Sources Affecting a Semi-Arid Coastal Region Using a Chemical Mass Balance Model

Chemical mass balance (CMB 8.2), a source apportionment model, was employed to identify the sources influencing the measured PM2.5 levels at an industrialized urban and a coastal rural site located in Corpus Christi, Texas. A speciated PM2.5 dataset consisting of 110 common sampling days with 25 key species, including elements, water soluble ions, organic and elemental carbon measured from 2003 to 2005, was used in this analysis. Based on the local and regional emissions characteristics, thirteen generic source profiles were selected from US EPA’s SPECIATE library for the CMB model application. Secondary sulfate was the major contributor at both sites with average concentrations of 3.45 μg/m (42% of the apportioned mass) and 3.06 μg/m (37%), respectively. Secondary organic aerosols were observed to be higher at the urban site (1.62 μg/m) than at the rural one (1.07 μg/m). Due to its location being closer to the Gulf of Mexico, the influence of marine aerosols was higher at the coastal rural site (1.15 μg/m) than at the urban one (0.37 μg/m). Unique sources, including the petroleum industry and industrial manufacturing, were found to influence the measured PM2.5 levels at the industrialized urban site, with average apportioned concentrations of 0.17 μg/m and 0.02 μg/m, respectively. The annual average PM2.5 concentrations showed a gradual increase in the secondary components, including sulfates and organic aerosols at both sites during the study period.


