Particle Number Emissions and Source Signatures of an Industrial

Dec 16, 2005 - The aims of the investigations were threefold: (a) the identification of the plant signatures in terms of particle size distributions i...
0 downloads 13 Views 1MB Size
Environ. Sci. Technol. 2006, 40, 803-814

Particle Number Emissions and Source Signatures of an Industrial Facility L. MORAWSKA,* G. R. JOHNSON, C. HE, G. A. AYOKO, M. C. H. LIM, C. SWANSON, Z. D. RISTOVSKI, AND M. MOORE† International Laboratory for Air Quality and Health, Queensland University of Technology, GPO Box 2434, Brisbane, QLD 4001, Australia

The work presented was conducted within the scope of a larger study investigating impacts of the Stuart Oil Shale project, a facility operating to the north of the industrial city of Gladstone, Australia. The aims of the investigations were threefold: (a) the identification of the plant signatures in terms of particle size distributions in the submicrometer range (13-830 nm) through stack measurements, (b) exploring the applicability of these signatures in tracing the source contributions at locations of interest, at a distance from the plant, and (c) assessing the contribution of the plant to the total particle number concentration at locations of interest. The stack measurements conducted for three different conditions of plant operation showed that the particle size distributions were bimodal with average modal count median diameters (CMDs) of 24 (SD 4) and 52 (SD 9) nm. The average of all the particle size distributions recorded within the plant sector at a site located 4.5 km from the plant, over the sampling period when the plant was operating, also showed a bimodal distribution. The modal CMDs in this case were 27 and 50 nm, similar to those at the stack. This bimodal size distribution is distinct from the size distribution of the most common ambient anthropogenic emission source, which is vehicle emissions, and can be considered as a signature of this source. The average contribution of the plant (for plant sector winds) was estimated to be (10.0 ( 3.8) × 102 particles cm-3 and constituted approximately a 50% increase over the local particle ambient concentration for plant sector winds. This increase in particle number concentration compared to the local background concentration, while high compared to the clean environment concentration, is not significant when compared to concentrations generally encountered in the urban environment of Brisbane.

Introduction Emissions from stationary sources such as industrial plants, power plants, and refineries are a major contributor to global atmospheric pollution. Inventories of emissions in relation to the pollutants regulated by air quality legislation are * Corresponding author phone: +61 7 3864 2616; fax: +61 7 3864 9079, e-mail: [email protected]. † National Research Centre for Environmental Toxicology, 39 Kessels Rd., Coopers Plains, Brisbane, QLD 4108, Australia. 10.1021/es048337e CCC: $33.50 Published on Web 12/16/2005

 2006 American Chemical Society

conducted routinely in most developed countries, and therefore, there is a relatively large body of information available concerning the relative contributions of individual facilities to the total emissions of these pollutants. However, there is much less information available on the contribution of individual industrial facilities to the actual airborne concentrations of the pollutants on different spatial scales, including in immediate proximity to and at various distances from a plant. There is even less information available on the contributions of industrial facilities to ambient concentrations of unregulated pollutants which are nonetheless potentially associated with health and/or environmental impacts, as is the case for submicrometer particles. The reason for the former is that source contribution to the pollutant concentrations is not spatially uniform owing to the effects of local meteorological conditions and topography. Such information could be obtained through extensive modeling; however, such modeling is not always conducted in association with, or as a part of, emission inventories. Validation of pollutant dispersion and concentration models would also require an adequate body of monitoring data. Yet, such information is of importance in relation to human exposure and risk assessment and, in turn, risk control. The need for information on airborne submicrometer particles has been highlighted in recent years by the emergence of scientific evidence that these very small particles may be more significant from the point of view of health than the larger particles or gaseous pollutants (1). A large variety of industrial facilities involve combustion of fossil fuels or biomass, and combustion processes are the most significant source of these particles in the air. Submicrometer particles are normally measured in terms of their number concentration: the numbers of particles emitted and present in the air are large, yet their mass is very small. Measurements of particle number concentration require state of the art instrumentation, and there are no standard procedures for conducting such investigations. Despite the complexities involved in measuring airborne submicrometer particles, the value of such measurements lies in the fact that they are conducted in real time and therefore provide data suitable for the investigation of spatial and temporal variations in particle concentrations. In addition, the size distribution of the particles can be used as a source signature, thus providing insight into contributions from individual sources. Source apportionment is complicated by the fact that ambient air contains a dynamic mixture of pollutants emitted from various sources. Therefore, it is a mixture which undergoes continuous change in time as the interactions between the pollutants take place and as the components of the mixture are removed from the air due to the presence of various sinks. This difficulty is further complicated by the fact that specific emission characteristics are rarely unique to a particular source. The particle physical characteristic which is most applicable as a source signature is the number size distribution. When a single pollution source is investigated and when it operates under steady-state conditions (for example, steady parameters of the combustion process), the size distribution obtained is likely to have one or two distinctive peaks. Those peaks are called modes of the distribution. In the case of a mixture of particles from various sources and of different size distributions, the measured size distribution may or may not display individual peaks from the contributing sources and thus may or may not be used for source identification. In addition to the mixture problem, processes including coagulation and atmospheric reactions VOL. 40, NO. 3, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

