Inversion of In Situ Light Absorption and Attenuation Measurements to Estimate Constituent Concentrations in Optically Complex Shelf Seas

A deconvolution approach is presented to use spectral light absorption and attenuation data to estimate the concentration of the major nonwater compounds in complex shelf sea waters. The inversion procedure requires knowledge of local material‐specific inherent optical properties (SIOPs) which are determined from natural samples using a bio‐optical model that differentiates between Case I and Case II waters and uses least squares linear regression analysis to provide optimal SIOP values. A synthetic data set is used to demonstrate that the approach is fundamentally consistent and to test the sensitivity to injection of controlled levels of artificial noise into the input data. Self‐consistency of the approach is further demonstrated by application to field data collected in the Ligurian Sea, with chlorophyll (Chl), the nonbiogenic component of total suspended solids (TSSnd), and colored dissolved organic material (CDOM) retrieved with RMSE of 0.61 mg m−3, 0.35 g m−3, and 0.02 m−1, respectively. The utility of the approach is finally demonstrated by application to depth profiles of in situ absorption and attenuation data resulting in profiles of optically significant constituents with associated error bar estimates. The advantages of this procedure lie in the simple input requirements, the avoidance of error amplification, full exploitation of the available spectral information from both absorption and attenuation channels, and the reasonably successful retrieval of constituent concentrations in an optically complex shelf sea. Key Points Inversion of constituent concentrations from in situ measurements of inherent optical properties in complex waters Local material‐specific inherent optical properties derived by linear regression of partitioned data set into optically dominant constituent Special attention is paid to different sources of measurement uncertainties and error propagation through the inversion model