Wave forecasting plays a major role in offshore operations and research activities. The accuracy of wave forecasting can be improved by wind input derived from a high-quality atmospheric model or by assimilating with wave observations. The present study uses an ensemble-based wave data assimilation scheme to improve the accuracy of wave parameters. Wave model setup over the Bay of Bengal (76⁰E to 100⁰E in longitudinal and 24⁰N to 0⁰S in latitudinal directions). The numerical simulation of the wind vector obtained from ECMWF ERA5 wind datasets with continuous six-hourly data of 0.25⁰x0.25⁰ resolution is used to force the SWAN (wave) model. An ensemble of wave height is generated by perturbing this wind vector from initial states. Ensemble-based data assimilation scheme was applied to improve (update) the significant wave height (Hs) derived from the first guess SWAN model output. This scheme distributes the error forecast over the entire model domain using a gain (G) weight matrix. Gain contours for different observation locations have been reported. Considerable improvements in the wave height measurements are highlighted. This scheme produces an efficiency of 30–50% reducing root mean square error wave height at observation and validation locations.

Keywords: Wave forecasting, Data assimilation, Gain matrix, Significant wave height.