Hi, I saved 3-hourly coupler history output from a coupled (BG) run and then forced a land-only (IG) run with this. Doing this, I expected the temperature and precip fields to be the same in both runs. However this does not seem to be the case.
For precip, there are only rounding differences before repartitionining, which is good. However after repartitioning, there are differences in the SNOW/RAIN fractions, mainly along the coast of Greenland. This could be due to a different initial file for CISM, where different elevations could trigger different temperature lapse rates and thus repartitioning. I still plan to test that.
Yet the biggest differences is in 2-m temperature (TSA), which differs over 2 K, annually averaged. See attached figure. As you can see, the largest differences are over ice sheets and other ice covered regions, which happen to be my areas of interest.
At first, I thought this may be related to the difference in coupling interval (3 hourly vs. 30 min). However this does not make sense because then this would manifest globally, not just over ice.
Second, a different CISM initial file could cause temperature differences over Greenland, but certainly not over mountain glaciers or Antarctica. So there must be something else at play.
Another difference between the two runs is the different CO2 forcing. For the IG run to work properly I had to resort to a constant CO2 forcing. However, there is no link between CO2 and TSA (right?) and that would also manifest globally.
In summary, I'm kind of puzzled right now, any ideas welcome!
CLM version: 4.5.14_r225
What I did was branch off #144, turn on the coupler history output (histaux_a2x3hr = .true.) and ran a BG simulation for 4 years.
Then, I set up an IG run that is forced by the coupler output. As the coupler output was daily, I had to concatenate this to monthly files using NCO. I use user_datm.streams.txt.* files to refer to the forcing files , as the relevant DATM option did not seem to work properly (see this CIME bug report).
The initial condition is set to be the same, using finidat.