RSS, UAH, GISS Comparison
Posted by Jeff Id on January 9, 2009
I’ve been reading papers for the last couple of days on satellite data and corrections. It’s pretty amazing what’s been done and with my improved skills in climatology they read a hell of a lot faster. Still it takes quite a bit of time to learn and the conclusions you can draw are incremental.
A lot of my readers have already dug into the papers at the NOAA about sat temp info. What isn’t easily found in the papers is the simple stuff which has so much meaning. Today I have a graph and a link, just a simple graph of the different data sources and a link which includes several important papers. The first time I read these papers was before I started blogging, they make more sense to me now but I got the gist back then.
NCDC Link here– you have to copy and paste it yourself because the level of censorship here seems to prevent me from adding links.
http://www.ncdc.noaa.gov/oa/climate/research/msu.html
And here’s a graph which I plotted from the three metrics GISS, UAH and RSS, with a 6 month filter.
The blue line is the difference between the RSS – UAH 6 month filtered data, lines are from LS fits. From my last post we know that the red trend and the black trend being similar in slope is not a confirmation of either dataset. As I understand it the RSS and UAH satellite data measures should have slopes greater than the GISS ground stations.
If you do take time to read the RSS and UAH papers at the link above you will learn that much of the slope difference between UAH and RSS comes from NOAA-9 satellite. One of a dozen in a series of short lived satellites which combine to make the temperature series. I found the blue line (difference between the two) interesing in that there is a step a 1992, before that time the difference trend is quite flat and after that time the difference trend is flat. I have been looking into what happened at this point for a future post.
If both metrics have the same trend before and after the discontinuity and if the discontinuity can be explained by satellite transitions, then the trend of before and the trend after the discontinuity should be able to identify the more accurate trend dataset.
DeWitt Payne said
It almost looks like two signals at the same frequency, but with the phase shifting with time. Ignoring the step change, they were in phase in the early 1990’s but were out of phase earlier and later. At a guess, it probably has something to do with converting the actual temperatures to anomalies.
Eric Adler said
As Tamino mentioned, there is a cyclical difference in addition to a difference in slope.
The cyclical difference is annual, strongest since 2000, and could be related to a seasonal difference observed in the Champ paper I linked on the previous thread. They found that the difference was in agreement was seasonal different in summer than in the other seasons.
One hypothesis that fits with this data is that the temperature scaling versus observed signal is off, and the difference is larger in one season than the other. THey didn’t say whether it was related to seasons in one or both hemispheres. It would be interesting to look at the raw data or a 1 month filter, instead of a 6 month filter in order to see precisely at what point in the year the differences show up the strongest.
John F. Pittman said
There are two annual upward bumps in temperature. One for the NH and one for the SH. There is a post somewhere on the net (I am not too skilled at searches) that discusses this and how to correct for it. GISS and Hadcrut correct for monthly such that each month’s anomoly is with reference to that month’s average of the base period.
Mike said
Hi Jeff,
A little OT, but I’ve been exploring GISS for my area of Europe, and it’s shocking.
Centred on the only station in the Netherlands, here’s the data that GISS has:
http://data.giss.nasa.gov/cgi-bin/gistemp/findstation.py?lat=52.1&lon=5.18&datatype=gistemp&data_set=0
km (*) De Bilt 52.1 N 5.2 E 633062600000 33,000 1880 – 1991
0 km (*) De Bilt 52.1 N 5.2 E 633062600001 33,000 1949 – 1990
0 km (*) De Bilt 52.1 N 5.2 E 633062600002 33,000 1961 – 1980
0 km (*) De Bilt 52.1 N 5.2 E 633062600003 33,000 1987 – 2008
146 km (*) Essen 51.4 N 7.0 E 617104100000 7,452,000 1951 – 1991
146 km (*) Essen 51.4 N 7.0 E 617104100001 7,452,000 1987 – 2008
156 km (*) Uccle 50.8 N 4.3 E 606064470000 1,055,000 1880 – 1980
156 km (*) Uccle 50.8 N 4.3 E 606064470001 1,055,000 1949 – 1990
156 km (*) Uccle 50.8 N 4.3 E 606064470002 1,055,000 1951 – 1991
156 km (*) Uccle 50.8 N 4.3 E 606064470003 1,055,000 1987 – 2008
160 km (*) Aachen 50.8 N 6.1 E 617105010000 242,000 1880 – 1893
191 km (*) Emden-Hafen W.Germany 53.3 N 7.2 E 617102030000 54,000 1880 – 1991
191 km (*) Emden-Hafen W.Germany 53.3 N 7.2 E 617102030001 54,000 1987 – 1997
