Wednesday, January 30, 2008

Chart of the Day

Climate "TOOTSIFs" insist supposed global warming will produce temperature extremes, especially more frequent record highs. Do the data support the claim? Coyote Blog's Warren Meyer explains (at 58-60):
That seems logical enough, but Bruce Hall actually compiled the data and found something interesting. He created a data base for each state which shows in what year that state's monthly temperature records were set. So for each state, he has the years when the twelve monthly high temperature records were set. . .

[E]ven I, a skeptic, would expect a disproportionate number of the all-time high temperatures to be in the last decade. . . The global warming folks would argue that the effect should be doubly pronounced, since they claim that we are seeing not just a general heating, but an increase in volatility (ie more extreme variation around the mean). But Hall doesn't find this when he graphs the data. Take the 600 state monthly high temperature records that exist on the books today (50 states times 12 months) and graph the distribution of years in which these records were set:

source: Hall of Record blog

Assuming about 120 years of data, you should expect to see a high temperature record on average in a database of 600 records at 5 per year, which is precisely where we have been of late and well below the record years in the thirties (remember the dust bowl?) and the fifties. It seems to actually show a reduction in temperatures or volatility or both.


OBloodyHell said...

I'd also like to see information on when the record was "matched" included, but in general, another nail in the GW coffin.

OTOH, I do note the high similarity between the pre-40s running average and the current one. I'm curious if this isn't a (gasp!) long-term running cycle of (gasp!) around 70 years (assuming 1894-1974), which I seem to recall is close to what has been suggested before.


Carl said...


I asked; here's Bruce Hall's answer:

"The same way NOAA handles the records, the tie always goes to the later year. So if 1998 tied a record set in 1934, the count would be taken away from 1934 and given to 1998. That way there is always a bias toward the latest year and a bias toward an increasing trend line.

Just trying to be consistent with our government...."

So his methodology is conservative and thus quite defensible.