SNAP is with this PR able to handle bomb-decay of the form proposed by Glasstone-Dolan: d_1 = d_0*t^-1.2 The usefulness of this method is still under testing.
Diffusion is now independent on model timestep by assuming a gaussian distribution and therefore being dependent on sqrt(timestep). This enlarges the diffusion slightly for shorter timesteps, i.e. < 15min for vertical diffusion and < 60min for horizontal diffusion.
The vertical diffusion for timesteps < 15min and within the boundary layer has been improved by using the real boundary-layer height in eta-levels rather than scaling the eta-levels to the boundary-layer height in m. In addition, the diffusion within the boundary layer allows now a 10% entrainment into the free atmosphere. This was documented already, but only implemented in the code for timesteps > 15min/full boundary layer mixing.
The new diffusion code was tested against the etex case, giving a much improved bias from nmb > 2 to nmb ~ 0.2 (and a minimal improvement against full boundary-layer mixing code (0.17 vs 0.23)).
SNAP is with this PR able to handle bomb-decay of the form proposed by Glasstone-Dolan: d_1 = d_0*t^-1.2 The usefulness of this method is still under testing.
Diffusion is now independent on model timestep by assuming a gaussian distribution and therefore being dependent on sqrt(timestep). This enlarges the diffusion slightly for shorter timesteps, i.e. < 15min for vertical diffusion and < 60min for horizontal diffusion.
The vertical diffusion for timesteps < 15min and within the boundary layer has been improved by using the real boundary-layer height in eta-levels rather than scaling the eta-levels to the boundary-layer height in m. In addition, the diffusion within the boundary layer allows now a 10% entrainment into the free atmosphere. This was documented already, but only implemented in the code for timesteps > 15min/full boundary layer mixing.
The new diffusion code was tested against the etex case, giving a much improved bias from nmb > 2 to nmb ~ 0.2 (and a minimal improvement against full boundary-layer mixing code (0.17 vs 0.23)).