Hi. As far as I have ever been able to tell from CF (pre 1.6), there has never been a proscription of missing values in auxiliary coordinates. True coordinate variables (as I think of them) contain definitions of independent axes, which are not acquired data. As a result, it makes no sense for them to have missing values. Auxiliary coordinate variables may very well contain acquired data values, and properly able to have elements marked as missing, independent of whether or not other variables are missing values for the corresponding elements.
Grace and peace, Jim Visit us on Facebook Jim Biard Research Scholar Cooperative Institute for Climate and Satellites NC North Carolina State University NOAA's National Climatic Data Center 151 Patton Ave, Asheville, NC 28801 e: [email protected] o: +1 828 271 4900 On Feb 6, 2014, at 11:42 AM, Jonathan Gregory <[email protected]> wrote: > Dear Rich and Seth > >> In this dataset, lat and lon aren't coordinate variables, they're >> auxiliary coordinate variables. There's no requirement that auxiliary >> coordinate variables be monotonic or lacking in missing_values. Your >> coordinate variables will be the ones associated with the I and J >> dimensions of the arrays. >> >> To make this data CF compliant, I believe all you'll have to do is: >> >> 1) Make sure there are coordinate variables for the I and J dimensions. >> (These can just have nondimensional dummy values.) > > In fact they don't have to be included at all. It is OK just to have > dimensions > without coordinate variables. > > Auxiliary coordinate variables are not supposed to have missing data values. > However, we relaxed that requirement in CF 1.6 for discrete sampling > geometries. We could relax it generally, but I suppose it would make sense > to permit missing data in aux coord vars only at points where the data > variable(s) concerned has missing data. > > Best wishes > > Jonathan > >> On 2/5/14 1:49 PM, Signell, Richard wrote: >>> CF folks, >>> >>> Many ocean models have curvilinear grids, but most of the ones I've >>> encountered have 2D coordinate variables lon,lat that are monotonic >>> and have no missing_values. >>> >>> The NOAA forecast model for the St. Johns River Operational Forecast >>> System, however, has a complex geometry, and used a grid generator >>> that allowed for "cuts" in the grid, so that in the resulting grid >>> system lthe coordinate variables on/lat are not monotonic and have >>> missing values. >>> >>> I did an experiment to see if I could fill in the missing lon/lat >>> values via interpolation, but this exercise only made it clear that >>> this cannot work, as the data being interpolated are not monotonic. >>> >>> Check out the first figure here to see the issue: >>> https://www.wakari.io/sharing/bundle/rsignell/SJROFS >>> >>> Is there any way this operational forecast data could be made to be CF >>> compliant? >>> >>> Thanks, >>> Rich >>> >> _______________________________________________ >> CF-metadata mailing list >> [email protected] >> http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata > > ----- End forwarded message ----- > _______________________________________________ > CF-metadata mailing list > [email protected] > http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
_______________________________________________ CF-metadata mailing list [email protected] http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
