“We estimate that the warming slowdown (< 0.1 K/decade trend beginning in 1998) could persist, due to internal variability cooling, through 2020, 2025 or 2030 with probabilities 16%, 11%, and 6%, respectively.” – Knutson et al.
A new paper by Tom Knutson, Rong Zhang and Larry Horowitz of NOAA GFDL has just been published in Nature Communications [link to full manuscript]. The authors take a well-balanced approach to seeking possible explanations for global mean surface temperature (GMST) increasing at a much lower rate from around the turn of the century than over the late 20th century, and consider the possibility that it may continue for some time.
The abstract reads:
Observed global mean temperature over 1998 to 2015 increased at a slower rate (0.1 K decade-1) compared to the ensemble mean (forced) projected by Coupled Model Intercomparison Project 5 (CMIP5) models (0.2 K decade-1). Here we investigate the prospects for this slower rate to persist for a decade or more. The slower rate could persist if the transient climate response is overestimated by CMIP5 models by a factor of two, as suggested by recent low-end estimates. Alternatively, using CMIP5 models’ warming rate, the slower rate could still persist due to strong multidecadal internal variability cooling. Combining the CMIP5 ensemble warming rate with internal variability episodes from a single climate model – having the strongest multidecadal variability among CMIP5 models – we estimate that the warming slowdown (< 0.1 K decade-1 trend beginning in 1998) could persist, due to internal variability cooling, through 2020, 2025 or 2030 with probabilities 16%, 11%, and 6%, respectively.
The authors say that CMIP5 and earlier generation climate models’ historical forcing simulations, in the ensemble mean, reproduce well observed late 20th century warming and a hiatus during around 1940–1970, in response to anthropogenic and natural forcings. However, these models do not reproduce the rapidity of the observed early 20th century warming, suggesting that internal (unforced) climate variability played a minor role in the two later episodes but a significant one in the early warming.
Although this comparison with observations does at first sight suggest such a conclusion, the satisfactory reproduction of the 1940–1970 hiatus and late 20th century warming by the CMIP5 ensemble could have other explanations. Whilst in the ensemble mean of climate models the mid-century hiatus and late century warming no doubt are almost entirely forced, there is reason to believe that in the real world multidecadal internal variability –linked to the Atlantic multidecadal oscillation (AMO) and the Pacific Decadal Oscillation (PDO) – played a significant role, depressing the GMST trend from 1940 to the early/mid 1970s and increasing it from then until the turn of the century.[i] In many CMIP5 climate models, excessive strengthening of negative aerosol forcing (cooling) during the 3rd quarter of the 20th century may have served to mimic the effect on GMST of missing multidecadal internal variability. And excessive transient climate response (TCR) levels in CMIP5 models may have compensated, in the late 20th century, for their lack of a positive influence of internal variability on GMST.
As cited by the authors, some recent studies suggest that misrepresentation of forcing accounts for the rapid post-2000 warming in CMIP5 models exceeding the observed warming.[ii] However, these studies focused on individual forcings, such as solar and stratospheric (volcanic) aerosols. When all forcings are taken into account – some of which grew more positively after 2005 than in the RCP scenarios – there is little evidence that the rate of increase in actual total net forcing was overestimated in CMIP5 models. In a thorough investigation of the situation in the NorESM1-M model, it was estimated that over 1998–2012 the actual forcing trend was about 20% higher than that in the RCP8.5 scenario.[iii]
Sensibly, Knutson et al. effectively discount explanations of the early 21st century warming slowdown based on misstated forcing changes, missing temperature data in warming regions such as the Arctic, and the like. Instead they “explore how long an early 21st century hiatus or slowdown could last, assuming a strong internally generated cooling superimposed upon ongoing anthropogenic warming.”