Indoor Air Quality of Typical Malaysian Open-air Restaurants

This paper reports the indoor air quality state of typical open-air restaurants in Malaysia. The measured air pollutant parameters include respirable coarse particulate matter (PM10), carbon monoxide (CO) and microorganisms (bacteria and fungi). We determined the effects of occupancy, number of vehicles on nearby roads, temperature, wind speed and relative humidity on the indoor concentrations of PM10, CO and microorganisms. The indoor air quality of the restaurants tested was moderate, only in the 75th percentile, and the CO concentrations were slightly elevated indoors. Among the ambient parameters measured, only wind speed and temperature affect the PM10 concentrations. The indoor and outdoor values of wind speed and temperature were similar. We observed a strong positive correlation between the PM10 concentrations and concentration of airborne microorganisms. Further microbiological analyses showed that Gram-positive bacteria were abundant compared to Gram-negative bacteria. Gram-positive cocci (micrococci, streptococci, staphylococci and diplococci) were the dominant microbial morphologies, followed by pathogenic Gram-negative enterobacteriaceae and Gram-positive bacilli.


Introduction
Indoor air quality in working, dining and residential locations concerns a considerable number of people because they spend increasing fractions of their lives indoors (Lee & Chang, 2000;Zhao & Wu, 2007).Both scientific and public interest in indoor air pollutants have recently increased due to the negative impact of poor air quality on environmental and occupational health (Rajasekar & Balasubramanian, 2011).
Criteria air pollutants, such as particulate matter (PM) and carbon monoxide (CO), are important indicators of both indoor and outdoor air quality due to their harmful effects on human health.PM originates from a variety of sources and may contain particles with different shapes, sizes and physicochemical compositions (El-Fadel & Massoud, 2000).PM has been linked to cardiovascular and respiratory diseases (Wan Mahiyuddin et al., 2013).Coarse particulate matter (PM10) consists of particles with aerodynamic diameters between 2.5 and 10 µm.Indoors, this class of PM, also known as respirable particulate matter (RSP), originates from cooking and smoking activities (Wallace, 1996).Table 1 lists studies reporting particulate matter concentrations (including PM10 and PM2.5, which have particle aerodynamic sizes of < 2.5 µm) in different building types.CO is a colourless, tasteless and odourless gas generated from the incomplete combustion of fuel and biomass, e.g., in gas stoves, and it causes asphyxiation at high concentrations.Although not a criteria pollutant, airborne microorganisms, or bio-aerosols, are also an important air quality parameter (Reanprayoon & Yoonaiwong, 2012), especially in restaurants, where they can increase the risk of food contamination and may originate from occupants, microbial growth and organic waste (Fabian et al., 2005).In Malaysia and many other Asian countries, numerous restaurants and cafes are open or semi-open air buildings and often situated beside roads, factories and construction sites.The served food is commonly exposed to air, which could be contaminated due to poor indoor and outdoor air quality.Studies of indoor air quality in these dining establishments are scarce, and thus, no guidelines or best practices can be developed.This study attempts to investigate the indoor air quality state of three typical open-air restaurants by conducting an assessment of respirable coarse particulate matter (PM10), carbon monoxide (CO) and PM-laden microorganisms together with ambient temperature, wind speed, relative humidity, number of occupants and number of vehicles on nearby roads.
The objectives of this study are (1) to assess PM10 and CO concentrations at three open-air restaurants, (2) to quantify and identify airborne microorganisms and (3) to determine the relationships between the concentrations of PM, CO and airborne microorganisms and ambient (or surrounding) wind speed, temperature and relative humidity.

Site Description
Restaurant #1 (RF): Open-air Restaurant Far from a road This open-air canteen had 28 tables with four chairs each and seated a maximum of 112 occupants at a time.The restaurant consisted of a 26 m 2 kitchen and a 164 m 2 dining area, for a total area of 190 m 2 .The floor of the cafe was made of tiles; the layout of the cafe is illustrated in Figure 1  This 30-table Thai food restaurant was situated beside a busy four-lane highway.The restaurant had an area of 240 m 2 (60 m × 40 m) and was located 5 m from the road (Figure 1 (b)).This restaurant was also located near a car workshop.Restaurant 2 was open for 9 hr a day from 17:00 LST to 02:00 LST.For ventilation purposes, the restaurant employed five standing fans.The floors were made of cement.The restaurant employed ten workers, including waiters and cooks, and the kitchen contained four gas stoves for cooking.The kitchen was openly connected to the dining area.This restaurant sporadically used charcoal burners for seafood barbecues, but the main styles of cooking were open-wok frying and deep-frying.
Restaurant #3 (RNR): Open-air Restaurant Near a Road This restaurant had 70 tables with four chairs each, able to cater to a maximum of 280 occupants.The total area of the restaurant was 279 m 2 (Figure 1 (c)).The floor was made of tile, and for ventilation, there were 11 ceiling fans.The restaurant had ten full-time workers and remained open 24 hr daily.In the kitchen, there were five gas stoves, and the style of cooking was deep-frying.The kitchen was separated from the dining area by a wall.Food was served on a countertop with covers."Sheesha" (or water-pipes) were available only between 18:00 to 00:00 LST daily.

