Date of Award

Fall 2004

Project Type


Program or Major

Earth Sciences: Oceanography

Degree Name

Doctor of Philosophy

First Advisor

Janet W Campbell


The ocean color as observed near the water surface is determined mainly by dissolved and particulate substances, known as "optically-active constituents," in the upper water column. The goal of ocean color modeling is to interpret an ocean color spectrum quantitatively to estimate the suite of optically-active constituents near the surface. In recent years, ocean color modeling efforts have been centering upon three major optically-active constituents: chlorophyll concentration, colored dissolved organic matter, and scattering particulates. Many challenges are still being faced in this arena. This thesis generally addresses and deals with some critical issues in ocean color modeling.

In chapter one, an extensive literature survey on ocean color modeling is given. A general ocean color model is presented to identify critical candidate uncertainty sources in modeling the ocean color. The goal for this thesis study is then defined as well as some specific objectives. Finally, a general overview of the dissertation is portrayed, defining each of the follow-up chapters to target some relevant objectives.

In chapter two, a general approach is presented to quantify constituent concentration retrieval errors induced by uncertainties in inherent optical property (IOP) submodels of a semi-analytical forward model. Chlorophyll concentrations are retrieved by inverting a forward model with nonlinear IOPs. The study demonstrates how uncertainties in individual IOP submodels influence the accuracy of the chlorophyll concentration retrieval at different chlorophyll concentration levels. The important finding for this study shows that precise knowledge of spectral shapes of IOP submodels is critical for accurate chlorophyll retrieval, suggesting an improvement in retrieval accuracy requires precise spectral IOP measurements.

In chapter three, three distinct inversion techniques, namely, nonlinear optimization (NLO), principal component analysis (PCA) and artificial neural network (ANN) are compared to assess their inversion performances to retrieve optically-active constituents for a complex nonlinear bio-optical system simulated by a semi-analytical ocean color model. A well-designed simulation scheme was implemented to simulate waters of different bio-optical complexity, and then the three inversion methods were applied to these simulated datasets for performance evaluation.

In chapter four, an approach is presented for optimally parameterizing an irradiance reflectance model on the basis of a bio-optical dataset made at 45 stations in the Tokyo Bay and nearby regions between 1982 and 1984. (Abstract shortened by UMI.).