Simple statistical formulas for estimating biogeochemical properties of suspended particulate matter in the southern Baltic Sea potentially useful for optical remote sensing applications

Abstract

Simple statistical formulas for estimating various biogeochemical properties of suspended particulate matter in the southern Baltic Sea are presented in this paper. These include formulas for estimating mass concentrations of suspended particulate matter (SPM), particulate organic matter (POM), particulate organic carbon (POC) and total chlorophyll a (Chl a). Two different approaches have been adopted. The first approach was to use the available empirical material (the results of field measurements and laboratory analyses of discrete water samples) and find statistical formulas for estimating the biogeochemical properties of suspended particulate matter from those of inherent optical properties (IOPs), which are potentially retrievable from remote sensing measurements. The second approach was to find formulas that would enable biogeochemical properties of suspended particulate matter to be estimated directly from spectral values of the remote-sensing reflectance Rrs. The latter was based on statistical analyses of a synthetic data set of Rrs obtained from numerical simulations of radiative transfer for which the available empirical material on seawater IOPs and biogeochemistry served as input data. Among the empirical formulas based on seawater IOPs that could be used as a step in two-stage remote sensing algorithms (the other step is estimating certain IOPs from reflectance), the best error statistics are found for estimates of SPM and POM from the particulate backscattering coefficient bbp in the blue region of light wavelengths (443nm), and for estimates of POC and Chl a from the coefficient of light absorption by the sum of all non-water (i.e. suspended and dissolved) constituents of seawater an, in the blue (443nm) and green (555nm) parts of the spectrum respectively. For the semi-empirical formulas under consideration, which could serve as starting points in the development of local onestage (direct) remote sensing algorithms, the best error statistics are found when SPM, POM and POC are estimated from the same blue-to-red band reflectance ratio (Rrs(490)/Rrs(645)) (with estimated SPM reaching a better precision than estimated POM and POC), and when Chl a is estimated from the green-to-red band ratio (Rrs(555)/Rrs(645)).