Measurements of Coarse Particulate Matter (PM10) and Carbon Monoxide (CO)
Dustmate (Turnkey Instruments, UK) was used to measure the concentrations of PM10 in the indoor air.Dustmate operates by continuously pulling in air samples through a nephelometer, a light-scattering based sensor.
The nephelometer analyses the particles passing through its laser beam to determine particle concentrations and relative size fractions.The particle size range of Dustmate is 0.5 -20 µm, and the detection limit of PM10 is 0.01 µg/m 3 .The Dustmate sampled air at a flow rate of 0.6 L/min.Readings were taken at an interval of 5 min over a total sampling period of 30 min for RF and 2 hr for RBR and RNR.
A Crowcon Tetra gas detector (Crowcon Detection Instruments, UK) was used to measure carbon monoxide (CO) concentrations in parts per million (ppm).The Crowcon Tetra can measure CO in the range 0 -500 ppm with a resolution of 0.1 ppm.
The Dustmate was stationed at a height of 1 m above the ground.This height was chosen to ensure the samples collected were representative of the air breathed by a seated person (Baek et al. 1997).At site RF, this was performed thrice a day at 08:00 LST (morning; occupants are present), 11:00 LST (late morning; occupants are present) and 18:00 LST (evening; cafe closed, occupants not present).At site RBR, measurements were taken from 21:00 LST to 23:00 LST and at site RNR, from 17:30 LST to 19:30 LST.All measurements were collected within four weeks from the initial sampling time for each location.Both indoor and outdoor measurements were taken, where the "indoor" location was a point in the centre of the restaurant, while the "outdoor" location was a point 2 m from the edge of the restaurant's boundaries.
A gravimetric measurement of PM10 using MiniVol Portable Air Sampler (Airmetrics, USA) was also employed to calibrate the Dustmate measurements.In this study, we used a size-selector impactor of < 10 μm with a sampling flow rate of 5 L/min for 2 hr.A 5-decimal analytical balance (OHAUS Discovery, USA) was used to measure the initial and final weight of Pallflex® fibre glass filter papers (Pall, USA).
Gravimetric-based PM measurements are more accurate than light-scattering based measurements (Cambra-López et al. 2012).We found that the relation between PM10 concentration using Dustmate (light-scattering) and MiniVol (gravimetric) is linear (Figure 2) and can be expressed by Equation (1).
where C is the PM10 concentration.Subsequent PM10 Dustmate results were corrected using Equation (1).removed from the incubator, and the growth patterns and morphologies of the cultured bacteria were studied to identify potentially pathogenic species.
Gram staining was carried out on the cultured bacteria according to the method described in Microbiological Applications (Benson, 2001) to determine bacteria morphology and Gram characteristics.Staining was repeated for each separate bacterial colony that was cultured.The glass slides with the bacterial smears were then examined under an immersion oil microscope (Nikon Eclipse E200, Japan) and reported based on their genus level and Gram characteristics.

Data and Statistical Analyses
Exploratory and statistical analyses were completed using RStudio© (R Development Core Team 2012).Sample populations were compared using t tests according to Welch's t test methodology.RStudio© was also used to conduct linear regression analyses and correlation analyses, and the Pearson correlation coefficients (1 > R > -1) between all parameters measured were calculated.A correlation coefficient of (R ≈ 1) indicates a strong positive relationship, whereas a correlation coefficient of (R ≈ -1) indicates a strong negative relationship.A value of (R ≈ 0) indicates a weak relationship.

