Interesting how eager people are to believe Eschenbach without any "auditing".
Anyway, check this comment and this comment in the McArdle thread before jumping to conclusions.
"the homogenization process is a fully automatic statistical treatment for 7000 stations - it has no biases for higher or lower temperatures. The homogenization is based on the records of the nearest 5 stations, which can have a higher or lower temperature so treatment is not biased. Darwin 0 has a higher adjustment due to higher temperature records in the neighboring stations. In the cases where the "neighbors" of the 7000 stations have lower temperatures, there is an automatic downwards adjustment.
"In fact, handpicking adjustments for individual stations, which is what Eschenbach suggests for Darwin 0 in 1941, would be a method far more prone to temptations to bias the result desired."
Raw observational data series must adjusted when instrumentation changes. The step change shown at your link is obviously an artifact. Removing such artifacts (in whichever direction) is a big part of the problem of getting a temperature record from imperfect surface observations. This just shows people doing their jobs.
McArdle knows nothing about these matters and doesn't seem to have consulted anyone who does. Have you?
Normally I don't respond to people who use "Thhpt!" in their argumentation. Do you think that helps?
Parent is stunningly confused. We actually greatly remediated the acid rain and ozone hole problems by regulation.
Michael Tobis, Ph.D.
Atmospheric and Oceanic Sciences, U Wisconsin-Madison 1996
+1
Michael Tobis,
Ph.D. Atmospheric and Oceanic Sciences 1996 at U Wisconsin-Madison
Part of the problem is that every government has a hand in it, and that since people don't like to pay taxes, many of the governments involved fund the research by reselling the data to private forecast entities. If they open source it (which I fully agree that they should) that revenue stream dries up, a number of businesses are threatened, and your taxes go up. In any case, AIUI the Climate Research Unit was under contractual obligation to the various contributory agencies (in MANY countries) NOT to reveal the information, so all the FOI requests amounted to nothing more than harassment.
In the end, robust code is more expensive than quick hacks. The purloined code has quick hack flavor, no doubt, and in a few places shows somebody who is stuck in a Fortran mentality where a proper scripting language would have been far superior. Whether it was suitable for purpose for said code to be a quick hack is not something I see being discussed anywhere.
Let's stipulate for argument that it was not at an appropriate level of rigor for the task and consider what it means. What it doesn't mean is dishonesty.
I know lots of scientific programmers who find the idea of having to learn Perl or Python terrifying. Pity, but really these are untrained programmers though trained scientists. Anyway, acquiring trained programmers and training them in science or acquiring trained scientists and training them in programming costs a lot of money, and despite what you may have heard, money is very tight in climate science. That said, riskinbg doing things wrong because it's cheaper doesn't make a lot of sense. In other words, I agree with the sentiments expressed here for the most part but readers should understand that most of them cannot be achieved on a shoestring.
The loss of credibility in science described in the leader is realistic and not without foundation. Science has problems which need to be addressed. An accusatory and adversarial stance, though, will simply throw the baby out with the bathwater. And the CO2 continues to pile up, with consequences that we can anticipate may be very serious.
Anyway I find it odd that the parent article refers to "our" government. Presumably parent author is British?
Michael Tobis,
Ph.D. Atmospheric and Oceanic Sciences 1996 at U Wisconsin-Madison
Texas is wonderful in many ways but wondrous strange as well.
Old programmers never die, they just hit account block limit.