How Accurate Are Climate Change Models?
Ignore the hype - Simple models with valid inputs accurately predict temperature rise
Many world leaders maintain that Anthropogenic (man-made) Climate Change (CC) presents the greatest threat to humanity with the possibility of a climate catastrophe. The United States and other developed countries have agreed to radically transform their economies in an effort to limit CO2 and other greenhouse gases (GHG) emissions and further global warming to address this threat.
Much of the concern relies on the predictions of CC models. It is therefore rational to assess how accurate CC models have been in the past before we rely heavily on them in determining our future.
It is important to first acknowledge that the predictions from CC models are frequently (and understandably) met with skepticism by large segments of the population for several reasons. First, mainstream media has presented a highly distorted, sensationalized picture of climate science and predictions. A perfect example of this is the constant effort in the media to attribute weather events to CC, despite the fact there is no rigorous scientific justification. In a recent Pure Science Substack post I detailed how there is no evidence of increasing tornado power over the last 70 years despite incessant media claims to the contrary. Similarly, the increase in hurricane damage over time is falsely attributed to CC, and even weather events such as record cold or snowfall amounts are wrongly attributed to global warming!
Additionally, good CC modeling will give a range of possible outcomes which depend upon currently unknown parameters; the media frequently will ignore predictions of the most likely outcome, and instead focus on unlikely but more sensational possible futures, such as the planet warming by 4 or 5 degrees C this century (extremely unlikely). It does not help that the scientific community is usually very quiet with regards to such hyperbolic claims and when Al Gore famously predicted in his 2006 movie “An Inconvenient Truth”:
The arctic summer sea ice would vanish at some point by 2014 (wrong)
The glacier on Mt. Kilimanjaro would vanish by 2016 (wrong)
Increasing tornado and hurricane power and frequency (wrong)
Having said all that, simple climate models with reasonable, non-politicized assumptions, have done well in predicting global temperature increases (regional trends not so much).
Recent research by Zeke Hausfather shows that 14 out of 17 climate models produced predictions that matched (within uncertainty) actual, observed global surface temperature increases when the predictions are adjusted to account for variations between predicted and actual GHG increases over time.
Hausfather’s analysis is partially confirmed by examining the predictions of one of the most highly cited and influential papers from a prominent climate scientist, James Hansen (and co-authors), in the 1981 Science paper, “Climate impact of increasing atmospheric carbon dioxide”. This paper is very informative because it was written at a time prior to the politicization of the CC issue. In fact, at the time of writing, the global temperature trend had been flat for 25 years and there was still concern by some that the Earth might be heading towards another Ice Age!
Hansen used a simple energy balance model to project future warming. The model had one critical uncertain parameter, now referred to as the Climate Sensitivity, which is the expected amount of mean global surface temperature increase resulting from a doubling of CO2 from pre-industrial levels of 280 ppm to 560 ppm. Figure 3 below shows the historical temperature anomaly (change relative to 1977) up to 1977, and 3 of the Hansen model temperature anomaly predictions going forward from 1977, each with a different assumed Climate Sensitivity. Fortunately for comparison purposes, Hansen assumed what he called a “slow growth” scenario for CO2 emission increases of 2 %/year which is slightly greater, but close to what actually occurred.
The climate model predicted a temperature increase of between 0.44 and 0.72 deg. C from 1977 to 2020. This corresponds to 0.10 to 0.17 C/decade, which is slightly below but consistent with measured temperature increases over the time period from earth stations, satellites, ocean measurements and weather balloons of 0.18 C/dec. (range 0.14 to 0.21 C/dec.).
Hausfather included in his analysis another paper published by Hansen and colleagues in 1988, which represented one of the first modern climate models. It divided the world into discrete grid cells of eight degrees latitude by 10 degrees longitude, with nine vertical layers of the atmosphere. It included aerosols, various greenhouse gases in addition to CO2, and basic cloud dynamics. Figure 4 below shows the temperature anomaly predictions in that paper for three different CO2 emission scenarios.
