An important radar technology introduced to radars used by forecasters in the U.S. in 2013 is called dual-polarization. Scientists are able to use this technology to infer additional information about what a radar is "seeing", like whether it's heavy rain or hail, that previously was difficult. At the same time, a vexing problem in the severe storms community is that the most severe storms (called "supercells") that produce tornadoes look very similar on radar to the supercells that do NOT produce tornadoes. As a result, a forecaster sees a supercell on radar and still cannot precisely determine if it is about to produce a tornado; if they did know, they could issue a tornado warning before the tornado formed, better informing the public of the danger of the storm. The goal of this project is to look at a large number of supercells on radar and compare the dual-polarization data of the storms that produced tornadoes with the storms that did not produce tornadoes so that we may determine if there is any radar information that is directly connected to tornadoes. A key aspect of this study is that it is based on climatologies instead of looking at just 1 or 2 or 5 cases; this allows us to better generalize the relevant signals because strange things can happen in a single case.

To do this, we used the radar data archived over a 5+ year period and had 100-200 cases that we investigated to come to conclusions. We used well-vetted and accepted techniques to make sure the data were of high quality. Then, we set out to determine what these data can tell us about supercells and tornadoes. Here is what we found:

• That standard radar techniques (NOT using the dual-polarization data) by themselves can be used to predict that a tornado is about to dissipate (die off) within 10 minutes. Forecasters can simply look at simplified ways to estimate the strength of the tornado and how it is moving via radar data to do so. 
• Previous studies of a few cases showed that using dual-polarization radar to estimate the sizes of raindrops may be used to predict tornado formation and dissipation. However, when looking at 130 cases, a much larger sample, we found that, unfortunately, that was not the case. However, it is still valuable for researchers to know this is a dead-end and concentrate their efforts elsewhere.
• Some dual-polarization radar variables were also found to be related to tornado dissipation. Combined with results summarized in the first bullet point, we found that when there were multiple radar "indicators" of tornado dissipation, this almost always signaled that tornado dissipation was imminent. We provided guidelines for forecasters to implement this in the future when it may be easier to automate these radar indicators. 
• Our most significant finding is that one dual-polarization feature, called the ZDR column, may be used to predict tornado formation and/or indicate that, if a tornado were to form, it would be more or less likely to be a very intense tornado. The feature serves as a proxy for the area of rising air in a storm called the updraft. We found that the size of the ZDR column, which is related to the size of updraft, is greater in storms that went on to produce stronger tornadoes compared to storms that produced either no tornado or weaker tornadoes. We provided guidelines to forecasters as to how they might use this information right now in their forecasting of severe weather. 
• When radars are run to monitor the weather around the U.S., forecasters, as of ~2015, have the ability to have the radar scan the lower part of the atmosphere above the ground more often; this may be important because close to the ground is where tornadoes and other hazards are impacting people. However, in the past 5-6 years, it had not yet been investigated whether tornadoes, large hail, and strong winds were being better forecasted when more of the near-ground scans were used. We used 6 years of data to investigate. We found that the extra low-level scans were indeed leading to more accurate severe thunderstorm and tornado warnings by forecasters compared to when they were not being used. This result provides confidence to forecasters that they should continue to use this radar data collection strategy.

Last Modified: 07/29/2023
Modified by: Michael French