Due to their coarse horizontal resolution, present day climate models must parameterize deep convection (see: Storer, R. L., Griffin, B. M., Höft, J., Weber, J. K., Raut, E., Larson, V. E., Wang, M., and Rasch, P. J. (2015), "Parameterizing deep convection using the assumed probability density function method", Geosci. Model Dev., 8, 1-19, doi:10.5194/gmd-8-1-2015).
Thus while Mauritsen & Stevens (2015) consider deep convention, the parameterization in their coarse grid model filters out the positive feedback mechanism identified by Trenberth et al (2015); who can identify the signal for this vapor related positive feedback mechanism because they are working with raw, un-filtered, data.
The models Mauritsen use have a finer resolution than the analysis Trenberth performs. Trenberth performs an analysis on the planet as a whole, and captures the impact of small scale feedbacks on average.
In the same way Mauritsen performs an analysis of the CERES data to determine what the effect of small scale feedbacks are on average. The average effects of these small scale feedbacks are then included in Mauritsen's model.
[/quote]
First, let me note that the objective the Trenberth et al (2015) paper was different than the objective of the Mauritsen & Stevens (2015), so there was no need for the different teams to use comparable models.
Furthermore, as Trenberth et al were not estimating climate sensitivity, I am not concerned about their model; however Mauritsen & Stevens (2015) do estimate climate sensitivities and thus their models run into the same limits discussed by Jean-Pascal van Ypersele (professor of climatology and environmental sciences at the Université catholique de Louvain in Belgium & the likely new chair of the Intergovernmental Panel on Climate Change (IPCC), to succeed Dr Rajendra Pachauri), in the interview with Carbon Brief below, where he notes that correctly modeling the physics of clouds requires high resolution models, special computer codes and a lot of computer time, as was the case for the CESM-H run. All of these consideration are not adequately addressed by the Mauritsen & Stevens (2015) model.
Hopefully, the first stage of the ACME program (to be completed in another 2.5 years) will reduce this uncertainty with modeling clouds before society follows a pathway that it comes to regret.
http://www.carbonbrief.org/blog/2015/04/the-carbon-brief-interview-jean-pascal-van-ypersele/Extract: "CB: So, as a climate scientist, what areas of new research excite you the most, which questions would you like to see more than any others?
JY: Probably one of the key questions is the role of clouds. I mean, the main reason behind, let's face it, the large range in the climate sensitivity, the equilibrium climate sensitivity - sorry for using jargon here, but probably the readers of this will know what we're talking about - climate sensitivity is simply the amount of warming that you get at equilibrium when you double the amount of CO2 in the atmosphere. But the range for the number has been basically [the same] for the last 40 years; 1.5 to 4.5C with some fluctuations. It's a situation with nuances, but basically it's a large range - it's a factor of three. It would make a big difference to reduce that range and to know better if, for a doubling of the concentration, the warming would be 2C, 3C, or 4C. It would make a big difference for policymakers as well when they discuss risk management because the risk would be better known. And the main factor behind that is cloud microphysics and the way clouds interact with other elements in the climate system. And relatively little progress has been made actually. When Charney published in 1975 his first assessment of the range of climate sensitivity, it was 1.5 to 4.5 and it's still the same today. So little progress has been made, and the main factor is the uncertainties around clouds.
CB: So, my understanding of clouds - cloud feedbacks - is that there's quite a lot of evidence that there is a positive feedback, which means it would amplify the warming that you would get just purely for a doubling of carbon dioxide. There are suggestions that it might be a negative feedback, but there isn't a lot of evidence to support that, in fact I'm not sure of any. Having identified clouds as an issue, does that lessen the possibility that's it's at the lower end of that range?
JP: Things are even a little more complicated than that. Because it's true that overall water vapour - and clouds is one of the manifestations of water vapour - increase the warming for an increase in the CO2 concentration, so overall the feedback is positive. But inside that big envelope there are different behaviours for different kinds of clouds. I'm not a cloud expert, but still, I know that, for example, high-level clouds like cirrus, have a warming effect if there's an increase in their number. Low-lying clouds, low in the atmosphere, on the contrary, have a cooling effect because they reflect more sunlight to space, while the upper clouds have a larger greenhouse effect by trapping heat radiation. So, when you talk about clouds, clouds is not just a large cloud - let's use that word! - of water vapour in different forms. The altitude and type of clouds and the microphysics is quite different for different levels in the atmosphere. And to understand the overall effects of clouds and changing cloudiness in warming climates depends on the understanding, the detail of the understanding, of the microphysics of those individual layers of clouds at different layers. For some layers, it's a positive feedback and for some layers it's a negative feedback. And the balance is positive indeed, but how positive? How to quantify that depends on the details of the microphysics of clouds, and there there is still much progress to do. One difficulty being that it's very hard to observe what is happening inside clouds at their level, and the other difficulty - or one other difficulty - is the difficulty to model with high resolution - the high resolution that would be needed to resolve clouds. It's very difficult to do that in climate models because we are limited by computer power, and that's one of the difficulties."