cc: Phil Jones date: Wed, 7 May 2008 13:27:18 +0100 from: Ian Harris subject: Re: wetdays... yippee! to: Tim Osborn Hi Tim, On 7 May 2008, at 13:03, Tim Osborn wrote: > Hi Harry and Phil, > > Not a cry for some rain... instead relief that we now appear to > have a usable wetday file. :-) > > The patterns of means, standard deviations and correlations with > CRU TS 2.1 data all look reasonable (obviously with a slightly > lower correlation over some, e.g. Russian, regions because 2.1 has > the step-change in wetdays due to the data inhomogeneity). > > The country-mean, seasonal time series are attached. The step in > 2.1 (the red lines) is clearly avoided in the new 3.0 (the black > lines). > > The version I've used is the usual CRU TS approach from 1901-1989 > -- i.e., the results of gridding the station wetday data together > with the 2.5-degree grid of synthetic wet day -- and then switching > to completely synthetic wet days calculated directly on the 0.5- > degree grid from the 0.5-degree pre, for 1990-2005 (I've ignored > 2006, but the same should be done for that and beyond, until the > post-1989 station wetday data is sorted out at some future time). > > Assuming you're both happy with the attached country-mean time > series, then can we settle on this as *the* CRU TS 3.0 wetday data > set? I expect that's pragmatic; but I'm not sure Canada looks right, and the dry African states/countries have a different unrealistic pattern to 2.1.. > In which case, I can give it to our QUEST partners, and Harry can > you make up a new 1901-2006 pair of files containing this final > combination? Currently I'm reading the separate files and > combining them as described above, but a single file would be > easier for me (and others) to deal with. Yup. > Well done for your work on all these multiple iterations/approaches > Harry. > > Assuming you're happy with wetdays, then that means all CRU TS 3.0 > is done except CLD and FRS (the latter is made but I've not checked > it, so its not impossible that there's some factor-of-10 error). Agreed entirely. > Harry -- FYI Allan Spessa has asked overall QUEST (i.e. not our > particular QUEST project) if they'll pay a month's salary to us to > help us get CLD working. As you noted before, it is not just down > to money, but rather to time, given the demands of our QUEST > project for future scenarios now. If QUEST do find the money, > therefore, then I'll ask Douglas to do the work on recreating the > coefficients that Mark New/Tim Mitchell couldn't find for making > synthetic CLD. That seems an easily separable task from the actual > gridding stage. Once synthetic CLD is made, you would then of > course be involved in the gridding stage -- though if you were > documenting the CRU TS process, it might prove to be a useful trial > run for Douglas to attempt to follow your instructions and see if > they can get him through the process? One thing I'm not sure about > is the status of the station data base for CLD... assuming we're > not basing the updates (post-2002) entirely on synthetic data, then > presumably station CLD (estimated mostly -- or entirely? -- from > station sunshine) updates are needed. Are these done already? If > not, do you have code ready to do this sunshine-->cloud > conversion? Am I barking up the wrong tree? It's very complicated, there's a whole raft of related programs I'm afraid, and dozens of data files. I can't remember what stage I got to but I do have notes. And Douglas is an excellent choice ;-) > Sorry for the long email... nearly there! > > Thanks for all your hard work on this Harry. I'd like to be able > to let you take a bit of a break and catch up with non-project > things. Unfortunately, after 6 weeks hard work from both of us, > we've now reached the stage that I already thought we'd reached at > Easter, when I was panicking about how far behind we were! :-( I did guess. I was going to ask for a week to sort out a non-IDL gridding package ;-) > So... when you've made the final wet files, please can you get back > to the climate change pattern diagnosis from the GCM runs. You'll > probably need to get up to speed with where you'd got to. I have > patterns for 'tas' (i.e., near-surface mean temperature) and > 'pre' (including relative and exponential ones) from you. For > those GCMs with min and max temp data, could you follow the same > procedure as for 'tas', to make patterns for tmin, tmax and also > repeat it using tmax-tmin to get patterns for dtr. Understood. Will spin up the disks and re-read the project diary! > Then I'll give you some guidance for diagnosing precip variability, > cloud and vapour pressure patterns as they differ a bit from the > other variables. OK. I will still have to fiddle with the automation of the TS process though; I'll do that in non-work time. As you'll have seen, BADC are getting edgy, and high-level interrupt routines may be kicking in shortly! Cheers H Ian "Harry" Harris Climatic Research Unit School of Environmental Sciences University of East Anglia Norwich NR4 7TJ United Kingdom