Characterization of the phytoplankton in Lake Constance for classification by remote sensing

Abstract

Phytoplankton influences the spectral reflectance of most waters, thus it is accessible to remote sensing. This study investigates how the phytoplankton in Lake Constance (Bodensee) must be characterized to enable a quantitative determination of its concentration by remote sensing. A classification based on specific absorption spectra and neglecting species-specific differences in backs-cattering was found to be suitable. When an irradiance reflectance model is applied which accounts for the influence of additional absorbing and scattering water con-stituents and which is insensitive against moderate sunglint contamination of reflectance measu-rements, the concentrations of four phytoplankton spectral classes can be determined. The threshold concentration of a phytoplankton spectral class that is required before it can be distinguished from the other classes lies between 12 and 25 %, depending on the natural variability of its specific absorption spectrum. The accuracy of the determination of the total chlorophyll-a concentration is improved by 45 % compared to other remote sensing algorithms which neglect absorption differences between phytoplankton classes and determine the chlorophyll-a concentration from phytoplankton absorption at a single best-suited wavelength. The remaining error of chlorophyll-a determination is ± 31 % for the concentration range of 3.6-64 µg/l. It is comparable with the error when the concentration is derived from a microscopic analysis of water samples by considering the biovolumes. Below 3.6 µg/l the model cannot determine the chlorophyll-a concentration reliably from reflectance measurements.