Bidirectional Modeling of Surface Winds and Significant Wave Heights in the Caribbean Sea

Though the ocean is sparsely populated by buoys that feature co-located instruments to measure surface winds and waves, their data of vital importance. However, due either minor instrumentation failure or maintenance, intermittency can be a problem for variable. This paper attempts mitigate loss valuable from two opposite but equivalent perspectives: conventional reconstruction significant wave height (SWH) Caribbean Sea buoy-observed wind speeds (WSP) inverse modeling WSP SWH using long short-term memory (LSTM) network. In direction, LSTM strongly able recreate variable its counterpart with lowest correlation coefficient (r2) measured at 0.95, highest root mean square error (RMSE) 0.26 m/s WSP, 0.16 m SWH. The absolute percentage errors (MAPE) are 1.22% 5%, respectively. Additionally, in event complete instrument absence buoy specific area, Simulating WAves Nearshore (SWAN) model first validated used simulate extreme before, during, after passage Hurricane Matthew (2016). Synthetic then fed joint SWAN—LSTM model, corresponding reconstructed compared observations. Although highly accurate (r2 > 0.9, RMSE < 1.3 m/s, MAPE 0.8%), there remains great room improvement minimizing capturing high-frequency events.