James,
I think I would assign each to midnight on the first of each month of the
current year, e.g., 2014-01-01 00:00:00, 2014-02-01 00:00:00,...
Easy to remember and keeps the months ordered in time-space. If your
calibration period was in a different year, it may be advisable to use
that year--you then could accommodate another calibration period in a
different year without disrupting your database.
Clint
Clint Bowman INTERNET: clint@xxxxxxxxxx
Air Quality Modeler INTERNET: clint@xxxxxxxxxxxxx
Department of Ecology VOICE: (360) 407-6815
PO Box 47600 FAX: (360) 407-7534
Olympia, WA 98504-7600
USPS: PO Box 47600, Olympia, WA 98504-7600
Parcels: 300 Desmond Drive, Lacey, WA 98503-1274
On Tue, 26 Aug 2014, James Adams - NOAA Affiliate wrote:
I want to store values in NetCDF as climatological statistics, and I haven't found a
good example or explanation of how this is best done for my use case.
For each lon/lat point in a grid I compute distribution fitting values for each
of the twelve months, and I want to store these values for later use when
running
various calculations on data sets that share the same grid. So for each lon/lat
point in the grid there are twelve values, one per month. The time values
therefore don't adhere to the normal conventions where any specific calendar
dates are applicable, but it's not clear from any of the documentation I've
seen how
you best represent generic month values as time coordinate axis values. For example how
do you represent "January" using a valid udunits for the time
dimension's/variable's units attribute? The data values I'm storing do have a pair of
associated "calibration period" dates associated with them, in that they're
computed from data beginning at one date and ending at another (i.e. what we refer to as
the "calibration period") -- maybe I just use the calibration period start
date within the udunits string, such as "days since <calibration_start_date>"?
For my purposes I can easily work with ordinal values (i.e. 0 = January, 1 =
February, etc.) but in order to conveniently use the data sets within my Java
code as GridDatasets it appears that the time coordinate variable needs to
adhere to
a conventional representation of time values with standard udunits, and it's not
clear what (if any) best practice for this is.
Can anyone advise as to how this is best represented in NetCDF for my use case?
Thanks in advance for any suggestions or insight.
--James
