Science of the Total Environment 482–483 (2014) 336–348

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Long term assessment of air quality from a background station on the Malaysian Peninsula Mohd Talib Latif a,b,⁎, Doreena Dominick b, Fatimah Ahamad a,b, Md Firoz Khan a,b, Liew Juneng a,b, Firdaus Mohamad Hamzah c, Mohd Shahrul Mohd Nadzir a,b a b c

School of Environmental and Natural Resource Sciences, Faculty of Science and Technology, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, Malaysia Centre for Tropical Climate Change System (IKLIM), Institute for Climate Change, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, Malaysia Department of Civil Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, Malaysia

H I G H L I G H T S • • • •

We analysed air quality data recorded at background station on Malaysian Peninsula. Principal component regression and sensitivity analysis have been employed. Wind direction influences the transport of air pollutants to the background station. Diurnal variations of major air pollutants contribute by motor vehicle emissions.

a r t i c l e

i n f o

Article history: Received 24 November 2013 Received in revised form 6 February 2014 Accepted 27 February 2014 Available online 21 March 2014 Keywords: Background stations Major air pollutants Meteorological factors Monsoon

a b s t r a c t Rural background stations provide insight into seasonal variations in pollutant concentrations and allow for comparisons to be made with stations closer to anthropogenic emissions. In Malaysia, the designated background station is located in Jerantut, Pahang. A fifteen-year data set focusing on ten major air pollutants and four meteorological variables from this station were analysed. Diurnal, monthly and yearly pollutant concentrations were derived from hourly continuous monitoring data. Statistical methods employed included principal component regression (PCR) and sensitivity analysis. Although only one of the yearly concentrations of the pollutants studied exceeded national and World Health Organisation (WHO) guideline standards, namely PM10, seven of the pollutants (NO, NO2, NOx, O3, PM10, THC and CH4) showed a positive upward trend over the 15-year period. High concentrations of PM10 were recorded during severe haze episodes in this region. Whilst, monthly concentrations of most air pollutants, such as: PM10, O3, NOx, NO2, CO and NmHC were recorded at higher concentrations between June and September, during the southwest monsoon. Such results correspond with the mid-range transport of pollutants from more urbanised and industrial areas. Diurnal patterns, rationed between major air pollutants and sensitivity analysis, indicate the influence of local traffic emissions on air quality at the Jerantut background station. Although the pollutant concentrations have not shown a rapid increase, an alternative background station will need to be assigned within the next decade if development projects in the surrounding area are not halted. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Background stations can provide invaluable information on pollutant exposure to humans and vegetation at distances from a few km2 to a few thousand km2. A rural background station is also useful in providing air quality information on a regional scale (USEPA, 1998; EU, 2005). Such a station must be located in an area with a natural ⁎ Corresponding author at: School of Environmental and Natural Resource Sciences, Faculty of Science and Technology, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, Malaysia. Tel.: +60 3 89213822; fax: +60 3 89253357. E-mail address: [email protected] (M.T. Latif).

http://dx.doi.org/10.1016/j.scitotenv.2014.02.132 0048-9697/© 2014 Elsevier B.V. All rights reserved.

ecosystem, low population density and be a good distance from anthropogenic emission sources (EU, 2005). Hence, continuous air quality monitoring data collected from a rural background station allows the observation of regional trends in air pollutant concentrations with minimal enhancement resulting from local emissions. The expansion of greater urban areas, however, has led to the movement of pollutants from city centres towards suburban areas, which not only affects the level of air pollutant concentrated in the city centres but also that of background areas (Agrawal et al., 2003; Grawe et al., 2013). For example, a study by Donnelly et al. (2011) showed that even cities located more than 50 km away from a background station can influence the NO2 concentrations at such stations. The concentration of pollutants

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348

(for example SO2) from industrial sources also revealed changes occurring at a rural background station when emission control measures were carried out in industrial and densely populated areas (Lin et al., 2012). In Malaysia, the central focus of development programmes since its independence in 1957 has been the Klang Valley area where Kuala Lumpur, the capital city, is located. In other parts of the Malaysian Peninsula, the capital of each state is usually the central pollutant emission source, aside from industrial areas located outside the city centre. The movement of air pollutants from the city and industrial areas to rural areas is predominantly dependent on seasonal winds and local conditions. These include land and sea breezes, the valley effect and also the transference of wind from urban to rural areas which results from the urban heat island effect (Sani, 1990; Juneng et al., 2009, 2011). Such developments also result from increases in deforestation and biomass burning through agricultural activities, as soil dust derived from deforestation and smoke from biomass burning contribute to the high amount of particulate matter existing, particularly in suburban and rural areas (Dominick et al., 2012). The Department of Environment, Malaysia (DOE) had designated Jerantut, Pahang as its site for a rural background station so as to monitor the general background concentrations of selected air pollutants in the Malaysian Peninsula. This continuous monitoring station is located in a rural area close to a national forest reserve. Many studies on air pollutant concentrations in Malaysia have included Jerantut for comparison with observations from other rural, urban and suburban monitoring stations (Azmi et al., 2010; Latif et al., 2012; Banan et al., 2013). However, there are as yet, no studies which focus on the long-term pollutant concentration trends at this station. Although the site was originally chosen because of its distance from local anthropogenic emission sources, there are some concerns regarding its continuing viability as a rural background station in the coming years as a result of increasing development of the surrounding areas. This study aims to fill this gap by analysing continuous monitoring data collected over a period of 15 years and to assess the trends in pollutant concentrations, primarily in terms of local influences on the background pollutant concentrations in the Malaysian Peninsula. Assessments were also undertaken in order to determine the feasibility of this site being used for background data collection in the future, given that rapid industrialisation and urbanisation over the past

337

decade may have introduced significant local or mid-range emission sources to the rural location where the background station is located. 2. Materials and methods 2.1. Location of sampling station The air quality data used in this research was collected from Batu Embun, Jerantut, Pahang Station, which has been established as a background air monitoring station by the Department of the Environment (DOE), Malaysia. This station is located near the middle of the Malaysian Peninsula with coordinates N03° 58.238′, E102° 20.863′ (Fig. 1) and is surrounded by natural forest and agricultural areas, as well as traditional Malaysian villages. It lies within a 2 km radius of Sungai Pahang, one of the longest rivers in Malaysia, and its confluence, Sungai Teh. The nearest town is Jerantut, which is about 7 km from Batu Embun. Some short-term studies have shown that the air quality level at this monitoring station is influenced by local open burning, soil dust and a low number of motor vehicles (Azmi et al., 2010; Banan et al., 2013). This may be due to the presence of a two-lane paved road close to the sampling site (b1 km) and villages, as well as modern housing areas within a 10 km radius south of the station. 2.2. Data collection The air quality data for the analysis period, January 1997 to December 2011, was obtained from the Air Quality Division of the DOE; the Ministry of Natural Resources and Environment of Malaysia. A total of 1,368,715 hourly observations from the 15-year (1997–2011) dataset, which consisted of 14 variables, were initially arranged based on month and year. The 14 variables studied were divided into three groups: major air pollutants, organic pollutants and meteorological parameters. The major air pollutant group consisted of ground level ozone (O3), carbon monoxide (CO), nitrogen oxide (NO), nitrogen dioxide (NO2), oxides of nitrogen (NOx), sulphur dioxide (SO2) and particulate matter with a diameter size of less than 10 µm (PM10). The organic pollutant group consisted of methane (CH4), total hydrocarbon (THC) and non-methane hydrocarbon (NmHC) while the meteorological parameter group consisted of wind speed, ambient temperature, relative

Fig. 1. Location of the Jerantut air monitoring station.

