Univariate Deep Learning Models Show Diminishing Returns for Wave Height Forecasting

A new arXiv paper explores the limitations of univariate deep learning models for forecasting significant wave height (Hs). The study found that while models like DLinear, LSTM, PatchTST, ResAttLstm, and Mamba2 can outperform persistence forecasting, architectural differences yield minimal gains on a large dataset. The research suggests that current univariate models have reached diminishing returns, with cross-buoy variance being a larger factor in forecast error than model architecture. Future work should focus on incorporating atmospheric covariates and decomposing Hs into swell and wind-sea components. AI

IMPACT Suggests current univariate deep learning approaches for wave height forecasting have reached their limits, indicating a need for multivariate or physics-informed models.

RANK_REASON Academic paper detailing research findings on model performance. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.LG →

• arXiv
• buoy 41009
• DLinear
• long short-term memory
• Mamba2
• PatchTST
• ResAttLstm

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →