Words: Correction, Compensation, Adjustment, Proxy, Trick

[BPSDB] Years ago, when I was a researcher, I used a technique called ‘infrared spectroscopy’ a lot.  It involves passing infrared radiation through a sample of a substance, then splitting the infrared radiation up into its various frequencies (a spectrum) and measuring which frequencies are absorbed by the substance. Most substances (though not all) absorb IR radiation.

The resulting pattern of peaks, called bands, can be used in analytical chemistry to identify compounds, if their IR spectra are known. Under favourable conditions, such as the vapours at low pressure I often worked with, you can get information that will help (along with other techniques) to discover the actual shapes of the molecules, and even to discover facts about the internal electronic structure.

So, in theory, all you do is pass the IR through the sample and record the spectrum, right? No. Between the lamp and the sample are other things that absorb IR. These could include the solvent (if you’ve had to dissolve the sample), but, in particular, there is carbon dioxide, which is a component of air that strongly absorbs IR. So your spectrum includes bands from the interfering substances, which will complicate matters and likely obscure details of the spectrum you want to see.

But there is a clever trick (that word came immediately to my mind writing this, in the sense of a clever thing to do). That is, you split the IR beam into two, and pass one half through the sample and the other through a path of the same length containing all the things (air, solvent, etc.) except the sample itself. Then you can electronically subtract the signal from the reference beam (the one without sample) from the signal from the sample beam. And you are left with the signal that comes from the sample alone. Clever, eh?

This is called correction or compensation. There is nothing dishonest about it. The result is a genuine spectrum. If anyone is being cheated, it is nature being tricked into giving up her secrets, as without the compensation or correction you would have much less, if any, information about the sample spectrum.

I mention this, as hordes of global warming ‘sceptics’ have been crying ‘fraud’ simply through seeing words like ‘correction’, ‘adjustment’ and even ‘trick’ in some otherwise unremarkable stolen emails containing informal correspondence between climate scientists. These words do not in themselves imply anything nefarious. Now there may have been improper behaviour on the part of the scientists, but that would require some contexts for these words, providing some actual evidence. But no evidence for serious wrong-doing (apart, apparently, from getting irritated with ignorant harassment) has been presented.

Corrections are not limited to the lab. Whenever you press the ‘tare’ button on a set of kitchen scales, you are making a correction, by subtracting the weight of the bowl from the flour you are weighing out. Without that correction you will have a less satisfactory cake. Corrections, or compensations, or adjustments, are an essential part of the tricks needed to extract real data from the raw data of the real world.

In science, most things are not measurable directly. Temperature is one of them. When you measure a temperature, you are actually measuring an electric current that changes numbers on a display or moves a needle on a dial. Or you are measuring the expansion of a liquid (alcohol or mercury) along a tube. Neither method measures temperature directly, which is, in fact, impossible. What you are measuring is a proxy for temperature: another measurable quantity that is related to temperature in a way that is known from theory and tested by practice. But the manufacturer still has to make corrections to ensure that the fixed points of the measuring device correspond to the fixed points of the standard temperature scale. Building these corrections in is called calibration.

Actual measurements of temperature, using thermometers, go back only a short time, at most a couple of centuries in relatively few places. That means that scientists trying to study temperature way back in history or beyond have to use proxies – whatever they can find, wherever they can find them. The proxies include annual tree-ring growth, ratios of isotopes of oxygen in ancient snow, and the patterns of annual deposition in lakes (varves). These may not vary in a straight line relationship with temperature. Also, nature may inconsiderately have placed the proxies in different places at different times, such as trees growing at different heights.

To try to make these separate items match up, you need to calibrate, or adjust, the data, so they represent the temperatures correctly. For example, you may need to allow for trees growing at different rates at different heights. You can imagine that a great part of research is given over to debate about these corrections. But there is nothing essentially dishonest about doing this. If you find that numerous different proxies, when compared over the same period of time, agree pretty well, then you can have reasonable confidence that the proxies are measuring the same thing. So it is with the so-called ‘hockey-stick’ graphs.