Indoor PM_2.5 in Santiago, Chile, spring 2012: Source apportionment and outdoor contributions

Highlights

• First source apportionment of indoor PM_2.5 conducted at Santiago, Chile.

• Outdoor and indoor sources each contribute half of the measured indoor PM_2.5.

• Traffic and indoor cooking are the strongest sources of indoor PM_2.5.

• Indoor concentrations of PM_2.5 were affected by socioeconomic status.

Abstract

Indoor and outdoor PM_2.5 sampling campaigns were carried out at Santiago, Chile (6 million inhabitants, 33.5°S, 70.6°W) in spring 2012. A pair of samplers was placed inside each household studied and an additional pair of samplers was placed at a fixed outdoor location for measuring trace elements and elemental (EC) and organic carbon (OC) in Teflon and quartz filters, respectively. A total of 47 households in downtown Santiago were included in this study. Mean outdoor and indoor PM_2.5 concentrations were 19.2 and 21.6 μg/m³, respectively. Indoor concentrations of PM_2.5 were affected by socioeconomic status (p = 0.048) but no such evidence was found for PM_2.5 species, except lead (p = 0.046). Estimated species infiltration factors were 0.70 (±0.19), 0.98 (±0.21), 0.80 (±0.12) and 0.80 (±0.03) for PM_2.5, OC, EC and sulfur, respectively. Estimated household infiltration factors had a median of 0.75, mean of 0.78, standard deviation of 0.18 and interquartile range (IQR) 0.67–0.86.

For the very first time, Positive Matrix Factorization (PMF3) was applied to an indoor PM_2.5 chemical composition data set measured at Santiago. Source identification was carried out by inspection of key species and by comparison with published source profiles; six sources were identified. Three of them were outdoor contributions: motor vehicles with 5.6 (±0.7) μg/m³, street dust with 2.9 (±0.5) μg/m³ and secondary sulfates with 3.4 (±0.5) μg/m³. The indoor sources were: indoor dust with 1.6 (±0.3) μg/m³, cleaning and cooking with 2.3 (±0.3) μg/m³ and cooking and environmental tobacco smoke with 6.1 (±0.7) μg/m³. There is potential for further reducing PM_2.5 population exposure in the short term —by improving ventilation of indoor air and controlling indoor sources — and in the long term — with filtration of outdoor air and household improvements to reduce air change rates.

Introduction

Indoor suspended particulate matter (PM) consists of ambient particles that infiltrate indoors and remain suspended, particles emitted indoors (primary), and sometimes particles formed indoors (secondary) through reactions of gas-phase precursors emitted both indoors and outdoors (Weschler and Shields, 1997). Outdoor particles can enter indoor environments by convection (through an open window or by the air conditioning system) or by infiltration through cracks and fissures in the housing envelope. These two combined mechanisms determine the residence time of air — or its reciprocal, the air exchange rate, AER — within a household. Outdoor PM_2.5 can be a significant contributor to indoor particle concentrations, especially when AERs are high (Abt et al., 2000, Meng et al., 2005a). When indoor sources are present, indoor PM concentrations can be substantially higher than outdoor PM concentrations (Ruiz et al., 2010, Zhang et al., 2010). Indoor anthropogenic PM sources include smoking, cooking, unvented space heaters, cleaning, washing and walking (Chao et al., 1998, Abt et al., 2000, Zhao et al., 2006, Zhao et al., 2007, Abdullahi et al., 2013).

Several mechanistic and statistical models have been applied to quantitatively describe factors modifying indoor PM. There are two approaches that have been applied: i) models that assume a well-mixed indoor air volume, like the steady state, single-zone mass balance (Abt et al., 2000, Ott et al., 2000), and ii) models that regard each household as several compartments (rooms) which may or may not be well-connected depending on when and how internal doors are opened. Both approaches are described in the following paragraphs.

The single zone mass balance model describes households as completely mixed flow reactors, where the indoor PM concentration depends on the outdoor PM concentration in the following way (Abt et al., 2000, Meng et al., 2005a):

$$C_{\text{I}}=\frac{Pa{C}_{\text{O}}}{a+k}+\frac{Q_{\text{I}}}{V(a+k)}\equiv F_{\text{INF}}{C}_{\text{O}}+C_{\text{IG}}$$

where C_I and C_O are the measured indoor and outdoor PM_2.5 (μg/m³), respectively, P is the penetration coefficient (dimensionless), a is the household air exchange rate (AER, h⁻¹), Q_I the rate of indoor generation and resuspension of PM (μg/h), k the rate of removal of PM by reaction or surface deposition (h⁻¹), and V the household volume (m³). The term Pa/(a + k) is called the infiltration factor (F_INF) and it quantifies the fraction of C_O that is found indoors; C_IG accounts for indoor-generated concentration. Contributions of outdoor sources to indoor PM_2.5 concentrations of 23–67% have been estimated in previous studies (Abt et al., 2000, Meng et al., 2005a).

