Hi all,
Is there something like a “common rule” for setting the atm- and rad timestep or the vertical levels for different resolutions in a nesting setup?
Does the timestep, for example, simply double when I double the resolution? (e.g. R02B04 720s, R02B05 360S)
Thanks a lot!
According to the recommended configuration (the typical case as NWP-Nesting with two domains), dt_conv and dt_sso should be reduced by half when horizontal resolution doubles, while other slow physics (e.g., rad and gwd) remain constant. Regarding the vertical levels, operational configuration sets model top height as 75 km with 90 levels using SLEVE coordinates but only 60 levels (~23 km) in nesting domains with lvert_nest = .TRUE..
However, I don’t know how much dt_xxx is best for different resolutions. I’m still confused by my diverse experiment results with different choices of dt_conv in subseasonal forecast. 
Ok thank you!
Is the vertical nesting a must for nested experiments or would it also be possible to run a nested simulation with lvert_nest = .FALSE. ?
This is possible. A nested simulation with lvert_nest = .FALSE. simply means that the parent domain and the child domain are using the same vertical level configuration, i.e. the same number of levels and over flat terrain the same height. The height of the actual levels will slightly differ due to differences in topography.
Regarding the time steps of slow physics components, I am not aware that a recommendation is documented somewhere. This is always a compromise between performance and the impact on the results. Furthermore, the intended application might also play a role in these considerations. Using the recommended NWP configurations is a good starting point.