Re: [ferret_users] Possible bug in Ferret 6.7 vs 6.64

[Kelin Zhuang <kzhuang@xxxxxxxxxxxxxxxxx>]

Hi all,
The present Ferret has trouble to read some CMIP5.

The problem now is not the curvilinear grid itself. We are missing
>  sufficient data series. Let me describe the data in detail:
>  For instance, hfds_Omon_NorESM1-M_rcp26_r1i1p1_200601-210012f.nc
>  name     title
>  I         J         K         L
>      LAT      latitude coordinate                            1:320
>  ...       1:384     ...
>      LON      longitude coordinate                         1:320
>  ...       1:384     ...
>      HFDS     Downward Heat Flux at Sea Water  1:320     ...       1:384 ...
>  We have only one data here. The data should span from 200601 through
>  210012, we should have 1140 hfds data but now we have only one. It is of
>  note that it is monthly data instead of average. Obviously we are lack
>  of enough data.
>  Things are similar to NorESM1-M historical, rcp45, and rcp85.
>  Another example is hfds_Omon_CNRM-CM5_rcp26_r1i1p1_209601-210012.nc
>  LAT      latitude coordinate              1:362     ...       1:292     ...
>      LON      longitude coordinate             1:362     ...
>  1:292     ...
>      LAT_VERTICES
>                                                1:362     1:4
>  1:292     1:362
>      LON_VERTICES
>                                                1:362     1:4
>  1:292     1:362
>      HFDS     Downward Heat Flux at Sea Water  1:362     ...       1:292 ...
>  We should have 60 hfds (5yearX12month=60) but we only have one hfds
>  here. Things are similar to CNRM-CM5 historical, rcp45, and rcp85. We
>  are lack of enough data again.
>  Conversely, let us have a good data here with a curvilinear grid. Let's
>  say MIROC5.
>  hfds_Omon_MIROC5_rcp26_r1i1p1_200601-210012.nc
>      name     title
>  I         J         K         L
>      LAT      latitude coordinate                     1:256     1:224
>  ...       ...
>      LON      longitude coordinate                 1:256     1:224
>  ...       ...
>      LAT_VERTICES
>
>  1:4       1:256     1:224     ...
>      LON_VERTICES
>                                                                   1:4
>  1:256     1:224     ...
>      HFDS     Downward Heat Flux at Sea Water  1:256     1:224     ... *1:1140*
>  Look at L(1:1140). We are now having the expected 1140 hfds data
>  (95yearX12month=1140).

And also in tos files I met with the same question. Even Ferret 6.72 can read the data now but read latitude mistakenly. Hope the next update will correct these bugs.

