Emission Rate of Particulate Matter and Its Removal Efficiency by Precipitators in Under-Fired Charbroiling Restaurants

In order to explore the potent role of meat cooking processes as the emission sources of particulate matter (PM), emission rates and the associated removal efficiency by precipitators were estimated based on the on-site measurements made at five under-fired charbroiling (UFC) restaurants. The emission patterns of PM for these five restaurants were compared after having been sorted into the main meat types used for cooking: beef (B), chicken (C), intestines (I), and pork (P: two sites). The mass concentrations (μg m-3) of three PM fractions (PM2.5/PM10/TSP) measured from these restaurants were 15,510/15,701/17,175 (C); 8,525/10,760/12,676 (B); 11,027/13,249/13,488 (P); and 22,409/22,412/22,414 (I). Emission factors (g kg-1) for those PM fractions were also estimated as 3.23/4.08/4.80 (B), 3.07/3.82/3.87 (P), 8.12/8.22/8.99 (C), and 6.59/6.59/6.59 (I). If the annual emission rate of PM10 is extrapolated by combining its emission factor, population, activity factor, etc., it is estimated as 500 ton year-1, which corresponds to 2.4% of the PM10 budget in Seoul, Korea. Removal efficiencies of PM10 via precipitators, such as an electrostatic precipitator (ESP), bag filter (BF), and the combination system (ESP + catalyst), installed in those UFC restaurants ranged between 54.76 and 98.98%. The removal efficiency of PM by this control system was the least effective for particles with <0.4 μm, although those in the range of 0.4–10 μm were the most effective.


INTRODUCTION
Certain cooking processes are now known to release undesirable pollutants into the atmosphere. In this respect, meat cooking processes are suspected as one of the major emission sources of particulate matter (PM) and associated compounds, e.g., secondary organic aerosols (SOA), organic carbon (OC), elemental carbon (EC), etc. [1,2]. The potent role of under-fired charbroiling (UFC) restaurants has been, in fact, recognized as the greatest emitter of PM and volatile organic compounds (VOCs) [3]. Schauer et al. [1] showed that emissions from meat charbroiling and frying can account for about 20% of all fine organic

The Selection of Target Restaurants
In order to select the target restaurants for study, we conducted an initial survey of all UFC restaurants registered in the city administrative office of Seoul as of 2008. The results showed a total of 10,103 UFC restaurants in its boundary, among which 1,649 can be classified as large-scale UFC restaurants (i.e., the area exceeding 100 m 2 ). The fuels used in those large-size restaurants consisted of charcoal (960 sites), liquefied petroleum gas (669), wood (11), briquette (7), and straw (2). Out of those 1,649 restaurants, 14 UFC restaurants were identified as having pollution control systems. Table 1 shows the list of five UFC restaurants selected randomly (out of 14) for this investigation that are equipped with precipitators such as an ESP (three sites; 4,000-10,000 m 3 h -1 ), ESP + catalyst (one site; 3,000 m 3 h -1 ), and BF (one site; 4,800 m 3 h -1 ). All of the five selected restaurants used Korean charcoal as the main fuel. The main meat types cooked for charbroiling at these restaurants were chicken (C) (one site), beef (B) (one site), pork (P) (two sites), and intestines (I) (one site). The collection of PM precipitators was made through the hood and duct system. The hood and duct systems at each restaurant were installed to cover as many as 12-30 table sets for each group of dining teams (up to four persons for each table).

