3506B24 Final Report
Download 11.49 Mb.
|
Figure 19: TEOM flow paths with installed Sample Equilibration System (SES). In addition to the relative humidity control (achieved 15 ±3 %), the sample air was heated to 40 °C inside the sensor unit (labeled “case” and “air” temperature, while the “cap” temperature remained uncontrolled), where the conditioned sample air stream passed through a filter made of Teflon-coated borosilicate glass fiber, which is weighed every two seconds. The difference between the filter’s current weight and the filter’s initial weight (as automatically measured by the instrument after the installation of the filter) gives the total mass of the collected PM. These instantaneous readings of total mass were smoothed exponentially to reduce noise. In general, the mass rate is calculated by taking the change in the smoothed total mass between the current reading and the immediately preceding one and expressing as mass rate in g s-1. This mass rate is also smoothed exponentially to reduce noise. Finally, the mass concentration (µg m-3) is computed by dividing the mass rate by the flow rate (corrected to EPA’s STP and expressed in ambient m3 s-1), and then multiplying the result by 106 to convert from g m-3 to µg m-3. All measurement and temperature functions of the instrument are controlled by a dedicated microcontroller. This computer was digitally interfaced with the external data collection systems (ESC data logger model 8816). The tapered element at the heart of the mass detection system, is a hollow tube, clamped on one end and free to oscillate at the other. The exchangeable Teflon-coated glass fiber filter is placed over the tip of the free end. The 3 lpm sample stream is drawn through this filter, and then down the tapered element. The tapered element oscillates precisely at its natural frequency, much like the tine of a tuning fork, being in essence, a hollow cantilever beam with an associated spring rate and mass. An electronic control circuit senses this oscillation and, through positive feedback, adds sufficient energy to the system to overcome losses. An automatic gain control circuit maintains the oscillation at a constant amplitude. A precision electronic counter measures the oscillation frequency within a 2 s sampling period. As in any spring-mass system, if additional mass is added, the frequency of the oscillation decreases. In a spring-mass system the frequency follows the equation: f = (K / M)0.5, where f is the frequency (radians/s), K the spring rate, and M is the mass. In actual operation, the TEOM always measures the entire mass of the system using the equation: M = K0 / f2. At the end of the instrument’s 30 min flow and temperature equilibration period, the monitor averages the frequency for a short period and uses this frequency to compute the baseline mass. Until the next time the unit is reset or taken out of its data collection mode, the frequency is sampled every two seconds and the system mass is calculated. The difference between this mass and the baseline mass is the mass change of PM collected on the filter cartridge. The TEOM’s detection limit typically averaged 0.8 ±0.2 μg m-3, determined from sampling HEPA filtered zero air for at least 1 h, and its precision and accuracy levels were estimated from previous inter-comparisons within ±10 % each [e.g. Baumann et al. 2003a]. During the study, the TEOM’s signal was averaged over the discrete PCM filter measurements and compared as illustrated in Figure 20. Between January 20 and May 31, 2003, 31 observations were made with sample durations ranging between 5 and 24 h that are subject to this comparison. The TEOM appears to have a positive offset (see intercept of +2.4 µg m-3), which possibly points to a positive artifact due to adsorption of water vapor and other condensable semi-volatile species, since the PCM Teflon filter samples were i) collected downstream from gas-removing denuders, and ii) desiccated before their gravimetric mass determination (see section below). The 1-σ uncertainty of this offset is ±1.3 µg m-3 (P-value = 0.064), and the corresponding uncertainty of the slope of 0.86 is ±0.07 at 95 % confidence level of the assumed normal distribution. If neither method is assumed to have an offset, the TEOM mass concentration averages 2 ±0.02 % low relative to the PCM filter method, at a correlation coefficient R2 = 0.834. 
