Using principal component analysis and fuzzy c–means clustering for the assessment of air quality monitoring

Determining whether a reduction can be made in the total number of monitoring stations within the Air Quality Monitoring Network is very important since in case of necessity, the devices at one group of stations having similar air pollution characteristics can be transferred to another zone. This would significantly decrease the capital investment and operational cost. Therefore, the objective of this study was grouping the monitoring stations that share similar air pollution characteristics by using the methods of principal component analysis (PCA) and fuzzy c– means (FCM). In addition, this study also enables determining the emission sources, evaluating the performances of the methods and examining the zone in terms of pollution. In the classification of monitoring stations, different groups were formed depending on both the method of analysis and the type of pollutants. As a result of PCA, 5 and 3 classes have been determined for SO2 and PM10, respectively. This shows that the number of monitoring stations can be decreased. When reduced classes were analyzed, it was observed that a clear distinction cannot be made considering the affected source type. During the implementation of the FCM method, in order to facilitate comparison with the PCA, the monitoring stations were classified into 5 and 3 groups for SO2 and PM10, respectively. When the results were analyzed, it was seen that the uncertainty in PCA was reduced. When the two methods are compared, FCM was found to provide more significant results than PCA. The evaluation in terms of pollution, the results of the study showed that PM10 exceeded the limit values at all the monitoring stations, and SO2 exceeded the limit values at only 3 of the 22 stations.