338

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348

humidity and ultraviolet-B radiation (UVb). Due to data limitations, the datasets for CH4, NmHC, THC, UVb and relative humidity were collected from 1998 to 2009, 1997 to 2009, 1997 to 2010, 1998 to 2009 and 2003 to 2011, respectively. No imputation procedure was adopted for the treatment of missing data. The instrument used to monitor PM10 was a BAM-1020 Beta Attenuation Mass Monitor (Met One Instrument, Inc., USA) which is equipped with a cyclone and PM10 head particle traps, fibre glass tape, flow control and a data logger. This instrument has a fairly high resolution of 0.1 μg m−3 at a 16.7 L min−1 flow rate, with lower detection limits of b4.8 μg m−3 and b1.0 μg m−3 for 1 h and 24 h, respectively. The instruments used to monitor SO2, NOx, CO and O3 were the Teledyne API Model 100A/100E, Teledyne API Model 200A/200E, Teledyne API Model 300/300E and Teledyne API Model 400/ 400E respectively (Teledyne Technologies Inc., USA). SO2 determination was based on the UV fluorescence method where the lowest level of detection was at 0.4 ppb. NOx determination used the chemiluminescence detection method with the same detection limit as the SO2 analyser. CO was determined using the non-dispersive, infrared absorption (Beer Lambert) method with 0.5% precision and the lowest detection of 0.04 ppm. Ozone was determined through the UV absorption (Beer Lambert) method with a detection limit of 0.4 ppb. The measurements of SO2, NOx, CO and O3 were at a precision level of 0.5%. CH4 and NmHC were measured using a Teledyne API M4020 (Teledyne Technologies Inc., USA) equipped with a flame-ionisation detector (FID) with a measurement accuracy of 1%. To record the meteorological variables, a Met One 010C sensor was used to measure wind speed and a Met One 020C sensor to measure wind direction. Ambient temperature and relative humidity were measured using a Met One 062 and Met One 083D sensor respectively (Met One Instrument, Inc., USA). 3. Statistical analyses 3.1. Cluster, sensitivity analysis and principle component regression In this study, all 14 variables were statistically analysed using XLSTAT 2013 software. Hierarchical agglomerative cluster analysis (HACA) was applied to the data sets in order to study temporal patterns. This is the most common technique used to classify observed data into clusters with high homogeneity (similar) levels within groups and also heterogeneity (different) levels between groups (McKenna, 2003). The number of clusters produced was visualised in the form of chart called a ‘dendrogram’ (tree diagram). The diurnal data used in this study was classified by its unique characteristics, and thus, helped in interpreting the data more efficiently. HACA was performed using Ward's method and the Euclidean distance was used to measure any similarities (Adams, 1998; Lau et al., 2009). The Euclidean distance was calculated based on the single linkage method, based on the minimum distance or the nearest neighbour rule. The distance between two clusters was computed as the distance between the two closest elements in the two clusters (Ibarra-Berastegi et al., 2009). The Euclidean distance being the ratio of linkage distance over maximal distance: [(D link / D max) × 100]. In order to standardise the linkage distance represented by the y-axis, the quotient is generally multiplied by 100 (Singh et al., 2004, 2005; Shrestha and Kazama, 2007; Lau et al., 2009). Sensitivity analysis is a very useful method for ranking the importance of input variables by assessing their contribution (percentage) to the variability of the model output (Manache and Melching, 2008). A study by McRae and Seinfeld (1983), emphasised how effectively sensitivity analysis indicates the degree in which overall uncertainty in the model predictions can be associated with the individual uncertainty of each model input. The knowledge of the model's sensitivity to different variables is necessary, however, in deciding where emphasis should be placed when estimating total uncertainty (Hanna, 1988). In this study, the one-factor-at-a time (OAT) regression analysis technique was

undertaken in order to study the level of importance of each input (CO, NO, NO2, NOx, O3, PM10, SO2, THC, CH4 and NmHC) on the output (vehicles). Morris (1991) introduced the OAT approach that analyses the sensitivity of an individual parameter over its entire range by repetitive application. Only one parameter can vary at a time, ignoring the effects of parameter interactions and multi-response interdependences (Liu et al., 2004). The standardised regression determination coefficient (R2) values were used to estimate this sensitivity. Principal Component Regression (PCR) analysis, which combines PCA (Principle Component Analysis) and OLS (Ordinary Least Squares regression), is one of the best ways to study statistical relationships between pollutants and meteorological factors and can reduce the multicollinearity–collinearity in datasets (Sousa et al., 2007). The presence of multicollinearity among the independent variables will produce invalid results in terms of the model's predictions and determination of the significant independent variables (i.e. the distribution of each independent (X) variable on the dependent (Y) variables). PCA is used to extract significant factors (principal components; PCs) so as to analyse the relationships between the observed variables (Liu et al., 2009). Previous studies, for example, Kim and Mueller (1987); Kumar et al. (2001); Dominick et al. (2012); Han et al. (2013) chose factors with significant eigenvalues (N1.0) in order to gain a better understanding of the correlation relationship between the variables. The significant factors (independent variables) obtained were regressed against dependent variables using OLS analysis to estimate the relationship. This important feature, in effect, reduces the complexity of the analysis since a lower number of input variables is required compared to multiple linear regression. The ratio values of pollutant concentrations behave similarly to correlation coefficients where they are less dependent on temporal variations in sources and meteorological conditions. Therefore, by taking this important and useful relationship into consideration, the ratio values between selected pollutant concentrations, such as SO2, CO and NOx were used in this study as a guideline to examine the pollutants' sources. Previous studies undertaken by Aneja et al. (2001) and Goyal and Sidhartha (2003) emphasised that mobile emission sources (e.g. cars, buses, motorcycles as well as light and heavy duty trucks) are characterised by high concentrations of CO and NOx. Point source emissions are characterised by high SO2 and NOx concentrations. Hence, high CO/NOx and low SO2/NOx ratios values indicate mobile sources while high SO2/NOx and low CO/NOx ratio values typically indicate point sources, e.g. from industrial activities. Furthermore, the variation in NO/NO2 is useful in examining the criterion for background stations. This is because, in areas with high traffic density, NO concentrations will be greater when compared to NO2. During diffusion of NO from traffic hot spots to the surrounding area, NO is converted to NO2. Therefore, the general trend for the NO/NO2 ratio is urban traffic N urban background N suburban background N rural background area. This useful ratio knowledge has been applied in several studies with similar focuses (e.g. Coppalle et al., 2001; Chaloulakou et al., 2008; Mavroidis and Ilia, 2012; Melkonyan and Kuttler, 2012). 3.2. HYSPLIT and wind rose Back trajectory analysis, performed using the Hybrid-Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Draxler and Rolph, 2003), was undertaken so as to determine the origin of air mass arriving at Jerantut. The Global Data Assimilation System (GDAS) from NOAA's Air Resources Laboratory was used as input data for the model. This model was used to calculate the 3-day back trajectories for each month between 2009 and 2011. The trajectories were driven using gridded meteorological data at six-hour time intervals with calculations set for 500 m above ground level. The results are presented as the mean trajectories of the clusters obtained for each month. A wind rose predominantly represents the circulation and distribution of local wind patterns. Detailed and comprehensive monthly