Sampling and Analysis
In order to measure the concentrations of PM and to develop its emission factors, the procedures described in the EPA method 201A were applied with a slight modification for this study. The EPA method 201A is the in-stack measurement of PM equal to or less than an aerodynamic diameter of nominally 10 (PM 10 ) and 2.5 µm (PM 2.5 ), along with total suspended particle (TSP). To comply with the sampling train of method 201A, the PM 2.5 cyclone is inserted between the PM 10 cyclone and the filter. The PM 10 cyclone collects the fraction greater than PM 10 , while the PM 2.5 cyclone collects the PM fractions between 2.5 and 10 μm. Thus, TSP is collected on the front side of the PM 10 cyclone, while PM 10 and PM 2.5 are collected on the in-line sampling cylinder (between the two cyclones) and on the in-line filter (Quartz), respectively. For the isokinetic collection of in-stack samples, isokinetic coefficients were maintained at a constant value of 95-110%, which is recommended by the Korean standard testing method. During our field campaigns, the collection of the PM samples from each facility was carried out one time between 19:00 and 21:00 in light of intensive cooking activities in that time band. More details for sampling conditions in facilities, such as sampling date, sampling time, suction flow rate, etc., are described in Table 2. Filters were placed in a weighing chamber over 24 h, both before and after sampling, and weighted with a microbalance (UMX2, Mettler Toledo Inc., SWI). The removal efficiency of PM is affected by several governing factors, including concentration, size distribution, etc. Therefore, removal efficiencies can be estimated in relation to the particle size distribution. To estimate the removal efficiency of PM, the optical aerosol spectrometer (OAS: Grimm 1.109, Grimm Aerosol Technique, Germany) equipped with isokinetic sampling probes (Grimm 1.152, Grimm Aerosol Technique, Germany) was employed to count PM levels, both up-and downstream of the precipitator. This OAS measures the particle count with 31 channels (diameter range: 0.25-32 μm) on a real-time basis at every 6 sec. As there is no international standard for the calibration of the OAS, its calibration was made by referring to ISO 12103 A3 (intercomparison method). As part of the calibration, a statistical comparison was made between this analyzer and a precalculated traceable reference unit (mother unit). The precalculation step was carried out with a combination of different monodisperse Poly Styrene Latex traceable standard particles (Duke Scientific Inc., U.S.) and standard equipment of LAS-X aerosol particle counter (PMS Inc., U.S.). The statistical comparison is then made against mean values of the particle count at each channel. This calibration is repeated automatically by a computer software that adjusts the data with the accuracy of ±2%.

Theoretic Basis
Emission factor (EF) is typically defined as the amount of a model pollutant emitted per unit mass of fuel burned [16]. It is often expressed in terms of mass-based EF with a mass/mass ratio unit (e.g., g kg -1 ). The concept of EF is generally used for the calculation of emission inventory. In this study, we adopted the procedures of Lee et al. [17] to apply mass-based EF.
where C and Q are the concentration of PM and the flow rate of air (passing through the duct), respectively. The latter term is calculated by the equation of V (velocity)  A (area of inlet duct) over 30 min of sampling time. Wt is the weight of charbroiled meat.
The emission rates for each meat type were then computed by the following equations: where k is activity factor, EF is emission factor of PM, CR is consumption rate of each meat type, and P is the population. The individual parameters, such as k, CR, and P, were derived for the entire city of Seoul as of 2008 [18,19,20].
Removal efficiencies of PM by all the precipitators were also evaluated as follows: where PM(i) and PM(o) denote the particle number concentration (counts) for up-(inlet) and downstream (outlet), respectively. Table 3 shows the mass concentrations of all PM fractions measured at each of the five UFC restaurants. The concentrations (μg m -3 ) of three particle fractions (PM 2.5 , PM 10   indicate that the UFC restaurant cooking with large intestine was the most distinct contributor of all particle fractions (PM 2.5 , PM 10 , and TSP) in terms of mass concentrations. In order to assess the effect of natural ventilation on PM concentration, the ratios between the indoor and outdoor PM concentrations had to be evaluated. These values indicate how well the building ventilation system disperses the indoor-generated PM and thus affects the outdoor PM concentration [21]. To this end, the outdoor concentrations of PM in this study were represented by those samples collected separately through the hood and duct system. In a previous report by the EPA[9], particle size distribution was found to be quite comparable between different meat types, while PM 2.5 and PM 10 constituted 80 and 20%, respectively. Welch and Norbeck [22] also confirmed that a majority fraction of particles was PM 2.5 on the basis of the measurements made at 18 commercial cooking restaurants. Furthermore, Lee et al. [17] showed that the accumulated mass concentrations of PM 10 from the UFC process were 92.2-99.5%, with their dominant fraction in particle diameter range of 2.0-2.5 μm. Park et al. [23] also reported that the median diameter of PM released from meat cooking was 2.4 μm Therefore, it is reasonable to infer that PM released from UFC restaurants should primarily consist of PM 2.5 .

