Assimilation of ZDR Columns for Improving the Spin-Up and Forecast of Convective Storms in Storm-Scale Models

The assimilation of radar data into storm-scale models is frequently performed to reduce spin-up time and improve short-term forecasts. One common approach is using a cloud analysis, which inserts hydrometeors and perturbations of temperature and moisture deduced from empirical relations with radar reflectivity (Z) in order to induce and sustain updraft circulations in the model. Dual-polarization radar data can provide additional insight into the dynamical and microphysical structure of convection. In particular, “Z_DR columns” are ubiquitous signatures associated with updrafts in deep moist convection and thus areas of saturation and latent heating. Thus far, however, relatively little has been done to leverage these data for numerical weather prediction.

In this study, the Advanced Regional Prediction System’s cloud analysis is modified from its original Z-based formulation to assimilate real polarimetric data. An algorithm is implemented to identify observed Z_DR columns, which is then used to isolate areas of added latent heating and moisture in the cloud analysis. A simple drying scheme is also introduced to lessen the impacts of over-moistening outside of the observed Z_DR columns. Results are compared to the original cloud analysis for a series of cycled model runs for two proof-of-concept experiments for high impact weather cases: the 19 May 2013 central Oklahoma tornadoes and the 25 May 2016 north-central Kansas tornadoes.

Results indicate a substantial improvement in the analyses and forecasts when using the modified cloud analysis. The analyzed vertical velocity fields feature less noise and spurious convection and a more coherent updraft path. One-hour forecast tracks of convection agree better with observed tornado tracks including more realistic reflectivity structures. The improvements are also seen quantitatively in improvements in the equitable threat score and generally reduced biases. These encouraging results warrant further exploration of the assimilation of dual-polarization radar data into storm-scale models.