Characterizing and predicting coarse and fine particulates in classrooms located close to an urban roadway

Abstract

The PM₁₀, PM_2.5, and PM₁ (particulate matter with aerodynamic diameters <10, <2.5, and <1 μm, respectively) concentrations were monitored over a 90-day period in a naturally ventilated school building located at roadside in Chennai City. The 24-hr average PM₁₀, PM_2.5, and PM₁ concentrations at indoor and outdoor environments were found to be 136 ± 60, 36 ± 15, and 20 ± 12 and 76 ± 42, 33 ± 16, and 23 ± 14 μg/m³, respectively. The size distribution of PM in the classroom indicated that coarse mode was dominant during working hours (08:00 a.m. to 04:00 p.m.), whereas fine mode was dominant during nonworking hours (04:00 p.m. to 08:00 a.m.). The increase in coarser particles coincided with occupant activities in the classrooms and finer particles were correlated with outdoor traffic. Analysis of indoor PM₁₀, PM_2.5, and PM₁ concentrations monitored at another school, which is located at urban reserved forest area (background site) indicated 3–4 times lower PM₁₀ concentration than the school located at roadside. Also, the indoor PM₁ and PM_2.5 concentrations were 1.3–1.5 times lower at background site. Further, a mass balance indoor air quality (IAQ) model was modified to predict the indoor PM concentration in the classroom. Results indicated good agreement between the predicted and measured indoor PM_2.5 (R² = 0.72–0.81) and PM₁ (R² = 0.81–0.87) concentrations. But, the measured and predicted PM₁₀ concentrations showed poor correlation (R² = 0.17–0.23), which may be because the IAQ model could not take into account the sudden increase in PM₁₀ concentration (resuspension of large size particles) due to human activities.

Implications:

The present study discusses characteristics of the indoor coarse and fine PM concentrations of a naturally ventilated school building located close to an urban roadway and at a background site in Chennai City, India. The study results will be useful to engineers and policymakers to prepare strategies for improving the IAQ inside classrooms. Further, this study may help in the development of IAQ standards and guidelines in India.

Introduction

Several epidemiological studies in the recent past reported an increase in occurrence of adverse health effects in populations living and working in enclosed environments near major roadways (Harrison et al., 1999; Pearson et al., 2000; Venn et al., 2000; Garshick et al., 2003; Heinrick et al., 2005; Bayer-Oglesby et al., 2006). Many studies have been reported on the impact of outdoor vehicular pollutants on indoor air quality of residential (Kingham et al., 2000; Chao and Wong, 2002; Lawson et al., 2011), commercial (Koponen et al., 2001; Riain et al., 2003; Lim et al., 2011), and school (Chithra and Nagendra, 2012; Pegas et al., 2012; Tran et al., 2012) buildings. In residential and commercial buildings, pollutants can arise from a range of sources, such as environmental tobacco smoke, cooking, domestic chemicals, and furnishings. Classrooms normally lack typical indoor sources such as smoking and cooking. Yet, several researchers reported that the particulate matter (PM) concentrations measured inside classrooms were higher than PM concentrations measured inside residences and commercial buildings (Lee et al., 2002; Oeder et al., 2012).

Motor vehicles are one of the major contributors of PM in urban school buildings. In general, vehicles emit both coarser particles (aerodynamic diameter >2.5 μm) and finer particles (aerodynamic diameter <2.5 μm). The former is emitted by resuspension of road dust and abrasion of brake and tire and the latter is directly from the tail pipe. Finer particles are of great concern to human health, since they can penetrate deep into the respiratory system, take longer time to remove from the body (Miller, 1999), and are associated with many respiratory and cardiovascular diseases (Pearce and Crowards, 1996; Pope et al., 2002; World Health Organization [WHO], 2003; Neuberger et al., 2004; Wallenborn et al., 2009; Mate et al., 2010).

Many researchers have analyzed the effect of PM sources in the immediate vicinity of school (Poupard et al., 2005; Fromme et al., 2007; John et al., 2007) by monitoring PM in more than one school located in different sites such as urban, rural, near traffic, and industrial areas. Lee and Chang (2000) investigated the indoor and outdoor air quality at five schools in Hong Kong and found that high level of PM₁₀ was due to vehicle exhaust emissions, followed by emissions from industrial processes or construction activities. Janssen et al. (2001) also observed that PM_2.5 and soot concentrations in both indoor and outdoor air at schools in the Netherlands were significantly increased with increasing density of truck traffic. Gadkari (2010) studied the indoor fine PM among school communities in mixed urban-industrial environment in India and reported that school located near the steel plant have shown 5–6 times higher values of respirable PM compared with National Ambient Air Quality Standards (NAAQS). Several investigators have compared the PM concentration in different indoor environments such as classrooms, library, administrative office, laboratory, etc., in same school building (Sawant et al., 2004; Diapouli et al., 2008; Gaidajis and Angelakoglou, 2009) and concluded that the resuspension of particles due to occupants’ activities plays an important role in indoor coarse particle concentration.

