviser: Haonan Chen
Co-Adviser: N/A
Non-ECE Member: Yanlin Guo, Civil and Environmental Engineering
Member 3: Anura Jayasumana, Electrical and Computer Engineering
Addional Members: N/A

Publications:
N/A

Program of Study:
CS510
ECE571/575
ECE578
CS430
ECE556
MATH569A
MATH569B
MATH569C