Impacts , e long-term impacts of PM 10 and PM 2 . 5 particles from construction works on surrounding areas †

Department of Civil and Environmental E Physical Sciences, University of Surrey, Gui surrey.ac.uk; Prashant.Kumar@cantab.net; 1483 682762 Environmental Flow (EnFlo) Research Cen Sciences, University of Surrey, Guildford GU MRC PHE Centre for Environment and He 9NH, UK † Electronic supplementary information S1–S4 and Fig. S1–S5. See DOI: 10.1039/c5 Cite this: Environ. Sci.: Processes Impacts, 2016, 18, 208


Introduction
Construction developments in both the developing and developed world are common. However, the impact of particulate matter (PM) emitted in the coarse (PM 2.5-10 ; between 2.5 and 10 mm) and ne (PM 2.5 # 2.5 mm) particle size range from such activities on the surrounding areas is poorly understood. Construction and demolition of structures is known to result in higher local concentrations of PM 10 , which contains a wide variety of toxic organic substances and may adversely affect the respiratory health of nearby residents. [1][2][3][4][5] There is also an increased interest in PM 2.5 because it penetrates deeper into the lungs and is of even greater concern for human health. [6][7][8] For this reason, exposure to PM 2.5 is globally the 9th most powerful risk factor for disease burden. 9,10 Until recently, only limited study has focused on the exposure to PM 10 and even less research on exposure to PM 2.5 fractions arising from outdoor construction activities and understanding their potential impact on local air quality (Table 1).
Besides construction activities, PM 10 concentrations are also affected by the emissions arising from local fugitive sources such as road works, [11][12][13][14][15] vehicle exhaust [16][17][18][19][20] and non-vehicle exhaust sources. [21][22][23][24][25][26] At the same time, many activities associated with air and sea transportation produce particles across the range of PM 10 and PM 2.5 . [27][28][29] A few studies have investigated the PM 10 emissions arising from industrial sources such as waste transfer stations. 30 There are also a few studies concerned with PM 10 emissions arising from outdoor construction activities. [31][32][33][34][35] However, there is still very little work focused on PM 2.5 fractions arising from construction activities. 36 The importance of particle exposure from construction sources is expected to increase with the ever growing world population. 37,38 In addition to concerns associated with the short-term exposure to airborne PM at the time of construction activities, there is also the potential for long-term exposure to PM that settles across the nearby community, which is then available for inhalation or ingestion aer resuspension. 32,39,40 The European Union 41 set the targets to limit the daily and annual mean values of PM 10 at a European-wide level for the years 2004 and 2010. 13 The legal limit by 2005 was to achieve a daily mean PM 10 concentration of 50 mg m À3 , not exceeded on more than 35 occasions per year and annual mean values of 40 mg m À3 . Moreover, the target by 2010 was to achieve a daily mean PM 10 concentration of 50 mg m À3 , not exceeded on more than 7 occasions per year and annual mean concentrations of 20 mg m À3 . These target values, to be met by 2010, were not carried forward in Directive 2008/50/EC. 42 Fuller and Green 13 noted that the PM 10 emissions generated by building and road works in and around London breached the EU limits for the daily mean PM 10 concentrations (50 mg m À3 ) on several occasions. In this study, a series of PM 10 and PM 2.5 measurements at 17 monitoring stations around construction sites were carried out during 2002-2013 to assess their impact on the air quality in the surrounding areas.

Description of the construction sites
Measurements were carried out around three outdoor construction sites, which are referred hereaer as CS 1 , CS 2 and CS 3 . CS 1 , CS 2 and CS 3 covered an area of about 260 Â 10 4 , 54 Â 10 4 and 3 Â 10 4 m 2 , respectively ( Fig. 1). There were 17 monitoring stations (i.e. MS 1 -MS 17 ) around these three outdoor construction sites (CS 1 , CS 2 and CS 3 ), which represent a diverse range of construction activities. The locations of the monitoring stations around these sites are shown in Fig. 1, but the specic details about the location have been kept anonymous for the protection of condential information.

