Forcing subtleties

[Joshua Rady]

I'm trying to figure out some subtleties of the forcing XML settings. Here is the scenario: I run a simulation (HIST competition) from 1901-1990 with the the following settings:
DATM_CLMNCEP_YR_ALIGN: 1901
DATM_CLMNCEP_YR_START: 1901
DATM_CLMNCEP_YR_END: 2014
RUN_STARTDATE: 1901-01-01

Then I run two more simulations both starting in 1990 for 25 years using the restart file from the previous case as initial conditions. In the first I only change:
RUN_STARTDATE: 1990-01-01

In the second I change:
DATM_CLMNCEP_YR_ALIGN: 1990
DATM_CLMNCEP_YR_START: 1990
RUN_STARTDATE: 1990-01-01

When I look at the forcing history output (TBOT, PCO2, etc.) from these two 1990 runs they are identical except for the first and last months. What I think is happening is that in the former case 1990-01-01 is being connected (interpolated) to 1989-12-01 and 2014-12-31 is being connected to 1901-01-01 while in the later case 1990-01-01 is being connected to 2014-12-31 and 2014-12-31 to to 1990-01-01. Is this correct?

I usually use the former approach but I was curious if the changes would have an effect. If I had used DATM_CLMNCEP_YR_ALIGN = 1989 for the last case would 1990-01-01 have been identical for the two cases?

Finally, I can't seem to figure out what DATM_CLMNCEP_YR_ALIGN does. In previous experiments I have seen that changing DATM_CLMNCEP_YR_ALIGN does not seem to do anything if the forcing dates overlap the simulation dates. For example, 1995 seems to align with 1995 if I'm using 1900-2000 forcing no matter what, although I am less confident about this statement.

The differences in forcing are subtle in these examples and are not likely to answer changing but understanding these settings would be helpful for some more exotic simulations.

 

[oleson]

Regarding your first question, I think that is correct. The datm needs a time slice from those files for interpolation. Of course, you can verify this by looking at the atm.log files.
Regarding your second question, no. Setting DATM_CLMNCEP_YR_ALIGN = 1989 means that you are aligning model year 1989 with 1990 forcing. So when you start the run at 1990, the datm will be using data from 1991.
Another example of the use of these settings is for a case where we want to run a full historical simulation (1850-2014), but we only have datm data for 1901-2014 (GSWP3V1). In this case, we run the model from 1850 through 1900 by looping over datm forcing for 1901-1920, and then continue the run for 1901-2014 using the forcing for 1901-2014. The settings for the 1850 through 1900 run are:

DATM_CLMNCEP_YR_ALIGN: 1901
DATM_CLMNCEP_YR_START: 1901
DATM_CLMNCEP_YR_END: 1920
RUN_STARTDATE: 1850-01-01

In this case the model years and datm years align like this:

# RUN Year : 1850 ... 1860 1861 ... 1870 ... 1880 1881 ... 1890 ... 1900 1901-
# FORCE Year : 1910 ... 1920 1901 ... 1910 ... 1920 1901 ... 1910 ... 1920 1901-

Note that since we have told the datm that we eventually want model year 1901 to line up with 1901 datm forcing, the datm will actually start the forcing at 1910 for model year 1850 so as to end up with 1901 model and datm year lining up at the end

 

[Joshua Rady]

OK, so DATM_CLMNCEP_YR_ALIGN is the forcing year matched to DATM_CLMNCEP_YR_START. This finally make sense. I though it set the first year in the forcing years to start the loop from, but that never worked right. I could never figure out how the model knows to start with with 1910 at 1850.

When I wrote DATM_CLMNCEP_YR_ALIGN = 1989 I meant to DATM_CLMNCEP_YR_START = 1989, but your answer to my mistake was much more edifying so that worked out well. Thanks for the help.

 

[Joshua Rady]

I looked at the atm.log file, which I have never done before, and I can now see the interpolation at the ends. Quite useful.