Assimilation of a ERS scatterometer derived soil moisture index in the ECMWF numerical weal-her prediction system

The European Centre for Medium-Range Weather Forecasts (ECMWF) currently prepares the assimilation of soil moisture data derived from advanced scatterometer (ASCAT) measurements. ASCAT is part of the MetOp satellite payload launched in November 2006 and will ensure the operational provision of soil moisture information until at least 2020. Several studies showed that soil moisture derived from scatterometer data contain skillful information. Based on data from its predecessor instruments, the ERS-1/2 scatterometers we examine the potential of future ASCAT soil moisture data for numerical weather prediction (NWP). In a first step, we compare nine years of the ERS scatterometer derived surface soil moisture index (Theta(S)) against soil moisture from the ECMWF re-analysis (ERA40) data set (Theta(E)) to (i) identify systematic differences and (ii) derive a transfer function which minimises these differences and transforms Theta(S) into model equivalent volumetric soil moisture Theta(S)*. We then use a nudging scheme to assimilate Theta(S)* in the soil moisture analysis of the ECMWF numerical weather prediction model. In this scheme the difference between Theta(S)* and the model first guess Theta(FG), calculated at 1200 UTC, is added in 1/4 fractions throughout a 24 h window to the model resulting in analysed soil moisture Theta(NDG). We compare results from this experiment against those from a control experiment where soil moisture evolved freely and against those from the operational ECMWF forecast system, which uses an optimum interpolation scheme to analyse soil moisture. Validation against field observations from the Oklahoma Mesonet, shows that the assimilation of Theta(S)* increases the correlation from 0.39 to 0.66 and decreases the RMSE from 0.055 m(3) m(-3) to 0.041 m(3) m(-3) compared against the control experiment. The corresponding forecasts for low level temperature and humidity improve only Marginally compared to the control experiment and deteriorate compared to the operational system. In addition, the results suggest that an advanced data assimilation system, like the Extended Kalman Filter, could use the satellite observations more effectively. (C) 2008 Elsevier Ltd. All rights reserved.