127
Ensemble Data Assimilation in the Antarctic Mesoscale Prediction System (AMPS) during CONCORDIASI (2010)
Ensemble Data Assimilation in the Antarctic Mesoscale Prediction System (AMPS) during CONCORDIASI (2010)
![](images/flag.png)
![](images/blue_ribbon.png)
Tuesday, 4 February 2014
Hall C3 (The Georgia World Congress Center )
The Antarctic continent and Southern Ocean are exceptionally challenging regions for numerical weather prediction (NWP) models in part due to the limited quantity and quality of observations of this remote and extreme environment. Thus, a greater degree of uncertainty exists as to the accuracy of the analyses utilized for the model initialization of model-based predictions over this region, which ultimately reduces the confidence of forecasts. The Antarctic Mesoscale Prediction System (AMPS) is currently the only operational mesoscale NWP model employed for the polar regions of the Southern Hemisphere. AMPS uses a specially modified version of the Advanced Research Weather and Research Forecasting (ARW-WRF) model with polar physics modifications, and is initialized using three-dimensional variational (3D-VAR) data assimilation techniques. In this study, we evaluate the impact of ensemble-based data assimilation in the AMPS model by utilizing an Ensemble Adjustment Kalman Filter (EAKF) within the framework of the Data Assimilation and Research Testbed (DART) for an intensive observation period (IOP; September-December 2010) during the CONCORDIASI project. We refer to this experimental data assimilation configuration as AMPS-DART.
We focus our results by first evaluating differences between the analyses using the operational AMPS model (3D-VAR) and AMPS-DART (EAKF). While both configurations utilize a subset of quality-controlled observations from the CONCORDIASI IOP, AMPS-DART overall assimilates far fewer observations and is limited to assimilating surface (METAR), radiosonde (wind, temperature, and relative humidity), aircraft (ACARS), cloud satellite wind, ship, and GPS (COSMIC) observations. Comparisons are subsequently evaluated to assess the impact of the special observation types, such as dropsondes obtained during IOP, on model initialization and forecast skill.