Figure 20: Linear regression of the TEOM PM2.5 mass concentrations averaged over 31 discrete PCM gravimetric mass concentrations from desiccated Teflon filter substrates collected at OLC between January 20 and May 31, 2003. 5.2 VOC Sampling Sampling of Volatile Organic Compounds (VOC) took place by means of evacuated whole-air stainless-steel canisters with internal volume of 2 l. The cans were prepared, shipped and analyzed by the Department of Chemistry at the University of California, Irvine under the direction of Prof. Don Blake. In order to capture potential differences in fuel type and local background air pollution loads, sampling took place both in the Fort Gordon – Augusta area and the Fort Benning - Columbus area. OLC served as the background or receptor site for the investigation of Fort Benning burns, whereas the FAQS site at Riverside Park (RP) north of Augusta, at 15 Dolphin Way, Evans, GA 30809, about 20 km north-northeast from Fort Gordon served as a similar reference for the Fort Gordon burns. A sampling strategy had been developed that captured the direct emissions from the open flaming and subsequent smoldering stages of various burns at both locations. The strategy included the collection of samples upwind and downwind of the burn source; the upwind sample was particularly important reference the source samples and determine the emission strength of certain VOC species. The sampling strategy took furthermore into account the uncertainty of predicting the upwind and downwind locations for each of the studied burns, and allowed the identification of samples worth analyzing. As a result, 92 can samples were analyzed that had been collected between February 3rd and April 30th 2003. In order to minimize contamination and unwanted artifacts, a guideline with detailed instructions for the filling of the evacuated canisters accompanied the samples into the field. Table 9 summarizes the number of samples taken at different locations, i.e. upwind, or before the ignition of fires, directly at the source of the prescribed burns at either the flaming or smoldering stage of the fires, downwind from the fires, and at both the comprehensive measurement sites, i.e. at the Oxbow Meadows Environmental Learning Center (OLC) near Columbus and at the Riverside Park (RP) site near Augusta. Note that of the total 92 samples, one was taken from the emissions of the burning torch used at Fort Benning, on April 15th. Table 9: Distribution of the total number of 92 VOC samples (incl. 1 torch sample) taken at specified locations before and during the conduct of the prescribed burns.
Table 10 lists all 42 VOC species that were analyzed for the total of 92 can samples taken between February 3rd and April 30th 2003. In addition to the listed VOC species, all samples were also analyzed for [CO], [CO2] and [CH4]. Note that ethyne (acetylene) and 1,3-butadiene (potential carcinogen) are grouped under the alkenes category, although they belong to the family of alkynes and dienes, respectively. Table 10: Summary of 42 VOC species detected and quantified during the whole-air canister sampling conducted between February and April 2003 (excl. CO, CO2, CH4).
Based on previous successful field experiments in urban environments [e.g. Blake and Rowland 1995; Chen et al. 1999, Colman et al. 2001], standard operating procedures are in place for the sample collection and laboratory analyses. Laboratory analyses follow principle methods of gas chromatography (GC) with flame ionization detection (FID), electron capture detection (ECD), and mass spectrometry (MS). Compounds are analyzed by using a loop that is filled with glass beads and immersed in liquid nitrogen in order to pre-concentrate the less volatile sample components while the more volatile components (e.g. nitrogen, oxygen, methane, and argon) are pumped away. The flow is regulated by a mass flow controller and be kept below 500 cm3 min-1 to ensure that the less volatile components are completely trapped. The pre-concentration loop is isolated before being warmed in a hot water bath (at 80ºC) to re-volatilize the gases. The content of the loop is flushed into a helium carrier gas flowing with 48 psig (~430 kPa) head pressure. The sample flow is reproducibly split into five streams, with each stream sent to a different column-detector combination. By using sub-ambient temperatures for each column, the sample is cryogenically recollected in each chromatographic system after being split into the five streams. The reproducibility of the split ratio is monitored by examining the calculated mixing ratios for a compound that gives a large signal, has good chromatographic characteristics, and is quantified on multiple detectors. Further details are given by Colman et al. [2001]. The accuracy for the above analyses is ±5 % and the analytical precision is ±3 % or 3 pptv whichever is larger. Note, that as concentrations approach the DL the precision becomes worse, hence an actual mixing ratio is stated in addition to the % precision estimate. For halocarbons the accuracy ranges from 1-20% depending on the gas and the analytical precision also depends on the gas, in part because there is such a large range in concentrations. For gases above 50 pptv the measurement precision is 1-3% while gases between 50 pptv and 10 pptv are 3-5%. Gases below 10 pptv are 5% or 0.1 pptv, whichever is larger. During the study, accuracy was assessed for the CO measurements at OLC by comparison of the GC analyses of the whole-air can samples with the instantaneous 1 min measurement from the continuous gas analyzer. The comparison is shown in Figure 21 as linear regression, yielding an average 11 ±3 % lower values for the can CO when forced through zero. When the intercept is not fixed, however, the can CO seems 16 ±3 % higher than the instantaneous (1 min) IR absorption signal of the continuous gas analyzer, which then appears to have a 95 ±12 ppbv offset. This disagreement is likely a combination of i) the IR analyzer’s sensitivity to water vapor, ii) possible reactions depleting CO in the can during the time between sampling and analysis (which could be several weeks), and iii) inaccuracies in the recordings of time when the cans had been filled. 
