“We performed [an] analysis that told us explicitly what effect different model sensitivities have on simulated temperature trends,” Jochem Marotzke of the Max Planck Institute for Meteorology in Germany told environmentalresearchweb. “Answer: very little, for the trends considered here.”

Instead, the important factors seemed to be chance events and model forcing. Over the past 15 years, the global mean surface temperature has risen at only one-third to one-half of the average rate for the second half of the 20th century and climate models have not projected this trend.

“Call it a piece of unfinished business from IPCC Assessment Report 5, which was published in September 2013,” said Marotzke. “We assembled the assessment of the published literature on the surface-warming slowdown – or ‘hiatus’ – in the chapter that I co-coordinated (Working Group I Chapter 9). On a number of important issues, we were hamstrung by a lack of published material.”

In particular, according to Marotzke, there was little that indicated to what extent the difference between model simulations and observations during the hiatus period was due to random chance, or happened because the models were not driven correctly, or whether the models responded too sensitively to perturbations such as rising greenhouse-gas concentrations.

“By some quarters, the difference between simulations and observations over the hiatus period has been interpreted along the lines of ‘I told you all along that models warm too much. This is proof positive’,” said Marotzke. “We have shown that this notion is unfounded.”

Together with Piers Forster of the University of Leeds, UK, Marotzke analysed CMIP5 climate models and observations of temperature from HadCRUT4 back to the turn of last century. “Because random chance plays such a big role, we performed all analyses for all trend periods since 1900, to provide context,” Marotzke said. The team looked at trends over 15 years and over 62 years, the time periods used by the IPCC fifth assessment report, employing a multiple linear regression based on the global surface energy balance. “Combining the regression approach with the whole-century perspective was a new step that has no precedent,” according to Marotzke.

The analysis indicated that some models overestimate the cooling caused by volcanic eruptions and the subsequent warming recovery. “Viewed over the entire period since 1900, the differences between simulated and observed 15-year trends in global-mean surface temperature are dominated by internal variability and hence arise largely by coincidence,” write the researchers in Nature, adding that their analysis cannot rule out a small contribution from a systematic forcing bias in the models. Volcanic forcing could have contributed up to 15% of the difference between temperature observations and simulations between 1998 and 2012. What’s more, the CMIP5 models could be missing a cooling contribution from the radiative forcing over this timeframe or an enhancement of ocean heat uptake.

For the 62-year trends the difference between simulations and observations was dominated by the spread in the radiative forcing trend in the models.

“For those willing to engage in rational discourse, our results should put worries at rest that the hiatus might signal something fundamentally wrong with climate models,” said Marotzke. “It does not!”

Now Marotzke says there are major unanswered questions about exactly where and how the chance events that are affecting climate projections unfold, and how the model forcing comes about.

Related links

Related stories