Salviati: That was delicious! Thank your wife for such a wonderful meal, and thanks again for having us over for the long weekend.

Simplicio: Yes indeed — your wife is such an excellent cook.

Sagredo: It’s my pleasure — I enjoy your company so much. I’m so glad you were both able to visit for a few days. It’s not easy to find such excellent conversationalists. And I have to agree about my wife’s cooking.

Simplicio: She’s gone to her choir rehearsal, did you say? That leaves us the rest of the night to entertain ourselves.

Sagredo: I’d like to hear more about a topic I know the two of you disagree on.

Salviati: You mean, global warming?

Sagredo: Of course. Salviati, I know you think it’s real, it’s caused by human activity, and it’s very dangerous. Simplicio, I know you think the idea is nonsense.

I’m not sure what to think, so I’d like to hear what two of my favorite thinkers have to say about it — especially after the amazing December we’ve just had. Wasn’t it the warmest on record? And not just here in New England, wasn’t it the warmest on record in England as well, and other places too?

Salviati: Yes it was. Here in the U.S.A., 29 of the 48 lower states broke the record for hottest December on record. More to the point, they didn’t just break the record by a little, but by a country mile.

Sagredo: How much did they break the record in England?

Salviati: Let’s have a look, I’ve got it here on my computer … The monthly average this December broke the record by 1.6 degrees C, which is almost 3 degrees F.

Simplicio: I think you’re referring to the “Central England Temperature” record. It’s the longest continuous record in the world, and I’ve got it here on my computer too. If you had looked back to the year 2010 you would have found a December average temperature of minus 0.7 degrees C. I was in England then, and all people could talk about was how cold it was. If you had seen a recent “Horizon” program over there, you would have watched a group of Climate scientists attempting to explain a succession of bitterly cold winters.

Salviati: What a coincidence, I happened to be there too. That December was quite cold, but it certainly didn’t out-do all other Decembers by 1.6 degrees C, in fact it wasn’t even the coldest on record. And as I recall, the English talked about plenty of things, including the Queen’s oldest granddaughter getting engaged to her rugby-star boyfriend.

As for “a succession of bitterly cold winters,” that’s the kind of exaggeration too many people indulge in. If we use the standard climatological definition of winter — December, January, and February — then the coldest recent winter in England was 2010 but it wasn’t even close to their record coldest winter. It was tied for 57th-coldest.

Sagredo: That hardly seems like anything out of the ordinary.

Simplicio: But December 2010 was the 2nd-coldest out of 357 English Decembers on record. I’d say that qualifies as “out of the ordinary.”

Salviati: Does it really? Even if I agree that it’s not “ordinary,” it’s not the record. And it only beat out 3rd place by 0.2 degrees C, while December 2015 took 1st place by a whopping 1.6 degrees C — eight times as much.

Sagredo: I have to admit, that’s a lot more impressive. Quite a lot, actually.

Salviati: Besides, I’ve tried to emphasize many times that what really counts is the trend. Temperature is a combination of two factors: the underlying trend, and the fluctuations. The fluctuations go sometimes up, sometimes down, sometimes a little, sometimes a lot, but the one thing they don’t do is stop. So, even with global warming we’ll still have cold months and hot months, but because of the trend, the hots are hotter and they’re more frequent.

Simplicio: Well, the Central England Temperature record does offer a very simple way to test whether the trend is upward. Nate Silver, in his book “The Signal and the Noise,” suggested that, at the end of a decade “if temperature changes are purely random and unpredictable, an increase and a decrease in (the next decade) temperatures are equally likely.” The Central England Temperature (CET) record covers 34 decades, from the end of the first decade in 1679.

The results are illuminating, at least from my sceptical point of view. The first 22 decades, to 1889, show 10 warming and 12 cooling decades, randomly scattered, with negligible overall change. The next 6 decades to 1949 are all warming as temperatures climb out of the Little Ice Age, from about 9.1 to 9.6 degrees C.

For the 6 decades to 2009, we are looking for the impact (or non-impact) of exponentially increasing CO2. There are 3 warming decades and 3 cooling decades. If we include the six complete years of the next decade, to 2015, there are 3 warming and 4 cooling. There is absolutely no sign of any steady increase from 1949 onwards.

Now, admittedly, the temperature increases in the last two decades were significantly greater than any previous, taking the decadal averages above 10 degrees for the first time. The average temperature in the decade ending 2009 was 10.4 degrees C, and in the incomplete decade to 2015 the average was 10.01.

There is an increase since 1999 of about one half of a degree which I make no attempt to explain. We can, however, be reasonably certain that this sudden increase in the decade ending 2009 has nothing to do with increasing CO2 in the atmosphere.

Salviati: I’m afraid you’ve completely misunderstood what Silver was doing in his book, and how to apply it. He doesn’t use that method to test whether or not a trend is upward. He uses it to illustrate how Bayes’ theorem can update an estimated probability that the trend is upward.

Using it depends on comparing the chance of getting an increase from one decade to the next if there is an upward trend, to the chance if there is not an upward trend — which, if temperature is just random, is indeed 50-50. But if the upward trend is real, the chance of an increase from one decade to the next depends on the amount of noise, i.e. the size of the fluctuations. Your using it for CET is misleading because the noise level is huge compared to the noise level for the global temperature that Nate Silver was talking about.

We can see that easily by graphing the two, CET and global temperature — which I’ll take from NASA — on the same plot. I’ll use a black line for CET and a red line for global temperature (the global data don’t start until 1880):


The noise level in CET is so large, compared to global temperature, that the chance of a decrease from one decade to the next is nearly the same, whether there’s an upward trend or not. For global temperature, the noise level is low enough that the chance of a decrease from one decade to the next is much smaller with the upward trend than without — Silver estimates it at a 15% chance, compared to the 50-50 chance when temperature is just random.

Because of the large chance of a decade-to-decade decrease in CET, just because of the noise level, the method is useless even for Silver’s stated purpose. And for testing whether or not there’s an upward trend, it’s about the weakest test you could devise. You really should have paid more attention to Nate Silver’s title — there’s not just signal, there’s also noise.

You also seem to imply that there somehow has to be a “steady increase from 1949 onwards.” Not only does this ignore the effect of the noise which is so large in CET, it quite ignores the fact that greenhouse gases like CO2 aren’t the only influence on climate. There’s also the effect of what are called “sulfate aerosols,” which cool the planet. If you read Nate Silver’s book, you’d know that a big source of sulfate aerosols is industrial activity, especially burning coal, and that after world war II it increased dramatically. If not for the warming effect of greenhouse gases, that would have made the world cool off substantially. Once we stopped increasing sulfate aerosols — since they also cause acid rain — global warming really took off.

( to be continued …)

If you like what you see, feel free to donate at Peaseblossom’s Closet.

3 responses to “Dialogue

  1. Nice! Playing the Galileo card, I see… ;-)

  2. I like that. It has grace in it, not the bitterness of mutual insults. And I like also the noise comparison between regional and global temperatures. We knew it somehow, yes, but to see it in a plot is completing the picture, so to say.
    I have only one small nitpicking: the notion, that there is the trend, and there is the noise, like those tow are separated in an objective manner. To me, the separation of trend and noise is arbitrary, and done by the choice of the smoothing time constant for the fit curve. So we separate the trend out of the signal through a conscious decision of ours.
    But this is a more, say, philosophical cantankerousness.

  3. Mildly OT: Berkeley Earth is in for December. It’s another new record monthly anomaly, and by a fair bit … 0.06°C above their previous record, which was of course October 2015.