Radio has a way of making voices sound subtly different, probably because it chops off some frequencies, so I had a hard time placing the interviewee's voice on radio today. It was awfully familiar, and the topic was storm warnings, so it made sense that I would know him. The wrap-up cleared it up: it was Don Burgess, talking about the 30-minute lead time that the Greensburg, Kansas tornado warning sported [Listen to the NPR audio here]
The lead time statistic (increasing from an average of 2 minutes in the 1970s to about 13 minutes now) is flawed. The way the National Weather Service computes it, it assumes that tornadoes that are not warned have a lead time of 0 minutes. This sort of makes sense on a tornado-by-tornado basis, but computing an average where some tornadoes have an actual lead time and others have an presumed 0-minute lead time causes anomalies that make the lead time statistic hard to interpret. Most of the "increase" in lead time is really because of an increase in the likelihood that a tornado is warned -- that has increased from about 25% range in the mid-1980s to about 75% now (due primarily, as the NPR piece notes, to Doppler radar research at NSSL). The average lead time of warned tornadoes, however, has remained relatively constant, at around 18 minutes.
Knowing this background, I found Don's performance impressive. He answered the interviewer's questions about the 30-minute lead time but took care to explain it in terms of actually being able to predict the tornado. And in the end part of the segment, he put lead time in context. The amazing thing is that without my commentary on the lead time (unless of course you are knowledgeable about tornado warning statistics), you would be hard-put to see that Don was walking a fine line between keeping it simple (while giving the reporter the story she wanted) and remaining on the side of the angels as far as the science was concerned.