On the heels of yesterday's post, it looks like another active severe weather day in the central Plains.  The National Weather Service has issued a Particularly Dangerous Situation (PDS) declaration as part of Tornado Watch #356.  Northern Texas and central Oklahoma are most likely going to see an outbreak of violent storms:  the probability of at least two tornadoes forming within the watch area before 10 p.m. currently stands at 90%, and the probability of at least one strong tornado (EF2-EF5) at 70%.  Forecasts predicting the April outbreak listed similarly probabilities, and in both cases these chances are listed at higher probabilities than I can recall seeing prior to this year.  (Perhaps 2008, another uncommonly active year, saw a couple of days with higher than 60% probabilities as well.)

I was leafing through my Google Reader subscriptions this morning and found an interesting post by Andrew Revkin, who writes the Dot Earth blog for the New York Times website, in which he mentions an issue about false alarms and their relation to tornado deaths.  Areas that see a greater proportion of false positives tend to experience a higher number of fatalities when a tornado actually does hit.  Seemingly, the boy who cried wolf can also cry tornado.

You can read the abstract Revkin cites from the American Meteorological Society, but the full article requires a subscription.  Since I can't read it myself, I'm curious to know what constitutes a false positive:  for instance, is a warning that is issued when radar indicates strong rotation in a storm considered a false positive if that storm does not eventually produce a tornado?  It would seem to me that the precautionary principle should apply here, because supercells can produce tornadoes very quickly and with little warning.  Strong rotation in a storm (usually signified by a supercell's hook echo) probably should justify issuance of an alert.  However, I'm speculating a bit as to what parameters the authors used in identifying false alarms.

Here's an interesting bit from the abstract, though:

The casualty effects of false alarms and warning lead times are approximately equal in magnitude, suggesting that the National Weather Service could not reduce casualties by trading off a higher probability of detection for a higher false-alarm ratio, or vice versa.

So it sounds like there's a wash.  You make up with lead time what you lose due to the perception of false positives.

Tornado and storm forecasting will continue to improve, and ideally, we'll one day be able to rely on the accuracy of a warning without reservation.  But the detection and warning system in place today is still pretty damn good, and we need to know how to properly assess risk and avoid becoming jaded at the many warnings we see.  (Storm forecasting is a tough business, after all.) Rule of thumb: if a tornado warning is issued, don't take a chance on it.  Get somewhere safe.  You don't have to do what I do and hunker down in basement for six hours until the last drop of rain has fallen out of the sky, but being too casual about storm warnings can put you in a tight spot—maybe even cost you your life.  Minimize your risk, even though the odds are usually in your favor.

Because forecasting is not an exact science, let's hope the meteorologists goofed on this PDS for today.  I hope Oklahoma and Texas don't experience  the outbreak as predicted by the NWS.  But chances are they will, and the National Weather Service's watch and warning products will be the best tools the affected communities will have in order to stay as safe as possible.