Laser monitoring of structure of water masses in the coastal zone of the North Sea using neural network classification algorithms

Tatiana A. Dolenko, Sergey A. Burikov, Irina V. Fadeeva
Physical Department, Moscow State University, Moscow, 119899, Russia
tdolenko@radio-msu.su, burikov@lid.phys.msu.su, ifadeeva@mail.ru

Rainer Reuter
Institute of Physics, Carl von Ossietzky University, 26111 Oldenburg, Germany
rainer.reuter@uni-oldenburg.de

ABSTRACT

The investigation of the structure of water masses in natural reservoirs is a current problem of oceanology and hydrology.

In this paper, it is proposed to use dissolved organic matter (DOM) as an indicator for water masses. The idea is based on the well-known fact that DOM of different genesis has different properties. The DOM concentration is much higher in rivers and lakes than in seas. The DOM chemical composition, including the relation of humic and fulvo acids, differs also. Humic and fulvo acids determine the DOM fluorescence band structure. If one can register small differences in fluorescence band parameters, then DOM can be used as an indicator for water masses.

It is shown that using artificial neural networks (ANN) for the DOM fluorescence analysis can allow to determine the water masses type in natural reservoirs. DOM fluorescence spectra were registered and processed for three reservoirs of different types: for the costal zone of the North Sea – in the water mixing zone of the Weser river, in the Wadden Sea and in the open North Sea. The measurements were performed in May 2003 during a cruise of R/V "Heincke".

The classification algorithm based on ANN was used for the analysis of the DOM fluorescence spectra. The obtained classes are presented on a digital map of the region, where they are displayed against spatial distributions of temperature, salinity, and the values of DOM fluorescence intensity normalized to the values of water Raman band intensity. The temporal variability of the water masses structure in the water mixing zone of the river Weser and in the North Sea water was registered. This variability is connected with tides. Therefore, the suggested method of classification of water masses by features of the shape of DOM fluorescence band supplements the widely used method of TS diagrams. It is important that both these methods can be implemented in remote sensing.

Last Update: 2005-04-7