Airborne Particulate Matter Passive Samplers for Indoor and Outdoor Exposure Monitoring: Development and Evaluation Kholoud Abdulaziz, Kholoud Al-Najdi, Abdullah Kadri and Konstantinos E. Kakosimos
Abstract— The Middle East area is highly affected by air pollution induced by anthropogenic and natural phenomena. There is evidence that air pollution, especially particulates, greatly affects the population health. Many studies have raised a warning of the high concentration of particulates and their affect not just around industrial and construction areas but also in the immediate working and living environment. One of the methods to study air quality is continuous and periodic monitoring using active or passive samplers. Active monitoring and sampling are the default procedures per the European and US standards. However, in many cases they have been inefficient to accurately capture the spatial variability of air pollution due to the small number of installations; which eventually is attributed to the high cost of the equipment and the limited availability of users with expertise and scientific background. Another alternative has been found to account for the limitations of the active methods that is the passive sampling. It is inexpensive, requires no continuous power supply, and easy to assemble which makes it a more flexible option, though less accurate. This study aims to investigate and evaluate the use of passive sampling for particulate matter pollution monitoring in dry tropical climates, like in Middle East. More specifically, a number of field measurements has be conducted, both indoors and outdoors, at Qatar and the results have been compared with active sampling equipment and the reference methods. The samples have been analyzed, that is to obtain particle size distribution, by applying existing laboratory techniques (optical microscopy) and by exploring new approaches like the white light interferometry to. Then the new parameters of the wellestablished model have been calculated in order to estimate the atmospheric concentration of particulates. Additionally, an extended literature review will investigate for new and better models. The outcome of this project is expected to have an impact on the public, as well, as it will raise awareness among people about the quality of life and about the importance of implementing research culture in the community.
Keywords—air pollution, particulate matter, passive samplers. I. INTRODUCTION
A
ir quality has been a topic of interest recently in Qatar as the particulate matter were monitored to have high concentrations in the Middle East region in general [1, 2]. Passive sampling was found to be an efficient method of air monitoring as it is inexpensive and requires no power supply. K. Abdulaziz and K. Al-Najdi are with the Chemical Engineering Program, Texas A&M University at Qatar, Qatar (e-mail:
[email protected],
[email protected]). A. Kadri is with Qatar Mobility and Innovation Center, Qatar University, Qatar (e-mail:
[email protected]).
Passive samplers are easily deployed and are used to cover adequately special distribution with easy assembly. Passive sampling was first investigated through literature review along with the models of analysis used previously. Interferometry was introduced as a new method of scanning as a substitute for SEM and optical microscopy. The report documents the progress achieved in both the field and the analysis parts of the research, in addition to the processing of the results obtained. II. METHODOLOGY A. Background Previous studies have employed similar methodology of passive sampling such as Wagner who developed a miniature passive aerosol sampler to estimate average size distribution and concentrations [3]. In another study by Wagner, he employed passive sampling to measure ambient PM concentrations and particle types near agricultural burns in Imperial Valley [4]. Further development on the model was done by Ott who revised Wagner’s method to estimate ambient PM from samples collected by the passive sampler [5]. Ott conducted another study to enhance the ability to assess exposure to PM at a local scale by coupling passive sampling with effective sampling design [5]. Similarly, outdoor passive sampling was employed to measure the PM concentration levels in Greece [6]. The most commonly used methods of analysis are SEM and light microscopy which have some limitations including inaccuracy and lack of dimensional information. In the studies conducted by all the aforementioned groups scanning electron microscope and optical microscopes were used to scan the PM particles samples [4-8]. As a result, interferometry was introduced as a new method of analysis using the ZeGage. The interferometer provides numerical data of the xyz dimensions of the particles precisely. Scanned 3D maps are constructed by the interferometer, and then translated into MATLAB maps to obtain the PSD. Studies on the passive sampling have used the Schneider model [9] as a starting point for the PSD analysis, in which it takes in account the effects of electrical fields, particles charges, and turbulence on the particles. The model is further developed by K. E. Kakosimos is with the Chemical Engineering Department and Mary Kay ‘O Connor Process Safety, Texas A&M University at Qatar, Qatar (corresponding; phone: +974 44230678; e-mail:
[email protected]).
