Not quite a bug, but using the surface temperature parameterization for snow slows down our 2 year long run from ~96 SYPD to ~65 SYPD. The parameterization solves for the skin temperature of the snow, effectively, using a root solver.
We should be calling the parameterization once per step. How can it be this slow? Note that we solve for canopy temperature as a prognostic variable (implicitly) using Newton iterations to solve for the temp at the next timestep. I don't think that is as slow as this, perhaps because we restrict to 3 iterations.
Not quite a bug, but using the surface temperature parameterization for snow slows down our 2 year long run from ~96 SYPD to ~65 SYPD. The parameterization solves for the skin temperature of the snow, effectively, using a root solver.
We should be calling the parameterization once per step. How can it be this slow? Note that we solve for canopy temperature as a prognostic variable (implicitly) using Newton iterations to solve for the temp at the next timestep. I don't think that is as slow as this, perhaps because we restrict to 3 iterations.