date: Wed, 14 Mar 2001 09:40:46 +0800
from: Scott Rutherford <srutherford@virginia.edu>
subject: Re: results
to: Tim Osborn <t.osborn@uea.ac.uk>

Answers to questions:

>
>(1) I'm a little confused:
>crua6[/cru/u2/f055/data/paleo/tapio_recons] wc *
>          1       560      8961 1400-warm-mxd-nhmean.ascii
>        561    565040   9041201 1400-warm-mxd-recongrid.ascii
>        144    144271   2308480 verify-warm-mxd-grid.ascii
>          1       143      2289 verify-warm-mxd-nhmean.ascii

I haven't a clue what happened here but when I gunzipped the files 
and loaded them into matlab I had the same problem. I replaced the 
files gzipped them and gunzipped them and they worked ok. If it 
doesn't work this time I will put them there uncompressed.


>(2) The instrumental temperatures I'm using have the centre of each grid
>box at (e.g.) 177.5W, 77.5N.  Does your 180W, 80N refer to the top-left of
>the grid box, or are we really using slightly offset grids?

I go the data from http://www.cru.uea.ac.uk/cru/data/temperature/ and 
based on the information on the web site I took that to be top left 
corner.


>(3) I've just spotted some 'NaN' values in the file - are those the grid
>points at which there was never enough observed data to try to infill the
>values?  This makes it a little difficult to read in - though I guess I can
>just do a global edit and change NaN to some missing code.

That's correct.  There are a few gridpoints that have no data over 
the calibration period. In the warm season case, there are 6 
gridpoints.  Do you want me to change the NaN's first?


>(4) The values seem to have values as large as 170 or more.  What are the
>units?  Are they standardised anomalies or maybe anomalies in degrees C,
>but multiplied by some factor (maybe 100?).  What period are they anomalies
>from?  I assume from the the calibration period (1901-1960), but please
>confirm.

Yup, the values are all anomalies in C X 100.

>  >I'll leave the instrumental data in both of these so you can check
>>and make sure we are looking at the same gridpoints.  These are all
>>northern hemisphere warm-season, April-September.
>
>(5) Do you mean that those grid boxes that originally had an observed
>temperature value in them, at a particular year, still have that value?
>Presumably there are no such values in the verification run from 1856-1900
>- I guess that those are *all* infilled from the density only.  For the
>1400 run, I guess that *all* the values from 1856 onwards are
>instrumentally-based (though infilled), except for those few 'Nan' boxes
>that couldn't be infilled with either instrumental or MXD data.  Can you
>confirm all these things?


Everything you say is true.  The details are:
For the verification runs, all the available values from 1998-1901 
that were present in the original data are still there.  All values 
from 1856-1900 are imputed from the MXD proxies.  For the 1400 
reconstruction, all instrumental values from 1856 onward should be 
there unchanged.  Missing instrumental values have been infilled 
previously using the available data.

>(6) Is it possible to also send me the instrumental file from which you
>started the verification run (not the infilled version).  This is just to
>ensure that my observed version is the same (& same coverage) as the one
>you're using.

I placed on ftp joneswarmnhgrid.ascii which is the original raw data 
that I started with before filling anything. It is the April-Sept. 
average of the monthly data.  NaN's are missing values, format should 
be 143 X 1008 same as the others.


I hope that clarifies everything.  Please let me know if any of this 
still doesn't make sense or if you have additional questions.

In other news, I re-ran the multivariate verification stats using 
only the orginally available values in the calculations. It turns out 
it doesn't make too much difference. The verification stats typically 
are about the same either way, some go up, some go down.

Let me know if you need anything else.

Regards,

Scott

-- 
______________________________________________
		       Scott Rutherford

University of Virginia	University of Rhode Island
Environmental Sciences	Graduate School of Oceanography
Clark Hall			South Ferry Road
Charlottesville, VA 22903 	Narragansett, RI 02882
srutherford@virginia.edu	srutherford@gso.uri.edu
phone: (804) 924-4669	(401) 874-6599
fax: (804) 982-2137	(401) 874-6160