Indoor air quality monitoring can provide valuable information about the PM level and its sources in the microenvironment. However, practical difficulties, cost, and time involved in continuous measurements have led to the development of modeling techniques for prediction of indoor air pollutants. In the past, several indoor air quality models (IAQMs) ranging from simple regression to complex computational fluid dynamics (CFD) techniques have been developed. A simpler version of IAQM is the mass balance model, which is used for the prediction of indoor concentrations in single or multiple compartments. Mass balance models take a macroscopic view by calculating average concentrations in different zones of the building, usually by assuming well-mixed indoor environments (Hayes, 1989, 1999), where CFD models predict the spatial variation of air velocity and pollutant concentration in a room instead of the average concentration predicted by mass balance models. Because of highly complex nature of CFD models, mass balance models are widely used to predict indoor pollutant concentrations (Chaloulakoua and Mavroidis, 2002; Chang et al., 2003; Goyal and Khare, 2010).

At present, the data on indoor fine PM influenced by heterogeneous traffic flow (different types vehicles, i.e., cars, buses, trucks, auto, and scooters are moving in same lane) under tropical climatic conditions are very scanty. The aim of the present study is to characterize the indoor coarse and fine PM concentrations of a naturally ventilated school building located close to an urban roadway and at a background site in Chennai City, followed by development of IAQ model to predict the indoor PM concentrations in naturally ventilated school building.

Methods

Site characteristics and PM sampling

Chennai (13.04°N, 80.17°E), the capital city of Tamil Nadu, is the fourth largest metropolitan city in India. It is one of the most rapidly growing cities in the world, whose permanent residents had increased from 1.3 million (2001) to more than 4.6 million in 2011. This city is prone to be become a megacity (population more than 10 million) very soon. The rapid urbanization has also enhanced the number of vehicles in the city. Motor vehicle population has increased at a phenomenal rate during the last few decades, with an average growth rate about 9.7% per annum (Chennai Metropolitan Development Authority, 2008). With the rapid increase in the number of vehicles, the problem of air pollution has assumed greater significance. Many authors reported that PM is one of the major air pollutants in Chennai (Gupta and Kumar, 2006; Pulikesi et al., 2006; Gupta et al., 2010; Mohanraj et al., 2011; Srimuruganandam and Nagendra, 2011). Air quality data generated over the years under the National Air Monitoring Programme (NAMP) by Central Pollution Control Board (CPCB) reveal that PM concentrations in Chennai were exceed the NAAQS limits. Based on number of PM violations, CPCB declared Chennai as one among the nonattainment cities in India.

In the present study, two naturally ventilated school buildings were selected, viz., Kendriya Vidyalaya School building at Central Leather Research Institute (KV-CLRI school), located close to a busy traffic intersection, and Kendriya Vidhyalaya School building at Indian Institute of Technology Madras campus (KV-IITM school) as the background site. Figure 1 represents the study site. The KV-CLRI school was located adjacent to a road, which has an average traffic flow of about 174,000 vehicles per weekday and 136,000 vehicles per weekends. The study area is very close to a busy traffic intersection and surrounded by densely populated residential and commercial area. On the other hand, the KV-IITM school is located in a reserved forest area, which is less affected by traffic emissions. No significant pollution sources are located near the measurement site except a road (250 m away) with an average traffic density of 1454 vehicles per day on weekdays and 769 vehicles per day on weekends. Indoor and outdoor air quality monitoring was conducted mainly in KV-CLRI school. Two classrooms in KV-CLRI school and one classroom in KV-IITM school were selected for the study. Among the classrooms selected in KV-CLRI, one classroom (classroom A) is located close to the road (28 m from road) and the other one (classroom B) is located slightly away from the road (65 m from road), with same classroom dimensions (Figure 2). The classroom (classroom C) in KV-IITM is located in reserved forest area, which is considered as background site. Classroom characteristics of both of the schools were presented in Table 1. All the classrooms were equipped with wooden and metallic tables and chairs and a blackboard. The classrooms were cleaned daily. Occupants’ activities were closely monitored in both schools. All the three classrooms were occupied with pupils of age group of 8–9 years. Classrooms were occupied on weekdays, usually from 08:00 a.m. to 04:00 p.m. Significant movement of occupants occurs at the start of the school day (08:00 a.m.), lunch time (12:00 p.m.), and at the end of the school day (03:00 p.m.).

Figure 1. Monitoring site.

Figure 2. Layout plans of school buildings.

