knutfrode
commented
4 years ago I agree, the logic here is not too logical.
But yes, the idea behind that change was that if one uses o.set_config('vertical_mixing:diffusivitymodel', 'environment') and there are no readers to provide diffusivity (often the case), then it would be better to use a parameterisation based on wind (eg. Large1994) than a constant fallback value. Thus Large1994 is set to override fallback_values if environment is chosen and diffusivity is missing for all particles. But of course, this behaviour might not be always desired.
Yes, maybe another option diffusivity_model = 'constant' would be useful. You are welcome to make a pull request, otherwise I can anyway implement this change tomorrow.
Note btw that the mixing scheme was greatly simplified yesterday, and I expect this to be quite a bit faster for cases with many particles, where mixing might be a bottleneck comparable to obtaining data from readers. https://github.com/OpenDrift/opendrift/pull/388/files
simonweppe
gauteh
Hi team,
In some applications, I need to set 'ocean_vertical_diffusivity' to a constant value. I used to do something like this :
o.set_config('vertical_mixing:diffusivitymodel', 'environment') o.fallback_values['ocean_vertical_diffusivity'] = 0.001 # m2/sIt did the trick, but since this new commit https://github.com/OpenDrift/opendrift/commit/c59f8dde68e2c2effb2d7400433d26fb9ff818e3 (I think), the model now switches to Large1994 instead, since that bit below becomes false.'ocean_vertical_diffusivity' in self.environment_profiles and self.environment_profiles['ocean_vertical_diffusivity'].min() != self.fallback_values['ocean_vertical_diffusivity'] or self.environment_profiles['ocean_vertical_diffusivity'].max() != self.fallback_values['ocean_vertical_diffusivity']:Maybe an option to retain that feature would be to set the default 'ocean_vertical_diffusivity' to 0 ? and check if
self.environment_profiles['ocean_vertical_diffusivity'].min() !=0rather than fall back ? It does mean that the fallback will be zero if it is not specified otherwise by user, which may not be ideal. Other option would be to have adiffusivity_model = 'constant', which then would always use the fallback ?Let me know which approach seems best to you and I could make small pull request for that.