Technical writing rule #303: never, ever use that exclamation thing. Yet there are times when absurdity demands an ugly response. The meme that claims a “pause” in global warming has become so pervasive that even experienced scientists — who ought to know better — have been spouting it¹.
I’m going to show here that recent data suggests an acceleration in warming, not any kind of pause!
In a previous post I showed visually (as others have show statistically) that there really isn’t an observable pause. The chart there plotted monthly global temperature estimates from seven different sources, to avoid any argument about source veracity. To begin, lets simplify that by choosing just one of those, that from the Cowtan & Way (version 2), which is arguably the best of the current crop:
You can see that the post-1970 trend is nearly identical to that previously calculated for the mean of the seven estimates. Now, because I’m going to look at warming “this century” (since the beginning of 20002), let’s refine the reference trend to cover just the 30-year interval from 1970 to the beginning of 2000:
The shorter trend is almost the same as the near half-century long trend. Now let’s fit a “trend” to just the 14 complete years3 of the 21st century (since the beginning of 2000):
Well I’ll be darned, the famous pause! So what’s wrong here? Fact is, we just fitted a complex compound function comprising two straight lines separated by a step change, with no fundamental reason whatever for choosing to do that. That fit has a total of six parameters: the start date (1970, which is not the beginning of the data), the first trend intercept at that date, the first trend slope, the step change date, the second trend intercept at that date, and the second trend slope. Complex, multi-parameter fits are rarely justifiable. The father of modern computing, John von Neuman is reputed to have said, “With four parameters I can fit an elephant, and with five I can make him wiggle his trunk.” We have six!
How, properly, to look at this data? In my view a simple, largely non-parametric approach is appropriate. Accept the clear long term trend and test the recent data against it as part of that process, rather than as a new, completely different process (which is what we do when we try to “free-fit” a new trend to a little bit of data). Looking at the data since the start of 2000, we get this:
The residuals to the trend line4 (difference between data and trend) since 2000 skew above trend! The mean and median of the residuals are both positive. That means that, since 2000, global temperatures have been predominantly warmer than trend. By any reasonable definition, warming has accelerated.
Look, let’s not overstate this (as deniers are want to…). The positive bias is small, and unlikely to be statistically significant. But it is positive, meaning that, viewed this way, there is more evidence for acceleration than for a “pause”.
1. What England et al and Kosaka and Zie have elegantly shown is that the broad structure of recent multi-year variations of global temperature about trend is well explained by variations in Pacific Ocean circulation, and related effects. Essentially, they propose a compelling explanation (slightly different in each case) for one element of the large-scale noise in the global temperature record. That is all good and useful, but they err, in my view, in conflating that with a so-called “pause” which is not demonstrated. Others publishing on the topic have been careful to avoid that.
2. “No warming this century” is a popular version of the meme (“…more or less steady since 2001″ as England et al put it). Here I define the century as beginning at 12 midnight on 31 December 1999, as most ordinary folk do.
3. This interval is too short to define a meaningful trend, but that is what the meme requires.
4. That is to the full trend, not the 1970-2000 trend, which would show even greater positive 21st century bias.
5. I promise to stop using exclamation marks.