Backward modelling and buoyancy

[sadie-ds]

Hello,

To give a brief background, I'm using the Oceandrift model to backwards model from washed up objects to potential sources. I'm seeding along a transect with delayed releases even across time to cover spatial and temporal uncertainty. I have a few queries (please excuse if some of these are simple im new to particle modelling). im also seeding about 600m out from the bay due to ROMS coverage. i'm seeding hundreds to thousands of particles and creating a trajectory "envelope,"i.e. there is no endpoint. i just cut it off at about 2 weeks (per particle), but might do some sensitivity analysis and play around with that. coastline action: previous

1. currently, I'm seeding close to the seafloor due to particles having slightly more negative buoyancy than seawater and associated with the seafloor, yet it is currently just acting as a passive tracer. I don't have exact terminal velocity and could probably find a proxy; however, I've read vertical physics can be hard to interpret in reverse time. Does anyone have any advice on this? is it worth doing or is simplicity best? Id also like to do more dense objects at some point (very much negative buoyancy). Would it just be the case of adding a normal terminal velocity and vertical mixing? or do i have to switch the signs (+/-) for reverse time tracking?

2. back and forth -- is there a way i can find out where particles finish to seed forwards for sensitivity analysis (cropped at 14 days; seeding over a period of 14 days would mean those seeded first would be active for 28 days)?

3. i have seabed interaction as previous; is this the best course of action? shall i change this if terminal velocity is appropriate to add

Thank you so much for any and all opinions !

[knutfrode]

Hi, and sorry for the late reply.

These are interesting questions, and you could also consider posting them on the OpenDrift Slack channel.

With OpenDrift you can simply pass a negative calculation time_step to your run-method, and "everything" should be reversed automatically, e.g. changing sign of both vertical and horizontal advection and buoyancy terms.
The terminal_velocity that you provide will have the same interpretation, with positive being upwards. For OceanDrift, this terminal_velocity is kept constant, but other modules (e.g. oil and fish eggs) have defined method to update terminal velocity:
https://opendrift.github.io/_modules/opendrift/models/openoil/openoil.html#OpenOil.update_terminal_velocity

I agree that vertical physics in reverse should be interpreded with care. I have no proper justification to give, but "believe" that things should be ok as long as there is no "surfacing" where elements would stay at surface (as e.g. oil or plastics might do).
As this is a general question independent of OpenDrift, you could also consider joining the newly created CLAM community, and ask for opinions there:
https://clam-community.zulipchat.com/

Btw, OpenDrift will extrapolate ocean model data towards land, so it should not be strictly necessary to move your seeding positions offshore, but extrapolation into a bay is of course anyway a little questionable.

I am not sure what is exactly meant with this question.
But it is possible to set a "lifetime" of elements with config option, e.g. for 14 days:
o.set_config('drift:max_age_seconds', 14*24*3600)
Also not sure if relevant to your question, but it is possible to add a flag (integer) origin_marker (or eventually origin_marker_name which will be converted to integers 0, 1, 2, ...) as in this example:
https://opendrift.github.io/gallery/example_river_runoff.html
Thus if you have multiple seedings, you can track back which elements origin from the various seedings.
For this, I can recommend the TrajAn package, which makes such analysis very simple with well known Xarray workflow:
https://opendrift.github.io/trajan/

I guess previous should be ok. Here is an illustration of the three different options:
https://opendrift.github.io/gallery/example_seafloor_interaction.html