Environmental and signal processing conditions that negatively impact the performance of the WSR-88D Tornado Detection Algorithm

Within the Weather Surveillance Radar 1988 Doppler (WSR-88D) processing system is a suite of meteorological algorithms that includes a Tornado Detection Algorithm (TDA). During major tornado outbreaks the TDA performance has been shown to be quite good. For instance, during an outbreak in Central Oklahoma, USA, May 3-4, 1999 the Probability of Detection (POD) was 0.85 with a False Alarm Ratio (FAR) of only 0.13. Overall, the TDA has a POD of 0.43 and a FAR of 0.48.

The TDA false alarms are classified into two categories. The first category we call Type I false alarms. Type I false alarms are detections of false rotation signatures resulting from non-meteorological artifacts in the radar radial velocity data, and the TDA incorrectly diagnoses these detections as being tornadic (i.e., a false detection). The conditions that lead to Type I false alarms may be related to signal processing errors including receiver automatic gain control saturation, side lobe contamination due to insufficient clutter filtering, and anomalous propagation resulting in detection of shear due to the juxtaposition of stationary and moving ground clutter targets (e.g., airplanes, trains, automobiles). The second category we call Type II which is defined as an algorithm misdiagnosis of a non-tornadic signature. Type II “false alarms” are those detections of bonafide gate-to-gate and vertically correlated shear which are not associated with an observation of a tornado, and the TDA incorrectly diagnoses the detections as tornadic (i.e., false diagnosis). It is more appropriate to refer to Type II false alarms as misdiagnosed detections.

The purpose of this paper is to aid operational users to recognize Type I false alarms by showing specific examples and to suggest ways to mitigate them. This paper also discusses the impact of a recent change to the Velocity Dealiasing Algorithm that can contribute to the identification of Type I false alarms.