A hierarchy of 2 dimensional numerical models based on the rotating anelastic equations of motion are used to analyse the large scale flow response to short mountain waves (GWs) breaking. Interest focuses on the proportion of the large-scale flow (LSF) response which goes into inertial oscillations (IOs), inertio-gravity waves (IGWs), and balanced flow. In the simulations presented, the GWs have narrowband spectra in both frequency and horizontal wavenumber, they interact with the LSF at critical levels (CLs) located in a shear layer of limited vertical extent. In all cases considered, the minimum Richardson number of the background flow is large, so linear theory predicts that the GWs are absorbed and deposit momentum in the shear layer. Because the problem needs to embrace a wide range of scales, the nonlinear dynamics in the shear layer is controlled by large viscosity.

When the GWs amplitude is small, the ratio between unbalanced and balanced motion in the LSF response is well predicted by the temporal and spatial Rossby numbers associated with the large-scale forcing induced by the GWs. In the periodic case, the balanced motion is dominated by a transverse mean velocity that equilibrate the wave drag via the Coriolis torque. The unbalanced motion is an IO superimposed onto this balanced response. In the non-periodic case, the balanced motion is a large scale growing baroclinic wave, which patterns result from the steering by the shear of the potential vorticity dipole generated by the GWs drag at the location where the GWs break.

When the GWs are large enough that their dynamics is nonlinear, the amplitude of the GWs and of the changes they induce onto the LSF become comparable inside the shear layer. The LSF changes also make that the GWs can become nonlinearly reflected at the shear layer, so the GWs forcing can become less efficient in modifying the LSF then in the linear case. In the periodic case, a nonlinear feedback loop also makes that the ratio between IOs and balanced response in the mean flow much larger than in the linear case. In the non-periodic case, the ratio between balanced motion and IGWs in the LSF response is not much affected by nonlinarities. In this case nevertheless, the IGWs outside the of the shear layer can be large enough that they clearly make a very significant part(compared to the GWs) of the total wave signal.