Field measurements
Continuous air quality monitoring was carried out at 17 different monitoring stations around three construction sites to measure the concentration of PM 10 and PM 2.5 . The measurements of PM concentrations analysed in this study were during the periods of construction and there were no similar measurements made before and/or aer the construction works. Measurements were undertaken continuously and divided into working hours (referred to as working period) in weekdays between 08:00 and 18:00 h (local time) and nonworking hours (referred to as non-working period), which covered the weekdays between 18:00 and 08:00 h and the weekends. Data were collected over a period of about 4000 days for about 12 years between 2002 and 2013 at the 17 different monitoring sites around CS 1 -CS 3 ( Table 2). A diverse range of construction works during the different phases of the construction were anticipated at the studied sites. However, we did not have access to information of the different phases of the construction processes at each site, except the overall duration of the works. Data were analysed with reference to the EU Limit Values for annual and daily PM 10 concentrations. In addition, bivariate plots were drawn to qualitatively assess the effects of wind speed and direction on the measured concentrations in upwind and downwind directions from construction sites. The k-means clustering technique was then applied to assess contribution of probable local construction sources, which were identied through bivariate polar plots of paired monitoring stations (i.e. one in upwind and the other in downwind). The k-means clustering technique helped to identify the range of increases in particle mass concentrations due to the construction operations, including the resuspension and emissions from any on-site vehicles. 43 Moreover, the frequency and variation in PM 10 and PM 2.5 concentration in the prevailing wind direction were evaluated with changes in distance from sources by pairing the sites in downwind of the CS 1 -CS 3 to assess the decay prole of the PM emissions, which is important to understand the impact of the construction works on the air quality in surrounding areas.

Instrumentation
PM concentrations at CS 1 were collected using a Tapered Element Oscillating Monitor (TEOM 1400) and those at CS 2 and CS 3 were measured using a Turnkey Osiris instrument (model 2315). Practical constraints, such as space and cost, were important factors in the instrument selection.
The TEOM 1400 was used to measure mass of particles per unit volume of air in the size range of 0.1-10 mm. The sampling stream and lters were heated to 50 C to maintain a stable temperature and to eliminate interference from water on the lter. 44 The mass measurement relied on the measurement of the resonant frequency of an oscillating system that consists of the lter and glass element. A correction factor of 1.3 was recommended in the UK for comparison of PM 10 measurements from TEOM with the EU Directive 1999/30/EC 13,41 prior to the development of a dynamic correction system 45,46 and was applied in this study. Further details about the working principle and sensitivity of the TEOM 1400 can be found elsewhere. [47][48][49] The Turnkey Osiris instrument (model 2315) was used to measure the mass distribution of particles per unit volume of air in the 0.4-20 mm size range by light scattering technology in a mass concentration range of 0.1-6000 mg m À3 . 50 The Osiris instrument is a portable device that is capable of sampling and measuring particle concentrations in real-time with a high temporal resolution (1 s minimum). The air sample is continuously drawn into the instrument by a pump with a ow rate set by the microprocessor at a rate of 0.6 l min À1 through an inlet heated to 50 C to minimise the effects of water droplets and particle bound water. Over 20 000 particles per second can be sized before coincidence effects occur. Several size selective inlets are also available for the instrument. These can be used to collect a size selected gravimetric sample on the instrument's lter and will measure in mg m À3 . The Osiris instrument also allows wind speed and direction, temperature and relative humidity to be recorded at the same time. The Turnkey Instruments, OSIRIS monitors, have also been used for the assessment of indoor and outdoor PM levels as well as personal exposure in a number of past studies. 51,52 Meteorological data was produced taking a mean from a number of different monitoring locations across the monitoring stations and construction areas wherein the meteorological equipment is considered to be working well and the data shows a good correlation. The measurements were carried out using cup anemometers and wind vanes (as opposed to sonic anemometers) mainly made by Campbell Scientic. This equipment was located at a height of about 10 m.  Table 2. Fig. 2 shows the polar plots that were constructed by partitioning wind speed and direction data and their corresponding hourly mean PM concentration data into different wind speed and direction bins. 43 These plots are presented as smoothed surfaces showing the variations in concentration, depending on the local wind direction and wind speed at a receptor. 53 The results presented in Fig. 2 show evidence of increased concentrations levels of PM 10 and PM 2.5 when the wind direction was from the construction sites to the monitoring stations.

