Perturbations That Optimally Trigger Weather Regimes

The sensitivity of the onset of two weather regimes with respect to initial conditions is studied. The weather regimes are a Euro–Atlantic blocking regime and a Euro–Atlantic strong zonal flow regime. Both regimes are characterized by the same anomaly pattern but with opposite sign. Using a three-level quasigeostrophic T21 model and its tangent linear and adjoint versions, initial perturbations are computed that have the largest projection on this anomaly pattern at a prescribed forecast time. The tangent linear and adjoint techniques can be used only to describe linear error growth. However, with an iterative procedure, nonlinear error growth can be taken into account. In this way perturbations can be computed that trigger the onset of a weather regime in the linear range (even optimally) as well as in the nonlinear range. It is shown that moderate initial perturbations occasionally trigger a transition from a blocking regime to a zonal flow regime, or vice versa, within 3 days. For an optimization time of 6 days the iteratively computed perturbations generate such transitions for almost all investigated cases. The perturbations are compared with regional singular vectors, which are the linearly fastest-growing perturbations in the Euro-Atlantic area. In the linear range the perturbations project mainly onto the leading regional singular vectors. In the nonlinear range the projection onto 1inearly slower-growing regional singular vector is stronger. The method can easily be generalized to study the sensitivity for a transition to any weather regime or anomaly pattern. This approach can be useful in generating specific initial conditions for ensemble forecasting.