INTRODUCTION
Aerosol composition in an urban environment has adverse effects on human health causing increased mortality and morbidity (WHO, 2003(WHO, , 2005;;Polichetti, et al., 2009) and other detrimental effects on environment such as reduced visibility, acid rain and aesthetic damage (Grantz et al., 2003;Tsai and Tainan, 2005).Aerosols are classified on the basis of aerodynamic diameter into fine particulate matter -PM 2.5 (PM with diameter 2.5 m and less) and coarse particulate matter -PM 10 (PM with diameter 10 m and less).Epidemiological studies have associated PM 2.5 with adverse health effects primarily in susceptible population of young children below 13 years of age (Dreher and Costa, 2002;Parker et al., 2005;Maté et al., 2010).Some of these health effects include increased respiratory problems, growth retardation, birth defects, and cardiovascular problems (Gilboa et al., 2005;Pope and Dockery, 2006).The United States Environmental Protection Agency (USEPA) revised the National Ambient Air Quality Standards (NAAQS) for PM 2.5 on 17 December, 2006.The PM 2.5 standards currently in place are based on the 3-year average of the annual arithmetic mean and the 3-year average of the 98 th percentile value of the 24-hr averaged PM 2.5 concentrations and these values should not exceed 15 and 35 μg/m 3 , respectively (USEPA, 2006).
Local anthropogenic emissions and long-range transport from natural and anthropogenic sources under favourable meteorological conditions have been identified as the primary sources of PM 2.5 (Park et al., 2004;Cohen et al., 2008;Karnae and John, 2011).PM 2.5 is a non-specific pollutant class and its toxicity depends on the source as well as its chemical composition.Thus, the identification and quantification of influential sources is a critical component for designing effective control strategies for the implementation of PM regulations (Rizzo and Scheff, 2005;Gupta et al., 2007).Prior to the development of sourcereceptor models, apriori compiled emission inventories were used in source apportionment studies (Friedlander, 2000).However, emissions inventories compiled for a region did not account for secondary aerosols like sulfates, nitrates and organic aerosols formed by complex photochemical reactions of gaseous pollutants.Source-oriented models such as Community Multi-scale Air Quality (CMAQ) models and source receptor models based on statistical approaches accounting for the secondary pollutant formation have been developed and successfully employed for comprehensive source apportionment analysis of various pollutants.Source receptor models used in various source apportionment analysis of air pollutants include chemical mass chemical mass balance (CMB), principal component analysis/absolute principal component scores (PCA/APCS), UNMIX and positive matrix factorization (PMF).Chemical mass balance (CMB) apportions sources based on the source profile database provided as an input while principal component analysis/absolute principal component scores (PCA/APCS), UNMIX and PMF are data driven statistical models.PCA/APCS, UNMIX and PMF have been successfully applied to identify sources of PM 2.5 and PM 10 measured in urban areas with diverse geography (Lee et al., 2008;Begum et al., 2010;Kothai et al., 2011;Oh et al., 2011).
Chemical mass balance (CMB) was developed by Hidy and Friedlander (1971) based on effective-variance least squares approach.It has been approved by Environmental Protection Agency (EPA) as a regulatory air quality model used for source apportionment in the development of state implementation plans (SIP).It requires input of sourceprofiles and speciation data measured at the receptor site and has been successfully employed in various source apportionment studies of poly-aromatic hydrocarbons (PAHs) and volatile organic compounds (VOC) (Latella et al., 2005;Srivastava et al., 2005;Hanedar et al., 2011).It has also been applied for source apportionment of coarse particulate matter (PM 10 ) measured in Harbin, China by Huang et al. (2010), and in Taiyuan, China by Zeng et al. (2010) in combination with PCA/MLR.CMB 8.2 has been widely applied to apportion PM 2.5 sources impacting ambient urban air quality in U.S.A., India, Korea, Turkey, and China (Park and Kim, 2005;Srivastava and Jain, 2007;Lee et al., 2008;Yatkin and Bayram, 2008;Chen et al., 2010;Yin et al., 2010).CMB 8.2 has been shown to better capture longterm temporal variation of source impacts at a receptor site than the emissions-based model like CMAQ and data driven statistical model like PMF (Marmur et al., 2006;Yatkin and Bayram, 2008).Thus along with composite source apportionment, CMB 8.2 has been successfully applied to study the seasonal variations in the sources contributing to PM 2.5 concentrations measured during the wild fires of 2000& 2001in Missoula Valley, Montana (Ward and Smith, 2005) and in Delhi, the largest metropolis urban area in India (Srivastava and Jain, 2007).Vega et al. (2010) have applied the model to apportion PM 2.5 sources in Mexico City and used the results as inputs for dispersion models to study the spatial variability in the particulate matter pollution.
Corpus Christi, located in a semi-arid region, is the eighth largest Consolidated Metropolitan Statistical Area (CMSA) in Texas.It is a fast growing industrialized urban area with a cluster of petrochemical and oil refineries located along the ship channel adjacent to the sixth largest port in the USA.As an initial effort Karnae and John (2011) have applied PMF2 to identify and quantify sources contributing to the PM 2.5 concentrations measured during 2003 through 2008 at a monitoring site located within the urban airshed.Potential source contribution function (PSCF) analysis conducted as a part of this study identified the impact of long-range transport on the measured levels of PM 2.5 in the urban airshed during three regional air pollution events along with several local anthropogenic sources.The three regional air pollution events included agricultural burns during early spring in Mexico and Central America, sub-Saharan dust storm events of Africa during summer and haze transport from highly industrialized areas of Texas and surrounding states during the fall months.Texas Commission on Environmental Quality (TCEQ) operated an additional speciation site located in a rural part of the sea shore on Padre Island along the Gulf of Mexico during 2003 through August 2005.Based on the geographic location, this monitoring site was classified as a coastal rural site and it had minimal influence of local anthropogenic sources.Thus, the primary objective of this study was to identify and quantify the sources influencing PM 2.5 concentrations measured at the industrialized urban siteand at the coastal rural site during 2003 through August 2005 using CMB 8.2.As there were no published source profiles specific to the study area and creating one was beyond the scope of this study, generic source profiles from EPA's SPECIATE database were used.As suggested, the results from this study along with findings from earlier studies will provide policy-makers with tools to understand the influence of local sources and long-range transport during regional air pollution events affecting the urban airshed and to develop effective mitigation strategies.

