Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Modelling of Interaction in Environment and Cod Using a Neural Network

 
: Fuchs, F.

International Council for Exploration of the Sea -ICES-:
International Council for Exploration of the Sea. Annual Science Conference 1996. Proceedings
Reykjavik, 1996
International Council for Exploration of the Sea (Annual Science Conference) <1996, Reykjavik>
Englisch
Konferenzbeitrag
Fraunhofer IGD ()
fishery; neural network; pattern recognition; simulation; spatial visualisation

Abstract
The domain of fishery hydrography is mainly the analysis of fish-environment relationships. In this paper a software tool is presented that supports "fishery hydrographic pattern recognition". With other words, complex hydrographical pattern will be used to describe cod distributions in the Baltic Sea. An artificial neural network is the basis of modelling the interaction in environment and fishes. It is trained with measurements of hydrological and corresponding fishery investigations. The hydrological data were provided by the monitoring cruise of the Institut für Ostseeforschung Warnemünde, the young fish survey data originate from the investigations of the Institut für Ostseefischerei Rostock der Bundesforschungsanstalt für Seefischerei Hamburg. Spatial hydrographical patterns and total biomasses represent the input in the neural network. The output of the net is the distribution of cod. Annual learning pattern exist from 1978 to 1993. This is the basis of training the neural netwo rk. To get a simplified interpretation of the input and output of the model, we create an interactive graphical user interface. The heart of the simulation environment on the computer is the scientific visualisation of the data. The visualisation of hydrographical as well as the fishery data results in a 3-D-model of the Baltic Sea. All inputs of the neural network are graphically interactively modifiable. Therefore, an aimed change of the hydrographical conditions is possible. How the changed hydrography influences the cod distribution is calculates by the trained neural network. The presentation of changed cod distribution in the 3-D-Baltic Sea topology should allow the user to recognise interactions between environments and fishes.

: http://publica.fraunhofer.de/dokumente/PX-24645.html