incorporating it in the passive sampling analysis which was accomplished by Wagner and Leith [10]. In this study, the deposition velocity consisted of both theoretical and empirical parts. The theoretical part includes the gravitational, inertial, and diffusive mechanisms for flat surfaces with a mesh. Ott el al. [11] adapted Wagner and Leith’s model with some simplifications that include equating the aerodynamic diameter with the projected diameter of the particle, the use of new shelter, and obtaining the volumetric shape factor from optical microscopy. Those were assumed to eliminate the estimation of variables like dynamic shape factor, friction velocity, and the volumetric shape factor. The previously mentioned models are used for indoor conditions, and so Assael et al. [6] adopted Wagner and Leith’s and Yamamoto’s models with the modification of the correction factor so that it can be used for outdoor conditions. Those models have shown close agreement with active samplers under the conditions they were developed on. On the other hands, a model was developed by Lai and Nazaroff [12] that accounts for the turbulent and Brownian diffusion affects along with the gravitational settling. The model is developed for indoor smooth surfaces, for both vertical and horizontal orientations. The model was adapted by Zhao and Wu [13] where it was modified for fully developed turbulent duct flows. This model is mainly applied to take in account the ventilation present indoors that may influence the deposition of particles. The turbophoresis is prominent for particles of 1 µm and greater, otherwise the effect of turbophoresis is ignored. One of the latest deposition rate models developed is by Hussien [14] which adapted the models developed by Lai et al. and Zhao et al. This model takes in account turbophoresis and coagulation affects for a wide range of particle size from 30 nm to 5 µm. A Multi-Compartment and Size-Resolved Indoor Aerosal Model (MC-SIAM) was used to verify the results obtained from the developed model, and it shows great agreement between the two results. Subsequent paragraphs follows the aims of the project as outlined earlier. B. Inception and Preparation A case was needed to hold the stubs during particle collection, so a design was made using SolidWorks and manufactured with aluminum (Fig. 1). The case consists of a main base that holds three stubs. To maintain stability, a bag of small clay balls was attached below the base. A protective disk was made of acrylic plastic to protect the stubs from rain and other large obstacles, which extends further than the main base. A smooth air flow was needed so a symmetric design was maintained.
(a)
(b)
Fig. 1 The protective case (a) as designed in Solidworks® and (b) the manufactured prototype.
Stubs from RJ Lee Group were first chosen to be used for sample collection, but they needed to be finely polished by chemicals in order to clear any pits. Due to the small size of the stubs, a polishing case was designed and created in order to hold the stub safely (Fig. 2b). The stubs were grinded up to 1200 grid but it was noticed that a lot of material is being eaten from it that caused the first layer to disappear. A new design for the stub using SolidWorks was made with a thick first layer to allow polishing without problems (Fig. 2a). The new stubs were then polished finely down to 0.04 μm using colloidal silica, in which then they were cleaned by ultra-sonication (Fig. 2c).
(a)
(b)
(c)
Fig. 2 A polished stub used to collect samples (a), the polishing and holding case for the stub (b) and a view of the ultra-sonication bath (c).
C. Counting and Sizing Methodology To analyze the samples, two methods of analysis were implemented. The first method is interferometry. The students underwent a basic training session to learn how to scan the samples with the interferometer using a green light of wavelength 530 nm, and apply some of the functions to the scans through the Zegage software. A reference sample of silica particles of size 1.57 μm was purchased for calibration. The reference sample helps in determining the efficiency and accuracy of the methods of analysis. The reference sample and test samples from test trials were scanned using the 50x lens of the interferometer. The scan generates a 3-dimentional image (Fig. 3) of the particles on the stubs which are to undergo further analysis to determine the projected area and the volume of particles. There were issues faced regarding the maps constructed by the interferometer. The reference material was not detected by the scanner as the green light was not reflected well back to the scanner. An explanation for this issue would be that the interferometer microscopy is not suitable for silica particles because of the low reflectivity (high transparency). For this reason another in-house, size characterized, natural sample was used (Dolomite dust; a calcium-magnesium carbonate) and tested with the same process. This specific sample has been selected because parallel field studies and XRF/XRD analysis showed that this is the main surface material in Qatar.
