[Thread Prev][Thread Next][Index]

Re: Beware! Time assignment change from older versions of Ferret



Hi everyone,
After a few offline exchnages with Lynn, we've come to an understanding
of the files Lynn brought up.  Thanks also to Wolfgang for reporting what
you see.

There is a standard in use by NOAA/CDC, which applies the 2-day offset
when the time origin is T0 = 1-jan-0001:00:00, and the data starts after the
year 1582.   I believe this is the case with the Reynolds and NCEP data
Lynn mentions.  Version 5.4 of Ferret reads these files correctly, as does
Version 5.22 (5.33 apparently had a bug that caused it not to apply the
2-day offset.)  This calendar is a "blended" Julian/Gregorian calendar.

When the time origin is other than 1-jan-0001 Ferret uses a "proleptic"
Gregorian calendar, that is it uses the current standard calendar for all
time.  Thus the definition of a year is consistent for all time with no
offsets.

Lynn notes that both the NCEP and Reynolds data files indicate that they
are COARDS compliant in their header info.  The COARDS convention
says that the time variable should follow the udunits convention.  This
means that Ferret V 5.22 and V 5.4 give the "correct" times for these files.
So if those who create files that are COARDS compliant actually follow
through and use the udunits calendar one would know that Ferret 5.22
or 5.4 reads them correctly.

Ansley Manke

Lynn deWitt wrote:

