Two‐step linear regression based on spectral absorption to detect cyanobacteria in fresh water

Abstract

A simple spectral absorption method based on two‐step linear regression was used to identify and estimate the quantity of cyanobacteria (Cyanophyceae) among other phytoplankton species in mixed freshwater populations. We focused on four major algal groups, the Cyanophyceae, Chloro‐phyceae, Bacillariophyceae, and Dinophyceae, as typical representatives in the fresh waters of the kanto region of Japan, and dissolved organic carbon (DOC). In the first step, simple linear regression analysis was applied to determine the relationship between spectral absorption characteristics and concentration for each pure sample which contained only one of the four algal groups or DOC. In the second step, the resultant characteristics represented by gradient vectors were used to estimate concentrations of the four algal species and DOC in mixed samples by multiple linear regression analysis. We used the method described here to estimate the quantity of cyanobacteria among algal communities cultured in the laboratory under four different light conditions and in field samples. The method accounted for variations in the spectral characteristics of cyanobacteria owing to different light conditions, which were caused by changes in the ratio of phycocyanin to chlorophyll a. Variations in spectral characteristics of cyanobacteria under different light conditions were assessed geometrically using a vector subspace method based on principal component analysis (PCA). The vector subspace method allowed a detailed study of spectral variation of cyanobacteria recognition under different light conditions in 3‐D space.