Current interpolation with different landmasks

[SamTFredriksson]

Hi,
I´m looking into how to set the landmask the most appropriate way and how the interpolation will work if using the default GSHHG landmask looking into https://opendrift.github.io/gallery/example_model_landmask.html

In the enclosed plot of the southern part of Skagerrak I have two areas of interest with different questions with ‘coastline_action’ of ‘previous’.

In the orange box I have sketched two green particles, one full and one striped. I´ve also sketched two possible currents in red and yellow as can be seen with quite different, however, possible direction. The full particle is outside the ocean but still active according to the GSSHG coastline/landmask. The striped one is active both according to the ocean landmask and according to GSSHG landmask. How will the interpolated currents affect the two particles? Will both sides of the land be used for the interpolation of the current that in turn affects the particles?

In the yellow box there´s another problem where I´ve sketched a possible yellow particle that according to the GSHHG landmask possibly could take the inner path, sketched as a yellow trajectory. Is this possible (depending on the current forcing) to leave the domain like this and enter it again in another position?

What kind of landmask would be your general recommendation here? Using the ocean or the GSHHG landmask? Or actually create a unique one?

//Sam
pathAndInterpolation.pdf

[knutfrode]

Hi, and sorry for the delayed answer.

For the green particle, there is no model current at the given position, which is still not on land according to GSHHG. For this case, the ocean model data are extrapolated/interpolated with "binary_dilution", one pixel at the time, until it will get a value (but a maximum of 10 pixel/iterations).
For this case, it would chose the nearest, i.e. correct, ocean current vector from the eastern side. Thus I believe there is no danger right here to get currents from the western side. However, a little to the north there is just a single pixel of land, and then I guess we could in principle have the risk of getting a (wrong) current value from the "other side". I would however expect this to be a very rare case.

For the yellow box, OpenDrift will (as explained above) extrapolate up till a maximum of 10 pixels, but I believe there will still be a part in the inner fjord where particle would be stranded with no data/current. The extrapolation could in principle be increased, but it is of course dubious with this much extrapolation.
Again, I think in principle one could construct a case where a particle could leave the domain to re-enter slightly offset, but in most cases I believe particles would be deactivated due to "missing_data" (unless the model is set to allow having e.g. no current (fallback_value = 0), in which case particles are not deactivated, but just remaining stationary).
It should in principle not make a difference to this whether using GSHHG or ocean model landmask, and either should be fine.
My choice would be to use the GSHHG landmask, to avoid having stranded pixels in a sawtooth-pattern (pixel boundaries).

[SamTFredriksson]

Thanks for your reply. We did run the simlations according awaiting your choice with good results. It was, howvere, good to have your advice of experience to support that decision.