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348

339

4. Results and discussion

happens because of the regional transboundary transportation of pollutants especially from Southeast Asia. During the dry weather, some places in Southeast Asia experience a high number of burning events which enhances the regional transboundary transportation. The hourly average concentrations of gases, such as: CO, NO2, SO2 and O3, were found to be far lower than those recommended by RMAQG guidelines (Table 1). The range of CO hourly average concentrations, for example, was recorded as being between 285.0 ppb and 387.6 ppb compared to the RMAQG suggested concentration of 30000 ppb. The maximum concentration for CO, NO, NO2, SO2 and O3 were found to be 2660 ppb, 30 ppb, 21 ppb, 13 ppb and 75 ppb, respectively. The concentrations of CH4, NmHC and THC were recorded at higher levels compared to the major air pollutants, namely: O3, NO, NO2, NOx, SO2 and PM10, with the exception of CO. This was expected due to the dominance of biogenic sources contributing to the quantity of organic gases present, particularly in rural areas, such as Jerantut. CH4 can be emitted through anaerobic decomposition (Haszpra et al., 2008) as well as biodegradation processes, especially from peat swamp areas namely those along the Pahang River, which is very near to the sampling station. According to Sauvage et al. (2009), biogenic sources are the one of the main contributors of NmHC in the environment, in addition to vehicular and industrial emissions. Isoprene, an NmHC sensitive to UV radiation and temperature, is the major compound found in biogenic sources (Tang et al., 2007).

4.1. Descriptive statistics

4.2. Diurnal patterns

In order to extract the important basic statistical descriptions, univariate statistical analysis was applied to all of the hourly averages of the yearly concentrations, apart from PM10. The PM10 database is based on daily average (24 h) concentrations. In this study, the basic statistical descriptions used are: minimum, maximum, mean, 1st Quartile, 3rd Quartile, median and standard deviation (Table 1). These basic statistical values are important when explaining the distribution of the variables and also useful in determining the pollution status of the atmosphere. Pollutant concentrations were compared with the maximum allowable values detailed by the Recommended Malaysian Air Quality Guideline (RMAQG) (Table 1). The PM10 concentration range was between 9.4 μg m−3 and 276.5 μg −3 m (Table 1). The results show that the 24 h average concentration of PM10 (37.9 μgm−3) is below the RMAQG and WHO limits, which are 150 and 50 μg m− 3 respectively. The maximum PM10 concentration was recorded on 20th September 1997 in the year of the worst and the strongest El Niño of the 20th century (Khandekar et al., 2000). The southwest monsoon winds and the occurrence of the biomass burning influences the air quality in Malaysia (Radzi bin Abas et al., 2004) including Jerantut, the Malaysian air quality background station. This

The diurnal patterns for all the pollutants, along with meteorological parameters, are presented in Fig. 2. NOx, NO2 and NmHC show a distinct bimodal pattern while CO and NO show a much smaller secondary peak. The first peak occurs between 8.00 and 9.00 h while the second one occurs at around 19.00–21.00 h. These two peaks can be largely attributed to rush hour traffic in the morning and evening. NO2 was found to peak about one hour after NO. This pattern is consistent with the formation reaction of NO2 from NO. In addition, the diurnal trends of some gases may also be explained by dilution with free tropospheric air during the daytime boundary-layer growth. NmHC concentrations were recorded at the highest concentration around 8.00–9.00 h, they then declined towards midday and increased again between 18.00 and 20.00 h (Fig. 2). The concentration of CH4 dropped much lower than its night time concentration particularly between 12.00 and 17.00 h. This is partly due to the vertical dispersion of CH4 during the day as the boundary layer develops further once surface temperature begins to increase after 9.00 h. The small molecular mass of methane would facilitate its movement upwards, as the warmer near surface air begins to rise due to buoyancy. Temperature and UVb are also expected to influence the concentrations of CH4 and NmHC

wind rose plots were prepared using Igor Pro 6.0.5. Ground level wind speed (km/h) and direction (°) were measured hourly over a period of 3 years from 2009 to 2011 and the results obtained from the DOE were used as the input data. 3.3. Quality control The instruments used to measure air pollutants and meteorological parameters were calibrated by Alam Sekitar Sdn. Bhd. with a specific time schedule based on the type of measurements used. The procedures followed include at least the minimum standards as outlined by internationally recognised environmental organisations such as the United State Environmental Protection Agency (USEPA). Single calibration was conducted during regularly scheduled visits to perform one-point manual calibration through the sample part of NOx. Multi-points calibration was conducted in order to perform manual calibration through the sample point on all gas analysers. All gas analysers were autocalibrated daily using a specific gas calibrator. Daily zero and span was performed for each analyser (SO2, NOx, CO and O3). Data from the monitoring stations were checked via a “polling computer” located at ASMA's Headquarters (HQ) (ASMA, 2007).

Table 1 Overall descriptive analysis of the variables within the 15 years (1997 - 2011). Variables

Min

Max

Q1

Median

Q3

Mean

SD

Time

RMAQG

PM10 (μgm−3) CO (ppb) O3 (ppb) NO2 (ppb) SO2 (ppb) NOx (ppb) NO (ppb) CH4 (ppb) NmHC (ppb) THC (ppb) UVb (Jm−2 h−1) Temperature (°C) Humidity (%) Wind Speed (km h−1)

9.4 285.0 3.5 1.9 1.8 2.1 1.3 1704.0 83.7 1788 3.3 23.0 59.4 2.0

276.5 387.6 26.4 3.4 2.1 6.1 4.6 1797.7 115.1 1896.4 790.4 32.1 96.7 6.2

27.6 302.6 5.5 2.1 1.8 2.8 1.4 1726.2 87.0 1812.5 3.4 23.6 70.3 2.2

35.1 318.9 8.7 2.2 1.9 3.2 1.9 1760.3 93.0 1860.7 5.2 25.1 90.6 2.9

44.8 357.8 21.6 2.6 2.0 3.9 2.2 1782.8 102.7 1866.8 323.7 29.5 95.4 5.4

37.9 326.0 12.6 2.4 1.9 3.4 2.0 1755.4 95.7 1844.3 185.2 26.4 83.4 3.6

16.8 32.0 8.6 0.5 0.1 1.0 0.8 32.8 10.2 33.7 277.2 3.3 14.0 1.6

24 h 1h 1h 1h 1h N/A N/A N/A N/A N/A N/A N/A N/A N/A

150 30000 100 170 130 N/A N/A N/A N/A N/A N/A N/A N/A N/A

Note: RMAQG = Recommended Malaysian Air Quality Guideline; Min = Minimum; Max = Maximum; SD = Standard Deviation; Time = Time Averaging; Q1 = Quartile 1; Q3 = Quartile 3; N/A = Not Available; Temperature = Ambient Temperature; Humidity = Relative Humidity. The concentration of PM10 is based on daily average concentrations. The concentration of gases pollutants and meteorological variables are based on yearly average of hourly concentrations.

