Address negative enthalpy

[alastairyang]

Hi, I am wondering what is the best remedy to address 'negative enthalpy' error during "steadystate" solution? From my experiments, this arises due to various reasons, e.g. basal forcing, boundary condition, initial state, and these negative enthalpy can be located in a small region. On the other hand, while a simple thermal solve usually does not run into these convergence issue, steadystate solve is more mechanically consistent so I'd still like to fix this. Is there a way to treat, for example, negative enthalpy values during the steadystate solve iteration so that it doesn't snowball into non-convergence?

Thanks,
Donglai

[MathieuMorlighem]

Hello Donglai,
there is no magic want (otherwise it would be implemented :) ). What we do generally is we get our steady state manually, i.e., we solve for temperature based on a good estimate of ice speed, and then we solve a stress balance. Update md.initialization.vx/vy/vz and run a new enthalpy model, and we iterate like that 3-4 times. This is generally enough to be close to steady state and you get a chance to inspect all the fields as they are calculated.
Sorry that's all I can recommend...
Mathieu

[alastairyang]

Thank you Mathieu!