An Effective Approach for Assimilating Radar Reflectivity in a 4DVAR System

Reflectivity assimilation into NWP models remains a challenging issue because of its complex relationship with hydrometeors and hence simplification required for operational data assimilation systems. Challenges also arise from the lack of well-defined dynamical balances between the hydrometers and other meteorological variables and the short predictability time scale of the hydrometeors. A common practice to improve the effectiveness of reflectivity data assimilation is to artificially force the model a “warm start” by adjusting either the latent heat or the relative humidity. This type of “warm start” approaches has been shown to produce positive impact on short-term precipitation prediction for simpler data assimilation techniques such as 3DVAR. Theoretically speaking, the “warm start” approach is not needed when a 4DVAR technique is concerned, because the dynamical model provides the balance between the hydrometeor and the dynamical variables. However, in practice it has been found that the dynamical forcing resulting from the reflectivity data assimilation can be very weak to have a significant impact on the thermal and humidity fields. In this talk, we will present some preliminary results from our recent study on how to effectively assimilate reflectivity data using a 4DVAR technique. Our experiments are conducted using the WRFDA (WRF data assimilation) 4DVAR system. We will show that assimilating the derived 2D rainfall data from reflectivity results in larger analysis increments to the dynamical variables of temperature and water vapor, leading to improved short-term precipitation forecasts over the assimilation of the 3D reflectivity data. A set of single observation tests will be presented to compare the differences of the increments of the two methods. Real-data experiments will also be shown using a heavy rainfall event that occurred in Taiwan to compare the impact of the two methods on precipitation forecasts.