Table 1. Characteristics of the classrooms

Classroom Characteristics	KV-CLRI		KV-IITM
	Room A (Near Road)	Room B (Away from Road)	Room C
Number of occupants	43	40	44
Volume (m^3)	122.64	122.64	162.08
Floor area (m^2)	40.88	40.88	50.97
Number of windows Area (m^2)	4 1.56	4 1.56	2 2.37
Number of doors Area (m^2)	1 2.52	1 2.52	1 2.74
Number of fans	4	4	6
Board	Blackboard	Blackboard	Blackboard

For measuring the airborne PM, portable optical Grimm dust monitors (models 1.107 and 1.108) were used (Grimm Labortechnik Ltd., Ainring, Germany). This optical analyzer uses light-scattering technology for single-particle counts. The scattered signal from the particle passing through the laser beam is collected at approximately 90° by a mirror and transferred to a recipient photo diode. After reinforcement, the signal of the diode passes a multichannel size classifier. A pulse height analyzer then classifies the signal transmitted in each channel. These counts are then converted to mass distribution and in particular into the mass cuts of PM₁₀, PM_2.5, and PM₁ (PM with aerodynamic diameters <10, <2.5, and <1 μm, respectively), based on an empirical density factor suitable for the urban suspended PM (Grimm and Eatough, 2009). The ambient air drawn into the unit via an internal volume controlled pump at a rate of 1.2 L/min. Gravimetric calibration is accomplished by measuring the actual mass of dust collected on a back-up 47-mm polytetrafluoroethylene (PTFE), 0.2 μm pore size filter. The model 1.108 dust monitor will give both mass and number distribution of particles ranging from 0.23 to 20 μm (0.23–0.3; 0.3–0.4; 0.4–0.5; 0.5–0.65; 0.65–0.8; 0.8–1; 1–1.6; 1.6– 2; 2–3; 3–4; 4– 5; 5–7.5; 7.5–10; 10–15; 15–20). Indoor-outdoor sampling was done for 90 days at classroom A of KV-CLRI school during July (18 days), August (16 days), October (14 days), November (2 days), and December (8 days) of 2011 and January (4 days), February (8 days), April (18 days), and May (2 days) of 2012. In addition, 1-week monitoring was done simultaneously in both of the schools to study the influence of traffic on PM concentrations (6–12 March 2012). Similarly, 1-week simultaneous monitoring was conducted in classroom A and classroom B of KV-CLRI school (5–11 May 2012). The inlets of the samplers were placed in the breathing zone of the students, i.e., 1 m above the floor. Outdoor monitoring was done only at KV-CLRI school and instrument was kept at 1.5 m above the ground near kerbside. One-week size distribution measurements were also conducted in classroom A using Grimm model 1.108 dust monitor. In order to understand the diurnal and weekly variations of pollutants, monitoring was conducted for 24 hr, covering both weekdays and weekends.

To understand the effect of resuspension on indoor PM concentrations, floor dust loadings in three classrooms were estimated. The floor dust deposited over 24 hr was collected from an area of 1 m² using a vacuum cleaner. Then, the fraction of particles less than 70 microns were determined by sieve analysis (ASTM, 2006). Two sets of samples were collected and the average dust loading for entire classroom area per day was estimated.

IAQ model description

A single-compartment mass balance model ( eq 1) developed by Hayes (1989, 1991) was modified in the present study to predict the indoor PM₁₀, PM_2.5, and PM₁ concentrations. Since school building selected in our study is naturally ventilated, the outdoor makeup air and recirculation term in eq 1 were ignored. Equation 2 presents the modified form of Hayes’s model for naturally ventilated building. The model has been solved recursively by integrating it over an averaging time period T, assuming that the parameters in the equation remain constants over the averaging period (1 hr).

1

where C_o = outdoor concentration; C = indoor concentration; V = building room volume; A = interior surface area; a_M = infiltration flow rate; a_F = outdoor make up air flow rate; E_F = efficiency of the makeup air filter; a_R = flow rate of recirculated air; E_R = efficiency of the recirculated air filter; k = mixing factor; K = pollutant reactivity factor; and S(t) = indoor source generation.

2

Mixing factor (k) is the fraction of pollutant mass that is uniformly mixed throughout the indoor microenvironment. Uniform mixing was assumed in the present study, for which the mixing factor is 1. The pollutant reactivity factor (K) was interpreted as its deposition velocity. In this case, other factors such as coagulation, condensation, evaporation, etc., was not taken into account. Deposition velocity for PM₁₀ and PM_2.5 was taken as 3.26 and 1.10 m/hr, respectively (Thatcher and Layton, 1995). The deposition velocity for PM₁ was ignored because it behaves as perfect gas molecules.

The infiltration flow rate (a_M), which is expressed as air changes per hour (ACH), is the rate at which air passes into a building through open doors, windows, cracks, etc. In India, the recommended value of infiltration rate for classrooms ranges from 5 to 7 ACH (Bureau of Indian Standards [BIS], 2005). In this study, infiltration rate was estimated based on the procedure given by the American Society of Heating, Refrigerating, and Air Conditioning Engineers (ASHRAE) (2009). Natural ventilation and infiltration are driven by pressure differences across the building envelope caused by wind and air density differences due to indoor and outdoor air temperature variation (buoyancy effect). Rate of air forced through ventilation inlet openings by wind, Q_W (m³/sec) is given by eq 3.