PM 2.5 Monitoring and Chemical Analysis
The Texas Commission on Environmental Quality (TCEQ) maintains and operates several CAMS sites measuring ozone, PM, meteorological, and other parameters in the urban airshed.CAMS 635 and CAMS 314 are equipped with filter sample collectors for further speciation analysis.Thus, for this study PM 2.5 speciation data from CAMS 635 and CAMS 314 were acquired and utilized.CAMS 635 located at 5707 Up River road (27°48'43''N, 97°27'57''W) is classified as an industrialized urban site, while CAMS 314 located at 20420 Park road (27°25'37''N, 97°17'55''W) is classified as a coastal rural site.Fig. 1 shows the physical locations of each of the monitoring sites on a map of South Texas.
Twenty-four hour average PM 2.5 filter samples were collected on pre-conditioned and pre-weighed 47-mm diameter Whatman teflon and quartz filters at a frequency of once every six days by TCEQ at CAMS 635 andCAMS 314 during 2003-2005.Gravimetric analysis of Teflon filters was performed to identify the mass of PM 2.5 collected.The filters were then stored at 4°C and shipped by TCEQ to Research Triangle Institute (RTI) in North Carolina for further chemical analysis.RTI was selected by EPA for chemical speciation of PM 2.5 filter samples for the nationwide monitoring network.The laboratory and the instrumentation used for chemical characterization at RTI meets the Federal Reference Method (FRM) requirements for PM 2.5 analysis.Filter samples acquired were preconditioned to room temperature prior to chemical analysis.The pre-conditioned Teflon filters were analyzed for elements using energy dispersive x-ray fluorescence (EDXRF) and were extracted into water using an ultra sonicator for the measurement of anions and cations using ion chromatography (IC).The pre-conditioned quartz fiber filters were used to measure for carbon species (elemental and organic carbon) and its fractions using thermo-optical transmittance (TOT).Quality assurance and quality control of the speciation data was performed using the blank filter sample analysis and method detection limits (RTI, 2012).The validated data along with analytical uncertainty information was subsequently reported to TCEQ.

Data Acquisition
Chemical composition of samples measured at the receptor sites along with each source type with appropriate composition profiles were used as the key input parameters for the CMB model.Speciation data measured at CAMS 635 and 314 during 2003 through 2005 was acquired from TCEQ for this study.The data set consisted of 177 and 162 observations of concentrations for 55 species measured at CAMS 635 and CAMS 314, respectively.Quality assurance and quality control checks were performed to validate the dataset and to allow for the handling of missing values.To reduce the uncertainty and to provide better identification of source profiles, species with more than 50% missing observations and those below method detection limit were excluded from further analysis.Thus, 25 key species including elements (Al, K, Fe, Ca, Si, Mg, Ti, Zn, Cd, Cr, Mn, Cu, Ni, Pb, Sb, Sn, Sr and S), water soluble ions (Na + , Cl -, NH 4 + , NO 3 -, and SO 4 2- ), elemental carbon (EC) and organic carbon (OC) were included in this study.A total of 110 common observation days were selected for this study to quantify and compare the impact of various sources affecting the urban and coastal sites.Meteorological data including resultant wind speed, resultant wind direction, atmospheric temperature and humidity measured at these sites during the sampling days were also acquired from TCEQ.The data was used to study the predominant weather conditions during days with PM levels above the 75 th percentile and were classified as high PM days.

Chemical Mass Balance (CMB) Modeling
CMB modeling is a fundamental source receptor analysis technique developed to measure the chemical and physical characteristics of gases and particles measured at both the source and the receptor site.The basic assumption of the model is that ambient chemical concentrations measured at the receptor are expressed as the sum of the products of species and source contributions as shown in Eq. (1).

