Making Heads and Tails of It

I’m going to flip a coin twice. Can you guess the outcome? Not with any certainty. Probability favors heads once and tails once, but there’s still a good chance (at 25 percent each) of two heads or two tails. Now, if I were to flip the coin 10 million times, an uninterrupted string of heads or tails would be, on the whole, doubtful. It’s very, very likely that the distribution would be reasonably close to 50-50. You could make an educated guess, at least.

Scientists don’t like to talk about it, but there’s a certain amount of coin-tossing involved in building a climate model. A study published Wednesday in the journal Nature lifts the lid on the guesses that climatologists make and explains why the models are reliable, despite these inherent and necessary uncertainties.

The messy fact is that some factors are simply impossible to predict over the short term. The most obvious, and most discussed, is the El Niño Southern Oscillation (ENSO) Cycle, a fluctuation in temperature between ocean and air that occurs at uncertain times. An episode should last a little under one year, but it can go much longer. Volcanoes also have a nasty habit of erupting with little warning, and their gases can depress global temperatures for years. Unpredictable events have a tremendous impact on climate. According to the new study, chaotic and unforeseeable factors account for between 30 percent and 40 percent of global temperature trends.

Numbers like those may shake your faith in climate models, but they shouldn’t. Let’s go back to the coin toss. There’s no way to know what the ENSO cycle will do in the next two, five, or even ten years. Over the next 60 years, however, its results are easier to predict. The hot and cold cycles will more or less cancel each other out—the same way heads and tails roughly balance after thousands of coin tosses. This effect turns chaotic variables into background noise, which scientists can basically manage when projecting temperature trends 60 or 70 years into the future. When the chaos drops out of the equation, underlying trends, like the effect of greenhouse gases on temperatures, emerge.

To prove this, climatologists Piers Forster of the University of Leeds and Jochem Marotzke of the Max Planck Institute for Meteorology collected temperature data from 1900 to 2012, then sliced the 112 years of information into 15-year chunks. They then compared the real-world observations with what climate models predicted.

Most of the differences between the temperatures forecast and the temperatures that actually occurred were a result of what the authors call “quasi-random internal climate variability”—in other words, meteorological coin tosses. When the authors broke the data into longer periods of time, those coin tosses had little impact on how the models performed—demonstrating that random events tend to cancel out when you take a long view on climate.

“If we were to run one of our long-term simulations, it would get the right chaotic variations, but not necessarily at the right time,” says Forster. “The differences would be just by pure chance.”

Here’s the upshot: Marotzke and Forster showed that the so-called global warming hiatus—the unexpected slowing of the rate of surface air temperature rise between 1998 and 2012—is the result of climatologists guessing wrong on a coin toss. Climate models are reliable over the long term, though, and their predictions of radical global warming due to increasing concentration of greenhouse gases in the atmosphere are on the money.

Marotzke and Forster aren’t the first to demonstrate that the much ballyhooed and misunderstood “global warming slowdown” is a result of unpredictable natural phenomena, rather than the end of climate change. Their research, however, reinforces scientists’ understanding of the strengths and weaknesses of climate models.

This kind of repetitive research—reaching the same conclusions via different strategies—is what separates climatologists from global warming deniers, who rely on a tiny number of (usually poorly constructed) studies. It’s one more piece of evidence to share with your skeptical friends. Who knows? Maybe this will be the one that convinces them. Something has to, right? Right?