Polarimetric tornado detection

by Alexander Ryzhkov

composite radar plot of May 8 tornadic storm

Composite plot of Z, ZDR, and _hv at El=1.5o fora tornadic storm on 8 May 2003 (2229 UTC). Location of tornado reported on the ground is shown by red arrow.

"Is it possible to improve tornado detection using a polarimetric radar?" a television crew asked me back in 2001. My answer was, "Very likely, because tornadic debris is supposed to have polarimetric properties that are very different from those of hydrometeors." The follow-up question was: "Did you collect polarimetric data for the May 3, 1999 tornado?" Unfortunately, we had not. The research Cimarron polarimetric radar was lost due to a lightning strike about half an hour before the tornadic storm hit the Oklahoma City metropolitan area. The interview prompted me to revisit the Cimarron dataset for that day, and I discovered data collected during a less-destructive F3 tornado west of Chickasha prior to the radar shutdown. I found exactly what I was looking for: a well-pronounced tornado signature in the polarimetric data. Two polarization variables, differential reflectivity ZDR and cross-correlation coefficient_hv clearly indicated tornado debris in the hook echo close to the ground at the time and space where the F3 tornado was detected according to ground observations.

ZDR is a ratio of radar reflectivities at horizontal and vertical polarizations expressed in dB. Meteorological scatterers are usually characterized by positive ZDR (0.5 – 4 dB for rain) because their falling habit is such that their horizontal dimension is larger than vertical. In contrast, tornadic debris are randomly oriented, and their ZDR is very close to 0 dB. Another polarimetric variable, _hv, has even better discriminating power than ZDR. The crosscorrelation coefficient is very close to 1 for pure rain or pure snow, somewhat lower for a mixture of hydrometeors of different phases (e.g, rain / hail mixture, melting layer), and substantially smaller for non-meteorological scatterers like insects, birds, chaff, ground clutter, etc. This coefficient is anomalously low for tornadic debris because of their large sizes and very irregular shapes. Thus, anomalously-low values of _hv combined with ZDR close to zero observed in the hook echo very likely indicate the presence of a tornado.

Our research group was given the opportunity to confirm our previous findings with the polarimetrically upgraded KOUN WSR-88D radar on May 8-9, 2003, when two tornadic storms hit the Oklahoma City metropolitan area. We observed polarimetric tornado signatures for both events. The figure to the left shows a combined plot of radar reflectivity, differential reflectivity, and cross-correlation coefficient observed with the KOUN WSR-88D radar at an elevation of 1.5o on May 8, 2003, at 2229 UTC (1729 CT) when the F3 tornado was reported east of Interstate 35 in Oklahoma City. The tornado touchdown signature at the very tip of the hook is marked with ZDR close to 0 dB and _hv < 0.5 (red arrow). We identified a very similar signature for the tornadic storm on May 9, 2003.

This experimental evidence gives us much more confidence in the unique capability of the polarimetric WSR-88D radar to detect a tornado when it touches the ground and creates lofted debris. Polarimetric signatures can be detected with coarser resolution than Doppler radar. Also, signatures in the Doppler velocities are dependent on the viewing angle, while those detected with the polarimetric radar do not vary with direction. Heavy debris as well as large hailstones are not perfect tracers of air motions. Polarimetric identification of those scatterers will help with quality control and better interpretation of Doppler measurements.

Although tornado detection is important, prediction and early warning are even more critical. A cursory look into evolution of the 3D pattern of polarimetric variables prior to tornado touchdown reveals quite unusual and intriguing polarimetric signatures aloft that might be related to the subsequent tornado. Understanding and interpretation of these signatures could provide more insight into microphysical aspects of tornadogenesis.


Previous | Briefings Home | NSSL Home