Application of regression and artificial neural network analysis of Red-Green-Blue image components in prediction of chlorophyll content in microalgae

A set of equations for determining chlorophyll a (Chl a) and accessory chlorophylls b, c2, c1 + c2 and the special case of Acaryochloris marina, which uses Chl d as its primary photosynthetic pigment and also has Chl a, have been developed for 90% acetone, methanol and ethanol solvents. These equations for different solvents give chlorophyll assays that are consistent with each other. No algorithms for Chl c compounds (c2, c1 + c2) in the presence of Chl a have previously been published for methanol or ethanol. The limits of detection (and inherent error, +/- 95% confidence limit), for chlorophylls in all organisms tested, was generally less than 0.1 microg/ml. The Chl a and b algorithms for green algae and land plants have very small inherent errors (< 0.01 microg/ml). Chl a and d algorithms for Acaryochloris marina are consistent with each other, giving estimates of Chl d/a ratios which are consistent with previously published estimates using HPLC and a rarely used algorithm originally published for diethyl ether in 1955. The statistical error structure of chlorophyll algorithms is discussed. The relative error of measurements of chlorophylls increases hyperbolically in diluted chlorophyll extracts because the inherent errors of the chlorophyll algorithms are constants independent of the magnitude of absorbance readings. For safety reasons, efficient extraction of chlorophylls and the convenience of being able to use polystyrene cuvettes, the algorithms for ethanol are recommended for routine assays of chlorophylls. The methanol algorithms would be convenient for assays associated with HPLC work.