Mitigating Excessive Drying from the Use of Observations in Mesoscale Modeling

Abstract

AbstractObservations can be used to enhance mesoscale model forecasts by both improving the initial conditions and adjusting the model state toward the observations during its forward integration. Since observations are not available at the same density as the model grid points, spatial and temporal error correlations must be specified in order to effectively assimilate the observations into the model. Overdrying is found while applying the Weather Research and Forecasting Model to five cases centered over Southern California. This overdrying is caused by incorporation of observations in the initial conditions both at a discrete time and over a preforecast data assimilation period. Nonphysical dryness results from the assumption that model moisture error at the observation location is correlated with model moisture error at other locations in the model without regard to the relative magnitude of the moisture at those two locations. Modifying the model and its preprocessor to remove this assumption greatly reduces the occurrence of excessive model dryness without degrading the overall model performance in predicting moisture.