18th Conference on Weather and Forecasting, 14th Conference on Numerical Weather Prediction, and Ninth Conference on Mesoscale Processes

Wednesday, 1 August 2001: 10:45 AM
Sensitivity of the 3-4 May 1999 Tornadic Outbreak to Synoptic Scale Initial Conditions
Paul J. Roebber, Univ. of Wisconsin, Milwaukee, WI; and D. M. Schultz and R. Romero
On 3-4 May 1999, central and northern Oklahoma and southern Kansas were raked by 10 supercells that produced a total of 68 confirmed tornadoes. Despite the magnitude of the event, the operators of the National Weather Service (NWS) watch/warning system were able to capitalize on the outbreak geometry, with storm paths running from sparsely to highly populated regions. As a result, there were only 36 fatalities. A review of historical damage and fatality data from major U.S. tornadic outbreaks suggests greater than 1000 fatalities for an event of this magnitude in the absence of such a warning system. Hence, the performance and the resulting lead times in this instance must be considered a forecast success.

However, forecasters were not able to confidently predict details concerning either convective initiation or mode prior to the outbreak. Surface boundaries were not well defined and a thick cirrus overspread the region through much of the afternoon. A related problem concerned the actual magnitude of the deep layer shear and its effect on storm organization. Uncertainty in the initial analysis at 0000 UTC 3 May 1999 with respect to an upper level PV filament/jet streak emerged as an important factor that influenced the events of 3-4 May 1999 through several effects: deep layer shear, large-scale lift and high cloud production.

With current technologies, it is now possible to produce storm-scale resolution numerical guidance. However, the relative efficacy of this approach relative to other modeling strategies (e.g. lower resolution, ensemble forecasts) has not been established. In this paper, the ability of a quasi-operational storm-scale resolution (2 km grid spacing) numerical weather prediction (NWP) modeling system to provide useful conceptual guidance at 24h forecast range for convective initiation and mode will be studied for the 3-4 May 1999 outbreak. The sensitivities of the forecast to the uncertainties in the initial state are addressed, using PV inversion techniques to modify the initial analysis and factor analysis methodology to establish the relative contributions of relevant physical factors.

With respect to storm-scale forecasts, where the deterministic predictability limit may be reached over much shorter time scales, 24h forecasts may be most useful in providing a tool to aid forecasters in developing a conceptual model of the forecast situation rather than providing explicit predictions of individual storms. The overall philosophy of this approach suggests a forecaster-controlled ensemble strategy, similar in concept to but more comprehensive in scale than the procedure suggested by Brooks et al.

Supplementary URL: