| Forecasts generated by numerical weather prediction models continue to improve, but they are far from perfect. Forecast errors can be separated into those caused by shortcomings in the model (e.g., discretization and parameterization errors) and those caused by an imperfect estimate of the state of the atmosphere, oceans, and land surface (i.e., the initial conditions) given to the model, The goal of data assimilation is to eliminate the second class of errors to the greatest extent possible, given the observations at hand. Data assimilation is often treated as a statistical problem: Given a set of observations valid at some time, a previous forecast also valid at that time, and information about their error characteristics, what is the initial condition that is most likely to be closest to reality?; Morphological data assimilation is a complementary approach that seeks to use visual information, satellite data in particular, to help define the state of the atmosphere. The connection between satellite imagery and initial conditions in the form a model expects is made through the use of potential vorticity and its inversion.; The results presented in this dissertation document the construction of methods necessary to carry out morphological data assimilation in a more advanced and presumably accurate way than previous attempts. In particular, warping, wind partitioning, and potential vorticity inversion techniques are discussed within the context of morphological data assimilation. The potential vorticity inversion procedure is the most involved, and its discussion takes up the bulk of the work. Although the inversion procedure is not currently robust, prospects for the future are discussed that may lead to further improvement. |
