Assimilation of surface observations in improving numerical weather prediction

Surface observations are the main sources of conventional observations for weather forecasts. However, in modern numerical weather prediction, the use of surface observations, especially those with data over complex terrain, has been a unique challenge for sometime. There are fundamental difficulties in assimilating surface observations with three-dimensional variational data assimilation (3DVAR). Therefore, in both NCEP global and regional reanalyses, surface 2-m temperature observations were not assimilated.

Along with the rapid development of ensemble-based data assimilation, it is found that the ensemble Kalman filter (EnKF) can overcome some fundamental limitations that the 3DVAR has in assimilating surface observations over complex terrain. Specifically, through its flow-dependent background error term, the EnKF produces more realistic analysis increments over complex terrain in general. The EnKF clearly performs better than 3DVAR, because it is more capable of handling surface data in the presence of terrain misrepresentation.

With the EnKF method, recent results demonstrated that assimilation of surface observations lead to improved high-impact weather forecasts for landfalling hurricanes and flows over the mountainous areas.

For many years, Dr. Eugenia Kalnay has encouraged the author to work in this important area and other related areas. The presentation will combine both science experience and personal lessons learned from Eugenia. The roles of Eugenia as both mentor and role model will be recognized.