(a)
(b)
Fig. 3 2D produced maps from interferometer for the a) silica reference sample and b) in-house natural dolomitic sample.
In this case, a similar problem occurred and the maps constructed consist only from undefined areas. The justification is again related to the interferometry analysis, the green light is either absorbed or diffused. To investigate further the reasons behind this issue, SEM was used to see exactly the shape of particles and a literature search has been conducted to find out the reflectivity distribution of Dolomite.
(a)
(b)
(c)
Fig. 4 SEM Images of the standard samples a) 200x of Dolomite b) 8,000x of Dolomite and c) 20,000x of Silica.
SEM (Fig. 4) has showed irregular shapes with a lot of sharp edges which probably explains the diffusion of the green light on the tested natural material, but showed clear images of the silica particles with a diameter of exactly 1.57 μm. In addition the literature review showed that Magnesium and Calcite carbonates do exhibit high absorbance around the green light wavelength. Both the diffusion (owe to the particles surface) and the low reflectivity (owe to the particles composition) do justify the inability of green light interferometry on scanning such particles. This is one of the important conclusions of this study. On the other hand, this problem is not prohibiting the use of Interferometry as method of analysis of particulate passive samples. Since the natural materials are shown as holes in the constructed maps, the projected area can be obtained from them. Campaigns were conducted to collect dust samples indoors and outdoors, and they were all scanned using the interferometer. The maps were well constructed, and the dust particles were visible and so the projected area and volume were to be analyzed. A MATLAB code was needed to construct a map to interpret the data from the interferometer in order to get the projected area and the volume of the particles. Another option was to use
ImageJ, which is a program that analyzes scans of particles and gives results that include area and particle distribution. ImageJ required images of the particles which were not of high quality and so MATLAB was used for the analysis. The MATLAB code records the coordinates and the height of all the particles to generate a 3D map to be used for analysis with the adjustment of the reference height of the surface. The final is comprised of three parts: 1) The post processing of the interferometer maps and the estimation of the projected area of every particle 2) The calculation of the aerodynamic diameter and the projected diameter for every particle and then the PSD (number of particles per size bin) for every sample 3) Finally the calculation of the deposition velocity for every specific model (e.g. deposition model) and in other words the ambient concentration D.Field Measurements In order to get used to sample collection and analysis a number of test trials has been conducted. TABLE 1 presents briefly each trial. TABLE 1 DETAILS OF THE FIELD TEST TRIALS. Test Type of Date Location sampling A. Trial test Active and April 6 – Outdoor: CP2 1 passive April 10 roof B. Trial test Active and April 23 – Outdoor: CP2 2 passive April 28 roof C. Variation Active and May 18 – Outdoor: CP2 test 1 passive May 25 roof Indoor: room D. Variation Active and Start: June Outdoor: CP2 test 2 passive 8 roof End: Indoor: room 3 days: June 11 5 days: June 13 7 days: June 15 E. Indoor Active and Nov 20 – Indoor: lab test 1 passive Nov 27 F. Indoor test Active and Dec 4 – Indoor: lab 2 passive Dec 11
No. of stubs 3 3 3 3 3 3
3 3
Two test trials were conducted for 4 to 5 days each. In each trial, three stubs (Fig. 5a) were used for passive sampling along with a reference monitor (GRIMM Dust Monitoring EDM365) on the rooftop of the CP2 (Qatar Foundation Central Plant 2, Education City Doha, Qatar; Fig. 5b) building which is surrounded by construction. The stubs were then scanned and analyzed by the interferometer. Two other test trials were conducted, one to test the variations between the stubs and the other is to test for different durations of testing and both are in parallel with indoor sample collection which includes the reference online nephelometer Dylos 1700pro (Fig. 5c).The stubs were then also scanned and analyzed by the interferometer. Two campaigns were conducted to test for air quality indoors in a lab. Passive sampling with three stubs were used along with an active sampler for a duration of a week for each campaign. The samples obtained were then scanned and analyzed by the interferometer.
