August 2001 Environmental Technology Verification Report Met One Instruments BAM 1020 Particle Monitor Prepared by Baltelle . . . Putting Technology To Work Battelle Under a cooperative agreement with oEPA U.S. Environmental Protection Agency ETVElV EtV ------- I lll ENVIRONMENTAL TECHNOLOGY VERIFICATION PROGRAM J ETV wEPA ETV Joint Verification Statement Bafteiie U.S. Environmental Protection Agency putting Technotogy To Work TECHNOLOGY TYPE: Continuous Ambient Fine Particle Monitor APPLICATION: MEASURING FINE PARTICULATE MASS IN AMBIENT AIR TECHNOLOGY NAME: BAM 1020 COMPANY: Met One Instruments, Inc. ADDRESS: 1600 Washington Blvd. PHONE: 541-471-111 Grants Pass, OR 9 7526 FAX: 541-471-7116 WEB SITE: http://www.metone.com E-MAIL: info@metone.com The U.S. Environmental Protection Agency (EPA) has created the Environmental Technology Verification (ETV) Program to facilitate the deployment of innovative or improved environmental technologies through performance verification and dissemination of information. The goal of the ETV Program is to further environmental protec- tion by substantially accelerating the acceptance and use of improved and cost-effective technologies. ETV seeks to achieve this goal by providing high-quality, peer-reviewed data on technology performance to those involved in the design, distribution, financing, permitting, purchase, and use of environmental technologies. ETV works in partnership with recognized standards and testing organizations; with stakeholder groups that consist of buyers, vendor organizations, and permitters; and with the full participation of individual technology developers. The program evaluates the performance of innovative technologies by developing test plans that are responsive to the needs of stakeholders, conducting field or laboratory tests (as appropriate), collecting and analyzing data, and preparing peer-reviewed reports. All evaluations are conducted in accordance with rigorous quality assurance protocols to ensure that data of known and adequate quality are generated and that the results are defensible. The Advanced Monitoring Systems (AMS) Center, one of six technology centers under ETV, is operated by Battelle in cooperation with EPA's National Exposure Research Laboratory. The AMS Center has recently evaluated the performance of continuous monitors used to measure fine particulate mass and species in ambient air. This verification statement provides a summary of the test results for the Met One BAM 1020 ambient fine particle monitor. ------- VERIFICATION TEST DESCRIPTION The objective of this verification test is to provide quantitative performance data on continuous fine particle monitors under a range of realistic operating conditions. To meet this objective, field testing was conducted in two phases in geographically distinct regions of the United States during different seasons of the year. The first phase of field testing was conducted at the ambient air monitoring station on the Department of Energy's National Energy Technology Laboratory campus in Pittsburgh, PA, from August 1 to September 1, 2000. During the period, daily PM25 concentrations ranged from 61 (ig/m3 to 36.2 (ig/m3, with an average of 18.4 |_ig/nr\ The second phase of testing was performed at the California Air Resources Board's ambient air monitoring station in Fresno, CA, from December 18, 2000, to January 17, 2000. During this period, daily PM2 5 concentrations ranged from 4.9 (ig/m3 to 146 (ig/m3, with an average value of 74.0 (ig/m3. Specific performance characteristics verified in this test include inter-unit precision, accuracy and correlation relative to time-integrated reference methods, effect of meteorological conditions, influence of precursor gases, and short-term monitoring capabilities. The BAM 1020 reports measurement results in terms of PM2 5 mass and, therefore, was compared with the federal reference method (FRM) for PM2 5 mass determination. Additionally, comparisons with a variety of supplemental measurements were made to establish specific performance characteristics. Quality assurance (QA) oversight of verification testing was provided by Battelle and EPA. Battelle QA staff conducted a data quality audit of 10% of the test data, and performance evaluation audits were conducted on the BGI FRM samplers used in the verification test. Battelle QA staff conducted an internal technical systems audit for Phase I and Phase II. EPA QA staff conducted an external technical systems audit during Phase II. TECHNOLOGY DESCRIPTION The BAM 1020 is a beta attenuation monitor that measures the concentration (mg/m3) of particulate matter in ambient air. The BAM 1020 may be equipped with a sharp cut cyclone PM2 5 or a WINS PM2 5 sampling inlet for automatic monitoring of finer particulate matter. The BAM 1020 monitor can also be configured to monitor total suspended particulate matter. An internal data logger allows up to six additional air quality or meteorological measurements. At the beginning of the sampling period, beta ray transmission is measured across a clean section of filter tape. This tape is mechanically advanced to the sampling inlet. Particulate matter is drawn into the sample inlet and deposited on the filter paper. At the completion of the sampling period, the filter tape is returned to its original location and the beta ray transmission is remeasured. The difference between the two measurements is used to determine the particulate concentration. The mass density is measured using the technique of beta attenuation. A small 14C beta source (60 iCi) is coupled to a detector that counts the emitted beta particles. The filter tape is placed between the beta source and the detector. As the mass deposited on the filter tape increases, the measured beta particle count is reduced according to a known equation. The BAM 1020 consists of a detector/ logger, pump, and sampling inlet. Each of these components is self-contained and may be disconnected for servicing or replacement. The BAM 1020 is designed to mount in a temperature-controlled enclosure. The sampling inlet is designed to mount through the roof of the enclosure. The BAM 1020 operates at 100 to 230 volts alternating current and is 310 mm high x 430 mm wide x 400 mm deep. All operations of the unit are displayed with an 8 line by 40 character display. VERIFICATION OF PERFORMANCE Inter-Unit Precision: During Phase I, the regression results from duplicate BAM 1020 monitors (Monitor 2 vs. Monitor 1) showed r2 values of 0.873 and 0.986, respectively, for the hourly data and the 24-hour averages. The slopes of the regression lines were 0.932 (0.027) and 0.973 (0.044), respectively, for the hourly data and 24-hour averages; and no statistically significant intercept was observed in either case at 95% confidence. The calculated coefficient of variation (CV) for the hourly data was 20.6%; and, for the 24-hour data, the CV was 9.5%. During Phase II, the regression analysis showed r2 values of 0.991 and 0.999, respectively, for the hourly data and the 24-hour averages. The slopes of the regression lines were 1.011 (0.007) and 1.018 (0.011), respectively, for the ------- hourly data and 24-hour averages; and the intercepts were -0.0016 (0.0007) mg/m3 and -0.0022 (0.0010) mg/m3, respectively. The calculated CV for the hourly data was 9.9% and for the 24-hour data the CV was 6.4%. Comparability/Predictability: During Phase I, comparisons of the 24-hour averages with PM2 5 FRM results showed slopes of the regression lines for Monitor 1 and Monitor 2 of 1.169 (0.152) and 1.142 (0.138), respectively; and these slopes were statistically different from unity at the 95% confidence level. The regression results show r2 values of 0.909 and 0.921 for Monitor 1 and Monitor 2, respectively. During Phase II, comparison of the 24-hour averages with PM2 5 FRM results showed slopes of the regression lines for Monitor 1 and Monitor 2 of 1.09 (0.08) and 1.11 (0.08), respectively; both statistically different from unity at 95% confidence. No statistically significant intercept was observed in either case at the 95% confidence level. The regression results show r2 values of 0.964 and 0.967 for Monitor 1 and Monitor 2, respectively. Meteorological Effects: Multivariable analysis of the 24-hour average data for Phase I showed that the vertical wind speed, the relative humidity, and the solar radiation all had a statistically significant influence on the results of Monitor 1 at the 90% confidence level. Similarly, vertical wind speed, and the ambient air temperature at both 2 meters and 10 meters influenced the results of Monitor 2 relative to the FRM at the 90% confidence level. Under typical conditions during Phase I, the combined effect of these paramters was approximately 7% or less. Multivariable analysis of the 24-hour average data for Phase II showed that relative humidity had a statistically significant influence on the readings of both monitors relative to the FRM values at 90% confidence. Under typical conditions during Phase II, the effect was less than 1%. Influence of Precursor Gases: During Phase I, multivariable analysis of the 24-hour average data showed that none of the measured precursor gases had an influence on Monitor 1 at the 90% confidence level, but hydrogen sulfide had a statistically significant, but practically negligible, influence on Monitor 2. During Phase II, multi- variable analysis of the 24-hour average data indicated that none of the measured gases had an effect on either monitor at the 90% confidence level. Short-Term Monitoring: In addition to 24-hour FRM samples, short-term sampling was performed on a five- sample-per-day basis. The BAM 1020 results were averaged for each of the sampling periods and compared with the gravimetric results. Linear regression of these data showed slopes of 1.13 and 1.15, respectively, for Monitor 1 and Monitor 2. The intercepts of the regression lines were 0.002 and 0.000 mg/m3, respectively; and the r2 values were 0.939 and 0.936, respectively. Other Parameters: No operating problems arose, and no maintenance was performed on either monitor during testing. Gabor J. Kovacs Date Gary J. Foley Date Vice President Director Environmental Sector National Exposure Research Laboratory Battelle Office of Research and Development U.S. Environmental Protection Agency NOTICE: ETV verifications are based on an evaluation of technology performance under specific, predetermined criteria and the appropriate quality assurance procedures. EPA and Battelle make no expressed or implied warranties as to the performance of the technology and do not certify that a technology will always operate as verified. The end user is solely responsible for complying with any and all applicable federal, state, and local requirements. Mention of commercial product names does not imply endorsement. ------- August 2001 Environmental Technology Verification Report ETV Advanced Monitoring Systems Center Met One Instruments BAM 1020 Particle Monitor by Kenneth Cowen Thomas Kelly Basil Coutant Karen Riggs Battelle Columbus, Ohio 43201 ------- Notice The U.S. Environmental Protection Agency (EPA), through its Office of Research and Development, has financially supported and collaborated in the extramural program described here. This document has been peer reviewed by the Agency and recommended for public release. Mention of trade names or commercial products does not constitute endorsement or recommendation by the EPA for use. ii ------- Foreword The U.S. EPA is charged by Congress with protecting the nation's air, water, and land resources. Under a mandate of national environmental laws, the Agency strives to formulate and implement actions leading to a compatible balance between human activities and the ability of natural systems to support and nurture life. To meet this mandate, the EPA's Office of Research and Development provides data and science support that can be used to solve environmental problems and to build the