From the “confirmation bias affects Top. Men.” department…a modern equivalent of “Mike’s Nature Trick”.
WUWT readers have probably seen the various graphical depictions of temperature coming from Dr. Ed Hawkins, who in his climate-lab-book website has been publishing spirals, bars, and other depictions of rising temperatures. For example (click for animation):
Hawkins is no garden-variety graphical tinkerer, he’s a climate modeler at the National Centre for Atmospheric Science (NCAS) at the University of Reading and IPCC AR5 Contributing Author, soon to be shepherding IPCC AR6 as lead author.
His visualizations have gotten a lot of notice. Millions of views on social media by my estimate. Plus, Hawkins boasts:
Recently, in a Twitter campaign called #MetsUnite, TV meteorologists and weathercasters around the world sported ties, pendants, and coffee mugs to bring attention to global warming, using one of Hawkins temperature visualizations. I pointed out the hypocrisy of it.
But like many, since Hawkins was just producing dramatized visualizations, I didn’t pay much attention to them to check for accuracy. We’ve seen so many exaggerations from the climate crusaders, it was simply lost in the noise. Millions of others apparently didn’t notice either, including his climate science peers. Otherwise we’d have heard about what is about to be said.
Enter serendipity via social media. One person did pay attention, and asked Dr. Roy Spencer about a similar spiral graph depicting temperature increase. He writes on his website:
About a year ago, Finnish climate researcher Antti Lipponen posted a new way to visualize global warming, an animation he called the “temperature circle”. It displays the GISS land temperature data as colored bars for each country in the world radiating from a circle. As the temperature in a country goes up, the colored bar changes from a blue bar to a red bar, and gets longer…and wider:
I didn’t pay much attention to the ‘temperature circle’ at the time as it seemed rather gimmicky. But yesterday I was asked on social media about it, and I watched it again. The video has about 163,000 views on Twitter and 175,000 views on Youtube, and its impact on people’s perception is evidenced by some of the recent Youtube comments:
“Excellent presentation of a large mass of data. But the denialists will invent reasons to ignore it.”
“This is among the scariest presentations I have ever seen. Yes, I have kids.”
After thinking about the animation for a minute, it quickly became apparent why warming displayed this way looks so dramatic… and is so misleading. The best way to describe the issue is with an example.
Assume all the countries in the world were 2 deg. C below normal, and then at some later time all of them warmed to 2 deg. C above normal. Here’s the way the ‘temperature circle’ plotting technique would display them (ignore the displayed year and ‘real’ data, just focus on the blue and red segments I have superimposed):
Note that the coldest temperatures will have the smallest area covered by blue, and the warmest temperatures will have the largest area covered by red, even though the absolute sizes of -2 deg and +2 deg departures from average are the same.
I consider this very deceptive.
What this display technique does is cause a linear rate of warming to appear like it is non-linearly increasing, or accelerating. The perceived warming goes as the square of the actual temperature increase.
In fact, even if warming was slowly decelerating, it would still look like it was accelerating.
If this was a graphics artist playing around with data in various kinds of display software, I might be able to excuse it as artistic license.
But the fact that a climate researcher would do this is, well, surprising to say the least.
He’s right. The presentation is deceptive due to the geometry used.
Dr. Spencer sent out an email notifying a number of climate skeptics of his findings, including me, and I immediately wrote this back:
“Note that the coldest temperatures will have the smallest area covered by blue, and the warmest temperatures will have the largest area covered by red.”
Based on your reasoning, the same is true for this similar animation…?
It sure looks like it. Also done by a climate researcher, Ed Hawkins.
Here’s the original.
Same issue here, more surface area given to the warmer colors, makes the warmer colors dominate visually.
Dr. Spencer responded with:
sort of… the ‘spirals’ display exaggerates linearly… the line segments increase in length with warming according to Pi*r.
But the ‘temperature circle” exaggeration causes the displayed area to increase nonlinearly, by Pi*r-squared.
Dr. Spencer also added this during our discussion:
Imagine if the ‘spiral’ temperature scale was increasing inward. In that case the spirals would get smaller with warming, which would be much less dramatic. The perception of warming should not depend upon whether the scale is reversed, and is evidence that these new display techniques have been contrived to whip up alarm since (as the recent Gallup poll reminds us) most people aren’t very concerned about warming rates that are too small to feel in their lifetimes.
And that’s why you don’t see Hawkins produce the spiral in reverse. It would kill the effect of his visualization, making it far less alarming. Confirmation bias at work.
While Hawkins spiral graph uses colored line segments (much like Lipponen does), his are oriented perpendicular to the radius of the circle, with warmer temperatures being near the edge, and as the radius increases, the segments get longer. Hawkins also uses a temperature scale to color the lines, using a color scale called ‘viridis’ . Of course, the warmer temperatures tend to be depicted as yellows and oranges, and the cooler temperatures as blues and violets, as Hawkins states:
What do the colours mean? The colours represent time. Purple for early years, through blue, green to yellow for most recent years. The colour scale used is called ‘viridis’ and the graphics were made in MATLAB.
More on the color choice by Hawkins later (which has it’s own set of problems). For now, let’s concentrate on the geometry trick.
Here’s what we normally see – a linear graph, where all data points have the same weight:
Note that in the HadCRUT4 data shown above in a linear fashion. most of the temperature increase comes after 1950. Hold that thought.
As Spencer pointed out about Lipponen’s circular graph, due to the way surface area increases exponentially with radius, far more surface area is given to the warmer temperatures than the cooler ones.
