Integrated ocean observatories are increasingly recognized as essential infrastructure for scientific discovery, workforce development, and effective coastal resource management. A key enabling component of these observatories is data assimilation (DA), which integrates sparse observations with numerical models to improve estimates of ocean state, uncertainty, and observing system design. Despite its promise, widespread DA adoption in coastal systems has been limited by computational cost, model nonlinearity, and logistical challenges. This dissertation presents fast, model-independent DA methods and demonstrates their application in the Columbia River estuary and plume, enabling multi-year hindcasts, analysis of ecologically important circulation features, optimization of observational arrays, and development of a real-time forecast system suitable for broad coastal observatory implementation.
