Cluster analysis applied PM10 concentration data for determination of sources and spatial patterns in ambient air-quality of Khatmandu Valley

The statistical cluster analysis of particulate matter concentration, measured as PM10 in a network of air-quality monitoring stations located in Kathmandu Valley is presented. PM10 data for the period 2003-05 has been analysed. Spatial classification was attempted on the basis of ambient air-quality data. A hierarchical agglomeration schedule using linkages between groups by the Euclidean distance metric was used and Ward's strategy was followed to unite two clusters. Two distinct clusters were formed irrespective of the season. The pre-monsoon cluster of monitoring sites reflected similar rural characteristics. The other cluster comprising air-monitoring sites was typified in having urbanized areas with densely populated commercial characteristics with high motor vehicular traffic. The cluster characteristics of monitoring sites during the monsoon, post-monsoon and winter seasons reflected city-centric and distant area characteristics. Cluster analysis reiterates the fact that it is the nature of sources that matters, on the overall air quality of the area. Results of both cluster analysis and non-parametric tests suggest that in the Kathmandu Valley there is difference in pollutant levels across land-use types and topographical characters.