Saturday, February 21, 2009

Honesty in Climate Reporting and Discussion

I am a fan of Edward Tufte’s work. For most of you I suspect that will inspire nothing more than a Huh? But I want to use an example from the global warming debate to show how, by manipulating graphics, you can sell opposing messages with the same set of data. It is this subject (that of graphic manipulation) and the need for honesty in data presentation, about which I feel rather strongly as an academic, that has made me a fan of Edward’s writing.

At the urging of the Advocate, I am going to use a post by Nate Silver (my turn for a Huh?) as my example, and will compare his selected use of segments with, an overall temperature plot from the British Climate Research Unit at Hadley (HadCRU) and with a plot from the Goddard Institute for Space Studies (GISS) in New York., which James Hansen heads. The initial period I am going to look at is from 1940 to 1975. This is the period that initially George Will wrote about, commenting that, in 1975, scientists were looking at the trends and he quoted nine articles from that time with discussions on this trend. So does our friendly critic look at that time period in his commentary – er, well, not exactly!. Here is the GISS plot for about that period:
Source GISS

and this is the more simplified segment of the HadCRU plot.
Source Hadley Climate Research Unit

They’re both small segment of larger plots I will get to in just a minute. And if you look at both you can see (you don’t have to be a rocket scientist) that the trend is down. That was what the scientists were talking about in 1975. So is that what our good friend showed – of course not that wouldn’t allow him to make his point, so he expanded the range, backwards from 1940, into the previous segment of time, including enough points to suggest that the overall plot was upwards (and ergo all those scientists, not to mention George Will, were idiots).

Source FiveThirtyEight

Now you should be able to see that if you knock the data off from before 1940, you will see the same trend as in the earlier graphs.

Now the next bit is a little more interesting, because he pulls back to show that there was a warming trend before 1940, and takes the curve back to about 1860, and, depending on where you define the end points he puts a single warming trend line through the data.

That simplifies what actually went on, looking at both the GISS plot and then HadCRU:


And for the simpler HadCRU

You can see that the way in which the temperature was behaving was not quite the simple plot that has been suggested, but that it fluctuates up and down, and if we break the HadCRU plot into roughly 30-year bits, you can see it drops 1880-1910, then goes up, 1910-1940, then drops 1940 – 1970. So it was really not that easy, back in 1975, to decide where the temperature would go next.

Where it went is now well-publicized, and I will put all three plots up here in full for the next bit.

First the Silver plot:
Source FiveThirtyEight

Now the GISS plot:

and then the HadCRU plot (I showed their entire illustration, lest I be accused of cropping anything):

There are a couple of interesting things to the shape of the lower two curves, that are not clear if you just draw the simple line, that Mr Silver uses, through the data. The first is that it warmed about as much from 1910 to 1940 as it did from 1970 to 2000, in other words the rate of warming was about the same. Not much attention is usually paid to this part of the warming trend – well that is because of this other graph, which is the one that usually gets all the attention.
Source White House Initiative on Climate Change

Now we will forgo for now the discussion about measuring carbon dioxide over an active volcano but you will notice the different slopes for the curve back in the 1910 – 1940 period where we were getting the same rate as we have just recently – curious scientists want to know why? But actually, and more to the point, that graph is a very smooth curve from 1860, as you may note the real temperature data is not.

There are two final points to consider. The first is why I put up the HadCRU and the GISS plots. You may note that at the end of the HadCRU data the curve turns over and starts to go back down. That’s what the data shows. But the GISS plot does not carry the plot that far down. In fact earlier plots of data that just showed the last ten years of data (with 1998 at the peak) disappeared from the GISS site about the end of last year. The differences between the interpretations, and the fact that recent data (in the HadCRU form) does not allow the hysteria about the imminence of doom that comes from the GISS Head, is perhaps why Hadley are now starting utter some words of caution about this whole debate, and that we cannot make misleading claims.

I am going to close with a plot from the 1995 IPCC report that does not get a lot of attention these days, but might explain why the leveling off in global temperature is happening:
Source IPCC 1995 full report

I have written earlier about the Little Ice Age, and will again, and on not only the last warming period, but also the ones before it, for which there is much evidence. It turns out not to be as simple a temperature plot as the IPCC summarized in that plot, but the realities of this cyclic nature of global climate are recorded, not only in books and manuscript, but in archaeological sites, and beneath our feet.

So let's stop with the insults (if you attack a person rather than their science, it concedes the point that their science is stronger than yours) and discuss the volumes of real scientific data that make exactly where we are in the progress of global warming a whole lot more of a question that many people would like you to think.


1 comment:

  1. Now, see, this was the kind of response I was expecting.

    the Advocate