Results
We found that the indoor median and mean PM10 (96 μg/m 3 and 120 μg/m 3 , respectively) and CO (2.7 ppm and 3.1 ppm, respectively) concentrations in the open-air restaurants studied were lower than the standards of Malaysia's Department of Occupational Safety and Health (DOSH) (8 hr average PM10 = 150 μg/m 3 ) and Department of Environment (24 hr average PM10 = 150 μg/m 3 ; 8 hr average CO = 10 ppm).For PM10, the 75th percentile (3rd quartile) concentration value was slightly above the PM10 limit (see Figure 3 (a) and Table 2).The World Health Organisation's (WHO) 2005 limit is 50 μg/m 3 for 24 hr for PM10 and 26 ppm for 1 hr for CO.Based on this stricter limit, the PM10 concentration in the open-air restaurants was unhealthy.The indoor CO concentrations were higher than the outdoor concentrations, but this result was not statistically significant (Welch's t test; p-value = 0.1093).
We found no relationship between the number of occupants and the PM10 and CO concentrations for any site or between number of vehicles counted on the road and PM10 and CO concentrations for any site, indoors and outdoors.
Only the indoor ambient parameters, such as wind speed and temperature, affect the PM10 concentrations (refer to Table 3).Wind speed and temperature were inversely correlated with PM10 concentrations for the indoor sites (p-value = 0.0420 and 0.0427, respectively).The CO concentration was not affected by ambient indoor or outdoor parameters.
For open-air restaurants, the indoor and outdoor temperatures, wind speeds and relative humidities can be considered identical (refer Figure 4).Both the indoor and outdoor temperatures were related to the relative humidity and wind speed, but the relative humidity was not related to the wind speed (refer Table 3 and Table 4).The number of occupants within the restaurant was positively correlated with relative humidity, but it was negatively correlated with temperature and wind speed.This study used averaging times of 15 min and 30 min; therefore, the results are not directly comparable to national and international standards, which use 8-hr and 24-hr averaging times.However, these are the only available standards that can be used to assess air quality; thus, they are applied here.Numerous studies still adopt this comparison (Baek et al., 1997;Lee et al., 2002;Lee et al., 2001).The exposure time of occupants to PM and CO in the restaurants is less than 8 hr or 24 hr; therefore, the measured values can be considered as peak values within the larger averaging time set.
Restaurants have higher indoor air pollutant concentrations compared with homes offices.However, the open design of the restaurant promotes air circulation, and thus indoor and outdoor concentrations of PM10 and CO are often similar.However, CO concentrations can be slightly elevated indoors due to local sources, such as gas stoves (Figure 3 (b)).A study on the indoor air quality of restaurants showed that heating and cooking (using kerosene/gas stoves) activities and inadequate ventilation worsen indoor air quality (Baek et al., 1997), particularly PM10 and PM2.5 levels (Lee et al., 2001).
In the surveyed restaurants, kerosene/gas stoves and background sources of PM10 could have overwhelmed any contribution from occupants.CO originates from the incomplete combustion of organic material and accumulates in closed environments; thus, no significant relationship is expected between number of occupants and the CO level.The correlation matrix tables (refer to Table 3 (indoor) and Table 4 (outdoor)) show that occupants were not the main source of PM10 in open-air restaurants.Cooking activities may be the predominant source, though coarse particulate matter can also originate from human activities such as walking (Qian et al. 2008), depending on the floor type (Zhang et al., 2008).Because the restaurants' occupants were mainly seated, human activities were not likely a large contributor to the increased PM10 concentrations.
Although cooking can significantly increase PM concentrations (Chao & Wong, 2002), no cooking occurred at RF in the mornings; thus, RF's morning PM10 concentrations were solely from smokers.Smoking is a major contributor to the PM10 concentrations in Hong Kong shopping malls (Li et al., 2001).Other studies have shown that cigarette smoke is a major source of PM (Wallace 1996), but in confined areas with low ventilation rates (McDonald et al., 2003), its dispersion is limited (Saha et al., 2012).In a study assessing the effects of a ban on smoking in restaurants, researchers discovered that the ban greatly reduced fine PM concentrations (Bohac et al., 2010).However, the dataset on number of smokers and PM concentrations is limited because only one to three smokers were present during measurements.
Lunchtime pollutant concentration measurements were high relative to other times of day, such as breakfast and dinner, due to high occupancy.Given the open-air nature of the restaurant, the ventilation rate is more important to dispersion for the restaurants studied than for a closed building.When construction materials are moved, they are exposed to wind, causing resuspension of coarser PM from the materials (Muleski et al., 2005).
The results obtained showed that wind speed and temperature had negative correlations to indoor PM10 concentrations and did not affect CO concentrations.A study has shown that naturally ventilated buildings have higher pollutant levels than air-conditioned buildings (Kukadia & Palmer, 1998) but do not retain external air pollutants for long periods because of their high air-change rates.Studies have also shown that concentrations of larger particulates, such as PM10, are correlated with ambient temperature (Celik & Kadi, 2007;Zaharim et al., 2007).Ventilation systems, outdoor air pollutant penetration factors, particulate deposition and resuspension rates and background air pollutants affect indoor air quality (Fromme et al., 2007).In most Asian countries, where buildings have no ventilation systems, indoor air pollution is higher than outdoor air pollution, particularly for buildings located near roads and/or construction sites (Montgomery & Kalman, 1989;Chao & Wong, 2002) because outdoor air pollutant sources can affect indoor air quality in addition to indoor sources, such as cooking, cleaning and smoking activities.
Total colony count was likely correlated with PM10 because the large surface area of course particles facilitates microorganism attachment.The bacteria sampling method employed is based on the theory of sedimentation and can therefore be biased towards larger particles.One limitation of this method is that only a fraction of the total bio-aerosols collected may be culture-able.Allergens have been detected on indoor soot particle surfaces (Ormstad et al., 1998).High concentrations of PM10 indicates that microorganisms are transported faster and in larger quantities per unit area.High bacteria counts could be caused by high occupancy, lack of proper sanitation and inadequate ventilation rates (Lee et al., 2002).The effect of temperature on airborne biological contaminants is minimal relative to ventilation and human activities indoors (Balasubraman et al., 2011).
Other studies have shown that Gram-positive bacteria are dominant in both indoor and outdoor environments (Aydogdu et al., 2010).Cocci were found to be prevalent in 100 large U.S. office buildings (Tsai & Macher, 2005).Gram-positive micrococci, staphylococci and streptococci are commonly found on human skin and mucus membranes.These organisms disperse and spread through direct skin-to-skin contact and are expelled through respiration (Chikere et al., 2008).The presence of cocci in the air of the restaurants implies overcrowding combined with inadequate ventilation (Awad, 2007).Gram-positive bacilli are associated with outdoor sources, such as soil emissions, water, dust, air, faeces, vegetation, wounds and abscesses (Aydogdu et al., 2010).They are found indoors as a result of transport by wind-blown particulates (Yassin and Almouqatea 2010).Although the majority of these species are considered harmless to human health, certain species can cause infections, especially in individuals with weak immune systems (Mckernan et al., 2008).Enterobacteriaceae are related to E. coli and Salmonella (Northcutt et al., 2004), which are associated with food poisoning (Hirsh & Martin, 1983).Possible sources of these pathogens include improper handling, serving and preparing of food.The potential presence of these species in restaurants is a risk to human health, which could easily increase with the number of occupants.

