May 2013 - Nature Geoscience published a paper (Otto et al, 2013) which discussed a 'best-estimate' for transient climate response which led to widespread media coverage. In this guest article for Research News, the lead author of the paper, Alexander Otto, discusses the implications of the research.
We published a paper in Nature Geoscience on Sunday giving a new best-estimate of 1.3 °C for the Transient Climate Response, or the warming expected at the time carbon dioxide reaches double its pre-industrial concentration, using data from the most recent climate observations.This best-estimate is lower than the HadGEM2 TCR value of 2.5 °C and it is also 30% lower than the multi-model average of 1.8 °C of the CMIP5 models used in the current IPCC assessment. Does this mean that the Met Office's advice to government is based on a flawed model? Certainly not.
It is well acknowledged by all that the HadGEM2 model is at the top end of the range of TCR values in CMIP5, but we need a diverse range of TCR values to represent the uncertainties in our understanding of climate system processes. And the Met Office's advice to government, like any solid policy advice, is based on the range of results from different models, not just their own.
Our study implies a 5 - 95% confidence interval for the transient climate response of 0.9 - 2 °C compared to the range of 1 - 2.5 °C represented by the CMIP5 models. Acknowledging these uncertainties makes the differences look a bit less game-changing: results from the most recent decade appear to exclude the top one-third of the CMIP5 range, but the TCR range estimated from the 1970-2009 period as a whole (0.7 - 2.5 °C) does not, and we should always be careful not to over-interpret a single decade. The CMIP5 multi-model mean of 1.8 °C is well within our confidence interval, and only models with very high TCR values look potentially inconsistent with the most recent data, a conclusion consistent with e.g. Stott et al. (2013).
What are the implications of a TCR of 1.3 °C rather than 1.8 °C? The most likely changes predicted by the IPCC's models between now and 2050 might take until 2065 instead (assuming future warming rates simply scale with TCR). To put this result in perspective, internal climate variability and uncertainties in future forcing could well have more impact on the global temperature trajectory on this timescale.
First, I would always argue that for all practical, policy-relevant questions it is the transient climate response we should be interested in, but historically and academically the equilibrium response is still of high interest, so does our study tell us anything new about ECS?2
Our best estimate for climate sensitivity from the last decade of data alone is 2 °C, which is lower than the AR4 best-estimate of 3 °C, but in the range of other recent estimates using 20th century data (Aldrin et al, 2012; Libardoni and Forest, 2013; Lewis, 2013), so the fact that we get a strongly skewed distribution with a best-fit around 2 - 2.5 °C is not really news.
Using only the data from the decade from 2000-2009 we find a 5-95% confidence interval for equilibrium climate sensitivity of 1.2 - 3.9 °C. We compare the range to the range of the CMIP5 models of 2.2 - 4.7 °C saying that the range overlaps but is slightly moved to lower values. If we use the data from 1970-2009, also including the last decade, instead we find a 5 - 95% confidence interval of 0.9 - 5 °C for equilibrium climate sensitivity.
Comparing these ranges directly to the IPCC's range for climate sensitivity from AR4 is difficult. For one, the IPCC didn't directly give a 5 - 95% confidence interval (i.e. no upper 95% limit), and secondly, the IPCC range is not derived formally from an analysis of data, but is a consensus expert assessment of all the different lines of evidence underlying the IPCC report. Hence the IPCC's likely range of 2.0 - 4.5 °C is not directly comparable to a 17 - 83% confidence interval derived from our study. IPCC typically down-grades confidence levels from those reported in individual studies to account for "unknown unknowns".
For all investigated periods apart from the last decade alone our derived confidence intervals fully include the 2 - 4.5 °C range. They do extend below it, but that is not an inconsistency - which is why we conclude that, given all the uncertainties, our results are consistent with previous estimates for ECS.
The 'warming pause' over the recent decade does not show that climate change is not happening. And it certainly does not mean that climate scientists are "backing away" from our fundamental understanding. Every new decade of data brings new information that helps reduce uncertainties in climate forecasts. In some ways, the picture changes surprisingly slowly for such an intensely scrutinised problem: our new results may help rule out some of the highest-response models, but they are still consistent with the multi-model average, and our uncertainty range encompasses the bulk of the old range. It would certainly also help to test the physically based global circulation models against more than their transient climate response and climate sensitivity. Knowing how they perform in representing regional climate variables like air pressure at sea level (PMSL) or moisture variables is important and might lead to a narrowing of uncertainty eventually, and confidence in how they perform in projecting regional impact relevant quantities. This study highlights the importance of continued careful monitoring of the climate system, and also the dangers of over-interpreting any single decade's worth of data.Alexander Otto is Research Fellow in Climate Decisions at the Environmental Change Institute, University of Oxford.
Aldrin, M. et al. Environmetrics 23, 253-271 (2012)
Libardoni, A. G., and C. E. Forest, 2011: Sensitivity of distributions of climate system properties to the surface temperature dataset, 38 L22705.
Stott, P. A., Good, P., Jones, G., Gillett, N. P. & Hawkins, E. Environ.Res. Lett. 8, 014024 (2013).
Last updated: 14 August 2014