From the result, the answer is YES. Recently, seasonal forecasting has become a popular and formidable challenge. For forecasting, longer lead times are preferred. However, in traditional numeric models, those errors accumulate over time, and significantly reducing the credibility about ten days ahead. The biggest differences between AI and numeric models lies in the fact that AI doesn’t use known physical equations to predict the weather. Instead, it undergoes extensive data training to analyze the most likely outcomes for the next time. Coupled with powerful GPU units, AI can produce a one-week forecast within 2 seconds, with the results in 48 hours surpassing those of numeric models. There’s no doubt that AI technology represents a groundbreaking change, and meteorologists must adapt by learning new tools, such as Python or R. There’s also a heightened emphasis on quality and accuracy since AI relies on data for training. Expanding expertise in meteorology and climatology knowledge is crucial, because AI lacks physical processes, and we need to explore through other means what the physical implications behind AI predictions are.

Implemented by College of Science
Date: 2023/09/22