803

can make the signature less clear as the distance from the source increases. Despite these complications, in many cases the particle size distribution can be a very useful tool in source characterization (2). This is particularly true in areas relatively close to the source. The work presented was conducted within the scope of a larger study investigating impacts of the Stuart Oil Shale project, stage 1, which is a facility operating to the north of the industrial city of Gladstone, Australia. Emissions associated with the extraction of hydrocarbon products from oil shale were investigated. The extraction process is summarized as follows. Shale is crushed and dried before entering an Alberta-Taciuk processor (ATP). The ATP consists of a horizontal rotating cylindrical processor containing two concentric kilns. The inner kiln has two compartments for preheating and retorting the shale, driving off the kerogen component from which the hydrocarbon product is recovered by cooling. Heat for the process is provided by residual petroleum coke burning in the outer kiln which carries the combusted shale in the opposite direction to the flow in the inner kiln. The hydrocarbon contains approximately 0.4 wt % sulfur and 1.0 wt % nitrogen. In response to community concerns about pollution emitted by the plant and its possible links with health symptoms experienced by the community, investigations were undertaken to quantify the contribution of the emissions from the plant to the concentrations in the local airshed, as well as to investigate links between the signatures of the plant emissions and those identified at a community site during plant operation. While the program included investigations of a number of pollutants or factors, the focus of this paper is only on submicrometer particle characteristics and their application as source signatures. The aims of the investigations in this respect were threefold: (a) to identify signatures of the Stuart Oil Shale facility in terms of particle size distributions through stack measurements, (b) to explore the applicability of these signatures in tracing source contributions at locations of interest, at a distance from the plant, (c) to assess the contribution of the plant to the total particle number concentration at locations of interest, using particle monitoring data and meteorological data.

Experimental Section The experimental part of the study involved two processes: (1) measurements of the particle size distribution characteristics at the stack for different operating conditions of the plant and (2) monitoring of the particle size distribution and meteorological parameters at a location 4.5 km from the plant. The experimental data were analyzed to identify source characteristics of the plant, to compare them to the characteristics of the distributions at the field location, and to assess the contribution of the plant to the total particle number concentration at this location. Apparatus. Two TSI model 3932 scanning mobility particle sizers (SMPSs) were used to measure submicrometer particle size distributions. Both SMPS systems consisted of a TSI 3071A electrostatic classifier and a condensation particle counter (CPC). The CPC models incorporated by the systems were TSI 3010 and TSI 3022A for the ambient and the stack measurements, respectively. The flow rates of both SMPSs were 3 Lpm of sheath air and 0.3 Lpm of aerosol sample. The size ranges were 14-710 and 13-830 nm for the ambient and stack measurements, respectively. The size distribution scan duration was 6 min. The instrument calibration was verified prior to the measurements using NIST traceable standard size PSL particles. Comparison of the reported overall particle number concentration across identical size ranges for the two SMPS systems using the flow settings used in this study, for ambient and laboratory-generated aerosols, 804

