Source Apportionment of Air Pollution and Characteristics of Volatile Organic Compounds in a Municipal Wastewater Treatment Plant , North Taiwan

Monitoring air quality in the municipal wastewater treatment plants is an initial stage in preventing several problems related to air emissions. This study measured 103 volatile organic compounds (VOCs), total VOC (TVOC), and some prominent air pollutants (CO, CO2, NH3, H2S, PM1, PM2.5, PM7, PM10, TSP) in the municipal wastewater treatment plant X located in the north of Taiwan. Thirty-three VOCs were identified, which categorized as alkane, aromatic, alkene, ester, ether, haloalkane and ketone. Five dominant factors were determined from principal component analysis (PCA). The first factor involves indoor activities (such as particulate matter resuspension), and outdoor activities (such as vehicles exhausts), which explained 32.42% of total variance. Factor 2 was paint applications and domestic wastewater decomposition, with an explained variance of 17.23%. Factor 3 was solvent use, with an explained variance of 14.26%. Factor 4 was solvent use and road dust, with an explained variance of 8.54%. Factor 5 was a byproduct of the chlorination treatment process, with an explained variance of 6.92%. The five factors explained 79.37% of total variance. By applying absolute principal component scores (APCS), source apportionments were obtained having 35.21%, 26.04%, 16.13%, 7.03%, and 15.59% for factors 1, 2, 3, 4, and 5, respectively. Toluene had the highest ozone formation potential (OFP).


INTRODUCTION
Municipal wastewater treatment is the process of removing harmful pollutants from wastewater, and the main source of pollutants is domestic water use.Proper treatment of wastewater ensures that acceptable overall water quality is maintained.Although the main purpose of wastewater treatment plants is to remove harmful pollutants in the wastewater to protect humans and the environment health, there are several problems related to air emissions.
Emissions from municipal sewers are usually omitted from both volatile organic compound (VOC) and hazardous air pollutant (HAP) emission inventories (Jones et al., 1995).Industries, commercial facilities, public institutions, and residential households are the sources of VOCs in municipal wastewater collection.Exposure of workers to toxic chemicals during wastewater collection and treatment, emissions of toxic air contaminants, accumulation of explosive gases in sewers, and the release of photochemical precursors to the atmosphere are several problems related to the transfer of VOCs from aqueous and gaseous phases (Quigley and Corsi, 1995).
Taipei City, which is located in the Taipei Basin in Northern Taiwan, covers an area of 271.7997 km 2 with a population of 2.705 million in 2015.Taipei has two municipal wastewater treatment plants and one of them is plant X.Municipal wastewater treatment plant X is a complex on land reclaimed by straightening the Keelung River, has a sewage-collection scope with an area of 1,176 hectares and an average treatment capacity of 240,000 m 3 per day.Municipal wastewater treatment plant X was constructed semi-underground, and a recreational sports park was built on top of the wastewater treatment plant for the benefit of the local community (Taipei City Government, 2017).
This study measured 103 VOCs in the gaseous phase in the municipal wastewater treatment plant X in the winter, summer, autumn, and spring seasons from February until November 2016.VOCs are organic chemicals mainly composed of carbon atoms that are readily emitted to the atmosphere as gases due to the high vapor pressures resulting from their low boiling points (Chen et al., 2013).VOCs are also crucial precursors to secondary organic aerosols (SOAs) (Zheng et al., 2017).An individual VOC can react at different rates with different reaction mechanisms and has different O 3 formation potentials; therefore, the control of VOC emission sources needs to consider their emission amounts and chemical reactivities (e.g., maximum incremental reactivity) (Ou et al., 2015).
The present study also identified and measured prominent criteria air pollutants (CO, CO 2 , NH 3 , H 2 S, PM 1 , PM 2.5 , PM 7 , PM 10 , TSP) and the total volatile organic compounds (TVOCs).Surfer 10 program was applied to plot the TVOC seasonal concentration distributions.Particulate matter is one of the significant pollutants with adverse health impacts such as premature mortality and, lung and cardiovascular diseases (Rasheed et al., 2015).
The seasonal CO, CO 2 , NH 3 , H 2 S, PM 1 , PM 2.5 , PM 7 , PM 10 , TSP, TVOC, and VOC were measured and their sources were analyzed using principal component analysis (PCA) and absolute principal component scores (APCS).PCA is a technique of multivariate analysis that has been widely used for the identification of the potential sources of pollutants in the environment (Mari et al., 2016).The advantage of using PCA is that it has the ability to identify different sources without any prior knowledge regarding them (Madureira et al., 2016).The APCS method was used to quantify the contributions of all sources to each measured pollutant.One of the advantages of the APCS method is that it relies on multiple key tracers, yielding more interpretable, independent (uncorrelated and not collinear), and unique source categories (Thurston et al., 2011).The results of this study could provide information on the levels and sources of air pollutants in municipal wastewater treatment plants, which is especially important for improving the air quality by using such data as a reference for the selection of air pollution control technologies.Moreover, the results of this study can be useful in serving as a database for assessing health risks in the future.

