I'm trying to run init_atmosphere_model (MPAS v8.4.1) with the 10km quasi-uniform mesh x1.5898242 downloaded from the official mesh page. The model is compiled with Intel MPI (intel-mpi target, ifort 17.0.5). When running init_atmosphere_model, the bootstrap phase fails with:
ERROR: At least one fields to be read from the 'input' stream is dimensioned
ERROR: by 'R3', but the 'R3' dimension is not defined
ERROR: in the file x1.5898242.grid.nc
The grid.nc file does not contain the R3 dimension or variables like edgeNormalVectors, localVerticalUnitVectors, cellTangentPlane,
and coeffs_reconstruct. However, the static.nc file for the same mesh DOES contain these R3 fields with correct geometric values. I noticed that core_atmosphere/Registry.xml
(lines 535-537 and 566) declares these as persistent fields without package qualifiers, and init_atmosphere/Registry.xml (lines 444-447) declares them with packages="gwd_stage_in;vertical_stage_in;met_stage_in".
My questions:
1. Is there a way to make init_atmosphere read the R3 fields from static.nc
instead of grid.nc?
2. Is there an updated version of the 10km grid.nc that includes R3?
3. Is there a standalone tool to compute and add coeffs_reconstruct and edgeNormalVectors to an existing grid.nc?
ERROR: At least one fields to be read from the 'input' stream is dimensioned
ERROR: by 'R3', but the 'R3' dimension is not defined
ERROR: in the file x1.5898242.grid.nc
The grid.nc file does not contain the R3 dimension or variables like edgeNormalVectors, localVerticalUnitVectors, cellTangentPlane,
and coeffs_reconstruct. However, the static.nc file for the same mesh DOES contain these R3 fields with correct geometric values. I noticed that core_atmosphere/Registry.xml
(lines 535-537 and 566) declares these as persistent fields without package qualifiers, and init_atmosphere/Registry.xml (lines 444-447) declares them with packages="gwd_stage_in;vertical_stage_in;met_stage_in".
My questions:
1. Is there a way to make init_atmosphere read the R3 fields from static.nc
instead of grid.nc?
2. Is there an updated version of the 10km grid.nc that includes R3?
3. Is there a standalone tool to compute and add coeffs_reconstruct and edgeNormalVectors to an existing grid.nc?