The Random Component Superposition (RCS) statistical model (Ott et al., 2000) uses the linear regression of indoor on the outdoor PM concentration — equivalent to Equation (1)— to estimate means and distributions of the outdoor and indoor contributions to indoor PM concentrations. Other approaches include multivariate linear regression and receptor models (Abt et al., 2000, Meng et al., 2007, Zhao et al., 2006, Zhao et al., 2007).

Several approaches have been proposed to model the dynamics of indoor concentrations due to spatial and temporal variation of indoor sources, internal and external window and door opening activities, occupant's behavior, presence of air conditioning systems, etc. (Klepeis and Nazaroff, 2006, Sohn et al., 2007, Du et al., 2012, Fabian et al., 2012, McGrath et al., 2014). In this approach each housing unit is regarded as a set of rooms connected by internal doors or windows.

Briefly, multi zone models capture spatial and temporal variations in indoor pollutants that provide more information that the single-zone model. However the resources needed to apply those models are substantially higher: AERs need to be measured for each room and for different combinations of window and door positions, occupants' behavior and emission sources need to be resolved with high temporal and spatial detail and some physical parameters — deposition losses — need to be validated using actual measurements (McGrath et al., 2014).

The greater metropolitan area of Santiago, Chile (33.5°S, 70.7°W) is the sixth largest South American city in population (6 million inhabitants). Ambient PM_2.5 has been measured at Santiago since 1989. Despite steady economic growth, ambient PM_2.5 concentrations have continuously decreased in the last 24 years (Jorquera et al., 2004, Koutrakis et al., 2005, Sax et al., 2007). Nonetheless, PM_2.5 ambient concentrations still exceed the World Health Organization (WHO) daily and annual guidelines of 25 and 10 μg/m³, respectively (WHO, 2006). Fig. 1 shows the evolution of ambient PM_2.5 at four monitoring sites — from TEOM data, uncorrected for volatile losses. The decrease in ambient PM_2.5 across the city in the last years is ascribed to improvements of the public transportation system at Santiago, initiated in 2007.

Several studies have shown how the composition of outdoor PM_2.5 has evolved in the last years (Artaxo et al., 1999, Koutrakis et al., 2005, Sax et al., 2007, Jorquera and Barraza, 2012) with decreasing trends in sulfur, lead and other anthropogenic elements. Few studies have characterized indoor PM_2.5 at Santiago; below we summarize them.

Rojas-Bracho et al. (2002) measured personal, indoor and outdoor PM_2.5 concentrations for school children in Santiago's central and NE areas. Each participant carried a personal sampler, while Harvard Impactors (Marple et al., 1987), located in their homes, simultaneously collected 24 h samples in the winter of 1999 (N = 20). They found a slope (F_INF) of 0.61 (p = 0.0001) and an intercept (C_IG) of 18.9 μg/m³ (p < 0.0001) with R² = 0.54; the median I/O ratio was 0.95 and 10% of ratios were above 1.6.

Ruiz et al. (2010) measured 48 h indoor and outdoor concentrations of PM_2.5 and its chemical components in winter 2007 at downtown and NE areas in Santiago. A total of 16 households were measured — 13 apartments and 3 homes — to estimate contributions of unvented space heaters to indoor air pollution. Average outdoor PM_2.5 was 55.9 μg/m³, and average indoor PM_2.5 varied according to the type of space heater used. They found slope (F_INF) values of 0.64, 0.94, 0.66, 0.63 and 0.66 for PM_2.5, EC, OC, S and Al, respectively. The contributions of kerosene and LPG heaters to indoor PM_2.5 were 44.2 and 18.6 μg/m³, respectively. Elemental carbon (EC) was only generated by kerosene heaters with an average of 9.3 μg/m³ of PM_2.5 in those households. Organic carbon (OC) was generated by kerosene and LPG heaters in concentrations of 6.5 and 4.9 μg/m³, respectively.

Burgos et al. (2013) compared indoor and outdoor PM_2.5 in the west side of Santiago where families were relocated from slums to public housing apartments. The campaign was conducted in winter 2009 and 71 slum units and 98 public housing apartments were measured. Average 24 h indoor PM_2.5 concentrations were 55.7 and 77.8 μg/m³, and ratios of average I/O values were 1.08 and 1.18 for public housing and slums, respectively. They estimated an infiltration factor of 0.5 (±0.1) and the following contributions to indoor 24 h PM_2.5 concentrations (in μg/m³): allocation to public housing was −10.4 (±5.1), smoking more than 3 cigarettes was 29 (±11), using biofuels was 25.6 (±10.0), and presence of an infant was −9.5 (±4.6). Negative numbers represent negative contributors to these levels.

All these studies were conducted in the winter season when residential heating is active and households restrict opening windows and doors to minimize heat losses; therefore, air exchange rates are kept to their minimum possible values.

The goal of the present study was to conduct a source apportionment of indoor PM_2.5 at Santiago, Chile to identify the major contributing sources. The campaign was conducted in springtime when residential heating is turned off and households are more ventilated.