Tornadoes with Hurricane Floyd— A WSR-88D Algorithm Study

This study evaluated the performance of the WSR-88D Build 10 Tornado Detection Algorithm (TDA) and Mesocyclone (MESO) algorithms during Hurricane Floyd in September, 1999; and attempted to determine if the adaptable parameters can be adjusted for better algorithm performance during tropical cyclone events. Five tornadoes developed in the outer rain bands in the National Weather Service (NWS) Newport/Morehead City (MHX) county warning area during a nine hour period preceding the landfall of Hurricane Floyd. Of the five tornadoes that developed, the MHX staff issued tornado warnings on four of them.

Using WATADS (WSR-88D Algorithm Testing And Display System, build 10.2), the MESO and TDA algorithms were reprocessed from MHX WSR-88D Archive Level II data for four different sets of adaptable parameters. The algorithms sets included the Operational Support Facility (OSF) default (DEF), OSF tropical (OSF), MHX hurricane operations plan (HOP), and a hypothesized set (HYP), where the parameters were relaxed to account for the shallow (cell tops < 25000 feet) and weak (rotational velocities < 40 m/s) characteristics of the tropical cyclone storm cell features.

The alteration for the MESO algorithm focused on the threshold pattern vector. The minimum 3D feature depth, minimum 3D feature low-level delta velocity, and the minimum TVS delta velocity were altered for the TDA algorithm. The algorithms were scored based on detection of the tornadic rotations; moreover, the probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI) were computed for each set of adaptable parameters.

For the MESO algorithm, the HOP and OSF settings produced a better POD and CSI than the DEF and HYP settings. For the TDA algorithm, the HOP and HYP settings produced a better POD and CSI than the OSF and DEF settings. However, for both algorithms, the FAR remained high (>=0.90); and as the POD improved, the overall number of MESO and TDA detections increased (most of which were false). These results summarize that as the adaptable parameter settings for both algorithms are relaxed from the default settings (to account for weaker tropical cyclone storm cell rotations), the POD increases, resulting in a higher FAR.

Overall performance of the MHX warning staff was superior to the radar algorithms. Additional adaptable parameter studies may be able to improve the algorithms for use during tropical cyclones. However, since humans can manually detect rotating storms much better than the MESO and TDA algorithms, this study concludes by proposing new WSR-88D scan strategies in lieu of trying to improve the algorithms.