Accurate short-term significant wave height forecasting is critical for many activities including coastal navigation, offshore renewable energy operations, climate hazard preparedness, real estate price/insurance modulation, etc. This talk discusses and compares three frameworks for forecasting, i.e., statistical models, deep learning models (LSTMs), and modified large language models (LLMs). All models incorporate historical data on wave heights and coastal winds. The study offers practical insight into the relative strengths of statistical and AI-based approaches in a challenging real-world forecasting environment and is the results of several years of collaborative research at the University of Connecticut supported by the Connecticut Institute for Resilience and Climate Adaptation.
