Great thread and questions here. While there's a lot of talk — and rightfully so — about the merits of probabilistic tornado forecasting and its role in our warning system — I think the "extra lead time" question may be even more interesting.
If/when we're able to extend tornado warning lead time out to an hour or so (as suggested by Stensrud et al. 2009), we'd be pushing beyond our traditional concept of "short fuse" warnings. Given an average of 13 minutes' lead time for tornado warnings, that doesn't leave much time to do much more than what we know people do: hear the warning, confirm the threat, decide if/how to take shelter, and act. Further, the smaller number limits your options. Unless you suffer from limited mobility, are at some large gathering place, or are out in the middle of nowhere for some reason, that limit on your options may not be such a bad thing.
As Greg says, stretching that 13 minutes out to 60 poses some
really interesting social science questions. Will people use that extra 47 minutes to further evaluate their threat? Will they decide upon hearing the warning that they want to take shelter when the time comes but then forget or get sidetracked? Perhaps most concerning, will they opt to move to "best shelter" (e.g. from their mobile home to grandma's house a few miles away) only to get caught in the storm?
The problem is that we just don't have the answers to these questions. We can't even look to other similar life-or-death situations on this kind of time horizon to make comparisons because there aren't many. The closest I can think of would be tsunami warnings, but those are somewhat new, and even they contain a certain urgency that goes with "get as high as you can" combined with a potentially much larger number of people affected and trying to get to the same places. Going beyond that, we get into longer lead-time events (e.g. winter storms, hurricanes) where our track record for getting people to take protective action and/or evacuate is not what we would want it to be.
Greg's point about using the new science to specify event onset and conclusion envelopes for various points is interesting, but I don't know how useful that will be for most people. TV stations would probably still break in as soon as the warning is issued and track the storm just as they do now, and the onset/conclusion times wouldn't be dramatically different than what is already done with path casts in the NWS warnings or from radar software. The biggest benefit for those envelope calculations would be for location-aware mobile devices and for educated users or gatekeeper-types who know how long they need and a confidence interval at which they would act.
Great discussion!