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 3, 2006

FIGURE 1. Sampling train used for the stack measurements. showed that the overall particle number concentration reported by the two instruments agreed to within 20%. Sampling Methods: Stack Measurements. Studies of submicrometer stack emissions have been conducted previously, including one of a package boiler (3) and another of a municipal waste incinerator (4). The high particle concentrations in stack gases make it necessary to dilute the sample before measurements can be performed. As has been shown by Kasper (5), preventing the sample from cooling, before diluting it immediately with hot air, prevents the formation of nucleation mode particles during the dilution process. The technique is known as hot dilution. Hot dilution is also necessary to prevent water condensation on the particles, which could potentially lead to particle losses due to impaction and gravitational settling. In this study the sample probe was heated to maintain the elevated aerosol temperature up to the point of dilution. The heated sample was diluted using hot, filtered, dry air via an ejector diluter system, as this method provides stable dilution factors over long periods of time. Sample Train. The measurements were carried out using the hot dilution fixed volume sampling technique illustrated in Figure 1. Here, an ejection style diluter (Dekati diluter L6) provides a fixed dilution ratio of approximately 10 parts dry filtered air to one part stack gas sample. The sample is obtained by selecting an appropriate combination of aerosol flow rate and sampling nozzle dimension to allow the nozzle inlet velocity to be matched to the gas velocity at the sampling point centered in the stack at the sample port. The inlet nozzle leads via a heated (T ) 180 °C) tube to the ejection-type diluter where clean and dry pressurized dilution air, heated to 180 °C, entering the ejector cavity creates a pressure drop, drawing the sample into the diluter. The dilution air mixes with the sample air at the ejector cavity and further in the diffusion cavity, resulting in homogeneous dilution. Stack Access. Sampling was conducted from the 40 m platform on the stack using the existing sample ports. The instruments were transported to this level using a steel basket suspended by a cable from an electric winch controlled from the platform. Stack Particle Sampling Procedure. Sampling was conducted during three distinct conditions of plant operation, simulating a range of operating conditions at the plant. The establishment and maintenance of the operational conditions at the plant during the sample runs was conducted by the staff operating the plant. These operating conditions were defined as follows. “Condition 1”: intended to simulate the behavior of the plant when a fault causes the deactivation of the stack burner accompanied by reduced scrubber efficiency when the plant is operating at near maximum capacity. The effects of this condition include increased gas and particle concentrations in the stack accompanied by reduced temperature, moisture condensation, and reduced plume buoyancy. “Condition 2”: plant operating at feed rates

FIGURE 2. Map of the investigated area. of 170 t/h with low scrubber efficiencies and the stack burner lit. “Condition 3”: plant operating “under typical conditions” with feed rates of 140 t/h. The measurements of stack emissions were conducted over a period of 3 days, with the plant operating in a different condition each day. Sampling Methods: Stupkin Lane. Ambient aerosol samples were collected at the Queensland Environmental Protection Agency (EPA) monitoring station in Stupkin Lane, Targinie, located in a community residential environment about 4.5 km to the west from the plant. This site is referred to as “Stupkin Lane”. Figure 2 presents a locality map of the area. The first sampling run at Stupkin Lane from Feb 27 to March 2, 2002, was conducted in parallel to the stack emission testing for the first 3 days, with 1 additional day after the stack testing concluded. The plant was operating during this entire period. Measurements were conducted again at Stupkin Lane from May 15 to May 28, which was a period when the plant was shut down for maintenance, to develop a better understanding of the local background particle characteristics. The measurements at Stupkin Lane consisted of continuous SMPS particle size distribution scans, as well as wind speed and direction monitoring. Data Analysis. Stack Data Analysis. The sample was kept above its dew point throughout the measurement process. Measured volumes and concentration values, determined at the cooled and diluted sample temperature and pressure, were then transformed to a common temperature and pressure of 298 K. Sample dilution ratios were established from the dilution air pressure and the ejector diluter calibration and adjusted via the dilution factor to obtain the appropriate values for the undiluted sample, at the diluted sample conditions of pressure and temperature. Stupkin Lane Data Analysis. The SMPS data acquired from the instrument consist of a series of size distribution files. In cases such as this, where the instruments operate continuously over a period of days or weeks, sampling yields many thousands of size distribution files. Data analysis for these files consists of first transferring the information into a manageable data structure which allows the information from the SMPS and meteorological instruments to be collated and analyzed. Visual examination of the data was performed by plotting the SMPS and meteorological variables together on a time series graph. This allows unusual features caused by known instrument malfunctions to be detected. This process also

allows synchronous behavior in different instruments to be detected. For example, it is possible to identify size distribution features that may be unique to specific wind directions. Statistical Methods. A large database obtained from the study was subjected to comprehensive analysis, using various numerical and statistical techniques. In particular, comparisons were made using parametric or nonparametric statistical methods as appropriate for the distributions of the data being analyzed.

