The Intelligent Agent for the sustainable management of the territory: An experimental approach using an Expert System for the management of complex phenomena.

The knowledge of the territory and the evaluation of the impact of local anthropic activities on it need a dynamic and relational approach to represent the qualitative variables of the environment for providing decision makers an helpful tool in planning the management of the territory.

This paper deals with the development of a system for the definition and implementation of a Decision Support System (DSS) – Expert System (ES) carried out combining the algorithmic potentialities of DSS (based on "formal" information) with those resolutive of an ES (based on "informal" information i.e. heuristic, personal judgments).

The developed system can be considered as a useful planning tool, able to provide information on the environmental impact of anthropic activities by examining their effects on groundwater quality. Anthropic activities, in fact, through different mechanisms, can worst groundwater quality because of chemical-, physical- and biological-pollution rising from civil, industrial and agricultural activities.

In the present study the municipal area of the Bisceglie town (southern Italy) has been investigated.

Given the peculiarity of such as area, characterized from a prevailing agricultural landuse, the problem matched in this study has been the local groundwater pollution caused by the use of pesticides. Therefore a methodology has been developed to assess the intrinsic vulnerability of a specific local aquifer by implementing the mathematical equations contained both in standard procedures (CNR – GNDCI method) and in a parametric managerial model (SINTACS), with relationships describing the interactions between different chemicals and subsoils.

According to such methodology, when a phenomenon cannot be modeled using it is necessary the integration the outputs of the GLEAMS model (a mathematical simulation to evaluate the effects of the agricultural practices on the groundwater) ( Leonard, et al., 1987 ), and qualitative representation obtained from expert people.

To the Expert System, it has been specifically implemented with information concerning either the degradation of pesticides to form metabolites as well as their diffusion through unsaturated zone and groundwater. This approach allowed avoiding the construction of a complex computational model where the determination of the variables would result extremely difficult because of the interactions among different pollutants. The results of such approach have been compared with chemical analyses carried out in groundwater samples.

The used methodology makes the developed system capable to face unstructured problems characterized by an elevated degree of uncertainty, such as the one considered, and also to forecast, in an appropriate way, the decision of a decision maker, this represents the purpose of cognitive science i.e. the base of Artificial Intelligence.