Assessing long short-term memory network significant wave height forecast efficacy in the Caribbean Sea and Northwestern Atlantic Ocean
Abstract
Precise wave forecasts are crucial, but few studies have directly tested artificial intelligence forecast efficacies in different wave regimes. Using ten years of buoy observations and Simulating WAves Nearshore (SWAN) simulations, the wave climates of the Caribbean Sea (CS) and Northwestern Atlantic Ocean (NWAO) are studied from 2010 to 2019. SWAN simulations are used to replace fragmentary buoy observations and then forecasting using the Long Short-Term Memory (LSTM) network is initiated on six sites throughout the CS and NWAO. Although expected, results illustrate that regardless of test site, LSTM forecasts were highly accurate, reaching correlation values of >0.8, root-mean-square errors <0.4 m, and mean average percentage errors of <14% up to 12-hr forecast horizons. Location-specific geographic and metocean attributes led to divergent forecast outcomes between test sites. Forecast correlations were higher near, but not directly under the Caribbean Low-Level Jet, leading to the best forecast results in the western CS, followed by the central CS, and was poorest in the NWAO. It was conclusively determined NWAO and CS wave fields are sufficiently different to ensure that forecasting using information from either subregion on its counterpart would lead to low correlations and unacceptably high levels of error.