3

where C_V = effectiveness of openings (C_V is assumed to be 0.55 for perpendicular winds and 0.3 for diagonal winds); A_nat = free area of inlet openings, m² (since the building has more than one opening, the outlet and inlet areas are considered equal); and U = wind speed, m/sec.

Airflow caused by stack effect, Q_S (m³/sec), can be expressed by eq 4:

4

where C_D = discharge coefficient for opening (dimensionless); g = acceleration due to gravity (9.8 m/sec²); ΔH_NPL = height from midpoint of lower opening to neutral pressure level (NPL), m; T_i = indoor temperature, K; and T₀ = outdoor temperature, K.

Equation 4 applies when T_i > T₀. If T_i < T₀, replace T_i in the denominator with T₀, and replace (T_i − T₀) in the numerator with (T₀ – T_i). The discharge coefficient can be calculated according to the following equation:

5

Combined airflows from wind and stack effects, the airflow from natural ventilation Q was calculated as

6

Infiltration flow rate (a_M) is air flow rate, Q, divided by V.

Indoor source generation term for PM₁₀ and PM_2.5 was estimated as the intercept of linear regression analysis of respective indoor and outdoor concentrations. This method is widely used by many researchers (Dockery and Spengler 1981; Riain et al., 2003; Bennett and Koutrakis 2006; Goyal and Khare, 2010), although it has several limitations, including the assumption of well-mixed and specific for individual building. It was assumed that there is no indoor source for PM₁. For IAQ modeling, 70 days’ (12 days in July, 12 days in August, 10 days in October, 2 days in November, 8 days in December, 4 days in Januarary, 8 days in February, 12 days in April, and 2 days in May) PM concentrations in classroom A were used. In each month, 50% of data were used for model development (obtaining source term) and 50% of data were used for model performance evaluation.

IAQ model performance evaluation

The statistical indicators namely, coefficient of determination (R²), root mean square error (RMSE), and index of agreement (d) were used for the evaluation of model performance. The value of d is expressed as

7

where N is the number of data points, O_i is the observed data, P_i represents predicted data, and is the mean of the observed data. A d value of 1 indicates perfect agreement between the observed and predicted observations, whereas 0 denotes complete disagreement (Willmott, 1982).

Results and Discussion

Characteristics of indoor-outdoor PM concentrations

Table 2 summarizes the 24-hr average PM concentrations monitored in three classrooms (A, B, and C) and the indoor-outdoor PM concentrations and corresponding indoor-outdoor (I/O) ratios in classroom A. In classroom A, the 24-hr average indoor PM₁₀, PM_2.5, and PM₁ concentrations were found to be 136.11 ± 60.43, 36.30 ± 14.98, and 19.80 ± 11.74 μg/m³, with corresponding outdoor concentrations of 76.38 ± 41.96, 33.30 ± 16.42, and 23.02 ± 14.12 μg/m³, respectively. The I/O ratios for PM₁₀, PM_2.5, and PM₁ were found to be 2.25 ± 1.45, 1.15 ± 0.27, and 0.88 ± 0.16, respectively. Analysis of indoor-outdoor PM concentrations revealed that the indoor PM₁₀ concentrations were higher than the outdoor PM levels. On an average, the indoor PM₁₀ concentrations were about 2 times higher than the outdoor level. This confirms the significant contributions from the indoor source. However, during weekends the differences in indoor and outdoor PM₁₀ concentrations were not that significant (1.05 times). During holidays (out of 90 days 14 days were holidays), the 24-hr average indoor and outdoor PM₁₀ concentrations were 76.91 and 73.12 μg/m³, respectively. But the respective indoor and outdoor PM₁₀ concentrations during weekdays were 151.27 and 75.45 μg/m³, respectively. Higher indoor PM₁₀ concentrations can be linked to the resuspension of particles due to the movement occupants and cleaning activities. Use of chalk for writing on the blackboard could also be a dust source. Many authors also reported that human activities have significant effect on indoor coarser particles concentrations (Thatcher and Layton, 1995; Janssen et al., 1997; Janssen et al., 1999; Branis et al., 2005; Poupard et al., 2005; Fromme et al., 2007; Diapouli, 2008; Stranger et al., 2008; Goyal and Khare, 2009).

