Despite the advances in the quality of numerical weather forecasts of extratropical cyclones, notable forecast busts still occur. These forecast busts may have been caused by the deficiencies in the NWP models and their assimilation schemes or due to deficiencies in the initial conditions. The growth and propagation mechanism of the forecast error associated with these deficiencies has been recognized as very important problem because it not only is related to dynamics but also has implications on practical issues such as data assimilation or adaptive observations.

A diagnosis of error evolution in a quasigeostrophic model is performed using potential vorticity inversion and localized Eliassen-Palm flux diagnostics following an ideal cyclone development under the perfect model assumption. The error evolution is compared with the evolution of the singular vector and perturbation derived from the gradient sensitivity. The diagnostics reveal that the mechanisms for error growth and propagation depend upon projection of error onto singular vectors of the flow. The error which projects onto these singular vectors during the evolution is initiated by the barotropic propagation of wave activity from regions of small potential vorticity gradient to regions of large potential vorticity gradient (i.e., near the jet in the middle of the domain) along the upper and lower boundaries of the quasigeostrophic model followed by the baroclinic interactions between errors along the upper and lower boundaries. That error which does not project onto singular vectors does not grow or propagate during the evolution.

In the view of adaptive observations, sensitive regions can be identified by the error growth and propagation mechanism. A forecast measure may be considered sensitive in some regions if the forecast measure exhibits a large response to a small change in the initial conditions in those regions. The relationship between the error growth mechanism and the sensitive regions will also be presented to address the alternative adaptive observation strategy.