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348

45

CH4 (ppb)

PM10 (µgm-3)

340

40 35

1800 1700 1600

30 3

5

7

9

11 13 15 17 19 21 23

NmHC (ppb)

1 700

CO (ppb)

1900

500 300

1

3

5

7

9

11 13 15 17 19 21 23

1

3

5

7

9

11 13 15 17 19 21 23

1

3

5

7

9

11 13 15 17 19 21 23

1

3

5

7

9

11 13 15 17 19 21 23

1

3

5

7

9

11 13 15 17 19 21 23

1

3

5

7

9

11 13 15 17 19 21 23

1

3

5

7

9

11 13 15 17 19 21 23

150

100

50

100 1

3

5

7

9

11 13 15 17 19 21 23

10 5

7

9

2

1

3

5

7

9

NOx (ppb)

6.1 4.1 2.1 0.1 3

5

7

9

11 13 15 17 19 21 23

UVb (Jm-2h-1)

1 6 4 2 0 1

3

5

7

9

11 13 15 17 19 21 23

5 3 1 -1 1

3

5

7

9

112 62 12

11 13 15 17 19 21 23

8.1

NO (ppb)

1800

11 13 15 17 19 21 23

Humidity (%)

3

1

NO2 (ppb)

1950

1650 1

Temperature (OC)

-5

SO2 (ppb)

THC (ppb)

25

11 13 15 17 19 21 23

Hour

38 28 18 900 650 400 150 -100 -350

Wind Speed (kmh-1)

O3 (ppb)

40

8 6 4 2 0

Hour

Fig. 2. Diurnal patterns for yearly average concentration of 14 variables between 1997 and 2011.

due to photochemical processes, as determined through previous studies (e.g. by Wang et al. (2013)). The concentration of CH4 and NmHC was negatively influenced by the temperature and level of UVb during daylight hours. The concentrations of CH4 and NmHC are high around 1.00–8.00 h and 20.00–24.00 h when the temperature and UVb are low. At around 14.00–15.00 h, UVb and temperature reached their highest levels with 790.42 J m−2 h−1 and 32.09 °C respectively. Conversely, NmHC and CH4 recorded their lowest concentrations at 13.00 h (83.75 ppb) and 16.00 h (1704.04 ppb) respectively (Fig. 2). O3 has a unimodal concentration distribution which is closely linked to UVb intensity since its formation involves the photochemical reaction

of NOx (Sillman, 1999; Monks et al., 2009). The concentration of O3 rose in the study area at around 14.00–16.00 h, about 1 h after UVb achieved its highest intensity (Fig. 2). UVb can initiate O3 precursors, such as NO2 and Volatile Organic Compounds (VOCs), in ambient air and release the oxygen atom (O) which influences the reaction with oxygen molecule (O2) to form O3 (Kleinman, 1994). Without the high intensity of UVb and the high quantity of O3's precursors, particularly NO2, the amount of O3, especially at night, will be very limited (Kleinman et al., 2000). SO2 profiles showed increasing concentrations from around noon to late evening. The peak of SO2 concentrations was found to occur between 18.00 and 19.00 h (Fig. 2). Motor vehicle emissions and

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348

conditions of atmospheric stability may influence the concentrations of SO2 at this particular time. Two-stroke motorcycle engines used by residents are an important source of SO2 emissions as these gasoline fuelled engines emit higher amounts of SO2 than cars (Yasar et al., 2013). In the mornings, the concentration of PM10 was found to be slightly higher between 8.00 and 9.00 h but then reduced significantly towards the afternoon (Fig. 2). The increase in mixing height (as temperature steadily increases between 9.00 and 15.00 h) and its subsequent decrease (when temperatures drop until it reaches a plateau around 21.00 h) shows that physical processes, such as dilution with free tropospheric air, can reduce PM10 concentration. Further investigation on the diurnal pattern was undertaken where the temporal pattern was determined by clustering standardised hourly averages of yearly concentrations for 15 years (1997–2011). Seven important air quality variables (CO, O3, PM10, SO2, NO, NO2 and NOx) were applied to HACA. The resulting temporal pattern is illustrated in a dendrogram (Fig. 3). The outcomes are consistent with the results of the diurnal pattern (Fig. 2). The temporal pattern shows 4 groups named as Group 1, Group 2, Group 3 and Group 4 (Fig. 3). Group 2, which consists of 7.00, 8.00, 9.00 and 10.00 h, corresponds to the emissions contributed by early morning traffic while Group 4 (19.00, 20.00, 21.00 and 22.00 h) corresponds to the peak resulting from late evening traffic and air mass stability (temperature inversion). The late night and early morning hours correspond to reduced pollutant concentrations. Group 3 consists of 11.00, 12.00, 13.00, 14.00, 15.00, 16.00, 17.00 and 18.00 h while Group 1 consists of 1.00, 2.00, 3.00, 4.00, 5.00, 6.00, 23.00 and 24.00 h. Group 3 fits the time frame when the UVb intensity is sufficiently significant to initiate photochemical reactions and subsequent nonphotosensitive reactions.

4.3. Monthly and yearly patterns Generally speaking, the monthly variation patterns in this study are influenced by the seasonal monsoons, namely the northeast (NE) and southwest monsoon (SW). The NE monsoon occurs between December and March and is associated with the long-range transport of air mass from the east coast of Indo-China (Latif et al., 2012). The SW monsoon typically starts around June and ends around September. During the SW monsoon, air mass movement tends to originate from the southwestern region (Fig. 4). During the inter monsoon periods (April–May

Dendrogram 80

(Dlink/Dmax) x 100

70 60 50 40

Traffic Emission

30 20

Group 3

Group 4 Group 2

0

Group 1

18 17 14 15 16 11 12 13 19 20 21 22 10 7 8 9 4 5 6 2 3 23 1 24

10

Hour Fig. 3. Dendrogram of diurnal temporal pattern based on yearly average concentration of seven variables (PM10, CO, NO, NO2, SO2, NOx and O3) for 15 years period (1997–2011).

