Links to current knowledge of vertical association (VA) failures within NSSL WDDS algorithms is provided below. VA failures within SCIT, MDA, and TDA have been examined in three different case studies. Pat Burke has examined the May 3, 1999 tornado outbreak and found several VA failures for both the MDA and TDA algorithms. Travis Smith has examined modifications that can be made to SCIT to improve the DDPDA's detection algorithm. The same modification may help SCIT's VA of cells as well. Aurthur Witt has analyzed a Florida hailstorm and noted some MDA VA failures as well. The May 3rd case involves high shear/fast moving storms while Travis's study is of weak shear/pulse-type storms.
A detailed examination of the May 3, 1999 tornado outbreak has yielded several different VA failures. Links have been provided to Pat Burke's investigation of this case study.
May 3, 1999 Case:
Travis found that modifications were needed for the SCIT Algorithm in order to increase the probability of prediction within the Damaging Downburst Prediction and Detection Algorithm (DDPDA). His study examines pulse-type storms from the Damaging Wind Events Database (DWED) for "wet downbursts". SCIT currently identifies 7 reflectivity thresholds (every 5 dBZ from 30 dBZ to 60 dBZ) to identify two-dimensional cell components. In SCIT's VA scheme, cell's are linked in the vertical by examining common features separated by one elevation tilt and within 5 km quasi-horizontally. If no common component is found, the routine steps outward in 2.5 km increments and continues the search until a component, which hasn't already been assigned to a cell, is found or the search is completed for a 10 km radius from the current component's center. To maximize DDPDA's detection ability, SCIT's reflectivity thresholds are increased to 15 (every 2 dBZ from 26 dBZ to 54 dBZ). And the VA scheme is modified so that cell components are first examined within .75 km (instead of 5 km) at the next elevation angle. If no common cell components are found, the scheme steps outward at .75 km increments (instead of 2.5 km) with an outer search limit of 7.5 km (instead of 10 km).
Aurthur has examined the VA analysis of the MDA algorithm for a Florida hailstorm. His analysis of MDA's VA algorithm for this storm is: "3D association errors were fairly rare (overall, they occurred only 5% of the time). However, those that did occur tended to happen with the stronger and deeper signatures. The most significant failures were the result of a midaltitude "split" of a deep (whole depth of the storm) circulation. These splits occur when two 2D features are detected in close horizontal proximity in the middle part of the storm. The lower half of the circulation is 3D associated with one of these 2D features, while the upper half is 3D associated with the other 2D feature. Thus, instead of getting a single, deep 3D detection, one gets two overlapping 3D detections (lower and upper halves). One possible way to reduce this problem is to have a minimum distance threshold for 2D detections. If two 2D detections are "too close," then the weaker one is deleted."