Explicit Prediction of the Moore, OK Tornadic Supercell Using Single-Doppler Retrieved Fields Obtained from WSR-88D Level-II Data

As part of an ongoing research effort into the 3 May 1999 tornado outbreak, the CAPS single-Doppler retrieval/assimilation procedure has been used to initialize short-range, high-resolution numerical predictions of the Moore, Oklahoma tornadic supercell from a time-series of WSR-88D single-Doppler radar observations. The retrieval/assimilation procedure includes the Shapiro two-scalar wind retrieval, a variational velocity adjustment, a Gal-Chen type thermodynamic retrieval and a simple moisture specification step. Retrieved "pseudo-observations" are blended with traditional observations (surface, upper-air, and satellite data) within the ARPS Data Analysis System to create model initial fields that explicitly resolve the incipient tornadic supercell.

A variety of storm-scale prediction experiments are underway including 3-km and 1-km predictions using radar retrieved fields, as well as cycling experiments using retrieved data from successive applications of the retrieval package. In addition, the robustness of the prediction results will be evaluated by examining a simple set of ensemble predictions created using the scaled lagged-average forecast technique. The sensitivity of the predictions to the explicit inclusion of the incipient storm via the single-Doppler retrieval procedure will also be considered.

Results from a preliminary 3-km prediction are quite encouraging. Application of the single-Doppler retrieval procedure at 2200 UTC (using five successive radar data volume scans from the Oklahoma City WSR-88D radar) yields model initial fields that resolve the deviant horizontal wind fields and principal storm updrafts associated with both the dominant right-moving cell and the weaker left-moving cell. The numerical prediction initialized from the retrieved field agress quite well with the observed storm evolution during its long-lived tornadic phase. In particular, the model-predicted storm develops significant cyclonic rotation and a pronounced hook-like appendage on its southwest flank, and maintains these features from more than two hours. The model predicted storm does, however, move slighty faster than its observed counterpart and weakens too rapidly after 0000 UTC. An experimental 1-km prediction (also initialized at 2200 UTC) further illusrates the model's ability to reproduce the tornadic mesocyclone region.