Sample Collection
Sample collection was done not only in the municipal wastewater treatment plant (indoor) but also in the recreational sports parks (outdoor) that was built on the municipal wastewater treatment plant X.The sports park was divided into 11 sampling points (Fig. 1).The municipal wastewater treatment plant was divided into three areas, and samples were taken at several points from each of these, with 19 sampling points in total (Fig. 2).While VOCs samples were taken at one point for each area.Studies carried out by the United States Environmental Protection Agency (US EPA) of human exposure to air pollutants indicate that indoor air levels of many pollutants may be 2-5 times higher than outdoor levels, and occasionally more than 100 times (Alves et al., 2013).
The sampling unit was set at 1.2 m from the ground, and sampling was done in February 2016, May 2016, August 2016, and November 2016, representing the winter, spring, summer, and autumn seasons.Samples were taken once every season during the day at 08:00-10:00.In the area with the highest TVOC concentration was then measured the concentrations of various air pollutants (CO, CO 2 , NH 3 , H 2 S, PM 1 , PM 2.5 , PM 7 , PM 10 , TSP), the total volatile organic  compound (TVOC) and 103 VOCs for seven days in every season (a longer investigation).In this, samples were taken every day at 08:00-10:00, 16:00-18:00 and 22:00-24:00 for seven days.While VOCs samples were taken once at 22:00-24:00 for seven days.

Chemical Analysis and Data Treatment
Passive flow control canisters (volume 6 L; flow rate was fixed at 40 mL min -1 ) were used to collect VOC samples.All the canisters were cleaned, and then vacuumed using humid N 2 pure gas (99.999%) to ensure their vacuum quality before sampling.To quantify the target compounds, Photochemical Assessment Monitoring Station (PAMS) certified gas was diluted into different concentrations using a dilution system (Wang et al., 2016).During sampling, preservation, transportation, and analysis, the United States Environmental Protection Agency (US EPA) Method TO-15 was adopted.Air samples were then analyzed using a gas chromatograph (GC, Agilent 6890N) and a mass spectrometer (MS, Agilent 5973MSD).The GC oven temperature was set at 32°C, increased to 200°C, and then held for 3 min.The total time required was 31.5 min.Several standard gases were used to calibrate VOCs (Ou-Yang et al., 2017).For each species, calibration was conducted and a good linear fit was observed with R 2 > 0.99.The cryogenic preconcentrator was baked after each analysis, and the GC column was also baked after analysis of every 20 samples (Wei et al., 2014).The method of detection limit (MDL) for each species was determined according to US-EPA Test Methods SW-856 (Table 1).
The monitoring of CO and CO 2 was done using the Q-TRAK™ Indoor Air Quality Monitor 7575 (TSI, Shoreview, United States) instrument, which is able to detect CO and CO 2 concentrations in the range of 0-500 ppm and 0-5000 ppm, respectively.Calibration was done in the field.Upon calibration, the measurement of various parameters can be adjusted.The appropriate detachable probes must be attached to the instrument before field calibration, except for pressure and barometric pressure calibration.
Particulate matter (PM) was measured using the Met One Aerocet 531 particle profiler (Met One Instruments, Inc. Grants Pass, Oregon).Particulate fractions of PM 1 , PM 2.5 , PM 7 , PM 10 and TSP (total suspended particles) were in mg m -3 and within a concentration range of 0-1 mg m -3 .The Aerocet 531 was calibrated using NIST (National Institute of Standards and Technology) traceable polystyrene spheres.
Real-time TVOC measurements were done using a portable PpbRAE 3000, photo-ionization detector (PID) having a 10.6 eV photoionization lamp detector (RAE System Inc., San Jose, CA).The instrument is sensitive to 200 VOC compounds and capable of estimating TVOC concentrations in the range of 1-10,000 ppb (resolution of 1 ppb), with a response time of 3 s and a sampling flow rate of 500 cc min -1 .The TVOC monitor was calibrated using 100 ppm isobutylene and zero air following the manufacturer's recommendations (Singh et al., 2016).
NH 3 and H 2 S were measured using a MultiRAE Lite PGM-620X system (RAE Systems, San Jose, USA), within the concentration range of 0-100 ppm.Concentration maps derived from real ambient TVOC concentrations were obtained using Surfer 10, which was developed by Golden Software Inc., USA.The instrument was calibrated using zero air following the manufacturer's recomendations.(3) Not classifiable with respect to its carcinogenicity to humans.