Table 2. Statistics of 24-hr average indoor-outdoor PM concentrations and I/O ratio in classroom A and summary of PM concentrations in different classrooms

	Indoor-Outdoor PM Concentrations in Classroom A
	Indoor (n = 90)			Outdoor (n = 90)			I/O ratio
	PM10	PM2.5	PM1	PM10	PM2.5	PM1
Statistics	(μg/m^3)	(μg/m^3)	(μg/m^3)	(μg/m^3)	(μg/m^3)	(μg/m^3)	PM10	PM2.5	PM1
Min	49.80	15.90	7.21	19.16	10.86	7.43	0.38	0.70	0.58
Max	286.05	76.22	55.49	196.30	82.59	70.87	8.41	1.69	1.26
Mean	136.11	36.30	19.80	76.38	33.30	23.02	2.25	1.15	0.88
Median	125.52	30.78	16.85	68.97	27.28	19.22	2.10	1.10	0.86
SD	60.43	14.98	11.74	41.96	16.42	14.12	1.45	0.27	0.16
	Comparison of Indoor PM Concentrations in Different Classrooms
	Classroom A (n = 7)			Classroom B (n = 7)			Classroom C (n = 7)
	PM10	PM2.5	PM1	PM10	PM2.5	PM1	PM10	PM2.5	PM1
	(μg/m^3)	(μg/m^3)	(μg/m^3)	(μg/m^3)	(μg/m^3)	(μg/m^3)	(μg/m^3)	(μg/m^3)	(μg/m^3)
Min	123.70	60.44	25.55	46.78	32.33	17.13	—	—	—
Max	171.16	66.93	33.79	106.58	71.96	43.33	—	—	—
Mean	149.79	62.26	29.41	69.28	49.36	24.66	—	—	—
Median	141.20	62.10	30.71	49.65	40.09	25.07	—	—	—
SD	24.01	3.37	4.16	29.66	12.80	6.31	—	—	—
Min	76.57	36.69	17.75	—	—	—	24.51	21.62	12.79
Max	215.98	72.48	43.32	—	—	—	59.76	53.52	37.91
Mean	144.95	54.62	32.12	—	—	—	42.15	37.20	26.08
Median	145.67	56.75	35.41	—	—	—	46.05	38.57	25.16
SD	45.93	14.48	9.91	—	—	—	14.27	12.71	9.64

Notes: n = number of data points.

The indoor PM_2.5 concentration increases as the outdoor PM_2.5 concentration increases and vice versa. Similar trend was also observed for indoor and outdoor PM₁ concentrations. PM₁ concentrations were found to be higher in outdoor than indoor environment. Diurnal pattern of indoor and outdoor PM_2.5 and PM₁ concentrations were correlated with outdoor traffic flow. Indoor and outdoor fine PM concentrations exhibited two peaks during morning and evening hours. The morning peak occurred between 09:00 a.m. and 11:00 a.m., with an average vehicle count of 11,779/hr for weekdays and 7402/hr for Sunday. In weekdays, the average indoor and outdoor PM_2.5 concentrations were 59.09 and 41.48 μg/m³, respectively, and PM₁ concentrations were 23.30 and 24.86 μg/m³, respectively, during morning peak hours. The evening peak occurred around 06:00 p.m. to 08:00 p.m., with an average vehicle count of 10,671/hr for weekdays and 8476/hr for Sunday. Indoor PM concentrations were found to be maximum during weekdays and the minimum concentrations were observed on Sundays. It was found that the indoor PM_2.5 and PM₁ levels on Sundays were on average about 26% and 29% lower than that of weekdays, respectively.

The daily average PM₁₀ and PM_2.5 levels were compared with World Health Organization (WHO) air quality guidelines, NAAQS of U.S. Environmental Protection Agency (EPA), and CPCB, Ministry of Environment and Forests, Government of India. It was observed that the indoor PM₁₀ concentrations were significantly higher at the study area. The 24-hr average indoor PM₁₀ in the classroom A exceeded WHO air quality guidelines (50 μg/m³) for all 90 days monitored and its outdoor concentrations exceeded 73% of the time. Indoor PM₁₀ concentrations exceeded 67% and 41% of time for Indian (100 μg/m³) and EPA NAAQS (150 μg/m³) limits, respectively, whereas most of the time the outdoor PM₁₀ concentrations were within the limits of Indian (77%) and EPA (92%) ambient air quality standards. The indoor PM_2.5 concentrations exceeded 83%, 39%, and 11% of time, respectively, for WHO (25 μg/m³), EPA (35 μg/m³), and Indian (60 μg/m³) standards. Outdoor PM_2.5 levels also exceeded the WHO (68%), EPA (38%), and Indian (9%) standards.

The concentration levels measured in present study were comparable to the concentration levels reported in previous studies conducted worldwide (Table 3). PM concentrations in urban areas of India were much higher than those measured in European countries and elsewhere. In contrast, PM values of the present study were fairly comparable to those in Athens, Greece and Antwerp, Belgium (Diapouli et al., 2008; Stranger et al., 2008). But the indoor PM concentrations were lower than those reported previously for Delhi, India (Goyal and Khare, 2009). However, they were comparable to those of studies conducted in Chhattisgarh, India (Gadkari, 2010) and Hong Kong (Lee and Chang, 2000). The differences in PM concentrations were correlated with the location of school, seasons, and time period of measurement. For example, a study reported in semirural area of California, USA, showed low concentration of PM_2.5 (Aniket et al., 2004). Most of the studies reported higher PM concentration during winter season, expect that of Antwerp, Belgium (Stranger et al., 2008).

