randombio.com | commentary
Tuesday, July 18, 2017

When is it acceptable to retroactively ‘correct’ your data?

A new report challenges the global temperature adjustments made by NASA and NOAA.

T hermometers are pretty simple instruments. Not much, you'd think, can go wrong with a thermometer, right? Wrong! According to NOAA, in the old days, before computers told us that the Earth was warming up, most of the thermometers were miscalibrated. How do we know this? Because we know it's getting warmer.

That's not the conclusion of some mad climate skeptic. NASA admits using weather models to calculate global average surface temperatures. They've also “corrected” the old results for a variety of reasons.

A new report, available here[1] says that these corrections have preferentially lowered earlier global average surface temperature (GAST) estimates by as much as 0.18°C and raised later ones by as much as 0.16°C, thereby exaggerating the amount of global warming.

From the Summary and Conclusions:

While the notion that some “adjustments” to historical data might need to be made is not challenged, logically it would be expected that such historical temperature data adjustments would sometimes raise these temperatures, and sometimes lower them. This situation would mean that the impact of such adjustments on the temperature trend line slope is uncertain. However, each new version of GAST has nearly always exhibited a steeper warming linear trend over its entire history.

. . . The conclusive findings of this research are that the three GAST data sets are not a valid representation of reality. In fact, the magnitude of their historical data adjustments, that removed their cyclical temperature patterns, are totally inconsistent with published and credible U.S. and other temperature data. Thus, it is impossible to conclude from the three published GAST data sets that recent years have been the warmest ever –despite current claims of record setting warming.

NCDC temperature corrections
Plot of NCDC temperature corrections from page 16 of the report. The authors question the adjustments, saying they eliminate evidence of periodicity and add 0.15 degrees to the 20th century warming.

This report will not be a death blow to the global warming theory. The issue has been so politicized that it has become nearly impossible to distinguish fact from propaganda. This may be understandable, since literally trillions of dollars are involved. We may never know how much global warming is real and how much is due to bad scientific practice.

The question I want to raise is this: since when is changing old data standard practice in science? I've been a professional research scientist for 31 years, and I've never heard of such a thing. I was taught that you collect your data, and if you discover that it's not correct you throw it out. When did it become permissible to just go back and change old data when you decide the results are wrong?

The GAST report shows the risks: once you start tampering with old data, even if you know it contained systematic errors, and even if you have a good idea of what those errors may be, you have demonstrated that the original results were questionable.

You've now refuted your original data, which means your methodology was faulty, so why should people accept your corrections, which could also be faulty, especially when the corrections just happen to beef up your theory?

Empirical measurements are sacred

Just as religions have unchangeable texts, which they call scriptures, science too has unchangeable texts, which we call the original data. In science, the original data are sacred. If your old data contain errors, the measurement must be repeated. If it's impractically expensive, that's too bad. If it's impossible, that's also too bad. There is no way to change the numbers without being dishonest about what happened.

Now, maybe some people would say if you could find the old thermometer and prove it was miscalibrated, and if you could prove that the amount of miscalibration didn't change since 1938, maybe correcting the data would be acceptable. Or if you find that all the old pre-Depression-era thermometers were located near jet exhausts and computer cooling fans, you might have to re-weight your averages.

But if you discovered that your 1947 data were unusually warm because, say, a flying saucer emitted a beam of tachyons toward your thermometer and heated it up, changing the numbers would eliminate evidence of something potentially important.

Whereas, if you said, well, maybe using buckets to sample the ocean temperature wasn't such a good idea compared to the ARGO floating buoys we use now, so we'll just knock off, oh 0.15 degrees sounds about right, most people would agree that you've just added a fudge factor.

WWCD (What would chemists do)?

Nowadays I work mostly with an instrument called a mass spectrometer. It uses very small bits of tubing, and the main difficulty is keeping some other people, whose name I won't mention here, but whose initials are S.B., from getting it clogged up.

The second difficulty is keeping it calibrated. The mass spec tells you the mass of whatever molecule you put into it. Suppose I injected a sample that I knew had a mass of 386.3548, but the machine said it was 369.3521. I decide that this means the machine needs recalibration. Am I then entitled to go back and add the difference, 17.0027 amu, to all my previous results?

The answer is clearly no. I have to re-calibrate and repeat.

Courtroom example

In many ways doing science is like being in a courtroom every damn day. Suppose you were a prosecutor. Your vic, one Mr. H.H. Silver, claims he was in a furry bar when someone slammed his head into his food dish. Silver is rushed to a nearby animal clinic where he recovers. A reliable witness accuses Silver's former friend, Ed, of the crime. You want to nail Ed for Drunk & Disorderly, Assault & Battery, and violation of the leash laws.

Ed claims he left the bar at 10:13 pm, but the video shows him there at 10:20 when the crime took place. But Mr. Ed's lawyer claims the clock on the DVR was off by seven minutes. You check the clock and discover that, in fact, it was off by one hour and 17 minutes. What do you do?

With this new evidence, your theory is as hopelessly screwed up as this story. Is it okay to save it by going back and changing the time track on the recorder? The cost of throwing out the bad evidence would be high: if you can't prove your theory, the real culprit, Ed, who has by now dragged hay all over the station, will escape justice.

Science is the same way. The instrument reading is a composite of the physical factors that influence it. If you change those numbers, you're no longer talking about the real world, but about the world as you would like it to be. And that contradicts the whole point of doing science.

1. Wallace JP, D'Aleo JS, Idso CD (2017). On the validity of NOAA, NASA, and Hadley CRU Global Average Surface Temperature Data & The validity of EPA's CO2 endangerment finding: Abridged research report.

created jul 18, 2017; last edited jul 20 2017, 7:39 am

See also

How to lose a scientific discussion
Browbeat your opponents, call them names, and use lots of pie charts.

An autopsy of the late global warming movement

What is the value of computer modeling?
If mathematical models are done badly, they will discredit an entire branch of science. It's happened before.

On the Internet, no one can tell whether you're a dolphin or a porpoise
Name and address
book reviews