ON THE USE OF SCATTEROMETER WINDS IN NWP Ad Stoffelen KNMI, de Bilt, the Netherlands, Email: Stoffelen@knmi.nl Lars Isaksen, Didier le Meur ECMWF, Reading, UK ABSTRACT Over the last years the processing of ERS scatterometer winds has been refined. Subsequently, High Resolution Limited Area Model, HIRLAM, and ECMWF model data assimilation experiments have been carried out to assess the impact of one scatterometer, ERS-1 and of two scatterometers, ERS-1 and ERS-2, on the analyses and forecasts. We found that scatterometer winds have a clear and beneficial impact in the data assimilation cycle and on the forecasts. Furthermore, ECMWF has shown that ERS scatterometer data improve the prediction of tropical cyclones in 4Dvar, where unprecedented skillful medium-range forecasts result of potential large social-economic value. Nevertheless, scatterometer winds contain much sub-synoptic scale information where the smallest scales resolved are difficult to assimilate into a Numerical Weather Prediction, NWP, model. This is mainly due to the otherwise general sparsity of the observing system over the ocean. In line with this it is found that scatterometer data coverage is very important for obtaining a large impact. In that respect future scatterometer systems such as SeaWinds on QuikSCAT and ADEOS- II, and ASCAT on EPS are promising. 1. INTRODUCTION After the launch of ERS-1 much improvement has been made in the interpretation of scatterometer backscatter measurements and a good quality wind product has emerged (Stoffelen and Anderson, 1997a, 1997b and 1997c). The consistency of the scatterometer winds over the swath makes them particularly useful for nowcasting purposes and several examples of the usefulness of the direct visual presentation of scatterometer winds to a meteorologist can be given. However, we will focus in this work on the data assimilation impact in numerical weather prediction (NWP) as gathered with the ERS-1 and ERS-2 scatterometer winds. As such, the scatterometer data preprocessing system (PRESCAT) has been further developed at KNMI and ECMWF and implemented in the High Resolution Limited Area weather forecasting Model (HIRLAM) and the ECMWF model. Impact experiments are carried out to test the usefulness of the data for the analyses and for short-range forecasts up to two days in HIRLAM. A meteorologically active period where both ERS-1 and ERS-2 scatterometer data are available has been selected to test the impact of an increased data coverage. In Stoffelen and Beukering (1997) these developments are described in detail, and here a brief summary is provided. At ECMWF, ERS-1 and ERS-2 scatterometer winds were used to carry out similar experiments, but at the global scale and for medium-range forecasting. The scatterometer proves beneficial in dynamic weather conditions, though the static 3D assimilation systems are probably not optimal for surface wind assimilation in such conditions. ECMWF carried out 4Dvar scatterometer impact experiments, where the model tendencies are taken into account in the data assimilation cycle, and a dynamically more consistent analysis results. It is shown that scatterometer winds in 4Dvar lead to unprecedented accurate medium-range forecasts of tropical cyclones. Figure 1 OI or 3Dvar analysis of the atmospheric state (vertical) over time (horizontal). © ESA, 1996.