Results and Discussion Particle Size Distribution and Concentration Sampled at the Stack. Figure 3 shows the average size distributions recorded at the stack for the three conditions of the plant operation. The bimodality is very apparent with clear intermodal troughs visible in the average size distributions for Feb 27 and 28. On March 1 the modes were less widely separated. Table 1 provides a summary of particle total average concentrations, concentrations in the modes, and mode count median diameters (CMDs) for the three different conditions of plant operation. The average size distributions were calculated such that the mobility channel concentrations are the averages of the corresponding channels from all of the relevant size distributions for each of the three plant operating conditions. It can be seen from Table 1 that the total number concentrations recorded at the stack were similar for the various plant operating conditions, though somewhat higher for the plant operating at higher feed rate, and lower for the simulated fault condition. The differences between the concentrations for the various operating conditions, however, did not reach statistical significance at the 0.05 level. The locations of modes 1 and 2 varied throughout the 3 day measurement period by 18% and 17%, respectively, while the concentration of the modes varied by 73% and 47%, respectively. Thus, while the position of the size distribution modes was quite stable, the fraction of the total submicrometer particle number falling within each of the two modes varied strongly to the extent that at times only one of the modes was visible. Owing to their locations on the size distribution graph, the first mode could be identified as nucleation, composed of secondary particles, and the second as accumulation, consisting of primary particles (see, for example, refs 6 and 7). The second mode is therefore likely to contain mainly VOL. 40, NO. 3, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

805

FIGURE 3. Average, undiluted size distributions and envelope of variability recorded at the stack conditions for the three conditions of plant operation. The error bars represent the envelope of variability ((1 standard deviation) of the particle number concentration at each diameter.

TABLE 1. Number Concentrations and Mode Diameters for Submicrometer Particle Size Distributions at the Stack plant operating conditions and date

total concn, cm-3 × 106

concn, cm-3 × 106

1 (Feb 27), av SD SD, % 2 (Feb 28), av SD SD, % 3 (March 1), av SD SD, % overall av SD SD, %

2.7 2.0 76 4.0 1.0 26 3.6 1.55 44 3.6 1.5 42

1.6 1.8 112 1.4 0.7 53 1.7 1.2 73 1.7 1.3 73

mode 1 GSD, cm-3

carbonaceous material, and the first condensed, semivolatile material. As the stack gases cool, semivolatile compounds in 806

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 3, 2006

1.23 0.04 4 1.18 0.06 5 1.28 0.04 3 1.24 0.07 5

CMD, nm

concn, cm-3 × 106

23 5 23 20 2 9 28 2 7 24 4 18

0.9 0.2 26 2.4 0.4 16 1.8 1.0 54 1.8 0.9 47

mode 2 GSD, cm-3 1.28 0.02 1 1.41 0.07 5 1.39 0.08 6 1.4 0.09 6

CMD, nm 54 11 21 46 6 13 55 9 16 52 9 17

the gas phase condense and the particles grow through condensation. If the existing particle surface area (in the

accumulation mode) is too small to accommodate the rapidly cooling and condensing gas-phase species, homogeneous nucleation occurs, creating a new peak which grows in diameter until the gas-phase concentration falls sufficiently to quench particle growth. The higher degree of variability of the concentration in the first mode at the stack supports this hypothesis and reflects variations in the condensing material vapor pressure due to variations in temperature, pressure, or gas-phase concentration and the existing particle surface area. The degree of variability in the particle concentration in the second mode, as expected for primary particles, is lower. This bimodal shape of the size distribution and the location of the two modes can be compared to those of the characteristic size distribution of particles emitted from other sources. The largest database of size characteristics of particles emitted from combustion outdoor sources exists for vehicle emissions. There were no measurements conducted specifically in Gladstone to characterize the size distribution of particles originating from the local fleet of vehicles on the road; however, there were extensive measurements conducted in Brisbane. It is expected that fleet and emission characteristics would be similar between these two subtropical cities, located in the same Australian state. The dynamometer emission testing conducted in Brisbane showed that emissions of vehicles operating on different fuels had distinctly unimodal size distributions with CMDs of diesel particles ranging from 63 to 68 nm (8), those of petrol ranging from 43 to 45 nm, and that of LPG equal to 60 nm (9). Another study comparing petrol and CNG emissions (10) showed that CMDs of emission peaks varied from 30 to 60 nm, with the variation most pronounced between different modes of vehicle operation, but less so between fuels. Thus, in summary, contrary to the plant emissions, the size distributions of vehicle emissions were in each case unimodal, and the locations of the modes within the size distribution graphs differed from those of the plant emissions. Unfortunately no data exist on the submicrometer particle size characteristics of any of the other industrial plants operating in Gladstone, or even Australia. Size distributions of particles emitted from a municipal waste incineration plant were obtained by a German group and showed that in most cases the distribution was also bimodal, with the nuclei mode located between 30 and 40 nm (depending on the plant operating conditions) and the accumulation mode between 80 and 140 nm (4). The nuclei mode was much less stable than the accumulation mode. In summary, the bimodal size distribution of particles emitted by the stack is unique compared to that of the other most common ambient anthropogenic emission source, which is vehicle emissions, and can be considered as a signature of this source. The limitation of this method is that while the location of the modes is more or less stable, the relative concentration of particles within the modes varies, and therefore, at times one or the other mode is less visible in the distribution, making the bimodality less pronounced. To overcome this limitation, longer term measurements at the receptor site may be necessary to obtain a sufficient amount of data, enabling statistical treatment and interpretation. Stupkin Lane Measurements. The particle concentration and size characteristics in ambient measurements can vary for many reasons, including background changes, variations in the contribution from the plant, and the meteorological conditions, of which wind speed and direction are the most important. Background for the purpose of this discussion is considered in a broad sense, and includes contributions from natural sources as well as from anthropogenic sources, other than the plant under investigation. Natural sources contributing to particles in the submicrometer size range include,