> Ferret users,
>
>      While writing up this posting I found the answer in the mail archives at:
>
> http://ferret.wrc.noaa.gov/Ferret/Mail_Archives/fu_99/tmp/msg00060.html
>
> but thought I'd post it anyway in case anyone else is having similar
> trouble --- and as a warning to long-time users who still have work
> around that they've done in older versions.  Note that the following
> only applies to files with reference start times before about the
> year 1500, but some commonly used files, such as some Reynolds sst
> files, have reference times like this.
>
>       I was trying to update some work I did a few years ago, and was
> having trouble reproducing the previous work.  I finally tracked the
> problem to a behavior change in assigning time stamps between v 4.91
> (which I was using at the time I did the previous work) and v 5.22
> which I'm using now.
>
>      Compare the listings below done on the same Reynolds sst file
> (downloaded from cdc.noaa.gov) using exactly the same commands but
> with different versions of Ferret.   The sst listed at each grid
> point in time is consistent, but the assigned day-month-year is
> different by two days between the two versions of Ferret.   For
> example (see complete listing and commands below) for identical grid
> points:
>
>        V 4.91 :                                       V5.22:
>   01-DEC-1984 00 / 162:  16.81  29-NOV-1984 00 / 162:  16.81
>   08-DEC-1984 00 / 163:  16.61  06-DEC-1984 00 / 163:  16.61
>   15-DEC-1984 00 / 164:  15.75  13-DEC-1984 00 / 164:  15.75
>
>    This two day change had a fairly significant affect on the results
> when performing various averaging!  The answer, as I interpret Steve
> Hankin's e-mail from the archives, is that they made a change in
> version 5.0 to be more consistent with the udunits convention.  This
> affects files with reference start times before about the year 1500.
> The particular Reynolds sst file I was using shows the following in
> and ncdump:
>
>          double time(time) ;
>                  time:units = "days since 1-1-1 00:00:00" ;
>                  time:long_name = "Time" ;
>                  time:actual_range = 723483., 726465. ;
>                  time:delta_t = "0000-00-07 00:00:00" ;
>                  time:avg_period = "0000-00-07 00:00:00" ;
>
> Now I just need to figure out whether the v 4.91 calendar (Gregorian)
> or the v5.0 (udunits) gives me the correct calendar day for the
> Reynolds SST datasets so I can combine the data in analyses with
> other data sets.  I would assume they use udunits.  Anyone know?
>
> Anyway, beware!
>
>     Lynn
>
> ----------Listings using v 4.91----------
>
>          NOAA/PMEL TMAP
>          Program FERRET (GUI)
>          Version 4.91 - 05/19/98
>          28-Feb-02 12:10
>
> yes? set data "Reynolds_sst/sst.wkmean.1981-1989.nc"
> yes? set region/i=231/j=126
> yes? list sst[t=1-dec-1984:1-feb-1985]
>               Weekly Means of Sea Surface Temperature (degC)
>               LATITUDE: 35.5N
>               DATA SET:
> /disk04/products/Derived/Reynolds_sst/sst.wkmean.1981-1989.nc
>                          129.5W
>                          231
>   01-DEC-1984 00 / 162:  16.81
>   08-DEC-1984 00 / 163:  16.61
>   15-DEC-1984 00 / 164:  15.75
>   22-DEC-1984 00 / 165:  15.51
>   29-DEC-1984 00 / 166:  14.93
>   05-JAN-1985 00 / 167:  14.98
>   12-JAN-1985 00 / 168:  14.83
>   19-JAN-1985 00 / 169:  15.06
>   26-JAN-1985 00 / 170:  14.35
>   02-FEB-1985 00 / 171:  14.06
>
> yes? list t[gt=sst,l=1:10]
>               T
>               axis TIME
>               DATA SET: Reynolds_sst/sst.wkmean.1981-1989.nc
>   31-OCT-1981 00 /  1:  723483.
>   07-NOV-1981 00 /  2:  723490.
>   14-NOV-1981 00 /  3:  723497.
>   21-NOV-1981 00 /  4:  723504.
>   28-NOV-1981 00 /  5:  723511.
>   05-DEC-1981 00 /  6:  723518.
>   12-DEC-1981 00 /  7:  723525.
>   19-DEC-1981 00 /  8:  723532.
>   26-DEC-1981 00 /  9:  723539.
>   02-JAN-1982 00 / 10:  723546.
> yes?
>
> ----------Listings using v 5.22----------
>
>          NOAA/PMEL TMAP
>          Program FERRET
>          Version 5.22 - 08/23/00
>          28-Feb-02 12:17
>
> yes? set data "Reynolds_sst/sst.wkmean.1981-1989.nc"
> yes? set region/i=231/j=126
> yes? list sst[t=1-dec-1984:1-feb-1985]
>               Weekly Means of Sea Surface Temperature (degC)
>               LATITUDE: 35.5N
>               DATA SET: Reynolds_sst/sst.wkmean.1981-1989.nc
>                          129.5W
>                          231
>   29-NOV-1984 00 / 162:  16.81
>   06-DEC-1984 00 / 163:  16.61
>   13-DEC-1984 00 / 164:  15.75
>   20-DEC-1984 00 / 165:  15.51
>   27-DEC-1984 00 / 166:  14.93
>   03-JAN-1985 00 / 167:  14.98
>   10-JAN-1985 00 / 168:  14.83
>   17-JAN-1985 00 / 169:  15.06
>   24-JAN-1985 00 / 170:  14.35
>   31-JAN-1985 00 / 171:  14.06
>
> yes? list t[gt=sst,l=1:10]
>               T
>               axis TIME
>               DATA SET: Reynolds_sst/sst.wkmean.1981-1989.nc
>   29-OCT-1981 00 /  1:  723483.
>   05-NOV-1981 00 /  2:  723490.
>   12-NOV-1981 00 /  3:  723497.
>   19-NOV-1981 00 /  4:  723504.
>   26-NOV-1981 00 /  5:  723511.
>   03-DEC-1981 00 /  6:  723518.
>   10-DEC-1981 00 /  7:  723525.
>   17-DEC-1981 00 /  8:  723532.
>   24-DEC-1981 00 /  9:  723539.
>   31-DEC-1981 00 / 10:  723546.
>
> yes?
>
> --------------------------------------------
> Lynn deWitt
>
> ldewitt@pfeg.noaa.gov
>
> Pacific Fisheries Environmental Laboratories
> 1352 Lighthouse Avenue
> Pacific Grove, CA 93950-2097
> (831)-648-9036

--
Ansley Manke  Pacific Marine Environmental Laboratory  Seattle WA  (206)526-6246




[Thread Prev][Thread Next][Index]

Dept of Commerce / NOAA / OAR / PMEL / TMAP

Contact Us | Privacy Policy | Disclaimer | Accessibility Statement