Anybody who has taken basic geometry in primary school knows this:
As seen in Figure 2, surface area increases exponentially with increasing radius.
To illustrate this with some basic geometry, I decided to take some measurements of Hawkins spiral graph. Since Hawkins spiral graph doesn’t have a reference scale on it, the only way I could get something to measure for radius was to import his graph into a graphics program and apply a pixel scale. I’ve done that in the image below from frame #1 of the animation and listed the values.
Because Hawkins didn’t provide 0.5°C and 1.0°C circle values, and because he offsets zero (to account visually for the possibility of negative anomaly values), it’s a bit different to work out, but I’ve done it in Table 1 below.
From Figure 3, the values are:
|Radius (pixels)||Surface area in pixels using πR2||Note|
|0.0°C||160||80424.7||(zero value offset is 160 px)|
|1.5°C||400||502654.8||(we aren’t there yet)|
|2.0°C||480||723822.9||(we aren’t there yet)|
TABLE 1: Values of temperature, radius, offsets, and surface area from Hawkins spiral plot.
Clearly, as the lines expand and get longer in Hawkins spiral graph, they extend into larger surface areas because the lines themselves are longer. Humans, when viewing the lines all massed together, tend to average them visually, assigning more weight to them because they cover more surface area of the circle.
It’s a visual trick, and one that is peculiar to Hawkins, because nowhere else in climate science do we see a linear graph of temperature turned into an exponential representation. I suspect Lipponen’s representation might be inspired by Hawkins.
To illustrate, for a Hawkins spiral circle, here in Figure 4 is how the linear values and the exponentially increasing surface area value graph out using the pixel values measured, including the 160 pixel offset from zero to handle negative anomaly possibilities. A polynomial curve fit is also added to illustrate the exponential increase in surface area for the values closest to the circumference of the circle.
It looks a lot like a hockey stick, doesn’t it?
When first I plotted Figure 4, and saw that the blue line didn’t follow the plotted path of a pure circle as seen in Figure 2, I thought I had made some sort of mistake. I looked at the data again, couldn’t find any errors in the way I measured it, then threw it out and started over again. I came up with the same result, again and again.
My conclusion? Hawkin’s 0.0, 1.5, and 2.0 reference circles aren’t accurate. I suspect they were some hand generated overlay, because they certainly don’t follow the surface area from increasing radii of a pure circle seen in figure 2. Either that, or he’s used some sort of non-linear scale for temperature that isn’t obvious when trying to reverse engineer his work. Not having his original MatLAB data and plots, I can’t say for sure. If I’ve erred someplace in measuring the original graph, please point it out in comments.
But one thing IS certain: by plotting HadCRUT 4.6 data using the circle/spiral method, he’s weighted post 1950 data far more heavily that data from 1850 to 1950, both in line length, as well as the surface area the pixels that make up those lines cover in the circle. Knowing this now, it is clearly obvious looking at his spiral graph endpoint in 2017:
In figure 5 above, note how the earlier lighter blues and pastel magentas are covered up by the more recent temperatures. Note also how the greenish yellows are the most prominent visual elements, both by color, and by surface area covered.
It’s basically “Mike’s Nature Trick” all over again.
The more recent graphic elements (post 1950) cover up the ones that the really didn’t want you to see. PLUS..the spiral presentation visually weights the more recent temperature data far more heavily than earlier data due to the increased surface area of the lines created by the most recent data. It’s a double-whammy of visualization bias.
Finally, remember earlier I said: “More on the color choice by Hawkins later (which has it’s own set of problems). ”
The problem is that the human eye does not perceive colors linearly, as this graph clear illustrates:
From colors in figure 6, we can clearly see that the end-frame of Hawkin’s spiral graph is mostly in the green to yellow range, and that the cooler blues and magentas aren’t just covered up, they don’t have the same visual color impact.
This is why fire trucks and other emergency vehicles are now painted a yellowish-green; it makes them more visible in traffic and easier to avoid.
The red fire trucks of days past weren’t as easy to see. It’s documented by a study: http://www.apa.org/action/resources/research-in-action/lime.aspx
(Added) Then there’s the longer lines near the edge of the circle. Because they are longer, it makes it appear in the animation as if they are moving faster due to the increased length (accelerating). This is not true, not at all.
So due to the color scale choice, and the faked-up acceleration in the animation, there’s a
TRIPLE-WHAMMY QUADRUPLE-WHAMMY of visualization bias in Hawkins graph.
In summary, Ed Hawkins spiral graph does the following.
- It gives post 1950 data far more visual weight due to increased line length and surface area of pixels that make up those lines.
- It it covers up older data with newer data, making it unavailable for visual comparison.
- The color scale choice visually weights the present data far more than the older data, shrinking it’s impact.
- (Added) It occurred to me shortly after publishing, that there’s a 4th bias. there’s the longer lines near the edge of the circle. Because they are longer, it makes it appear in the animation as if they are moving faster due to the increased length (accelerating). Title edited to reflect this.
This isn’t good science, it’s simple visual propaganda, and Ed Hawkins should retract it, in my opinion. As Dr. Spencer said:
I consider this very deceptive.
Dr. Hawkins probably won’t retract it since we’ve learned time and again that climate science often doesn’t care much about accuracy in presentations, it’s more about the messaging, and in that, he’s succeeded in pushing an alarming message. They are also exceedingly stubborn, and don’t like being shown to be wrong.
Even if Hawkins does retract it, it will be impossible to put the genie back in the bottle, since his graph is shared in millions of social media posts and tweets.
But, in the climate skeptic world, this fiasco will live on forevermore known as “Hawkins spiral trick”.