scientific knowledge base needed to manage our ecological resources wisely, to understand how pollutants affect our health, and to prevent or reduce environmental risks. The Environmental Technology Verification (ETV) Program has been established by the EPA to verify the performance characteristics of innovative environmental technology across all media and to report this objective information to permitters, buyers, and users of the technology, thus substantially accelerating the entrance of new environmental technologies into the marketplace. Verification organizations oversee and report verification activities based on testing and quality assurance protocols developed with input from major stakeholders and customer groups associated with the technology area. ETV consists of six technology centers. Information about each of these centers can be found on the Internet at http://www.epa.gov/etv/. Effective verifications of monitoring technologies are needed to assess environmental quality and to supply cost and performance data to select the most appropriate technology for that assess- ment. In 1997, through a competitive cooperative agreement, Battelle was awarded EPA funding and support to plan, coordinate, and conduct such verification tests for "Advanced Monitoring Systems for Air, Water, and Soil" and report the results to the community at large. Information concerning this specific environmental technology area can be found on the Internet at http://www.epa.gov/etv/07/07_main.htm. iii ------- Acknowledgments The authors wish to acknowledge the support of all those who helped plan and conduct the verification test, analyze the data, and prepare this report. In particular we would like to thank the staff at the Department of Energy's National Energy Technology Laboratory, including Richard Anderson, Don Martello, and Curt White, for their assistance in conducting Phase I of the verification test reported here. We would like to thank the California Air Resources Board for its assistance in conducting Phase II of verification testing. We would like to acknowledge the efforts of ETV stakeholders for their assistance in planning this verification test and for reviewing the test/QA plan and the verification reports. Specifically, we would like to acknowledge Judith Chow of Desert Research Institute, Jeff Cook of the California Air Resources Board, Tim Hanley of EPA, and Rudy Eden of the South Coast Air Quality Management District. We also would like to thank Tim Hanley of EPA for the loan of a BGI FRM sampler for Phase II. iv ------- Contents Notice ii Foreword iii Acknowledgments iv List of Abbreviations x 1. Background 1 2. Technology Description 2 3. Test Design and Procedures 3 3.1 Introduction 3 3.2 Test Design 3 3.3 Reference Method and Supplemental Measurements 4 3.3.1 PM2 5 Mass 4 3.3.2 Supplemental Measurements 5 3.4 Data Comparisons 6 3.5 Site Layout/Instrument Installation 7 3.5.1 Phase I 7 3.5.2 Phase II 7 4. Quality Assurance/Quality Control 10 4.1 Data Review and Validation 10 4.2 Deviations from the Test/QA Plan 10 4.3 Calibration and Parameter Checks of Reference Sampler 10 4.3.1 Flow Rate Calibration and Verification 11 4.3.2 Temperature Sensor Calibration and Verification 11 4.3.3 Pressure Sensor Calibration and Verification 12 4.3.4 Leak Checks 12 4.4 Collocated Sampling 13 4.4.1 Phase I—Pittsburgh 13 4.4.2 Phase II—Fresno 13 4.4.3 Summary 13 v ------- 4.5 Field Blanks 12 4.5.1 Phase I—Pittsburgh 15 4.5.2 Phase II—Fresno 15 4.6 Data Collection 15 4.6.1 Reference Measurements 15 4.6.2 BAM 1020 Monitors 16 4.7 Assessments and Audits 16 4.7.1 Technical Systems Audit 16 4.7.2 Performance Evaluation Audit 17 4.7.3 Audit of Data Quality 17 5. Statistical Methods 18 5.1 Inter-Unit Precision 18 5.2 Comparability/Predictability 18 5.3 Meteorological Effects/Precursor Gas Influence 19 5.4 Short-Term Monitoring Capabilities 20 6. Test Results 21 6.1 Phase I—Pittsburgh (August 1 - September 1, 2000) 21 6.1.1 Inter-Unit Precision 22 6.1.2 Comparability/Predictability 23 6.1.3 Meteorological Effects 27 6.1.4 Influence of Precursor Gases 28 6.2 Phase II—Fresno (December 18, 2000 - January 17, 2001) 28 6.2.1 Inter-Unit Precision 29 6.2.2 Comparability/Predictability 32 6.2.3 Meteorological Effects 34 6.2.4 Influence of Precursor Gases 34 6.2.5 Short-Term Monitoring 35 6.3 Instrument Reliability/Ease of Use 37 6.4 Shelter/Power Requirements 37 6.5 Instrument Cost 38 vi ------- 7. Performance Summary 39 7.1 Phase I—Pittsburgh (August 1 - September 1, 2000) 39 7.2 Phase II—Fresno (December 18, 2000 - January 1, 2001) 40 8. References 41 Figures Figure 2-1. Met One Instruments BAM 1020 Monitor 2 Figure 3-1. Site Layout During Phase I of Verification Testing 8 Figure 3-2. Site Layout During Phase II of Verification Testing 9 Figure 4-1. Comparison of Collocated PM25 FRM Samplers During Phase I of Verification Testing 14 Figure 4-2. Comparison of Collocated PM2 5 FRM Samplers During Phase II of Verification Testing 14 Figure 6-la. Hourly PM2 5 Concentrations from Duplicate BAM 1020 Monitors During Phase I of Verification Testing 23 Figure 6-lb. Correlation Plot of Hourly PM2 5 Data from Duplicate BAM 1020 Monitors During Phase I of Verification Testing 23 Figure 6-2a. 24-Hour Average PM2 5 Concentrations for Duplicate BAM 1020 Monitors During Phase I of Verification Testing 24 Figure 6-2b. Correlation Plot of 24-Hour PM2 5 Concentrations for Duplicate BAM 1020 Monitors During Phase I of Verification Testing 24 Figure 6-3a. Daily PM2 5 FRM Concentrations and the 24-Hour Average PM2 5 Concentrations from the Duplicate BAM 1020 Monitors During Phase I of Verification Testing 26 Figure 6-3b. Correlation Plot of the 24-Hour Average PM2 5 Concentrations from Duplicate BAM 1020 Monitors and the PM2 5 FRM Concentrations During Phase I of Verification Testing 26 Figure 6-4a. Hourly PM2 5 Concentrations from Duplicate BAM 1020 Monitors During Phase II of Verification Testing 30 vii ------- Figure 6-4b. Correlation Plot of Hourly PM2 5 Measurements from BAM 1020 Monitors During Phase II of Verification Testing 30 Figure 6-5a. Midnight-to-Midnight Average PM2 5 Concentrations from Duplicate BAM 1020 Monitors During Phase II of Verification Testing 31 Figure 6-5b. Correlation Plot of 24-Hour Average PM2 5 Concentrations from Duplicate BAM 1020 Monitors During Phase II of Verification Testing 31 Figure 6-6a. Midnight-to-Midnight Averages from Duplicate BAM 1020 Monitors and the PM25 FRM Results During Phase II of Verification Testing 33 Figure 6-6b. Correlation Plot from Duplicate BAM 1020 Monitors and the PM2 5 FRM During Phase II of Verification Testing 33 Figure 6-7. Correlation Plot of the Time-Weighted Average for the Short-Term Samples and the PM2 5 FRM 36 Figure 6-8. Correlation Plot of Short-Term Monitoring Results and the Corresponding Averages from the Duplicate BAM 1020 Monitors During Phase II of Verification Testing 36 Tables Table 6-1. Summary of Daily Values for the Measured Meteorological Parameters During Phase I of Verification Testing 21 Table 6-2. Summary of Daily Values for the Measured Precursor Gas Concentrations During Phase I of Verification Testing 21 Table 6-3. Linear Regression and Coefficient of Variation Results for Hourly and 24-Hour Average PM2 5 Concentration Values from Duplicate BAM 1020 Monitors During Phase I 25 Table 6-4. Comparability of the BAM 1020 Monitors with the PM2 5 FRM During Phase I 25 Table 6-5. Summary of Daily Values for the Measured Meteorological Parameters During Phase II of Verification Testing 29 Table 6-6. Summary of Daily Values for the Measured Precursor Gas Concentrations During Phase II of Verification Testing 29 viii ------- Table 6-7. Linear Regression and Coefficient of Variation Results for Hourly and 24-Hour Average PM2 5 Concentration Values During Phase II 32 Table 6-8. Comparability of the BAM 1020 Monitors with the PM2 5 FRM During Phase II 34 Table 6-9. Summary of PM2 5 Levels During Phase II of Verification Testing 35 Table 6-10. Regression Analysis Results for the Short-Term Monitoring 37 ix ------- List of Abbreviations ADQ audit of data quality AMS Advanced Monitoring Systems 14C carbon 14 CARB California Air Resources Board cm centimeter CI confidence interval CO carbon monoxide CV coefficient of variation DOE U.S. Department of Energy DPI digital pressure indicator DRI Desert Research Institute EPA U.S. Environmental Protection Agency ETV Environmental Technology Verification FRM federal reference method h2s hydrogen sulfide Hg mercury IMPROVE Interagency Monitoring for Protection of Visual Environments in. inch L/min liters per minute mg milligram mm millimeters NETL National Energy Technology Laboratory NIST National Institute of Standards and Technology NO nitric oxide no2 nitrogen dioxide N0X nitrogen oxides 03 ozone PPb parts per billion QA/QC quality assurance/quality control QMP Quality Management Plan R&P Rupprecht & Patashnick see Sharp Cut Cyclone SLAMS state and local air monitoring stations SFS sequential filter sampler S02 sulfur dioxide TOR thermal optical reflectance TSA technical systems audit WINS well impactor ninety six X ------- Chapter 1 Background The U.S. Environmental Protection Agency (EPA) has created the Environmental Technology Verification (ETV) Program to facilitate the deployment of innovative environmental technologies through performance verification and dissemination of information. The goal of the ETV Program is to further environmental protection by substantially accelerating the acceptance and use of improved and cost-effective technologies. ETV seeks to achieve this goal by providing high- quality, peer-reviewed data on technology performance to those involved in designing, distributing, permitting, purchasing, and using environmental technologies. ETV works in partnership with recognized testing organizations; with stakeholder groups consisting of regulators, buyers, and vendor organizations; and with the full participation of individual technology developers. The program evaluates the performance of innovative tech- nologies by developing test plans that are responsive to the needs of stakeholders, conducting field or laboratory tests (as appropriate), collecting and analyzing data, and preparing peer- reviewed reports. All evaluations are conducted in accordance with rigorous quality assurance protocols to ensure that data of known and adequate quality are generated and that the results are defensible. The EPA's National Exposure Research Laboratory and its verification organization partner, Battelle, operate the Advanced Monitoring Systems (AMS) Center under ETV. The AMS Center recently evaluated the performance of fine particle monitors for use in continuous monitoring of fine particulate matter in ambient air. This verification report presents the procedures and results of the verification test for the Met One Instruments BAM 1020 particle monitor. 