Intercomparison of WRF 3/4-Dimensional Variational (3/4D-Var) Data Assimilation Methods for Antarctic Applications over the Month of October 2007

This study examines the performance of WRF 3/4 dimensional variational (3/4D-Var) data assimilation techniques in Antarctic applications over a full month period of October 2007. The observations collected for the month are conventional observations, Automatic Weather Station (AWS) data, satellite imagery, GPS (Global Positioning System) refractivity and some satellite-retrieved products. WRF 3D-Var assimilates one-time observational data, and WRF 4D-Var assimilates observations in a 6-hour window. The first-guess for WRF 3/4D-Var data assimilations is from the previous 6-hr WRF forecast. A 3-day forecast using WRF V3.1 follows each data assimilation analysis. The domain configuration is exactly the same as the Antarctica Mesoscale Prediction System (AMPS), which has a mother domain with a resolution of 45 km and a 15-km nested domain covering all of Antarctica and the parts of the Southern Ocean. We execute the runs two times (at 0000 UTC and 12000 UTC) a day over the whole month of October 2007. In addition to the data assimilation experiments, we also carried out the control runs starting from GFS analysis. In our study, the forecast error statistics and verifications are conducted to evaluate the performance of WRF 3D-Var and 4D-Var in Antarctic applications. Based on their performance, some conclusions and insights are summarized in our presentation.