for example, bushfires, while anthropogenic sources include other industrial plants operating in the area. Therefore, the impact of the plant on particle characteristics at Stupkin Lane will be considered in the context of the background characteristics of the airshed and taking into account the wind direction. Wind Distribution during the Measurements. The predominant wind direction during the February measurements (plant operating) differed considerably from that during the background measurements conducted in May (plant not operating). Figure 4a shows the total time the wind was in each wind direction sector during the two sets of measurements. In this paper wind directions are given as angles in degrees referring to upwind directions. Wind direction data in the figures are grouped according to the widely used 22.5° interval standard compass bearings. These include north (0°), which ranges from 348.75° to 11.25°, north-northeast (22.5°) with the range 11.25-33.75°, northeast (45°) with the range 33.75-56.25°, etc. The plant is located at 83° relative to the Stupkin Lane site. The influences of local topographical features and meteorological conditions ensure that the trajectory followed by air moving between the plant and the Stupkin Lane site is not confined to the above direction. The term “plant sector” is used in this paper to refer to those wind directions ranging from 45° to 135°, which are the directions from which the plant emissions are most likely to be carried to the Stupkin Lane site. The directions ranging from 202.5° to 337.5° are referred to as the “nonplant sector” as they are unlikely directions for the arrival of plant emissions. A number of industries are located in the plant sector with respect to the Stupkin Lane site and are additional potential contributors to ambient particles for winds from this sector. The relative locations of these industries are summarized in Table 2. To investigate and characterize the overall contribution from these sources, a further set of background ambient size distribution measurements was conducted when the plant was shut down for maintenance. Because the shutdown occurred in May, which is late autumn and winds from the plant sectors are less frequent at this time than in summer (February), the measurement period was extended to two weeks, thus ensuring that these sectors were well represented in the data set. For either period (plant operating in February/March or plant not operating in May), the minimum cumulative time that the wind was within each of the standard compass bearings within the plant sector was 6 h. Figure 4b shows the average wind speed in each compass bearing during the measurements. As can be seen from the graph, the February wind speeds were on average lower for the plant sector than those recorded in May. Relationship between Time-Resolved Submicrometer Particle Size Distribution and Wind Direction at Stupkin Lane. A time series graph showing the submicrometer size distribution and wind direction during plant operation in February/March and during plant shutdown in May above a common time scale is presented in Figure 5. The plant sector is highlighted along with the times when the wind lay in this sector. The graph indicates particle concentration in terms of shades of gray, as shown in the concentration shade legend. The graph reveals an increase in concentration for diameters between 20 and 100 nm during the highlighted times in the section referring to the period when the plant was operating. The size distribution during these episodes is bimodal as shown in Figure 6a, which presents the average of the 10 highest concentration scans for two intervals when the wind was within the plant sector. On Feb 28 and March 1, modes 1 and 2 had CMDs of 27 and 53 nm and 30 and 54 nm, respectively, for the average of the 10 highest concentration scans in the plant sector at Stupkin Lane. Thus, the VOL. 40, NO. 3, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

807

FIGURE 4. (a) Total time wind was in each wind direction sector (deg): February (plant operating) and May (plant not operating). (b) Average wind speed in each wind direction sector (deg): February (plant operating) and May (plant not operating).

