Improvements in Ku-band scatterometer wind ambiguity removal using ASCAT-based empirical background error correlations

Wind scatterometry is an established technique for accurately measuring wind vectors over the world's oceans. Inversion of the geophysical model function (GMF) in general leads to a number of ambiguous solutions, and thus an ambiguity removal procedure for selecting the best solution is required. In this study, Two-Dimensional Variational Ambiguity Removal (2DVAR) is considered. 2DVAR produces an analysis of the ambiguous scatterometer wind solutions (observations, o) and collocated ECMWF forecasts (background, b), and then selects the ambiguity closest to the analysis. The properties of 2DVAR are determined by the observation and background error variances and by the background error correlations (BECs). Empirical BECs (EBECs) can be obtained from o − b correlations, and their application in 2DVAR is known to have a beneficial effect on ASCAT wind quality, notably in situations where there is a mismatch between the measured and forecasted position of mesoscale structures. In this study EBECs are applied to wind retrieval from Ku-band scatterometer systems, in particular RapidScat on the International Space Station and OSCAT on OceanSat-3. These systems have poor direction skill in the nadir part of the swath, which can be improved by letting 2DVAR take the full wind vector probability density function into account. It will be shown that for Ku-band systems 2DVAR with EBECs yields a wind product that correlates better with buoys and has less quality control flagging than the default product. This is due to greater detail in the 2DVAR analysis. BECs are defined in the wind potential and stream function domain. While EBECs are much broader than the default Gaussian BECs used in 2DVAR, their effect on the analysis in the spatial wind domain is determined by their much narrower second derivatives, as demonstrated by a single-observation analysis.