341

and October–November), which are associated with slower atmospheric circulation, the clustered trajectory for April and November shows that the distance travelled by the air mass is generally much shorter. The monthly concentrations for the majority of the air pollutants, such as: PM10, O3, NOx, NO2, CO and NmHC, were recorded at higher concentrations between June and September during the southwest monsoon (Fig. 5). This was expected due to the wind direction which was blowing in from Sumatra, Indonesia, as shown in Fig. 4. Biomass burning is predominantly a contributing factor when there is a high amount of particulate matter in the atmosphere. This has been shown by other studies which have indicated that the concentration of PM10 from biomass burning in Sumatra can be blown from its sources within several hours (Anwar et al., 2010; Juneng et al., 2011). During this season, the movement of air pollutants from more urbanised areas along the west coast of the Malaysian Peninsular, such as Kuala Lumpur city centre and Jerantut city, are also possible, as is shown by wind rose in Fig. 6. Urbanisation contributes to a high amount of air pollutants, such as NOx and NmHC, particularly from motor vehicles and industrial activity. The long-range movement of air has the ability to transport pollutants from city centres to background areas, such as Jerantut. The amount of NO2 and NmHC originating from the city centre could contribute to the level of O3 at Jerantut station. In contrast, other pollutants, such as SO2 and CH4, did not seem to be affected by the regional monsoon. SO2 itself, has a very short lifetime and can easily transform into SO24 − through oxidation and interaction with particles and water vapour (Kai et al., 2007). Whilst CH4 can originate from a variety of sources, particularly from: biogenic degradation and the agricultural activities which occur around the sampling station. Overall yearly patterns from 1997 show that the concentrations of air pollutants recorded at Jerantut station demonstrate an increasing trend, except for CO, SO2 and NmHC which show − 0.04, − 0.27 and −0.34 fit line (r) values respectively (Fig. 7). NOx (NO and NO2) is the main pollutant and shows the highest increasing trend with an rvalue of 0.83. NO2 is more dominant compared to NO, where the rvalues are 0.76 and 0.11, respectively. Other pollutants that have positive r-values are THC (0.73), O3 (0.68), CH4 (0.57) and PM10 (0.05). The fluctuation of air pollutants in different years is to be expected as a result of differing weather conditions, an increase in the number of motor vehicles on the road (Fig. 8), activities by residents such as burning of fallen leaves as well as regional emissions, particularly those resulting from industrial activity and biomass burning. The yearly average concentrations of O3, NO, NO2 and SO2 in this study were found to be lower than those obtained in studies undertaken by Meng et al. (2010) and Melkonyan and Kuttler (2012) (Table 2). Whilst the yearly average of PM10 (ranging from 31 μg m− 3 to 50 μg m− 3 ) is under the maximum limit as suggested by RMAQG for an annual average (50 μg m−3), it does however, exceed the level recommended by WHO guidelines for average annual PM10 concentrations (20 μg m−3). The yearly concentration of PM10 obtained in this study was almost identical to those previously recorded in rural areas of Rio de Janeiro, Brazil at 34 μg m− 3 (Gioda et al., 2011) and Hong Kong with an average of 37 μg m−3 (Wai and Tanner, 2005). Nevertheless, the concentration of PM10 in this study far exceeds those recorded in rural areas in Europe. Namdeo and Bell (2005), for example, recorded a PM10 concentration of around 14 μg m−3 at Rochester in the United Kingdom (Table 2). This big difference in yearly concentrations of PM10 between European and regional tropical countries is, however, caused by several factors. The main one is the higher temperature in tropical countries which leads to high evaporation and resuspension of particles in ambient air. High temperatures tend to cause hot weather and lower the relative humidity level, which in turn promotes local and regional biomass burning, and subsequently increases the quantity of particles in ambient air (Juneng et al., 2011). The yearly concentrations of organic pollutants, such as: THC, NmHC and CH4 were found to follow an upward trend, particularly after 2004, with the exception of NmHC. Additionally, all three of these pollutants

342

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348

Fig. 4. The mean trajectories for the air mass (%) calculated using HYSPLIT for the period between January 2009 and December 2011. Each trajectory groups are indicated by different line colours.

were recorded at the highest level in 2009 (Fig. 7). Development through deforestation and agricultural activities in this area since the 1990's is likely to contribute to the increasing trend of organic pollutants in the study area. An increasing number of oil palm plantations which were established in the middle of the Malaysian Peninsular, particularly in areas near to Jerantut, have led to the construction of mills to process the palm oil. Deforestation, resulting from the development and construction of new housing areas along with local biomass combustion activities, is also expected to have contributed to higher levels of organic gases around the study area. Decreases in the concentration of SO2, beginning in 2000, may be due to reductions in the percentage of sulphur in petrol and diesel, as asserted by Azmi et al. (2010). Fluctuations in SO2 levels may have been influenced by local and regional emissions of SO2, for example, from biomass burning, diesel engine and palm oil mills.

4.4. Traffic patterns and their effects on pollutant concentrations The 4-year traffic data: 2002, 2006, 2008 and 2010 was provided by the Malaysian Public Works Department (PWD) under the Malaysian Ministry of Works (MOW). The traffic data was recorded for 16 h daily (6.00–22.00 h) and it was noted that the total number of vehicles in the study steadily increased between 2002 and 2010 (Fig. 8). Sensitivity analysis, using the one-factor-at-a-time (OAT) technique, was undertaken so as to examine the level of importance for CO, NO, NO2, NOx, O3, PM10, SO2, THC, CH4 and NmHC. The standardised regression determination coefficient (R2) values were used to estimate the relationship between traffic and pollutant concentrations. The results show that the ten variables can be classified into three groups based on their level of importance. The ‘least important’ group, as influenced

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348

700

CO (ppb)

30

O3 ( ppb)

343

20 10

500 300 100

0 J

F M

A M

J

J

A

S

O

N

D

J

F M A M

J

J

A

S

O N D

J

F M A M

J

J

A

S

O N D

J

F M A M

J

J

A

S

O N D

J

F M A M

J

J

A

S

O

J

F M A M

J

J

A

S

O N D

PM10 (µgm-3)

NOx ( ppb)

7 5 3 1 F M

A M

J

J

A

S

O

N

50

D

CH4 ( ppb)

7

NOx ( ppb)

100

0 J

5 3

2000 1800 1600

1 J

F M

A M

J

J

A

S

O

N

D

NmHC ( ppb)

3

NO (ppb)

150

2 1

200 100 0

0 J

F

M

J

A M

J

S

A

O

D

N

N

D

6

THC (ppb)

SO2 (ppb)

2200 4 2

1900

1600

0 J

F M

A

M

J

J

A

S

O

N

D

Month

Month

Fig. 5. The box plot of monthly variations of air pollutants from 1997–2011. The edges of the box indicate the 25th and the 75th of the daily sample values. The rhombuses indicate the maximum and minimum values.

