Only local National Weather Service Forecast Offices can issue official tornado warnings. That’s not a job for the private sector, nor should it be. The confusion which would ensue with multiple warnings coming from multiple sources could cost lives.
So, how is the NWS doing with this life-saving part of their mission?
Recently, not as well as might be expected.
Lead warning time has been declining in recent years. It’s down from 13-15 minutes between 2005-2011 to about 8 minutes for all tornadoes in the last five years. But the news isn’t all bad. It’s believed part of the reason for this trend is the desire on the part of NWS administrators to reduce the false alarm rate. The NWS reported to the Washington Post the false alarm rate a decade ago was about 80 percent. That is, only one in five tornado warnings verified as actual tornadoes.
All of us in meteorology worry about the “boy who cried wolf syndrome.” The fear with tornado warnings was if that high rate of false alarms continued, people might start ignoring them unless and until they saw the tornado with their own eyes. A subtle pressure in the NWS developed to pull the trigger less often to bring down the false alarm rate. In that, the NWS has succeeded. The false alarm rate is down about 10 percentage points in the last 5 years. However, the detection rate is down as well. That’s troubling. Between 2005-2011, the NWS detected 65-75 percent of all verified tornadoes. The detection rate now averages 58 percent.
One reason for so many false alarms is tied to the huge technological advance of Doppler radar. The radar can track the motion of raindrops and hailstones to detect velocity of winds within a thunderstorm, and whether rotation is occurring. Previous conventional radar showed the reflectivity of the precipitation and the shape of the storm, but could not directly detect rotation inside a storm unless the entire storm was rotating. The majority of tornado warnings are tied to Doppler radar-INDICATED possible tornadoes. Meteorologists infer from the strength and speed of rotation within a thunderstorm and its approximate elevation whether a tornado probably exists. Sometimes it does, but the majority of the time the rotation aloft is not actually producing a tornado on the ground. The radar beam travels in a straight line, while the earth’s horizon curves. The farther from the radar antenna the storm, the higher within that storm the core of the beam strikes. If a storm is 100 miles away, the radar can’t detect whether the rotation is reaching the ground. In some supercell tornadoes, the storm is so strong a ball of debris starts rotating around the vortex closer to the ground. The debris ball can be detected if the tornado is close enough to the radar, and is a very good indicator of an actual large tornado. Weaker tornadoes produce much less of a detectable debris ball, if any.
NWS meteorologists do not issue tornado warnings every time Doppler radar detects rotation aloft; a range of standards, and researched radar algorithms trigger alarms and detected velocities when a tornado is becoming more likely. The paradox is this wonderful technology brings the ability to see this rotation aloft, but it also brings with it the risk of more false alarms.
National Severe Storms Laboratory researcher Harold Brooks told the Post one might assume the lower detection and lead time warning rates were tied mostly to weaker tornadoes, rather than the infrequent monsters spinning from a powerful supercell. But he says the lower detection and lead time warning rate also applies to those huge storms, not just the weak ones. The weak ones are more typical of what few tornadoes we see in Western New York. The weak tornadoes are usually short-lived, harder to see on radar, and may spin up and dissipate between the 6 minute scan time of the NWS’ powerful Doppler radars.
I’m puzzled by the lower detection rate for significant, large tornadoes. Brooks is a leader in this field. However, most other researchers point to a lower incidence of these large tornadoes during the last several years, with more small, brief tornadoes which are harder to detect. It appears the others are not reading the detection numbers on large tornadoes the same way Brooks is.
In any case, these numbers are a national average. Some regions’ local offices have better numbers than others. These numbers are not directly tied to the Tornado Watches issued by the Storm Prediction Center in Norman, Okla., in which the SPC tries to advise local NWS offices and the public how favorable conditions are likely to become for tornado formation over a defined geographic area. Those watches are valuable in providing needed heads-up to meteorologists and the public.
The NWS provides extensive training to their forecasters, and has exercises to test their level of knowledge and ability to interpret complex datasets quickly. The way in which warning decisions are made is being constantly re-evaluated at a NWS center in Norman, and training is frequently updated. It is possible training is part of the problem, but the level of experience also plays a role. I would venture a guess forecasters in the NWS Wichita, Norman and Birmingham offices have better pattern recognition skills than someone in Duluth. You get better skills when you encounter these storms more often.
In the meantime, we can all take some comfort knowing the tornado death rate per million population is down drastically in the decades since the deployment of Doppler radars. Fatalities have gone from 1.8 per million earlier in the 20th century to about .12 per million. Some of that decline is tied to the advent of Doppler in 1991, and some of it is tied to superior communication of warnings to the public by the broadcasting industry and more recent web-relayed warnings.
Yet, an outbreak in the deep south in 2011 which killed hundreds despite far in advance warning is a reminder some tornadoes are simply not survivable when adequate shelter is unavailable. Fortunately in Western New York, we seldom have to face a threat of that magnitude.