Re: climatology question

[Jonathan Callahan <callahan@xxxxxxxxxxxxx>]

All,

I'd just like to point out that NOAA's Climate Diagnostics Center has
done a ton of work creating derived products from the NCEP Reanalysis
data and all of it is available to Ferret users via DODS.  Here are some
quick instructions on how to find what you want and how to use it in
Ferret but you'll have to poke around a bit when you're at their web
site.

1) Here is where I start poking around looking for datasets:

http://www.cdc.noaa.gov/Datasets/

2) Here is a file that looks promising  (But don't click on the link!):

http://www.cdc.noaa.gov/Datasets/ncep/air.day.ltm.nc

I guess that this is file contains "AIR temp, DAilY Long Term Mean"

3) Now I insert "cgi-bin/nph-nc/" to create a DODS URL :

http://www.cdc.noaa.gov/cgi-bin/nph-nc/Datasets/ncep/air.day.ltm.nc

4) Tell Ferret to read this file (you may have to wait a few seconds):

yes? use
"http://www.cdc.noaa.gov/cgi-bin/nph-nc/Datasets/ncep/air.day.ltm.nc";
 *** NOTE: Coordinates out of order or missing on axis level at
subscript 2
 *** NOTE: A dummy axis of subscripts will be used

5) Now go ahead and use it:

yes? show data
     currently SET data sets:
    1>
http://www.cdc.noaa.gov/cgi-bin/nph-nc/Datasets/ncep/air.day.ltm.nc 
(default)
 name     title                             I         J        
K         L
 AIR      Long term mean daily air temper  1:144     1:73      1:13     
1:365
 LEVEL    Level                            1:13      ...       ...      
...
       (invalid coordinate axis)

yes? shade air[k=11,l=200]
...
 

Take home message:  DODS works well and the CDC has done a lot of work
for the rest of us.


-- Jonathan Callahan



Billy Kessler wrote:
> 
> Use climatological_axes to make a monthly climatology
> (average annual cycle). Then regrid back to daily.
> After all, unless you have REALLY a lot of years, a
> daily climatology does not make much sense. For example,
> the average of 43 November 22nds (1957-99) does not give
> a good estimate of the climatological November 22nd.
> The chances are it will differ significantly from the
> average Nov 21st over those 43 years, due to not having
> a sufficiently long record to average over the randomness
> of weather. A better estimate is formed by linearly
> interpolating the monthly averages to daily.
> 
> Do this by:
> 
> yes? let myvar_cl = myvar[gt=month_reg@mod]
> yes? let myvar_cl_daily = myvar_cl[gt=myvar]
> 
> Now myvar_cl_daily has exactly the same structure as
> the original myvar, which is useful in that you can
> subtract to get daily anomalies from the average
> annual cycle:
> 
> yes? let myvar_anom = myvar - myvar_cl_daily
> 
> Billy K