Estimation of Ozone Formation Potential
Ozone formation potential is used in the assessment of the roles of different VOC species in O 3 formation (Ou et al., 2015).VOC concentration influences the O 3 chemical production, and different VOC species make different contributions to O 3 formation (Geng et al., 2009).The OFP of an individual VOC species was calculated as follow (Liu et al., 2008): where OPF i is the total ozone formation potential of species i, VOCs i is the concentration of a species i, and MIR i is the maximum increment reactivity of species i. Updated MIR values from Carter (2010) were adopted in this study.

Principal Component Analysis (PCA)/Absolute Principal Component Scores (APCS)
Principal component analysis (PCA) was performed using the SPSS statistical packages (SPSS Inc, USA).PCA is often used in data dimension reduction to identify a small number of factors that explains most of the variance observed in a much larger number of manifest variables (Guo et al., 2004).In this study, VOC concentrations below the detection limits were replaced with 1/2 MDL, and extract factors with eigen-values greater than one were chosen (Lan et al., 2014).We decided that VOCs with values more than twenty percent below the detection limits were excluded from the analysis.The Kaiser Meyer Olkin and the Bartlett's test values also were used for analysis.Variables with factor loadings greater than 0.7 are considered relevant, which indicate a possible emission source.The correlation between the concentration of a particular pollutant and a component increases with its loading (Cheng et al., 2016).
PCA was operated on z-score transformed data, which converts the experimental data into zero (mean) and unity (variance) to neutralize the effects of multidimensionality and the different units of the parameters (Pandey et al., 2015).
The variance of individual factors in PCA indicate the relative magnitudes among the dominant potential sources (Huang et al., 2012).After determining the number and variety of potential sources, APCS was applied to estimate the contribution of each pollution source to each pollutant.

Air Pollutant Concentration
Table 2 shows the CO, CO 2 , PM, TSP, H 2 S, NH 3 , and TVOC concentrations for the longer period of investigation.
The level of CO for four seasons in the present study was below the standard values prescribed by WHO (32 ppm for 1 h).It is noted that the level of CO 2 observed in the wastewater treatment plant was below that by the NIOSH REL (5000 ppm).The overall concentration of PM 10 in the wastewater treatment plant exceeded the standard values suggested by the WHO for the annual mean (20 µg m -3 ).The overall concentration of PM 2.5 in the present study was below the standard value prescribed by the WHO for the annual mean (10 µg m -3 ).The observed H 2 S and NH 3 concentrations in the wastewater treatment plant were also below the standard values prescribed by NIOSH REL (25 ppm and 10 ppm for H 2 S and NH 3 , respectively).
Several studies have assessed the air pollutants associated with wastewater treatment.For example, Hamoda (2007) identified the presence of VOCs and other gaseous pollutants such as methane, ammonia, and hydrogen sulphide in air surrounding a municipal wastewater treatment plant, a petroleum refinery wastewater treatment plant, and a landfill site of solid wastes in the State of Kuwait.In some cases the levels exceeded the concentration limits specified by the air quality standards.However, the concentrations of CO, CO 2 , PM 2.5 , H 2 S, NH 3 in this study were lower than related standards.Lee et al. (2007) assessed the air quality of four wastewater treatment plants in Iowa, USA by monitoring the levels of hydrogen sulfide (H 2 S) and endotoxins.The results showed that the geometric means of the H 2 S concentration were less than 1 ppm, and the endotoxin concentrations ranged from 6-1247 EU m -3 .The H 2 S concentration in the current study is higher than in this earlier work.
According to an article published by Washington State Department of Ecology in 1998, acetone, toluene, methylene chloride, and chloroform were the most commonly reported VOCs in the wastewater influent and air emissions from wastewater treatment plants.The current study also identified the same species of VOCs having toluene, ethanol, acetone and isobutane as the most abundant species.

Level of TVOC
The concentration distributions of TVOC for every season in the sports park (outdoor) and municipal wastewater treatment plant (indoor) are shown in Figs. 3 and 4  respectively.From Fig. 3, it can be seen that higher TVOC concentrations, except for winter, were distributed in the eastern sampling sites, which were near roads and commercial areas.While for the municipal wastewater treatment plant, it can be seen that higher TVOC concentrations, except for autumn, were distributed in the primary clarifier area (Fig. 4).Figs. 3 and 4 show that the TVOC concentrations in the wastewater treatment plant (indoor) are higher than those of the sports park (outdoor).This is because indoor levels are influenced by significant indoor sources or penetration from outdoor sources (Singh et al., 2016).In Table S1, the overall concentration of some parameters (PM 2.5 , H 2 S, NH 3 and TVOC) were higher in the primary clarifier.For that reason, a longer duration of measurement of the air pollutant concentrations (CO, CO 2 , NH 3 , H 2 S, PM 1 , PM 2.5 , PM 7 , PM 10 , TSP), total volatile organic compound (TVOC) and 103 VOCs in the primary clarifier were carried out for seven days in every season.

