“The goal of the paper was to produce a probabilistic forecast of the climate warming response to doubling of carbon dioxide concentrations in the atmosphere (Climate Sensitivity) from and ensemble of simulations.”

It presented results from the highly publicized distributed computing project which is available online at climateprediction.net (CPDN). The CPDN project relies on members of the public to donate their computer down time to run hundreds of thousands of climate simulations. The paper tackled a hot topic, which is the production of constrained, or probabilistic, forecasts of future climate warming. Also, I believe it was one of the earliest studies to suggest that climate sensitivity was intrinsically difficult to constrain and hence not the best choice of variable on which to base mitigation and adaptation policies.

This first multi-thousand member "perturbed physics" ensemble simulation of present and future climate, as completed by the distributed computing project at CPDN, was used to search for constraints on the response to increasing greenhouse gas levels among present day observable climate variables. The search was conducted using a systematic statistical methodology to identify correlations between observables and the quantities we wished to predict, namely, the climate sensitivity and the climate feedback parameter. A sensitivity analysis was conducted to ensure that results were minimally dependent on the parameters of the methodology.

The goal of the paper was to produce a probabilistic forecast of the climate warming response to a doubling of carbon dioxide concentrations in the atmosphere (climate sensitivity) from an ensemble of simulations. In particular, a methodology was devised to minimize the role of the prior, that is, the initial choice of simulations that constitute the ensemble.

You cannot forecast whatever you like. Climate sensitivity does not scale with observations but the inverse does. A direct consequence, though difficult to explain in layman’s terms, is that probability distributions for climate sensitivity are skewed toward higher values, that is, they have a sharp cut-off at the low end and a gentle decrease at the high end, to the representation of the origins of model-data discrepancy. This means that even small changes in the distribution amount to significant changes in the upper 90% cut-off value (upper limit) which is what matters most.

I got involved with CPDN after moving to the University of Oxford’s Department of Physics. It is in the nature of projects like CPDN to face constant obstacles. Some can be overcome, while others may lead us to newer and even more exciting opportunities.

I am now part of the "Water and global change" (WATCH) project, which is funded by the European Union. I am developing methodologies to transfer climate model uncertainties into hydrological models to help quantify the vulnerability of the water cycle to climate change.

Adaptation and mitigation policies should not be based on poorly constrained estimates of the equilibrium response of the climate system such as Climate Sensitivity. The attempt to specify a "safe" level of greenhouse gas concentrations that will avoid a Climate Sensitivity >2 is no different. Governments should focus on alternative policy targets that depend on better-constrained aspects of the climate system, such as the transient response.