At least two new papers are discussing our ability to constrain climate sensitivity (the holy grail in the title). Klocke et al. (2011, 10.1175/2011JCLI4193.1) caution, that even though we do weight multi model ensemble estimates of climate sensitivity by the statistical relationships of individual models with observed data, this may not be admissible due to quite basic model shortcomings. Put drastically, even in one model we are in principle able to tune the climate sensitivity according to our wishes. The authors state as an implication of other published results, “that all models are formally unlikely,” and thus “weighting an ensemble … is essentially asserting that incorrect models are more reliable than even-more-incorrect models.”
Yoshimori et al. (2011, 10.1175/2011JCLI3954.1) on the other hand describe, how the climate sensitivity – in a perturbed physics ensemble with the japanese MIROC3.2 model – depends on the forcing that is applied and the climate background state due to differences in the feedback processes. That is, firstly, the estimate of the climate sensitivity differs for greenhouse gas forcing or a forcing thought to represent the last glacial maximum (LGM) conditions. Secondly, the feedbacks and thus the sensitivity is not the same for a doubling of CO2 in the recent climate and a doubling of CO2 in a LGM climate.
While the results by Klocke et al. (2011) stress the importance of improving our models and of bringing to our own mind the shortcomings of our models, Yoshimori et al. (2011) to some extent give a theoretical foundation for studies of dynamic processes in the climate system under (really) cold or warm climates and under more recent (moderately) cool and warm conditions like the little ice age or the medieval climate anomaly / medieval warm period.