Smoothing window for realistic LESbrary geostrophic base state

[ali-ramadhan]

Right now in the "realistic" pilot simulation the geostrophic base state U_geo(z, t), V_geo(z, t) is computed for each day then linearly interpolated in time.

@rafferrari pointed out a couple of issues with this approach:

1. We may need to add the time derivative of the base state ∂t(U⃗_geo) to the momentum equation and linear interpolation would introduce discontinuities in ∂t(U⃗_geo) which wouldn't be good.
2. Over long time scales, e.g. over a seasonal cycle, the temporal variability of U⃗_geo is important. But over short time scales, e.g. a week or two, the temporal variability may not be so important and may just end up creating extraneous/spurious turbulence (as most perturbations/variability tends to do in LES).

So especially for short runs we may want to smooth the geostrophic base state. @rafferrari suggested a smoothing window or 1-2 weeks as this is roughly the geostrophic eddy timescale (I think?).

cc @adelinehillier @sandreza