(a)
(b)
(c)
Fig. 5 Photos from the field trials (a) the protective case with 3 sampling stubs, (b) the online dust monitor GRIMM EDM365 and (c) Dylos 1700pro a portable nephelometer.
E. Model Development After conducting a literature review of the previously used models, three models were chosen and the results obtained from them are compared. The models of Wagner and Leith, Assael, and Lai and Nazaroff were chosen for the analysis of the PM concentration. 1) Wagner and Leith The deposition velocity model of Wagner and Leith consists of a theoretical component and an empirical component. The theoretical component incorporates gravitational, inertial, and diffusive mechanisms. The deposition velocity is considered as an ambient deposition velocity for flat, smooth surfaces and a mesh factor which is an empirical correlation to account for the effects of the sampler and the mesh. The study focused on particles of a size larger than 0.1 µm. The model of Wagner and Leith along with the model of Yamamoto were further modified by Assael to be used for outdoor conditions. Therefore Assael’s model was chosen to analyze the samples obtained outdoors. The concentration of the atmosphere can be defined by the following equation: 𝐶 = 𝐹/𝑢𝑑𝑒𝑝 = 𝐹/(𝑢𝑎𝑚𝑏 𝛾𝑚 ) (1) Where, C is the particle concentration in the atmosphere (in micrograms per cubic meters), F is the particulate flux (in micrograms per square meter per second), 𝑢𝑑𝑒𝑝 is the deposition speed (in meters per seconds), 𝑢𝑎𝑚𝑏 is the actual deposition velocity of PM (in meters per second), and 𝛾𝑚 is the dimensionless correction factor. 𝑢𝑎𝑚𝑏 can be calculated from the following equation: 𝑢𝑎𝑚𝑏 = −𝑢𝑡 /[[(1 − 0.67𝜏 0.49 𝑢∗ −0.02 𝜈 −0.49 𝑢𝑡 ) ⅇxp(−𝑢𝑡 𝐼) ] − 1] (2) Where, 𝑢𝑡 is the terminal particle velocity (in meters per second), 𝑢∗ is the friction velocity (in meters per second), 𝜈 is the kinematic viscosity (in square meters per second, 1.51 x 105 m2/s), and along with the rest of the variables, they can be calculated by the following equations: 𝑢∗ = (𝑘𝑢𝑧 )/𝑙𝑛(𝑧/𝑧0 ) (3) Where, k is the Von Karman constant which is 0.4, 𝑢𝑧 is the wind speed at height 𝑧 (in meters) (in meters per second), 𝑧0 is the roughness height (in meters). 𝑢𝑡 = 𝜏𝑔 (4) 2 𝜏 = (𝜌𝑝 ⅆ𝑎 𝐶𝑐 )/18𝜇 (5)