View Article Online
Closer inspection of polar plots at all 17 monitoring stations around each of the three sites indicates the following: rst, whenever there are monitoring stations in the downwind side of the construction sites, high concentrations of PM 10 (Fig. 2) and PM 2.5 (Fig. 3) are observed, indicating a potential contribution from the construction activities (Fig. 2). Second, pockets of high PM concentrations can also be observed in some cases (for example, observe MS 2 for PM 10 and MS 9 for PM 2.5 in Fig. 2a  and 3, respectively) despite the monitoring stations being in the upwind of the east and south-west wind directions. This was expected due to long-range transport of PM 10 during easterly winds from European countries 54,55 and the effect of generated sea salt on PM 2.5 from the south-westerly winds. 55,56 This observation also suggests that the concentrations measured downwind of the construction sites include some contribution of emissions from these sources and are not solely from the emissions of the construction activities. However, this analysis was inadequate to conclusively report that the measured downwind emissions are from the construction sites. Therefore, paired-site (Section 3.2) and k-means (Section 3.3) analyses were performed to better understand the contributions of the construction emissions during varying wind directions.

Assessment of the paired sites for examining differences in PM concentrations
We paired the monitoring stations opposite to each other, upwind and downwind of the construction sites to assess the relative changes in the concentrations that may have been contributed by the construction emissions (Fig. 4). We found two pairs of paired monitoring stations each for PM 10 and PM 2.5 around CS 1 (Fig. 4a) and another two pairs for PM 10 around CS 2 (Fig. 4b), giving a total of 6 paired monitoring stations. This pairing allowed us to measure changes in the concentrations (i.e. DPM 10 and DPM 2.5 ) as air mass crosses the construction sites and the results are presented in Fig. 4. For example, the hourly mean differences in PM 10 and PM 2.5 at CS 1 measured in the two pairs of opposite monitoring stations (MS 1 , MS 4 and MS 7 , MS 8 ), which were estimated as MS 4 minus MS 1 and MS 7 minus MS 8 . Likewise, the hourly mean differences at CS 2 were calculated using MS 14   suggest that the differences are larger for PM 10 when compared with PM 2.5 , suggesting a relatively greater variability in PM 10 emissions than those in PM 2.5 from construction works. Similar observations were reported by previous studies 57 wherein they found greater increases in PM 10 compared with PM 2.5 from road widening works in London.

k-Means cluster analysis
To identify and independently assess the contribution from local sources, k-means cluster analysis was applied on the 6 paired monitoring stations that were identied and discussed in Section 3.2. Eight different clusters were chosen that were found to be optimal for separating the local source contribution from external sources, based on the recommendations from previous studies. 53,58-60 Fig. 5 and 6 show the contribution of each cluster in the polar plots of DPM 10 and DPM 2.5 . The temporal variation of DPM 10 and DPM 2.5 contributed by each cluster on an hourly, weekly and monthly basis are also shown. Based on the DPM 10 and DPM 2.5 concentrations showing the high concentration peaks in the polar plots (Fig. 4), clusters 5-7 can be identied to represent the concentrations of DPM 10 (Fig. 5) and DPM 2.5 (Fig. 6) due to construction sources. If we observe these clusters in the temporal variation plots, peaks can be observed during the weekdays, which are missing during the weekends. This is also demonstrated by the increases in the PM 10 and PM 2.5 concentrations during 08:00 and 18:00 h, which we referred to as "working hours". The temporal plots on a monthly basis were examined and the identied clusters (i.e. 5-7) showed relatively lower concentrations during the cold months (i.e. December, January and February) compared with the rest of the months. There could be two possible reasons for these lower concentrations: (i) less construction activity compared to normal and (ii) the weather conditions suppressing the emissions and transport of particles due to relatively wetter conditions than the other months and also affecting the normal construction due to adverse weather conditions. For example, the mean precipitation and relative humidity is expected to be higher during winter months (e.g. 256 mm and 85% to 213 mm and 70% during summer) and low temperature (e.g. mean $4 C to 15 C during summer). 61 Past studies have found wet conditions such as water spraying an effective method to suppress coarse particle emissions by up to 13-times during construction works. 62 Detailed receptor modelling studies could help further in drawing rm conclusions. Fig. 7 shows the annual mean PM 10 and PM 2.5 concentrations at the three construction sites. The annual average in PM 10 concentrations were found to be 22.9 AE 3.3 mg m À3 , 18.8 AE 2.2 mg m À3 and 34.9 AE 2.8 mg m À3 at CS 1 (Table 3), CS 2 ( Table 5) and CS 3 (Table 6), respectively, whereas the annual average PM 2.5 concentrations were 14.0 AE 1.7 mg m À3 at CS 1 ( Table 4). These averages include both the working and non-working hours and the averages for these separate durations are presented in ESI Fig. S1a and S2 and described in ESI Sections S1-S2. † Depending on the source and distance from the monitoring stations, the values of PM 10 and PM 2.5 varied and the concentrations in all cases increased as the working period started (ESI Fig. S1-S4 †). In general, the concentrations observed during the working hours were higher than those during non-working hours (ESI Tables S1-S4 †), presumably due to construction activities and the other emission sources such as road vehicles in operation during working hours. Moreover, exhaust and non-exhaust construction sources were at rest during the non-working hours and therefore these are unlikely to contribute to the observed variations during the night. Because there was no major roadway around CS 1 , the variation in particle mass concentrations (PMCs) between the three construction sites during working and non-working hours could be attributed to the variability in meteorological conditions (mainly wind speed and direction; Fig. 2 and 3) during the different years of the measurements. Overall, the PM 10 values were about $24%, 18% and 120% larger during the working periods when compared with those observed during the non-working periods at CS 1 , CS 2 and CS 3 , respectively. Moreover, at CS 1 , there was an increase of about 11% in PM 2.5 values during the working period when compared with the non-working periods (ESI Fig. S2 †).