Figure 21: Linear regression of CO measured by a subset of 17 whole-air canister samples collected at OLC between January 20 and May 31, 2003, with coincident 1 min average CO measured by the continuous IR absorption analyzer. 5.3 Laboratory Analyses 5.3.1 PCM Samples for Major Ions and Carbon The following describes the laboratory procedures and quality assurance and control (QAQC) measures that were executed for the analyses of the discrete samples collected by means of the Particle Composition Monitor (PCM) and the High-Volume Samplers (HVS) described in section 5.1.1 above. - Gravimetry The PM2.5 mass concentrations were gravimetrically determined from the mass increase of Teflon membrane filters before and after sampling. Un-ringed Teflon-membrane filter (Zeflour P5PJ047, Gelman, Ann Arbor, MI) were used here. The filter has a 2m pore size, and a 47mm diameter. Each filter was given sufficient time (at least one month) to equilibrate to constant levels of relative humidity (33% 3%) and temperature (21C 1C) inside a controlled clean air room (class 1000), that has the micro-balance installed. The clean air room is part of the AREC analytical laboratory in Atlanta, and is kept under slightly positive pressure by temperature and humidity controlled filtered air. Positive pressure is maintained by introducing a small amount of ambient air into the otherwise closed-loop circulation. The air is filtered by citric acid and activated charcoal beds prior to a high efficiency particle air (HEPA) filter. The filtered air enters the room via a membrane diffusion ceiling plenum creating a near-laminar flow at the work surface. Mass had to be recorded before and after a filter was used for sampling, and because of the specific properties of the filter material humidity and temperature had to be controlled. The following is the standard operating procedure (SOP) exercised for the mass measurements made here. The Teflon filters were removed from their original packaging in the clean room using Teflon tweezers. They were placed into small individual partially open petri-dishes labeled with a piece of autoclave tape. The filters remained in a climate-controlled clean room, where they were unsealed, for no less than one month. The humidity inside the room is maintained at 33 3%, the temperature at 21 1C. The filters were serialized. Each subset of filters consisted of the filter to be weighed on a particular day. The piece of tape was marked with the weighing date and lot number. The filter was now serialized. Masses were determined using a Mettler Toledo MT5 Electronic Balance. Each time that the balance was turned on, an internal automatic calibration was preformed. Immediately after the balance was zeroed, two standards (200mg each) were repeatedly weighed. If the weighed standards lie outside of the accepted precision of the balance; then the internal calibration and zeroing procedure is repeated and the standards are weighed again. Once the balance had “warmed up”, three filters were weighed three consecutive times, alternating in a round robin between the three. NOTE: The balance is activated with nothing on the scale after every third weighing to recheck the zero weight. If the weight of the balance is not zero, then the balance is re-zeroed. At the entrance to the balance chamber a radioactive strip has been placed to minimize the amount of static electricity. The strip is oriented so that when each filter is placed in to the balance chamber the strip is close enough to act on the filter; but not to hinder the filter. Each filter was placed in its own filter holder and stored and shipped in accordance with the SOP for storage and shipping. When the filters had returned they were weighed as soon as possible. Because of our particular sampling procedure 2 Teflon filters were associated with each day of sampling. So just back from sampling each of the two filters, and the associated blank got it first post sampling weight. All filters were then placed uncovered into a desiccator. The desiccator contains anhydrous calcium sulfate. It is opened only during loading and unloading; and is not evacuated. The second and third weights were taken after two consecutive periods of 1 to 2 days in the desiccator. Associated sample filters and blanks were always weighed at the same time. After verifying that the last two weights lied within 10%, the Ch1 filter of the Ambient PCM and half of the Ch1 filter of the Source PCM were extracted and analyzed via IC according to the below. Both sample and field blank filters were treated the same. The Mettler Toledo MT5 Electronic Balance has an experimentally determined instrumental detection