Enhancing Surface Data Assimilation Through Removing Systematic Forecast Biases

Near-surface variables, such as temperature and dewpoint temperature at 2-m height (namely, 2-m temperature and dewpoint temperature) and winds at 10-m height (namely, 10-m wind) are essential variables for near-surface weather prediction. However, recent studies and operational practice demonstrated that large errors commonly occur in near-surface variables, even if in many cases when forecasts in the middle and upper levels of the atmosphere are reasonable.  Assimilation of surface observations (e.g., 2-m temperature and 10-m wind) could help improving near surface weather analysis and forecasts, but the impacts of data assimilation are usually limited by the representation of the data and the unknown model errors. 

In this study, a series of data assimilation experiments are conducted to improve the surface data assimilation by removing the systematic forecast biases in the surface weather forecast within the data assimilation cycles. Both ensemble Kalman filtering and variational methods are used and compared with the mesoscale community Weather Research and Forecasting (WRF) model. The problem, preliminary results, as well as potential developments, will be presented with a comprehensive discussion.