Source apportionment of volatile organic compounds in the north-west Indo-Gangetic Plain using positive matrix factorisation model

A high temporal resolution dataset of 32 volatile organic compounds and trace gases measured at a suburban site in the North-West IGP in May 2012 was applied to investigate VOC sources using the EPA PMF 5.0 Model and comparison with three existing emission inventories. Six source factors were resolved by the PMF model. The profile of factor 1, "biofuel use and waste disposal" correlated best with the measured VOC source speciation profiles of domestic cooking (R=0.9), leaf-litter burning (R=0.8) and smouldering garbage fires (R=0.7). Factor 2, "wheat-residue burning" correlated with flaming wheat-residue fires (R=0.9), factor 3 "cars" with the source signatures of petrol-fuelled cars (R=0.6) and traffic-junction ambient air samples (R=0.8) and factor 4 "two-wheelers" with petrol-fuelled 2-wheelers (R=0.7). Factor 5 was identified as "industrial emissions and solvent use" while Factor 6 contained "mixed daytime sources" comprising of biogenic VOC emissions and photo-chemically formed VOCs. The six factors accounted for 23.9%, 23.6%, 22.5% (both traffic sources combined), 14.6% and 13.8% respectively, of the total VOC mass and 25.6%, 27.1%, 28.3%, 6.7% and 12.3% respectively, of the total O3 formation potential. We find that the PMF traffic factor contribution (26.6%) was comparable to GAINS (24%), whereas EDGAR (10.9%) has lower and REAS (36.2%) higher traffic emissions, respectively. The PMF results for industrial sources and solvent use (17.2%) compared best with EDGAR (13.7%), while REAS (22%) and GAINS (23.2%) overestimate the sum of industrial emissions and solvent use slightly. Wheat residue burning (27.9%) was absent in REAS and underestimated by GAINS (7%) and EDGAR (4.7%). The importance of biofuel use and waste disposal activities (28.3%) was overestimated by all three emission inventories accounting for 67.9%, 44.2% and 41.7% of the VOC emissions in EDGAR, GAINS and REAS, respectively. The results highlight need for improved representation in emission inventories over N.W. India.