I’ve often said that the modern global warming era starts in 1975. Denialists seem to love to accuse me of cherry-picking that time. In fact I’ve received some really nasty comments to that effect, most of which went straight into the trash-bin. The fact is that I didn’t pick 1975 out of thin air, nor is it cherry-picked. It’s an estimate of the time at which the trend in global temperature took its modern value. I’m not interested in answering fabricated, ignorant objections. But sincerely interested readers might want to know where the 1975 comes from.
A glance at the global temperature during the instrumental era indicates that the trend hasn’t been constant over the last 130 years or so:
The smooth line (a lowess smooth) emphasizes the longer-term variations. Clearly there’s a lot of random jitter superimposed on the trend, so no delineation of the trend can be perfect.
But it’s still a pretty good approximation to model global temperature as a piecewise-linear function. That’s just a set of straight lines. Of course, that leaves open the question when to stop one straight line (at one slope) and start a new one (at a new slope) — these are the “change points” of the slope. A long time ago I devised a method to do so. First, choose some approximate change points. Doing so by eye is good enough. Use them to split the data into intervals. In this case, visual inspection indicates there are four “episodes” of different slope, so we’ll need three change point times.
Then fit straight lines over each separate interval, independently. Extrapolate these lines to find when they intersect with the lines of their neighboring intervals. These intersection times define new change point times. Iterate this procedure.
Sometimes the process converges, which defines estimates of the change times unambiguously. Other times, the set of change points goes into an infinite loop. But often, even when it loops the differences are small and we can get good estimates of the change point times by taking the average of the “limit cycle.”
And that’s how I determined that the change point for the rate of global warming is at about 1975. It was years ago I performed this calculation, and we’ve acquired more data since then so it’s worthwhile to repeat the calculation to see whether or not the best-estimate change point times have changed.
Using monthly data from GISS, the average change point time for the start of the most recent slope is 1974.732. By this calculation the “modern global warming era” starts with October 1974. But let’s face it, we can’t really pin it down with that much precision. We should round to the nearest integer year, and accept the fact that we can’t really pin it down that precisely either. It’s just an estimate, which turns out to be — no surprise — 1975.
We can do the same calculation using annual averages rather than monthly data. This gives a slightly different result; using annual data the most recent episode (the modern global warming era) starts in 1973. This only serves to emphasize that we can’t nail down the change point time to the nearest year with certainty. But we can certainly approximate it, and both 1973 and 1975 are about as good estimates as I’d believe.
Just because we can’t be certain when the new slopes begin to the nearest year, doesn’t mean there’s any uncertainty at all that the slopes in different episodes really are different. We can take the episodes as defined by the annual data and estimate the slope, and its uncertainty, over each. The uncertainties are computed using a white-noise approximation, but for annual rather than monthly data that’s not a bad estimate. And here they are:
There’s absolutely no doubt — none at all — that the slopes in different intervals are different. And the slope in the modern global warming era (whether you choose to start it at 1973 or 1975) is the fastest.
I often say that the modern global warming era starts in 1975. I didn’t pull that number out of a hat, that’s the year which was chosen by the data itself. If you want to dispute that it’s not accurate to the nearest year, fine — I’ll agree with that. If you want to accuse me of cherry-picking it, then you’re another one of those who doesn’t know when a choice is justified and when it isn’t.