Estimation of global ocean surface winds blending reanalysis, satellite and buoy datasets

Abstract

Data derived from reanalysis projects and satellites are often used to increase the ocean surface wind data's spatiotemporal resolution available at buoy stations. All available datasets present variable accuracies and different performances on wind estimations due to: (i) the proximity to the coast; (ii) the wind strength; (iii) the geographical location; and (iv) the assumption of near-neutral atmospheric stability. An assessment of the effect of these aspects on estimating winds was carried out. To partly reduce the observed inaccuracies, a methodology is proposed to correct available spatiotemporal gridded datasets of ocean surface winds. It resulted in a hybrid model blending reanalysis and satellite, as usual, but additionally enriched with information from buoy measurements. This model shows potential to correct reanalysis datasets even for cases when reanalysis outperforms satellite wind accuracy. This model led to accuracy improvements ranging from 9% to 161% compared with wind estimations from newer and older generations of reanalysis models. The blended model was calibrated and validated between 2011 and 2016 using observations at 155 ocean-moored buoys located in 6 oceanic regions. The estimated wind fields presented the highest accuracy for mid-to-high winds (5–20 m/s) with an overall Mean Absolute Error (MAE) of 0.96 m/s and 11.6° for wind speed and wind direction, respectively.