Source Apportionment of Total Suspended Particulate Matter in Coarse and Fine Size Ranges Over Delhi

Source apportionment of total suspended particulate matter (TSPM) and associated heavy metals has been carried out for the city of Delhi using the Chemical Mass Balance Model, Version 8 (CMB8), as well as principle component analysis (PCA) of SPSS (Varimax Rotated Factor Matrix method) in coarseand fine-size mode. Urban particles were collected using a five-stage impactor at six sites in the winter of 2005-06. The impactor segregates the TSPM into five different size ranges (viz. > 10.9, 10.9– 5.4, 5.4–1.6, 1.6–0.7 and < 0.7 μm). Four samples were collected from six different sites every 24 hours. Samples were analyzed in five size ranges gravimetrically and chemically for the estimation of SPM and metals. The five different size ranges were divided into two broad categories: coarse (1.6 to > 10.9 μm) and fine (< 1.6 μm). The CMB8 and PCA were executed separately for both coarse and fine size ranges. Results obtained by CMB8 indicate the dominance of vehicular pollutants (62%), followed by crustal dust (35%) in the fine size range; while in the coarse size range crustal dust dominated (64%) over vehicular pollution (29%). Little contribution from paved-road dust and industrial sources was observed. Results of PCA (or factor analysis) reveal two major sources (vehicular and crustal resuspension) in both coarse and fine size ranges. The correlations of factors (sources) with the metals show that in the coarse size range the dominant source is crustal re-suspension (68%) followed by vehicular pollution (23%). However, this is reversed in the case of the fine size range factor analysis where vehicular pollution (86%) dominated over crustal re-suspension (10%).


INTRODUCTION
Delhi, the capital of India is one of the 10 most polluted cities in world. It is the second largest Indian city with a population of over 14 million, with 1.3% of India's population. The annual average growth rate of Delhi's population is 3.85%, but the annual average vehicular growth rate is 5.85% (Economic Survey of Delhi, 2006) Model. In the present investigation, we have applied both CMB and PCA for the source apportionment of TSPM in the ambient air of Delhi. CMB models are fundamental receptor models and very useful for coarse and fine particle source apportionment (CMB8, 1997).
CMB models estimate source contributions by determining the best-fit combination of emission source chemical composition profiles needed to reconstruct the chemical composition of ambient samples (Watson et al., 1991;Watson et al., 1994;Schauer et al., 1996).
While the PCA method focuses on cleaning up the factors, it also produces factors that have high correlations with one smaller set of variables and little or no correlation with another set of variables (Stevens, 1996).
However, in India studies on source apportionment by CMB are rather limited. Patil (1992, 1994) used CMB and PCA respectively for source apportionment of aerosols in Bombay. Srivastava (2004) and  carried out source apportionment of ambient VOCs in Mumbai and Delhi, respectively, using CMB8. Srivastava and Jain (2007a,c) carried out source apportionment studies using CMB and PCA, respectively, on data collected in 2001. In the present study, we assessed the air quality of Delhi at a time when almost all the public transport is compressed natural gas (CNG) fueled and metro rails have been introduced on some fixed routes. In this paper, an attempt has been made to assess the various sources of fine and coarse particulates in Delhi using CMB8 and PCA methods.  Table 1.

Sampling procedure
Sampling was done with a five-stage cascade particulate separator (Kimoto Electric Co. Ltd. Japan). It was operated at an average flow rate h to remove any moisture content, and then weighed by a precision micro balance (Mettler AE 50). Afterwards, they were mounted on the air sampler. After sampling, the filters were immediately transferred to vacuum desiccators to again de-moisturize in the same manner and then again weighed. The difference in pre-and post-sampling weights were used for the estimation of TSPM.

Analysis
Acid digestion, required for the metals determination by AAS, was carried out according to a standard procedure (Katz, 1977).
Acid digestion was performed in Teflon bombs following these steps.
Step 1: Sub samples of dry filters were dissolved in 3 mL HF, 6 mL HNO 3 and 1.5 mL HCl in Teflon bombs at 120℃ for 1 h; Step 2: Digestates were evaporated to dryness at 70℃ for 1/2 h; Step 3: Residues were re-dissolved in 10 mL of 10 M HNO 3 and evaporated to dryness.
Step 3 was repeated until residues were fully dissolved. A series of blanks were prepared using the same     (Stevens, 1996).

RESULTS AND DISCUSSION
Source apportionment was carried out for both coarse and fine size ranges of TSPM using
in both coarse and fine size fractions; i.e., 3 and 2%, respectively. The contribution from paved road dust is somewhat more for coarse particles (4%) than fine particles (1%). The results obtained in this study are not in agreement with those from the previous study by Srivastava and Jain (2007a) in which it was observed that vehicular pollution dominates in both coarse and fine size ranges. This study was carried out before the implementation of compressed natural gas vehicles (CNG).
Therefore, we can conclude that after the implementation of CNG, although vehicles continue to be the largest polluters of Delhi's ambient air, their over-contribution is now confined to fine particles only, while soil and crustal dust are now the predominant contributors to coarse particles.

Source apportionment by PCA
To validate the results obtained by CMB, a PCA was performed by the Varimax Rotated Factor Matrix method of SPSS for both coarse and fine size fractions. Sites were assumed to be repetitive of data, while metals were treated as the variable. A total of nine components were obtained, out of which two were extracted as principle components (Eigenvalue > 1) for both coarse and fine size ranges. The detailed results are given in Table 4. In case of coarse particles, two PCs were obtained, first and second PCs contribute about 68 and 23%, respectively.
Considering a correlation coefficient of ≥ 0.5, the metals correlated with factor 1 points it to be the crustal re-suspension source and factor 2 as the vehicular source. It is important to note that the metals generated from various sources Crustal resuspension Vehicular also become part of crustal dust over time (Khemani et al., 1985;Balachandran et al., 2000;Anju and Banerjee, 2003;Monkkonen et al., 2004;Khillare et al., 2004). The condition is reversed in the case of fine particulate source apportionment. Again, two PCs were obtained, contributing about 86 and