Texas
Corpus Chris where, C ij is the concentration of the i th element (i = 1, 2, 3, …, m) measured in the j th sample (j = 1, 2, 3, …, n) with m as the total number of species or elements and n as the total number of samples.f ik is the fractional source composition for the i th species from the k th source (k = 1, 2, 3, …, p) with p as the total number or sources and S kj is the mass concentration of the species from the k th sources contribution to the j th sample.Source contributions are identified using ambient concentrations and source profiles as input data for Eq. ( 1).Consistency of source composition over the entire sampling period was one of the major model assumptions considered during the study.The other assumptions included zero reactivity of chemical species, all chemical species are additively linear and the number of sources were less than or equal to the number of species (Li et al., 2003;USEPA, 2004).The degree to which these assumptions are met depends to a large extent on the properties measured at each of the source and receptor sites.CMB version 8.2 with effective variance least squares estimation method was employed in this study to identify and quantify the sources.It incorporates precision estimates for all of the input data into the solution, thus any errors in the input data are propagated to the model outputs.The use of regression analysis minimizes the differences between the measured and model predicted values and the fitting statistics of the model results are provided (Ramadan et al., 2000).A calculation is considered a good fit if coefficient of determination (R 2 ) falls in the range of 0-1.0.Chi-square values less than 1 indicates a very good fit to the data, while values between 1 and 2 are in the acceptable range.Various source apportionment studies using CMB also employed similar model evaluation parameters (Ramadan et al., 2000;Li et al., 2003).

RESULTS AND DISCUSSION
The Corpus Christi urban area is currently in compliance with both the primary and secondary National Ambient Air Quality Standards (NAAQS) for PM 2.5 .The annual average concentrations of PM 2.5 (total mass from filter samples) at CAMS 635 (urban site) during 2003-2005 were observed to be 7.80 μg/m 3 , 8.09 μg/m 3 and 10.64 μg/m 3 , respectively, while at CAMS 314 (coastal site), the concentrations were observed to be 7.19 μg/m 3 , 8.06 μg/m 3 and 10.77 μg/m 3 , respectively.As shown above the annual average filter mass concentrations were also below the primary PM 2.5 NAAQS of 15.0 μg/m 3 but exhibited a gradually increasing trend.The two monitoring sites were situated in distinctly different locations with impact from varying sources.A detailed analysis of the measured chemical characteristics and prevailing meteorological conditions was conducted.The results were then used to enhance the source apportionment analysis using CMB 8.2.

Chemical Characteristics of Measured PM 2.5
Chemical composition data acquired from TCEQ was used for the identification of the dominant chemical species observed at CAMS 635 and CAMS 314.Twenty five key species were selected with less than 50% of the observations with missing values and/or below the method detection limit.These included elements (Al, K, Fe, Ca, Si, Mg, Ti, Zn, Cd, Cr, Mn, Cu, Ni, Pb, Sb, Sn, Sr and S), water soluble ions (Na + , Cl -, NH 4 + , NO 3 -, and SO 4 2-), elemental carbon (EC) and organic carbon (OC) and these accounted for 97.1% and 93.3% of the filter mass at CAMS 635 and CAMS 314, respectively.
At CAMS 635, sulfates were found to be the major component of PM 2.5 levels accounting for 33.1% of the PM 2.5 mass.OC from field burning or biomass burning was observed to be the second largest component accounting for 19.7% followed by ammonium ion (8.8%), sulfur (11.6%) and EC (2.0%).The other significant species included sodium ion (2.4%), nitrate ion (2.2%), silicon (1.9%), and chloride ion (1.7%).The other measured species accounted for the remainder 1.7% of the total measured PM 2.5 mass.Similar chemical characteristics were observed at CAMS 314 with sulfates being the single largest component accounting for 33.9% of the measured PM 2.5 mass concentrations.The other major components included OC (25.9%), ammonium ion (10.5%) from marine aerosols and vessels, sulfur (11.7%),EC (4.1%), followed by sodium ion (4.5%), nitrate ion (2.3%) and chloride ion (5.3%).Silicon was found to account for 2.1% of the total measured PM 2.5 mass.Higher marine aerosol compositions as shown by the presence of sodium and chloride ions were observed at CAMS 314 primarily due to its coastal location as compared with CAMS 635, which is located slightly inland and within an industrialized urban region.
Descriptive statistics including mean, standard deviation, minimum and maximum of total PM 2.5 mass and species specific data observed at CAMS 635 and CAMS 314 are shown in Table 1.Pearson correlation coefficients (r) between the key species selected were estimated using Statistica  software as shown in Table 2. Strong correlations between ammonium vs. sulfur (0.92) and ammonium vs. sulfates (0.93) were observed at CAMS 635 suggesting the influence of local petrochemical refineries, industrial manufacturing sources and marine aerosols.The correlation coefficients between ammonium vs. sulfur and ammonium vs. sulfates at CAMS 314 were observed to be 0.88 and 0.72, indicating slightly lower influence of local sources such as petrochemical refineries and industrial manufacturing.Similar results were reported in a study conducted by Chen et al. (2001).Being located in an industrialized urban area, higher correlation coefficient (0.98) between sulfur and sulfates was observed at CAMS 635 indicating stronger influence of local anthropogenic sources.Whereas at the coastal site (CAMS 314), the correlation coefficient was observed to be 0.76 suggesting slightly lower contribution from local anthropogenic sources.Higher correlation coefficients were observed between sodium and chloride at CAMS 314 as shown in Table 2, suggesting a stronger influence of marine aerosols.Significantly high correlation coefficients were estimated between elemental species