1 ------- Chapter 2 Technology7 Description The following description of the BAM 1020 is based on information provided by the vendor. The BAM 1020 is a beta attenuation monitor that measures the concentration of particulate matter in ambient air. The BAM 1020 may be equipped with a sharp cut cyclone PM2 s or a WINS PM2 5 sampling inlet for automatic monitoring of fine particulate matter. The BAM 1020 monitor can also be configured to monitor total suspended particulate matter. An internal data logger allows up to six additional air quality or meteorological measurements. At the beginning of the sampling period, beta ray transmission is measured across a clean section of filter tape. This tape is mechanically advanced to the sampling inlet. Particulate matter is drawn into the sample inlet and deposited on the filter paper. At the completion of the sampling period, the filter tape is returned to its original location and the beta ray transmission is remeasured. The difference between the two measurements is used to determine the particulate matter concentration. The mass density is measured using the technique of beta attenuation. A small 1 'C beta source (60 jiCi) is coupled to a detector that counts the emitted beta particles. The filter tape is placed between the beta source and the detector. As the mass deposited on the filter tape increases, the measured beta particle count is reduced according to a known equation. The BAM 1020 consists of a detector/logger, pump, and sampling inlet. Each of these compo- nents is self-contained and may be disconnected for servicing or replacement. The BAM 1020 is designed to mount in a temperature-controlled enclosure. The sampling inlet is designed to mount through the roof of the enclosure. The BAM 1020 operates at 100 to 230 volts alternating current and is 310 mm high x 430 mm wide x 400 mm deep. All operations of the unit are displayed with an 8 line by 40 character display. Figure 2-1. Met One Instruments BAM 1020 Monitor 2 ------- Chapter 3 Test Design and Procedures 3.1 Introduction The objective of this verification test is to provide quantitative performance data on continuous fine particle monitors under a range of realistic operating conditions. To meet this objective, field testing was conducted in two phases in geographically distinct regions of the United States during different seasons of the year. Performing the test in different locations and in different seasons allowed sampling of widely different particulate matter concentrations and chemical composition. At each site, testing was conducted for one month during the season in which local PM2 5 levels were expected to be highest. The verification test was conducted according to the procedures specified in the Test/QA Plan for Verification of Ambient Fine Particle Monitors.(1) The first phase of field testing was conducted at the ambient air monitoring station on the Department of Energy's (DOE's) National Energy Technology Laboratory (NETL) campus in Pittsburgh, PA. Sampling during this phase of testing was conducted from August 1 to September 1, 2000. The second phase of testing was performed at the California Air Resources Board's (CARB's) Air Monitoring Station in Fresno, CA. This site is also host to one of the EPA's PM2 5 Supersites being managed by Desert Research Institute (DRI). This phase of testing was conducted from December 18, 2000, to January 17, 2001. 3.2 Test Design Specific performance characteristics verified in this test include ¦ Inter-unit precision ¦ Agreement with and correlation to time-integrated reference methods ¦ Effect of meteorological conditions ¦ Influence of precursor gases ¦ Short-term monitoring capabilities. To assess inter-unit precision, duplicate BAM 1020 monitors were tested in side-by-side operation during each phase of testing. During Phase I, the monitors used were Serial Number Y3402 and Y2863. During Phase II,the monitors used were Serial Number Y3402 and Y3330. Collocation of the BAM 1020 monitors with reference systems for time-integrated sampling of fine particulate mass and chemical speciation provided the basis for assessing the degree of agreement and/or correlation between the continuous and reference methods. Each test site was 3 ------- equipped with continuous monitors to record meteorological conditions and the concentration of key precursor gases (ozone, nitrogen oxides, sulfur dioxide, etc.). The data from the meteorological and gas monitors were used to assess the influence of these parameters on the performance of the fine particle monitors being tested. Reference method sampling periods of 3, 5, and 8 hours were used in Phase II of this test to establish the short-term monitoring capabilities of the continuous monitors being tested. Statistical calculations, as described in Chapter 5, were used to establish each of these performance characteristics. Additionally, other performance characteristics of the technologies being verified, such as reliability, maintenance requirements, and ease of use, were assessed. Instrumental features that may be of interest to potential users (e.g., power and shelter requirements and overall cost) are also reported. 3.3 Reference Method and Supplemental Measurements Since no appropriate absolute standards for fine particulate matter exist, the reference methods for this test were well established, time-integrated methods for determining particulate matter mass or chemical composition. It is recognized that comparing real-time measurements with time- integrated measurements does not fully explore the capabilities of the real-time monitors. However, in the absence of accepted standards for real-time fine particulate matter measurements, the use of time-integrated standard methods that are widely accepted was necessary for performance verification purposes. It should be noted that there are necessary differences between continuous and time-integrated, filter-based techniques. For example, in time-integrated sampling, particulate matter collected on a filter may remain there for up to 24 hours, whereas continuous monitors generally retain the particulate sample for one hour or less. Thus, the potential for sampling artifacts differs. Also, in the case of particle mass measurements, the mass of particulate matter is determined after equilibration at constant temperature and humidity, conditions that are almost certain to differ from those during sampling by a continuous monitor. The BAM 1020 reports measurement results in terms of PM2 5 mass and, therefore, was compared with the federal reference method (FRM) for PM2 5 mass determination.® Additionally, comparisons with a variety of supplemental measurements were made to establish specific performance characteristics. Descriptions of the reference method and supplemental measurements used during the verification test are given below. 3.3.1 PM2SMass The primary comparisons of the BAM 1020 readings were made relative to the FRM for PM2 5 mass determination, i.e., the 24-hour time-averaged procedure detailed in 40 CFR Part 50.(2) This method involves manual sampling using any of a number of designated commercially available filter samplers, followed by gravimetric analysis of the collected sample. In this method, a size- selective inlet is used to sample only that fraction of aerosol of interest (i.e., < 2.5 |im aero- dynamic diameter). The air sample is drawn into the sampler at a fixed rate (16.7 L/min) over 24 hours, and the aerosol is collected on a Teflon filter for gravimetric analysis. After equilibration 4 ------- of the sample and filter in a temperature- and humidity-controlled environment, the sample is weighed on a microbalance. The particulate matter sample weight is determined by subtracting the weight of the filter alone, determined prior to sampling after similar equilibration. Protocols for sample collection, handling, and analysis are prescribed by the EPA(2) and were followed for this verification test. Filter samples for the PM2 5 FRM were collected daily during each phase of the testing using a BGI FRM sampler (RFPS-0498-116), and the PM25 mass was determined according to the procedures mentioned above. In Phase I, a single BGI FRM sampler (SN 311) was operated daily from noon to noon to collect the FRM samples. During Phase II, two BGI FRM samplers (SN 287 and SN 311) were used and were operated on alternate days to facilitate a midnight-to- midnight sampling schedule. Collocated samples were collected during each phase to establish the precision of the FRM. A discussion of the collocated sampling is presented in Section 4.4 of this report. 3.3.2 Supplemental Measurements Various supplemental measurements were used to further establish the performance of the continuous monitors being tested. Meteorological conditions were monitored and recorded continuously throughout each phase of the verification test. These measurements included temperature, relative humidity, wind speed, direction, barometric pressure, and solar radiation. These data were provided to Battelle for Phase I by DOE/NETL and for Phase II by DRI. Likewise, the ambient concentrations of various precursor gases including ozone and nitrogen oxides also were measured continuously during the verification test and used to assess the influence of these parameters on the performance of the monitors tested. Continuous measure- ments of sulfur dioxide, hydrogen sulfide, nitric oxide, nitrogen dioxide, nitrogen oxides, and ozone were provided for Phase I by DOE/NETL; and continuous measurements of carbon monoxide, ozone, nitric oxide, nitrogen dioxide, and nitrogen oxides were provided for Phase II by DRI. These gases were of interest as potential chemical precursors to aerosol components, and as indicators of ambient pollutant levels. During Phase I, samples for chemical speciation were collected using an Andersen RAAS speciation sampler configured with five sample trains (one channel at 16.7 L/min and four channels at approximately 8 L/min). The 16.7 L/min channel was operated with a Teflon filter for PM2 5 mass determination. Samples for carbon analysis were collected at 8 L/min on quartz filters and analyzed by the IMPROVE thermal optical reflectance method at DRI. Nitrate and sulfate samples were collected on nylon filters downstream of a magnesium-oxide-coated compound annular denuder, and analyzed by ion chromatography at Consol. To supplement the 24-hour samples, additional samples for PM2 5 mass were collected at the Fresno site over shorter sampling periods (i.e., 3-, 5-, 8-hour) to assess the capabilities of the monitors being tested in indicating short-term PM2 5 levels. A medium-volume sequential filter sampling (SFS) system sampling at a flow rate of 113 L/min was used to collect the short-term mass and speciation samples during Phase II. The SFS was configured to take two simultaneous samples (i.e., Teflon-membrane/drain disk/quartz-fiber and quartz-fiber/sodium-chloride- 5 ------- impregnated cellulose-fiber filter packs) at 20 L/min through each sampling port. Anodized aluminum nitric acid denuders were located between the inlets and the filters to remove gaseous nitric acid. The remaining 73 L/min required for the 113 L/min total inlet flow was drawn through a makeup air sampling port inside the plenum. The timer was set to take five sets of sequential samples every 24 hours. Solenoid valves, controlled by a timer, switched between one to five sets of filters at midnight each day. A vacuum pump drew air through the paired filter packs when the valves were open. The flow rate was controlled by maintaining a constant pressure across a valve with a differential pressure regulator. The filters were loaded at the DRI's Reno, NV, laboratory into modified Nuclepore filter holders that were plugged into quick-disconnect fittings on the SFS. One filter pack contained a 47-mm- diameter Teflon-membrane filter with quartz-fiber backup filter. A drain disc was placed between the Teflon-membrane and quartz-fiber filters to ensure a homogeneous sample deposit on the front Teflon-membrane filter and to minimize fiber transfer from one filter to the other. The Teflon-membrane filter collected particles for mass and elemental analysis. The other filter pack contained a 47-mm-diameter quartz-fiber filter with a sodium-chloride-impregnated cellulose-fiber backup filter on a separate stage. The deposit on the quartz-fiber filter was analyzed for ions and carbon. The sodium-chloride-impregnated cellulose-fiber backup filter was analyzed for nitrate to estimate losses due to volatilization of ammonium nitrate from the front filter during sampling. This sequential filter sampler was operated from midnight to 5:00 a.m. (0000-0500), from 5:00 a.m. to 10:00 a.m. (0500-1000), from 10:00 a.m. to 1:00 p.m. (1000-1300), from 1:00 p.m. to 4:00 p.m. (1300-1600), and from 4:00 p.m. to midnight (1600-2400). These short-term sampling measurements were appropriately summed over 24 hours for comparison with the corresponding 24-hour results of the FRM reference samplers to establish the relationship between the two sets of measurements. 