Where, 𝜌𝑝 is the PM density (in kilograms per cubic meter, 2700 kg/m3), ⅆ𝑎 is the aerodynamic diameter (in meters), 𝜇 is the viscosity of air (in kilograms per meter per second, 1.81 x 10-5 kg/m.s), and 𝐶𝑐 is the slip correction factor coefficient according to Cunningham, which can be found by: 𝐶𝐶 = 1 + 𝜆/ⅆ𝑝 (2.514 + 0.8exp(−0.55 ⅆ𝑝 /𝜆)) (6) 𝜆 is the mean free length of air molecule, 𝜆 = 0.066 𝜇𝑚 If 𝑢∗ is smaller than 0.4 m/s, then 𝑢𝑎𝑚𝑏 ≅ 𝑢𝑡 . As for 𝐼 which is the integral (in seconds per meter), it can be calculated using the following equation: 𝐼 = 𝐼 + /𝑢∗ (7) 𝐼 + = [(3√3)/(29𝜋) 𝑆𝑐 (−2/3) + 6.2 × 10−4 (𝜏 + )2 ]−1 (8) Where, 𝑆𝑐 is the Schmidt number which can be calculated by: 𝑆𝑐 = 𝜈/𝐷 (9) Where, D is the Brownian diffusivity coefficient is in terms of kB (Boltzmann coefficient), T (absolute temperature, 293 K), and 𝜇 the dynamic viscosity of air (1.81 x 10-5 kg/m.s). 𝐷 = (𝑘𝐵 𝑇𝐶𝐶 )/(3𝜋𝜇ⅆ𝑝 ) (10) kB =1.38 x 10-23 As for 𝜏 + is can be calculated by: 𝜏 + = (𝜏𝑢∗ 2 )/𝜈 (11) Part of the particle concentration in the atmosphere is the correction factor which can be calculated as follows: 1, ⅆ𝑎 < 1.63 𝜇𝑚 𝛾𝑚 = { (12) −0.539 0.0027𝑅ⅇ𝑝 , ⅆ𝑎 ≥ 1.63 𝜇𝑚 Where, 𝑅ⅇ𝑝 is the Reynold number defined as: 𝑅ⅇ𝑝 = (ⅆ𝑝 𝑢𝑧 )/𝜈 (13) The last variable needed is the particulate flux, F. The flux can be calculated as follows: 𝐹 = ∑𝑁 (14) 𝑖=1 𝑚𝑖 /𝐴𝑡 Where, N is number of particles, A is the collection surface area or microscopically analyzed area (in square meters), t is the sampler’s exposure time interval (in seconds, 1209600 s), and 𝑚𝑖 is the total mass flow equation: 𝑚_𝑖 = (3𝜋𝜇𝑆𝑑 )/(𝛾𝑚 𝑔𝑆𝑣 𝐶𝑐,𝑑𝑒𝑣 ) ⅆ𝑝𝑎,𝑖 (15) Where, 𝑆𝑑 is the dynamic shape factor (1.5), 𝑆𝑣 is the volume shape factor (1.5), and ⅆ𝑝𝑎 is the PM diameter (in meters). The aerodynamic diameter mentioned previously, can be calculated using the following equation: ⅆ𝑎 = ⅆ𝑒𝑣 ((𝜌𝑝 𝐶𝐶,𝑑𝑒𝑣 )/(𝜌𝑎𝑖𝑟 𝐶𝐶,𝑑𝑝𝑎 ) 1/𝑆𝑑 )1∕2 (16) Where, ⅆ𝑒𝑣 = ⅆ𝑝𝑎 /𝑆𝑣 , 𝜌𝑝 is the density of the particle, and 𝜌𝑎𝑖𝑟 is the density of air. 2) Lai and Nazaroff Lai and Nazaroff’s model considered the turbulent intensity, turbophoresis and the significance of coagulation. A wide range of 30 nm to 8 µm particle sizes was considered in their study. These models are used for indoor conditions and thus are opted to analyze the samples collected indoors. For a horizontal surface, the deposition velocity can be calculated from the following equation 𝜐𝑑 = 𝜐𝑠 /(1 − ⅇ𝑥𝑝((−𝜐𝑠 𝐼)/𝑢∗ ) ) (17) In which 𝜐𝑠 , the settling velocity, can be calculated as follows: |𝜐𝑠 | = 𝐶𝑐 (4/3 (𝑔ⅆ𝑝 )/𝐶𝐷 ((𝜌𝑝 − 𝜌)/𝜌))1/2 (18) Where, CC is Cunningham coefficient, ⅆ𝑝 is the particle diameter (input), CD is the drag coefficient, 𝜌 is the density of air (1.21 kg/m3), 𝜌𝑝 is the density of the particle (2700 kg/m3).
