Comparing two methods for wind analysis: Numerical simulations of a severe convective event

One technique that modelers use for mapping observations onto a model grid is distance-dependent weighted averaging (DDWA), whereby observations close to a particular gridpoint have a greater influence on that gridpoint's value than observations farther away. Although this type of analysis for initializing numerical models enjoys less favor than it did back in the 1960s and 1970s, it remains popular for diagnostic studies. This is not to say that traditional DDWA successive correction methods are no longer suitable for numerical prediction models. In fact, it has been argued that such methods may be quite appropriate, especially if there are plentiful data. For example, an option within the current version of the Pennsylvania State University-National Center for Atmospheric Research (PSU-NCAR) Mesoscale Model (MM5) is to update a first-guess field by applying a Cressman weighting function to the observations.

Using analytically-specified observations, Schaefer and Doswell (1979; hereafter SD79) have shown that this "traditional" application of a DDWA successive corrections analysis technique for mapping observations (specifically, the wind components) to a grid is inferior to their proposed variational adjustment technique, which "forces the analyzed wind field to have the measured divergence and vorticity while conserving the area-averaged velocity of a preliminary analysis."

It has been suggested that improved analyses of divergence might be important for improved simulations of heavy rainfall, cumulus convection, and cyclogenesis, for example. Improvements in numerical forecasts may be particularly evident for precipitation events, since an inaccurate specification of the divergence within the model initial conditions is partly responsible for poor precipitation fields during the first few hours of the forecasts.

In this study, we perform two simulations of a severe convective event with the MM5. One simulation uses the "traditional" technique for wind analysis, whereby a DDWA scheme is applied directly to the wind observations to obtain a gridded wind field. The second simulation uses the SD79 approach for wind analysis, whereby the gridded wind field is reconstructed from gridded fields of kinematic variables (along with a first-guess field and suitable boundary conditions) such that the divergence and vorticity associated with the analyzed wind field match those derived from observations. Results from the simulations are compared to determine whether the SD79 wind analysis provides a better simulation of the convective episode.