The information content of different optical spectral ranges for remote chlorophyll estimation in coastal waters

Abstract

Simple and successful methods have been presented for deriving chlorophyll a concentrations from optical reflectance spectra of seawater, using information from only a few wavelengths. The characteristic vector method allows all wavelengths in a set of (low resolution) spectra to be analysed for signatures characteristic of different, independently varying phenomena and can therefore be used to evaluate the information content of different spectral ranges for chlorophyll determinations.

This paper describes an eigenvector analysis of 56 sea surface water reflectance spectra acquired from a boat in protected coastal waters of British Columbia, Canada, using the IOS multichannel silicon diode spectrometer, during the period 29 July to 15 October 1981. Supporting measurements made at the same time show that the phytoplankton pigment and phaeopigment content varied between 4 and 14 mg m⁻³, and up to 200 mg m⁻³ in red-tide conditions.

Four eigenvectors are sufficient to account for over 99 per cent of the variance in all cases. The effects of chlorophyll absorption, particulate scattering and chlorophyll fluorescence are all evident in the eigenvectors. Linear combinations of the associated scaling factors can be used as estimators of chlorophyll content, and so provide algorithms for deducing chlorophyll a concentrations that make use of the full range of spectral information present in the data.

Such algorithms were derived for different spectral ranges in the data and the results were compared with those from the presently used radiance ratio and fluorescent line height algorithms. The characteristic vector method derives a better correlation between chlorophyll concentration and variations in the blue and green spectral region than is given by the use of a single ratio, showing that at least for the present data, a more complex algorithm can give improved results. For the red spectral range most of the information appears to be in the observed fluorescence line height.

The analysis derives as good a correlation with chlorophyll a using data from the red spectral region as from the blue/green region, showing that as much information is present at the longer wavelengths, in spite of the lower signal levels there. A suitably designed spectral scanner should be able to make use of this information for improved remote mapping of surface chlorophyll a concentrations in oceans and lakes.