Regional-Scale Analysis and Forecasting (RAP) Report

The development of the RAP at the halfway point of the contracted effort is described. The RAP is based on optimum interpolation and rigorous selection by forward stepwise regression of a small subset of the observations available for interpolation to a point. The RAP scheme produces analyses that are nearly as accurate as analyses produced by interpolating the full set of observations to the point. Another unique feature of the RAP is a buddy check that very carefully determines those observations that should be eliminated from affecting the analysis. The scheme used simulated error correlations, which were a function of distance between observations, but will be using error correlation models developed from real data. The correlation models will be stratified by region, season, standard atmospheric levels, forecast lengths, and sensors. The RAP will be used in observing system simulation experiments OSSEs. The simulated observations will be obtained from perfect FGGE-2b databases after insertion of observation errors. The perfect FGGE-2b observations were derived from a 16-day nature forecast run from the ECMWF. The OSSEs will be used to determine the goodness of RAP and the validity of the developed correlation models.