date: Thu, 26 Feb 2004 11:56:46 -0000
from: "Rob Wilson" <rjwilson_dendro@blueyonder.co.uk>
subject: Fw: When Jones Meets MSU
to: <K.briffa@uea.ac.uk>

   Hi Keith,

   I am on this climate sceptics e-mail group that you might have heard of. Please don't
   groan. I am only signed on so that I can sit on the fence and see both sides of the
   argument.



   The e-mail below just came through and it is very interesting. I have already read about
   the difference between ground-based temperature data and the MSU satellite lower
   troposphere temperature data in the literature, but never really thought about it. However,
   having read this e-mail, I can't help think that IF there is a 'significant' problem with
   the ground-based data-set, that they suggest, it would partly explain the divergence
   problems between TR data and temperature data.



   Any comments?

   I did not send this e-mail to Phil as I do not know how testy he is on the subject.



   All the best

   Rob



   ----- Original Message -----
   From: [1]Willis Eschenbach
   To: [2]climatesceptics@yahoogroups.com
   Sent: Thursday, February 26, 2004 10:20 AM
   Subject: [Climate Sceptics] When Jones Meets MSU

   A work in progress, offered for your comments and criticisms.
   w.
     ______________________________________________________________________________________

   There has been a lot of talk, investigation, and speculation about the differences between
   the Jones (HadCRUT2 combined land and ocean) ground-based temperature data and the MSU
   satellite lower troposphere temperature data. So I thought I'd look at them to see what I
   could find. I looked at the 20 year period from 1979 to 1998.
    Unfortunately, what I found was that the Jones dataset is full of holes. Of the 2,592
   cells (5 degrees x 5 degrees covering the world) of the dataset, a full 60% of them are
   missing one month or more of data. Given this, I don't know how we can really compare the
   net change in global temperatures of the two sets -- the Jones dataset is trying to give a
   change in the global temperature using only 40% of the globe. Given that, here's the
   general picture.

                          [cid:005501c3fc5f$9bee3c40$a7671f3e@Desktop]
                 Fig. 1 - Comparison of satellite, GHCN, and balloon anomalies.

   From this, we can see that the MSU and balloon data agree closely, while the ground station
   data are quite different.
   Having looked at the general trends, here is my first comparison of average temperatures
   for 1979-1998 of the two datasets, a scatterplot of the two temperatures. I only used cells
   for which there were full Jones data for the entire period.

                          [cid:005601c3fc5f$9bee3c40$a7671f3e@Desktop]
         Fig. 2 - Scatterplot of the Jones and MSU full datasets, with linear trend line

   There were several surprises in this for me. The first was that the MSU temperatures are
   colder than the Jones temperatures, but not as much colder as I might have expected.
   The second was the width of the scatter of the data. I had expected a much closer
   correspondence between the Jones and the MSU data. In some places, for the same MSU
   temperature the Jones temperatures are as much as 12 degrees apart.
   The biggest surprise, though, was that a 1 degree difference in MSU temperature does not
   correspond to a 1 degree difference in Jones temperature, as I had expected. A 1 degree
   difference in MSU temperature corresponds to a 1.22 degree difference in Jones temperature.
   Since the linear trend does not reveal the true relationship between the datasets, I took a
   Gaussian average of the scatterplot. To do this, I converted from Cartesian to Polar
   coordinates (x,y to r,theta). Then I sorted the data by r, and took the gaussian average of
   the theta values. Finally, I converted back to Cartesian coordinates. This procedure is
   necessary to avoid distortion of the gaussian average, especially at the hottest and
   coldest ends of the data.

                          [cid:005701c3fc5f$9bee3c40$a7671f3e@Desktop]
           Fig. 3 - Scatterplot of the Jones and MSU data with Gaussian average (red)

   As you can see, the Gaussian average does not follow the trend line. What can we deduce
   from its wanderings?
   First, the temperate regions of the planet, where most of the people live, is where the
   Jones data is higher than the trend line. This supports the idea that local and urban
   warming have exaggerated the Jones temperature in these regions.
   Next, at the top end of the temperatures, the Gaussian average goes steeply upward. Since
   this means that we are getting much higher Jones temperatures than we would expect, this
   supports the idea that temperatures in hot city areas and areas like canyons in deserts are
   giving very high Jones temperatures. Here is a closeup of that region:

                          [cid:005801c3fc5f$9bee3c40$a7671f3e@Desktop]
         Fig. 3 - Scatterplot of the high temperature end of the Jones and MSU datasets

   You can see that how the Gaussian average is rising very steeply in the hottest parts of
   the world.
   DISCUSSION
   1)  The wide scatter of data is most likely due to differences in the Jones data, as the
   MSU data is internally more consistent. There are no areas of the world for which the MSU
   are very different from the surrounding area, but there are in the Jones data. The MSU data
   "flows" from area to area while the Jones temperature has discontinuities, especially over
   the oceans as can be seen below. I have used the baseline average datasets for both the MSU
   and Jones data (although it is not clear to me how the Jones average dataset was
   constructed given that 60% of the cells have missing data).

    [cid:005901c3fc5f$9bee3c40$a7671f3e@Desktop] [cid:005a01c3fc5f$9bee3c40$a7671f3e@Desktop]

   You can see that the MSU dataset is cooler, going down to -71 degrees and peaking out at
   about +9 degrees. It is also much more even, as we would expect given that it is not
   subject to local heating as is the ground record.
   2)  The signature of urban and local warming can be seen in two ways. One is that the
   temperatures in the areas above freezing (273 degrees K), where trees, houses and roads are
   more prevalent, are higher than the trend; on the other hand, at temperatures below
   freezing they are lower than the trend. Here is Jones temperature vs. deviation from the
   trendline:

                          [cid:005b01c3fc5f$9bee3c40$a7671f3e@Desktop]

   The other is that the data are "clumped" along the upper edge of the range, with few
   outliers, and the bottom of the range has a number of outliers. As there are few local
   conditions which lower local temperature and many conditions which raise it (trees,
   windbreaks, buildings, paving, etc.), this suggests that many of the Jones temperatures are
   reading higher than they should.
   3.  Given the slope of the trendline, we would expect that temperature changes in the Jones
   dataset should be larger than those in the MSU dataset. This seems to be the case -- for
   the 21 year period shown in Figure 1, the annual Jones delta t / annual MSU delta t is 1.25
   +/- 0.56 (1 std. dev.). However, this does not explain the rise in the Jones temperature
   compared to the MSU temperature over the 1980 - 2001 period.
     ______________________________________________________________________________________

   Yahoo! Groups Links
     * To visit your group on the web, go to:
       [3]http://groups.yahoo.com/group/climatesceptics/

     * To unsubscribe from this group, send an email to:
       [4]climatesceptics-unsubscribe@yahoogroups.com

     * Your use of Yahoo! Groups is subject to the [5]Yahoo! Terms of Service.

   Embedded Content: image.jpg: 00000001,38c1c2c5,00000000,00000000 Embedded Content:
   image1.jpg: 00000001,700521da,00000000,00000000 Embedded Content: image2.jpg:
   00000001,27488102,00000000,00000000 Embedded Content: image3.jpg:
   00000001,5e8be017,00000000,00000000 Embedded Content: image4.jpg:
   00000001,15cf3f3f,00000000,00000000 Embedded Content: image5.jpg:
   00000001,38561bca,00000000,00000000 Embedded Content: image6.jpg:
   00000001,6f997adf,00000000,00000000

