[BUG/ISSUE] Fullchem run failure in 12.7.0+ at c180+ due to reduced timesteps

[geoschem]

[BUG/ISSUE] Fullchem run failure in 12.7.0+ at c180+ due to reduced timesteps

[lizziel]

I also did runs of GEOS-Chem online in GEOS (old MAPL) with two different timesteps. Both had warm GEOS-Chem restart settings the same so no issue:

 Doing warm GEOS-Chem restart
 GEOS-Chem phase            1 :
 DoConv   :  T
 DoDryDep :  T
 DoEmis   :  T
 DoTend   :  T
 DoTurb   :  F
 DoChem   :  F
 DoWetDep :  F

[lizziel]

To further clarify the issue, here is a breakdown of current (12.7.1) behavior of GCHP in terms of what components are run during each timestep. For simplicity I only list chemistry and dynamics with the understanding that each represents a set of GEOS-Chem subcomponents.

Current (and correct) GCHP behavior with default low resolution timesteps (10/20):

warm restart: dynamics, chemistry
00:10: dynamics
00:20: dynamics, chemistry
00:30: dynamics
00:40: dynamics, chemistry
00:50: dynamics
etc.

Current (and incorrect) GCHP behavior with default high resolution timesteps (5/10):

warm restart: dynamics
00:05: dynamics, chemistry
00:10: dynamics
00:15: dynamics, chemistry
00:20: dynamics
00:25: dynamics, chemistry
etc.

This appears to be a MAPL alarms issue since IsChemTime is set from the alarm, alternating between timestep, and directly setting whether chemistry is done in addition to dynamics. Somehow the reduction of the timestep makes the alternating IsChemTime out-of-phase when it should be independent.

[lizziel]

I have verified this behavior is also present in GCHP versions prior to the MAPL upgrade in 12.5.

[lizziel]

This bug is fixed in commit geoschem/geos-chem@8d2a8c2. I added updates to address the problem for the recommended timesteps only. These are autoset in runConfig.sh, and are 10/20 min (dyn/chem) for resolutions below c180, and 5/10 min above. It is a kludge and may not work with other timestep combinations.

The fix is made up of two parts:

1. Update GIGCchem_REFERENCE_TIME seconds in GCHP.rc from 0 to 1 if using reduced timesteps. This prevents the shift of the alternating IsChemTime logical value returned by MAPL by one timestep.
2. Force all components to be on during the first timestep.