A Bayesian Estimation Technique for Improving the Accuracy of SCATSAT-1 Winds in Rainy Conditions

Abstract

In this study, we present a mean-of-posteriors Bayesian estimation technique for improving rain contaminated wind speed estimates from the SCATSAT-1 scatterometer, particularly at low and moderate wind speeds. The performance of the algorithm is assessed using the SCATSAT-1 Version 1.1.3 data, collocated National Centre for Medium Range Weather Forecasting (NCMRWF) analysis winds, measurements from the Advanced Scatterometer (ASCAT), and buoy measurements for a 6-month period from April to September 2018. The validation results show a reduction in wind speed bias from 2.6 to $-$ 0.1 m/s and a reduction in root mean square difference (RMSD) from 4.8 to 2.9 m/s relative to NCMRWF. The results from comparisons with ASCAT and buoy data are also qualitatively similar; the RMSD relative to ASCAT reduces from 2.9 to 1.8 m/s after correction, whereas the RMSD relative to buoy reduces from 4.1 to 2.2 m/s. Different methods such as histogram comparisons, across-track, wind speed dependent, and spatial variation of wind speed errors have also been presented to demonstrate the overall improvement achieved by this technique. This algorithm presents two main advantages: 1) the errors in wind speed estimates are significantly lower after correction and 2) this methodology can be applied on both the inner and outer SCATSAT-1 swath.