date: Thu, 11 Mar 2004 17:00:30 +0200
from: Timothy Carter <tim.carteratXYZxyzaristo.fi>
subject: Re: Vapour pressure scenarios
to: Phil Jones <p.jonesatXYZxyz.ac.uk>
Thanks for the prompt reply. I am copying this to Dave Wilson and the ATEAM
co-ordinators. I suspect there may be some others in ATEAM who are using
Penman and haven't looked into the humidity changes in this detail. It's
probably too late to expect them to do new runs. One recommendation might
be simply to assume no change in VP, but this would introduce internal
inconsistency in the scenarios (though that's been the practice in many
applications before humidity scenarios were being provided to impact
Apart from the weaknesses in the GCMs themselves, I wonder if the empirical
conversion methods used to derive VP from temperature/SLP have contributed
to some of the rather odd results. Let's see what Tim has to say.
At 15:26 11/03/04, you wrote:
> Tim and Mike,
> I've sent an email to Tim Mitchell for his thoughts (and asked him
> what the new job is like).
> I'm not surprised by what you've found - i.e. the large inter-model
> differences. In the EU-project
> SWURVE, we've gone back to calculating PET (assuming this is why you
> want a humidity
> type variable) with Thornthwaite and Blaney/Criddle as they only depend
> on temperature.
> This is being written into project final report and the special
> issue of HESS (Hyd. and Earth
> System Science). Project run by Chris Kilsby and he's arranged this
> issue. Even with HadCM3
> with small changes in vapour pressure (well in HadAM3P/HadRM3P - same
> there also), the
> increasing temperature means that vapour pressure deficit becomes very
> large, so PET
> calculated with Penman formula is ridiculous.
> If this is why you want vapour pressure I would suggest you go down
> this route also.
> Happy for you to forward this to Nigel as he'll understand what I'm on
> about. Hydrologists
> know that Penman should be best, but not with models. Even for 1961-90
> the problem can
> be seen in the warmer summers.
> Basic problem is that all models are wrong - not got enough middle
> and low level clouds.
> Problem will be with us for years, according to Richard Jones. Chris has
> talked to him about
> it at length. It looks as though CSIRO2 may be the best one. CGCM2 looks
> most odd.
> The HC think their variable tile parameterization may help. This can
> keep some small
> portion of open water in each box, so the whole thing doesn't dry out.
> There was a paper in Science a year or so ago, that showed PET (from
> going down recently in many regions !
> I'll let you know what Tim thinks. Omitted the two pdfs as they were
> large. The ppt plot
> gives the essence of the message.
> I'm assuming here that Tim hasn't made a mistake - the HadCM3 plots
> look like the
> ones Declan produced for SWURVE a while ago and similar to ones Marie
> has produced
> for RM3P and AM3P.