His method? A favorite climate denier trick: the one called “cherry-picking.”
Clutz uses yearly average sea ice extent in the Arctic according to satellite data, from 1979 (the first complete year with satellite data) through the 2017 (the most recent complete year). He plots the data like this:
The white vertical bars are the yearly values, while the red line is Clutz’s idea of the “trend.” He has already chosen a dreadful graph to show the data. The entire amount of change over the nearly 40-year time span is compressed into a slice only 1/7th the height of the graph, which makes it much harder to see the real changes, which, it seems to me, serves Ron Clutz’s purpose perfectly because I don’t think he wants you to see the real changes. If you can see up close how his claimed “trend” matches the data, you might know how fake it is.
Then Ron Clutz gives us his interpretation:
There was a small loss of ice extent over the first 15 years, then a dramatic downturn for 13 years, 6 times the rate as before. That was followed by the current plateau with virtually no further loss of ice extent. All the fuss is over that middle period, and we know what caused it. A lot of multi-year ice was flushed out through the Fram Strait, leaving behind more easily melted younger ice. The effects from that natural occurrence bottomed out in 2007.
Let’s look at the data more closely, and instead of deciding what we want it to say, let’s listen to what it’s telling us. Here’s a vastly better (i.e. more informative) graph of the same data:
One way to get an estimate of a non-linear trend is with a good smoothing method, and I like my own program for a lowess smooth because it also compute the rate of increase or decrease at each moment of time. It gives this (the trend estimate shown as a red line):
There are many other choices, and with a competent analyst I could have a fruitful, perhaps even heated (but civil) discussion about their merits. But not with Ron Clutz.
Here’s his idea of the trend (the blue line):
He has split the time span into three segment. The first is from 1979 through 1994, but he doesn’t estimate its trend by least squares regression, or by any other sensible method, but simply by the line from the first to the final data points. Essentially his method completely ignores everything that happened in between. My opinion: it’s because Ron Clutz wanted to ignore that stuff; only by doing so can he get the early rate of change he wants. He’s free to choose a high point as his ending year, in order to make the earlier trend rate less negative and the following trend rate more negative. That’s what he wants the data to say — not what the data are trying to say themselves. Most of the data values in that time span say otherwise, lying below Clutz’s trend line, but his method ignores them. I prefer what the data say.
His second segment is from 1994 through 2007, and his claim of what the trend was doing during that time is … well, there’s no polite way to put this … stupid. Of the 12 values in between those years, 11 of them are above his trend line. To me, it’s just obvious that Clutz chose this line not because it reflects what the data are saying, but because by choosing an extra-low ending point he can make his “fast decline” episode longer than it should. I prefer what the data say.
And of course, using 2007 makes the next segment start on an extra-low point, helping him impose what he wants rather than what the data are trying to say. His final segment, from 2007 through 2017, also utterly fails to reflect the actual trend during this time span, mainly by starting with the extreme low value in 2007. That’s not just cherry-picking, it’s cherry-picking supreme. Shame on you, Ron Clutz.
I’ll also repeat my opinion: part of the reason he used his original dreadful graph is that it makes it so hard to see the details of his purported “trend” that he can sucker people with his fake trend. When you can see clearly how Clutz’s trend compares to the data, you can see clearly how ridiculous his claim is.
Compare my trend estimate (a lowess smooth) to Ron Clutz’s (which depends on deliberately choosing time points and ignoring everything in between them):
Which do you think gives a more realistic portrayal of the data?
As I mentioned, my lowess program returns the estimate rate of change at each moment, and it also estimates the uncertainty in that rate estimate. That enables me to track how the rate of change (the rate of Arctic sea ice loss) has changed over time. If I compare my estimate (red line, with dashed red lines showing the 2-standard-deviation uncertainty range) with Ron Clutz’s rate estimates (blue line) I get this:
This shows us clearly just what Ron Clutz has done with his “trend.” From 1979 to 1994, he chose points that would give an artificially low (slower decline) trend estimate. From 1994 to 2007 his chosen (cherry-picked) points give an artificially high (rapid decline) estimated rate. From 2007 to the present his cherry-picked starting point, combined with the amateurish attempt to estimate the trend by just connecting the endpoints, enables him to get an artificially low (slower decline) trend estimate. That’s how he justifies his entire “analysis.”
I too can play the “pick two moments to get what you want” game. Here, for instance, is an alternative choice (I’ll call it the “other cherry” choice):
Just as Ron Clutz can choose times to make his third episode decrease most slowly of all, I can choose them to make the third episode decreasing fastest of all. Just as Ron Clutz can choose times to make this second episode decrease fastest of all, I can choose them to make the second episode decrease most slowly of all. Ron Clutz’s “pick a few time points and connect them with lines” strategy, when done to get what you want instead of to get the truth, can produce just about any result you want. My opinion: that’s exactly what Ron Clutz did.
Heck, I can even take his time points — 1994 and 2007 — and estimate the trend, not by connecting endpoints, but by finding the best-fit continuous piecewise-linear trend by least squares, then use those to estimate the rates of change. That gives this:
Clearly, even if we allow Ron Clutz’s misleading choices of change points for the trend, you still have to use an ignorant and misleading way to make a trend out of them to get what Ron Clutz wants.
Here’s my opinion:
We’ve dealt with Ron Clutz before, in another example of pretending to do “analysis” but not really doing it. Clutz finds it easy to justify ludicrous opinions because he doesn’t know how to do the analysis. We get the same kind of thing from Cliff Mass — repeatedly. They, and others, push out garbage “analysis” regularly, and even when it’s proved wrong they refuse to admit any error. Frankly, it’s very difficult to have a productive discussion about analysis with someone who knows nothing about it but think he does. Honesty is off the table.
Anyone can make mistakes. For example, Roy Spencer made a doozy of a mistake some time ago, but when I called him out on it he did not insist that his bonehead mistake was correct. As far as I know, he hasn’t mentioned it since, and I suspect that’s because he knows — and admitted to himself — that it was wrong. But too many others, like Ron Clutz and Cliff Mass and Anthony Watts and Christopher Monckton etc. etc. etc., will put out ideas that are so idiotic they’re embarrassing but refuse to admit error even when it’s proved. You can’t reason with those people.
But hey, that’s just my opinion.
This blog is made possible by readers like you; join others by donating at My Wee Dragon.