A Statistical Comparison of Methods for Determining Ocean Surface Winds

Abstract

The performance of various techniques which determine ocean surface winds using information from large-scale analyses and forecast models is discussed. The techniques evaluated are the geostrophic relation, a simple empirical law, National Meteorological Center (NMC) 1000-mb winds, a two-region analytically matched boundary layer, a two-region boundary layer based on Rossby number similarity theory, and the Fleet Numerical Oceanography Center (FNOC) marine winds. Statistical comparisons of the model winds were made with observed buoy and ship winds for wind speed, wind direction, and the vector wind. This study is based on analyses and 24-h forecasts made once a day at 0000 UTC from 3 December 1985 through 6 January 1986 on a 2.5 ? 2.5 degree latitude, longitude, grid. The statistical results indicate that no one Model was clearly the best. The absolute wind speed difference between all the models and observations is, on the average, about 3 m s?1, and the RMS difference is about 4.O m s?1. However, the geostrophic relation was definitely the poorest, as would be expected. Model wind speeds and directions compared better with buoy data (lower RMS differences) than ship data. Furthermore, the study indicated that comparisons with buoys for wind speed were better over the northwest Atlantic than over the northwest Pacific, but the reverse was true for direction. For high wind speed reported by ships (> 22.5 m s?1) all model winds were comparatively lower.