Particle mass concentrations during working and nonworking hours
A comparison of the 24 hour average concentrations of PM 10 with the EU Directive 2008/50/EC, 42 as described in Table 7, suggests the number of exceedences each year (Table 8 and ESI Fig. S5 †). However, these exceedences are not expected to be due to construction works alone, given that the winds were blowing from various directions ( Fig. 1 and 2) and the presence of nearby sources could also have made a contribution to these exceedences. Therefore, we ltered the data based on the wind direction on an hourly basis at each monitoring station (Fig. 8). The  Table 8 shows these exceedences possibly due to the construction activities, which were, except on two occasions in 2003 Table 3 The annual average concentrations of PM 10 , including the working and non-working periods at CS 1 ; AE refers to standard deviation and "-" to the unavailability of data     (Table  7). Unlike previous studies 13 where the exceedences of daily mean PM 10 concentrations were reported over the EU limit value of 50 mg m À3 on several occasions during the monitoring of emissions from road and building works in London, our exceedences are within the regulatory limits and could also be attributed to the construction works, given that the paired polar roses and k-means clusters analysis in Sections 3.1-3.3 suggesting a clear contribution of the construction works on the downwind monitoring stations.

Decay proles of PM 10 and PM 2.5
The purpose of this section is to evaluate the variation in concentrations of PM 10 and PM 2.5 at different distances from the construction sites. This analysis assisted in understanding how far the PM concentrations can go to affect the surrounding areas as well as help in planning for emission mitigation measures, particularly for construction sites close to sensitive areas such as hospitals or schools. Fig. 9 shows the decay proles of the PM 10 and PM 2.5 concentrations with the changing distance from CS 1 and CS 2 . Both the logarithmic (Fig. 9a) and exponential (Fig. 9b) best-t functions were applied to our DPM 10 and DPM 2.5 . The logarithmic decay function (Fig. 9a) was chosen as a best t to our data based on better R 2 values than those given by an exponential decay prole as 0.79, 0.90 and 0.89 for PM 10 (CS 1 ), PM 10 (CS 2 ) and PM 2.5 (CS 1 ), respectively (Fig. 9b). The differences between the hourly averages of PM 10 and PM 2.5 concentrations (DPM 10 and DPM 2.5 ) during the working and non-working time periods provided the net concentrations from the construction activities, which were then used to draw decay proles (Fig. 9).  Table 8 The number of exceeded days from the EU standard limit and UK government objective (AQS). Please note that the exceedences presented in the parenthesis against each exceedance number represent the exceedences belonging to the 24 h periods when the wind was blowing from construction to the monitoring stations. This represents the possible exceedences due to construction activities. "-" refers to unavailability of data  5 18 ( - Furthermore, the calculated concentrations for PM 10 and PM 2.5 were ltered on the basis of prevailing wind direction. The decay prole of the PM 10 concentrations at CS 1 was drawn using the data measured at MS 4 , MS 5 , MS 7 and MS 3 , which were 100, 200, 500 and 1000 m away from CS 1, respectively. Moreover, the data measured at MS 4 , MS 6 and MS 7 were used to draw the decay prole of PM 2.5 at CS 1 , which were 100, 200 and 500 m away from CS 1 , respectively. Furthermore, the decay proles of the PM 10 concentrations were measured at 100, 200 and 400 m away from CS 2 at MS 10 , MS 11 and MS 15 , respectively. Because of atmospheric dilution, the mass concentration dramatically decreased with an increasing distance from the construction site to approximately half of its value at a distance between 100 and 200 m. The best tting logarithmic decay equations for PM 10 were drawn, which gave R 2 values of 0.92 and 0.91 for CS 1 and CS 2 , respectively (Fig. 9a). A much higher rate of change in the PM concentrations can be observed close to the construction site when compared with those at greater distances. For instance, the rate of change in PM 10 (CS 1 ) concentration with per meter distance was 0.06 mg m À3 in between 100 and 200 m, which decreases to 0.030 and 0.013 mg m À3 per meter distance in the 200-400 m and 400-1000 m range, respectively (Fig. 9a). These observations suggest to measure the PM within a few 100 meters distance from the construction sites to capture the rapid decay in PMCs. The total  Although studies measuring the decay of the PM concentrations around the construction sites are rare, we tried to compare our data with the most relevant studies. For example, Hitchins et al. 63 determined the PM 10 concentration at increasing distances from a road at two sites in Australia. They found that PM 2.5 and ultrane particles decayed by up to half of their maximum initial concentrations within a distance of 100-150 m from the road. Likewise, Buonanno et al. 64 found the PM 10 concentration values to decrease exponentially away from the freeway in Italy during weekly traffic conditions.