The following equations are further calculations of the variables. 𝐶𝐶 = 1 + 𝜆/ⅆ𝑝 (2.514 + 0.8exp(−0.55 ⅆ𝑝 /𝜆)) (19) 𝜆 is the mean free length of air molecule, 𝜆 = 0.066 𝜇𝑚 𝐶𝐷 = 24/𝑅ⅇ𝑝 , 𝑅ⅇ𝑝 < 1 (20) 2/3 𝐶𝐷 = 24/𝑅ⅇ𝑝 (1 + 0.15 𝑅ⅇ𝑝 ), 1 < 𝑅ⅇ𝑝 < 1000 (21) Where, 𝑅ⅇ𝑝 is the Reynold number, in terms of 𝜈 the kinematic viscosity (1.51 x 10-5 m2/s). 𝑅ⅇ𝑝 = (𝜐𝑠 ⅆ𝑝 )/𝜈 (22) The variable u* from the deposition velocity equation is the friction velocity which can be calculated from: 1/2
𝑢∗ = (𝜈(ⅆ𝑈/ⅆ𝑦)|𝑦=0 ) (23) Where, U is the average air speed parallel to the surface, L is the plate length, 𝑈∞ is the freestream velocity. (ⅆ𝑈/ⅆ𝑦)|𝑦=0 = (0.074/(𝜌𝑎 𝜈))((𝜌𝑎 𝑈∞ 2 )/2) ((𝑈∞ 𝐿)/ 𝜈)−1/5 (24) While I from the deposition velocity equation is the integral term that can be calculated from: 2 3
𝐼 = 3.64 𝑆𝑐 (𝑎 − 𝑏) + 39 (25) Where, Sc is Schmidt’s number, a and b are calculated as follows: 𝑎 = 1/2 𝑙𝑛((10.92 𝑆𝑐 −1/3 + 4.3)3 /(𝑆𝑐 −1 + 0.0609)) + √3 tan−1 ((8.6 − 10.92 𝑆𝑐 −1/3 )/(√3 10.92 𝑆𝑐 −1/3 )) (26) 𝑏 = 1/2 𝑙𝑛((10.92 𝑆𝑐 −1/3 + 𝑟 + )3 /(𝑆𝑐 −1 + 7.669 × 10−4 (𝑟 + )3 )) + √3 tan−1 ((2𝑟 + − 10.92 𝑆𝑐 −1/3 )/ (√3 10.92 𝑆𝑐 −1/3 )) (27) 𝑟 + , a dimensionless variable can be calculated as follows: 𝑟 + = ⅆ𝑝 (2.6𝑢∗ )/2𝜈 (28) 𝑆𝑐 = 𝜈/𝐷 (29) The Brownian diffusivity coefficient is in terms of kB (Boltzmann coefficient), T (absolute temperature, 293 K), and 𝜇 the dynamic viscosity of air (1.81 x 10 -5 kg/ms). 𝐷 = (𝑘𝐵 𝑇𝐶𝐶 )/(3𝜋𝜇ⅆ𝑝 ) (30) kB = 1.38 x 10-23 The flux J can be calculated by Fick’s Law: 𝐽 = −(𝜀𝑃 + 𝐷)(ⅆ𝐶/ⅆ𝑦) (31) Where, D is the Brownian diffusion of the particle, 𝜀𝑃 is the particle eddy diffusivity in the boundary layer, C is the average particle concentration, y is the distance from the surface. Alternatively, it can be calculated by the following equation: 𝐽 = ∑𝑁 (32) 𝑖=1 𝑚𝑖 /𝐴𝑡 Where, m is the mass of the particle, N is the number of particles, A is the area of the sample, and t is the duration of the testing (e.g. 7 days x 24 hours x 3600 seconds). To get the concentration PM in the atmosphere, the following equation is used. Since 𝐶∞ is used in the calculations of 𝐶 + , the equations should be solved implicitly. 𝐶∞ = 𝐽|𝑦=0 /𝜐𝑑 (33) 3) MATLAB Code Development The equations used by the chosen models were written in a MATLAB code. The main input into the code were the properties of the particle and air such as the densities and the viscosities. The code also takes the diameter of the particles as an input and processes it through the two models chosen using the equations. Those give the deposition velocity and then calculates the concentration of the particles. Since the particles
behave differently according to their sizes, the diameters were separated to 0.5 – 2.5 µm, 2.5 – 4 µm, and 4 – 10 µm groups. As for the implicit equations, iterations were made to get a relatively close answer to the variable. The developed code is attached to the Final Report documents. III. RESULTS AND DISCUSSION Six test campaigns have been performed in order to assess the following four objectives: 1) Level of PM ambient concentrations in outdoor and indoor environments 2) Use of green light interferometry as new method for counting particles from passive samples 3) Available models for deposition velocities, applicable to passive samples 4) The overall applicability of passive samples as a tool to assess indoor and outdoor PM pollution levels These are discussed with more details in the following paragraphs, however because of the short duration of the study and a number of issues that occurred the overall conclusion is mostly based on qualitative criteria.
