Using clear-air adjoint-method wind retrievals from the WSR-88D network to initialize a mesocale model in a pre-convective environment

A clear-air adjoint-method (CAAM) wind retrieval scheme is developed to accurately depict wind flow over a mesoscale domain using WSR-88D data. The CAAM is based upon the simple adjoint method that incorporates predictive equations for reflectivity and radial wind, treating them as passive scalars. These equations are used to retrieve the horizontally varying time-mean wind that gives the best estimate of the observed fields during the retrieval period. The optimal simulation of the observed field is obtained through minimizing, in a least squares sense, a cost function representing discrepancies between the estimated and observed quantities such as reflectivity and radial wind.; The CAAM is applied to both idealized atmospheric regimes and real-data cases containing various configurations of atmospheric structures. Accurate retrieval solutions are shown to depend on the initial conditions of the technique and the existence of coherent gradients within the radar observation fields.; Data void regions exist between WSR-88D radars even though they are sensitive enough to detect clear-air returns up to and sometimes greater than 100 km away. Thus, when initializing a mesoscale numerical weather prediction model, observation information from multiple CAAM wind retrievals over a mesoscale region must be spread across the entire model domain. This is accomplished using a Cressman successive correction method of objective analysis with an “air mass” weighting scheme. The weighting scheme does not allow retrieved wind information to be spread across a detected boundary but only within the same air mass in which it resides.; Application of the assimilation technique is demonstrated in a numerical simulation of a mesoscale convective system initiated along a stationary boundary in weak, synoptic flow using retrieved wind information from seven WSR-88D radars. The assimilated retrieved winds significantly affect the horizontal convergence along the stationary boundary producing a more realistic spatial distribution of precipitation in the first couple of hours of the simulation. Furthermore, the orientation of the resultant cold pool and outflow boundary generated by the convective system is modified. Hence, the location and amount of lift along the outflow boundary is altered resulting in a more accurate simulated evolution of subsequent convection.