What land-sea mask do we use?

[ccarouge]

CABLE takes the assumption the met. forcing defines the grid and the land-sea mask. This means the simulation runs exactly on the grid of the met. forcing and this forcing needs no interpolation/filling etc. operation.

JULES takes a different approach where the land-sea mask comes from a separate file or the land cover.

Which approach do we want to adopt? It seems avoiding regridding the met forcing before simulation is better but we need to decide.

[matlipson]

I agree it will be better to avoid regridding met-forcing and adapting new ancils to match. We can use tools such as ants.analysis.make_consistent_with_lsm() to a) remove land data outside met forcing region, b) fill missing land points (ants.analysis.FillMissingPoints) using a "spiral-search" nearest neighbour approach.