date: Fri, 21 Nov 2008 15:46:03 -0000
from: "Nigel Arnell" <n.w.arnell@reading.ac.uk>
subject: RE: ClimGen vap
to: "Tim Osborn" <t.osborn@uea.ac.uk>

Tim,

Thanks - I'll digest this on the way to China.

Thanks for your input to the meeting on Wednesday - the meeting was
extremely helpful, and I'm really pleased with progress and direction.
One immediate question - when do you think you could have the extra
scenario data (or code+pattern)? It influences the timing of the rest of
the project.

Regards

Nigel 

Professor Nigel Arnell
Director
Walker Institute for Climate System Research
University of Reading
Earley Gate
Reading
RG6 6BB
UK
 
+44-118-378-7392
 
www.walker-institute.ac.uk

-----Original Message-----
From: Tim Osborn [mailto:t.osborn@uea.ac.uk] 
Sent: 21 November 2008 15:42
To: Nigel Arnell
Subject: ClimGen vap

Hi Nigel,

sorry I had to leave Wednesday's meeting early, but please let me know
anything that arose after I left (or at the stakeholder meeting) that
needs some input from me.

Regarding the VAP scenarios for those GCMs that haven't provided surface
humidity output...

Harry has now completed his calculations, using specific humidity on
pressure levels in the atmosphere, together with the surface pressure
(note: different from sea level pressure over the land, of course).
Surface pressure is used to find the two lowest pressure levels that are
above the land surface, and then to extrapolate from the spec. hum. at
these two levels down to the pressure at the land surface, and then
convert this estimated surface spec. hum. to surface vapour pressure.
This then gets put into the pattern estimator (pooling across multiple
ensemble members, averaging into 30-year running means, and regressing
against 30-year running mean global temperature change).

You probably don't need to know all this detail!  Instead, please look
at the attached PDF which compares our results for the
cccma_cgcm3_1 model where we have both surface and pressure level
humidity data -- so we can compare the estimated patterns with the
correct surface patterns.

One page per season.

Top-left is the correct VAP change pattern (hPa per degC of warmning)
calculated from model surface humidity (called "v2p").

Top-right is "vaplev" (our estimate from the pressure level humidity).

Bottom-left is the difference (vaplev-v2p).  Green is essentially zero
difference, light blue and grey are small differences.  The browns and
pinks are those of most concern.

Bottom-right is a scatter plot showing v2p vs. vaplev values for all
grid boxes, plus the correlation and slope of a best-fit (least squares
regression) line.  The black line is the perfect y=x line, while the
blue line is the best-fit line.

I've only shown the land -- the fit is even better over the oceans.
Over the land, the pattern correlations are 0.95, 0.93, 0.90, and 0.93
for the four seasons.  Very good, but the scatter plots and the
difference maps show some significant variations.  Most problems are
where vaplev has underestimated the correct changes (points above the
line in the scatter plot).

I've been trying to think why the extrapolation from levels would fail,
but can't seem to get an explanation that fits the pattern.  For
example, some of most extensive differences are over the Mediterranean
in JJA.  Here this model shows strong summer drying, which leads to
decreased soil moisture.  If anything I'd expect that the drier soils
would mean that near-surface humidity would increase less than away from
the surface at the pressure levels, yet the surface VAP patterns shows
more increase than the pattern estimated from the pressure level data!

What is your opinion about this?  Can we use vaplev estimates on the
basis of the >0.9 pattern correlations?  Are they better than the
alternative(s), e.g. the constant RH assumption?

Best regards

Tim






No virus found in this incoming message.
Checked by AVG - http://www.avg.com
Version: 8.0.175 / Virus Database: 270.9.8/1800 - Release Date:
11/19/2008 6:55 PM