Table 3. Indoor-outdoor PM concentrations of classrooms in previous studies

Location	Site Characteristics	Sampling Period	Averaging Time	Indoor PM (μg/m^3)	Outdoor PM (μg/m^3)	Reference
Amsterdam, The Netherlands	Urban, nonindustrial	February–April; January–May	24 hr	PM10 = 45.9–74.4	PM10 = 32–34.5	Janssen et al., 1997
Hong Kong	Urban, residential, industrial, and rural	November–January	24 hr	PM10 = 21–617	—	Lee and Chang, 2000
	Natural and mechanical ventilation
California, USA	Semirural Mechanically ventilated	September–January	24 hr	PM2.5 = 16.3	—	Aniket et al., 2004
Prague, Czech Republic	Urban	October–November	12 hr (day)	PM10 = 42.3	PM10 = 39	Branis et al., 2005
	Naturally ventilated			PM2.5 = 21.9
				PM1 = 13.7
Munich, Germany	Urban, residential	Winter	5 hr (class	PM10 = 105	PM10 = 38	Fromme et al., 2007
	Naturally ventilated		hours)	PM4 = 71.9
				PM2.5 = 23
		Summer		PM10 = 71.7	PM10 = 26
				PM4 = 44.8
				PM2.5 = 13.5
Athens, Greece	Urban	Winter	8 hr (class hours)	PM10 = 229 ± 182	PM10 = 166 ± 133	Diapouli et al., 2008
	Naturally ventilated			PM2.5 = 82 ± 56	PM2.5 = 56 ± 26
Antwerp, Belgium	Urban	Winter	12 hr	PM2.5 = 57 ± 10	PM2.5 = 53 ± 13	Stranger et al., 2008
	Naturally ventilated	Summer		PM2.5 = 61 ± 12	PM2.5 = 72 ± 12
Kozani, Greece	Suburban	September–October	12 hr	PM10 = 107	PM10 = 28	Triantafyllou et al., 2008
	Naturally ventilated
Delhi, India	Urban roadside	Non-winter	6 hr (class hours)	PM10 = 410.6 ± 196.4	PM10 = 162.4 ± 71.3	Goyal and Khare, 2009
	Naturally ventilated			PM2.5 = 71 ± 37.3	PM2.5 = 37.7 ± 22.6
				PM1 = 45.8 ± 26.5	PM1 = 27.9 ± 19.1
		Winter		PM10 = 1181.1 ± 578.3	PM10 = 531.1 ± 297.3
				PM2.5 = 359.9 ± 187.7	PM2.5 = 244.7 ± 190.5
				PM1 = 293.3 ± 180.3	PM1 = 221.7 ± 175.3
Xanthi, Greece	Suburban	October	24 hr	PM10 = 32–181	—	Gaidajis and Angelakoglou, 2009
	Naturally ventilated			PM2.5 = 25–144
Oporto, Portugal	Urban	February 2004–March 2005	8 hr	RSPM = 88 ± 34	RSPM= 20 ± 9.9	Madureira et al., 2009
	Naturally ventilated
Chhattisgarh, India	Suburban roadside	Summer, rainy, and winter	24 hr	RPM = 188.8± 43.9	RPM = 56.8 ± 9.32	Gadkari, 2010
	Naturally ventilated

Spatial variations in PM concentrations

The simultaneous measurements of PM mass concentrations in classrooms A and B were analyzed to understand the spatial variation at the study site. Results indicated that the PM concentrations (PM₁₀ = 149.79 μg/m³; PM_2.5 = 62.26 μg/m³; PM₁ = 29.41 μg/m³) inside classroom A, which is close to road side, were higher than the PM concentrations (PM₁₀ = 69.28 μg/m³; PM_2.5 = 49.36 μg/m³; PM₁ = 24.66 μg/m³) in the classroom B, which is away from the road (Figure 3). However, the diurnal variation of PM₁₀, PM_2.5, and PM₁ in classroom B were found to be similar to that of classroom A. Even though the PM₁₀ concentration depends on occupants’ activities (resuspension of particles due to occupants’ movements), a significant reduction of PM₁₀ concentration was observed in classroom B (54%) when compared with classroom A. This may be due to the fact that coarser particles contributed by vehicles (resuspension of road dust and abrasion of brake and tire) will not get transported to longer distances because of the higher settling velocities of those particles when compared with finer particles. The floor dust loading (for resuspension) in classroom B was found to be 247 g/day, which is less when compared with classroom A (450 g/day) because of the distance from the road. As a result, we noticed less PM₁₀ concentrations at classroom B, which is located far from the urban road. Morawska and Salthammer (2004) pointed out that there is relatively little information on the distribution of particles in the indoor settled dust. The floor dust loading in the present study was very high when compared with dust loading reported for office environments in Israel by Chudnovsky and Ben-Dor (2008). To the best of our knowledge, there are no studies reported about indoor settled dust in India. The indoor PM_2.5 and PM₁ at classroom B have also reduced about 21% and 14%, respectively. The location of monitoring station with respect to the adjacent road has been found to be a main factor affecting the PM concentrations at classroom B. When compared with previous studies, it was found that in most of the outdoor measurements, the PM concentrations showed sharp drop with distance away from roadway (Kinney et al., 2011; Levy et al., 2003). Kinney et al. (2011) observed that the average concentrations at the roadside (0 m) were more than 4 times higher than average concentrations 30 m away, and over 6 times higher than those 100 m away.

