The goal of this research is to increase our understanding of the theoretical and practical limits on atmospheric predictability to guide the development of new strategies for observing and utilizing data in an optimal manner, including defining the scientific principles for the development of an adaptive observation capability for the Navy's environmental prediction systems. In principle, this adaptive approach could revolutionize the methodology for determining initial conditions for numerical environmental prediction by coupling the data assimilation process interactively to the observation process. The objective of this research is to investigate the origin and growth of amplifying disturbances that profoundly influence atmospheric predictability in the 1-3 day time range using adjoint and singular vector methods. Increased understanding of rapid error growth is essential for identifying and controlling structures in forecast initial conditions that lead to large forecast errors, exercising intelligent control of our environmental observing systems, and improving methods of data assimilation. These objectives include the development and use of adjoint- and inverse-based tools to diagnose rapidly growing analysis errors in a post-time setting, which may have applications for real-time analysis corrections. The improvement of forecasts of high-impact weather, including landfalling tropical cyclones, is a major focus of this work.