The predictability of cyclogenesis evolving into two distinct scenarii is addressed within a bi-periodic spectral quasi-geostrophic model with uniform potential vorticity.

An initial condition has been carefully built up using finite amplitude top and bottom temperature anomalies. Weak amplitude upper-level modifications lead to dramatic modifications of the scenario (see the figure below)

A Monte-Carlo ensemble (2000 runs) has been run. It suggests the presence of two likely scenarii, the first with a well developped frontal-wave and the second one with an anticyclone (see the figure below).

Two ensembles of reasonable size (i.e. about 20 members) suitable for operational purpose have been run. The first one (yellow on the figure) , based on singular vectors as initial perturbations, fails at capturing the frontal wave. The second ensemble, based on initial low- and -upper level features perturbations in terms of amplitude and position, samples the pdf of the attractor pretty well (Plu and Arbogast, 2005).