Meteorological Characteristics of Sampling Days
Formation of particulate matter by the gas-to-particle conversion processes has been studied to be a dominant source of particulate matter in addition to direct emissions (Meng et al., 1997;Lee et al., 2008).Higher temperatures leads to an increase in the photochemical reactions converting gas phase SO 2 into sulfates, while higher humidity will accelerate the conversion of aqueous phase SO 2 to sulfates as well.Gas and particulate phase conversions of volatile species including NH 3 andHNO 3 into nitrates are limited by the thermodynamic equilibrium.Thus, it is critical to study the influence of the prevailing meteorological conditions including ambient air temperature and relative humidity.Annual ambient air temperatures ranged from 6.4°C to 35.6°C at CAMS 635 and 6.6°C to 36.7°C at CAMS 314, respectively.Temperatures ranging between 20°C to 35°C and humidity ranging from 50% to 60% were observed during the days when high PM filter mass concentrations were measured (days with PM 2.5 filter mass concentrations greater than 75 th percentile of the total observations) suggesting relatively warm and moderately humid conditions responsible for the formation of secondary aerosols by photochemical oxidation.
In addition to local direct emissions and gas-to-particle phase conversions, the influence of long range transport from surrounding polluted areas has also been observed in this region.The prevailing wind conditions during the sampling days were analyzed using wind rose as shown in Fig. 2. At CAMS 635, 77% of the times, winds with speeds between 2.1-3.6 m/s from south and south-east were observed during the warm months.Southerly winds during agricultural burning events in Mexico and Central America during April and May and southeasterly winds coinciding with sub-Saharan dust transport from Africa during June and July influenced the study region.At the coastal rural site (CAMS 314) dominant winds from south, south-south-east and south-east were observed as shown in Fig. 2.About 62% of the observed wind speeds ranged from 0.5 m/s to 3.6 m/s, while 37% of the winds measured during the sampling days had higher than average speeds of 3.6 m/s-5.6 m/s.

