A key property of the equations governing atmospheric motion is their instability relative to small perturbations. Fields which are arbitrarily close to each other at some initial time evolve during an integration to the point where the difference between them (or the "error") is as large as the difference between two randomly chosen states of the system. In numerical weather prediction, uncertainty in the initial state grows as the forecast proceeds until details of the forecasted weather are no more valid than weather patterns for an arbitrarily chosen day. The objective of this research is to determine the sensitivity of a realistic atmospheric model to localized errors in the initial conditions. There are two main areas of investigation: (1) identifying the locations in which the errors grow most quickly in short and long range forecasts and (2) determining the extent to which a reduction in the initial error can extend predictability (the period for which a forecast is useful). The principal investigator will employ a simplified version of a global numerical model, developed by the National Meteorological Center, to study the relationships among the processes of instability, nonlinearity, and latent heat release during the course of the forecast. In addition, concepts from dynamical systems theory will be employed in an effort to understand the role of scale interactions.
