Saturday, April 14, 2012

I don't know and neither does anyone else!

Today, with a "High Risk" of a major tornado outbreak in the forecast, some recent experiences prompt me to discuss a tornado-related topic that seems to be springing up everywhere, sort of like playing a game of Whac a mole.  The question is ... wait for it ... Will global climate change cause an increase in the frequency and/or intensity of tornadoes?

I recently had an extended email discussion  on this topic with a famous climate scientist after I posted a blog on the subject.  That scientist, not a specialist in severe convective weather, seems reluctant to accept the arguments of a fellow scientist who has spent 40 years studying tornadoes and tornado climatology, or to do anything to correct "misstatements of his position" on the subject in the popular media.  Whatever - I have no means to control the words/actions of others, nor do I wish to have such control.

The more time I've spent looking at our data regarding the occurrence of tornadoes, the more concerned I've become about the relationship between those data and the unknown true time-space distribution of tornadoes in the USA.  I've said on many occasions that the USA has the best data regarding tornado occurrence in the world, but our data still pretty much suck!  They're dominated by the heavy hand of non-meteorological influences:  population density, challenges to rating tornado intensity, reporting practices, politics of various sorts, and so on.  The effects of these "secular" factors only become evident as one works with the data.

It's indeed logically possible that some of the changes we see over time in the space-time distribution of tornado reports are the result of global climate change.  I certainly have no basis for ruling that out.  Unfortunately, I can't rule out the possibility that global climate change has, as yet, had no effect, either!  Given the very strong signal of a increasing tornado frequency, it seems inevitable that people examining the data for the first time would be impressed most by that very obvious trend in the data. Even a relatively sophisticated newcomer capable of using tools like statistical analysis on the data couldn't avoid being dominated by that trend.   The problem is that deeper examination of the data eventually reveals large biases that clearly have nothing to do with meteorology.

Our tornado data simply do not permit the drawing of conclusions with high confidence about possible changes in the space-time distributions of tornadoes.  No simple adjustments of the data - say, accounting for population bias - can remove the multitude of secular changes that resist being characterized easily in time and space.  Population bias is only one amongst the multitude and it's relatively easy to accommodate, whereas many of the other factors are much more difficult to account for, or even to characterize.  Given the potential significance of the effects arising from anthropogenic global warming, it's easy to understand why someone would jump on a superficial analysis of the tornado occurrence data to make unjustifiable, speculative statements about the future for tornadoes.  But the simple fact is I don't know what's going to happen and neither does anyone else!  Anyone who asserts that they do know what's going to happen with respect to future tornado occurrences has no scientific foundation upon which to stand. 

I'm not happy about this situation, but scientific integrity demands that I acknowledge it and behave accordingly.  I see little or no potential for the development of an extensive data base about tornadoes free of these secular influences any time in the next several decades.  I know of no way to obtain highly accurate data that would reveal the true space-time distribution of tornadoes.  Even if some such solution did exist, we would have to use it for many decades to develop a sufficient sample of events to justify making confident statements about temporal changes in the space-time tornado distribution.

Therefore, this problem isn't going away any time soon, barring some sort of unforeseen new developments in the climatological analysis of tornado occurrence data.  Absent highly reliable information about that true distribution, any hypothesis based on the historical record of tornadoes regarding the future of tornado occurrences as the global climate changes is going to remain speculative - which is to say, outside of the domain of good science.