by motor vehicles, consisted of CO, SO2, NO2 and NmHC. This group showed a distribution range between 0.4% and 3.3% (Fig. 9). The ‘moderately important’ group, which consisted of NO, NOx and PM10, had a distribution range between 8.2% and 10.8%. The ‘most important’ group, which was influenced by motor vehicle emissions, consisted of THC, O3 and CH4. Motor vehicle emissions were found to influence the THC concentration the most, while affecting NmHC the least, at 26.1% and 0.4%, respectively (Fig. 9). This outcome strongly suggests that the air quality level in the study area has yet to be seriously impacted by the increasing volume of traffic it now experiences, since the air pollutant indicators for motor vehicles (e.g. CO, NmHC, NO2 and NOx) were classified in the low-moderate groups (Fig. 9). 4.5. Significant ratio values for selected air pollutants The standardised hourly average of yearly concentrations for the 15year data set was applied to index analysis (ratios) in order to examine changes in the pollutants source structures. Selected air pollutants, such as: CO, NOx, SO2, NO and NO2, were used because they are significant variables relating to mobile and point source emissions (Aneja et al., 2001). Within the 15-year observation period, the air quality level in the study area was dominantly influenced by the CO/NOx ratio values (1998–2011) with ratio values ranging from 92 to 190. The SO2/NOx ratio values were only higher than CO/NOx in 1997 (Fig. 10). Correlation

analysis (p b 0.01) showed that CO and NOx have a strong positive significant relationship with the regression determination coefficient (R2) value = 0.829 and correlation coefficient (r) value = 0.911. For SO2 and NOx, they showed a negative relationship but there was no significant correlation between these two variables (R2 = 0.043, r = −0.209) at p b 0.01. Furthermore, the calculated NO/NO2 ratio values in this study were less than 2 (0.6 ≤ NO/NO2 ≤ 1.1) compared with those determined in the study undertaken by Liu (1991). The relationship between NO and NO2 showed a positive correlation but lacked any significant correlations (R2 = 0.054, r = 0.233). These ratio value results indicate that air pollution in the study area of Jerantut station, is not yet seriously influenced by mobile sources, as according to Liu (1991), if the NO/NO2 ratio value is greater than 2, then the air pollution in the study area can be determined as being the result of traffic. This result is consistent with others presented in Section 4.4, which discusses traffic patterns and their effect on pollutant concentrations. 4.6. The impact of meteorological factors on pollution levels In this study, the hourly averages of the yearly concentrations for the 15-year (1997–2011) dataset were applied to the PCR analysis in order to investigate the statistical relationship between the air pollutant variables and meteorological parameters. The air pollutant variables used in this analysis were: CO, O3, PM10, SO2, NO, NO2, NOx, CH4, THC and NmHC

344

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348

Fig. 6. Ground level wind speed (km/h) and direction (°) measured hourly for a span of 3 years in 2009–2011.

whilst the meteorological parameters were: temperature (T), humidity (H), wind speed (W) and UVb. Principal component analysis (PCA) after varimax rotation together with Kaiser Normalisation, were applied to the air pollutant variables. Only strong factor loadings (N 0.70) were selected for the interpretation of results. Two factors with eigenvalues of more than 1.0 with a cumulative variability of 89.39% were produced (Table 3). Factor 1 (F1) explained 60.67 % of the variability of the dataset consisting of O3, SO2, NO, CH4 and THC whereas Factor 2 (F2) consisted of CO, PM10, NO2, NOx and NmHC. Next, standardised multiple regression analysis was performed on significant PCs against the meteorological variables using OLS analysis. In this study, the dependent variables were the significant PCs while the meteorological variables (i.e. temperature (T), humidity (H), wind

speed (W) and UVb) were the independents. The equations for the two factors are illustrated in Eqs. (1)–(2). F1 ¼ −2:919ðTÞ þ 0:831ðHÞ þ 0:417ðWÞ þ 1:084 ðUVb Þ   2 þ20:667 R ¼ 0:982

ð1Þ

F2 ¼ 4:628ðTÞ–2:336ðHÞ–0:021ðWÞ–3:480 ðUVb Þ   2 –32:599 R ¼ 0:700

ð2Þ

Meteorological parameters influenced the F1 variability the most (98.2%) with a coefficient of determination (R2) value of 0.982.

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348

345

Fig. 7. Long-term yearly variations pattern for air pollutions from 1997 to 2011: (a) Yearly variation for major pollutants; (b) Yearly variation for organic pollutants.

Conversely, meteorological parameters only influenced 70.0% of the F2 variability, resulting in an R2 value of 0.700. In F2, T, H and UVb generated stronger standardised coefficient values compared to values generated in F1 while W shows a stronger standardised coefficient value in F1 compared to F2 (Eqs. (1)–(2)). However for both F1 and F2, T and UVb showed the strongest influence. Since the dominant pollutants particularly in F2 are mostly active precursors (NOx, NO2, CO) for secondary pollutants and some are photochemically sensitive species, they would be strongly influenced by meteorological parameters such as UVb and T. The relative influence of individual meteorological parameters to the pollutant species and pollutant groups can be explained by relating

Number of Vehicles

60000 50000

between the results generated in the equations above and the rotated factor loadings (Table 3). All the pollutants in F2 (PM10, CO, NO2, NOx and NmHC) show positive association with T, but negative association with H, W and UVb. In F1, T has a positive association with O3 and SO2 but a negative association with NO, CH4 and THC. These results shows that meteorological parameters play an important role in air pollution variability. They are closely associated to pollutant concentrations as they influence pollutant dispersion and chemical reactions in photosentive reactions. 5. Conclusion This study shows the long-term pattern of major air pollutants at a background station in Malaysia. The results demonstrate that the concentrations of all pollutants, apart from PM10, are lower than the recommended levels suggested by RMAQG and WHO. Tropical weather conditions are believed to influence the yearly average concentration

40000 Table 2 Comparison of the yearly average between this study and other studies. All concentrations are in ppb except for PM10 (µg m-3)

30000 20000

References

10000 0 2002

2006

2008

Year Fig. 8. Traffic patterns for 2002, 2006, 2008 and 2010.

2010

This study Namdeo and Bell, 2005 Meng et al., 2010 Melkonyan and Kuttler, 2012 Kulshrestha et al., 2009

Variables O3

NO

NO2

SO2

PM10

NOx

11.8 – – 54.1 –

1.3 8.6 – 3.65 –

1.9 – 11.5 13.76 –

1.1 – 7.55 – –

38.2 14.0 – – 148.4

3.1 – – – –

346

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348

NO 10.04%

NO2 8.30%

CH4 13.55%

CO 3.11%

NOx 8.30%

SO2 0.56%

Other 11.69%

NmHC 0.37%

O3 23.82% PM10 7.66%

THC 24.29% Fig. 9. Influenced of motor vehicles on the distribution of the variables.

of PM10 (38 μg m−3) which exceeds the average annual level recommended by WHO. Long-range transport may also influence the formation of surface ozone due to the movement of ozone precursors in the study area. Monthly air quality data shows that the long-range transport of air pollutants, especially from regional biomass burning episodes in Southeast Asia, can contribute to elevated PM10 concentrations in the study area. Whilst no other pollutants exceeded RMAQG, their variability is still important in indicating possible emission sources. Although the sampling site chosen is located in a rural area close to a forest reserve, localised anthropogenic emission sources (such as motor vehicles) are major contributors to air pollution in the study area. This can be shown by the amount of air pollutants (NO2 and NmHC) present during peak traffic hours in the morning (7.00–9.00 h) and late afternoon (17.00– 19.00 h) at the times people generally commute to and from work. The ratio values between major air pollutants also indicate that air pollution in the study area is moderately influenced by mobile sources. Results of the study demonstrate that there in an upward trend for organic gases, such as: THC and CH4, in the study area. Deforestation resulting from the expansion of agricultural activity and urbanisation is also accredited as being a contributing factor to organic gas compositions in the study area. Furthermore, the sensitivity analysis demonstrates that an increase in the number of motor vehicles had the most positive influence on the concentration of THC. Added to this,