Figure 3. Box-Whisker plot of hourly average PM concentrations in different classrooms. The bottom of the box is the 25th percentile and the top is the 75th percentile. Vertical line inside the box is median. The whiskers extend to the highest and lowest observation. n = number of data points.

A Student’s t test (single-tailed) was also performed to determine whether the PM₁₀, PM_2.5, and PM₁ concentrations were significantly higher at near-field region compared with far-field region. The hypothesis employed was no difference between the two classrooms with respect to PM concentrations. Test was conducted for hourly average PM₁₀, PM_2.5, and PM₁ concentrations for all sampling days at 5% significance level. The estimated t values between paired data of classroom A and classroom B rejected the null hypothesis (PM₁₀ = 18.89, PM_2.5 = 9.24, PM₁ = 15.15; t_critical = 1.68). This confirms that the indoor PM concentrations were significantly higher at classroom near road compared with far-field region.

Comparison of PM concentrations with background site concentration

Figure 3 shows the simultaneous measurements of hourly average PM concentration in two schools. At background site (KV-IITM), the average indoor PM₁₀ concentrations were found to be 53.46 ± 24.21 and 35.02 ± 15.29 μg/m³, respectively, during working (08:00 a.m. to 4:00 p.m.) and nonworking (04:00 p.m. to 08:00 a.m.) hours. The average PM_2.5 and PM₁ concentrations were found to be 37.57 and 28.08 and 29.27 and 20.78 μg/m³, respectively, for working and nonworking hours. Comparison of indoor PM₁₀, PM_2.5, and PM₁ concentrations measured at classroom A (KV-CLRI) and C (KV-IITM) clearly indicated that PM concentration in classroom A was much higher when compared with classroom C at background site (Figure 3). Even though the PM₁₀ concentration depends on occupants’ activities, a significant reduction in PM₁₀ was observed in classroom C. This may be due to the less dust loading (for resuspension) in classroom C (182 g/day) when compared with classroom A (450 g/day). Further, classroom C is located in reserved forest area, which is free from vehicular pollution. The estimated t values between paired data of classroom A (KV-CLRI) and classroom C (KV-IITM) rejected the null hypothesis (PM₁₀ = 12.17, PM_2.5= 21.96, PM₁ = 16.82; t_critical =1.65). Janssen et al. (2001) and Lee and Chang (2000) also reported higher PM concentrations in school buildings located near traffic than schools located at industrial and rural areas.

Size distribution of indoor PM concentrations

The PM size distribution data measured at classroom A were grouped into two periods, i.e., working (08:00 a.m. to 04:00 p.m.) and nonworking (04:00 p.m. to 08:00 a.m.) hours. Figure 4 shows hourly average particle mass size distribution in classroom A of KV-CLRI school during working and nonworking hours. The upper and lower limits of the concentration error bars represent standard deviations of particle mass concentrations. The particulate size distribution was found to be bimodal, viz., coarse (>1 μm) and accumulation mode (0.1–1 μm). Modality of size distribution arises due to the particles coming from the different sources. The coarse mode was found to be around 3–4 μm and accumulation mode was found to be 0.3–0.4 μm. It can be also seen that there were differences between working and nonworking hours size distributions. Coarse mode was dominating during working hours. It confirms the resuspension of particles due to movement of occupants. A sharp drop in coarse mode was observed during nonworking hours. During this period, 65% of PM mass was less than 2 μm, whereas on working hours 38% of PM mass was less than 2 μm. The present study was comparable to a study conducted on indoor PM size distribution of suburban area of Prague (Smolík et al., 2008). These authors also observed that the mass size distributions were predominantly bimodal, with fine mode peaks at about 0.3 μm. The coarse mode was centered at about 5 μm, which was around 3–4 μm in the present study. Many previous PM size distribution studies conducted in indoor environments also reported an increase in fine particles with outdoor traffic and increase in coarse mode particles with occupant activities (Eileen et al., 2000; Tippayawong et al., 2009; Guo et al., 2010).

Figure 4. Hourly average PM size distribution in classroom A during working and nonworking hours.

