UAH September 1997 to August 2009:
-0.00421516884611802
Model Minimum Trend Values for 144 month periods in A1B runs, Jan 2001 to Dec 2020:
BCCR BCM2.0
0.003878423
CGCM3.1 (T47)
0.000822331
CGCM3.1 (T63)
0.008297988
CNRM CM3
-0.002661171
CSIRO Mk3.0
-0.004071267
CSIRO Mk3.5
0.002891261
GFDL CM2.0
0.010607289
GFDL CM2.1
-0.009373989
GISS AOM
0.010586626
GISS EH
3.72635E-05
GISS ER
0.006881269
FGOALS g1.0
-0.008261236
INGV ECHAM4
-0.000885811
INM CM3.0
0.019439775
IPSL CM4
0.000612523
MIROC3.2 (hires)
0.017552506
MIROC3.2 (medres)
0.00368745
ECHO G
-0.009114739
ECHAM5/ MPI-OM
-0.008325985
MRI CGCM 2.3.2
-0.006205251
CCSM3.0
0.012407194
PCM
-0.008955884
UKMO HadCM3
0.012840406
UKMO HadGEM1
0.002333876
As you can see, I’ve highlighted the models which don’t have any 12 year periods of negative trends at all. That’s fifteen of 24, or 62.5%
Pretty bad I’d say. This is just a preliminary test, and more formal tests are on the way (as soon as I master R) and I should emphasize that I don’t know if these 15 models are more sensitive/warm more in the long run than the rest. But note the influential names.
Hypothesis Testing 
September 7, 2009 at 8:01 pm |
Some of these models may not be running as warm when the numbers are crunched relative to the actual coverage of observational data. GISS interpolates over the Arctic where warming should be the strongest. The other two surface data sets don’t. When comparing model data to observational data, it may be necessary to mask out the Arctic and significant portions of the Southern Hemisphere. I’ve had this comparison on my list for sometime, but became bogged down with other things (like the far more difficult task of generating synthetic MSU data). When you become more comfortable with R, it may be worth your time to download the tas temperature data from PCMDI. All the data (20C3M/SRESA1B) is at most 10 GB or so.
September 7, 2009 at 9:28 pm |
That’s a good point and one of the subtle difficulties comparing models with data. This is particularly tricky with radiosonde data, where the coverage is pretty poor all around. I suspect that the difference in the trend distributions wouldn’t be huge though. But maybe.
September 13, 2009 at 3:35 pm |
Regarding the 24 climate models… Do all have sunspots modeled correctly? Do those that show a negative trend have few sunspots (2009 to 2015) and those that show a positive trend have a constant # of sunspots each year?
September 14, 2009 at 10:27 pm |
David-I’m not sure what you mean. In twentieth century runs, solar irradiance is used as an input. However, I don’t believe it varies in the projections, but is instead held constant or something (I’m NOT sure about this, though). BUT that would hardly matter. The reason is because the solar irradiance variation is very small, and so in models (not in reality) the solar cycle signal is tiny compared to the variability that is in there inherently.
September 14, 2009 at 11:54 pm |
Thanks… sort of what I thought. But I was talking more about sunspots than solar irradiance. Here is what Dr. David H. Hathaway
(NASA Marshall Space Flight Center) said on Aug. 5, 2009.
“The 0.1% change in the Total Solar Irradiance seen over the last three solar cycles only produces a 0.1° C temperature change in climate models. However, the Sun seems to have a bigger impact. Two other mechanisms (besides direct forcing by the Total Solar Irradiance variations) are under study: 1) solar ultraviolet and extreme ultraviolet variability and 2) Cosmic Ray modulation on cloud cover. “
September 15, 2009 at 5:38 pm |
David: The Sunspot number is not know ahead of time and is difficult to predict, so I imagine projections don’t include it. However, if you allow me a quick comment on Hathaway’s comment:
To my knowledge the solar cycle is not even .1 C in models (in fact, that’s the observed effect. However it is true that “1) solar ultraviolet and extreme ultraviolet variability and 2) Cosmic Ray modulation on cloud cover. ” Are not included in most models. I think there are a few which include the ultraviolet effects, though.