meteorological factors such as: wind speed, temperature, humidity and UVb were found to influence the concentration of air pollutants in the study area and it was also determined that the SW monsoon has the capacity to transport air pollutants from urbanised areas on the west coast of the Malaysian Peninsular, including Kuala Lumpur city centre, to the rural background study area. The results obtained from Jerantut station, the Malaysian air quality background sampling station, are especially valuable since they not only aid in the assessment of air quality levels on a regional scale and allow comparison of pollutant levels in areas directly affected by pollution sources, but also provide comparative data between the pre- and post-modernisation status of an area. The results of this study demonstrate that for the time being, Jerantut station can still be effectively utilised as the background station for Malaysia. Nevertheless, in time, increases in motor vehicle numbers will influence the concentration of pollutants at this station. This being the case, pollution levels at this station should be monitored on a regular basis in order to maintain the position of Jerantut station as a background station, particularly since at specific times, for example during peak hour traffic flow, observations show the concentration of some pollutants being higher than usual. Local pollution emission sources which result from increased urbanisation and the expansion of industrial activity also need to be controlled if emissions from any point source are to be significantly reduced around the study area. Any future policy focusing on managing Malaysian air

200

150 1.0

100 0.5 50

0.0

0 97'

98'

99'

00'

01'

02'

03'

04'

05'

06'

07'

08'

09'

10'

Years CO/NOx

SO2/NOx

NO/NO2

Fig. 10. Hourly averaged value of CO/NOx, SO2/NOx and NO/NO2 ratio values from 1997 to 2011.

11'

(CO/NOx)

[(SO2/NOx, NO/NO2)]

1.5

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348 Table 3 Factor loadings for principle component analysis after varimax rotation. Variables

F1

F2

O3 CO SO2 NOx NO NO2 PM10 CH4 THC NmHC Eigenvalue Variability (%) Cumulative %

−0.92 0.35 −0.89 0.59 0.73 0.01 0.42 0.99 0.93 −0.13 6.07 60.67 60.67

−0.30 0.92 0.32 0.72 0.37 0.96 0.74 0.06 0.35 0.98 2.87 28.72 89.39

Note: Strong factor loading values (N0.7) are in bold.

quality should in effect take into consideration the changes in pollutant concentrations around this background station. A background sampling station located on a remote island or in a remote highland area, which is minimally influenced by human activity, would be a viable option should an alternative or additional background station in Malaysia be required. Acknowledgements The authors would like to thank Universiti Kebangsaan Malaysia for the Research University Grants (UKM-AP-2011-19 ICONIC-2013-004 and DIP-2012-020) and the Ministry of Education for the Fundamental Research Grant (FRGS/1/2013/STWN01/UKM/02/2). We are grateful to the Malaysian Department of the Environment (DOE) for permitting us to use the air quality data. The authors also gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and/or READY website (http://www.arl.noaa.gov/ready.html) used in this publication. Special thanks to Ms K Alexander for proofreading this manuscript. References Adams M. The principles of multivariate data analysis. London: Blackie Academic & professional: Ashurst & M.J.Dennis (Eds), 1998. Agrawal M, Singh B, Rajput M, Marshall F, Bell JNB. Effect of air pollution on peri-urban agriculture: a case study. Environ Pollut 2003;126:323–9. Aneja VP, Agarwal A, Roelle PA, Phillips SB, Tong Q, Watkins N, et al. Measurements and analysis of criteria pollutants in New Delhi, India. Environ Int 2001;27:35–42. Anwar A, Juneng L, Othman MR, Latif MT. Correlation between hotspots and air quality in Pekanbaru, Riau, Indonesia in 2006–2007. Sains Malaysiana 2010;39:169–74. ASMA. Standard Operating Procedure for Continuous Air Quality Monitoring. Selangor: Alam Sekitar Sdn. Bhd. Shah Alam; 2007. Azmi SZ, Latif MT, Ismail AS, Juneng L, Jemain AA. Trend and status of air quality at three different monitoring stations in the Klang Valley, Malaysia. Air Qual Atmos Health 2010;3:53–64. Banan N, Latif MT, Juneng L, Ahamad F. Characteristics of surface ozone concentrations at stations with different backgrounds in the Malaysian Peninsula. Aerosol Air Qual Res 2013;13:1090–106. Chaloulakou A, Mavroidis I, Gavriil I. Compliance with the annual NO2 air quality standard in Athens. Required NOx levels and expected health implications. Atmos Environ 2008;42:454–65. Coppalle A, Delmas V, Bobbia M. Variability of NOx and NO2 concentrations observed at pedestrian level in the city centre of medium sized urban area. Atmos Environ 2001;35:5361–9. Dominick D, Juahir H, Latif MT, Zain SM, Aris AZ. Spatial assessment of air quality patterns in Malaysia using multivariate analysis. Atmos Environ 2012;60:172–81. Donnelly A, Misstear B, Broderick B. Application of nonparametric regression methods to study the relationship between NO2 concentrations and local wind direction and speed at background sites. Sci Total Environ 2011;409:1134–44. Draxler R, Rolph G. HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory). model access via NOAA ARL READY website (http://www.arl.noaa.gov/ready/ hysplit4.html) Silver Spring. Md: NOAA Air Resources Laboratory; 2003. European Union (EU). Directive on Ambient Air Quality and Cleaner Air for Europe (the “CAFE” Directive) (COM(2005) 447); 2005. Gioda A, Amaral BS, Monteiro ILG, Saint'Pierre TD. Chemical composition, sources, solubility, and transport of aerosol trace elements in a tropical region. J Environ Monit 2011; 13:2134–42.