VOCs Characteristics during the Sampling Period
Table S2 shows the mean concentration for one day investigation of VOCs in the sports park and each area of the municipal wastewater treatment plant for every season.Toluene, ethanol, and acetone were the most abundant species.Table 3 shows the mean concentrations in the primary clarifier with a longer period of investigation in every season.Toluene and ethanol were the most abundant species, followed by acetone and isobutane.Fig. 5 shows the boxplots of VOC concentrations by category.Based on the 103 VOCs observed in this work, 33 VOCs were identified and categorized as alkane, aromatic, alkene, ester, ether, haloalkane, and ketone.The most abundant VOCs were aromatic, followed by ester and ketone.

Analysis of Ozone Formation Potential
Ozone is a secondary pollutant that is generated in photochemical reactions which are often more toxic than   (Yang et al., 2016).Table 4 lists speciated VOCs and their OFPs.It was observed that the highest OFP values for winter, spring, summer, and autumn were for toluene.This shows that toluene had the highest mean concentration compared to the other VOC species.Toluene is one of the most significant precursors to ozone formation, which greatly affects the air quality and human health (Chen et al., 2017a).The second highest OFP was ethanol for all seasons followed by m-Xylene for winter and spring and, 1,2,4-Trimethylbenzene for summer and autumn (Table 4).Although m-Xylene and 1,2,4-Trimethylbenzene had mean concentration of less than 10 ppbv, they had a higher MIR value than other compound.

Source Apportionment Using PCA/APCS
Table 5 shows five factors with the application of PCA.According to Kaiser's criteria, only the factor with eigenvalues ˃ 1 were included.The Kaiser-Meyer-Olkin value ˃ 0.5 and the Bartlett's test reached an extremely significant level (0.000) in this work, showing that the factor analysis was adaptive (data not shown).Factor 1 explained 32.42% of total variance, while the five factors explained 79.37% of the total variance.
The source profiles were collected from literatures to establish the fingerprints of VOCs and other pollutants.The characteristics of the five sources are shown in Fig. 6.Indoor activities (such as particulate matter resuspension) and transport from outdoor activities (such as vehicle exhausts) (35.21%) were the first factor, correlated with PM 2.5 , PM 7 , PM 10 , TSP and CO.Particulate matter, and especially PM 10 , levels might be explained by the mixed sources from indoor and outdoor activities (Madureira et al., 2016).Vehicle exhausts are one of the PM 2.5 sources (Chen et al., 2017b).
For the second factor (26.04%), high factor loadings were found for toluene, CO 2 , H 2 S, and NH 3 .Paint applications were correlated with toluene (Huang et al., 2012).CO 2 levels were related to the natural ventilation in the building (Madureira et al., 2016) while H 2 S and NH 3 were related to domestic wastewater decomposition (Jeon et al., 2009).
Solvent use was the third (16.13%) and fourth factor (7.03%) correlated with ethanol and acetone, respectively.Non-chlorinated solvents (ethanol and acetone) are present   in a large range of products, such as household cleaners, car shampoos and degreasing agents from vehicle maintenance and production (Gendebien, 2010).The fourth factor also correlated with PM 1 , and one of its sources was road dust (Titos et al., 2014).The chlorination treatment process was the fifth factor (15.59%) correlated with methylene chloride, which is formed as a volatile organic byproduct of the chlorination treatment process (Oullette, 1994).

CONCLUSION
PCA/APCS were applied to identify different air pollution sources and calculate the contribution of each in the municipal wastewater treatment plant X in the north of Taiwan.Air pollutants including CO, CO 2 , PM 1 , PM 2.5 , PM 7 , PM 10 , TSP, H 2 S, NH 3 , TVOC and VOCs were investigated.A total of 33 compounds were identified from 103 VOCs and were categorized as alkane, aromatic, alkene, ester, ether, haloalkanes and ketone.Five factors were determined using PCA, and this explained 79.37 % of the total variance.The calculation of each pollution source according to APCS indicates that the contribution rates of these five factors were 35.21%, 26.04%, 16.13%, 7.03% and 15.59%, respectively.The MIR method was applied for the analysis of OFP, with toluene having the highest OFP value.

Fig. 1 .
Fig. 1.Location of sampling in the sport park X (outdoor).

Fig. 2 .
Fig. 2. Location of sampling in the municipal wastewater treatment plant X (indoor).
Fig. 4. TVOC concentration (ppbv) contour maps of the municipal wastewater treatment plant X (indoor) in four seasons.