Abstract
A method of simultaneous determination of temperature and salinity of water by Raman spectra was tested and validated on natural waters from the White Sea area. The basis of the method is the solution of this multi-parametric inverse problem by a modern technique of pattern recognition: artificial neural networks. Testing of the presented method was carried out on natural waters of unique meromictic lakes of the White Sea coast. The accuracy of determination of natural water parameters is 0.1°C for temperature and 0.2 psu for salinity in laboratory conditions in the investigated ranges of variation of parameters.