A COMPARISON OF STRATEGIES TO COUPLE GLOBAL AND LIMITED-AREA DATA ASSIMILATION

An atmospheric limited area model uses time-dependent lateral boundary conditions provided by a global atmospheric model. In current practice, the initial conditions for the limited area model are either analyses prepared using the global model and interpolated to the higher resolution grid of the limited area model, or analyses prepared by using a data assimilation system specifically designed to produce initial states for use by the limited area model. In the latter case, the analysis inside the limited area domain is obtained independently of the global analysis: the global analysis affects the limited area analysis only through the lateral boundary conditions. In this report, we consider the scenario in which we have access to both the global and the limited-area model and a model-independent data assimilation system. Our goal is to begin to addressees the problem of finding that configuration of the coupling between these three components of the forecast system, which provides the best global and limited area model forecasts. In particular, we compare the forecast performance of the system for different potential coupling strategies using both simulated and operationally used observations of the atmosphere. In our experiments, the global model is the model component of the Global Forecast System (GFS) of the National Centers for Environmental Prediction (NCEP) integrated at a T62L28 (about 150 km ) horizontal resolution, the limited area model is the Regional Spectral Model (RSM) of NCEP integrated at 35 km and L28 resolution, while the data assimilation system is the Local Ensemble Transform Kalman Filter.