Thank you, Alan, for pointing out this issue. I posted a CESM bug report (http://bugs.cgd.ucar.edu/show_bug.cgi?id=1710) that reads as follows...
Alan Di Vittorio wrote:
[...] the cell area and landfrac variables in the land surface files have different values than are output in the history files. The pftmask variables I checked were the same. I didn't directly check the landmask variable because it corresponds with the landfrac variable. I am working with what I think is the equivalent of CESM 1_0 (whatever version that was used for the CMIP5 runs).
I found that the land surface files (particularly from mksurfdat, and including the initial state file pointed to by fsurdat in the namelist) significantly overestimate total global land area, and that the values from the history files seem to be correct. This is a problem when developing and debugging a piece of code where the intermediate files need to be examined and no history files have been written.
It appears that the 'actual' values output in the history files come from somewhere else, presumably the atmosphere grid and landfrac files. With two sets of these values floating around, which ones are used internally? And is one set used exclusively, or is there mixed use? Is there a reason for these inconsistencies? I think that the surface files should contain the correct values. Otherwise this leads to confusion and errors (especially if there is mixed use), not least because the incorrect values are bundled together with the actual information on land units and pfts. For example, my initial estimates of global forest area using the land surface file values were 10 million km^2 larger than they should have been.
---
Erik and I confirmed the following:
AREA in fsurdat comes from a mapping file used by the mksurfdata code and is not used by the CLM. It appears in fsurdat for convenience.
The area in h0 files comes from a domain file and I think is also not used by the CLM because CLM calculations are area-independent on the gridcell level.
LANDFRAC_PFT in fsurdat comes from the pft data and is not used. It appears in fsurdat for convenience.
The landfrac in h0 files is used by the coupler I think, because I'm pretty sure that the CLM does not need to use it.
Ideally we would make mksurfdata consistent with the CLM so that fsurdata would contain the same info as h0 files. Such a modification will change answers and should happen at a time when we're ok with that. I post this issue just for the record. We do not plan to correct it at this time.Sam Levis
Alan Di Vittorio wrote:
[...] the cell area and landfrac variables in the land surface files have different values than are output in the history files. The pftmask variables I checked were the same. I didn't directly check the landmask variable because it corresponds with the landfrac variable. I am working with what I think is the equivalent of CESM 1_0 (whatever version that was used for the CMIP5 runs).
I found that the land surface files (particularly from mksurfdat, and including the initial state file pointed to by fsurdat in the namelist) significantly overestimate total global land area, and that the values from the history files seem to be correct. This is a problem when developing and debugging a piece of code where the intermediate files need to be examined and no history files have been written.
It appears that the 'actual' values output in the history files come from somewhere else, presumably the atmosphere grid and landfrac files. With two sets of these values floating around, which ones are used internally? And is one set used exclusively, or is there mixed use? Is there a reason for these inconsistencies? I think that the surface files should contain the correct values. Otherwise this leads to confusion and errors (especially if there is mixed use), not least because the incorrect values are bundled together with the actual information on land units and pfts. For example, my initial estimates of global forest area using the land surface file values were 10 million km^2 larger than they should have been.
---
Erik and I confirmed the following:
AREA in fsurdat comes from a mapping file used by the mksurfdata code and is not used by the CLM. It appears in fsurdat for convenience.
The area in h0 files comes from a domain file and I think is also not used by the CLM because CLM calculations are area-independent on the gridcell level.
LANDFRAC_PFT in fsurdat comes from the pft data and is not used. It appears in fsurdat for convenience.
The landfrac in h0 files is used by the coupler I think, because I'm pretty sure that the CLM does not need to use it.
Ideally we would make mksurfdata consistent with the CLM so that fsurdata would contain the same info as h0 files. Such a modification will change answers and should happen at a time when we're ok with that. I post this issue just for the record. We do not plan to correct it at this time.Sam Levis