The motivation
Currently for each job (day) we do a 12 hour spin up. This requires 50% additional compute to the day and can cause small differences between the final timestep from the day before and the spin up.
The alternative is to use a restart file to start the model in a known good condition. The spin up should only be required for the first day of a new domain.
The proposed solution
Allow running in restart mode. If a restart file is found use that instead of spinning up.
Alternatives
Do nothing
Additional context
The motivation
Currently for each job (day) we do a 12 hour spin up. This requires 50% additional compute to the day and can cause small differences between the final timestep from the day before and the spin up.
The alternative is to use a restart file to start the model in a known good condition. The spin up should only be required for the first day of a new domain.
The proposed solution
Allow running in restart mode. If a restart file is found use that instead of spinning up.
Alternatives
Do nothing
Additional context