Of late, it's become fashionable to discuss the issue of the efficacy of warnings in terms of the need to do social science-oriented research. In fact, I've just returned from the Weather Ready Nation Workshop in Birmingham, AL where the main goal was to define research objectives for a multidisciplinary approach to the challenge posed by large death tolls in spite of reasonably timely and accurate warnings. Hopefully, the outcome of this workshop will be a major step forward in getting the needed multidisciplinary research done.
For decades, the NWS has been focused, naturally enough, on the weather! During most of my career, most forecasters operated under the notion that they have a forecasting task to accomplish that basically ends when they push the "send" button on their computers and their forecast products (including tornado warnings) leave the office (mostly electronically). They've not considered it their responsibility to understand what happens when it leaves the NWS office - and it shouldn't be their job. Weather forecasting is tough enough! My long-time friend and colleague, retired NWS operational forecaster Alan R. Moller made me aware of the inadequacy of this situation: even a perfect forecast is useless if the forecast information:
- Fails to reach its intended users
- Is not understood by the intended users
- Is not believed by the intended users
- Is of no value in helping intended users to make critical decisions
I've long been an advocate on behalf of this research and I continue to believe we need those answers before we start meddling with the warning process. No doubt in reaction to all the angst, the NWS recently initiated a public "experiment" in tiered warnings. This is an ill-advised, ill-designed experiment. It "recycles" some old terminology and represents nothing particularly new. Its experimental design isn't founded on new understanding of how the public receives, understands, and uses warnings to make decisions. Moreover, it shouldn't have been carried out in a "public" mode. The ability to issue skillful tiered warnings should have been tested in a non-public, "testbed" mode before this project was undertaken - the HWT at the Norman Weather Center would have been an ideal venue for a series of internal experiments establishing the extent of forecast skill before going public. And the physical and social scientists should have been given the opportunity to review this experiment and offer recommendations before the ill-advised decision to go ahead with this was ever made.
Having said all this, I now want to throw some cold water onto the entropy generated by all this anxiety over the warnings. First, in my opinion, there's no magic bullet that will transform the existing warnings into something where all the challenges described above are overcome.
Is it our goal to turn the public into automatons who shuffle off like zombies to shelter every time we issue a warning? Personally, I don't think so. We have no business telling our forecast users what to do. The meteorologists need to issue their warnings, but the users must make their decisions about what to do with our weather information on the basis of their own complex needs, which vary from one user to the next. This diverse set of user needs can't be accounted for in NWS warnings and so I'm opposed to such things as call-to-action statements. If people need to know what to do in specific situations, the time to learn that was before the threat ever materialized, not in the last few minutes before a tornado strikes.
Although the existing warning system is far from perfect, its impact on life-saving over the years has been dramatic. I have no doubt that thousands of lives have been saved by the existing imperfect warning system. Before we start tinkering with the warnings, we need to be as certain as possible that any proposed changes won't result in even less effective warnings than they are now! First of all, do no harm!!
Second, it's obvious to me that mere wording changes aren't likely to have any lasting impact on the warnings. We have to reconsider the whole process, end-to-end (to use a phrase that once was popular in NWS management circles), and make changes only on the basis of clear and compelling new findings about the warning process. Wordsmithing is of little or no value.
Third, the existing warnings do their best with respect to the tornadoes that are most likely to take lives: long-track violent tornadoes. We're already doing pretty well there and the rest of the tornadoes only account for a small fraction of the fatalities. When large, violent, long-track tornadoes hit populated areas, people will be killed. Most of the time, tornadoes miss population concentrations here in the USA. A comprehensive explanation for the massive death tolls in 2011 has yet to be offered, and likely involves some complex interactions among factors leading to casualties, but it's my belief that the primary "explanation" is that our luck ran out in 2011. The threat has been there all along, but we were mostly unlucky in 2011, whereas we've mostly been lucky in the preceding decades. This has been characteristic of tornado fatalities over time - sporadic disasters decades apart, with much lesser events dominating in between. We need to review and consider who is being killed in deadly tornadoes and the reasons for their deaths. I don't believe the "answer" to preventing fatalities is longer warning lead times by tornado forecasters. As noted, we're already doing a pretty good job with most of the really deadly tornadoes, paradoxically.