Numerical modelling of natural water colour: implications for remote sensing and limnological studies

Based on suggested earlier optical models of Lakes Ontario and Ladoga, numerical modelling experiments have been carried out to the effect of revealing the responsiveness of chromaticity coordinates (x, y, z), dominant wavelength (lambda(dom)) and associated spectral purity (p) to the abundance in water of optically active components (OAC), i.e., phytoplankton (chl), suspended minerals (sm) and dissolved organic carbon (doc). It has been shown that highly turbid waters (i.e., waters with high chi and sm concentrations) with low content of doc display colour varying from green to brownish. High turbidity or large doc concentration invariably is characteristic of waters with brown colour. With growing OAC content in water, the chromaticity coordinates and, consequently, the dominant wavelength lambda(dom) tend asymptotically to respective limit-values that seem to be intrinsically characteristic of natural waters. It is also shown that the colour purity p asymptotically tends to values of about 35-45 per cent (with the only exception for waters containing exclusively chl and this in small quantities (less than or equal to 0.5 mu g l(-1))) when concentrations of one or more OAC are over 10 (in respective concentration units). These findings clearly indicate that neither panchromatic nor two-channel ratio images could be meaningful for an unambiguous inference of water quality parameters. Furthermore, the correspondence between water colour physical characteristics (i.e., x, y, z, lambda(dom) and p) and water colour scales traditionally used in limnology is established.