Predicting air pollution using fuzzy genetic linear membership kriging in GIS

Predicting air pollution is an important prerequisite for estimating, monitoring and mapping unknown pollution values. We can use fuzzy spatial prediction techniques to determine pollution concentration areas in practical situations where our observations are imprecise and vague. Fuzzy membership kriging with a semi-statistical membership function is an example of this type of technique. The implementation of fuzzy membership kriging extracts semi-statistical membership functions from data, and applies these functions to an indicator kriging model. Such functions, which can be linear or nonlinear, transform fuzzy data into membership degrees and grades.Evolutionary genetic algorithms (GAs) can improve prediction efficiency and make it easier to choose an optimum membership function for air pollution applications. In this paper, we used a GA to determine the threshold parameters for a fuzzy membership kriging function based on preprocessed data from Tehran, Iran. We measured particulate matter with a mass median aerodynamic diameter of less than 10 μm (PM10) concentrations at 52 sample stations in Tehran to identify areas that are dangerous for human health. After we predicted the PM10 data, our results showed that GAs reduce the estimated error (3.74) compared to linear functions (8.94 and 12.29). This study indicates that using a GA for optimizing membership functions can get higher estimated accuracy than fuzzy membership kriging for modeling uncertainty in the prediction process of PM10 data.