Prediction of indoor PM concentrations

Table 4 summarizes the input parameters used for predicting PM concentrations inside naturally ventilated school building. It was observed that the indoor source rates for both PM₁₀ and PM_2.5 cover a wide range. The maximum S(t) values for PM₁₀ (120.8 μg/m³) and PM_2.5 (24.2 μg/m³) were observed during sweeping, i.e., at 4:00 p.m. The PM₁₀ and PM_2.5 deposited on the floor will get resuspended at a higher rate during sweeping. The minimum S(t) values for PM₁₀ and PM_2.5 were observed during night time. During nighttime, resuspension of PM may be occurring due to natural airflows entering through the vents and cracks in the classrooms. Goyal and Khare (2010) also reported a wide range of source generation terms for PM₁₀ and PM_2.5 in school building. Even though PM₁ is emitted from traffic, source generation (due to resuspension) for PM₁ was assumed as zero, because it has negligible settling velocity.

Table 4. Summary of input parameters used in the IAQ model

Infiltration Flow Rate, aM (ACH)	Mixing Factor, k	Pollutant Reactivity Factor, K (m/hr)	Indoor Source Generation, S(t) (μg/m^3)	Interior Surface Area, A (m^2)	Building Room Volume, V (m^3)
3.8–19.0	1	3.26 (PM10)	2.9–120.8 (PM10)	40.88	122.64
		1.10 (PM2.5)	0.16–24.2 (PM2.5)
		0 (PM1)	0 (PM1)

The scatter plots of measured and predicted indoor PM₁₀, PM_2.5, and PM₁ concentrations in the classroom are presented in Figure 5. The predicted hourly average PM₁₀, PM_2.5, and PM₁ concentrations were found to be 79.60 ± 57.50; 35.67 ± 25.03, and 25.09 ± 22.57 μg/m³. Performance of the model was evaluated on the basis of statistical indicators listed in Table 5. It can be seen that the results show good agreement between the predicted and measured indoor PM_2.5 (R² = 0.72–0.81) and PM₁ (R² = 0.81–0.87) concentrations, whereas the measured and predicted PM₁₀ showed poor correlation (R² = 0.17–0.23). The index of agreement for PM₁₀ indicates that 65% of predictions were error free, which is considered as satisfactory.

Figure 5. Scatter plot of 1-hr and 24-hr average measured and predicted PM₁₀, PM_2.5, and PM₁ concentrations in classroom A.

Table 5. Performance evaluation of IAQ model in classroom A

	Measured		Predicted		1-Hr Average			24-Hr Average
Pollutant	Mean (μg/m^3)	SD (μg/m^3)	Mean (μg/m^3)	SD (μg/m^3)	R^2	d	RMSE	R^2	d	RMSE
PM10	115.02	91.89	79.60	57.50	0.171	0.65	65.80	0.225	0.72	40.50
PM2.5	38.04	24.14	35.67	25.03	0.721	0.91	9.99	0.812	0.94	4.92
PM1	21.97	17.55	25.09	22.57	0.816	0.92	7.90	0.870	0.95	3.76

In general, the IAQ model underestimated indoor PM₁₀ concentrations mainly because of sudden increase in PM₁₀ concentration due to human activities (resuspension), which was not taken into account by the model. The model utilizes both measured outdoor PM₁₀ concentrations as well as predicted indoor PM₁₀ concentrations from the previous time step; hence, when the model has calculated the next concentration value, the prediction accuracy may be poor for a short burst of PM emission. Chaloulakou and Mavroidis (2002) also reported poor performance of the mass balance model where rapid concentration changes were observed. In the present study, PM₁ concentrations were overestimated by the IAQ model. Particle reactivity factors such as coagulation, condensation, evaporation, etc., were not included in the model, which may be a reason for overestimation of indoor PM₁ concentrations. It was documented that the semivolatile aerosols became volatilized more quickly and adheres to room surfaces when the particles were brought into an indoor environment, causing a significant net loss of PM mass (Lunden et al., 2003; Sawant et al., 2004; Saliba et al., 2009).

Conclusion

In this study, the indoor PM₁₀, PM_2.5, and PM₁ concentrations were monitored at an urban roadside and an urban background sites. The results indicated a higher indoor PM concentration in classroom close to urban road than at background site. Spatial variation of PM concentrations in classroom close to urban road and away from the road clearly indicated the influence of traffic emissions. Size distribution of PM in classroom indicated dominance of coarse mode during working hours and fine mode during nonworking hours. Increase in PM₁₀ concentration coincided with resuspension of particles due to occupants’ activities. The mass balance–based IAQ model results reveals that the model can predict fine PM concentration with better accuracy than coarse PM, mainly because the model fails to account for short-term resuspension phenomena due to occupants’ activities.

Acknowledgment

The authors express their sincere thanks to the students, teachers, and staffs of the Kendriya Vidyalaya School, CLRI and IIT campuses, Chennai, for their support and cooperation to carry out this study.

Funding

The authors wish to thank the Ministry of Environment and Forests, Government of India, New Delhi, for funding this study.

Additional information

Notes on contributors

V.S. Chithra

V.S. Chithra is a Ph.D. Research Scholar and S.M. Shiva Nagendra is an Associate Professor in the Department of Civil Engineering, Indian Institute of Technology Madras, Chennai, India.