cc: "Collins, Matthew" , Keith Briffa date: Wed, 1 Feb 2006 11:16:23 +0000 from: Thomas Kleinen subject: Re: HadCM3 control run results to: Tim Osborn Hi Tim and Matt. > (1) On the comparison with observations for near-surface temperature, > you get large "errors" over winter sea ice regions. It might be > worth checking what the HadCRUT climatology represents over the > oceans - is it marine air temperature or SST. If the latter, then it > won't fall below -2 deg C and this would explain the big "errors" > when comparing with HadCM3 1.5m air temperature. Even if this is not > the case, and HadCRUT climatology purports to be marine air > temperature, then I wouldn't believe the values over sea ice areas > anyway! They'll have been extrapolated from a very few data points > or only from Arctic buoys that didn't cover the whole 1961-90 period. Supposedly it is air temperature, according to the documentation I could find (our data webpage isn't documented all THAT well), but I am not too worried about that anyway. So far I am mainly concerned with having a model configuration that gives output similar to the MetOffice's one. And that it seems to do. > (2) For the UEA control minus Hadley control results, does the > difference pattern stay constant between, say, the early part of the > UEA run and the later part. This would imply a systematic > problem. So how does UEA(30-59)-HAD(3160-3189) for DJF SLP compare > with the UEA(60-89)-HAD(3160-3189) pattern that you showed? I guess > the statistical testing you have done implies that it is sampling > variability rather than systematic, but it'd still be nice to see how > the two periods' results compared for DJF SLP. No, those differences are random. The pattern doesn't stay contant, from what I've seen. We'll have more data next week, so I can do some more checks, but my main concern right now is whether our ocean is ok. More on that next week. Cheers, Thomas