Source Apportionment Using CMB 8.2
The key to successful application of CMB is to identify the source profiles that are consistent with the measurements at the receptor locations.Source profiles are the mass abundances (fraction of total mass) of a chemical species in source emissions and are regarded as a category of sources  rather than individual emitters (Eatough et al., 2000;Watson and Chow, 2001;McDonald et al., 2003;USEPA, 2004).In most of the studies conducted on source apportionment using CMB, the source profiles were either compiled as a part of the study or generic source profiles from EPA's SPECIATE database were used (Park et al., 2001;Song et al., 2001;Ward and Smith, 2005).Since there were no local source profiles available for the Corpus Christi urban airshed, and creating one was beyond the scope of this study, a generic set of source profiles adaptable to the study region was developed using EPA's SPECIATE database.The set of source profiles identified included heavy duty vehicles, light duty vehicles, petroleum industry, residential wood combustion, field burning, secondary sulfates, secondary organic aerosols, paved road dust, crustal material, unpaved road dust, marine aerosols and industrial manufacturing.A detailed evaluation of EPA 's SPECIATE source profiles showed sulfates to be the major chemical component of marine vessel sources based on the weight percentages while in the case of heavy duty vehicles the key species included organic and elemental carbon with traces of Ni, Zn and Cu.In the current study no specific source profile was selected for the sea-faring vessels due to limited local emissions information on this source category.
CAMS 635: The model performance indices including the coefficient of determination (R 2 ) was found to range between 0.83 to 1, chi-square (χ 2 ) ranged from 0.00 to 5.37 and the percent mass explained ranged from 69% to 128% as shown in Table 3.The values were in agreement with those observed in other source apportionment studies (Park et al., 2001;Song et al., 2001).Annual average source contributions estimated during 2003-2005 are shown in Table 4, and Fig. 3 shows the percent contributions.Secondary sulfate was identified to be the major contributor accounting for 42% of the measured PM 2.5 concentrations as shown in Fig. 3.The other sources identified included secondary organic aerosols (19%), light duty vehicles (13%), heavy duty vehicles (8%), marine aerosols (4%), crustal matter (4%), paved road dust (2%), petroleum industry (2%), residential wood combustion and field burning (2%) and unpaved road dust (1%).Over the duration of the study period a gradual increase in the annual apportioned concentrations of secondary sulfate was observed suggesting the increased contribution of local anthropogenic sources.An increase in marine aerosol contributions was also observed which can be associated with increased influence of stronger winds from the southeast.In contrast to an increase in contribution of the above sources a considerable decrease in contributions of light duty and heavy duty vehicles was observed between 2003 and 2005.A decrease in the concentrations of secondary organic aerosols was observed and this was possibly due to a decrease in local anthropogenic sources.
Species compositions quantified by CMB 8.2 were used to reconstruct the total PM 2.5 mass and this was intercompared with the observed mass concentrations to evaluate the performance of the model.Regression analysis of the measured and predicted PM 2.5 concentrations was performed and is shown in Fig. 4. The coefficient of determination for CAMS 635 was observed to be 0.81, indicating good model performance.
CAMS 314: The model indices including the coefficient of determination at the coastal rural site was observed to range from 0.91 to 1, while chi-square (χ 2 ) was observed to range from 0.00 to 5.63 and the percent mass explained was observed to range from 70% to 124% as shown in Table 3.The major contributors of PM 2.5 measured at CAMS 314 during 2003-2005 were observed to be secondary sulfates accounting for 37% of the apportioned mass, followed by light duty vehicles contributing to 21% as shown in Fig. 4. The other sources apportioned included marine aerosols   (14%), secondary organic aerosols (13%), field burning and residential wood combustion (3%), crustal matter (3%), heavy duty vehicles (2%), paved road dust (2%), and unpaved road dust (1%).An increase in contributions of secondary sulfate was observed during the study period and is reflected in Table 5.This monitoring site has minimal impact of local anthropogenic sources and thus such annual trend could be attributed to the direct increase in contribution from long-range transport of photo chemically aged air mass under favorable wind conditions.Marginal increase in the source contributions were noted for heavy and light duty vehicles, while the marine aerosol source exhibited no trend at all.A general decrease in the OC concentrations was also observed.PM 2.5 concentrations measured at CAMS 635 located in the industrialized urban area was heavily influenced by emissions from petroleum refining industries as well as light duty and heavy duty vehicles from local traffic sources.Thus, the contribution of secondary sulfate (3.45 ± 0.31), secondary organic aerosols (1.62 ± 0.34) and heavy duty vehicles (0.62 ± 0.21) were higher at the urban site than at the coastal rural site (secondary sulfate: 3.06 ± 0.27, secondary organic aerosols: 1.07 ± 0.26, heavy duty vehicles: 0.18 ± 0.08).On the other hand, being located closer to the beach, the source contributions from light duty vehicles and marine aerosols were higher at the coastal site than at the urban site.

