LWA Diagnostics

Development of a new diagnostic for large-amplitude eddies

We developed a diagnostic for Rossby Wave packets based on Local Finite Amplitude Wave Activity (LWA). LWA is proportional to the meridional displacement of Potential Vorticity (PV) from its equivalent latitude circle and therefore this diagnostic is able to evidence all the large scale meteorological features which are related to a deviation of PV with respect to a zonally symmetric state, such as Rossby Wave packets (towards which our interest is mainly focused), atmospheric blockings and PV cut offs.

Furthermore, we applied a smoothing filter to LWA in the zonal direction in order to remove its phase dependence (an intrinsic characteristic of LWA) and allow to diagnose large amplitude Rossby Waves as whole wave packets. An example of LWA diagnostic for a Rossby Wave packet case (April 2011, data from ECMWF ERA Interim reanalysis) is shown below:

Local finite amplitude wave activity (a) smoothed LWA (b) and potential vorticity (c) at 320K isentropic surface obtained using data from the ECMWF ERA Interim reanalysis for the 8th Apr 2011 at 00 UTC. Coordinate in the meridional direction in the first two panels is equivalent latitude.


Investigation on error growth using LWA diagnostic

The LWA diagnostic is applied to two ICON simulations which differ only for a perturbation in the stochastic convection parameterization scheme. The figures below show the smooth LWA for each run at day 1 (starting time), 7 and 14 and it can be seen how after some time the two fields start to differ significantly. This is due to the upscale error growth propagation from the small (convective) to the large (planetary waves) scales.
Run 1

Run 2

Smooth local wave activity on the 330 K isentrope for two ICON simulations at day 1 (starting time of both simulations), day 7 and day 14. Here only a portion of northern hemisphere is shown.
The following figure shows the RMSE (Root Mean Square Error) between the two runs averaged over the whole earth, evidencing how for this particular simulation the error saturation (and therefore the limit of predictability) at the planerary scale occurs after about 17 days.
Root mean square error averaged over the whole domain (global) for the smooth local wave activiy. The RMSE is calculated as the difference of the LWA field between the two ICON runs at each timestep.