limit of 1.2 .02 μg. This was determined by repeated weighing of the same item; the standard deviation of those weights is considered to be the detection limit. An instrumental precision of 0.37 % of 1μg was determined by repeated weightings of a 1 mg standard weight. The instrumental accuracy was determined by weighing standards (ranging from 1 – 500 mg) multiple times and creating a linear plot. The slope of this line times 100 is considered to be the percent accuracy yielding 99.99956 %. To assure this level of accuracy and precision special care must be taken when weighing the Teflon Filters. Teflon has a tendency to build up a static charge that can impede weighing. To eliminate the potential static charge on from the filter before weighing two Staticmaster Ionizing Units (Model number 2U500) are used. One adhered inside the balance chamber to the roof; and the other placed just outside the balance chamber door. The optimal distance the filter should be place from the unit is 1 to 1.25 inches. And each unit has an effective lifetime of about six months. - Ion Chromatography (IC) The concentrations of selected anions and cations collected on both sodium-carbonate coated glass fiber (GSC) and phosphorous-acid-impregnated cellulose fiber (PPA) filters, as well as Na2CO3- and phosphorous-acid-coated denuders (DSC, DPA) were determined by extraction and subsequent IC analysis. Filters and denuders were extracted with distilled deionized water (DDW) according to a specific SOP. The quantitative analysis requires the determination of field blanks to account for systematic sampling artifacts. All IC measurements were recorded in the IC Log Book and the appropriate Sample and Field Blank Logs. In addition all chromatograms were cross-referenced to the IC Log Book and filed. Findings from additional laboratory and field tests performed during the U.S. EPA sponsored Atlanta Supersite Experiment in August 1999, helped quantify the positive interference that ambient NO2 and O3 has on the nitric acid collection with Na2CO3-denuders. These findings were instrumental in defining the calculations and presumptive treatment of the individual denuder and filter extracts described below. The water soluble ion content of the various extracts was quantified applying a dual-channel Dionex DX-500 ion chromatograph with two separate EG40 eluent generators; KOH for anions, methane-sulfonic acid (MSA) for cations, controllable to within 0.1 and 100 mM, and Ion-Pac analytical columns AG11-HC for anions and CS12A for cations, both in the 2 mm ID micro-bore format. Each channel operates a self-regenerating SRS-ULTRA suppressor in external DDW regeneration mode, a CD20 conductivity detector, and a GP50 gradient pump. The applied micro-bore system allows economical analyte flow rates of 0.25 ml/min for cations, and 0.35 ml/min for anions. DDW is supplied by a Barnstead E-Pure at a resistivity of 18.0 0.3 M or better, and fed directly to the EG40. Degassing is performed on-line immediately after the eluent is added to the DDW well upstream of the injector. The extracts are submitted to the IC in pre-cleaned 5 ml vials and injected via Dionex Autosampler. As part of the SOP, standard solution is prepared for each ion that is being examined. A total of ten different standard concentrations are prepared. The concentrations are prepared so that their concentrations lie on either side of the expected ion sample concentration. For example, a very typical sulfate sample concentration is 15 g/ml. The standard concentrations that are used to calibrate the IC for sulfate, therefore are 1.7, 3.3, 5.7, 8.2, 11.4, 14.5, 19.1, 25.3, and 31.2 g/ml (incl. 0 from pure DDW). On average 5 of the 10 standards are run per day; sometimes more. A calibration plot is generated, and the produced data is adjusted for the actual concentration of the standards versus the IC calculated standard concentration. Once the standards have been run, approximately 1 ml of sample is injected into the IC for testing. The IC uses a 10 l sample volume; so the rest is used as a rinse and is then expelled to waste. Since the standards have already been run, and the peak range for specific ions has already been recorded in the computer, all that remains is to apply the calibration equation to the PeakNet output. The output gives the concentration of each ion in area and height. Table 11 summarizes the accuracy, relative and absolute precision achieved for the IC analyses of the samples collected during the study. Table 11: Accuracy, relative and absolute precision (P) achieved for the laboratory IC used during the analyses of the samples collected between January 20 and May 31, 2003.
