sigma coordinates

[ana-machado]

Hello everyone,
Can anyone update me on the status of the implementation of sigma coordinates in parcels? And do you know of any studies successfully using parcels with ROMS outputs? I have some scripts from last summer, to study larval dispersal with ROMS outpus, and I am now getting back to them (finally have the ROMS simulations) and would like to know if there were any advances. I know someone was working on implementing sigma coordinates varying in time. Is this done yet?
Thank you very much :)
Ana

[reint-fischer]

Hi @ana-machado ,

I am not working with ROMS myself, so I am not sure about the exact compatibility at the moment. To confirm, last spring this (#765) is where the discussion was left? I know @DEHewitt was working on some simulations recently, is that working for you? I also know that Parcels supports time-varying grids through this implementation, although I don't know how that works with a spherical mesh and the grid indexing used in ROMS. I believe @cjongedijk was working with the time-varying grids: are those sigma grids Cleo?

[DEHewitt]

Hi @ana-machado,

@HaydenSchilling and myself have been working on using recent ROMS output to model larval dispersal in the east Australian region. We have been attempting 3d-advection which has been made tricky given the sigma-coordinates, but we have got a combination of #765 and a buoyancy kernel we wrote to do what we wanted (although it is very hacky). The ROMS model we're using doesn't have vertical velocities (W) either.

@HaydenSchilling and @jaseeverett published a paper last year using just the surface layers of a ROMS model which can be found here, and the associated repo here. If you can justify only using the surface layers this would be much easier, see section 2.1 and Fig. S1 and S2 in this paper for an example of what I mean.

I hope this helps!

[ana-machado]

Hi @DEHewitt and @reint-fischer

Thank you so much for your reply. I have also been using #765 together with some buoyancy and diurnal vertical migration kernels that I wrote. It seems to be working ok. I am considering the depth of the sigma coordinates does not change in time, I think it was under development the possibility of considering them varying in time, but it seems there were no advances in this.

I am now trying to add diffusion to my simulations, I was thinking of using AdvectionDiffusionM1 kernels, but just noticed @HaydenSchilling and @jaseeverett used instead BrownianMotion2D. What are you using in your simulations, any of these? and why?

Thank you again :)
Ana

[DEHewitt]

Hi @ana-machado, yeah we've stitched together something simple to get particles to rise at a constant rate. It's a little messy though.

I'm currently using DiffusionUniformKh, which I think replaced BrownianMotion2D. With this diffusion scheme you add zonal and meridional diffusion constants to the field. @HaydenSchilling and @jaseeverett define the constants on lines 52 and 53, and then add them to the field on lines 89 and 90.