(a)
(b) Fig. 6 Measured Particle size distribution using the active measuring equipment (GRIMM Spectrometer) for all 7 sampling periods (see TABLE 1) a) normal scale and b) log scale for the scaled particle density.
A. Levels of Particulate Pollution Fig. 6 shows the scaled number concentration for specific diameters for all six campaigns. It is rather interesting that both outdoor (Tests A-D) and indoor (Tests E-F) campaigns agree on the fine size levels and differentiate as expected in the bigger particles. However, the distribution of the outdoor particles was expected to demonstrate a mode at coarse particles (i.e. 10μm and above) which would have shown that typical natural sources are dominating. On the contrary outdoor measurements show quite similar behavior with minimum seasonal variation (Tests A-D span from April to October). Note that outdoor measurements have been conducted at the rooftop of a building so the station is isolated from local sources (e.g. distinct traffic) and represents a wider area. The lack of a coarse mode can attributed to the different type of surface material than neighboring dry-arid areas (carbonates instead of quartz) and the intense construction activities which produce finer particles. The mass concentrations in all cases were considered high. PM10 values ranged from 80 to 180 μg.m-3 for the outdoor measurements while the PM2.5 from 20 to 80 μg.m-3. These values are lower than the National Ambient Air Quality Standards but still higher than the World Health Organization recommended values. The indoor PM10 and PM2.5 showed very small differences and around 20 μg.m-3, dominated by PM2.5. B. Green Light Interferometry as new Particles Counting Method The use of green light interferometry showed really promising in the preliminary tests. Typical counting methods so far were based on monochromatic pictures from optical microscopes or tedious SEM analysis. Interferometry is capable to add the 3D aspect in the measurements and thus improve the identification and sizing of particles. Fig. 7 shows the steps of analysis that the team has developed using an interferometer map as a starting point. The 3D map of particles is initially converted to 2D image in order to identify the particles and the “height” is added back to the calculation procedure.
Fig. 7 Steps of post-processing of a 3D map as collected from the green light interferometer (arbitrary sample).
A rather extended investigation on the specific samples for this study revealed the weak point of “specific wave length” interferometry. The reflectivity of the studied material is very
important and for this particular methodology cannot be unknown. On the other hand, the advantages of employing the interferometric analysis of such samples are still valid when using a white light instrument. In that case and even if a specific reflectivity of the material is low, the sample will still be visible if it reflect significantly on other wavelengths. Unfortunately, there was no white light interferometer available to try this assumption with the existing samples. C. Deposition Models The analysis concerning the deposition models concluded that the two most appropriate ones are the model Schneider as it has been adapted by Wagner and Assael, and the model of Lai as it is implemented by himself and Nazaroff. Both models have been compared (Fig. 8) and no significant variations have been observed at the indoor experiments. On the other side the outdoor tests showed no clear results, so it is necessary to extend the outdoor datasets for a conclusion; this is probably attributed to the ambient wind velocity dependence and more specifically the friction velocity which in any case is not strictly defined in the urban environment. IV. CONCLUSION Measurement of the ambient particulate pollution by passive sampling is strongly based on the overall levels of PM. Therefore, its use in dry and arid climates should be more robust and accurate than in temperate and green environments where there are local and regional PM sinks. This study focused on this aspect and showed that it is possible to apply passive sampling but new measuring and analysis protocols are necessary in order to address the new challenges. Probably this is equally important to all measuring techniques. More specifically the conclusions are: 1) High levels of ambient PM or isolated dust events are possible to overload the passive sampler and sustain impossible the use of the sample. 