3.4 Data Comparisons The primary means used to verify the performance of the BAM 1020 monitors was comparison with the 24-hour FRM results. Additional comparisons were made with the supplemental meteorological conditions and precursor gas concentrations to assess the effects of these parameters on the response of the monitors being tested. The short-term monitoring results from Fresno in Phase II of the verification test also were used to assess the capabilities of the BAM 1020 monitors to indicate short-term levels of ambient PM2 5. The comparisons were based on statistical calculations as described in Section 5 of this report. Comparisons were made independently for the data from each phase of field testing; and, with the exception of the inter-unit precision calculations, the results from the duplicate monitors were analyzed and reported separately. Inter-unit precision was determined from a statistical inter- comparison of the results from the duplicate monitors. 6 ------- 3.5 Site Layout/Instrument Installation In each phase of testing, the two BAM 1020 monitors were installed in Battelle's instrument trailer, which is a converted 40-foot refrigerator semi-trailer. The BAM 1020 monitors were placed on a counter top, with each monitor directly below a 7.6-cm (3 in.) port through the roof of the trailer. Separate inlet tubes, approximately three meters (10 feet) in length, were installed vertically through the sampling ports and secured on the trailer roof using tripods. A PM10 head and PM2 5 Sharp Cut Cyclone (SCC) were used with each BAM 1020 to provide particle size selection. Data generated by the BAM 1020 monitors were recorded internally and downloaded several times throughout each phase of testing as described in Section 4.6.2. 3.5.1 Phase I Phase I verification testing was conducted at the DOE/NETL facility within the Bruceton Research Center. The facility is located in the South Park area of Pittsburgh, PA, approximately 7 miles from downtown. The air monitoring station where testing was conducted is located on the top of a relatively remote hill within the facility and is impacted little by road traffic. The layout of the testing facility is illustrated schematically in Figure 3-1. For this test, Battelle provided temporary facilities to augment the permanent facilities in use by the DOE/NETL air monitoring staff. These temporary facilities included a temporary Battelle/ ETV platform (16-foot by 14-foot scaffold construction) and a Battelle instrument trailer. The Battelle trailer was positioned parallel with, and approximately 25 feet from, the DOE/NETL instrument trailer. The Battelle/ETV platform was located between the two trailers, with the surface at a height of approximately 2 meters (6 feet). Most of the DOE/NETL continuous monitoring equipment, including the continuous precursor gas monitors, was located inside the DOE/NETL instrument trailer. A DOE/NETL Rupprecht & Patashnick (R&P) Co. Partisol FRM sampler used to evaluate FRM precision was located outside on a DOE/NETL platform. The BAM 1020 monitors were installed inside the Battelle trailer, and the BGI FRM sampler was installed on the Battelle/ETV platform. A vertical separation of approximately 2 to 3 meters and a horizontal separation of approximately 3 meters existed between the inlets of the BAM 1020 monitors and the BGI FRM sampler. A 10-meter (33-foot) meteorological tower was located approximately 20 meters (65 feet) to the north of the DOE/NETL instrument trailer. 3.5.2 Phase II Phase II of verification testing was conducted at the CARB site on First Street in Fresno. This site is located in a residential/commercial neighborhood about three miles north of the center of Fresno. The two BGI FRM samplers and a 3-meter (10-foot) meteorological tower were located on the roof of the two-story building housing the CARB office. Continuous precursor gas monitors were located inside the CARB office space and sampled through a port in the roof of the building. The two BGI FRM samplers were located on the southernmost edge of the rooftop to be as close as possible to the instrument trailer. The Battelle trailer used during Phase I of this verification test also was used during Phase II. For Phase II, the Battelle trailer was located in the 7 ------- DOE/NETL Instrument Trailer DOE/NETL Platform Battelle/ETV Platform N Figure 3-1. Site Layout During Phase I of Verification Testing (not drawn to scale) parking lot adjacent to the building in which the CARB site is located. The trailer was positioned approximately 25 meters (80 feet) to the south of the building, as shown in Figure 3-2. The BAM 1020 monitors were located in the Battelle trailer and installed in the same fashion as in Phase I of the verification test. A difference in elevation of approximately 6 meters (20 feet) existed between the top of the trailer and the roof of the building housing the CARB site and between the inlets of the BAM 1020 monitors and the BGI FRM samplers. In addition to the two BGI FRM samplers used to collect the reference samples, an R&P Partisol FRM sampler was operated on the rooftop by CARB. This sampler was positioned approximately 25 meters (65 feet) to the northeast of the BGI FRM samplers and was used to measure the precision of the FRM reference values. The sequential filter sampler used to collect the short-term samples was located near the R&P FRM sampler. Battelle Instrument Trailer 8 ------- 00 (T> Tl ~ 73 n co Q) 3 "D_ 0 (/) Roof of CARB Site ~ ~ CARB Samplers N Figure 3-2. Site Layout During Phase II of Verification Testing (not drawn to scale) 9 ------- Chapter 4 Quality Assurance/Quality Control 4.1 Data Review and Validation Test data were reviewed and approved according to the AMS Center quality management plan (QMP)(3) and the test/QA plan.(1) The Verification Test Coordinator or the Verification Testing Leader or designee reviewed the raw data, laboratory notebook entries, and data sheets that were generated each day and approved them by initialing and dating the records. Data from the BAM 1020 monitors were validated by a representative of Met One and reviewed by the Verification Test Coordinator before being used in statistical calculations. Data were checked for error flags and not used if flagged for power or instrument failure. Daily PM2 5 concentration averages calculated from the continuous BAM 1020 data were considered valid if the percent data recovery for the 24-hour sampling period (i.e., noon to noon for Phase I, or midnight to midnight for Phase II) was 75% or greater. 4.2 Deviations from the Test/QA Plan The following deviations from the test/QA plan were documented and approved by the AMS Center Manager. None of these deviations had any deleterious effect on the verification data. ¦ Calibration checks of the temperature and pressure sensors were not performed within one week of the start of Phase II. Subsequent checks of these sensors indicated proper calibration. ¦ The distance between the reference samplers and the monitors being tested was increased to approximately 25 meters to accommodate changes in the overall site layout for Phase II. In addition, although not formally a deviation from the test/QA plan, we note that the relative humidity of the continuing weighing room used by Consol in Phase I occasionally deviated from the specified limits. The impact of this occurrence was minimal, as noted in Section 4.4.1. 4.3 Calibration and Parameter Checks of Reference Sampler The BGI FRM samplers provided by Battelle for this verification test were calibrated using National Institute of Standards and Technology (NIST)-traceable flow meters and temperature and pressure sensors. The calibration and verification of these samplers are described below. 10 ------- 4.3.1 Flow Rate Calibration and Verification Prior to Phase I of the verification test, a three-point calibration of the sampler flow rate was performed on June 22, 2000. Flows were measured at three set points (16.7 L/min, and approx- imately + 10% and -10% of 16.7 L/min) using a dry gas meter (American Meter Company, Battelle asset number LN 275010, calibrated January 21, 2000). If necessary, the flows were adjusted manually until agreement with the dry gas meter fell within ±2% of the sampler's indicated flow reading. The on-site operators checked the flow rate of the BGI FRM sampler both before and after Phase I of the verification test using an Andersen Instruments Inc. dry gas meter (identification number 103652, calibrated March 30, 2000). The flow rate was checked prior to testing on both July 19, 2000, and July 30, 2000. In both cases, the measured flow rate was verified to be within 4% of the flow rate indicated by the sampler. After testing, the flow rate was again checked on September 11, 2000, using the same Andersen dry gas meter. In this case, the flow rate did not fall within the 4% acceptance limit. This failure is probably linked to the failure of the ambient temperature thermocouple, on September 7, 2000, after completion of the Phase I sampling (see Section 4.3.2). Prior to Phase II of the verification test, single point calibration checks of the duplicate BGI FRM samplers were performed at 16.7 L/min on December 15, 2000. These flow rate checks were per- formed using a BGI DeltaCal calibrator (BGI Inc., serial number 0027, calibrated October 24, 2000), and the measured flow rates were within 4% of the indicated flow on each sampler. Weekly flow rate checks also were performed throughout Phase II using the DeltaCal flow meter. In each case, the measured flow rates were within ± 4% of the indicated reading of the BGI FRM and within ±5% of the nominal 16.7 L/min setpoint. Calibration of the flow rate for the SFS used during Phase II, was maintained by DRI through daily flow checks with a calibrated rotameter, and independent performance evaluation audits conducted by Parson's Engineering. No additional flow verification was performed for this test. 4.3.2 Temperature Sensor Calibration and Verification Both the ambient temperature sensor and the filter temperature sensor of the BGI FRM sampler were checked at three temperatures (approximately 5, 22, and 45°C) on June 20, 2000. The sensor readings were compared with those from an NIST-traceable Fluke Model 52 thermocouple gauge (Battelle asset number LN 570068, calibrated October 15, 1999). Agreement between the sampler temperature sensors and the calibrated thermocouple was within ±2°C at each temperature. The temperature sensors also were checked at the DOE/NETL site both before and after Phase I of the verification test by the on-site operators. Prior to testing, the sensors were checked on July 19, 2000, and July 30, 2000, against the readings from a mercury thermometer (Ever Ready, serial number 6419, calibrated October 29, 1999). For these checks, agreement between the sensors and the thermometer was within ±2°C. After the verification period, the ambient temperature sensor suffered a malfunction on September 7. The filter temperature sensor was 11 ------- checked on September 11, 2000, and showed agreement with the mercury thermometer within ±2°C. The ambient sensor was replaced, after completing Phase I, with a new factory-calibrated sensor provided by BGI. The temperature sensors for the two BGI FRM samplers were checked on January 16, 2001, against readings from a Fluke Model 52 thermocouple gauge (Battelle asset number LN 570077, calibrated October 26, 2000). For each BGI FRM, both the ambient and filter temperature sensor readings agreed with the thermocouple readings within ±2°C. 4.3.3 Pressure Sensor Calibration and Verification Before Phase I, the barometric pressure sensor in the BGI FRM sampler was calibrated against an NIST-traceable Taylor Model 2250M barometer (Battelle asset number LN 163610, calibrated January 12, 2000) and an NIST-traceable convectron gauge (Granville-Phillips Co., Battelle asset number LN 298084, calibrated August 25, 1999) on June 17 and 18, 2000. The sensor was calibrated at ambient pressure and under a reduced pressure (approximately 100 mm mercury below ambient). Checks of the pressure sensor were performed at the DOE/NETL site both before and after Phase I of the verification test. The pressure sensor was checked on July 19, 2000, and July 30, 2000, using an NIST-traceable Taylor Model 2250M barometer (Battelle asset number LN 163609, calibrated January 12, 2000). On September 11, 2000, the pressure sensor of the BGI FRM sampler was again checked against the same barometer, but did not agree within the acceptance criterion of 5 mm mercury. This failure is possibly associated with the failure of the ambient temperature sensor on September 7, 2000. The ambient pressure sensor for both BGI FRM samplers used in Phase II was checked against the pressure readings of a BGI DeltaCal on January 16, 2001. Agreement between the BGI FRM pressure readings and those of the DeltaCal was within 5 mm mercury for both samplers. 