The explicit prediction of intense convective weather using high resolution numerical models is the principal goal of the Center for Analysis and Prediction of Storms (CAPS) at the University of Oklahoma. Of the many difficult questions being addressed in pursuit of this goal, one stands above the rest: to what extent are high-resolution observations, principally from the national WSR-88D network, as well as unobserved quantities retrieved therefrom, needed in order to obtain accurate forecasts of convective phenomena? This question is particularly relevant given that, at the present time, the only WSR-88D data available nationally in realtime and suited for use in numerical weather models are the spatially-
and quantitatively-degraded lowest four tilts from the NEXRAD Information Dissemination Service (NIDS).
To answer the important question posed above, we present results from a data impact study in which the intense, long-lived Lahoma, Oklahoma hailstorm of August 17, 1994 was simulated using the full-physics Advanced Regional Prediction System (ARPS) and a variety of radar data options:
Experiment #1: no radar data (i.e., the model initial state consisted only of a 3-D background field provided by the NCEP RUC analysis);
Experiment #2: the model initial state consisted of a 3-D background field provided by the RUC analysis, but with NIDS reflectivity data
used to specify the moisture and diabatic heating structure of the storm;
Experiment #3: the same as Experiment #2, but with Level II radial velocity and reflectivity data instead of NIDS; and
Experiment #4: the same as Experiment #2, but with Level II radial velocity and reflectivity data, along with the polar and azimuthal wind components and thermodynamic information retrieved using a single-Doppler velocity retrieval algorithm.
Only in Experiment #4 - with full radar data plus the retrieval - does the storm sustain itself and move as observed. As a result, we have demonstrated that, at least in some cases, not only are WSR-88D Level II data essential for the numerical prediction of intense convective weather, but radial velocity data alone are insufficient; one must apply single-Doppler retrieval techniques in order to obtain the complete 3-D flow structure and thermodynamic fields.
Based on these and similar results, we are about to implement a
prototype system for the realtime acquisition of WSR-88D Level II data in the southern Great Plains. Founded upon the RIDDS system though designed for extension to the NEXRAD Open Systems architecture, the prototype infrastructure will utilize the UNIDATA Local Data Manager (LDM) for data transport and organization within the framework of the Oklahoma OneNet network operated by the State Regents for Higher Education. The acquisition system will be scalable and flexible so as to be regionalized for broader
implementation. Raw data will be archived in native spherical coordinates, and as gridded 4-D assimilated datasets produced by CAPS. It is hoped that the first data will be flowing by spring 1999 in order to serve the realtime numerical forecast efforts of the 2nd Storm and Mesoscale Ensemble Experiment (SAMEX). Radars presently scheduled for inclusion are: Oklahoma City, Tulsa, Fort Smith, Wichita, Dodge City, Amarillo, Lubbock, Fort Worth, Fredrick, and Enid.