2) On the other side, this means that there is no need to expose the sampler for prolonged periods in order to get adequate sample. Specific tests showed it is possible to reduce the sampling period from a week (used so far in other studies) to 3-4 days. This increases the temporal resolution of the method 3) In the case, the composition of the sampled PM is unknown it is likely to get unreliable readings. Therefore, it is important that passive sampler’s application is followed, in some extent, by PM characterization. So the appropriate counting method can be selected. The present study successfully addressed the objectives set initially but also created new questions and more specific directions. Both can be summarized in the following two points: 1) Explore the use of white light interferometry for the analysis of PM passive samples, but also identify an appropriate standard or reference material for calibration
2) Outdoor and indoor PM concentration levels are high and therefore an extended campaign is necessary in order to characterize the sources and material of particulate matter ACKNOWLEDGMENT This project was made possible by UREP grant # [15 - 001 - 2 - 001] from the Qatar National Research Fund (a member of Qatar Foundation). The statements made herein are solely the responsibility of the authors. REFERENCES [1] J.P. Engelbrecht, E.V. McDonald, J.A. Gillies, R.K.M. Jayanty, G. Casuccio, A.W. Gertler, Characterizing mineral dusts and other aerosols from the Middle East - Part 1: Ambient sampling, Inhalation Toxicol, 21 (2009) 297-326. [2] V. Tsiouri, K.E. Kakosimos, P. Kumar, Concentrations, sources and exposure risks associated with particulate matter in the Middle East Area a review, Air Quality, Atmosphere & Health, (2014). [3] J. Wagner, D. Leith, Passive aerosol sampler. Part II: Wind tunnel experiments, Aerosol Sci. Technol., 34 (2001) 193-201. [4] J. Wagner, K. Naik-Patel, S. Wall, M. Harnly, Measurement of ambient particulate matter concentrations and particle types near agricultural burns using electron microscopy and passive samplers, Atmos. Environ., 54 (2012) 260-271. [5] D.K. Ott, N. Kumar, T.M. Peters, Passive sampling to capture spatial variability in PM10-2.5, Atmos. Environ., 42 (2008) 746-756. [6] M.J. Assael, D. Melas, K.E. Kakosimos, Monitoring Particulate Matter Concentrations with Passive Samplers: Application to the Greater Thessaloniki Area, Water Air Soil Poll, 211 (2010) 395-408. [7] D.K. Ott, W. Cyrs, T.A. Peters, Passive measurement of coarse particulate matter, PM10-2.5, J. Aerosol Sci, 39 (2008) 156-167. [8] J. Wagner, D. Leith, Passive aerosol sampler. Part I: Principle of operation, Aerosol Sci. Technol., 34 (2001) 186-192. [9] T. Schneider, M. Bohhard, A. Gudmundsson, A Semiempirical Model for Particle Deposition onto Facial Skin and Eyes. Role of Air Currents and Electric Fields, Aerosol Science, 25 (1994) 583-593. [10] J. Wagner, D. Leith, Passive Aerosol Sampler. Part I: Principle of Operation, Aerosol Science and Technology, 34 (2010) 186-192. [11] D.K. Ott, W. Cyrs, T.M. Peters, Passive measurement of coarse particulate matter, Journal of Aerosol Science, 39 (2008) 156-167. [12] A. Lai, W. Nazaroff, Modeling Indoor Particle Deposition from Turbulent Flow onto Smooth Surfaces, Aerosol Science, 31 (2000) 463-476. [13] B. Zhao, J. Wu, Modeling particle deposition from fully developed turbulent flow in ventilation duct, Atmospheric Environment, 40 (2006) 457-466. [14] T. Hussein, A. Hruška, P. Dohányosová, L. Džumbová, J. Hemerka, M. Kulmala, J. Smolík, Deposition rates on smooth surfaces and coagulation of aerosol particles inside a test chamber, Atmos. Environ., 43 (2009) 905914.
Fig. 8 Scatter plots for comparison of passive vs active measurements for the outdoor (circles) and the indoor (triangles) campaigns, and using the two formulations Assael et al [6] (full symbols) and the LaiNazarroff [12] (empty symbols)