Conclusions
We surveyed three typical open-air Malaysian restaurants and found that the indoor air quality of these establishments was moderate based on the Department of Environment Malaysia and Department of Occupational Safety and Health standards, only exceeding standards in the 75th percentile of measurements.Indoor PM10 is influenced by wind speed and temperature.The CO concentrations were higher indoors than outdoors, but they were not correlated with temperature or wind speed, suggesting that CO originated from within the restaurants.Indoor ambient parameters, including wind speed, temperature and relative humidity, were similar to outdoor parameters.We found a strong positive relationship between PM10 and CFU per plate.The majority of bacteria detected were Gram-positive (84.3%) and non-pathogenic to humans, with minute percentages of Gram-negative bacteria, enterobacteriaceae (4.2%).The presence of airborne bacteria indicated overcrowding and inadequate ventilation within the open-air restaurants.Wind speed is the sole ambient parameter to consider when trying to improve indoor air quality in open-air restaurants.

Table 1 .
Respirable particulate matter (PM10 and PM2.5) concentrations in different building types from published literature (n is number of data points)

Table 2 .
Range of indoor and outdoor air PM10 and CO concentrations collected at three open-air restaurants; n is number of data points RBR -Open-air Restaurant beside a Road (averaging time 15 min) RNR -Open-air Restaurant near a Road (averaging time 15 min) LOD -Lower than Detection limit (detection limit of 0.1 ppm)

Table 3 .
Indoor Pearson correlation matrix of air quality parameters measured for two open-air restaurants: RBR (open-air Restaurant beside a Road) and RNR (open-air Restaurant near a Road); bolded values are correlated

Table 4 .
Outdoor Pearson correlation matrix of air quality parameters measured for two open-air restaurants: RBR (open-air Restaurant beside a Road) and RNR (open-air Restaurant near a Road); bolded values are correlated; note that no correlations are statistically significant