Remember the evidence is overwhelming, and deniers deny the evidence. But in Oct 2012, two atmospheric scientists were reporting, yet again, the models are wrong. Twenty years after we started looking for the fingerprint of the amplification required to make the CO2 theory of global warming work, it still isn’t there. Forgive me for harping on. It’s still The Most Major Flaw in climate models.
Never heard of “the Hot Spot”? See the first post on the hot spot argument. The models are wrong (but only by 400%!) See how climate scientists admit it’s important and missing. See how they stoop to changing color scales on graphs to pretend they’ve found it and ignore 28 million weather balloons. Or just read the summary with scientific references I wrote in May.
Background: The assumption that was wrong
Researchers made an assumption that water vapor would amplify the direct warming of extra CO2 from a small harmless amount to a large catastrophe. They started with the theory that relative humidity would stay constant in a warmer world and the thicker layer of water vapor would warm the world even more. Greenhouses gases in this instance means mainly water vapor; the assumption is that extra water vapor is heating up the upper troposphere (both by displacing colder drier air, and by condensing and releasing the latent heat absorbed in evaporation). It was predicted by James Hansen in 1984, is repeated by all the climate models and by the IPCC in AR4.
The graphs from this recent paper show once again that the models are wrong, the observations lie far outside most of the models. No matter how many ways they reassess the same data and rejig the models, they aren’t getting a match.
The problem in a nutshell: If they drop the assumptions about amplification by upper tropospheric water vapor, the models will match reality but they won’t predict a crisis.
The weather balloons produced the dramatic images showing just how “missing” the hot spot is. But people have been searching with satellites too. The satellites don’t have the vertical resolution of the weather balloons, because they measure large thick bands of sky. So while researchers won’t find the “hot spot” exactly with a satellite, they hope to find the right ratio of trends in the upper atmosphere compared to trends in lower bands. (More cynically, one might say they hope to get a vague fit to the models by using the less precise and more fuzzy satellite data rather than the higher resolution data from the weather balloons.)
Does this topic matter? These climate researchers think so:
Given the importance of both models and observations, it will be important to continue to investigate this discrepancy between models and observations.The representation of upper tropospheric warming in models is important to climate sensitivity and thus future projections of anthropogenic global warming.
In other words, the models can’t calculate climate sensitivity and future temperatures without getting this right. This is central.
In the words of the researchers: “… most atmospheric models exhibit excessive tropical upper tropospheric warming”.
In this graph, the vertical line up from 1.05 and 1.1 are the satellite measured ratios, show where the yellow bars (the model predictions) ought to be if they matched the satellites.
Figure 3. Histogram of the ratio of the T24 trend to the TLT trend over 1981–2008 from AMIP ensemble members in the tropics (20S–20N). The T24 to TLT trend ratios for RSS and UAH are
shown for comparison. The T24/TLT trend ratios under the histogram bins represent the bin center values.
The black circles and crosses (below) are supposed to fall around the blue square and the red square. This is what 90% likely looks like if you are the UN, and you want more money.
…
Figure 4. Decadal versus interannual amplification of T24 to TLT from both AMIP and coupled GCM simulations and MSU observations in the tropics (20S–20N) between 1981 and 2005.
The decadal amplification is defined as the T24 trend divided by the TLT trend. The interannual amplification is defined as the standard deviation of the de-trended monthly T24 anomaly time series divided by the standard deviation of the de-trended monthly TLT anomaly values. Each cross or circle represents the ensemble mean for each model. The mean of all models is given by the bold
symbols. Note that the MIROC-ESM-CHEM model is not contained in this plot as it has a relatively large decadal amplification value (table 1), likely related to biases after the Mt. Pinatubo eruption in 1991 (Watanabe et al 2011).
Curiously, these same authors had a run-in with Roy Spencer and John Christy earlier in 2012, when they published a paper suggesting that UAH ought be adjusted up to fit better with the models. See Spencer’s response at his site, (repeated on WUWT). This later paper came out in October. Hmm.
Again, those empty circles are supposed to be close to the red and blue squares.
Figure 5. Decadal amplification (as in figure 4) versus the annual mean T24 temperature over 1981–2008 for AMIP models. The relationship is statistically significant (95% confidence) and the r-value is 0.56. The annual mean T24 temperatures are also presented for RSS and UAH for reference. Note that much of the focus for MSU groups has been on relative changes and not on
absolute temperature calibration (e.g. Mears et al 2011).
—————————————————————————
AMIP models are better than GCM’s, but they are still wrong
The AMIP models use sea ice and sea surface temperature (SST) observations to constrain the workings of the models, so the evaporation off the oceans should at least be about right in those simulations, so naturally they are better than the GCMs — see the top band here, where the AMIP ones get it mostly right, and the bottom band where GCM’s get it wrong even though they have a wide “error band”.
Figure 1. Times series of TLT monthly temperature anomalies in the tropics (20S–20N) for AMIP GCM (top) and coupled GCM (bottom) simulations (black) and the average of RSS and UAH (red). The model spread is shaded. Many coupled GCM simulations only extended to 2005, while AMIP runs included here extended to 2008.
Atmospheric Model Intercomparison Project (AMIP).
Page 2 AMIP model simulations use observational records of sea ice and SST evolution (e.g. Hurrell et al 2008) as boundary conditions for atmospheric GCMs. Thus the AMIP model’s SSTs have the same variability and trends as observations. Previous studies have demonstrated that AMIP style runs can closely reproduce the observed tropical tropospheric temperature variations (Hurrell and Trenberth 1997). The use of AMIP models allows us to closely examine simulated changes of the vertical temperature structure in the tropical troposphere in GCMs using the observed SST evolution.
But even the AMIP models are still unable to get the results that match UAH especially. The dots in the first graph below should center around zero (which means ‘no difference’ with UAH). They are so far away that the error bars too, often end far from zero.
In this graph, the vertical line up from zero is where the dots (which are the model predictions) ought to be if they matched the satellites. The broad horizontal lines are the error margins.
REFERENCES:
Po-Chedley S. and Fu Q. (2012) Discrepancies in tropical upper tropospheric warming between atmospheric circulation models and satellites, Environ. Res. Lett. 7 044018 (http://iopscience.iop.org/1748-9326/7/4/044018) [Full text PDF (323 KB]
All Jo’s harping posts about the Hot Spot. 🙂
H/t to The HockeySchtick
Thanks to Richard Courtney.