Re: Why Would Anyone Trust NASA's Climate Data Now?

On Aug 20, 1:50 pm, Sam Wormley <sworml...@xxxxxxxxx> wrote:
d....@xxxxxxxxxxx wrote:
August 20, 2007

Why Would Anyone Trust NASA's Climate Data Now?

Because the data comes from working scientists trying to get is as
accurately as they can.

Um. Leaving aside the "proof by authority" involved in this.

What is your basis for this claim, Sam? In all seriousness, the
data at hand tends to suggest that this is not the case. There may
be some good scientists in the pile. But there are clearly also a
significant number of political hacks who are trying to drum up as
much hysteria as possible in order to pad grant proposals.

If I tried to submit reports of this quality to the regulator in my
industry, I would be looking for another job within days. And it
would matter not a bit whether my answers were right or not.

Where are the QA reports on the software that was used to do
the adjustments to the raw data? Who verified and validated that
software? Where is the source code for that software?

Where are the qualification reports for the sensors and data
files that contain the raw data? Exactly how important are the
reports at http://www.surfacestations.com/ ? Why has nobody
done this kind of survey before? Where are the equivalent
surveys for the weather stations used for the rest of the globe?
Where is the data indicating where those weather stations are?

It's quite possible that getting this data wouldn't change anything.
It's even possible that it would produce a picture that had the
global temperature rising faster than has been reported.

But it might, as the data in the US now indicates, show no
particular trend over the 20th century.

The point is, until that stuff is available, the entire enterprise
is suspicious.

I've had enough experience with this sort of thing to know how it
goes. Some grad student someplace whomped together the
program to combine the data from the weather stations. He did
it in four or five coffee-addled days. The code is not just spaghetti,
but an entire Italian meal. And it won't have comments. And it
won't have a user manual nor a developer's manual. Cripes, there
might not even *be* source code anymore if it was originally
written before Jan. 1, 2000. Then the original code would get
modified by a series of other grad students, each with a different
style, a different understanding of the code's operation, and a
different understanding of the data sources. And all with no code
comments, no user manual beyond possibly a three line "help"
file, and no developer manual.

I've seen kaka of this nature so many times. Academics get taught
how to do these calculations, but not how to check them. They
show up in my office, fresh out of school, and they think they know.
Then I look at their source code. Then I ask for their documentation
and it's "Documentation? What for?" Then when I insist they read
some remedial "how to program" books, they get upset. And when
their first two years in the industry consists entirely of doing QA,
they mope and whine and complain.

Sigh. So, Sam, what's your evidence that they are trying to get it
right? What's your evidence they know what "right" means?

Keep in mind what "undocumented feature" is slang for.
Socks

.

Quantcast