Best pratice to modify p.dt inside a kernel ?

[quai20]

Hello all !

Within my custom kernel, I would like to change particle.dt because in some part of the advection, I want to refine or enlarge the sampling rate. I see that I can do particle.update_next_dt(new_dt) but :

• does it apply to all the kernels ? (I have kernels = custom_kernel + AdvectionRK4 + periodicBC)
• How do I set the file writing ? because my dt is not constant, so what about the ParticleFile outputdt ? Is it possible to write all the data, when irregularly sampled ?

I tried combining particle.update_next_dt(new_dt) in my kernel, and leaving outputdt empty but I only get the first and last values of the simulation in my output file, butI think it's normal with the default outputdt=np.infty.

I you have any recommandation of example of varying, thanks very much.

Kevin

[erikvansebille]

Dear @quai20. Yes, I think that what you are doing is fine, and that Parcels should be able to indeed adapt dt on the fly. outputs will be respected no matter how you change dt, I think. But there's only one way to find out I guess, by just trying and see what happens?