A paper out this week by Friedrich et al. (2016) is the latest in what seems like a series this year determining climate sensitivity using the palaeoclimate record (see here). This is a very powerful approach but has its difficulties many of which are discussed in our previous post and here. The Friedrich et al (2016) study used a new empirical estimate of Surface Air Temperature (SAT) based on a compilation of Sea Surface Temperatures (as did Snyder recently) and a complete assessment of the processes “forcing” climate change over the last ~800 thousand years (e.g. CO2, land-ice albedo, and dust) to identify that climate sensitivity changed as a function of climate state: they found it was ~1.8 K per CO2 doubling when the Earth was substantially colder than today and ~5K per doubling when the Earth was only a little bit colder than the pre-industrial. We @theFosterlab were just involved in a review on this subject (here) and a summary of the literature from that paper is shown below.
Friedrich et al. go further than other studies and apply the palaeo-sensitivity to predict our warm future (see figure below). They correct for the fact that over the next 100 years the climate isn’t in equilibrium with its forcing and show that the high “warmer-paleo” sensitivity yields temperature in 2100 AD, given a business as usual RCP8.5 emissions scenario, that could be as high as 6 K (5-7 K) compared to 5 K (3 -6 K) from the CMIP5 models (see figure below).
The accuracy of such a paleo-sensitivity approach to predict the future climate is: (i) very dependent the accuracy of the temperature record used to determine sensitivity in the past (and others limitations, see http://julesandjames.blogspot.co.uk/); (ii) but is reliant on the assumption that the sensitivity estimated from climates slightly warmer and substantially cooler than today is applicable to Earth temperatures up to 5 K warmer.
I think it’s a great finding of this paper that climate sensitivity is state dependent over the Pleistocene glacial cycles, but this also means it’s probably an over simplification to assume that a similar state dependency doesn’t characterise the system when temperatures are substantially higher than the pre-industrial (see top figure).
We attempted to investigate this in this paper in 2015 using boron isotope based CO2 data from the Pliocene. We didn’t examine the data for a state dependency within the Pliocene (see here for why) but determined that the average climate sensitivity in the Pliocene was ~3.7 K per CO2 doubling (with an uncertainty of 2-6 K). This overlaps with what Friedrich et al. determine as “warm-paleo” sensitivity. However, my feeling is that, given the current uncertainties in determining climate sensitivity in the past, rather than indicating the Earth system is more sensitive than the future, studies like that of Friedrich et al. (2016) and our own (http://www.nature.com/nature/journal/v518/n7537/full/nature14145.html) are very important validators of our understanding of the behaviour of the climate system encapsulated by the CMIP5 models. These are very different ways to understand the climate system yet they give the same results – this is very powerful and is a great illustration of the utility of palaeoclimate research. As a community we now need to work hard to reduce the inherent uncertainties and get the most out of the rock record of past climate changes.