Modeling of Marine Light Absorption and Ocean Color PDF Download

Author: Guangming Zheng
Publisher:
ISBN: 9781303001406
Category :
Languages : en
Pages : 203

Get Book Here

Book Description
Partitioning of the total non-water absorption coefficient of seawater, anw(lambda), into phytoplankton, aph(lambda), and non-phytoplankton, adg(lambda), components is important to research in ocean optics, biology, and biogeochemistry. I developed a partitioning model based on stacked-constraints approach, which requires weakly restrictive assumptions about the spectral slope of adg(lambda) and the spectral shape of aph(lambda). With a comprehensive set of inequality constraints, the model first derives a wide range of speculative solutions for adg(lambda) and aph(lambda) and then identifies feasible solutions. Final model outputs include the optimal solutions that agree well with measurements (with biases typically within ±5%), and a range of feasible solutions that encompasses the measured adg(lambda) and aph(lambda) with a probability> 90% at most wavelengths. I also developed another model for partitioning the spectral absorption coefficient of suspended marine particles, ap(lambda), into phytoplankton, aph(lambda), and non-algal, ad(lambda), components based on the stacked-constraints approach. Partitioning results of the model generally agree well with measurements and are superior in terms of error statistics compared with previous partitioning models. These results support the prospect for the applications of the partitioning models using the input data of anw(lambda) and ap(lambda) collected from various oceanographic and remote-sensing platforms. I also evaluated the performance of the Quasi-Analytical Algorithm (QAA) for deriving the spectral total absorption, a(lambda), and backscattering, bb(lambda), coefficients of seawater from input spectrum of remote-sensing reflectance, Rrs(lambda), using field data collected in the Arctic and lower-latitude open waters. The performance of QAA for estimating a(lambda) varies from very good to fair (bias on the order of ±10%) depending on light wavelength and the oceanic region. For bb(lambda), the QAA typically shows overestimation from small to as large as about 35%. A sensitivity analysis shows that the parameter u [-̲bb/(a+bb)] at the reference wavelength of 555 nm generally contributes the most significant bias to bb(lambda) at all wavelengths within the spectrum of visible light, whereas the interplay between u(555) and u(lambda) generally dominates the errors of QAA-derived a(lambda) except for the reference wavelength. Our findings provide guidance for future efforts towards refinement of the QAA and potentially also development of other inverse models.