We see significant difference compared to pyqg for low res sims. We have only changed the advection scheme and removed the exponential filter, so it would be nice to know which of thesse two modifications caused this (particularly now as the backscatter is captured even with nx=48 simulations, which is a dramatic difference to pyqg).
Pseudospectral method with 2/3 dealiasing should be equivalent to Arakawa - this scheme also conserves PV. We would need to add Smagorinsky here again, in place of the exponential filter, but it would be interesting to see if that system produces similar spectral characteristics to the Arakawa + smag. Pavel put a PR for pyqg here.
Chris-Pedersen
commented
11 months ago
We see significant difference compared to pyqg for low res sims. We have only changed the advection scheme and removed the exponential filter, so it would be nice to know which of thesse two modifications caused this (particularly now as the backscatter is captured even with
nx=48simulations, which is a dramatic difference to pyqg).Pseudospectral method with 2/3 dealiasing should be equivalent to Arakawa - this scheme also conserves PV. We would need to add Smagorinsky here again, in place of the exponential filter, but it would be interesting to see if that system produces similar spectral characteristics to the Arakawa + smag. Pavel put a PR for pyqg here.