Dept. of Environmental and Biomolecular Systems
Oregon Health & Science University
There is a building consensus among scientists, educators, managers, and politicians that integrated ocean observatories are a critical backbone to future scientific exploration, workforce training, and science-based management of coastal resources in the U.S. An integral part of the future infrastructure for these observatories is data assimilation (DA)-a mathematical technique that uses sparse observations of the ocean to constrain and improve the dynamics of a numerical model. The uses of DA in a coastal margin observatory are multiple and include an optimal estimate of the ocean state, an estimate of uncertainty for this state estimate, a suggestion for improving the design of the observational array, and a suggestion for improving the formulation of the numerical model. However, the wide application of DA in coastal margin observatories has been hampered, among other things, by the computational cost of existing algorithms, by the logistical difficulties in developing adjoint codes for rapidly evolving coastal models, by the strong nonlinearity of coastal circulation processes, and by our ignorance about the statistics of model and forcing errors. In this dissertation, we overcome many of the algorithmic and logistical challenges that impede wide application of advanced DA algorithms in coastal margin observatories. We demonstrate an application of the developed methods in an observatory for the Columbia River (CR) estuary and plume-an excellent test-bed for developing DA methods, with well documented, yet challenging dynamics. Once implemented, we used DA system in the CR to: 1) Assimilate in situ measurements of water levels, salinity and temperature into a multi-annual hindcast of the estuary. 2) Study the impact of DA on the dynamics of ecologically significant circulation features in the estuary and plume, such as the orientation and size of the plume, and the length of the salinity intrusion in the estuary. 3) Guide optimal placement of observational arrays in the estuary and plume. 4) Develop a real-time, assimilative forecast system for the estuary. Our successful application of the enabling algorithms in the CR suggests that the developed technologies for fast, model independent DA and for optimization of observational arrays can be applied in many other coastal margin and coastal ocean observatories, enabling the implementation of these observatories at low computational and personnel costs.
OGI School of Science and Engineering
Frolov, Sergey, "Enabling technologies for fast, nonlinear data assimilation in a coastal margin observatory" (2007). Scholar Archive. 156.