In defense of David Rose, he insists that “Despite Denial, Global Temperatures Are Dropping Fast.” He shows several data sets other than the one used by David Rose, to establish that “dropping fast” isn’t just about Rose’s choice, and the very first he displays is from NASA:
Does that look like it’s “dropping fast” to you? Or does it look like it’s fluctuating a lot while the overall trend is going up? What if we showed you some of what it did before 1997?
Does that look like it’s “dropping fast” to you? Or does it look like it’s fluctuating a lot while the overall trend is going up?
David Whitehouse doesn’t seem to like my mentioning that there’s data before 1997, when David Rose started his graph (and Whitehouse starts his).
The blogger shows the lower tropospheric data back to the start of the data set in 1979 and says showing the post-1997 “hiatus” data on its own is misleading as there is a clear trend from 1979 upwards. Except that there isn’t.
There is no way to reproduce the trend observed either during the period 1979 – 1997 or 1979 – 2016 in the post 1997-data (ie half the data set) because a straight line does not represent the data over its entirety. It is obvious that a straight line doesn’t work when one examines the residuals (the difference between observed and predicted data) which are not randomly distributed.
Let’s find out, shall we?
We’ll start with the NASA data he so kindly showed. Let’s see how the 1979-1997 trend estimate compares to the post-1997 trend estimate:
Golly gee. From that, it looks like from 1979-1997 to post-1997 global warming actually got faster!
But did it really? I’ve often emphasized that “looks like” isn’t proper statistical procedure. After all, those trend estimates are only estimates; we should actually compute the uncertainty associated with them. Of course that involves using some statistics, which may not be David Whitehouse’s forte.
Here are the estimated trends for the two time spans, together with their 95% confidence intervals (the range in which the actual trend is likely to be):
The fact that the confidence intervals overlap so much is telling evidence that statistically speaking, there’s no real evidence that the trend rate is any different after 1997 than it was before.
How do those trend estimates, for the two time spans, compare to the trend estimate for the entire span from 1979 to the present? Let’s add the after-1979 trend, as a dashed line, to see for ourselves:
Both of the confidence intervals we computed before include the estimated rate for the entire time span 1979-present. To put it succinctly, statistically speaking there’s no real evidence that the trend rate 1979-1997 or post-1997 is any different than it was for the whole span post-1979. The statistically best estimate is just this:
That of course is for the NASA data, which is a whole-earth (not land-only) data set but based on meteorological stations only, not using sea-surface temperatures. What about David Rose’s original choice, the lower-atmosphere data over land areas only from RSS? Comparing pre- and post-1997 trend estimates we get this:
For these data it does “look like” the trend rate went down. But as I said before (and have often emphasized), “looks like” is an invitation to error. Let’s once again compute uncertainty levels and compare confidence intervals:
For this data set, as well as for David Whitehouse’s chosen NASA data, there’s no real evidence that the trend rate is any different after 1997 than it was before. Spoiler alert: there’s also no real evidence that the trend rate 1979-1997 or post-1997 is any different than it was for the whole span post-1979.
But David Whitehouse insists that the whole-span linear trend can’t be right:
It’s obvious, he says, because the residuals are not randomly distributed.
With that, David Whitehouse decided to get mathematical. At this point I apologize to first-time readers who don’t like math, because I’m about to get mathematical myself.
This is where David Whitehouse reveals that statistics is definitely not his forte, because he really doesn’t know what he’s talking about. To the untrained eye, the residuals don’t look random:
But those who know what they’re talking about, know that not all random noise looks the same. In particular, we’ve known for decades that global temperature time series have a kind of randomness that isn’t the simplest kind (what’s called “white noise”), instead it’s called autocorrelated noise. In fact the autocorrelation can be quite strong, especially for satellite data for the lower atmosphere over land areas only.
I generated random noise with the same autocorrelation structure, to show you what it looks like. I didn’t generate hundreds and hundreds of samples until I got one that looked the way I wanted. I was going to make 10 of them, but the very first one — right out of the box — illustrates the point so well that I’ll show you that one:
That, my friends, is what autocorrelated noise looks like when it emulates satellite data for lower-atmosphere temperature over land areas only. Yes, it’s random. But it’s not the simplest kind of random.
David Whitehouse thinks it’s “obvious” that it’s not “randomly distributed.” That’s because he doesn’t know what he’s talking about.
By the way — David Rose indulged in cherry-picking in more ways than one. It was cherry-picking to conceal what happened before 1997, because that way he hides the part that makes the rising trend obvious. But it was also cherry-picking to choose the satellite lower-atmosphere data over land areas only from RSS. Remember how that NASA data didn’t even “look like” anything but fluctuation on top of a rising trend? David Rose picked his data set because at least it didn’t look like a rising trend if you leave out the stuff before 1997. But as I’ve said many times, “looks like” is a bad guide and when we do the math, the rising trend is still there — even in David Rose’s chosen data set, even leaving out what came before 1997.
David Rose’s original article has been a major embarrassment to him, and to the many climate deniers who have insisted on supporting his argument. In fact, maybe the best thing about David Rose’s original blunder is that this gift just keeps on giving. It spurred Ross Clark to embarrass himself, then inspired James Delingpole to embarrass himself, then (probably best of all) got Lamar Smith and the U.S. House of Representatives committee on science, space, and technology to embarrass themselves. And it didn’t quit then, Jaime Jessop joined the club, and now David Whitehouse and the GWPF have joined in.
If they were smart they’d quit talking about this, just “lay low until the heat’s off,” because every time they do another “defense” of Rose’s nonsense, I get to do another post showing how lame their defense is. More people read it. More and more people see just how ridiculous their claims really are. I feel like Christmas arrived early this year.
This blog is made possible by readers like you; join others by donating at Peaseblossom’s Closet.