Wave height prediction in the Western Mediterranean using genetic algorithms

This study explores the suitability of a soft computing technique to predict (forecast and hindcast) significant wave heights in some areas of the Mediterranean Sea. Using a 44-year database of meteorological and oceanographic records we train a genetic algorithm, to reconstruct a 44-year time series of significant wave height at different coastal points of the Balearic Islands (Western Mediterranean). The correlation obtained between the numerical model data and the algorithm output is 0.80≤r2≤0.970.80≤r2≤0.97. The equations obtained by the genetic algorithm are then used to predict significant wave heights measured at several buoys. We make use of the equation from the node closest to the buoy using as input wind data from the meteorological station at each buoy. We obtained that SWH at each buoy can be predicted with a 1-h ahead forecast time using only wind data from the previous 12 h with a correlation of 0.87≤r2≤0.910.87≤r2≤0.91.