Closed ali-ramadhan closed 5 years ago
Even hard coding in the kh
value gives issues as the closure/kernel is not pickleable so joblib errors =/
So I guess I'll use the built-in BrownianMotion2D
kernel but it slows the advection down by a factor of 12~15x which is kind of sad. Might open an Ocean Parcels issue.
Oh well, we can just use up more cores lol...
Not 100% sure why that was slow but just looping over the particles and manually adding in the diffusion works and takes up as much time as the advection so I'll just stick with this for now.
This should let us add a constant horizontal diffusivity. Ocean Parcels has one but it requires you to define
kh_zonal
andkh_meridional
fields which is a bit cumbersome.I'm trying to use a closure to generate isotropic diffusion kernels that SciPy can compile.
Two issues:
kh
for now.Resolves #21