Particularly as models move toward higher resolutions and more complex multi-physics interactions. The coupling between physical parameterizations and fluid dynamical cores shapes the major source of model error, numerical instability and uncertainties. The UFS-WM dynamics and physics are solved separately in a sequential way. And has a much larger time step and uses dt_atmost=k_split*dt_dyn, this means the UFS-WM calling dynamics multiple times before calling physics. The model error grows with the length of time step which means that the current UFS-WM is sensitive to the timestep. Then an new methodology is implemented as an option (PDC=.T./.F.) to account for the lack of physics tendencies in the dynamics which could briefly be on the second time step, we use the physics/coupling tendencies from the previous step and then allocate the tendency containers at a time. Reset the state variables to the saved state from the end of the previous time step, add the new tendencies piecemeal on each acoustic step.
Particularly as models move toward higher resolutions and more complex multi-physics interactions. The coupling between physical parameterizations and fluid dynamical cores shapes the major source of model error, numerical instability and uncertainties. The UFS-WM dynamics and physics are solved separately in a sequential way. And has a much larger time step and uses dt_atmost=k_split*dt_dyn, this means the UFS-WM calling dynamics multiple times before calling physics. The model error grows with the length of time step which means that the current UFS-WM is sensitive to the timestep. Then an new methodology is implemented as an option (PDC=.T./.F.) to account for the lack of physics tendencies in the dynamics which could briefly be on the second time step, we use the physics/coupling tendencies from the previous step and then allocate the tendency containers at a time. Reset the state variables to the saved state from the end of the previous time step, add the new tendencies piecemeal on each acoustic step.