Dear Paul – Risbey et al. aren’t _recommending_ the expectation of zero centennial trend, they are _reporting_ the state-of-the-art opinion amongst attribution experts. That’s what their study was, a questionairre and follow-up interviews with a range of attribution experts. A companion article in Bulletin of the American Meteorological Society was interesting too.

The expected mean trend in century-long series of (internally generated) natural variability is typically zero unless there are reasons to expect a prolonged cooling or warming on this scale.

assumes an answer for something that has be verified by measurement. At the level of the GCMs, even the warmers accept the idea that climate is a non-linear dynamical system, which means that on its own there is lots of internal variability even without a change in external forcing. Riseby et. al., doesn’t make any sense to me.

I’ve read your post a few times and it sure seems to me that if you’re accurately presenting “The standard assumption in climate research”, the whole thing is one big circular argument.

Sure “the expected mean trend of X on a century scale is 0″ can be said for many things, not just global temperatures. This sort of statement just serves as an implicit null hypothesis. That’s particularly so in a case such as you present where an aside like “unless there are reasons to expect a prolonged cooling or warming on this scale” is provided. In other words, the science is going to be based on assuming no major long-term increase in global temperatures. So when a proxy reconstruction like Mann98 is produced that shows essentially no century-scale changes in temperature before the advent of large amounts of human-based CO2 production, you can legitimately do one of two things, but not both. You can say that Mann98 confirms the null hypothesis or you can say that the null hypothesis makes the findings of Mann98 reasonable. But if doubt is cast on the findings of low variability in global climate based on the procedures of Mann98 then either doubt is also cast on the null hypothesis or the null hypothesis tends to reject Mann98 after all. You can’t use the initial agreement of Mann98 and the null hypothesis to amelorate the later findings of statistical doubt.