ASCAT wind products at 25 km grid resolution are assimilated into the Integrated Forecasting System (IFS) at the European Centre for Medium-Range Weather Forecasts (ECMWF). An error of 1.5 m/s is given to the zonal (U) and meridional (V) wind components and the observations are thinned every 100 km in both the along and the across track direction. An alternative way to thinning is to reduce the spatial representativeness error by aggregating (averaging) wind observations up to those scales that are representative of those resolved by the NWP model. This process is called superobbing in data assimilation. ASCAT superobbing research products were generated by averaging the nominal EUMETSAT ASCAT 25-km or 12.5-km grid product winds to lower resolution in an aggregation process. A 2-month test dataset of superobbing 50 km, 62.5 km and 100 km grid ASCAT products was generated and provided together with related bias correction estimation, quality control flags and situation-dependent observation/background errors estimation. Assimilation experiments were run at ECMWF using the nominal and the new superobbing products with a final resolution that ranges from 12.5 km to 100 km. Results showed a reduced standard deviation for analysis and background departures, compared to the ones for the corresponding 12.5 km and 50 km thinned products. The averaging of small-scale variance in the superobbing product will explain (some of) the reduction in standard deviations of the departures. To account for this effect in the observations and for the effect of sample variation due to quality control differences, all experiments were also validated against a fixed resolution and sample set of ASCAT winds in order to obtain a metric for the optimum weight and aggregation of the ASCAT winds for achieving NWP impact. The situation-dependent observation errors provided with the superobbing products were also tested, replacing the constant error value of 1.5 m/s as used in operations

2019 Living Planet Symposium, 13-17 May 2019, Milan, Italy