CONCLUSIONS
Source apportionment analysis was conducted employing CMB 8.2 on the PM 2.5 species concentrations measured at an industrialized urban site (CAMS 635) and a coastal rural site (CAMS 314) located within the Corpus Christi urban airshed of Texas.There were no prior studies conducted in the urban airshed focussing on the generation of source profiles specific to the airshed.Hence thirteen generic source profiles were selected from the U.S. EPA's SPECIATE database and applied to apportion twenty five key chemical species that accounted for 97.1% and 93.3% of measured PM 2.5 mass at CAMS 635 and CAMS 314, respectively.Overall, the model resolved 97% of the PM 2.5 mass at both the sites.Secondary sulfates were apportioned to be the major contributor at both the industrialized urban and coastal rural sites with average concentrations of 3.45 μg/m 3 and 3.06 μg/m 3 , respectively.Emissions from light duty and heavy duty vehicles with average concentrations of 1.74 μg/m 3 and 1.93 μg/m 3 were quantified to be the second largest source at CAMS 635 and CAMS 314, respectively.Higher concentrations of mobile sources at the rural site could be attributed to the near-by beach going traffic.Slightly higher levels of average secondary organic aerosol concentrations were apportioned at CAMS 635 (1.62 μg/m 3 ) as compared to CAMS 314 (1.07 μg/m 3 ) contributed by the local petrochemical industries and also due to longrange transport of photochemically aged air mass.Being in close proximity to the Gulf of Mexico higher levels of marine aerosols were quantified at CAMS 314 (1.15 μg/m 3 ) as compared to CAMS 635 (0.37 μg/m 3 ).The results showed slightly higher influence of field burning and residential wood combustion sources at CAMS 314 (0.27 μg/m 3 ) when compared to CAMS 635 (0.18 μg/m 3 ).CMB 8.2 apportioned average concentrations of 0.17 μg/m 3 and 0.02 μg/m 3 at CAMS 635 to the unique sources of petroleum industry and industrial manufacturing, respectively.The other source contributions apportioned by CMB 8.2 included earth's crust with higher levels at CAMS 635 (0.32 μg/m 3 ) than at CAMS 314 (0.22 μg/m 3 ), while the contributions from paved and unpaved road dust were similar at both sites.The annual average concentrations showed an increase in the secondary sulfates and secondary organic aerosol contributions during 2003 through 2005 at both the sites.Apportioned source categories of petroleum industry and industrial manufacturing also showed a gradual increase at the industrialized urban site during the study period indicating probable increase in local emissions affecting the measured PM 2.5 levels.

ACKNOWLEDGMENTS
The authors would like to acknowledge the Texas Commission on Environmental Quality (TCEQ) for providing the PM 2.5 speciation data.The authors would also like to show their appreciation to the anonymous reviewers for their valuable comments and feedback that were employed to improve this manuscript.

DISCLAIMER
Reference to any companies or specific commercial products does not constitute an endorsement by the authors of either the company or the product.

Fig. 1 .
Fig. 1.Monitoring site locations map in the Corpus Christi urban airshed.

Table 1 .
Statistical summary of PM 2.5 concentrations and the corresponding chemical species concentrations in μg/m 3 observed at CAMS635 and CAMS 314 (2003-2005).

Table 2 .
Pearson correlation coefficients between various chemical species in the observed PM 2.5 concentrations at (a) CAMS 635 and (b) CAMS 314.Bold underlined numbers indicate strong correlations.

Table 4 .
Source contribution of PM 2.5 at CAMS 635.

Table 5 .
Source contribution estimates of PM 2.5 at CAMS 314.
Note: value ± uncertainty is denoted here.