Convective Weather Forecasting: AutoNowcaster Enhancements
Figure 1. Radar reflectivity at (a) 0230 UTC and (b) 0430 UTC 06 June 2005. Contoured field is the 2-h forecasted vertical shear interest field valid at 0430 UTC 06 June 2005. This field indicates an empirical likelihood of intense convection for values > 0.6 based on the strength and orientation of the environmental shear relative to automatically detected preexisting linear convective features. Negative values indicate the likelihood of the weakening or dissipation of preexisting convective features at the indicated locations with values < -0.6 predicting a high likelihood of such decay.
The research and development efforts of RAL's Convective Weather Group are aimed at improving short-term (0 – 6 hour) thunderstorm forecasting and bridging the gap in skill between observation-driven expert systems and numerical weather prediction. Focus areas include basic studies of the dynamical and microphysical processes occurring at short-time (and small space) scales and their parameterization, advances in observing capabilities, data assimilation and numerical weather prediction, nowcasting (0 – 2 hours) and short-term forecasting (1 – 6 hours) techniques, and verification. One of the most important tools used by the group is the AutoNowcaster (ANC), an automated system developed at RAL that provides short-term forecasts of convective storms based on extrapolation of current radar echoes, model output and mesoscale fields in the path of convection. Efforts to enhance the capabilities of the ANC are on-going and are supported by the National Science Foundation, the FAA's Aviation Weather Research Program, and NOAA's National Weather Service.
FY06 Accomplishments:
Figure 2: An example of an elevated convection initiation likelihood field (filled color contours) at 850 mb at 0500 UTC on 23 September 2005. Superimposed are 30 dBZ radar reflectivity contour in white and 850 mb RUC winds in red arrows at the same time as that of the likelihood field. The warm colors of the likelihood field indicate high probability of elevated CI.
Theoretical and observational studies have illustrated the importance of the intensity and orientation of vertical shear relative to convective features on the subsequent evolution of the convection. Strong vertical shear perpendicular to convective lines is more often associated with strong convection than significantly weaker line-perpendicular shears. Based on that knowledge, an algorithm has been developed that approximates the magnitude of the lower-tropospheric (0 – 3 km AGL) environmental shear relative to preexisting linear convective features that are detected automatically using 5-min radar data. Using a fuzzy logic approach, the interest field output from this algorithm is being combined with other predictor fields to better incorporate the effects of the shear on the likely short-term evolution of convection (Fig. 1).
According to some recent studies, elevated convection is the cause for approximately 50% of convective storms. Unfortunately, the ANC in its present configuration lacks the ability to forecast elevated convection. A new effort is underway to remedy this situation by developing new ANC algorithms for the elevated convection regime. A set of possible 3D predictor fields for elevated Convection Initiation (CI) has been identified and the membership function that converts each predictor field into a likelihood field have been designed and tuned. The weighting of the individual likelihood fields into a combined field is now being refined. An example of the elevated CI likelihood field is shown in Fig. 2. A preliminary evaluation of this approach to elevated convection is encouraging. According to some recent studies, elevated convection is the cause for approximately 50% of convective storms. Unfortunately, the ANC in its present configuration lacks the ability to forecast elevated convection. A new effort is underway to remedy this situation by developing new ANC algorithms for the elevated convection regime. A set of possible 3D predictor fields for elevated Convection Initiation (CI) has been identified and the membership function that converts each predictor field into a likelihood field have been designed and tuned. The weighting of the individual likelihood fields into a combined field is now being refined. An example of the elevated CI likelihood field is shown in Fig. 2. A preliminary evaluation of this approach to elevated convection is encouraging.