Share this article

print logo


When it comes to accuracy, climate models are doing pretty good so far

Don Paul

One of the battle cries of climate change nonbelievers (not scientist skeptics, but genuine deniers) is that climate models have been wildly inaccurate from the get-go, the 1970s.

I have to admit I’m dumbfounded as to how anyone in the face of irrefutable increasing obvious impacts of warming can still deny that evidence, but this article isn’t about my personal impressions. It’s about a new study that has gone back to both the early and more sophisticated climate models and measure their performance over the decades. The results are visualized here.

The performance is far from perfect, but no one in the scientific community expected perfect. There are far too many variables and interactions between variables we are still learning about. Overall, however, models are performing rather well over a long period of time.

One statistical problem in measuring the performance of these models is there can be no controlled experiment. There is only the planet Earth, in real time.

The new study published in Geophysical Research Letters this past week demonstrates that the models have been doing an unevenly good to very good job. Earlier models in the 1970s and 1980s were more primitive, with a more incomplete understanding of the many interactions between sea and air and the actual distribution of human activity-caused greenhouse gas emissions. For example, while it was known the oceans would be an energy and carbon sink, absorbing heat and carbon dioxide, most earlier models underestimated how much absorption actually occurs. The latter has caused more acidification of the oceans, threatening and killing coral reefs. The more rapid ocean heating has hastened sea level rise and caused migration of mid-latitude fish farther north in some cases and decimation of certain species populations in other cases.

Uncertain weekend wintry mix; certain global warming
Winter prepares to retreat – again – this week
Saturday's strong winds to sweep in weekend lake-effect snow
Winter finally coming to WNY with somewhat of a vengeance

The advent of more sophisticated models with far greater computer power behind them is improving model performance across the board. Some of the poorer performance of earlier models involved an overestimation of how much greenhouse gas would be emitted by human activity, especially in the case of methane, and initially then overestimated the pace of warming. That is no longer the case and hasn’t been for many years. Carbon dioxide reached its highest level in human history this past month, and methane continues to rapidly increase due to warming releasing more trapped methane from cold sea beds and the melting permafrost, as well as from agricultural methods.

What about carbon dioxide? Another early model error was an underestimation of how much reduction in the gases that destroy ozone would actually occur as the result of an international agreement. The protocol essentially banned the use of freon and chemically related gases being used in spray cans as propellants. These substances were strong greenhouse gases and, in fact, that international protocol succeeded beyond expectations. Its success helped initially slow the rate of atmospheric warming. Early models, in a sense, didn’t count on so much international cooperation.

Early models also didn’t foresee a temporary stabilization in methane emissions in the first decade of this century, now long since gone with rapid increases in methane.

So, newer and more complex models are now making projections that can be viewed with greater confidence, but they are still a long way from elusive perfection. Lead author Zeke Hausfather of the University of California at Berkeley gave an underlying viewpoint to the Washington Post: “Hausfather says the study found 'no detectable consistent overestimation or underestimation of future warming' in the models that were examined. … For the study, researchers evaluated climate models on how they performed regarding the amount of projected global mean temperature change compared to what actually took place over time.

"On this count, 10 out of the 17 models examined were virtually indistinguishable from observations, the study found — in other words, a near-perfect match.”

Precipitation changes linked to a warming climate are more difficult to track and predict, so much of what is examined in this study is directly concerned with the warming itself rather than its effects.

Coming back to my view, these findings may also suggest recent media reports that the models have badly underestimated overall warming may, in some quarters, be somewhat overblown. That’s not to suggest the climate crisis isn’t a crisis — not for a moment. The measured observations and impacts are bad enough. The model projections are bad enough, and effects will continue to worsen for the foreseeable future. Current modeled trends and observed data suggest we are on track toward some of the worst-case scenarios if we don’t act quickly to make drastic reductions in the burning of fossil fuels. There are feedback mechanisms, some tied to the drastic warming in the arctic, which cannot be reversed and which are actually amplifying.

The commission of this study did not occur with a goal of debunking unscientific denialism. The study’s purpose was to make a detailed analysis of how well the models have been performing. That the performance has been rather good and the study debunks denialism is an side effect of the hard data. In the meantime, a number of newer, more powerful climate models will be coming on line in the next two years, in advance of the sixth Intergovernmental Panel on Climate Change report.

There are no comments - be the first to comment