TABLE 2. Industry Locations Including the Oil Shale Processing Plant Relative to the Stupkin Lane Monitoring Site industry type

wind direction, deg

distance, km

oil shale processing plant cement clinker production sodium cyanide production sodium cyanide, ammonium nitrate, and chloralkali production coal-fired power station

83 90 127 131 121

4.5 4.8 4.8 7.7 13.0

CMDs of the nuclei mode on both days were comparable and slightly larger than that of the average nuclei mode at 808

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 3, 2006

the stack of 24 nm. The location of the accumulation mode was almost identical on both days and almost the same as

FIGURE 5. Stupkin Lane submicrometer particle size distribution and wind direction versus time for the periods when the plant was operating (labeled here as “plant active”) and when the plant was not operating (labeled “plant inactive”). the average location of this mode at the stack of 52 nm. While the apparent very small growth of the nuclei mode between emission from the stack and arrival at the Stupkin Lane site is within the standard deviation of the CMD of this mode for the stack, there is no sign of accumulation mode growth. This observation is consistent with particle growth theory. Assuming that condensation processes occurring in the stack and affecting the particle size distribution stop immediately after the plume is emitted from the stack, the remaining growth mechanism for both modes is coagulation. It has approximately 1 h (on the basis of the distance to

Stupkin Lane and the average wind speed during the measurements) to act before the sample reaches Stupkin Lane. Considering particle concentrations in the diluted plume, which are on the order of 103 to 104 particles cm-3, coagulation is not expected to significantly change the particle size distribution between emission from the stack and measurement at Stupkin Lane (11). Increases in concentration are also evident in the section of the graph representing the measurement period when the plant was not operating; on May 26 and 28, however, in this case the size distributions contain a single mode as shown VOL. 40, NO. 3, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

809

FIGURE 6. Average of the 10 highest concentration submicrometer particle number size distributions and the corresponding envelopes of variability for high-concentration periods with wind in the plant sectors. The error bars represent the envelope of variability ((1 standard deviation) of the particle number concentration at each diameter. (a1, a2) Plant-operating measurement period. (b1, b2) Measurement period when the plant was not operating. in Figure 6b. The CMD for this mode was 43 and 39 nm on May 26 and 28, respectively. The difference in the location of this pronounced peak compared to the location of the peaks associated with the plant indicates a strong source, presumably combustion, different from the plant (the plant was not operating during this time). It was not possible to identify the source within the scope of this work. Parts a and b of Figure 6 also show the corresponding cumulative concentration curves for the same average size distributions. Relationship between the Submicrometer Particle Size Distribution and the Concentration and Wind Direction at Stupkin Lane. Parts a and b of Figure 7 show the average particle size distributions recorded at Stupkin Lane during the measurements in the February/March (plant operating) and May (plant not operating) periods, respectively. The size distribution data were grouped by wind direction and averaged over the entire period of the measurements (4 and 14 days, respectively). Particle concentrations were higher for most wind sectors during the background sampling in May. Rural fire service reports indicted fires during the May sampling period occurring in the district. No evidence of fires was observed 810

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 3, 2006

during the February/March sampling period. Biomass burning tends to produce a broad, unimodal size distribution in the submicrometer size range centered below 200 nm (12). Figure 7b shows such a mode in the distribution, with the peak located at 113 nm, in May, when wind is in the 337.5° direction. Unfortunately, the fire occurrence and locality data for the region are not complete, and attempts to draw more detailed conclusions linking reported fire locations to wind direction could not be conducted. The three-dimensional presentation of the relationship between wind direction and size-classified particle concentrations (Figure 7) gives a good general picture of this relationship; however, it makes some features of the distributions (modes) less pronounced. Modality in the distributions is more evident in Figure 8, which shows the average of all the particle size distributions recorded within the plant sector over the sampling period when the plant was operating (February/March) and not operating (May). Modes 1 and 2 are located at 27 and 50 nm, respectively. These compare well with the average CMDs for the modes for the stack (24 and 52 nm, respectively) and with the positions of the modes for the average of the 10 highest concentration scans in the plant sector, as discussed above.

FIGURE 7. (a) Average submicrometer particle number size distribution versus wind direction for Stupkin Lane: plant operating. (b) Average submicrometer particle number size distribution versus wind direction for Stupkin Lane: plant not operating.

