AbstractStochasticPhysics and SPPT

[milankl]

Implements a general interface for stochastically perturbed parameterization tendencies SPPT as well as inputs. Running the same model twice now gives already different trajectories

one can however create a random process with a non-zero seed that would sample the same sequence of random numbers for reproducibility (default no)

random_process = SpectralAR1Process(spectral_grid, seed=123)

(seed=0 means seed from Julia's global RNG)

@minqi6 this would give you another somewhat adhoc stochastic physics baseline than the stochastic convection you are working on!

@matt-graham maybe that's the added stochasticity you'd need?

@CakeMixCore It sounded like you were more interested in perturbing parameters but if you just want to create ensembles then this is a much better way of doing it!

To be used like

using SpeedyWeather

spectral_grid = SpectralGrid()
random_process = SpectralAR1Process(spectral_grid)
sppt = StochasticallyPerturbedPhysicsTendencies(spectral_grid)

model = PrimitiveWetModel(spectral_grid; random_process, stochastic_physics=sppt)
simulation = initialize!(model)
run!(simulation)

default remains: random_process = NoRandomProcess() and stochastic_physics = NoStochasticPhysics()