TGS4 said:
the outcomes of the GCMs predictions diverge from reality. Full stop - they made predictions and their predictions failed
This overly simplistic, under-informed statement is common amongst skeptics but needs unpacking. When do climate models “fail”? If climate models “fail”, does the theory fall apart?
----When Do Climate Models “Fail”----
This is a bit of a subjective question. An easy answer would be when observed temperature trends fall outside the predicted range of climate model runs. 1)Which temperature data set? 2) Which confidence range does it need to lie outside? 3) Which climate model runs? 4) For how long do they have to lie outside the confidence range? The answers are usually : 1) average of GISS, NCDC and CRU, 2)95%, and 3) average of all CMIP3 runs. 4) ranges from 1 day for “skeptics” to multiple years to a decade, let’s say 3 years.
So, let’s look to see if
the average of the major temperature data sets lies outside the 95% confidence interval for the average of all model runs for at least 3 years. From AR5 SPM, Figure 1.4 (Dark Blue = GISS, “Mustard” = NOAA, Bright Green = CRU)
[image
]
“But…but…the leaked draft version of that graph is the right one! The IPCC tried to cover it up with the other one! You’re cherry-picking!” said the “Skeptics”.
Let’s make sure we understand why it was changed –
it was wrong. In the leaked draft version, all series (models and observations) were aligned at 1990, a single year. Observations inherently include the random year-to-year fluctuations, whereas models do not as they are the average of numerous runs off numerous models. So to aligned to a single year, a hot one at that, makes the projections of the models appear hotter than they actually are. In other words, they took the projected trend from the models and attached it to the 1990 data point (which was hotter than the trend leading up to that). What should happen, and what did happen in the final report, is the projected trend line should be attached to the
trend line leading up to 1990. That’s not “masking the truth”, that’s correcting a statistical error. If you’re curious, more info can be
found here.
This is what the draft version did:
[image
]
And here is the same projected trends but, correctly, applied to the TREND leading up to 1990 (it looks almost identical to the IPCC final report):
[image
]
Or another way of looking at it, the trends since 1990:
[image
]
So,
NO, climate models have not “failed” to predict temperatures accurately.
----If Models “Fail” (which they haven’t), does the theory fall apart?----
We could stop here but let’s address the consolation argument, “well they may not have “failed” but they are certainly
trending away from observations”. So let’s revise are second major question to, “If Models momentarily diverge from observations, does the theory fall apart?”. Again, we need to unpack this statement – 1) What is causing the divergence? 2) Is that error a crucial aspect of the theory? 3) If yes, is the divergence long enough and/or drastic enough to point to a serious flaw in the theory? We will take each question one at a time
1) What is Causing the Divergence?
We come back to the “pause”. If that pesky short-term memory issue is coming up again, please re-read any of my 13 previous posts on the subject. The basic answer is that we are in an negative PDO where La Nina events dominate, creating a short-term decrease in surface warming. Models do not attempt to accurately match ENSO periods with reality as ENSO is stochastic. However, the PDO is also cyclical and is purely about the short term storage and release of heat in the oceans –
it cannot effect long-term trends.
Another stochastic element of climate that cannot be predicted is volcanoes. Again, volcanoes only have a short-term effect.
The last element that is difficult to predict is human actions. We appreciate this when discussing CO2 emissions and no one has an issue with their being different trends that represent different emission scenarios – we don’t expect the models to tell us what the emissions will be. However, few “skeptics” also realize this applies to other emissions such as aerosols. Recently, anthropogenic aerosol emissions have been higher than estimated, largely contributed to faster than expected growth in the developing world. More aerosols means less energy enters the troposphere which has caused incoming solar forcing to be less than estimated.
What do temperature trends vs model predictions look like when we account for these effects? The first image is the as-is model (CMIP5) vs observation trends (average of GISS, NOAA and CRU):
[image
]
Now when we account for the La Nina dominated period by inputting the actual ENSO values into climate models, we get the following:
[image
]
When we update the aerosol and volcano values to match the actual observed trends, solar forcing gets adjusted and the trend looks like this:
[image
]
2) Is the Error (ENSO, Aerosols) a Crucial Aspect of the Theory?
Some of you I’m sure are saying “well you’ve just tweaked the models to get what you wanted!”. That is incorrect. All that was changed was:
- ENSO was set to match observations
- Volcano and aerosol (solar) values were set to match observations.
No one, including climates models, can predict when ENSO and volcano events will happen. Both of these are short term and have no effect on long-term trends. Aerosol adjustments came primarily from updated anthropogenic emission numbers (they were higher than previous projections). Again, models cannot predict emission trends from either CO2 or aerosols, you have to tell them what you think they’ll be and they tell you what temperature that will lead to.
So all that we have “tweaked” are aspect that are not predictable and, in the case of ENSO and volcanoes, have only short term effects. We have not adjusted anything related to CO2 sensitivity, cloud feedbacks, albedo effects, water vapour feedbacks, etc.
The manner in which the models calculate all the interactions between CO2 and the environment, the various forcings and feedbacks and the way energy is moved around the system remain untouched – these are the “crucial” elements of the theory and the accuracy of climate models. Therefore, without making any adjustments to any crucial elements of our understanding of climate science, models are shown to be in excellent agreement with observations.
Beyond providing a definite NO to the question “Is the Error a Crucial Aspect of the Theory?”, we can also conclude that the crucial elements of the CO2 theory, when modeled, do an excellent job predicted temperature trends. The estimated effect of feedbacks and forcings are, firstly, very accurate and, secondly, well represented within climate models.
3) If Yes, Is the Divergence Long or Drastic Enough to Point to a Serious Flaw in the Theory?
Given that the answer is a clear NO to 2), this question doesn’t apply. Furthermore, reviewing the data actually says that the theory is even stronger than we thought.
I’d also like to illustrate just how spurious of an argument the “pause” is by showing the temperature trends of various sets vs a true pause (a flat line) vs a continued warming trend from 1997 on.
[image
]
[image
]
That’s your “pause” for ya.
![[ponder]](/data/assets/smilies/ponder.gif "[ponder] [ponder]")
