The disciplines of meteorology and oceanography are both vital to understanding the earth system. Throughout most of the last half century, meteorology has largely been a prognostic discipline. Forecasts made by meteorologists have been widely used and scrutinized, allowing for countless opportunities to test and improve ideas about atmospheric circulation and physics. Since weather forecasts involve integrating numerical models and updating the model state via data assimilation, forecasting demands frequent use of the principles of Bayesian inference. This requirement essentially confronts the physics contained within numerical models at recurring intervals and can reveal systematic model bias.

In contrast, prognostic applications have been less prevalent in oceanography. Oceanographic forecasts are much rarer than atmospheric forecasts and, perhaps as a consequence of this disparity, many ideas concerning oceanic circulation have not been tested to the same degree as ideas concerning atmospheric circulation. This dissertation presents three methods for testing oceanographic ideas: applying common methodologies to analogous regions of different ocean basins; creating synthetic time series to mimic the properties of oceanographic time series in order to construct null distributions for hypothesis testing; and using water mass census information to interpret the results of water mass transformation analysis.