Big update

[kate]

I can't reply to the thread with the ROMS announcement, but wanted to note one change to the boundary conditions:

* In the old way, ROMS would read the 'time' attribute on each boundary variable from the NetCDF file. This caused me grief when one of my tools added a C end-of-string character after the 'bio_time' or 'dye_time', but otherwise made sense. ROMS was ignoring the 'time' attribute set in varinfo.dat.

* In the new way, ROMS uses the 'time' attribute from varinfo.dat and ignores that in the NetCDF file. Check your varinfo.dat!

By the way, the report from a mismatch is not so very intuitive:

Code: Select all

 NETCDF_GET_FVAR_1D - error while reading variable:  zeta_west
                      in input file:  /wrkdir/kate/NEP5/Files/NEP5_bdry1_1985.nc
                      call from:  get_cycle.F

 GET_NGFLD   - error while reading variable: zeta_west   at TIME index =    0

 ERROR: Abnormal termination: NetCDF INPUT.
 REASON: NetCDF: Index exceeds dimension bound

This is because the time strings never match, so it assumes there are no records.

[arango]

Let me check this in the debugger. First, we want to use the time-variable specified in the time attribute. If the variable does not have a time attribute, we will use then the variable specified in varinfo.dat.

[kate]

I spoke too soon. It turns out ROMS is now depending on you having proper dimension coordinates. If your dimension is called bio_time, you'd better have a variable of the same name. Having a dimension of 'time' with an associated variable 'ocean_time' is a bad thing.

[arango]

Yes, this part of the CF compliance. ROMS follows the CF 1.0 metadata conventions.