Method of manufactured solutions
Idea:
- Take continuum function space (space + time). Ensure obeys boundary conditions (maybe hard?).
- Posit equations of motion (maybe plus hyperviscosity?).
- Posit several solution fields from continuum function space.
- Derive forcing terms necessary to make that solution a steady state.
- Give dynamics interface to sample with arbitrarily high spatiotemporal frequency.
This proves code correctness (are we solving the equations that we want to solve?).
Possibilities: take high res atmosphere runs with diabatic forcing, project into continuum basis (e.g. 5 days of high-res amip run). Realistic-ish atmosphere with forcing term backed out by projection.