4.3.4 Leak Checks Leak checks of the BGI FRM and SFS samplers were performed every fourth day during Phase I of the verification test. These leak checks were conducted immediately following the cleaning of the WINS impactor and were performed according to the procedures in the operator's manual for the BGI FRM sampler. All leak checks passed the acceptance criteria provided in the operator's manual. Leak checks of the BGI FRM and SFS samplers were performed daily during Phase II of the verification test. These leak checks were conducted during set-up for each 24-hour sampling period. All leak checks passed before the sampler set-up was completed. 12 ------- 4.4 Collocated Sampling 4.4.1 Phase I—Pittsburgh To establish the precision of the PM2 5 FRM, the BGI FRM sampler was collocated with an R&P FRM sampler for Phase I, including a period of two weeks prior to and one week after Phase I of the verification test. During the sampling periods before and after Phase I, the BGI and R&P FRM samplers were located on the same platform and within 4 meters of one another. During the Phase I testing period, these samplers were separated by a distance of approximately 25 meters. The samples from the BGI FRM sampler were collected and analyzed by Consol, and the samples from the R&P FRM sampler were collected and analyzed by on-site Mining Safety and Health Administration staff. Figure 4-1 shows the results of the collocated FRM sampling conducted for Phase I. These data were analyzed by linear regression; and the calculated slope, intercept, and r2 values are 0.939 (0.033), 1.28 (0.66) |ig/m3, and 0.957, respectively, where the values in parentheses are 95% con- fidence intervals (CIs). Despite completely independent operations (i.e., separate sampling staff and weighing facilities), these data show very good agreement between the BGI FRM and the R&P FRM samplers. The data also indicate that, although the humidity in the conditioning/ weighing room at Consol was not always within the specified FRM limits, the influence of the elevated humidity was not severe. 4.4.2 Phase II—Fresno During Phase II of testing, duplicate BGI FRM samplers (SN 287 and SN 311) were used to collect the 24-hour FRM reference samples. These samplers were operated one at a time on alternate days to facilitate midnight-to-midnight sampling. Likewise, an R&P Partisol sampler was used by CARB to collect 24-hour FRM samples. The R&P FRM sampler was located approximately 25 meters from the BGI FRM samplers. The same on-site operators performed the sampling for the FRM samplers; however, DRI performed the gravimetric analyses for the BGI FRM samplers and CARB performed the analyses for the R&P FRM sampler. Figure 4-2 shows the results for the collocated FRM sampling conducted for Phase II. Only 12 days of collocated sampling were available from the Fresno site. The linear regression of these data shows a slope of 1.096 (0.047) and intercept of-1.0 (2.1) |ig/m3 and r2 value of 0.982, where the numbers in parentheses indicate the CIs. 4.4.3 Summary The results from the collocated FRMs in both Pittsburgh and Fresno show agreement that is consistent with the goals for measurement uncertainty of PM2 5 methods run at state and local air monitoring stations (SLAMS). These goals are identified in Appendix A to 40 CFR Part 58, Section 3.5(4) which states: "The goal for acceptable measurement uncertainty has been defined as 10 percent coefficient of variation (CV) for total precision and ± 10% for total bias." Since the collocated FRMs in both Pittsburgh and Fresno were operated by independent organizations, a 13 ------- 40 y = 0.939x + 1.282 r2 = 0.957 O) o_ 20 0 10 20 30 40 BGI FRM Sampler (ug/m3) Figure 4-1. Comparison of Collocated PM2 5 FRM Samplers for Phase I of Verification Testing 80 y= 1.096x- 1.040 r2 = 0.982 60 O) U- 40 20 0 0 20 40 60 80 BGI PM2.5 FRM (|ig/m3) Figure 4-2. Comparison of Collocated PM2 5 FRM Samplers for Phase II of Verification Testing 14 ------- comparison to the SLAMS data quality objectives for PM2 5 is an appropriate way to assess whether the measurement systems were producing data of acceptable quality. In both Pittsburgh and Fresno, the results of the collocated sampling meet the data quality objectives for the total bias. In Fresno, the collocated sampling results show a CV of 6.3%, which meets the data quality objectives for precision. In Pittsburgh, the calculated CV was 10.5%. However, this value is driven largely by scatter in the low concentration regimes. When a single data pair is removed, the CV becomes 9.1%, which meets the data quality objectives for total precision. (It should be noted, as well, that the Fresno collocated results consist of only 12 data points.) Thus, the collocated FRM results from Pittsburgh and Fresno show that the reference measurements were suitable for verifying the performance of continuous fine particle monitors. 4.5 Field Blanks 4.5.1 Phase I—Pittsburgh During Phase I, at least 10% of the collected reference samples were field blanks. The observed filter mass difference of the field blanks ranged from -7 |ig to 16 |ig, and the corresponding PM2 5 concentrations (which were determined using an assumed sample volume of 24 m3) were all less than 0.0007 mg/m3, averaging 0.00015 mg/m3. FRM results for Phase I were not blank corrected. 4.5.2 Phase II—Fresno During Phase II, at least 10% of the collected reference samples (for both the BGI FRM samplers and the DRI sequential filter sampler) were field blanks. The results were added to a database containing historical field blank data. On average, these blanks showed mass differences of 2 |ig, with a standard deviation of 8 |ig. Assuming a sample volume of 24 m3 (i.e., FRM value), these blanks account for approximately 0.0001 mg/m3. Assuming a sample volume of 3.6 m3 (i.e., three- hour short-term sample from sequential filter sampler), these blanks account for approximately 0.0006 mg/m3. These blank values were negligible, even for the short-term sampling periods, in comparison with the PM2 5 mass levels that were present during the Phase II testing (see Section 6.2). FRM results for Phase II were blank corrected using the data available from the historical database. 4.6 Data Collection 4.6.1 Reference Measurements During Phase I, daily records of the sampling activities for the BGI FRM sampler were recorded on individual data sheets by the on-site operators, and summary data from the BGI FRM sampler were downloaded daily using portable data logging modules. Information recorded on the data sheets included identification of the sampling media (i.e., filter ID numbers) and the start and stop times for the sampling periods. Summary data from the sampler included the parameters listed above, in addition to the sampling duration, volume sampled, and average temperature and pressure readings. 15 ------- During Phase II, summary data from the BGI FRM samplers were logged daily on sampling sheets by the on-site operators. These data included sample identification, start times for the sampling period, sampling duration, volume sampled, and average temperature and pressure readings. 4.6.2 BAM 1020 Monitors Hourly data from each of the BAM 1020 monitors were recorded in an internal memory buffer throughout each phase of the verification test. For each day, the data were stored in tabular format with hourly values reported for PM2 5 concentration (mg/m3), sampled volume (m3), wind speed (knots), room temperature (C), relative humidity (%), barometric pressure (inches of mercury), and ambient temperature (C). Additionally, the average values for each of these parameters and the percent data recovery were generated by the BAM 1020 for each midnight to midnight period. The recorded data were downloaded directly onto a laptop computer and saved as text files. These files were imported into a spreadsheet for analysis, and copies of the data were stored by the Verification Test Coordinator on a floppy disk, as well as on a computer hard drive. 4.7 Assessments and Audits 4.7.1 Technical Systems Audit Phase I—Pittsburgh The technical systems audit (TSA) ensures that the verification tests are conducted according to the test/QA plan(1) and that all activities associated with the tests are in compliance with the ETV pilot QMP.(3) The Battelle Quality Manager conducted an internal TSA on August 3, 2000, at the Pittsburgh test site. All findings noted during this TSA were documented and submitted to the Verification Test Coordinator for correction. The corrections were documented by the Verifica- tion Test Coordinator and reviewed by Battelle's Quality Manager, Verification Testing Leader, and AMS Center Manager. None of the findings adversely affected the quality or outcome of this phase of the verification test. All corrective actions were completed to the satisfaction of the Battelle Quality Manager. The records concerning this TSA are permanently stored with the Battelle Quality Manager. Phase II—Fresno An internal TSA was conducted by the Battelle Quality Manager on January 9, 2001, at the Fresno test site. An external TSA was also conducted concurrently by EPA quality staff, Ms. Elizabeth Betz and Ms. Elizabeth Hunike. All findings noted during these TSAs were documented and submitted to the Verification Test Coordinator for corrective action. None of the findings adversely affected the quality or outcome of this phase of the verification test for the BAM 1020. All corrective actions were completed to the satisfaction of the Battelle Quality Manager and the EPA. 16 ------- 4.7.2 Performance Evaluation Audit Phase I—Pittsburgh The reference sampler provided by Battelle for this verification test was audited during Phase I to ensure that it was operating properly. During Phase I of the verification test, the flow rate of the BGI FRM sampler was audited on August 28, using a dry gas meter (American Meter Company, Battelle asset number LN 275010, calibrated April 17, 2000). The measured flow rate was within the ±4% acceptance criterion with respect to the internal flow meter and within the ±5% acceptance criterion with respect to the nominal flow rate. Both temperature sensors in the BGI FRM sampler were checked on August 28, using a Fluke 52 thermocouple (Battelle asset number LN 570068, calibrated October 15, 1999). Agreement between each sensor and the thermocouple was within the ±2°C acceptance criterion. Phase II—Fresno A performance evaluation audit was conducted to ensure that the two BGI FRM samplers used during Phase II of testing were operating properly. The flow rates of the samplers were audited on January 16 and 17, 2001, using a dry gas meter (Schlumberger, SN 103620, calibrated July 6, 2000). For each sampler, the measured flow rate was within the ±4% acceptance criterion with respect to the internal flow meter and within the ±5% acceptance criterion with respect to the nominal flow rate. The temperature readings for the two samplers were checked with a mercury thermometer (Fisher Scientific, SN 7116). Agreement between each sensor and the thermocouple was within the ±2°C acceptance criterion. The pressure sensors for the two samplers were checked against a Druck digital pressure indicator (DPI) (SN 6016/00-2, calibrated June 28, 2000). Agreement between each sensor and the DPI was within the acceptance criterion of ±5 mm mercury. 