347

Goyal P, Sidhartha. Present scenario of air quality in Delhi: a case study of CNG implementation. Atmos Environ 2003;37:5423–31. Grawe D, Thompson HL, Salmond JA, Cai XM, Schlünzen KH. Modelling the impact of urbanisation on regional climate in the Greater London Area. Int J Climatol 2013;33: 2388–401. Han NM, Latif MT, Othman M, Dominick D, Mohamad N, Juahir H, et al. Composition of selected heavy metals in road dust from Kuala Lumpur city centre, Environ Earth Sci; 2013. http://dx.doi.org/10.1007/s12665-013-3008-5. Article in press at: http:// link.springer.com/article/10.1007/s12665-013-3008-5. Hanna SR. Air quality model evaluation and uncertainty. J Air Pollut Control Assoc 1988; 38:406–12. Haszpra L, Barcza Z, Hidy D, Szilágyi I, Dlugokencky E, Tans P. Trends and temporal variations of major greenhouse gases at a rural site in Central Europe. Atmos Environ 2008;42:8707–16. Ibarra-Berastegi G, Sáenz J, Ezcurra A, Ganzedo U, Díaz de Argandoña J, Errasti I, et al. Assessing spatial variability of SO2 field as detected by an air quality network using Self-Organizing Maps, cluster, and Principal Component Analysis. Atmos Environ 2009;43:3829–36. Juneng L, Latif MT, Tangang FT, Mansor H. Spatio-temporal characteristics of PM10 concentration across Malaysia. Atmos Environ 2009;43:4584–94. Juneng L, Latif MT, Tangang F. Factors influencing the variations of PM10 aerosol dust in Klang Valley, Malaysia during the summer. Atmos Environ 2011;45:4370–8. Kai Z, Yuesi W, Tianxue W, Yousef M, Frank M. Properties of nitrate, sulfate and ammonium in typical polluted atmospheric aerosols (PM10) in Beijing. Atmos Res 2007;84: 67–77. Khandekar M, Murty T, Scott D, Baird W. The 1997 El Nino, Indonesian forest fires and the Malaysian smoke problem: A deadly combination of natural and man-made hazard. Nat Hazards 2000;21:131–44. Kim JO, Mueller CW. Introduction to factor analysis: What it is and how to do it. Quantitative applications in the social sciences series. Newbury Park: Sage University Press; 1987. Kleinman L. Ozone formation at a rural site in the southeastern United States. J Geophys Res 1994;99:3469–82. Kleinman LI, Daum PH, Imre DG, Lee JH, Lee YN, Nunnermacker LJ, et al. Ozone production in the New York City urban plume. J Geophys Res D Atmos 2000;105:14495–511. Kulshrestha A, Satsangi PG, Masih J, Taneja A. Metal concentration of PM2.5 and PM10 particles and seasonal variations in urban and rural environment of Agra, India. Sci Total Environ 2009;407:6196–204. Kumar AV, Patil KS, Nambi KSV. Source apportionment of suspended particulate matter at two traffic junctions in Mumbai, India. Atmos Environ 2001;35:4245–51. Latif MT, Huey LS, Juneng L. Variations of surface ozone concentration across the Klang Valley, Malaysia. Atmos Environ 2012;61:434–45. Lau J, Hung WT, Cheung CS. Interpretation of air quality in relation to monitoring station's surroundings. Atmos Environ 2009;43:769–77. Lin W, Xu X, Ma Z, Zhao H, Liu X, Wang Y. Characteristics and recent trends of sulfur dioxide at urban, rural, and background sites in North China: Effectiveness of control measures. J Environ Sci 2012;24:34–49. Liu CM. A study of air quality characteristics and trend in Taiwan area. R.O.C.-EPA; 1991 [Contract No. EPA-81–E3L1–09–02]. Liu Y, Gupta HV, Sorooshian S, Bastidas LA, Shuttleworth WJ. Exploring parameter sensitivities of the land surface using a locally coupled land‐atmosphere model. J Geophys Res Atmos (1984–2012) 2004;109. Liu Y, Chen L, Huang Q-H, Li W-Y, Tang Y-J, Zhao JF. Source apportionment of polycyclic aromatic hydrocarbons (PAHs) in surface sediment of the Huangpu River, Shanghai, China. Sci Total Environ 2009;407:2931–8. Manache G, Melching CS. Identification of reliable regression and correlation based sensitivity measures for importance ranking of water quality model parameters. Environ Model Softw 2008;23:549–62. Mavroidis I, Ilia M. Trends of NOx, NO2 and O3 concentrations at three different types of air quality monitoring stations in Athens, Greece. Atmos Environ 2012;63:135–47. McKenna Jr JE. An enhanced cluster analysis program with bootstrap significance testing for ecological community analysis. Environ Model Softw 2003;18:205–20. McRae GJ, Seinfeld JH. Development of a second-generation mathematical model for urban air pollution—II. Evaluation of model performance. Atmos Environ (1967) 1983;17:501–22. Melkonyan A, Kuttler W. Long-term analysis of NO, NO2 and O3 concentrations in North Rhine-Westphalia, Germany. Atmos Environ 2012;60:316–26. Meng Z-Y, Xu X-B, Wang T, Zhang X-Y, Yu X-L, Wang S-F, et al. Ambient sulfur dioxide, nitrogen dioxide, and ammonia at ten background and rural sites in China during 2007–2008. Atmos Environ 2010;44:2625–31. Monks PS, Granier C, Fuzzi S, Stohl A, Williams ML, Akimoto H, et al. Atmospheric composition change—global and regional air quality. Atmos Environ 2009;43:5268–350. Morris MD. Factorial sampling plans for preliminary computational experiments. Technometrics 1991;33:161–74. Namdeo A, Bell MC. Characteristics and health implications of fine and coarse particulates at roadside, urban background and rural sites in UK. Environ Int 2005;31:565–73. Radzi bin Abas M, Oros DR, Simoneit BRT. Biomass burning as the main source of organic aerosol particulate matter in Malaysia during haze episodes. Chemosphere 2004;55: 1089–95. Sani S. Urban climatology in Malaysia: An overview. Energy Build 1990;15:105–17. Sauvage S, Plaisance H, Locoge N, Wroblewski A, Coddeville P, Galloo JC. Long term measurement and source apportionment of non-methane hydrocarbons in three French rural areas. Atmos Environ 2009;43:2430–41. Shrestha S, Kazama F. Assessment of surface water quality using multivariate statistical techniques: A case study of the Fuji river basin, Japan. Environ Model Softw 2007; 22:464–75.

348

M.T. Latif et al. / Science of the Total Environment 482–483 (2014) 336–348

Sillman S. The relation between ozone, NOx and hydrocarbons in urban and polluted rural environments. Atmos Environ 1999;33:1821–45. Singh KP, Malik A, Mohan D, Sinha S. Multivariate statistical techniques for the evaluation of spatial and temporal variations in water quality of Gomti River (India)—A case study. Water Res 2004;38:3980–92. Singh KP, Malik A, Sinha S. Water quality assessment and apportionment of pollution sources of Gomti river (India) using multivariate statistical techniques—A case study. Anal Chim Acta 2005;538:355–74. Sousa SIV, Martins FG, Alvim-Ferraz MCM, Pereira MC. Multiple linear regression and artificial neural networks based on principal components to predict ozone concentrations. Environ Model Softw 2007;22:97–103. Tang JH, Chan LY, Chan CY, Li YS, Chang CC, Liu SC, et al. Characteristics and diurnal variations of NMHCs at urban, suburban, and rural sites in the Pearl River Delta and a remote site in South China. Atmos Environ 2007;41:8620–32.

U.S. Environmental Protection Agency (USEPA). Quality assurance handbook for air pollution measurement systems. Part I: Ambient air quality monitoring program quality system development, vol. II. Research Triangle Park, NC: Office of Air Quality Planning and Standards; 1998. Wai KM, Tanner PA. Extreme particulate levels at a Western Pacific Coastal City: The influence of meteorological factors and the contribution of long-range transport. J Atmos Chem 2005;50:103–20. Wang JL, Chew C, Chang CY, Liao WC, Lung SCC, Chen WN, et al. Biogenic isoprene in subtropical urban settings and implications forair quality. Atmos Environ 2013;79: 369–79. Yasar A, Haider R, Tabinda AB, Kausar F, Khan M. Short communication a comparison of engine emissions from heavy, medium, and light vehicles for CNG, diesel, and gasoline fuels. Pol J Environ Stud 2013;22:1277–81.