FIGURE 8. Average submicrometer particle number size distributions recorded at Stupkin Lane during the two measurement periods (plant operating and plant not operating) for the plant sector. The stack size distribution is shown for comparison. Figure 7 shows that compared to the localized sources to the east of Stupkin Lane the background is relatively flat. The plant lies to the east of the Stupkin Lane site, and further east, immediately beyond the plant, lies the Pacific Ocean. Clean marine environments typically display very low submicrometer aerosol number concentrations, well below 1000

cm-3. The background concentration for the western sectors was at the upper end of this range. During both sampling periods the winds displayed a regular diurnal pattern of alternation between sea and land breezes as can be seen in Figure 5. Background continental air moved out to sea when the winds were from the nonplant sectors, and this air VOL. 40, NO. 3, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

811

FIGURE 9. (a) Average submicrometer particle number concentration (cm-3) versus wind direction sector (deg). (b) Amount by which the average submicrometer particle number concentration for each wind direction sector (deg) exceeds the average for nonplant sectors. returned to the mainland during the sea-breeze (plant sector wind) conditions. The sea-breeze air which prevailed during plant sector winds was essentially recirculated continental air similar to the nonplant sector air. Figure 9a shows the influence of wind direction on the average submicrometer particle concentration at Stupkin Lane. The average concentration for the nonplant sector for the measurements when the plant was operating was 610 cm-3, and when the plant was not operating, it was 1300 cm-3. Assuming that the diurnal recirculation pattern implies that these averages represent the average background concentrations for all sectors for the respective measurement periods, the localized contributions as a function of direction 812

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 3, 2006

can be roughly approximated by subtracting these averages from all of the measured concentrations for the corresponding period. The result is shown in Figure 9b. Figure 10 shows the difference between the two results displayed in Figure 9b and represents the overall contribution of the plant during the February/March plant-operating measurement period. It can be seen that the average submicrometer particle number concentration contribution of the operating plant reaches its maximum value in the direction of the plant. The result is to some degree affected by differences in meteorological conditions in February and May, and in particular wind speed. As shown in Figure 4b, wind speed

FIGURE 10. Apparent contribution of the plant to the average submicrometer particle concentration at Stupkin Lane versus the wind direction sector (deg). in the plant sectors was slightly higher in May, which would result in greater dilution of source plumes when the plant was not operating and, in turn, in an overestimate of the contribution of the plant during operation by 15% of the May concentration for the relevant sector. The average concentrations for the western sectors, treated here as the background, varied from 10.3 × 102 to 16.8 × 102 cm-3 (a range of 6.5 × 102 cm-3) in the case where the plant was not operating and from 4.9 × 102 to 8.7 × 102 cm-3 (a range of 3.8 × 102 cm-3) in the case where the plant was operating. The uncertainty in the difference values portrayed in Figure 10, based on the uncertainty in the background concentrations of the western sector concentrations, is 3.8 × 102 cm-3. This is the propagated error (square root of the sum of the squares of the errors) in calculating this difference, where the uncertainty in the background level exceedance during the two periods is assumed to be half of the corresponding ranges discussed above. Taking into account the results presented in Figure 10 and the approximated 15% of the May background level exceedance as discussed above, the average contribution of the plant for plant sector winds is estimated to be (10.0 ( 3.8) × 102 particles cm-3. In comparison to the average plant sector concentration of 18.3 × 102 cm-3 during the period when the plant was not operating, the estimated plant contribution is statistically significant and amounts to about a 50% increase in the particle ambient concentration for plant sector winds. Source Emission Signatures. To control and mitigate particle emissions with a view of health and environmental risk reduction, a good understanding is necessary of the relative and absolute contribution from various emission sources. This understanding could only be achieved by developing source signature libraries through direct emission measurements from the sources, on one hand, and by measuring particle concentrations in the air and apportioning them to the specific local and distant sources using the signatures, on the other hand. The latter is a particularly complex process as ambient air contains a dynamic mixture of pollutants emitted from various sources, undergoing continuous change in time due to the interactions taking place as well as removal from the air due to the presence of