4.7.3 Audit of Data Quality Battelle's Quality Manager ensured that an audit of data quality (ADQ) of at least 10% of the verification data acquired during the verification test was completed. The ADQ traced the data from initial acquisition, through reduction and statistical comparisons, to final reporting. Reporting of findings followed the procedures described above for the Phase I TSA. All findings were corrected to the satisfaction of the Battelle Quality Manager, and none of the findings adversely affected the quality of the verification test for the BAM 1020 monitors. 17 ------- Chapter 5 Statistical Methods Performance verification is based, in part, on statistical comparisons of continuous monitoring data with results from the reference methods. A summary of the statistical calculations that have been made is given below. 5.1 Inter-Unit Precision The inter-unit precision of the BAM 1020 monitors was determined based on procedures described in Section 5.5.2 of EPA 40 CFR 58, Appendix A, which contains guidance for precision assessments of collocated non-FRM samplers. Simultaneous measurements from the duplicate BAM 1020 monitors were paired, and the behavior of their differences was used to assess precision. For both the hourly and the 24-hour PM2 5 measurements, the CV is reported. The CV is defined as the standard deviation of the differences divided by the mean of the measurements and expresses the variability in the differences as a percentage of the mean. As suggested by the EPA guidance, only measurements above the limit of detection were used in precision calculations. Inter-unit precision was assessed separately for each phase of the verification test. 5.2 Comparability/Predictability The comparability between the BAM 1020 results and the PM2 5 FRM was assessed, since these monitors yield measurements with the same units of measure as the PM2 5 FRM. The relationship between the two was assessed from a linear regression of the data using the PM2 5 FRM results as the independent variable and the BAM 1020 monitor results as the dependent variable as follows: Q = ji + PxRt + Sj (1) where Rt is the i'h 24-hour FRM PM2 5 measurement; C, is the average of the hourly BAM 1020 measurements over the same 24-hour time period as the i4 reference measurement; |i and P are the intercept and slope parameters, respectively; and Sj is error unexplained by the model. The average of the hourly BAM 1020 measurements is used because this is the quantity that is most comparable to the reference sampler measurements. Comparability is expressed in terms of bias between the BAM 1020 monitor and the PM2 5 FRM and the degree of correlation (i.e., r2) between the two. Bias was assessed based on the slope and 18 ------- intercept of the linear regression of the data from the PM2 5 FRM and the BAM 1020 monitor. In the absence of bias, the regression equation would be Ct = Rt + Sj (slope = 1, intercept = 0), indi- cating that the 24-hour average of hourly BAM 1020 measurements is simply the PM2 5 FRM measurement plus random error. A value of r2 close to 1 implies that the amount of random error is small; that is, the variability in the hourly measurements is almost entirely explained by the variability in the PM2 5 FRM measurements. Quantities reported include r2, intercept, and slope, with estimates of the 95% CI for the intercept and slope. Comparability to the FRM was determined independently for each of the two duplicate BAM 1020 monitors being tested and was assessed separately for each phase of the verification test. 5.3 Meteorological Effects/Precursor Gas Influence The influence of meteorological conditions on the correlation between the BAM 1020 monitors and the PM2 5 FRM reference samplers was evaluated by using meteorological data such as temperature and humidity as parameters in multivariable analyses of the reference/monitor comparison data. The same evaluation was done with ambient precursor pollutant concentrations as the model parameters. The model used is as follows: Q = ji + PxRt + SyjxXji + Si (2) where is the meteorological or precursor gas measurement for the i"1 24-hour time period, Yj is the associated slope parameter, and other notation is as in Equation 1. Comparability results are reported again after these variables are adjusted for in the model. Additionally, estimates and standard errors of Yj are provided. Meteorological effects and precursor gas interferences were assessed independently for each of the duplicate BAM 1020 monitors tested and were assessed separately for each phase of the verification test. In conducting these multivariable analyses, a significance level of 90% was used in the model selection. This significance level is less stringent than the 95% level used in other aspects of the verification, and was chosen so that even marginally important factors could be identified for consideration. Note that the multivariable model ascribes variance unaccounted for by linear regression against the FRM to the meteorological or precursor gas parameters. The model treats all candidate parameters equally. The model discards the least significant parameter and is rerun until all remaining variables have the required significance (i.e., predictive power). The results of the model should not be taken to imply a cause-and-effect relationship. It is even possible that the parameters identified as significant for one unit of a monitoring technology may differ from those identified for the duplicate unit of that technology, due to differences in the two data sets. 5.4 Short-Term Monitoring Capabilities This assessment was based on linear regression analysis of results from the BAM 1020 monitors and the short-term (3-, 5-, and 8-hour) sampling results from the two BGI FRM samplers 19 ------- generated in Phase II only. The analysis was conducted, and the results are reported in a fashion identical to that for the comparability results for the 24-hour samples described in Section 5.2. These comparisons were made only after establishing the relationship between the short-term sampling results and the corresponding 24-hour results. The relationship between the two sets of reference measurements was made by linear regression using the weighted sum of the results from the short-term sampling as the dependent variable and the 24-hour FRM results as the independent variable in the regression analysis. Comparability was assessed using Equation 1, replacing the average of hourly measures with the average of short-term sampler measurements. The short-term sampling results also have been used to assess the effects of meteorological conditions and precursor gas concentrations on the response of the monitors. These short-term results were used in place of the 24-hour measurements in the analysis described in Section 5.3 for Phase II only. Independent assessments were made for each of the duplicate BAM 1020 monitors, and the data from each phase of testing were analyzed separately. 20 ------- Chapter 6 Test Results 6.1 Phase I—Pittsburgh (August 1 - September 1, 2000) Samples were collected daily between August 1 and September 1, 2000, using a PM25 FRM sampler. During this period, the daily PM2 5 concentration as measured by the BGI FRM sampler ranged from 6.1 |ig/m3 to 36.2 |ig/m3, with an average daily concentration of 18.4 |ig/m3. Typically, the PM2 5 composition was dominated by sulfate and carbon species. On average, the measured sulfate concentration, determined by ion chromatography, accounted for approximately 47% of the daily PM25mass. Total carbon, as measured by the IMPROVE thermal optical reflectance (TOR) method, accounted for approximately 38% of the PM2 5 mass, with elemental carbon contributing approximately 22% and organic carbon contributing approximately 77% of the total carbon. Additionally, nitrate contributed about 8.3% of the daily PM2 5 concentration. Table 6-1 summarizes the meteorological conditions during Phase I, and Table 6-2 summarizes the observed concentrations of the measured precursor gases during this period. Table 6-1. Summary of Daily Values for the Measured Meteorological Parameters During Phase I of Verification Testing Vertical Air Solar Wind Wind Wind Temp. Air Temp. Radiatio Total Speed Speed Direction @ 10 m @ 2 m RH n Press. Precip. (mph) (mph) (degrees) (C) (C) (%) (W/m2) (mbar) (in.) Average 3.35 0.09 196 20.0 16.6 89.4 162.8 979.7 0.0014 Max 6.45 0.29 298 24.1 22.5 95.8 246.1 986.7 0.03 Min 1.88 -0.03 106 14.6 12.1 80.2 47.9 974.5 0.00 Table 6-2. Summary of Daily Values for the Measured Precursor Gas Concentrations During Phase I of Verification Testing S02 (ppb) H2S (ppb) NO (ppb) NOz (ppb) NOx (ppb) 03 (ppb) Average 6.9 1.5 3.1 10.1 13.0 24 Max 12.8 2.9 10.4 17.4 27.4 51 Min 2.7 -0.6 0.14 5.3 5.3 5 21 ------- 6.1.1 Inter- Unit Precision The hourly mass concentration readings from the two BAM 1020 monitors for Phase I of the verification test are shown in Figure 6-la. (Note: The PM2 5 concentrations are presented in mg/m3, which are the same units as reported by the BAM 1020 monitors.) Breaks in the data indicate periods during which power outages occurred at the test site (August 6, 7, and 10-11). The two traces in Figure 6-la appear barely distinguishable. In Figure 6-lb these same data are plotted against one another to illustrate the correlation between the two monitors. For comparison with the PM2 5 FRM reference measurements, the hourly data were averaged from noon to noon for each day to correspond with the 24-hour sampling periods used in Phase I of the verification test. In Figure 6-2a, the noon-to-noon averages for Phase I of the verification test are presented for the two BAM 1020 monitors. A correlation plot of these data is shown in Figure 6-2b. The hourly BAM 1020 data were analyzed by linear regression, and the results of this analysis are presented in Table 6-3. The CV for these values was also determined according to Section 5.1, and the calculated CV is shown in Table 6-3. The regression analysis of the hourly data shows a correlation of r2 = 0.873 between the duplicate monitors. The results of the regression analysis indicate a bias between the two monitors, with Monitor 1 generally reading higher than Monitor 2 [slope = 0.932 (0.027)], where the number in parentheses is the 95% CI. A Student's t-test also shows a statistically significant bias between the duplicate BAM 1020 monitors with Monitor 1 reading 0.0017 mg/m3 higher than Monitor 2 on average for the hourly data. The calculated CV for the hourly data is 20.6%, much of which may be the result of the observed bias between the duplicate monitors. The regression results for the hourly data show that the intercept of the correlation plot [-0.0004 (0.0007)] includes zero at the 95% confidence interval. For the 24-hour average concentration results, the regression results in Table 6-3 indicate an r2 value of 0.986. The calculated CV for the 24-hour averages is 9.5%. As with the hourly data, a Student's t-test indicates a bias between the duplicate monitors. However, the slope of the correlation plot [0.973 (0.044)] is not statistically different from unity at the 95% confidence level. These data do show a negative intercept of 0.0013 (0.0010) mg/m3, which is statistically significant at the 95% confidence level. 