The way that the NDOWN program works is that a previous WRF output file from the coarse grid with multiple time periods is used to provide lateral boundary conditions to the finer-grid domain. A single time period of the WRF input data, at the fine resolution, is used for the initial conditions.
The namelist settings for the NDOWN program require that the time_control namelist record has the io_form_auxinput2 and fine_input_stream defined:
These stream columns are associated with the specific domains (column 1 = the coarse grid, column 2 = the fine grid). The stream numbers are attributes that are associated with most state arrays. For example, look at the latitude field in the Registry:
The "1" says is the location informing us that the field is latitude (nameed XLAT).
The "2" says that this is an "i0" variable: it goes into WRF from the real program.
The "3" says that this is ALSO an i2 variable: it is a masked field that will be required for the IC file in NDOWN.
The "4" indicates that this is a "h" array: it is output from WRF.
For NDOWN to properly function, you may remove all of the state arrays from the model output that meet this condition: the field is output but the field is not input. For practical considerations, use ncview to identify two lists of 3d arrays: a list that is in the wrfinput_d01 file, and a separate list of fields that is in the wrfout_d01 file. You may remove all of the "h" fields (the wrfout fields) that are NOT in the wrfinput file.
You should do a before vs after test to insure that you are getting bit-wise identical results with your WRF simulation that follows NDOWN (remove optimization, only a few time steps are required, use the utility WRF/external/io_netcdf/diffwrf).
The fields may be removed from the WRF output file via capabilities listed in:
http://www2.mmm.ucar.edu/wrf/users/tutorial/201907/gill_registry.pdf
Look at page 1 of the PDF, slide #4.
(This response originally comes from this thread).
The namelist settings for the NDOWN program require that the time_control namelist record has the io_form_auxinput2 and fine_input_stream defined:
Code:
&time_control
io_form_auxinput2 = 2
fine_input_stream = 0, 2,
/These stream columns are associated with the specific domains (column 1 = the coarse grid, column 2 = the fine grid). The stream numbers are attributes that are associated with most state arrays. For example, look at the latitude field in the Registry:
Code:
1 2 3 4
state real XLAT ij misc 1 - i0123rh0156{22}{23}du=(copy_fcnm) "XLAT" "LATITUDE, SOUTH IS NEGATIVE"The "1" says is the location informing us that the field is latitude (nameed XLAT).
The "2" says that this is an "i0" variable: it goes into WRF from the real program.
The "3" says that this is ALSO an i2 variable: it is a masked field that will be required for the IC file in NDOWN.
The "4" indicates that this is a "h" array: it is output from WRF.
For NDOWN to properly function, you may remove all of the state arrays from the model output that meet this condition: the field is output but the field is not input. For practical considerations, use ncview to identify two lists of 3d arrays: a list that is in the wrfinput_d01 file, and a separate list of fields that is in the wrfout_d01 file. You may remove all of the "h" fields (the wrfout fields) that are NOT in the wrfinput file.
You should do a before vs after test to insure that you are getting bit-wise identical results with your WRF simulation that follows NDOWN (remove optimization, only a few time steps are required, use the utility WRF/external/io_netcdf/diffwrf).
The fields may be removed from the WRF output file via capabilities listed in:
http://www2.mmm.ucar.edu/wrf/users/tutorial/201907/gill_registry.pdf
Look at page 1 of the PDF, slide #4.
(This response originally comes from this thread).