On mesoscale analysis and ASCAT ambiguity removal.

The two-dimensional variational ambiguity removal (2DVAR) method provides a spatial analysis of the sampled ocean vector winds to resolve the local Advanced Scatterometer (ASCAT) dual wind vector ambiguity. Like other variational meteorological data assimilation systems in numerical weather prediction (NWP), 2DVAR combines ASCAT observations with prior NWP background information, in this case from the European Centre for Medium-range Weather Forecasts (ECMWF). Although 2DVAR is generally effective, it may select the wrong ambiguity under certain conditions, e.g. when the background mislocates frontal areas or low-pressure centres, or when it misses convective systems. The relative influence of the ASCAT and ECMWF wind fields in the resulting 2DVAR analysis field can be controlled by adjusting the background error spatial correlation structure, and the background and/or observation error variances. In this article an adaptive 2DVAR approach is proposed to improve ASCAT ambiguity removal, using background error spatial correlations estimated from the autocorrelation of observed scatterometer wind components minus ECMWF forecasts, and using observation and background errors estimated from triple collocation analysis on collocated buoy, ASCAT and ECMWF data. The triple collocations are segregated into several categories according to the ASCAT-derived parameters that have proven to be effective in detecting the correct position of frontal lines and low-pressure centres. Verification using a typical cyclone case and collocated ASCAT and buoy winds shows that the 2DVAR analysis as well as the ASCAT ambiguity removal is improved significantly by putting more weight on the ASCAT observations using empirically determined spatial background error structure functions and situation-dependent observation/background error variances. [ABSTRACT FROM AUTHOR]