[MITgcm-support] spurious energy sink in AIM?

[Brian Rose]

To those MITgcmers familiar with the atmospheric physics pkg/aim_v23:

I may have found evidence of a bug violating energy conservation, and it seems to be related to the large-scale precipitation parameterization.

Briefly, the model consistently equilibrates with a global mean imbalance in radiative fluxes at the TOA (top of atmosphere) of order 1 - 2 W/m^2 (excess absorbed shortwave over outgoing longwave).  A large ensemble of runs shows that the TOA imbalance varies systematically with the global mean large-scale precip rate.

Some details:
- I'm running AIM and pkg/thsice on a global C24 grid with aquaplanet geometry.  The same TOA imbalance appears in slab ocean runs (AIM coupled to a 60 m mixed layer) and in fully coupled runs that have reached equilibrium. 
- I compute TOA imbalance based on AIM diagnostic output fields TSR and OLR averaged over 20 year snapshot runs, taking an area-weighted global average of TSR-OLR.
- I have a large array of runs in a slab ocean configuration.  I vary a control parameter that leads to changes in global mean temperature.
-  All runs have had ample time to equilibrate, usually about 100 years or more.  There is no temperature drift.
- The TOA imbalance (net flux in to the system) systematically decreases as the climate warms (see first figure below).
-  There is no such imbalance in the net heat flux across the sea surface.
-  These runs are all warm enough to be sea ice-free (although pkg/thsis is enabled)
- After some sleuthing, I figured out that there is a systematic shift from large-scale to convective precip as the climate warms.  There is a very strong correlation between the TOA imbalance and the large-scale precip rate (see second figure).
-  The slope of the graph in figure 2 suggests that the radiative imbalance would tend to zero in the limit of zero large-scale precip.
- It thus appears that the large-scale precip is behaving as an energy sink (and also therefore acting a spurious positive feedback in the model).

I tried two simple tests to reduce the imbalance:
(1)  set aim_energPrecip=.FALSE.  (normally I have this set to .TRUE.)
(2)  increase QSMAX by an order of magnitude over its default value
Both tests were run out to equilibrium.  Both had basically no effect on the TOA imbalance.

So I know the imbalance is very well correlated with large-scale precip, but I don't understand why.
Any bright ideas?

I am attaching a typical data.aimphys file for my setup.  (there are a few unfamiliar switches in here, belonging to my custom implementation of a specified 'q-flux' to the ocean slab)

Thanks
Brian

---------------------------------------------
Brian E. J. Rose, PhD
NOAA Climate and Global Change Postdoctoral Fellow
Department of Atmospheric Sciences
University of Washington

office:  ATG 318   phone:  (206) 543-4596
email:  brose at atmos.washington.edu
---------------------------------------------
-------------- next part --------------
A non-text attachment was scrubbed...
Name: TSmean_radImbalance.pdf
Type: application/pdf
Size: 23234 bytes
Desc: not available
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20110526/46ad9a66/attachment-0002.pdf>
-------------- next part --------------


-------------- next part --------------
A non-text attachment was scrubbed...
Name: PRECLS_radImbalance.pdf
Type: application/pdf
Size: 25105 bytes
Desc: not available
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20110526/46ad9a66/attachment-0003.pdf>
-------------- next part --------------

-------------- next part --------------
A non-text attachment was scrubbed...
Name: data.aimphys
Type: application/octet-stream
Size: 7465 bytes
Desc: not available
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20110526/46ad9a66/attachment-0001.obj>
-------------- next part --------------