date: Thu Apr 27 14:01:28 2000 from: Mike Hulme subject: schneider's comments to: new Mark, Some useful comments from Schneider (although at times he lapses into a Wigleyesque view of history, i.e., there is nothing new under the sun and I did it first anyway!). p.1: I am happy to qualify what is meant by "unknowable" (Hulme and Carter put this term in quotes, which at least signifies we did not mean it literally). But I accept SS's main thurst here. p.2: I leave you to decide on Figrue 1 p.3: fine, let's include 1-2 references on Bayesian methods p.4: I have checked and Moss/Schneider did *not* cite 1.5-4.5 as the 90% range. They recommended confidence intervals be cited (e.g. 90%) and then elsewhere discussed the 1.5-4.5 range. My memory put the two together. Let's quote Morgan and Keith which is a famous study. A 90% range is not unreasonable, but it is our judgement. p.5: not sure this is the point. The evolution of future climate is difficult because signal/noise are entangled. But in principle is the signal quasi-linear (assumption of pattern-scaling methods) or not? Even if it is, then the precise evolution of future climate will *still* be difficult; and if it is not, then it will be harder still and pattern-scaling methods may on occasions be quite misleading. p.7: not sure I agree. T-P correlations or not fall out of fundamental physics/synoptics, many of which will not change. Seems highly unlikely to me that under GHG forcing, say, UK cold/dry-mild/wet winter regime will change to cold/wet-mild/dry regime. p.7, bottom: agree that a caveat could be put here. Of course, 2100 is the IPCC time-scale which we follow and our main thrust is not to discuss the 'danger' issue. That should be left for another time. p.10: these are subjective judgements. Perhaps better to say that these events are low probability, but not known to an order of magnitude (i.e., could be 10%, 1% or 0.1%). We need for formal methods to quantify these. I haven't got the references here - perhaps Pim could send them or else Oxford's library might help. Hope this helps. Mike