Emission Factors
We developed emission factors of PM based on in situ data collected during real-life operations at all selected restaurants. The results of our measurements made at the UFC restaurants were then used for the estimation of emission inventory of PM. Table 2 summarizes the sampling conditions of PM at each restaurant, e.g., measured velocity, flow rate, sampling time, and weight of meats charbroiled (as parameters for calculating emission factors). The ranges of air velocity and flow rate at each restaurant were 4.06-6.72 m sec -1 and 0.4-1.6 m 3 sec -1 (1,440-5,760 m 3 h -1 ), respectively ( Table 2). The suction flow rates for dining tables connected to a hood and duct system were 120-384 m 3 h -1 . From all five restaurants, sampling of PM was made for the duration of 30 min. Information concerning the total quantity of charbroiled meats was obtained from the manager of each restaurant before investigation. The weights of charbroiled meat samples for the test were 3.50-9.18 kg for the five UFC restaurants.
Our emission factors for UFC restaurants using charcoal fuels were compared with those quantified previously in three separate studies ( Table 4). All of these emission factors were computed by considering the contribution of both meat and charcoal fractions. However, emission factors of the charcoal were not divided separately, as the relative contribution is trivial under most circumstances. The EPA [9] reported that charcoal did not contribute significantly to PM emission levels. In addition, Lee et al. [17] reported that emission factors of charcoal were 0.01-0.02 (PM 2.5 ), 0.01-0.03 (PM 10 ), and 0.02-0.05 g kg -1 (TSP) on the basis of the experimental results derived by the hood method. The basic concept of this method is to equip a hood and exhaust duct above the cook stove to collect all emissions [16]. In this study, PM 2.5 /PM 10 /TSP emission factors (g kg -1 ) of C, B, P, and I were 8.12/8.22/8.99, 3.23/4.08/4.80, 3.07/3.82/3.87, and 6.59/6.59/6.59, respectively. Relative ordering in emission factors were, in some senses, consistent between B and P across different particle fractions. It is also interesting to note that emission factor values of C were the largest in both particle fractions. In addition, all emission factors developed in this study were low relative to the previous studies (Table 4). It is suspected that all these differences should be ascribed at least partially to different sampling approaches. Note that the basic sampling methods of PM can be classified into two categories: sourcelevel sampling and ambient-level sampling. Source-level sampling, equivalent to the in-stack sampling method (EPA method 201A), was applied in this study. In contrast, ambient-level sampling techniques, employed in previous studies [3,9,24], required a very wide range of space and high costs. The general sampling and analytical principles employed in the analysis of concentration, size, and composition of PM are comparable between the two types of measurement approaches. However, each specific approach differs greatly due to the differences in the prevalent environmental conditions (temperature, concentration, etc.) of sampling [25]. In the case of the PM measurements from stationary sources, sourcelevel sampling is currently recommended in the context of regulatory enforcement, Korean standard testing method [26]. In contrast, ambient-level sampling based on the dilution chamber method is accepted for PM measurement from mobile sources.

Emission Rates
To estimate emission rates of each PM fraction (PM 2.5 /PM 10 /TSP), the parameters for calculating emission rates were derived from a number of previous reports: (1) consumption ratio (CR)[18], (2) activity factor (k) [18,19], and (3) population (P) of Seoul [20]. However, the CR reported in the survey by the KMTA[18] involved the total amount of meat consumed both in all places of meat consumption (e.g., wholesale market, meat processing factory, restaurants, etc.) and by all cooking methods (e.g., charbroiling, soup, stew, frying, etc.). For an accurate computation of the consumption ratio (CR; kg person -1 year -1 ) for each meat type in this study, activity factors (k) for only UFC were taken into consideration. This "k" value was expressed as the proportion (%) representing specific activities. In order to derive the "k" value, we attempted to document both (1) the relative proportion (%) of three meat types (B, C, and P) consumed in restaurants among all the consumption pathways and (2) the ratio of those meats processed via the UFC method among all the cooking approaches. Both activities were applicable to restaurants cooking meat types of B, P, and C. The KMTA [18] reported that the former activities (%) for each of the three meat types were 25.7, 21.7, and 37.0, respectively. In addition, the NIER [19] reported that the latter activities (%) for those three meat types were 61.6, 46.0, and 41.7, respectively, among 1,124 restaurants in Korea. Consequently, their activity factors were calculated as 0.158, 0.100, and 0.154, respectively. Table 5 summarizes activities and the corresponding activity factors. If one assumes that UFC restaurants cook only those three meat types (B,P, and C), their emission rates of PM 2.5 , PM 10 , and TSP are computed as 214, 240, and 260 ton year -1 , respectively. (Refer to the column of subtotal values in Table 6.) Among 10,103 UFC restaurants in Seoul, the relative proportion of restaurants cooking those three meat types was 48%, as shown in Fig. 1. Hence, through an extrapolation of our experimental results for all the total UFC restaurants in Seoul, their emission rates of the three PM fractions in Seoul are estimated to be 446, 500, and 542 ton year -1 , respectively.

