Developing and Evaluation of An Agent-based GIS to Identify the Impacts of Air Pollution over the Environmental Risk Areas

The aim of this paper, is designing and evaluating an agent based GIS to identify the environmental risk areas based on the smoke plumes, derived from NOAA satellite images, and the online modeling of impacts over an environmental phenomenon. Air pollution is an event that produces risk areas and effects on the environment in many ways. Therefore, it is difficult to determine full impacts of this phenomenon for risks. Whereas, several aspects of this phenomenon such as growth change of vegetation covers, can be considered as an urgent application to make the spatial decisions. To solve this problem, it can be considered the application of environmental damages of the smoke plumes of Kuwait's oil well fire on the southwest forest lands of Iran's territory during the 1991 Persian Golf war. By checking of NOAA-AVHRR midday images, it could be confirmed the movement of oil pollution via south west of Iran territory. The polluted inland areas can be outlined as natural environmental resources. Then, it is necessary to online track, quick estimate and evaluate of risk values in terms of natural forest covers due to the mentioned atmospheric pollution. This purpose will be accessible, if an integrated intelligent system was designed for making decision. Thus, GIS and Remote sensing tools and data, are used to identify dangerous sites prior to undertaking further analyses or field investigations. In this paper, one of objectives is to evaluate different methods in a spatial agent based system in order to assess environmental risks for air pollution application. In this system, different genetic learning methods are evaluated to determine proper rules for reasoning. With respect to the uncertain nature and indeterminate boundaries of input spatial data (pollution plumes and risk area), it is necessary to use predictive mode of knowledge base for inference. Then, for each snapshot, the stored inference rules are learned and tuned based on the uncertain conditions. Thus, it is essential to design an appropriate agent based system to use in the final inference process to obtain risk areas. Therefore, risk areas are determined, and different genetic learning methods are evaluated using Landsat TM satellite images sample points.