various sinks. Despite the complexities involved, source signatures provide powerful tools for source identification and apportionment, and this study demonstrated that particle size distribution measurements have the potential to be used as a source signature of industrial emissions. However, signature matching of field sample size distributions based solely on the shape of the size distribution has its limitations, and to use such signatures to obtain more definite answers to the question of plant impact on particle characteristics at the surrounding residential sites, longer term measurements would be required to provide a larger data set enabling better statistical analysis, as well as further development of the statistical method to test the significance of the perceived similarities. Comparison with Other Environments. While not the focus of this work, a brief comparison can be made between particle concentration levels measured at Stupkin Lane and concentrations reported in other parts of Australia. In general, the background concentration of particles in the investigated area is relatively low, with the average submicrometer number concentration of 1.5 × 103 particles cm-3 (with a maximum of 14 × 103) being several times lower than the average reported for the capital of the state, Brisbane, of 7.40 × 103 particles cm-3 (the maximum reported for the air monitoring station in Brisbane was 41 × 103) (13). The concentrations reported 150 m downwind from a busy road (14) were even higher, an average of 18 × 103 particles cm-3 and a maximum of 60 × 103 particles cm-3. Yet the concentrations are about twice as high as those reported at Cape Moreton (Moreton Island, located off the east cost of Australia, relatively free from anthropogenic influences, particularly for easterly winds from the Pacific Ocean), an average of approximately 0.8 × 103 particles cm-3 and a maximum of ∼4 × 103 particles cm-3. Thus, the significant increase in the particle number concentration of about 50% compared to the local background concentration, and even higher compared to the clean environment concentration, is not significant compared to the concentrations encountered in urban environments.

Acknowledgments This project was supported by the Department of State Development, Queensland Government.

Literature Cited (1) WHO. Guidelines for concentration and exposure-response measurement of fine and ultrafine particulate matter for use in epidemiological studies; World Health Organization: Geneva, Switzerland, 2002. (2) Morawska, L.; Zhang, J. Combustion sources of particles. 1. Health relevance and source signatures. Chemosphere 2002, 49, 1045-1058. (3) Miller, C. A.; Linak, W. P.; King, C.; Wendt, J. O. L. Fine particle emissions from heavy fuel oil combustion in a firetube package boiler. Combust. Sci. Technol. 1998, 134, 477-502. (4) Maguhn, J.; Karg, E.; Kettrup, A.; Zimmermann, R. On-line analysis of the size distribution of fine and ultrafine aerosol particles in flue and stack gas of a municipal waste incineration plant: Effects of dynamic process control measures and emission reduction devices. Environ. Sci. Technol. 2003, 37, 4761-4770. (5) Kasper, M. Sampling and Measurement of Nanoparticle Emissions for Type Approval and Field Control; SAE Paper No. 200526-013; Society of Automotive Engineers: Warrendale, PA, 2005; pp 313-327. (6) Kittelson, B. D. Engines and Nanoparticles: a Review. J. Aerosol Sci. 1998, 29, 575-588. (7) Kittelson, D. B.; Watts, W. F.; Johnson, J. Diesel Aerosol Sampling MethodologysCRC E-43: Final Report; Report for the Coordinating Research Council; University of Minnesota: Minneapolis, MN, 2002. (8) Morawska, L.; Bofinger, N. D.; Kocis, L.; Nwankwoala, A. Submicrometer and supermicrometer particles from diesel vehicle emissions. Environ. Sci. Technol. 1998, 32, 2033-2042. VOL. 40, NO. 3, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

813

(9) Ristovski, Z. D.; Morawska, L.; Bofinger, N.; Hitchins, J. Submicrometer and Supermicrometer Particulate Emission from Spark Ignition Vehicles. Environ. Sci. Technol. 1998, 32, 38453852. (10) Ristovski, Z. D.; Morawska, L.; Ayoko, G. A.; Johnson, G.; Gilbert, D.; Greenaway, C. Emissions from a vehicle fitted to operate on either petrol or compressed natural gas. Sci. Total Environ. 2004, 323, 179-194. (11) Baron, P. A., Willeke, K., Eds. Aerosol Measurement: Principles, Techniques and Applications, 2nd ed.; van Nostrand Reinhold: New York, 2001. (12) Hays, M. D. Speciation of Gas-Phase and Fine Particle Emissions from Burning of Foliar Fuels. Environ. Sci. Technol. 2002, 36, 2281-2295.

814

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 3, 2006

(13) Morawska, L.; Thomas, S.; Bofinger, N. D.; Wainwright, D.; Neale, D. Comprehensive characterisation of aerosols in a subtropical urban atmosphere: particle size distribution and correlation with gaseous pollutants. Atmos. Environ. 1998, 32, 2461-2478. (14) Hitchins, J.; Morawska, L.; Wolff, R.; Gilbert, D. Concentration of submicrometer particles from vehicle emissions near a major road. Atmos. Environ. 2000, 34, 51-59.

Received for review October 26, 2004. Revised manuscript received June 28, 2005. Accepted November 3, 2005. ES048337E