Contribution Ratio of PM 10
In order to assess the relative contribution of UFC restaurants to atmospheric PM levels in Seoul (as of 2008), the latest emission ratio data (as of 2007) for all available air pollutants in Seoul were obtained [27]. As shown in Fig. 1B

Removal Efficiency of PM
The removal efficiencies of each PM fraction were estimated and compared between different types of precipitators (Table 7). Most importantly, the collection efficiency of precipitators is affected very sensitively by particle size distribution. Therefore, the removal efficiencies of PM 2.5 , PM 10 , and TSP (by precipitator) were calculated by integrating the particle count data for each particle size. The integrated particle count data obtained by the OSA were arbitrarily classified into (1)   Particle count data were integrated in the ranges of (1) 0.4-2.5 μm (the first stage: 11 channels), (2) 2.5-10 μm (the second stage: eight channels), and (3) 10-32 μm (the third stage: seven channels). These data were then applied to the computation of removal efficiencies for each PM fraction of PM2.5, PM10, and TSP, respectively.
The removal efficiency of both ESP and BF can be theoretically >99% for all size fractions of PM [28]. However, all the facilities investigated in this study showed values lower than such theoretical reference values. We suspect that there were many interfering factors towards the removal efficiency, including the alterations in operating conditions and maintenance of each PM precipitator, types of meats charbroiled, test conditions, etc. In addition, oil mist and vapor generated from fat droplets can be an obstacle that gradually decreases the removal efficiency. They can be accumulated easily on pipelines and walls of precipitators via coagulation and solidification, when oil mist and vapor fly through the precipitators. Furthermore, fat droplets are vaporized or splattered into the smoke aerosol because the liquid grease droplets melt from fat deposits due to heating [29]. Fig. 2 shows the removal efficiency of PM according to particle diameter distribution (31 channels) measured by an OAS. PM <0.4 μm were removed with the least efficiency, while PMs with a diameter <2.5 μm were removed almost completely. In the case of restaurants A, B, and E, the removal efficiency of PM was fairly stable in the overall particle size distribution. However, those of restaurants C and E showed different patterns. The removal efficiency of the latter two was relatively low at PM about <1 μm and >20 μm, respectively.

CONCLUSION
In this research, the concentrations of PM 2.5 , PM 10 , and TSP were measured from five UFC restaurants, and their concentration data were used to derive the removal efficiency between different control facilities. The mean mass concentrations (μg m -3 ) of three PM fractions (PM 2.5 /PM 10 /TSP) measured from five restaurants were 15,510/15,701/17,175 (C); 8,525/10,760/12,676 (B); 5,168/7,084/7,084 (P1); 16,886/19,413/19,891 (P2); and 22,409/22,412/22,414 (I). A brief glance over the PM data indicates that the concentrations of large intestine tend to peak compared to other meat types. Furthermore, PM released from UFC restaurants in this study was mainly composed of PM 2.5 ; 0.73-1.00 (PM 2.5 /PM 10 ) and 0.85-1.00 (PM 10 /TSP). As a result, the control of PM 2.5 appears to be very important, as its dominant fraction can otherwise be exposed to the customer. Considering the fact that PM 2.5 was the major component of PM emissions in UFC restaurants, one needs to establish the control strategy of PM 2.5 , as an a priori target.
In order to estimate emission rates of PM 2.5 , PM 10 , and TSP, their emission factors were developed for each UFC restaurant investigated in this study. Emission factors of C were the largest in all particle fractions. According to our calculations, the emission rates of PM 2.5 , PM 10 , and TSP were estimated to be 446, 500, and 542 ton year -1 in Seoul, respectively. Furthermore, if the contribution of UFC restaurants is extrapolated from our study, it is estimated to comprise 2.4% of PM 10 emissions from all different sources in Seoul. Removal efficiencies of PM against the major control facilities operated at each restaurant were estimated to be in the range of 54.62-98.98% (PM 2.5 ), 54.76-98.98% (PM 10 ), and 89.61-99.96% (TSP). In addition, PM with a diameter <10 μm was removed almost completely by the control facilities at most restaurants.
This study was undertaken to describe the basic features of PM emissions from UFC restaurants in Seoul on the basis of PM data measured at each size fraction. The measurement data were used to estimate their emission rates and removal efficiencies against precipitators. The results of our study should be helpful in establishing a database for regional and national emission inventories of PM from UFC restaurants, and to offer appropriate strategies to control air quality in those public facilities.