Climate heroes

Eli Kintisch interviews Richard Muller, whose BEST project has been causing something of a stir in recent days. Muller certainly knows how to get attention...

I realized that Watts was doing something that was of importance. The issues he raised needed to be addressed. It made me seriously wonder whether the reported global warming may be biased by poor station quality. Watts is a hero for what he's done. So is [prominent skeptic blogger] Steve McIntyre.

ABC in Australia is also looking at BEST and Anthony W.

Comments

[HaroldW]

Theo Goodwin --
I think Muller did his best to state the appropriate caveats in his report about the preliminary results, which "don’t yet address many of the known biases." Predictably, such nuances were lost in the general reporting, perhaps because the oral testimony is not as detailed as the written report in the above link. Or more likely because the typical reporter likes to boil down the story to a black or a white, rather than deal with describing shades of gray. I'd like to think that he [Muller] will eventually address Dr Pielke Snr's concerns with biases expressed here and here.

I'm not sure if you're still asserting that breaking a long record into two shorter records, when there is evidence of e.g., a station move, or instrument change, constitutes altering data. It is a method of trying to rationalize measurements which were made under different circumstances, and which logically require an adjustment to make them commensurate. Suppose I had a medical chart, in which a patient's oral (mouth) temperature was taken for several days, and later the axillary (armpit) temperature was recorded. It's well known that axillary temperatures tend to run lower than oral temperatures. So if you saw a sequence of temperatures thus (in degF): 99, 99.2, 99.4, 99.6, 99.8, 99.0 [1st axillary], 99.2, 99.4, 99.6 -- two possibilities present themselves. One, the increase is actually constant, but the temperature *reading* dropped by 1 degree on the 5th day because of the method; or two, the patient recovered on the fifth day, but the fever returned shortly thereafter. Or something in between might be correct. With one patient, I don't think one can be sure. On the other hand, if those are the averaged readings of a thousand patients with the same syndrome, it seems more likely that it's the readings which are misleading, and the increase was constant.

As to the lawn-mowing reference, it never happened. Thanks BBD, and Theo too...civility in our disagreements is always a good thing, although it can be difficult to maintain enough distance to try to see the other point of view. I would like to correct an apparent misapprehension though -- it seems that you (Theo) believe that I'm British. In fact I'm American, and have lived in the U.S. my entire life, with the exception of several delightful months in England some 20 years ago.

[Theo Goodwin]

HaroldW,

Maybe some Yanks are as civil as Brits. Though the care with which you state your points strikes me as more Brit than Yank.

As regards your example of the medical chart, I find it acceptable and I have no criticism of it. If the BEST people could use the same care, I would be as accepting of their work. But their processes of snipping and dropping will be automated. Individual snips and drops will make no reference to historical fact. The point is that with BEST you accept the automated process or you do not, and you cannot defend your argument by appeal to individual facts. To me, that is just more statistical hand waving with no regard for fact. I cannot but distrust it. Especially when Anthony Watts has just what is needed, documentation of historical events showing that measuring instruments were mismanaged. What Anthony has shown beyond the shadow of a doubt is that eighty percent of the historical records used by NOAA should be thrown out. That is what I want to do, along with creation of a reliable measurement regime that might give us something to work with in forty years. No amount of statistical magic is going to change my mind on this matter.

[diogenes]

Theo

You say:

"What Anthony has shown beyond the shadow of a doubt is that eighty percent of the historical records used by NOAA should be thrown out"

However, this does not seem to be what Mr Watts indicates in his paper:

"Temperature trend estimates vary according to site classification, with poor siting leading to an overestimate of minimum temperature trends and an underestimate of maximum temperature trends, resulting in particular in a substantial difference in estimates of the diurnal temperature range trends. The opposite-signed differences of maximum and minimum temperature trends are similar in magnitude, so that the overall mean temperature trends are nearly identical across site classifications. Homogeneity adjustments tend to reduce trend differences, but statistically significant differences remain for all but average temperature trends. Comparison of observed temperatures with NARR shows that the most poorly-sited stations are warmer compared to NARR than are other stations, and a major portion of this bias is associated with the siting classification rather than the geographical distribution of stations. According to the best-sited stations, the diurnal temperature range in the lower 48 states has no century-scale trend."

This seems to suggest that although the min and max temperatures are affected by siting issues, the mean is not and therefore you can still derive a useful trend.

[HaroldW]

Theo Goodwin -
As I understood it, BEST will separate long-term records into shorter-term records whenever there is an entry in the metadata which indicate events such as station moves, equipment changes, or observation changes. Here they describe fragmenting records but don't indicate what triggers fragmentation:

Data split: Each unique record was broken up into fragments having no gaps longer than 1 year. Each fragment was then treated as a separate record for filtering and merging.

So it's possible I have that wrong; perhaps they're only fragmenting whenever there's a too-large gap in recording. But they'll have to deal with the events listed above in *some* fashion. We may need to wait to read the paper.

In addition, it sounds as though they're proposing to have some sort of automated mechanism which flags unusual jumps in data. I don't know for certain whether dividing the records at such points will be an automated or manual decision. From the BEST website:

In addition to the uncertainty adjustments detailed above, we might also search for statistically significant discontinuities in the record that could indicate undocumented station moves and similar problems. The impact of discontinuities can be resolved by partitioning such data into two time series with independent baseline estimators. Hence the corrections for such discontinuities can become part of the simultaneous solution to the larger averaging problem rather than a series of local adjustments.

The use of the conditional makes it seem (to me) that they're not certain how reliable such a detection might be; perhaps they'll publish one reconstruction "as is" and another with break points added.

I agree that it would be preferable to have temperature measurements only from well-sited instruments, all made with identical equipment and with identical sampling methods. But, lacking a time machine*, we're stuck with the records we have -- and the question is, how good a reconstruction is possible. As with the Antarctic temperature data sets of Steig v. O'Donnell fame, it's possible to make some conclusions despite limited data. The key is not to underestimate the uncertainties in the resultant analysis.

*Even if I did have a time machine, improving the quality of historical temperature measurements is not the first use I would make of it.