215 km (*) Guetersloh 51.9 N 8.3 E 617103200000 77,000 1880 – 1920
223 km (*) Lille 50.6 N 3.1 E 615070150000 171,000 1880 – 1897
236 km (*) Clervaux 50.0 N 6.0 E 629065850000 rural area 1971 – 1980
243 km (*) Gorleston 52.6 N 1.7 E 651034960010 50,000 1931 – 1975
256 km (*) Bentwaters 52.1 N 1.4 E 651035900010 rural area 1951 – 1967
258 km (*) Bitburg 50.0 N 6.6 E 617106070010 rural area 1952 – 1967
258 km (*) Woodbridge 52.1 N 1.4 E 651036960020 rural area 1954 – 1967
259 km (*) Spangdahlem 50.0 N 6.7 E 617106070020 rural area 1953 – 1967
266 km (*) Felixstowe 52.0 N 1.3 E 651036960010 19,000 1931 – 1960
271 km (*) Echternach 49.8 N 6.5 E 629065970010 rural area 1971 – 1980
280 km (*) Hunsruck 50.0 N 7.3 E 617106160010 rural area 1953 – 1967
282 km (*) Trier-Petrisb 49.8 N 6.7 E 617106090000 100,000 1880 – 1981
282 km (*) Trier-Petrisb 49.8 N 6.7 E 617106090001 100,000 1971 – 1990
282 km (*) Trier-Petrisb 49.8 N 6.7 E 617106090002 100,000 1993 – 2008
284 km (*) Luexmbourg/Town 49.6 N 6.1 E 629065890010 78,000 1961 – 1980
285 km (*) Luxembourg/ 49.6 N 6.2 E 629065900000 78,000 1951 – 1991
285 km (*) Luxembourg/ 49.6 N 6.2 E 629065900001 78,000 1951 – 1990
285 km (*) Luxembourg/ 49.6 N 6.2 E 629065900002 78,000 1987 – 2008
293 km (*) Helgoland 54.2 N 7.9 E 617101200000 rural area 1977 – 1990
298 km (*) Laon 49.6 N 3.5 E 615070610010 27,000 1952 – 1966
305 km (*) Geisenheim 50.0 N 8.0 E 617106280000 rural area 1951 – 1991
308 km (*) Kassel 51.3 N 9.4 E 617104380000 206,000 1880 – 1991
308 km (*) Kassel 51.3 N 9.4 E 617104380001 206,000 1987 – 2002
309 km (*) Hannover 52.5 N 9.7 E 617103380000 553,000 1880 – 1991
309 km (*) Hannover 52.5 N 9.7 E 617103380001 553,000 1987 – 2008
310 km (*) Sculthorpe 52.9 N 0.8 E 651034820030 rural area 1950 – 1964
316 km (*) Lakenheath 52.4 N 0.6 E 651034820020 rural area 1949 – 1967
317 km (*) Wiesbaden 50.0 N 8.3 E 617106330010 251,000 1949 – 1967
321 km (*) Wethersfield 52.0 N 0.5 E 651036830010 rural area 1952 – 1967
322 km (*) Mildenhall 52.4 N 0.5 E 651034820010 rural area 1950 – 1967
330 km (*) Frankfurt/ 50.0 N 8.6 E 617106370000 636,000 1971 – 1990
330 km (*) Frankfurt/ 50.0 N 8.6 E 617106370001 636,000 1961 – 1981
332 km (*) Frankfurt A Main W.Germany 50.1 N 8.7 E 617106400010 636,000 1880 – 1961
342 km (*) Ramstein 49.4 N 7.6 E 617106140010 101,000 1952 – 1967
Note how few go to the present day!
Now, looking at De Bilt itself, we get:
Raw and Adjusted : http://img440.imageshack.us/my.php?image=debiltyj4.jpg
Adjustments (adj-raw) : http://img249.imageshack.us/my.php?image=debiltadjustmentsic1.gif
metadata are:
OK, the metadata for de bilt are:
pre-16/05/1950 – thermograph in large pagoda cabin 2.2m over a mown field
17/05/1950-28/06/1961 – thermograph in stevenson screen 2.2m above mown field (note downward step change in data)
29/06/1961-25/06/1993 – resistance measurement in stevenson screen 1.5m above mown field
26/03/1993-present day – electric sensor in dish cabin 1.5m above mown field (note step up).
So, how did the Hansomatic do?
-1950 step not adjusted
-1961 change is not visible in data
-1993 step not adjusted
-artificial trend introduced
HOW can this be accepted as “science”? Just look at the adjustment graph! How can this be justified? It ignores the step changes due to instrumentation changes and then adds a significant trend itself! And, because this is the only long(ish)-record station within 100(ish) miles, it affects a huge amount of the record. The next nearest station to the South, Uccle (actually spelt Ukkel now…) is in a suburb of Brussels.
This needs to be looked at by the big boys (yourself, Anthony Watts & Steve Mc) who I can’t contact directly…
page48 said
Hi Jeff,
How’s the other side of the world these days.
“NCDC Link here– you have to copy and paste it yourself because the level of censorship here seems to prevent me from adding links.”
Let’s all hope this kind of thing doesn’t happen in the US!
Take care,
Page48
BarryW said
OT but I wanted to pass this by someone.
I was thinking about the super el nino in 1998 and I noticed that the trend before it occurred seemed very different than that afterwards. I did OLS trend lines for the UAH data up to the start of the peak and after the end and got the plot in this graph. The trend before is 0.003 deg / year (purple) and after (green) is 0.011 while the overall is 0.012 (red). The red solid line is what I ignored (I used the zero crossing points to decide where to cut). What it looks like to me is that the el nino caused some sort of damped oscillation in the temp data. Notice that the extended purple line starts to intersect with the data after the event as if the pulse is decaying back to the .003 trend line. My guess is that the temps will go back to the trend line or “bounce” once more before they do.