22 ------- 0.08 Monitor 1 Monitor 2 0.07 - 0.06 - E 0.05 " 0.03 0.02 - 0.01 - 8/1/00 8/6/00 8/11/00 8/16/00 Date 8/21/00 8/26/00 8/31/00 Figure 6-la. Hourly PM2 5 Concentrations from Duplicate BAM 1020 Monitors During Phase I of Verification Testing 0.10 y = 0.932x - 0.000 r = 0.873 O) E 0.08 ^ 0.02 0.00 0.00 0.02 0.04 0.06 0.08 0.10 Hourly PM2.5 Readings - Monitor 1 (mg/m3) Figure 6-lb. Correlation Plot of Hourly PM2.5 Data fromDuplicate BAM 1020 Monitors During Phase I of Verification Testing 23 ------- 0.05 0.04 - O) Ł | 0-03 -i—' as -i—' c CD O o 0.02 O 0.01 - 0 fid, ¦ Monitor 1 ~ Monitor 2 8/1/00 8/8/00 8/15/00 Date 8/22/00 8/29/00 Figure 6-2a. 24-Hour Average PM2 5 Concentrations for Duplicate BAM 1020 Monitors During Phase I of Verification Testing 0.05 y = 0.973x-0.0013 r2 = 0.986 0.04 0.03 CO 0.02 0.01 0.00 0.00 0.01 0.02 0.03 0.04 0.05 24-Hour Average PM2 5 Concentration - Monitor 1 (mg/m3) 24 ------- Table 6-3. Linear Regression and Coefficient of Variation Results for Hourly and 24-Hour Average PM2 5 Concentration Values from Duplicate BAM 1020 Monitors During Phase I Parameter Hourly Data 24-Hour Average Data Slope (95% CI) 0.932 (0.027) 0.973 (0.044) Intercept (mg/m3) (95% CI) -0.0004 (0.0007) -0.0013 (0.0010) r2 0.873 0.986 CV 20.6% 9.5% 6.1.2 Comparability/Predictability In Figure 6-3a, the noon-to-noon averages of the BAM 1020 measurements are shown along with the PM2 5 FRM measurements for Phase I of the verification test. These PM2 5 concentration values were analyzed by linear regression according to Section 5.2 to establish the comparability of each of the BAM 1020 monitors with the PM2 5 FRM sampler. The resulting comparisons are plotted in Figure 6-3b; and the calculated slope, intercept, and r2 value of the regression analyses are presented in Table 6-4 for each monitor. For the regression results show r2 values of 0.909 and 0.921, respectively, for Monitor 1 and Monitor 2. For Monitor 1, the slope of the regression line is 1.169 (0.152) and the intercept is -0.0013 (0.0031) mg/m3. For Monitor 2, the slope is 1.142 (0.138) and the intercept is -0.0028 (0.0028). In both cases, the slopes are statistically different from unity and the intercepts are statistically indistinguishable from zero at 95% confidence. Table 6-4. Comparability of the BAM 1020 Monitors with the PM2 5 FRM During Phase I Regression Parameter Monitor 1 Monitor 2 Slope (95% CI) 1.169 (0.152) 1.142 (0.138) Intercept (mg/m3) (95% CI) -0.0013 (0.0031) -0.0028 (0.0028) r2 0.909 0.921 25 ------- 0.05 Monitor 1 Monitor 2 PM2.5 FRM 0.04 O) '.02 '.01 '.00 8/1/00 8/8/00 8/15/00 8/22/00 8/29/00 Date Figure 6-3a. Daily PM2 5 FRM Concentrations and the 24-Hour PM2 5 Average Concentra- tions from Duplicate BAM 1020 Monitors During Phase I of Verification Testing 0.05 ~ Monitor 1 ~ Monitor 2 0.04 O) E Linear (Monitor 1) Linear (Monitor 2) g 0.02 =3 O X ™ 0.01 0 0.01 0.02 0.03 0.04 0.05 PM2 5 FRM (mg/m3) Figure 6-3b. Correlation Plot of the 24-Hour Average PM2 5 Concentrations from Duplicate BAM 1020 Monitors and the PM2 5 FRM Concentrations During Phase I of Verification Testing 26 ------- 6.1.3 Meteorological Effects A multivariable model, as described in Section 5.3, was used to determine if variability in the readings of the BAM 1020 could be accounted for by meteorological conditions. This analysis involved a backward elimination process to remove from the analysis model those parameters showing no statistically significant influence on the results. This analysis indicated that vertical wind speed, relative humidity, and solar radiation all had a statistically significant influence on the readings from Monitor 1 relative to the FRM values at the 90% confidence level. Likewise, the results of Monitor 2 were dependent upon vertical wind speed and ambient temperature at both 2 meters and at 10 meters. The regression analysis indicates the following relationships: Monitor 1 = 1.08*FRM - 0.024*VWS - 0.00066RH - 0.00044Rad + 0.069 mg/m3 and Monitor 2 = 0.990*FRM - 0.027*VWS - 0.00048*T10 + 0.00056*T2 + 0.00147 mg/m3. In these equations, FRM is the PM2 5 FRM results in mg/m3, VWS is the vertical wind speed in mph, RH is the relative humidity in percent, Rad is solar radiation in watts per square meter, T10 and T2 are the ambient air temperature in Fahrenheit at 10 and 2 meters, respectively. Using the average values for PM2 5 and the various meteorological parameters during Phase 1 (Table 6-1), the equation above would predict an average PM25 reading of 0.0205 mg/m3 for Monitor 1. Monitor 1 = 1.08*0.0184 - 0.024*0.09 -0.00066*89.4 - 0.000044*162.8 + 0.069 = 0.0205 mg/m3. Based on the linear regression results (Table 6-4) and the average PM2 5 concentration during Phase 1, Monitor 1 would read, Monitor 1 = 1.169*0.0184 - 0.0013 = 0.0202 mg/m3 i.e., a difference of approximately 1.5%. 27 ------- The multivariable model would predict a PM2 5 reading of 0.0170 mg/m3 for Monitor 2. Monitor 2 = 0.990*0.0184 - 0.027*0.09 - 0.00048*20 + 0.00056*16.6 + 0.00147 = 0.0170 mg/m3 whereas the linear equation would predict Monitor 2 = 1.142*0.0184 - 0.0028 = 0.0182 mg/m3 i.e., a difference of approximately 7%. 6.1.4 Influence of Precursor Gases As described in Section 5.3, a multivariable analysis was performed to determine if precursor gases had an influence on the readings of the BAM 1020. This analysis involved a backward elimination to remove from the analysis model those parameters showing no statistically significant influence on the results. This analysis showed that none of the measured gases influenced Monitor 1 at the 90% confidence interval but that hydrogen sulfide had a statistically significant influence on the results of Monitor 2 relative to the FRM at the 90% confidence level. The regression analysis indicates the following relationship: Monitor 2 = 1.24*FRM - 0.0024[H2S] - 0.0024 mg/m3 where the concentration of hydrogen sulfide is in ppb. Using the average hydrogen sulfide concentration and PM2 5 concentration during Phase II, the multivariable equation above would predict an average value of 0.0168 mg/m3, whereas the linear equation would predict 0.0182 mg/m3, i.e., a difference of approximately 8%. 6.2 Phase II—Fresno (December 18, 2000 - January 17, 2001) During Phase II, daily 24-hour PM2 5 concentrations averaged 74 (ig/m3 and ranged from 4.9 (ig/m3 to 146 (ig/m3. A strong diurnal pattern was observed in the PM2 5 concentration, with the peak levels occurring near midnight. Particle composition was dominated by nitrate and carbon. On average, the overall PM2 5 concentration comprised 22% nitrate and 40% total carbon. Sulfate accounted for only about 2% of the daily PM2 5 mass. Both nitrate and sulfate were determined by ion chromatography, and carbon was determined by the IMPROVE TOR method. Table 6-5 summarizes the meteorological conditions during Phase II and Table 6-6 summarizes the observed concentrations of the measured precursor gases during this period. 28 ------- Table 6-5. Summary of Daily Values for the Measured Meteorological Parameters During Phase II of Verification Testing Wind Speed (mps) Wind Direction (Degrees) Change in Wind Direction (Degrees) Air Temp. (C) RH (%) Solar Radiation (W/m2) Press. (mmHg) Average 1.43 186 34.2 8.3 75.4 88.2 756.2 Max 4.18 260 48.8 12.8 92.0 123.5 761.7 Min 0.91 116 21.3 4.6 51.6 17.1 747.3 Table 6-6. Summary of Daily Values for the Measured Precursor Gas Concentrations During Phase II of Verification Testing CO (ppm) 03 (ppb) NO (ppb) N02 (ppb) NOx (ppb) Average 1.9 13 61.8 32.6 94.4 Max 3.3 28 119.9 50.3 170.2 Min 0.4 6 4.9 14.8 18.9 6.2.1 Inter- Unit Precision The hourly mass concentration readings from the two BAM 1020 monitors for Phase II of the verification test are shown in Figure 6-4a. In Figure 6-4b, these data are plotted against one another to illustrate the correlation between the two monitors. As was the case in Phase I, the two BAM 1020 monitors gave nearly indistinguishable readings of PM2 5 mass. For comparison with the PM2 5 FRM reference measurements, the hourly data were averaged from midnight to midnight for each day to correspond with the 24-hour sampling periods used in Phase II of the verification test. In Figure 6-5a, the midnight-to-midnight averages for Phase II of the verification test are presented for the two BAM 1020 monitors. A correlation plot of these data is shown in Figure 6-5b. The results of a linear regression analysis of these data are presented in Table 6-7. The CV for the hourly and the noon-to-noon average values were also calculated and are shown in Table 6-7. 29 ------- 0.35 0.25 - o O 0.15 - 0.05 - — Monitor 1 — Monitor 2 7'2000 12/24/2000 12/31/2000 01/07/2001 01/14/2001 Date Figure 6-4a. Hourly PM2 5 Concentrations from Duplicate BAM 1020 Monitors During Phase II of Verification Testing rz CM O y = 1.011x-0.002 0.99 0.25 - 0.15 - 0.05 - Monitor 1 - Hourly PM2.5 Readings (mg/m ) Figure 6-4b. Correlation Plot of Hourly PM2 5 Measurements from BAM 1020 Monitors During Phase II of Verification Testing 30 ------- 0.20 0.15 - 0.10 - 0.05 - 0.00 r ¦ Monitor 1 ~ Monitor 2 I mJ I 12/15/2000 12/22/2000 12/29/2000 Date 01/05/2001 01/12/2001 Figure 6-5a. Midnight-to-Midnight Average PM2 5 Concentrations from Duplicate BAM 1020 Monitors During Phase II of Verification Testing 0.20 y = 1.018x-0.002 r = 0.999 o) 0.10 0.00 0.00 0.05 0.10 0.15 0.20 24-Hour PM2 5 Averages - Monitor 1 (mg/m3) Figure 6-5b. Correlation Plot of 24-Hour Average PM2 5 Concentrations from Duplicate BAM 1020 Monitors During Phase II of Verification Testing 31 ------- Table 6-7. Linear Regression and Coefficient of Variation Results for Hourly and 24-Hour Average PM2 5 Concentration Values During Phase II Parameter Hourly Data 24-Hour Average Data Slope (95% CI) 1.011 (0.007) 1.018 (0.011) Intercept (mg/m3) (95% CI) -0.0016 (0.0007) -0.0022 (0.0010) r2 0.991 0.999 CV 9.9 % 6.4 % The hourly data from the duplicate monitors show a slope of 1.011 (0.007); intercept of-0.0016 (0.0007) mg/m3, and r2 = 0.991. The calculated CV for the hourly data is 9.9%. A Student's t-test shows a statistically significant bias between the duplicate BAM 1020 monitors, with Monitor 1 reading approximately 0.0007 mg/m3 higher than Monitor 2 on average for the hourly data. Although a statistical bias existed between the monitors, Figure 6-4a illustrates that the two monitors track each other very well. In fact, only because the duplicate monitors track each other so closely is it possible to determine such a small statistical difference between the two. The 24-hour average concentration results show a correlation between the duplicate monitors of r2 = 0.999. The calculated CV for the 24-hour averages is 6.4%. The agreement between the duplicate monitors is shown by a slope of 1.018 (0.011) and intercept of-0.0022 (0.001) mg/m3. 6.2.2 Comparability/Predictability In Figures 6-6a and 6-6b, the midnight-to-midnight averages of the BAM 1020 measurements are shown, along with the PM2 5 FRM measurements for Phase II of the verification test. These PM2 5 concentration values were analyzed by linear regression according to Section 5.2 to establish the comparability of each of the BAM 1020 monitors with the PM2 5 FRM sampler. The calculated slope, intercept, and r2 value of the regression analyses are presented in Table 6-8 for each monitor. The r2 values of the regression analyses of the 24-hour averages were 0.964 for Monitor 1 and 0.967 for Monitor 2. For Monitors 1 and 2, the slopes of the regression lines were 1.094 (0.080) and 1.111 (0.078), respectively. In each case, the slope was statistically different from unity at the 95% confidence level, indicating a positive bias relative to the FRM. The intercepts of the regression lines were not statistically different from zero at the 95% confidence level. 