Towards the characterization and removal/mitigation of scatterometer wind sampling errors

Abstract

Surface winds derived from Earth Observation satellites are increasingly required for use in operational monitoring and forecasting of the ocean. A drawback of space-borne wind observing systems, such as scatterometers, is that they provide time and space scales unsuitable for, among others, high- resolution ocean model forcing. As such, blended ocean forcing products combining scatterometer data and numerical weather prediction (NWP) output, are being developed over the past few years [Bentamy et al., 2009]. These products, which provide global coverage at increased temporal resolution, however generally resolve spatial scales closer to NWP (200 km) rather than scatterometer (25 km) scales. More recent techniques include the use of a simplified Boussinesq-type dynamical model, which is constrained by the ECMWF pressure field on the larger scales, to spatially and temporally propagate scatterometer winds [Harutyunyan and Stoffelen, 2011]. Preliminary results on the blended wind product show that it captures shorter scales than the ECMWF wind output, although it does not yet resolve the scatterometer scales. So far, a limitation of the blended product is that only the C-band ASCAT wind data have been used in the assimilation step. With the recent availability of the Ku-band Oceansat-2 (OSCAT) and HY-2A (HSCAT) scatterometer wind data, a substantial improvement of scatterometer data coverage, and thus of the blended product quality, is expected. Thus, to achieve a high resolution mapping of the daily cycle we intend to merge wind data from several scatterometers, and by that beable to resolve spatial scales of about 25 km. Prior to merging different scatterometer data sources, a comprehensive characterization of the scatterometer sampling errors is required. Using data from 2013, we provide an assessment of the sampling errors for the tandem scatterometer dataset composed by ASCAT-A/B, Oceansat-2 and HY-2A, which, so far offers the most complementary orbits in terms of global daily coverage. We analyse these results both globally and for different oceanic regions, namely the Mediterranean Sea and the North Atlantic Ocean, and for each satellite swath, also addressing the relative importance of temporal versus spatial sampling. Preliminary results on different strategies used to remove/mitigate these sampling errors will also be presented