Kelin

On 02/09/2012 09:58 AM, Ansley Manke wrote:
> Hi all,
> Gary has had a chance to test a beta version of Ferret which includes
> some further work on this issue of creating the grids for variables, and
> he's able to read the datset successfully in Ferret. This will be
> included in the next Ferret release, and will keep us going until the
> extra-dimensional Ferret is ready to go.
>
> Ansley
>
>
> On 2/9/2012 8:12 AM, Ansley Manke wrote:
> > Hi,
> > In addition to all of this, we've been working with Patrick Brockmann
> > on some similar examples. I'll talk with Gary off-line and see if we
> > can get Ferret working with his files.
> >
> > In addition, we have plans for adding dimensions to Ferret, to make it
> > 6-dimensional.  This will allow for an ensemble dimension, or say a
> > forecast time dimension. It will let us handle these variables with
> > grids that have a dimension "before" the X dimension; in Gary's
> > original report, older Ferret executables created these grids from
> > what is in the dataset:
> >
> >   LAT      latitude coordinate              1:320     1:384
> > ...       ...
> >   LON      longitude coordinate             1:320     1:384
> > ...       ...
> >   LAT_VERTICES                              1:4       1:320
> > 1:384     ...
> >   LON_VERTICES                              1:4       1:320
> > 1:384     ...
> >
> >
> > which indicates that if the LON and LAT variables have grids of (x,y),
> > then the vertices variables have grids which are something like
> > (v,x,y).  Ferret's grids don't have a slot for a dimension to the left
> > of X.
> >
> > Ansley
> >
> > On 2/9/2012 12:00 AM, Hein Zelle wrote:
> > > Dear Gary Strand,
> > > > I'm getting an error from Ferret 6.7 that doesn't occur when using
> > > > Ferret 6.64. This is a straightforward netCDF file that happens to
> > > > have a field defined on a curvilinear grid that meets CMIP5
> > > > requirements.
> > > Have you tried ferret 6.7.2?  There were relevant changes in 6.7,
> > > 6.7.1: improvements in recognizing axes names and dealing with
> > > dimensions.  These changes initially broke some things for us with
> > > specific variables in WRF curvilinear files, very similar to what
> > > you're seeing.  Since ferret 6.7.2, this works "as expected" for us,
> > > again.
> > >
> > > Could you post an ncdump -h of your file?  I'd like to see the
> > > attributes of the LEV dimension / variable.
> > >
> > > Kind regards,
> > >        Hein Zelle
> > >
> > >
> > > > Under Ferret 6.64:
> > > >
> > > >           NOAA/PMEL TMAP
> > > >           FERRET v6.64
> > > >           Linux(gfortran) 2.6.9-89.0.20.ELsmp - 09/16/10
> > > >            8-Feb-12 08:48
> > > >
> > > > yes? use thkcello_fx_CCSM4_lgm_r0i0p0.nc
> > > >    *** NOTE: Units on axis "j" are not recognized: 1
> > > >    *** NOTE: They will not be convertible:
> > > >    *** NOTE: Units on axis "i" are not recognized: 1
> > > >    *** NOTE: They will not be convertible:
> > > >    *** NOTE: If no missing_value or _FillValue attribute on
> > > > variables, will use NaN
> > > > yes? show d
> > > >        currently SET data sets:
> > > >       1>   ./thkcello_fx_CCSM4_lgm_r0i0p0.nc  (default)
> > > >    name     title                             I         J
> > > > K         L
> > > >    LAT      latitude coordinate              1:320     1:384
> > > > ...       ...
> > > >    LON      longitude coordinate             1:320     1:384
> > > > ...       ...
> > > >    LAT_VERTICES
> > > >                                              1:4       1:320
> > > > 1:384     ...
> > > >    LON_VERTICES
> > > >                                              1:4       1:320
> > > > 1:384     ...
> > > >    THKCELLO Ocean Model Cell Thickness       1:320     1:384
> > > > 1:60      ...
> > > >
> > > > Under Ferret 6.7:
> > > >
> > > >           NOAA/PMEL TMAP
> > > >           FERRET v6.7
> > > >           Linux(gfortran) 2.6.32-71.24.1.el6.x86_6 - 05/06/11
> > > >            8-Feb-12 08:50
> > > >
> > > > yes? use thkcello_fx_CCSM4_lgm_r0i0p0.nc
> > > >    *** NOTE: Units on axis "j" are not recognized: 1
> > > >    *** NOTE: They will not be convertible:
> > > >    *** NOTE: Units on axis "i" are not recognized: 1
> > > >    *** NOTE: They will not be convertible:
> > > >    *** NOTE: Axis LEV marked as UD
> > > >    *** NOTE: but it has different directions in grids. Axes in grids
> > > > may be inconsistent.
> > > >    *** NOTE: Axis LEV marked as UD
> > > >    *** NOTE: but it has different directions in grids. Axes in grids
> > > > may be inconsistent.
> > > >    *** NOTE: Axis LEV marked as UD
> > > >    *** NOTE: but it has different directions in grids. Axes in grids
> > > > may be inconsistent.
> > > >    *** NOTE: Axis LEV marked as UD
> > > >    *** NOTE: but it has different directions in grids. Axes in grids
> > > > may be inconsistent.
> > > >    *** NOTE: Axis LEV marked as UD
> > > >    *** NOTE: but it has different directions in grids. Axes in grids
> > > > may be inconsistent.
> > > >    *** NOTE: Axis LEV marked as UD
> > > >    *** NOTE: but it has different directions in grids. Axes in grids
> > > > may be inconsistent.
> > > >    *** NOTE: Axis LEV marked as UD
> > > >    *** NOTE: but it has different directions in grids. Axes in grids
> > > > may be inconsistent.
> > > >    *** NOTE: Axis LEV marked as UD
> > > >    *** NOTE: but it has different directions in grids. Axes in grids
> > > > may be inconsistent.
> > > > yes? show d
> > > >        currently SET data sets:
> > > >       1>   ./thkcello_fx_CCSM4_lgm_r0i0p0.nc  (default)
> > > >    name     title                             I         J
> > > > K         L
> > > >    LAT      latitude coordinate              1:320     ...
> > > > 1:384     ...
> > > >    LON      longitude coordinate             1:320     ...
> > > > 1:384     ...
> > > >    LAT_VERTICES
> > > >                                              1:320     1:4
> > > > 1:384     1:320
> > > >    LON_VERTICES
> > > >                                              1:320     1:4
> > > > 1:384     1:320
> > > >    THKCELLO Ocean Model Cell Thickness       1:320     ...
> > > > 1:384     1:60
> > > >
> > > > Help!
> > > >
> > > > The file is available via anonymous ftp at
> > > >
> > > > ftp://ftp.cgd.ucar.edu/pub/strandwg/ferret/thkcello_fx_CCSM4_lgm_r0i0p0.nc
> > > >
> > > >
> > > > Gary Strand
> > > > strandwg@xxxxxxxx