Complexity and the power to understand it: Artificial Neural Networks Modelling approach on the evaluation of coastal environmental scenarios

Nuno Neves and Marco Freire

University of Évora (PT)

Key words: Complexity, coastal areas, artificial neural networks, spatial analysis, geographic data model

The management of coastal resources implies an adequate knowledge about the complex system of spatial relations involved. The complexity scenarios in spatial interactions are frequently very difficult to describe and even more to model in a computational based simulation.

This paper is an attempt to deal with the complexity of ecological effects on the environment of coastal areas resulting from the propagation of industrial and domestic pollutants through the drainage system associated to an area to be evaluated.

Classical modelling approach for the evaluation of runoff is usually based in watershed models that do not accurately represent the eminently local micro relations in space. In a pro-holistic perspective the watershed must be considered, but also is fundamental an attempt to describe the cellular spatial relations both locally, and integrating the general area effects.

Coastal areas are a good example of a complex spatial system involving a large set of dynamic phenomena interactions. Some of this interaction are relatively well known and so can be used as a basis for an exploratory analysis process.

Artificial Neural Networks (ANN) development is directly related to the process of knowledge acquisition in a similar way as considered to be performed by human brain. This inspiration for the emulation of the human neural system aims to develop "machines" capable to perform like humans on knowledge acquisition and in the use of that knowledge.

Artificial neural networks are commonly used on poorly defined problems with significant lack of knowledge on the processes involved. Usually ANN are used as a statistical estimator without a explicit spatial relation. This paper includes also a methodological discussion on the definition of a set of spatial variables to be combined in an exploratory knowledge discovery system.

A geographic model was defined and implemented considering Water Framework Directive (WFD) reference sites, Directive 2000/60/EC, who establishes a framework for Community action in the field of water policy, aiming good water status for all waters by 2015. The water classification system followed is focused on its ecological status expressed in ways of disturbance related to the quality of a reference site.

This approach is expected to improve the knowledge on coastal systems spatial relations, to be used in the definition of policies for coastal areas management ensuring the comprehensiveness of the conflicting complexity scenarios.