Re: [ferret_users] Time axis

["Ansley C. Manke" <ansley.b.manke@xxxxxxxx>]

Hi, Arijeet,

You'll want to define a calendar axis, for instance here's a simple short one.  You could use the MODULO function to define

yes? define axis/t/t0=15-dec-2000/units=months taxis_apr_may = {4,5,16,17,28,29}
           *** NOTE: /UNIT=MONTHS is ambiguous ... using 1/12 of 365.2425 days

yes? show axis/t taxis_apr_may
 name       axis              # pts   start                end
 TAXIS_APR_MAY TIME             6 i   15-APR-2001 17:56    16-MAY-2003 16:03
T0 = 15-DEC-2000
   Axis span (to cell edges) = 26
 
       L     T                   TBOX      TBOXLO                TSTEP (MONTHS)
       1>  15-APR-2001 17:56:24  1          31-MAR-2001 12:41:51  4
       2>  16-MAY-2001 04:25:30  6          30-APR-2001 23:10:57  5
       3>  15-APR-2002 23:45:36  6          30-OCT-2001 14:05:33  16
       4>  16-MAY-2002 10:14:42  6          01-MAY-2002 05:00:09  17
       5>  16-APR-2003 05:34:48  6          30-OCT-2002 19:54:45  28
       6>  16-MAY-2003 16:03:54  1          01-MAY-2003 10:49:21  29

This will be fine for plotting - see the attached image

yes? let var2 = var[gt=taxis_apr_may@asn
yes? plot/symbol=19/siz=0.2/line/thick/color=blue var2


But if you're doing any computation or comparisons with data on another time axis, you'd need to be careful.  There are discussions of all this under "calendar" in the Ferret manual.

It happens that I was just talking with somebody who was doing something similar.  In the SHOW AXIS/T output above, look at the TBOX column.  Notice that the first timestep has a TBOX size of 1 month while the rest, up to the last have a box size of 6 months.  Ferret axes are contiguous, so that the upper grid-cell edge of each grid cell coincides with the lower cell limit of the next grid cell.  Here, each April cell (after the first one) starts at the end of October or start of November the previous year, and ends at the end of April; the May cell starts at the start of May (or end of April) or and ends at the end of October the next year.  For a regular axis like this, the "months" are equal fractions of a year, so the edges don't exactly meet up with the starts and ends of months.  Here's a little sketch of what is going on, with coordinate locations at ---*--- and grid-cell edges shown by ---|---

 
  A   M   J   J   A   S   O   N   D   J   F   M   A   M   J   J   A   S   O   N   D   J   F   M   A   M   J
|-*-|-*---------------------|---------------------*-|-*---------------------|---------------------*-|-*----



This will be fine for just plotting your data, but if you do something like regrid this to a time axis with every month represented, maybe to compare values with another dataset, you'd want to be careful to use a transformation such as @NRST which will not take data from the May grid cell that lies in September and put it onto the monthly grid. 

Ansley

On 5/16/2017 9:38 AM, Arijeet Dutta wrote:

Hi all

I have a time series from 2001 to 2012 and contains only april and may months of each year. I did the following to define the time axis which gives numeric value for the months on the axis

define axis/t=1:24:1 t24

var1 = var[gt=t24@asn]

But is there any way I can properly label the axis with april and may without using ANNOTATE?

Regards

Arijeet

Attachment: taxis.png
Description: PNG image