Scenario B assumed GHG emissions which most closely resemble the actual GHG emissions that followed. Scenario B predicted a temperature increase from 1988 to 2020 of approximately 0.8 C (0.25 C/dec.), which is about 40% too high.
Hansen’s 1988 model assumed a Climate Sensitivity of 4.2 C, which is much higher than the Climate Sensitivity value of 2.8 C considered most likely in the 1981 paper. Recall that the Climate Sensitivity is the amount of temperature increase expected with a doubling of CO2. Therefore, models with higher Climate Sensitivity predict greater temperature increases.
It is probably not a coincidence that the issue of Global Warming was rebranded as Climate Change when Dr. Hansen testified in front of Congress in 1988 AND Hansen and other climate scientists took on additional roles as advocates for CC prevention.
Models since 1988 have become much more complex with much smaller grid sizes. The Climate Sensitivity is not an input parameter but instead is determined by numerous other model parameters. For some reason, possibly bias by researchers, most of the models used by the IPCC (Coupled Model Intercomparison Project (CMIP)) since then produce Climate Sensitivity values higher than justified, resulting in the models predicting rising temperature trends in excess of observed rates. Fig. 5 below shows a 2017 comparison of observed temperature trend vs. CMIP5 models in the lower troposphere (0 to 10 km above earth). The yellow band is the 5% to 95% confidence range of output from CMIP5 climate simulations. Clearly, the observed temperature trend is barely in the model confidence range.
In his book “Unsettled” Steven Koonin, who served as Undersecretary for Science in the US Department of Energy under President Obama, explains that the CMIP climate models have dozens of free, uncertain variables that are tuned to match past climate. But just like underdetermined systems in mathematics there are an infinite number of possible solutions. Figure 6 shows twenty-six CMIP6 models compared with actual observed temperature since 1880. Given the wide variability between model predictions, many scientists like Koonin and myself question the reliability of any single model or ensemble of models to predict not only the magnitude of future temperatures, but other climate characteristics like regional and precipitation trends.
Consequently, despite 4 decades of increasingly sophisticated climate modeling we are left relying on simple energy balance models of the earth to predict future temperatures. A recent paper by Lewis and Curry reanalyzed the earth’s energy balance with the latest data and concludes the Equilibrium Climate Sensitivity (ECS) of 1.5 C (1.05 - 2.45 range.) If they are correct, when CO2 atmospheric concentration reaches 560 ppm later this century, the mean global surface temperature anomaly would have increased only around +0.41 C to 1.5 C from the current 1.09 C. This appears too low: extrapolating the current rate of +2.5 ppm/yr increase in atmospheric CO2 concentration from the current 417 ppm predicts reaching the 560 ppm CO2 doubling point in 2079; if the current observed linear temperature trend continues until then, the temperature anomaly would increase another 1.02 C from the present value of 1.09 C and be +2.11 C relative to the end of the 19th century. This ad-hoc 2.1 C ECS value is consistent with the two energy balance models presented here (Hansen 81, Lewis& Curry 2018) and is almost exactly half way between the ECS values found for each.
Conclusion: Energy balance climate models have accurately predicted rising global temperatures. Combining updated ECS values and extrapolating present trends leads to a prediction of another 1 C global mean temperature increase by the time atmospheric CO2 reaches 560 ppm later this century.
One point you alluded to isn't made often enough generally, I think -- the basic power of the simple *early* models to pretty accurately predict temperature response to CO2 is a robust indicator that the physics is right, later refined by more complicated calculations. I can imagine that if only recently had we been able to hone in on the temperature signal, critics could say that the temperature rise is only one of a multitude of possible consequences of so many input parameters. The simple case of energy in = energy out and an IR-absorbing layer (warming lower atmosphere, cooling upper atmosphere) is incredibly hard to argue against.