January 16, 2018

Should you believe climate models? Part 1

After a Twitter exchange on the validity and quality of climate models, I was asked to briefly comment on the first chapter of the Nongovernmental International Panel on Climate Change (NIPCC) report Climate Change Reconsidered II, which is the chapter devoted to climate models (the PDF is here). The lengthy report is positioned as the skeptical antidote to the reports issued by the Intergovernmental Panel on Climate Change (IPCC) and it emphasizes papers and reports that contradict mainstream climate narratives and comparisons between climate models and observations.

The central claim of the NIPCC chapter is that IPCC authors are too confident in the ability of climate models to make claims about what is causing climate change or make reasonable projections about it, but I’m left unconvinced. The case it makes is limited, indirect, or based on faulty studies.

It is too large an exercise to go through the whole chapter, but for an example of the NIPCC arguments, we can dial in on section 1.1.5. That section addresses the question of how much temperatures will increase due to higher CO2 levels. Given a further increase in atmospheric CO2, that is a question of climate sensitivity, or the temperature change after the climate adjusts to a doubling of atmospheric CO2. Climate models surveyed by the IPCC produce sensitivities between 2 and 4.5 °C, and the IPCC offers a likely range from 1.5-4.5 °C.

The NIPCC report argues that climate models are too sensitive, based on estimates of climate sensitivity from a cherry-picked group of studies that infer climate sensitivity from observed climate changes. That figure offers a slanted, but not ridiculous, survey of climate sensitivity estimates that are lower than climate models. I’ve addressed such a figure before and shown how relying on it gives a biased view of climate sensitivity. I would be cautious in using it to inform my opinion about the large body of research on sensitivity.

Moreover, since the NIPCC was published, we have seen that the methods used to diagnose climate sensitivity from measurements of recent temperature change were faulty and biased in ways that underestimate climate sensitivity as it is calculated by models. Eliminating the bias resolves any apparent difference between modeled and measured sensitivity. Reversing directions, we have also seen that if you compare modeled apples-to-measured apples, there is again very little difference.

Section 1.1.5 also highlights a few specific estimates of climate sensitivity that are lower than the range implied by climate models. The first are from Lindzen and Choi and were first presented in 2009, and later refined in 2011, showing equilibrium climate sensitivity is almost surely below 1 °C. If those studies are true, then climate models would be far too sensitive to CO2 and their predictions wrong. But those studies are of suspicious quality, having had problems being published and later having been directly refuted by multiple authors. None of that is mentioned or challenged in the NIPCC paper. Instead the studies are held as authoritative, even though they are demonstrably wrong.

As far as I can tell, there is no reckoning with the weakness of the alternative arguments offered in this section or in the chapter as a whole. There is also no reckoning with the evidence that led the IPCC authors to conclude that climate models are suitable for diagnosing recent warming and predicting future changes. So while the chapter provides an alternative view, it doesn’t convincely argue for it.

The report is not a total bust. Some of it is textbook-style description of what a climate model is, how one works, how different climate models compare, and how they can be used and misused. Some of what it says is consistent with or heavily cites mainstream thinking. And on the fundamental questions—like whether or not human emissions of CO2 will affect climate—I would notionally, if not quantitatively, agree. However, when you set about to uphold skeptical papers and ideas that are long in the tooth, you will invariably end up with a product that makes scattershot, inconsistent, naive, and incorrect claims.  

And while I don’t want to beg the question too much, it should be noted that confidence in the models is something of a strawman. The 5th IPCC report has an entire chapter on the evaluation of climate models, and comparisons with observations, which does not hide the ball when climate models struggle and discusses many of the same comparisons highlighted in the NIPCC report.

In a follow-up post, I’ll do a deep dive into learning from model-data comparisons.