Sensitivity of Perturbation Growth to Flow Characteristics and Sampling Strategy

The chaotic nature of the weather/climate attractor intrinsically limits the deterministic skill of weather forecasts by promoting rapid growth of errors. In this study, such error growth is simulated by artificially perturbing the atmosphere at initial time, and its sensitivity to the chosen perturbation methodology and to the flow characteristics is investigated. The different simulations are integrated with the limited-area model LM run on a convection-resolving grid. Results demonstrate that the locations of growing disturbances are insensitive to the definition of the initial temperature perturbation. This can be explained through an analysis of the perturbation growth and propagation mechanisms. In particular, rapid radiation of the imposed initial disturbance through a sound wave and presence of specific flow characteristics (e.g. convective instability) appear to force localized error growth far remote from the initial perturbation.