- Elemental and Organic Carbon (EC, OC) The elemental carbon and organic carbon (EC/OC) contents of the collected PM2.5 samples were determined from the Q and XQ filter samples by the thermal-optical transmittance (TOT) method. This method is based on the thermo-optical transmittance properties of a quartz filter containing carbon residue. Figure 22 shows a schematic of the instrumentation used in this analysis [Birch and Cary, 1996]. Prior to sampling, the sample Q and XQ filters (both Pallflex #2500 QAT-UP) had been baked at 600 oC for at least 2 hours and stored in polystyrene petri-dishes at 1 °C until sampling. All sampled quartz filters, including the field blanks, were stored in same petri-dishes at –19 °C until TOT analysis, which is in accordance with NIOSH. TOT analysis is performed on a square 1 x 1.45 cm2 section punched out of the 47 mm diameter quartz filter. The following briefly describes the TOT procedure. 
Figure 22: Schematic of the thermal-optical transmission (TOT) instrument for the analysis of elemental and organic carbon (EC+OC) in PM25 quartz filter samples [Birch and Cary, 1996]. Figure 23 shows an example of the instrument output. The four traces appearing in this figure correspond to temperature, filter transmittance, and detector response at two different gain settings (FID1 and 2). The analysis proceeds essentially in two stages. In the first, organic carbon is volatilized from the sample in a pure helium atmosphere as the temperature is stepped from initially ~45 °C to 340 °C, 475 °C, 615 °C to 870 °C within ~4.5 minutes. Evolved carbon is catalytically oxidized to CO2 in a bed of granular MnO2 (held at ~900 °C), reduced to CH4 in an Ni/firebrick methanator (~500 °C) and quantified as CH4 by a Flame Ionization Detector (FID). The FID operates through the use of a hydrogen flame that is placed in an electrical field. A potential difference is created on ionization of particles in this field, which is proportional to the number of molecules generated in the analysis. During the second stage of the analysis, pyrolysis correction and EC measurement are made. The oven temperature is reduced, an oxygen (10%)-helium mix is introduced, and the oven temperature is then raised from 550 °C, 640 °C, 720 °C, 790 °C, and 870 °C to 890 °C within another period of approximately 4 minutes. As oxygen enters the oven, pyrolytically generated EC is oxidized to CO2 and a concurrent increase in filter transmittance occurs as the darkened carbon (soot) is driven off. A He/Ne laser light (633 nm) is passed through the filter to monitor transmittance during the course of the run. Correction for the pyrolytically generated EC (or char) contribution is accomplished by measuring the amount of char oxidation required to return the filter to its initial transmittance value. The point at which the filter transmittance reaches its initial value is defined as the “split” between organic and elemental carbon. Carbon evolving prior to the split is considered “organic” (incl. carbonate), and carbon volatilized after the split and prior to the peak used for instrument calibration (final peak) is considered “elemental”. 
