[asa] Successful reproduction of Eschenbach's results - and a possible explanation?

From: Iain Strachan <igd.strachan@gmail.com>
Date: Sat Dec 12 2009 - 12:39:33 EST

I received yesterday an email out of the blue from Robert Wittig, who had
read of my unsuccessful attempts to reproduce Eschenbach's results while
browsing the ASA archive.

He showed me how to reproduce them; apparently the "adjustments" I had
downoaded were the GISS adjustments and not the GCHN adjustments. He
enclosed an excel spreadsheet, where he had reproduced the graphs.

I have always gone into this with an open mind and desire to get honestly at
the truth. It appears Eschenbach was not fabricating his results (though
his interpretation of those results is of course up for debate). I have
independently reproduced Eschenbach's results (and Wittig's) by downloading
the raw data, and plotting them out using Matlab. The following five links
give the plots I obtained:

http://picasaweb.google.com/IGD.Strachan/NOAADarwin#5414394772197509010

shows the "Raw" data for all five stations - and as can be seen they are all
in close agreement.

http://picasaweb.google.com/IGD.Strachan/NOAADarwin#5414394772652507794

shows the "Adjusted" data - adjustments only exist for stations 0 1 and 2.
It is curious and does to me require some explanation as to why these are
all quite different, whereas in the raw data they are in agreement.

http://picasaweb.google.com/IGD.Strachan/NOAADarwin#5414394774309127298

This is the most controversial one and appears also in Eschenbach's
article. It shows the raw and adjusted data for Station zero (of "smoking
gun" fame). As can be seen there are step adjustments in 1979,1962,1951 and
1930. The first two of these appear to be exactly 2/3rds of a degree ( the
data-tips show 0.6667 and 1.3333). The other two adjustments are of a
similar order. This looks suspicious because there appears to be a rising
trend from 1940 onwards in the adjusted data, in line with the adjustments.
This is what Eschenbach calls a clumsy effort to create a trend where none
was present.

From the above link you can navigate, via arrows to the other to pictures in
the album which show the similar traces from stations 1 and 2.

I obtained the datasets from the following website (provided in a link in
the comments to Eschenbach's article, and which Robert Wittig has helpfully
provided for me:

http://www1.ncdc.noaa.gov/pub/data/ghcn/v2/

The relevant datasets to download are v2.mean.z, v2.mean_adj.z,
v2.temperature.inc, and v2.temperature.readme files

The compressed .Z files can be unpacked using something like WinZip (I used
PowerArchiver). The readme file shows how to interpret the lines of data,
and the v2.temperature.inc file gives the station code to search for.

Robert used an Access database to extract the data for Darwin; I wrote a
Matlab script to parse the data, and hence extracted it independently of
him. I then produced the plots using Matlab. In the datasets, the
temperature value of -9999 is used to indicate missing data. I computed the
yearly averages (each line has 12 readings) by averaging the values that
were not marked -9999.

My explanation:

Clearly this cries out for explanation. Eschenbach suggests that this was a
crude attempt to massage the data to manufacture a trend.

However, I don't think there is foul play here, but instead I think there is
an error in the dataset - I believe the most logical explanation is that the
adjusted and raw data have been mistakenly put into the wrong datasets, and
that the REAL adjusted data should be in v2_mean.z and not v2_mean_adj.z,
which was where I extracted it.

This means that the true raw data (which is in the "adjusted" set) shows
disagreements between the stations for 0 1 and 2 which could be down to
anything (siting, covering on the instrument etc), which after correction
agree with all the rest of them.

This would make sense, because the original dataset I pulled off the GISS
website stated:

"Raw GHCN data + USHCN corrections" (5 stations)

so I'm guessing that the USHCN corrections are the ones we see in
Eschenbach's diagram, and these are NOT the "inhomogeneity" adjustments.

Note that the "Raw GHCN data + USHCN corrections" graph ( see mine at:
http://picasaweb.google.com/IGD.Strachan/Darwin#5413595233568220978 ) agrees
with the "Raw" data from the NOAA website.

Hence, I suggest that the data for stations 0 1 and 2 raw and adjusted on
the NOAA website have been mistakenly put in the wrong archives. This then
reverses the conclusion that Eschenbach drew. It would appear to me that
before the adjustments, the discrepancies between the stations showed a
trend that wasn't real, and the adjustments removed the false rising trend
for this station.

Now... if I'm totally honest, I don't know if my explanation is the right
one. The alternative one is that the data raw and adjusted are as they are,
and that it does look suspiciously like the data has been massaged to make a
trend. I think the experts on the list had better judge for themselves. I
am quite suspicious of those figures 0.6667 and 1.3333, two successive
adjustments of exactly 2/3 of a degree. That didn't look like it was
computed by a machine because it wouldn't be so close to a simple ratio (not
twice over).

I'm awaiting the comments from Rich Blinne with great interest.

Many thanks to Robert Wittig (cc'd in on this) for providing me the relevant
explanation.

Regards,
Iain

-- 
-----------
Non timeo sed caveo
To unsubscribe, send a message to majordomo@calvin.edu with
"unsubscribe asa" (no quotes) as the body of the message.
Received on Sat, 12 Dec 2009 17:39:33 +0000

This archive was generated by hypermail 2.1.8 : Sat Dec 12 2009 - 12:40:24 EST