Uncertain spatial reasoning of environmental risks in GIS using genetic learning algorithms

Area of Science:

• Environmental Science
• Computer Science
• Artificial Intelligence

Background:

• Managing ecosystem damage from air pollution requires advanced modeling.
• Spatial reasoning systems with learning capabilities are crucial for complex environmental analysis.
• Genetic fuzzy systems offer flexible and robust rule learning for environmental risk assessment.

Purpose of the Study:

• To design and evaluate a spatial genetic fuzzy system for assessing environmental risks of air pollution.
• To specifically analyze air pollution from oil well fires during the Persian Gulf War.
• To compare the precision of different genetic learning (GL) algorithms for rule tuning.

Main Methods:

• Utilized a spatial genetic fuzzy system incorporating expert knowledge and genetic learning (GL) algorithms.
• Extracted and monitored dynamic pollution areas using NOAA satellite imagery stored in a spatial database.
• Evaluated various GL methods (e.g., GL (0.7, 0.03), GL (0.98, 0.03)) against Cordon learning using Landsat data.

Main Results:

• The GL (0.98, 0.03) algorithm demonstrated superior precision in learning and tuning rules for the air pollution risk assessment.
• Different GL parameter combinations yielded varying levels of precision in the environmental risk modeling.
• The system successfully integrated spatial data and expert knowledge for inference.

Conclusions:

• The spatial genetic fuzzy system, particularly with GL (0.98, 0.03), is an effective tool for environmental risk assessment of air pollution.
• Genetic fuzzy approaches enhance the precision and robustness of spatial reasoning systems in ecological management.
• Accurate modeling of air pollution impacts is vital for ecosystem protection and management strategies.