Prediction of coastal surface currents using numerical model and soft computing model

Accurate information of surface currents is crucial to a variety of economic and environmental operations relating to marine renewable energy extraction. Although numerical models based on fluid mechanics are capable of providing forecasting information, its establishment process is a challenge for researchers due to difficulty in accurately defining initial and boundary conditions, grid structure and so on. In this paper, a soft computing approach Random Forests (RF) was adopted to predict surface currents covered by a radar system with high density in Galway Bay. The RF model was trained based on taking use of outputs from numerical model Environment Fluid Dynamics Code (EFDC) and observations from a Coastal Ocean Dynamic Application Radar (CODAR) system. Input variable structure was examined in details through experiments. Sensitivity experiments on input variable structure were performed to establish the best RF models for estimating surface currents. Results indicated that the RF algorithm is a promising means to generate satisfactory surface currents over a long prediction period.