Weather and climate agencies around the world have been almost unanimous in declaring 2014 the hottest year on record — something that has promoted considerable chagrin among climate change doubters. That’s because these “skeptics” have long sought to cast doubt on man-made global warming by pointing to an alleged global warming “pause” or “slowdown” — going on to suggest that the computerized climate models that scientists use to project future temperatures are flawed, and overestimate carbon dioxide’s warming effect.
So, is that true? Do the models consistently overestimate the warming effects of greenhouse gases like CO2?
As a recent study suggests, the answer is no. While many models didn’t predict the relatively modest surface-warming “hiatus,” it’s not because they’re biased in favor of greenhouse-gas emissions’ warming effects. Rather, researchers report in Nature, these computer simulations just struggle to predict “chaotic” (or random) short-term changes in the climate system that can temporarily add or subtract from CO2 emissions’ warming effects.
It’s true that air temperatures have increased slower in the past 15 years or so, and climate models on average instead predicted much more warming. And scientists are slowly beginning to figure out why temperatures didn’t rise quite as much as expected.
One probable contributor is pure natural variability: Cyclical processes in the Earth’s climate and temporary changes in the amount of solar radiation that reach the Earth’s surface can introduce “blips” into the Earth’s warming trend. Right now, oceans may be temporarily sucking up more heat from the atmosphere than they normally do. Moreover, a temporary downturn in solar output and an increase in light-reflecting aerosol pollution (acting like a chemical sunblock of sorts) could also have partially masked CO2-driven warming.
But researchers Jochem Marotzke of the Max Planck Institute of Meteorology and Piers M. Forster of the University of Leeds also wanted to check whether climate models are biased, by testing how their temperature predictions stack up against reality. So the researchers tested how 114 model simulations that underpin last year’s assessment report of the U.N. Intergovernmental Panel on Climate Change (IPCC) performed — not just for the 15-year period from 1998-2012 but for all 15-year periods stretching back to 1900. If this analysis were to show that models consistently overestimated or underestimated the amount of warming that actually occurred, then they must have some sort of systematic bias.
As it turns out, however, the models did pretty well. In each 15-year period, the model simulations produced a range of predictions. But each 15-year interval’s actual temperature trend always fell somewhere in the models’ prediction range. Moreover, even when 15-year actual temperature trends did fall toward the edges of the corresponding predicted ranges, they weren’t consistently at the higher or lower edges. Basically, when the models were missing the mark, they weren’t doing so consistently in one direction.
So, it’s true that the IPCC model runs didn’t predict the recent warming slowdown. But as these findings show, they didn’t accurately predict certain other 15-year periods of warming accelerations or slowdowns in the past either, and it’s not because they were always overestimating warming. Indeed, in some 15-year periods, the models underestimated warming. Essentially, that means climate skeptics are cherry-picking when they point out that climate models didn’t predict the recent 15-year hiatus.
That doesn’t entirely explain why the model simulations in a given year produced varying results to begin with, though. Was it due to differences in the underlying physics coded into the models? (The models differ slightly in terms of how much light they assume hits the Earth, how “sensitive” temperatures are to changes in CO2, and how much heat the oceans suck up.) Or was it just random fluctuations in the climate system? Or a combination? The researchers did a statistical analysis to answer that question.
In the end, none of those physical reasons was a major factor. Random fluctuations had 2.5 times the impact on the model predictions’ variations as all those physical factors together did, the researchers found. Only when the researchers used longer-term intervals (of more than 60 years) did differences in sunlight amount, ocean heat trapping or climate sensitivity start to make a big difference.
So climate models may not provide the perfect picture of what will happen to temperatures in a given short-term period (on 10- or 20-year scales). But maybe they simply can’t, due to the random ways in which climate can temporarily fluctuate. That doesn’t mean that climate models aren’t valuable to us. They still give us good sense of the long-term picture, the one that is more important for us to worry about anyway: that temperatures are increasing, and that natural factors can’t explain this increase.
As the researchers argue, then, their findings ought to put to rest assertions by climate “skeptics” that climate models overestimate how much warming we’re going to get.