Figure 23: Thermogram of a quartz filter sample with front oven temperature, two differently amplified FID signals, and the He/Ne laser transmittance signal. Note, the final peak is from the internal CH4 calibration. As part of routinely performed overall instrument performance checks, standard sucrose stock solution is made using 10.00 ± 0.01 g of sucrose weighed using an analytical balance and transferred into a 1000 ml volumetric flask, which is filled with DDW to marked level. This yields a sucrose concentration of 4.2067 µg C / µl solution. This stock solution is replaced every 6 months. Using an appropriate syringe, 2 µl of standard sucrose solution is spiked onto a blank punch. The spiked punch is placed on the sample boat and into the TOT analyzer while in idle mode to dry. After 30 min of drying, the TOT analysis program is started and the front oven temperature starts to increase as programmed. The above steps are repeated for spiking with 4 and 6 µl of standard sucrose solution. The analyzer’s performance was validated if the following range of results in µg cm-2 was achieved for the 2, 4, and 6 μl standard solution, respectively, assuming a filter punch area of 1.5 cm2: 4.49-6.73, 9.53-12.91, and 15.14-18.52 μg cm-2, centered nominally at 5.61, 11.22, and 16.83 μg cm-2, respectively. The ranges varied with nominal concentration due to the uncertainty in the determination of 2, 4 and 6 μl standard solution, i.e. ±20, ±15, and ±10 % of the nominal values, respectively. According to EPA protocol, when a reference standard for a complex mixture of OC in the particle-phase is lacking, instrument precision is estimated as bias by comparison of two discrete methodologies using ambient data. Under the assumption of homogeneous particle deposition onto the filter substrate during sampling, the comparative analysis and evaluation of two side-by-side measurements meets this requirement. Hence, the precision for EC and OC (but also for the ionic species) is: Bias = 1/n  [((M1i-M2i)/((M1i+M2i)/2))] . 100
where M1i and M2i are the ith measurement of the two methodologies (M1 and M2, here side-by-side filter samples) being subjected to comparison. The use of the average of the two methodologies in computing bias recognizes that a primary standard is not available. Precision measurements were made on a subset of samples during the study period. The measurement uncertainties for EC, OC and SVOC (i.e. OC from the XAD-coated backup adsorbing filters) expressed as bias are listed in Table 12 and averaged 14 ±15 %, 8 ±13 %, and 16 ±15 %, respectively. Table 12: Precision (bias) assessment for elemental carbon (EC), organic carbon (OC), and semi-volatile OC (SVOC, from XAD4-coated adsorbing quartz backup filter behind denuded quartz front filter) from side-by-side measurements performed at OLC during the study period.
An evaluation of the field blanks was also performed in order to estimate the lower detection limits for the EC, OC, SVOC and all ionic species concentrations of fine PM, encompassing the overall measurement method. The minimum detection limit (MDL) is defined as a statistically determined value above which the reported concentration can be differentiated, at a specific probability, from a zero concentration. The MDL for the discrete samples extends from the instrumental DL by including the handling, shipping and storing of the sample collection media, i.e. the quartz fiber filters in the case of the TOT analysis, and all other filters plus denuders for the IC analyses. Information on the overall effect of these procedures can be gained by the field blanks that were carried out on almost every sampling day. The MDL is then given by the following equation: MDL =  + s t (n-1,1-
where is the average of the field blank analyses, s is the standard deviation of the field blank analyses, t is the students t value appropriate to a 95% confidence level and a standard deviation estimate with n-1 degrees of freedom. Table 13 includes the results of the field blank analysis as part of all data quality indicators associated with the PCM gas- and particle-phase data, by means of denuder efficiency, detection limit, precision (expressed as bias from side-by-side runs), and accuracy estimates from previous field and laboratory assessments.
Table 13: Summary of PCM Data Quality Indicators (DQI) for gas- and particle-phase species determined during the analysis of samples collected between January and May 2003. D (pa)… phosphorous acid-coated denuder; D (sc)… sodium carbonate-coated denuder; G… Gelman 61631 glass fiber filter, pre-baked at 580°C for 2 h, P… Whatman 41 cellulose [paper] filter; Q… Pallflex 2500 QAT-UP Quartz fiber filter, pre-baked at 600°C for 2 h; T… Zeflour P5PJ047, un-ringed Teflon membrane, 2μm pore size; XAD… XAD-4, porous macro-reticular, non-polar, polystyrene-divinyl-benzene resin (725m2g-1). * from linear regression with cont. SO2 UV absorption measurements. ** from NIST standards. *** from linear regression with TEOM measurements. # from linear regression with PILS measurements [Baumann et al., 2002]. Gas Phase
Download 11.49 Mb. Share with your friends:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||