Originally posted on Apr 2, 2015 at Bishop Hill

April 1st is an interesting choice of date on which to release a new paper on climate sensitivity, but nevertheless that is the choice of Nature Climate Change. The new estimate has been produced by a team led a new name in this area: Daniel Johansen of Chalmers University of Technology in Gothenburg. The only member of the team who may be familiar to readers is Claudia Tebaldi.

The article’s headline conclusion is that ECS cannot be lower than 2°C.

Here we analyse how estimates of ECS change as observations accumulate over time and estimate the contribution of potential causes to the hiatus. We find that including observations over the hiatus reduces the most likely value for ECS from 2.8 °C to 2.5 °C, but that the lower bound of the 90% range remains stable around 2 °C. We also find that the hiatus is primarily attributable to El Niño/Southern Oscillation-related variability and reduced solar forcing.

And of course with conclusions like that it hasn’t taken long to pick up some media attention (Carbon Brief). But of course it’s always worth getting another opinion: Nic Lewis has studied the paper and has emailed me some thoughts:

[The study] estimates aerosol forcing using temperature data that is not at all latitudinally-resolved (the data is global only, but with land and ocean values separated – which has little effect). I believe that the use of latitudinally-unresolved data results in the flattening off of the increase in aerosols after the mid 1970s being conflated with the upswing in the Atlantic Multidecadal Oscillation (AMO), resulting in an excessively negative estimate for aerosol forcing. A more negative aerosol forcing estimate inevitably leads to a higher estimate for climate sensitivity.

Good previous studies such as Aldrin et al (2012) used hemisherically-resolved temperature data and, since aerosols are emitted very largely in the northern hemeisphere, were able to do a reasonable job of separating out the influence of aerosols and the AMO, and to arrive a modestly negative estimate for aerosol forcing. By contrast, Johansson obtains a very negative best estimate for aerosol forcing: -1.38 W/m2 including black-carbon-on-snow. This equates to close to -1.50 W/m2 for aerosol forcing alone (taking out BC-on-snow forcing at the AR5 efficacy-adjusted best estimate) as opposed to the AR5 best estimate of -0.9 W/m2 and Bjorn Stevens new estimate of -0.5 W/m2 (all changes since preindustrial).

Additionally, the study uses uniform Bayesian priors for ECS and effective ocean vertical diffusivity (their EVD). Doing so doesn’t seem to have a huge effect here, except when using data only  to 2001 or earlier. I imagine that is because the study uses low estimates of observational uncertainty, particularly for ocean heat content.

Another, minor, source of bias is that no allowance is made for the ocean being in imbalance in the early part of the 1765-2011 period. The Earth was recovering from the Little Ice Age then and ocean heat uptake was probably not negligible. In conjunction with the high aerosol forcing estimate, this accounts for all or most of the difference between the ECS estimate in this paper and that in Lewis & Curry 2014.

Oh well.