High Resolution Ocean Winds Detection using Machine Learning Models

This research presents a new solution for providing increased demand for high-resolution ocean wind forcing needed to support realistic ocean modeling and climate studies. As part of this effort, a novel forcing product is proposed, ERA. ERA, implements algorithmic adjustment to the ERA-Interim (ERAi) reanalysis dataset created by the European Centre for Medium-Range Weather Forecasts (ECMWF), using scatterometer-based wind observations. Methodology aims at correcting local wind vector errors in ERAi globally by fusing data from state-of-the-art scatterometers ASCAT-A/B and OSCAT, cross-validated against independent sources HSCAT. The method steers clear of over-smoothing and retains small-scale, high-frequency wind variability through the merging of several collocated scatterometer estimates within narrow time windows. In addition, spectral analysis is used to enforce geophysical consistency in the corrected wind fields, U10S, that have spatial resolutions as high as 50 km a substantial increase from the ~150 km resolution of global numerical weather prediction models. Compared to ERAi, the new product decreases vector root-mean-square (RMS) differences by around 10%, with significant enhancements occurring particularly in the tropics. This correction scheme improves the simulation of ocean-atmosphere coupling and offers a crucial basis for enhanced maritime safety, weather prediction, and climate modeling. By providing high-resolution ocean wind fields, ERA enhances the accuracy of boundary layer flux estimates, which are the focus of many ocean and climate simulation studies.