[ncl-talk] calendar_decode2: is this a bug?

[Dave Allured - NOAA Affiliate]

Stefan,

It is not an NCL bug.  I would call it a design problem.

Since months in the real calendar have different lengths, a unit
specification of "months since ..." is ambiguous.  Several NCL calendar
support functions are based on the UDUNITS software package, and UDUNITS
has chosen a fixed non-integer value for the length a month.  This is the
reason for your odd result values.

Please refer to the ut_calendar function documentation, in the section
"Caveats of UDUNITS-2", for more details.  It starts by saying:

   "The length of a month is fixed at 1/12 of a tropical year
   or 2629743.831225 seconds."

This is the same issue for both UDUNITS-1 and UDUNITS-2 versions.  I think
this is also discussed somewhere in UDUNITS original documentation:

   http://www.unidata.ucar.edu/software/udunits/

But please notice, this is not really a UDUNITS problem either.  In my
opinion, best practice is to never use variable length base units such as
month and year.  Such usage is dangerous and leads to much user confusion
and subtle errors.  Stick with "days since", "hours since", etc. to avoid
problems.

There are artificial "model calendars" with fixed length months or years,
e.g "360_day".  Only in these calendar systems does it make sense to use
"months since" or "years since".  HTH.

--Dave A.
Author of calendar_decode2

On Wed, Nov 19, 2014 at 2:00 PM, Stefan Muthers <ncl at z107.de> wrote:

> Dear all,
>
> I have a problem with calendar_decode2 and do not know whether this is a
> bug or whether I maybe missunderstand something.
>
> Consider the following example:
>
> --------------------------
>
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/gsn_code.ncl"
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/gsn_csm.ncl"
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/contributed.ncl"
> load "$NCARG_ROOT/lib/ncarg/nclscripts/contrib/calendar_decode2.ncl"
>
> a=1
> a at units = "months since 1600-01-01 00:00:00"
> printVarSummary(a)
> ;
> ;Variable: a
> ;Type: integer
> ;Total Size: 4 bytes
> ;            1 values
> ;Number of Dimensions: 1
> ;Dimensions and sizes:   [1]
> ;Coordinates:
> ;Number Of Attributes: 1
> ;units :       months since 1600-01-01 00:00:00
>
> print(""+calendar_decode2(a, -5))
> ;(0,0)   1600
> ;(0,1)   1
> ;(0,2)   31
> ;(0,3)   10
> ;(0,4)   29
> ;(0,5)   3
>
> --------------------------
>
> I would expect the calendar_decode2 function to return 1600-02-01. Why
> is 1600-01-31 returned here?
>
> thanks you and best regards,
>  Stefan
>
>
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20141119/0c6d1f19/attachment.html