A Computer Method for Separating Hard to Separate Dye Tracers

Abstract

Tracer tests are an irreplaceable tool for hydrogeologists. They are used to determine the paths of water flow between two spots in a catchment below the surface of the earth. Usually, hydrogeologists carry out tracer tests with only one tracer at a time, but sometimes two or more fluorescent substances are simultaneously injected into different spots and collected in a spring. Then, the resulting cocktail is analyzed by optical methods (fluorescence spectrometer) to separate the tracers and calculate their concentrations. Molecules with sufficiently different excitation spectra are easily separated. But two among the most frequently used tracers, uranine (Na fluoresceine) and eosine, are very close in this respect. Their separation is well-known to be difficult. Other examples are sodium naphthionate and amino G acid, two very useful tracers since they are colorless and therefore unnoticed in surface waters. The eluent of charcoal bags (fluocapteurs) is another example. Beside the released tracer, there is a very high fluorescence background of dissolved organic matter (DOM) from which it must be optically separated. The shape of the excitation spectrum of a fluorescent tracer can be approximated by a Gaussian curve. This curve is completely described by three parameters: peak wavelength, height, and width. The spectrum of a cocktail of two tracers is the sum of two such Gaussian curves. To separate these two curves, we use an algorithm based on the steepest descent in the parameter space to find the best set of 2 × 3 = 6 parameters of the model that best fits the measured curve. We achieve good separation even with a concentration ratio smaller than 1:10.