Again With the David Rose Hole

David Rose has another feptic article in the Daily Mail, with false claims about global warming not meeting projects, etc. Phil Plait does a thorough debunking here. One thing Rose wrote is

The graph confirms there has been no statistically significant increase in the world’s average temperature since January 1997 – as this newspaper first disclosed last year. 

This is misleading in two ways.

If you take the HadCRUT4 global data, the ordinary least squares linear regression trend since January 1997 is 0.045 ± 0.036 C/decade, where the uncertainty is the OLS uncertainty representing the 5-95% confidence level.

That's positive (i.e warming) and statistically significant.

This wasn't true just few months ago, when I wrote about the David Rose hole.

It's now true.

If you include autocorrelation (as you ought to), the trend (from the handy SkS trend calculator) is 0.046 ± 0.124 C/decade, where the uncertainty is 2σ.

That's warming, but with a statistical significance of only 54%. Canonical climate science usually considers 95% to be the standard, though if you're worrying about the future of civilization you might relax that a little.

The large uncertainty that accompanies the autocorrelated result says that your time interval is too short to make statistically meaningful conclusions.

A short time interval is always going to have large trend uncertainties like this -- because, you know, the climate has noise in it -- and unless your warming is always greater than about 0.124 C/decade, or your cooling is always less than about -0.124 C/decade, you're usually not going to be able to make statistically significant conclusions about the trend. Period.

Here is some history of the 15-yr trend for the Hadley data:

It varies a great deal. That makes it not very useful. What would feptics have said in 2007, when it was 0.29 C/decade? I suspect they would have said, yabbit it's abnormally high because you had an El Nino in 2002 and 2004 and 2006 a big one in 1998, and those are skewing the results upward.

Well, now the 15-yr trend is abnormally low, because there was a large warming El Nino on the back end of the interval, and a big La Nina on the front end, and this is skewing the results downward.

The 15-year trend is for suckers. It measures weather, not climate, except the weather is in the fluid ocean and not just the fluid atmosphere.

This is all going to change in 6-12 months, when the 1997-98 El Nino falls out the backend of the 15-year window. In fact, looking at the chart above you see this already starting.

Like I said, the 15-year trend is for suckers. [And of course it's a lousy place to look for an energy imbalance anyway.]