Skip Monte Carlo computations for g-points with high gasous optical depths

[MennoVeerman]

G-points with high gaseous optical depth barely show any impact of clouds, but introduce a significant amount of noise. User can define a threshold ratio between the maximum gas and cloud optical depths (per-cell, not vertically integrated). If the ratio exceeds this threshold, use 1D fluxes instead of MC ray tracing. This threshold defaults to 0.1, in which case less than half of the g-points are run in 3D

[Chiil]

Would it not make more sense in terms of physics to also include the ratio of mean free path to grid spacing? Whether you need 3D also depends on that.

[MennoVeerman]

Good idea, I agree that also makes sense and is closer to the physics. The attached image visualizes the potential for using mean free path. For each g-point I took the highest gaseous and cloud optical depth and plot the ratio between the two (f_tau) against the mean free path based on both optical depths. Dotted line shows the current grid spacing (200m), solid line the "optimal" f_tau based on visual inspection. The gas mean free path nicely meets the grid spacing at this optimal f_tau

[MennoVeerman]

Instead of using f_tau, we can switch to using the ratio between the gas mean free path and the grid spacing to decide whether to do monte carlo or not. This ratio can be provided by the user and defaults to ~1?