Ensemble Data Assimilation with the NCEP GFS model

Ensemble-based data assimilation techniques are now being actively explored as possible replacements for 3-dimensional or 4-dimensional variational analysis systems. With ensemble-based methods, parallel cycles of ensemble forecasts and updates to the observations are conducted; the ensemble of forecasts is used to estimate forecast-error statistics during the data assimilation step, and the output of the assimilation is a set of analyses, which are used as the initial conditions for the next ensemble of short-range forecasts. Are ensemble data assimilation schemes competitive with operational NCEP 3D-Var method using realistic numerical weather prediction models and observation data sets? Ideally, such a test would demonstrate the ability to assimilate the current full observational data set, including satellite radiances, using a high-resolution model and a large ensemble. However, computational expense of ensemble-based assimilation methods scale linearly with the number of observations, the number of ensemble members, and the dimension of the model, so a robust test like this is still not computationally feasible in a research environment. Accordingly, in this study we will explore ensemble-based data assimilation in a moderate-resolution general circulation model (GCM) assimilating a thinned set of observations.