Summary and conclusions
OSIRIS (model 2315) and TEOM 1400 were used to measure the mass concentration of particles in the 0.1-10 mm size range around three construction sites at 17 monitoring stations over a period of 12 years between January 2002 and December 2013.
The objectives were to understand the emission characteristics of coarse and ne particles from construction activities, identifying their contribution to the ambient levels of PM concentrations in the vicinity of these sites and their spatial decay away from the construction sites.
The following conclusions are drawn from this study: The source characteristics of PM 10 and PM 2.5 were investigated using bivariate concentration polar plots and k-means clustering techniques. The high concentrations of PM 10 and PM 2.5 were observed at the paired monitoring stations during the construction works when the winds were blowing from construction sites towards the monitoring stations. A k-means clustering technique provided a useful development to the bivariate polar plots to assess the contribution of construction and other local sources.
Three clusters (5, 6 and 7) from a total of the 8 selected clusters showed strong evidence of a downward increase in PM 10 and PM 2.5 levels during the weekdays. These clusters were identied to represent construction activities.
PM 10 were found about $24%, 18% and 120%, and PM 2.5 about 11%, larger during the working periods when compared with those during non-working periods at CS 1 , CS 2 and CS 3 , respectively. These increases were attributed to the construction works as indicated by the bivariate concentration polar plots and k-means clustering analysis. In addition, the downwind concentrations of PM 10 were found to be relatively more inuenced by construction works at CS 1 than the PM 2.5 concentrations.
The 24 h mean EU limit of value of 50 mg m À3 set by EU Directives for PM 10 not to be exceeded more than 35 times a calendar year was breached on two occasions due to construction operations on downwind monitoring stations during the measurements taken between 2002 and 2013.
Both the total PM 10 and PM 2.5 values during working periods in the downwind direction were found to be well correlated with distance with R 2 values over 0.90 in a logarithmic form. These concentrations reduced to half of their initial concentrations within a few 100 meters. This indicates that placing a monitoring station around a site within this peripheral distance could help capture the rapid decay of particles escaping from the construction sites.
The results presented in this study highlight the contributions of PM 10 and PM 2.5 from construction works. The increase in the concentrations of PM 10 and PM 2.5 at the downwind monitoring stations suggest that there is a need to design more detailed and appropriate risk mitigation strategies to limit the exposure to onsite workers and people that live in the surrounding environment. Further studies covering chemical ngerprinting of size fractionated PM from construction operations are recommended to understand their chemical composition as well as apportioning construction dust (e.g. using calcium and other similar minerals as a marker) from the exhaust emissions of construction machinery (e.g. using black carbon as a marker 65 ), together with allowing to differentiate between the properties of construction dust and the PM produced by the most common source (i.e. road vehicles) in urban environments. 16  In the fitting equations, x and y express the distance from principal construction site and the PM 10 values, respectively. The solid lines represent the best fitting decay curves.