OF the two points, the only one that really matters is that Harry contains spliced data from a period when it did not exist; the Gill thing is the sort of interesting detail that adds color but doesn’t really matter.

I noticed that GISS and Steig used the same data (See my first post), so I was interested in why Mannian RegEM yielded different results than GISTEMP – the sort of thing that you’d think that they would report in PRL. I noticed that Harry was a big outlier in the West Antarctic and decided to look at it. I have several GISS versions and compared versions – I’m not sure why. I’ve got a lot of relatively well-organized data and can do this sort of thing in a couple of minutes. A huge difference popped out.

I’ve done enough deconstruction of Team and Mannian stuff that I don’t overlook the possibility of really goofy things like accidentally splicing unrelated series – this is not the first time. If you’re fresh to this corpus of material, it wouldn’t occur to you to be alert for this sort of thing.

So I looked for info on Harry – I googled “Harry AWS” and got to the Wisconsin site. It showed that HArry didn;t exist prior to 1994. So I was pretty sure on Sunday afternoon that they’d done something goofy and asked readers to look for a provenance of Harry data before it existed. I posted up a script for scraping data from the Wisc website as well.

I went back just before bed and scraped all the files, the number of stations increased and varied by year, so the scraping was a little fiddly. I then looked through Jan-Feb values in a couple of years before Harry existed and found a match with Gill. I then downloaded all the Gill series and found that all the imaginary Harry data came from Gill.

There was another way. Gill proved to be one series above Harry in the BAS list. Tim C noticed that the two BAS series were identical.

I don’t know what Gavin did in his “independent” identification of the Gill provenance. However, as noted above, for BAS purposes, the provenance of the bad data doesn’t matter; only that the data is bad – something that Gavin obviously didn’t “independently” notice.

As observed elsewhere, we later learned that Monaghan et al 2008, cited by Steig, observed that Harry data was “suspicious” but apparently took no steps to investigate or pin down these suspicions.

I have not read all of the comments here or at RC and am hoping that someone will provide a little more detail on how the problem was detected and ultimately identified. The detail I am seeking is not so much who and when but how. In lieu of reading every comment I did some searching and noticed that “Tim C” identified the problem but did not find any associated code.

I must admit that I have spent several hours playing with GISTEMP step0 and did not catch the error and I would like to enhance my emulation by screening the source data for errors of this nature.

Thanks

Another thing…I keep a backyard weather station. With no formal training other than my own hobby reading, I know to go out after it snows to clear off the temperature sensor shelter to get the most accurate low readings. Maybe these guys also didn’t realize that anything on the surface of the ice in Antarctica also gets BURIED over time.

Let’s start a movement to see that they don’t.