Climate models are even more accurate than you thought

Global climate models aren’t given nearly enough credit for their accurate global temperature change projections. As the 2014 IPCC report showed, observed global surface temperature changes have been within the range of climate model simulations.

Now a new study shows that the models were even more accurate than previously thought. In previous evaluations like the one done by the IPCC, climate model simulations of global surface air temperature were compared to global surface temperature observational records like HadCRUT4. However, over the oceans, HadCRUT4 uses sea surface temperatures rather than air temperatures.

Thus looking at modeled air temperatures and HadCRUT4 observations isn’t quite an apples-to-apples comparison for the oceans. As it turns out, sea surface temperatures haven’t been warming fast as marine air temperatures, so this comparison introduces a bias that makes the observations look cooler than the model simulations. In reality, the comparisons weren’t quite correct. As lead author Kevin Cowtan told me,

We have highlighted the fact that the planet does not warm uniformly. Air temperatures warm faster than the oceans, air temperatures over land warm faster than global air temperatures. When you put a number on global warming, that number always depends on what you are measuring. And when you do a comparison, you need to ensure you are comparing the same things.

The model projections have generally reported global air temperatures. That’s quite helpful, because we generally live in the air rather than the water. The observations, by mixing air and water temperatures, are expected to slightly underestimate the warming of the atmosphere.

The new study addresses this problem by instead blending the modeled air temperatures over land with the modeled sea surface temperatures to allow for an apples-to-apples comparison. The authors also identified another challenging issue for these model-data comparisons in the Arctic. Over sea ice, surface air temperature measurements are used, but for open ocean, sea surface temperatures are used. As co-author Michael Mann notes, as Arctic sea ice continues to melt away, this is another factor that accurate model-data comparisons must account for.

One key complication that arises is that the observations typically extrapolate land temperatures over sea ice covered regions since the sea surface temperature is not accessible in that case. But the distribution of sea ice changes seasonally, and there is a long-term trend toward decreasing sea ice in many regions. So the observations actually represent a moving target.

When accounting for these factors, the study finds that the difference between observed and modeled temperatures since 1975 is smaller than previously believed. The models had projected a 0.226°C per decade global surface air warming trend for 1975–2014 (and 0.212°C per decade over the geographic area covered by the HadCRUT4 record). However, when matching the HadCRUT4 methods for measuring sea surface temperatures, the modeled trend is reduced to 0.196°C per decade. The observed HadCRUT4 trend is 0.170°C per decade.

So when doing an apples-to-apples comparison, the difference between modeled global temperature simulations and observations is 38% smaller than previous estimates. Additionally, as noted in a 2014 paper led by NASA GISS director Gavin Schmidt, less energy from the sun has reached the Earth’s surface than anticipated in these model simulations, both because solar activity declined more than expected, and volcanic activity was higher than expected. Ed Hawkins, another co-author of this study, wrote about this effect.

Combined, the apparent discrepancy between observations and simulations of global temperature over the past 15 years can be partly explained by the way the comparison is done (about a third), by the incorrect radiative forcings (about a third) and the rest is either due to climate variability or because the models are slightly over sensitive on average. But, the room for the latter effect is now much smaller.

As Hawkins notes, the remaining discrepancy between modeled and observed temperatures may come down to climate variability; namely the fact that there has been a preponderance of La Niña events over the past decade, which have a short-term cooling influence on global surface temperatures. When there are more La Niñas, we expect temperatures to fall below the average model projection, and when there are more El Niños, we expect temperatures to be above the projection, as may be the case when 2015 breaks the temperature record.

We can’t predict changes in solar activity, volcanic eruptions, or natural ocean cycles ahead of time. If we want to evaluate the accuracy of long-term global warming model projections, we have to account for the difference between the simulated and observed changes in these factors. When the authors of this study did so, they found that climate models have very accurately projected the observed global surface warming trend.

In other words, as I discussed in my book and Denial101x lecture, climate models have proven themselves reliable in predicting long-term global surface temperature changes. In fact, even more reliable than I realized.

There’s a common myth that models are unreliable, often based on apples-to-oranges comparisons, like looking at satellite estimates of temperatures higher in the atmosphere versus modeled surface air temperatures. Or, some contrarians like John Christy will only consider the temperature high in the atmosphere, where satellite estimates are less reliable, and where people don’t live.

This new study has shown that when we do an apples-to-apples comparison, climate models have done a good job projecting the observed temperatures where humans live. And those models predict that unless we take serious and immediate action to reduce human carbon pollution, global warming will continue to accelerate into dangerous territory.