The present work deals with the seasonal variations in the contribution of sources to PM2.5 and PM10 in Delhi, India. Samples of PM2.5 and PM10 were collected from January 2013 to December 2016 at an urban site of Delhi, India, and analyzed to evaluate their chemical components [organic carbon (OC), elemental carbon (EC), water-soluble inorganic components (WSICs), and major and trace elements]. The average concentrations of PM2.5 and PM10 were 131 ± 79 μg m-3 and 238 ± 106 μg m-3, respectively during the entire sampling period. The analyzed and seasonally segregated data sets of both PM2.5 and PM10 were used as input in the three different receptor models, i.e., principal component analysis-absolute principal component score (PCA-APCS), UNMIX, and positive matrix factorization (PMF), to achieve conjointly corroborated results. The present study deals with the implementation and comparison of results of three different multivariate receptor models (PCA-APCS, UNMIX, and PMF) on the same data sets that allowed a better understanding of the probable sources of PM2.5 and PM10 as well as the comportment of these sources with respect to different seasons. PCA-APCS, UNMIX, and PMF extracted similar sources but in different contributions to PM2.5 and PM10. All the three models extracted 7 similar sources while mutually confirmed the 4 major sources over Delhi, i.e., secondary aerosols, vehicular emissions, biomass burning, and soil dust, although the contribution of these sources varies seasonally. PCA-APCS and UNMIX analysis identified a less number of sources (besides mixed type) as compared to the PMF, which may cause erroneous interpretation of seasonal implications on source contribution to the PM mass concentration.
Keywords: PCA-APCS; PM10; PM2.5; PMF; Seasonal variability; UNMIX.