chris_fletcher@utoronto_ca
New Member
[[Apologies for dual-posting, but I originally posted this on the CAM namelist board, then realised it's really related to the ice model]]
This thread is a follow-up to this one: http://forum.cgd.ucar.edu/showthread.php?1878-How-to-use-another-specified-SST-data/page2
What I want to do is run an F compset case at T31 resolution forced with prescribed SSTs and ice output from a previous coupled T31 simulation. My code base is cesm1_0_2. I started from the FAMIPCN compset and then followed the instructions in the thread above:
When I run the case, it fails during the ice model initialization with the following error:
In this setup, the model didn't like the fact the input ice data were on the gx3v7 (irregular) grid. Is there something else I need to add or change to make the ice model accept these input data, or should I regrid the ice data to a regular grid (in this case 48x96)? [[Update: I regridded the gx3v7 ice data to 48x96 and ran the case without a problem, but ifrac in the resulting cice history file was significantly different from ifrac in the coupled simulation and therefore this does not solve the problem.]]
As a test, I reverted cice.buildnml.csh back to defaults (i.e. prescribed Hurrell ifrac data), but I left the docn.buildnml.csh file pointing to the SST from the coupled run on the gx3v7 grid. In this case the model runs fine. The reason seems to be that shr_strdata_nml can be changed for docn at runtime through docn_ocn_in, docn_in and docn.stream.txt, and somehow through this combination of files the gx3v7 data can be used. However, for the ice model the only namelist file is called ice_in, and it seems that shr_strdata_nml can't be changed at runtime.
Perhaps the problem is that the F compset uses cice in prescribed mode by default, rather than dice (the data ice model)? Should I create a new compset that uses dice? Based on the replies in the previous thread, I suspect what I'm trying to do can be achieved using the existing code base and the F compset. So I'm probably either doing something wrong, or I've omitted something. Could anyone point me in the right direction?
Thanks in advance.
This thread is a follow-up to this one: http://forum.cgd.ucar.edu/showthread.php?1878-How-to-use-another-specified-SST-data/page2
What I want to do is run an F compset case at T31 resolution forced with prescribed SSTs and ice output from a previous coupled T31 simulation. My code base is cesm1_0_2. I started from the FAMIPCN compset and then followed the instructions in the thread above:
- I extracted surface-level potential temperature and aice from the coupled ocn and ice history files, filled in missing data over land, corrected the time axis to CF compliant.
- I configured an F1850 case, edited docn.buildnml.csh and cice.buildnml.csh to point to the input files generated in step 1. Since the input SST and ice data are on an irregular grid (gx3v7), I set all the domain-related variables in the namelists (domfilename, dom[X,Y]varname etc.) to correspond to the domain file for the gx3v7 grid ($DIN_LOC_ROOT/ocn/docn7/domain.ocn.gx3v7.090903.nc, variable names xc, yc, area and mask).
When I run the case, it fails during the ice model initialization with the following error:
Code:
(shr_stream_set) size of filename = 1
(shr_stream_set) filename = /scratch/p/pjk/cgf/cesm1_0_2/inputdata/atm/cam/sst/ifrac_cesm1_comp-B_1850_CN_res-T31_g
37.rerun_gx3v7_1850_c120225.nc
(shr_strdata_init) calling shr_dmodel_mapSet for fill
(shr_strdata_init) calling shr_dmodel_mapSet for remap
(shr_map_checkGrids_global) ERROR Xsrc not increasing 11 2.900000000000 359.3000000000
(shr_map_checkGrids_global) ERROR Xsrc not regular lat,lon 1 5 323.3000000000 323.3000000000
(shr_map_checkGrids_global) ERROR Ysrc not regular lat,lon 2 50 0.2997398353064 0.2997382175421
(shr_sys_abort) ERROR: ('shr_map_checkGrids_global') ERRORIn this setup, the model didn't like the fact the input ice data were on the gx3v7 (irregular) grid. Is there something else I need to add or change to make the ice model accept these input data, or should I regrid the ice data to a regular grid (in this case 48x96)? [[Update: I regridded the gx3v7 ice data to 48x96 and ran the case without a problem, but ifrac in the resulting cice history file was significantly different from ifrac in the coupled simulation and therefore this does not solve the problem.]]
As a test, I reverted cice.buildnml.csh back to defaults (i.e. prescribed Hurrell ifrac data), but I left the docn.buildnml.csh file pointing to the SST from the coupled run on the gx3v7 grid. In this case the model runs fine. The reason seems to be that shr_strdata_nml can be changed for docn at runtime through docn_ocn_in, docn_in and docn.stream.txt, and somehow through this combination of files the gx3v7 data can be used. However, for the ice model the only namelist file is called ice_in, and it seems that shr_strdata_nml can't be changed at runtime.
Perhaps the problem is that the F compset uses cice in prescribed mode by default, rather than dice (the data ice model)? Should I create a new compset that uses dice? Based on the replies in the previous thread, I suspect what I'm trying to do can be achieved using the existing code base and the F compset. So I'm probably either doing something wrong, or I've omitted something. Could anyone point me in the right direction?
Thanks in advance.