Tuesday, January 26, 2010

The Devil's Climate Dictionary

At the dawn of the twentieth century, Ambrose Bierce published The Devil's Dictionary, a satiric (some say caustic) puncturing of much that needed puncturing. For example:
SENATE, n. A body of elderly gentlemen charged with high duties and misdemeanors.
I'm firmly of a mind that the whole Anthropogenic Global Warming oh noes Thermageddon business is firmly in need of puncturing. Sadly, my level of caustic snark is nowhere in Mr. Bierce's league, so instead I offer here some words and phrases that are absolutely central to the scientific discussions, but which you will never see in the Mastodon Main Stream Media. Feel free to leave your own caustic snark in the comments. [UPDATE: OK, snark is here.]

When you see discussions about temperature, you're almost never seeing the actual temperature. Rather, you're seeing the "anomaly". This is the difference between the temperature and the average temperature. While this is actually scientifically useful, "average" introduces all manner of opportunity for error, either intentional or not.

The best way to detect the error in this calculation is to look at the long term raw temperature data; trends will be clearly seen, and the long term view will give you context as to whether the current reported period is unusual or what you would expect. For example, here is the unmodified (raw) temperature from the 88 Siberian weather stations in the GHCN temperature data set:
Flat to maybe falling over 130 years. Here's the anomalies (the difference between these temperatures and the "average" temperatures) from the same stations over the same period.
FAIL. Something is out of whack with the adjustment process. How it's done is terribly important for you to know whether things are getting warmer or colder.
Decadal Oscillation
The Oceans are a massive heat sink, much, much more significant than the atmosphere. We've learned a fair amount about climate cycles in ocean temperatures, for example the El Nino warmings that we see periodically. I hear more about the Pacific Decadal Oscillation than the Atlantic one.

The "oscillation" refers to alternating cycles of warmer and cooler ocean water, and approximately 30 year intervals. The warming that was observed during warm PDO cycles matches warming periods in the twentieth century precisely (e.g. 1905-1947, 1977-1998), and the cold PDO cycles match the cooling periods in the twentieth century.
When the raw temperature is homogenized and gridded, the computer algorithms look for situations where the temperature shows sudden, unexpected changes. For example, if a weather station were moved to a different site, you might suddenly see the station reporting warmer or cooler temperatures. This is a discontinuity, and artificial modification to the temperature record.

The computer algorithms adjust temperature data when they find a discontinuity, to eliminate it. This is done automatically by machine, though, so it's very important that reported discontinuities are due to modifications to the instruments, and not an actual sudden change in temperature.

People are now investigating the weather service's computer programs that do this. Shockingly, it seems that these programs are misidentifying very large numbers of actual temperature changes as instrumental adjustment, and applying "corrections" to the temperature data. Even more shockingly, this adjustment causes older data to be adjusted downwards (made cooler) and newer data adjusted upwards (warmer). This process has amplified the reported warming over the twentieth century by a factor of four.
The whole debate about AGW is uninteresting if we only look at the last 100 years - we wouldn't know whether we were in the realm of normal variation or not. To me meaningful, we need to look at thousand-year climate variations, and to do that, we can't use thermometers (which were invented around 1600). Instead, we use "proxies" such as tree rings that give us a sort of view into what temperatures likely were back in the day.

The "hockey stick" pictures that you've all seen rely on tree ring proxy data stretching back a thousand years or more. These pictures usually switch to thermometer data once it's available (starting around 1850 for a decent part of the earth's surface).

But there's a problem: the reported thermometer data say that the temperature has been going up, up, up over the last 50 years; tree ring proxy data says that the temperature has been going down, down, down for the same period. This difference is called divergence, and is probably the most important scientific discussion going on today. "Hide the decline" refers to this.
What's the average temperature of the world? We have a bunch of theremometers measuring temperature at different locations - so how do you average them? Gridding is the process of breaking up the earth's surface into identically-sized (or as close as you can get) areas, so you can measure the temperature in each area. Then all the areas are averaged. Voila!

Except there are 8000-odd grids, and there are only 1000 thermometers. So what do you do? The software massages the data, interpolating results from surrounding grids where there are thermometers.

Lately, it's come to light that interpolation is being performed for locations where there are actually weather stations that are reporting data - data that is not being used. Shockingly, the interpolation uses data from much hotter grids to produce an estimated temperature that is much higher than the raw data shows - for example, record reported heat in mountainous Bolivia when there have been no actual measurements in Bolivia (in the data sets, anyway) since 1990.
Discontinuities make a long term data set "inhomogeneous". Computer algorithms manipulate the data to make the data sets long term homogeneous. However, this modification is poorly documented and introduces errors, like in Darwin, Australia, where the process changed a 0.7° decline over the course of the 20th Century into a 1.2° increase over the same period. While there were discontinuities in the data, Darwin had five separate stations recording temperature data, and all five agreed very closely with each other. None of them agreed with the 1.2° "homogenized" result.
There you go - you now know more about the scientific debate than 99% of world, and 98% of the media. Next time you hear someone mutter "the science is settled", these are the terms to toss out there to make him squirm. Specifically, here are some questions that he won't be able to answer:

"Why are the reported anomalies from Siberia increasing, while the temperature there is not?"

"Why do the tree ring proxies used in the IPCC AR4 report show declining temperatures after 1960, while the reported thermometer data show increasing temperatures? Are the proxies wrong now, or were they wrong in the past, or were they always wrong? How do you reconcile the different errors?"

"Why has the number of thermometers in the data sets fallen from 8000 to 1000? Does it make sense to homogenize data from stations located close to each other with data from stations a thousand miles away?"

Actually, I lied - he will be able to reply: You must be one of those Deniers. Git the heck off my lawn.

Dude, you shouldn't have a lawn. Don't you want to do your part against Global Warming?


Keith said...

Have you seen "Roger's Proffanasaurus"?

Borepatch said...

Keith, that's a riot - I'd never seen it before.

Definitely NSFW, but probably useful, especially when posting comments!