Piecewise Regression Modeling of CDOM Absorption Spectra

RUIZ, M. G.; LUTZ, V.; FROUIN, R.J.

Vancouver

Conferencia; Ocean Optics XXIII; 2016

The Oceanography Society

Chromophoricdissolved organic matter (CDOM) is an important light-absorbing constituent inseawater. Yet, discrepancies still exist in the laboratory methods used todetermine CDOM absorption coefficients, and in the modeling of its spectralshape. Our objectives were: 1) to test some empirical factors concerning CDOMacquisition with a spectrophotometer; 2) to test the existence of differentspectral shapes using a set of 145 CDOM spectra (MV1102 cruise), and to findthe best way to model them. No significant differences were observed in CDOMabsorption coefficients obtained by filtration through 0.2 µm membrane or GFFfilters, due to not even pore-size in the former and to fiber contractionduring combustion of the latter. Two spectral shape groups were found bycluster analysis on the MV1102 dataset, which were mainly associated to?coastal? and ?oceanic? environments. A single exponential model (SEM), thusone spectral slope, is generally used to describe the CDOM absorption spectra.This method, however, can provide variable results according to how it isapplied (e.g., spectral range) and it does not fit equally well all spectra.Here, a segmented regression model (SRM) was used to model the CDOM absorptionspectra. The slopes and breakpoints of the SRM, retrieved in a systematic way,fitted all samples with higher accuracy than the SEM and also allowed discriminatingspectral shapes. Future work on the application of the SRM will contribute tothe understanding of CDOM dynamics and to improving satellite algorithms forCDOM absorption estimation.