UC San Diego

Observations of ocean color from satellite imagery provide valuable opportunities to monitor biogeochemical systems of the ocean over broad spatiotemporal scales. To fully leverage the potential of these measurements, it is necessary to continue research on the development, assessment, and validation of inversion methods that estimate inherent optical properties (IOPs), which physically regulate the characteristics of ocean color. However, existing inversion models often face limitations in retrieving a complete suite of IOPs, particularly in the retrievals of absorption coefficients of seawater constituents. This limitation is relevant in the context of the recently launched NASA PACE satellite mission which has the capability to measure hyperspectral radiances from the near-UV (350–400 nm) through the visible (400–700 nm). To address these challenges, this dissertation first focuses on advancing an absorption partitioning model that is incorporated into the broader structure of a four-step Semi-Analytical Algorithm (4SAA) designed to retrieve a full suite of seawater optical properties from satellite measurements. In the first chapter, a high-quality dataset of 1610 field samples of constituent absorption coefficients collected from diverse oceanic environments is compiled to develop a model to extrapolate the non-phytoplankton absorption coefficient, adg(λ), the non-algal absorption coefficient, ad(λ), and the colored dissolved organic matter absorption coefficient, ag(λ), into the near-UV from the visible spectral region. This extrapolation model is assessed and validated using the compiled in situ dataset and provides a method to extend absorption partitioning models constrained to the visible spectral region. In the second chapter, the development of an absorption partitioning model, called ADG, is described. This model separates adg(λ) into its constituents, ad(λ) and ag(λ), across the near-UV through visible spectral range. The ADG model is assessed and validated with the compiled dataset described in the previous chapter and demonstrates a strong capability to partition adg(λ). This model consists of two variants, one of which incorporates the extrapolation model described in the first chapter and is implemented as the final component of the 4SAA model. The third chapter describes the analysis of a comprehensive performance and uncertainty assessment of the 4SAA model. This analysis utilized a recently published synthetic optical database to quantify the performance of each component model of 4SAA and investigate error propagation through the multi-step inversion sequence of 4SAA. A Monte Carlo approach was implemented to evaluate the uncertainty introduced by each component model as well as the propagation of uncertainty of the complete 4SAA model.