32 ------- 0.20 0.15 - CD c o 2 c CD O c o O 0.10 - 0.05 - 0.00 Monitor 1 Monitor 2 FRM 12/17/2000 12/24/2000 12/31/2000 Date 01/07/2001 01/14/2001 Figure 6-6a. Midnight-to-Midnight Averages from Duplicate BAM 1020 Monitors and the PM2 5 FRM Results During Phase II of Verification Testing 0.25 ~ Monitor 1 ~ Monitor 2 -- 1:1 Linear (Monitor 1 — Linear (Monitor 2) 0.05 0.1 0.15 0.2 0.25 PM2.5 FRM Concentration (mg/m Figure 6-6b. Correlation Plot from Duplicate BAM 1020 Monitors and the PM2 5 FRM During Phase II of Verification Testing 33 ------- Table 6-8. Comparability of the BAM 1020 Monitors with the PM2 5 FRM During Phase II Regression Parameter Monitor 1 Monitor 2 Slope (95% CI) 1.094 (0.080) 1.111 (0.078) Intercept (95% CI) (mg/m3) -0.0003 (0.0065) -0.0025 (0.0065) r2 0.964 0.967 Although both BAM 1020 monitors show a bias relative to the FRM, the collocated FRM sampling performed during Phase II showed a similar bias between the BGI and R&P FRM samplers. Additionally, flow audits performed during Phase II of testing showed a bias in the internal flow readings of the BAM 1020 monitors relative to the audit flows. In each case, the audit flow rates were between 3 to 5% higher than the displayed flow rates. Consequently, the calculated concentrations for the BAM 1020 monitors may be artificially high. 6.2.3 Meteorological Effects As with the data from Phase I, multivariable analysis was performed to determine if the meteorological conditions had an influence on the readings of the BAM 1020. This analysis involved a backward elimination process to remove from the analysis model those parameters showing no statistically significant influence on the results. This analysis indicates that, during Phase II, relative humidity had a statistically significant influence on the readings of both monitors relative to the FRM values at 90% confidence. The regression analysis indicates the following relationships: Monitor 1 = 1.13*FRM + 0.00040*RH - 0.033 mg/m3 and Monitor 2 = 1.14*FRM + 0.00034*RH - 0.031 mg/m3 where FRM represents the measured PM2 5 FRM values in mg/m3, and RH represents the average relative humidity in percent. For Monitor 1, the average PM2 5 concentration and the average relative humidity were used, the multivariable equation above would predict an average PM2 5 concentration of 0.0808 mg/m3, whereas the linear equation from Table 6-8 would predict a value of 0.0806 mg/m3. For Monitor 2, substituting the average values for Phase II into the multi- variable equation and linear equations give 0.0790 and 0.0797 mg/m3, respectively. In both cases, the effect of relative humidity on the BAM 1020 results were small (i.e., < 1%). 6.2.4 Influence of Precursor Gases Multivariable analysis was also performed to establish if a relationship exists between precursor gases (carbon monoxide, nitrogen dioxide, nitric oxide, nitrogen oxides, ozone) and the BAM 34 ------- 1020 readings relative to the FRM. This analysis showed no influence of the precursor gases on the readings of either monitor at the 90% confidence level. 6.2.5 Short-Term Monitoring During Phase II of the verification test, short-term monitoring was conducted on a five-sample- per-day basis throughout the test period. Table 6-9 presents the averages and the ranges of PM2 5 concentrations for these sampling periods during Phase II. Figure 6-7 shows the correlation between the time-weighted sum of the short-term measurements from the sequential filter sampler and the 24-hour FRM measurements. The slope and intercept of the regression line are 0.930 (0.077), and 2.2 (6.6) |ig/m3, respectively, with an r2 value of 0.960, where the numbers in parentheses are 95% CIs. Table 6-9. Summary of PM25 Levels During Phase II of Verification Testing PM2 5 Concentration (pg/in') Sampling Period 0000-0500 0500-1000 1000-1300 1300-1600 1600-2400 Average 81.0 52.2 56.8 46.7 87.7 Maximum 163.2 131.4 140.9 136.6 180.7 Minimum 3.4 7.7 4.8 2.2 7.2 In Figure 6-8, the averages of the BAM 1020 readings for all the short-term monitoring periods are plotted versus the corresponding PM2 5 concentration values from the sequential filter sampler. Linear regression analysis of these data was performed separately for each BAM 1020, and the results are presented in Table 6-10. Regression analyses also were performed separately for each of the five time periods during which the short-term samples were collected, (i.e., 0000-0500, 0500-1000, 1000-1300, 1300-1600, and 1600-2400). These regression results also are presented in Table 6-10. 35 ------- 0) Q. E 03 CD E a5 H i -c 0 -C CD °"E O) §> a5 < "O CD -t—< O) 'cd § 1 a) E 200 150 - 100 - y = 0.930x + 2.240 R2 = 0.960 50 100 150 200 BGI FRM PM2 5 Mass (ug/mJ) Figure 6-7. Correlation Plot of the Time-Weighted Average for the Short-Term Samples and the PM2 5 FRM 0.25 0.15 - o CM O < m Monitor 1 Monitor 2 -1:1 Linear (Monitor 1) Linear (Monitor 2) 0.05 - 0.05 0.1 0.15 0.2 0.25 Short-Term PM2.5 (mg/m ) Figure 6-8. Correlation Plot of Short-Term Monitoring Results and the Corresponding Averages from the Duplicate BAM 1020 Monitors During Phase II of Verification Testing 36 ------- Table 6-10. Regression Analysis Results for the Short-Term Monitoring Monitor 1 Monitor 2 Short-T erm Monitoring Period Slope Intercept (mg/m3) r2 Slope Intercept (mg/m3) r2 All 1.13 0.002 0.939 1.15 0.000 0.936 0000-0500 1.25 0.001 0.967 1.28 0.000 0.968 0500-1000 1.18 0.000 0.938 1.19 0.000 0.937 1000-1300 1.16 0.000 0.969 1.16 0.000 0.937 1300-1600 1.15 0.000 0.958 1.18 0.000 0.948 1600-2400 1.01 0.000 0.961 1.02 -0.001 0.963 The regression analyses for each of the five sampling periods show r2 values of 0.937, or greater, for both monitors. The slopes of the regression lines range from 1.01 to 1.25 for Monitor 1 and 1.02 to 1.28 for Monitor 2. (It should be noted that the reference measurements have not been corrected to account for the observed difference between the time-weighted average of the short- term samples and the FRM.) No statistically significant intercept was observed with either monitor for any of the five sampling periods. For both monitors, the best quantitative agreement (i.e., slope closest to 1.0) occurred during the 1600-2400 period. 6.3 Instrument Reliability/Ease of Use With the exception of three brief power outages between August 6 and 7, and an extended outage on August 10 and 11, 100% data recovery was achieved by each of the BAM 1020 monitors from the time of installation (August 1, 16:00) to the end of Phase I sampling (September 1, 12:00). After the power outages, the BAM 1020 monitors came back on line automatically and required no manual restart. No operating problems arose during Phase I of testing, and no maintenance was performed on either monitor during this phase. During Phase II of the verification test, 100% data recovery was achieved by each BAM 1020 monitor. No operating problems arose, and no maintenance was performed on either monitor during Phase II of testing. 6.4 Shelter/Power Requirements The BAM 1020 monitors were installed and operated inside an instrument trailer during each phase of testing. During each phase, a heater was used to condition the inlet of the BAM 1020 monitors to approximately 40°C. The monitors and pumps were run on a single 15 A circuit. 37 ------- 6.5 Instrument Cost The price of the BAM 1020 as tested is approximately $14,000. Filter tape for the BAM 1020 is the only consumable associated with this technology. Though not verified in this test, Met One suggests that a roll of filter tape is expected to last approximately 60 days of continuous monitoring at a 1-hour sample rate. 38 ------- Chapter 7 Performance Summary The BAM 1020 monitor is a semi-continuous particle monitor designed to provide hourly indications of the ambient particulate matter concentration. Duplicate BAM 1020 monitors were evaluated under field test conditions in two separate phases of this verification test. The duplicate monitors were operated side by side and were installed with a PM10 head and PM2 5 SCC to provide size selection of the aerosol. The results from each phase of this verification test are summarized below. 7.1 Phase I—Pittsburgh (August 1 - September 1, 2000) Regression analysis showed r2 values of 0.873 and 0.986, respectively, for the hourly data and for the 24-hour averages. The slopes of the regression lines were 0.932 (0.027) and 0.973 (0.044), respectively, for the hourly data and 24-hour averages; and no statistically significant intercept was observed in either case at the 95% confidence. The calculated CV for the hourly data was 20.6%; and, for the 24-hour data, the CV was 9.5%. Comparisons of the 24-hour averages with PM2 5 FRM results showed slopes of the regression lines for Monitor 1 and Monitor 2 of 1.169 (0.152) and 1.142 (0.138), respectively; and these slopes were significantly different from unity at the 95% confidence level. The regression results show r2 values of 0.909 and 0.921 for Monitor 1 and Monitor 2, respectively. Multivariable analysis of the 24-hour average data showed that the vertical wind speed, the relative humidity, and the solar radiation all had a statistically significant influence on the results of Monitor 1 at the 90% confidence level. Similarly, vertical wind speed and the ambient air temperature at both 2 meters and 10 meters influenced the results of Monitor 2 relative to the FRM at the 90% confidence level. On average, the combined effect of these parameters was below the approximately 10% uncertainty of the reference method. Multivariable analysis of the 24-hour average data showed that none of the measured gases had an influence on Monitor 1 at the 90% confidence level, but hydrogen sulfide had a statistically significant influence on Monitor 2. However, the combined effect of this gas on the instrument readings was approximately 8% on average, and below the approximate 10% uncertainty in the FRM reference measurements. 39 ------- 7.2 Phase II—Fresno (December 18, 2000 - January 17, 2001) Regression analysis showed r2 values of 0.991 and 0.999, respectively, for the hourly data and the 24-hour averages from Phase II. The slopes of the regression lines were 1.011 (0.007) and 1.018 (0.011), respectively, for the hourly data and 24-hour averages; and the intercepts were -0.0016 (0.0007) mg/m3 and -0.0022 (0.0010) mg/m3, respectively. The calculated CV for the hourly data was 9.9% and for the 24-hour data the CV was 6.4%. Comparison of the 24-hour averages with PM2 5 FRM results showed slopes of the regression lines for Monitor 1 and Monitor 2 of 1.09 (0.08) and 1.11 (0.08), respectively and these slopes were statistically different from unity at 95% confidence. No statistically significant intercept was observed in either case at the 95% confidence level. The regression results show r2 values of 0.964 and 0.967 for Monitor 1 and Monitor 2, respectively. Multivariable analysis of the 24-hour average data showed that relative humidity had a statistically significant influence on the readings of both monitors relative to the FRM values at 90% confidence. However, the effect was small in both cases and on average accounted for a change of approximately 1%. Multivariable analysis of the 24-hour average data indicated that none of the measured gases had an effect on either monitor at the 90% confidence level. In addition to 24-hour FRM samples, short-term sampling was performed on a five-sample-per- day basis. The BAM 1020 results were averaged for each of the sampling periods and compared with the gravimetric results. Linear regression of these data showed slopes of 1.13 and 1.15, respectively, for Monitor 1 and Monitor 2. The intercepts of the regression lines were 0.002 and 0.000 mg/m3, respectively; and the r2 values were 0.939 and 0.936, respectively. No operating problems arose, and no maintenance was performed on either monitor during testing. 40 ------- Chapter 8 References 1. Test/QA Plan for the Verification of Ambient Fine Particle Monitors, Battelle, Columbus, Ohio, June 2000. 2. "National Ambient Air Quality Standards for Particulate Matter; Final Rule," U.S. Environmental Protection Agency, 40 CFR Part 50, Federal Register, 62 (138) :38651- 38701, July 18, 1997. 3. Quality Management Plan (QMP) for the Advanced Monitoring Systems Pilot, Version 2.0, Battelle, Columbus, Ohio, October 2000. 4. "Quality Assurance Requirements for State and Local Air Monitoring Stations (SLAMS)." Appendix A to 40 CFR Part 58, Federal Register, 62 (138), p.65, July 18, 1997. 41 ------- |