Climate modelers are slippery devils. Why do I say that? McKitrick and Christy summed it up best in their paper, A Test of the Tropical 200- to 300-hPa Warming Rate in Climate Models, by relating the following about Global Climate Models (GCMs).
Read more Losing our religion: Most Americans don’t believe in virgin birth
GCMs embed countless minor hypotheses subject to continual testing and revision. Due to the sheer complexity of the climate itself, and of climate models, any number of observed discrepancies between projections from an individual model and some local feature of the real world can be accommodated, rationalized, or ignored, without calling into question the model itself, since the rejected component could be removed without the rest of the model ceasing to be a member of the class of GCMs.
It was this insufferable bobbing and weaving by climate modelers that made it difficult to pin them down. It did not help at all that their models were opaque to outsiders with no testable major hypothesis. McKitrick and Christy communicated their frustration on this point in the following passage.
We start from the assumption that there must, in principle, be at least one testable major hypothesis, the rejection of which would constitute failure of the model itself, in the sense that were the failed component to be removed, what remains would no longer be a GCM. We also start from the assumption that the measure we seek represents an emergent behavior from models based on both physical theory and modeler judgment, so that model integrations are genuinely expressions of a hypothesis.
In their paper, McKitrick and Christy came up with a testable major hypothesis. All GCMs have predicted specific magnitudes of temperature change in a reasonably rapid timescale in the tropical 200- to 300-hPa layer. It stands out as a bullseye in the middle of a temperature map of model results with pressure (or altitude) on the vertical axis and latitude on the horizontal axis. (That image can be found as Figure 1 under the “Figures” tab on the right, here.)
I don’t care what government scientists say, I am not at all convinced that the urban heat island (UHI) effect has been removed from the global surface air temperature record. This is for two very distinct reasons: 1️⃣ ALL climate models project that the strongest warming from greenhouse gas (GHG) forcing on the climate system should occur in the tropical mid-to-upper troposphere, not at the surface (McKitrick & Christy, 2020). Yet, most of the warming since the 1970s has been near the surface. These observations run counter to theory. 🔗 https://t.co/YFaxdZS1Rv 2️⃣ Rural and suburban GHCNd stations also exhibit spurious warming trends. In fact, the greatest rate of warming occurs when population just starts to increase beyond wilderness conditions (Oke, 1973), which is unaccounted for in the “pairwise homogenization” practice that is used by NOAA / NASA GISS. 🔗 https://t.co/PNQFHyICJT
— Chris Martz (@ChrisMartzWX) May 27, 2025
Unfortunately for the modelers, the observations did not cooperate. Examining the 60-year period from 1958 through 2017, McKitrick and Christy found that every one of the 102 CMIP5 model simulations predicted more warming in the tropical upper troposphere than was observed. The discrepancies were statistically significant in most individual comparisons and for the ensemble average. (Scroll over in the embedded tweet, or see Figure 3 at the same link.)
Read more Rubio: ‘A Generational Shift’
This was not some obscure discrepancy in an inconsequential corner of the climate system. It occurred precisely where the models predicted one of their strongest and most distinctive atmospheric warming responses. The significance of the finding was subsequently reinforced by the IPCC itself, which acknowledged in its Sixth Assessment Report that both CMIP5 and CMIP6 models continued to overestimate tropical upper-tropospheric warming.
McKitrick and Christy interpreted this failure as evidence against a major hypothesis embedded in contemporary GCMs: the parameterized representation of tropical atmospheric thermodynamics and greenhouse-induced warming. Their results call into question whether the models correctly represent the magnitude and vertical distribution of atmospheric warming. It raises legitimate questions about modeled climate sensitivity, and the reliability of projections derived from these models.
Subsequently, climate modelers have offered explanations involving imperfect parameterizations, but explanations should themselves produce independently testable predictions, rather than serve as unlimited opportunities to accommodate contradictory observations.
The essential issue is not whether climate models can be continually modified to reproduce the historical record. It is whether they can make sufficiently specific predictions that survive independent testing. When a major, theoretically important prediction repeatedly fails such a test, the burden falls upon model developers to demonstrate that their underlying assumptions remain sound. Scientific credibility ultimately rests on predictive performance, not on the capacity to explain away discrepancies after the fact.
Climate modelers do not like it when outsiders try to hold them accountable to normal scientific practices like formulating a testable hypothesis. Does that make them slippery creatures? It depends on one’s definition of a slippery creature.
Read more This election year, Columbus Day matters
Image generated by ChatGPT.