Thursday, May 26, 2011

How Much (Tornado) Warning is Enough?

In recent years, the emphasis in meteorology has been to gain "lead time" for tornado warnings (i.e., issue the warnings sooner relative to the storm's arrival time). The U.S. average now is around 14 minutes. In the devastating Alabama tornadoes on April 27th, the average lead time was 24 minutes. In Joplin, the lead time was 19 minutes.

It is an important question because there will always be a tradeoff between accuracy and lead time, i.e., tornado warnings with one hour of lead time (given today's science) will be inaccurate most of the time whereas 1 minute lead time (see below) will be extremely accurate.

AccuWeather decided to ask, via its Facebook page, how much "lead time" readers wanted. Here are the results:
While this represents an unscientific poll of about 800 people (voting still going on at the link above), it does suggest that 14 minutes is adequate.

I believe that the bigger problem is false alarms and would rather see meteorology focus on that issue rather than extending lead time further.


  1. I'm of the opinion that Storm Based Warnings have improved the specificity and reduced the falsely alarmed area. I remain concerned about the gerrymandering of SBW polygons to conform to county or CWA boundaries, though. What we need are better tools to get the data out to the public, and here's where a mix of technology (new set-top boxes which are aware of their geographical location, NOAA Weather Radio receivers with geospatial awareness (Midland is already working on this), and cellphone system notifications using the data broadcast capabilities of the cellular systems.

    I also know NSSL scientists are working on Warn On Forecast, to provide significantly earlier warnings, or, what in my mind might be, an intermediate level between today's watch, and short-fuse warnings.

  2. If there is a specific cell showing rotation headed my way, the more lead time the better. If you could tell me it will be at my house in 55-60 minutes, that's great. I realize you can't do that very reliably now, but if you could, then tell me.

    "an intermediate level between today's watch, and short-fuse warnings"

    This is the crux of it. There has to be a way to convey levels of risk. Perhaps we could adopt a color-coding system:
      [Red] confirmed tornado on the ground or funnel cloud.
      [Orange] Doppler radar shows rotation of a cell or visual confirmation of rotating wall cloud.
      [Yellow] Specific strong storms with clear potential to develop into Orange or Red. (Roughly equal to Severe Thunderstorm Warning.)
      [Blue] No specific R/O/Y warning has been issued, or those issued have passed, but other specific storms could still develop.
      [Green] No forecast for even the potential to form damaging storms. (Yes, I know that this is out of order based on the color spectrum, but we are conditioned to think of green as "all clear".)

  3. I might also add that the higher you are on that hierarchy, the smaller the geographic extent. A [Blue] watch might cover large chunks of states, spanning from Dallas to Des Moines or Salina to Sioux Falls, while the [Yellow] covers a handful of counties, [Orange] and [Red] might only affect a handful of town(ship)s.

  4. The problem with a color system is....Well it has to hit something or actually be confirmed for a red in the case above. So, what happens if you are in the first mile or so of the path? Then your lead time is cut way down. With the advent of dual radar, the ability to confirm a tornado is on the ground...without spotter or local EM input is going to drastically increase.

    Personally, I think the best thing that has happened in the past 25 some years since the 88 has been in the addition or change to polygon warnings. No longer are full counties issued a warning. Its more of the direct path of the cell.

  5. That should have said....With the advent of Dual Pol radar....sorry!