![Many benefits over traditional technologies
• Continuous, real-time readings
• Small & quiet (wearable! drones!)
• Battery and solar power capabilities
• New multi-wavelength optical technologies allow source
apportionment
• Many are low cost — less than $2500 (USEPA [Williams et al 2014])
• Often closer to $300 - $1000
• Surround a facility’s fence-line in sensors for the cost of a
reference monitoring station
Disadvantages and limitations
• Often measure proxies (eg., light scattering vs. actual PM mass)
• Require calibration, parameters of which can vary in time and space
• Often made for environmental monitoring, but small enough be worn
• Not always durable enough for harsh workplace settings
Sensor revolution
Kerry Klein, KVPR
Airborne dust sensing at an open-pit mine.
(Alvarado et al 2015)](https://image.slidesharecdn.com/hill-sensorsandanalyticsinexposureassessment-aihajan2020-forcirculation-200522044345/85/Air-monitoring-sensors-and-advanced-analytics-in-exposure-assessment-14-320.jpg)
Air monitoring sensors and advanced analytics in exposure assessment
- 1. Air monitoring sensors and advanced analytics in exposure assessment L. Drew Hill, PhD, MPH AethLabs American Industrial Hygiene Association - SF Bay Area Dinner Talk - January 28, 2020
- 2. • Common airborne occupational hazards • Traditional monitoring techniques • An emerging next generation of methods • The sensor revolution & applications in exposure assessment • The data revolution & applications in exposure assessment Overview
- 3. Common airborne occupational hazards
- 4. Common airborne occupational hazards Particulate Matter • Particulate matter of specific, health-relevant sizes • Total suspended particles • PM1.0 • PM2.5 • PM10 • Silica • Black Carbon (BC) Gases and vapors • Carbon dioxide (CO2) • Nitrogen and sulfur oxides (NO, NO2, SO, SO2) • Volatile organic compounds (VOC) • Benzene Metals • Chromium • Cadmium • Lead Greenstone and Qing Fan, U Chicago
- 5. Common airborne occupational hazards Particulate Matter • Particulate matter of specific, health-relevant sizes • Total suspended particles • PM1.0 • PM2.5 • PM10 • Silica • Black Carbon (BC) Gases and vapors • Carbon dioxide (CO2) • Nitrogen and sulfur oxides (NO, NO2, SO, SO2) • Volatile organic compounds (VOC) • Benzene Metals • Chromium • Cadmium • Lead Greenstone and Qing Fan, U Chicago
- 7. Traditional monitoring techniques Laboratory-analyzed filter samples • Obtrusive — filter measurement often require heavy belt pack pump • No real-time insights (time-integrated assessment only) • Traditionally employ large, expensive offsite analysis equipment • Gas chromatography • X-Ray Fluorescence spectrometry (XRF) • Fourier-Transform Infrared spectroscopy (FTIR) Colorimetric and other on-site badges • Interpretation may vary from person-to-person (color) • No real-time insights • Low sensitivity Gas adsorbent tubes • Often requires time consuming offsite analysis with large, expensive equipment Traditional real-time monitors • Little-to-no pollutant speciation • Expensive; often bulky and loud • Environmental vs. personal assessment USGS US CDC Gurung et al 2012
- 8. An emerging next generation
- 9. An emerging next generation Computers have hit critical size and cost thresholds. Pictorial Parade/Archive Photos
- 10. An emerging next generation Shrinking laboratory equipment • Portable and potentially wearable Fourier-Transform Infrared spectroscopy (FTIR) • Multiple wavelength monitoring now available in a handheld device Onsite—analyzed filter samples • Recent success with Fourier-Transform Infrared spectroscopy (FTIR) • Timely, but still no real-time insights (time-integrated assessment only) Smaller personal pumps • Typically employ an optical real-time mechanism • Still require analysis of in-line filter for pollution speciation London Mayor, Sadiq Khan, with a realtime black carbon monitor credit: Graeme Robertson “Small” belt-wearable sampling pump. credit: AIHA and SKC Ltd A simulated image of a wildfire fighter carrying a portable black carbon monitor into the field. credit: etserv.be
- 12. Sensing technologies are shrinking, too Monitor costs are dropping Monitor prevalence is growing nearly exponentially Dye 2018 Sensor revolution
- 13. Sensing technologies are shrinking, too Monitor costs are dropping Monitor prevalence is growing nearly exponentially PurpleAir (Jan 26, 2020) Sensor revolution
- 14. Many benefits over traditional technologies • Continuous, real-time readings • Small & quiet (wearable! drones!) • Battery and solar power capabilities • New multi-wavelength optical technologies allow source apportionment • Many are low cost — less than $2500 (USEPA [Williams et al 2014]) • Often closer to $300 - $1000 • Surround a facility’s fence-line in sensors for the cost of a reference monitoring station Disadvantages and limitations • Often measure proxies (eg., light scattering vs. actual PM mass) • Require calibration, parameters of which can vary in time and space • Often made for environmental monitoring, but small enough be worn • Not always durable enough for harsh workplace settings Sensor revolution Kerry Klein, KVPR Airborne dust sensing at an open-pit mine. (Alvarado et al 2015)
- 15. Particle counters: GRIMM vs. Alphasense OPC-N3 Sensor revolution GRIMM OPC-N3 Price (USD) $25,000 + ~ $350 Maintenance Costly, frequent “set it and forget it” (nearly) Features 0.25 - 32 um (31 size bins) 0.38 - 40 um (24 size bins) Reliability Federal Equivalence Method (FEM) Take a look Wearability Minimal High SCAQMD 2018OPC-N3 envirotech-online directindustry.com
- 16. Gas-Phase Sensors Sensor revolution SCAQMD AQ-SPEC
- 18. Carbon nanotubes NIOSH REL for carbon nanotubes (CNT): 7 µg/m3 • Set to estimated upper LOQ of Method 5040 for elemental carbon (EC), organic carbon (OC), and total carbon (TC) Horne and O’Shaughnessy (2013) • Handheld sensors can be used as surrogate method for 5040 • When CNT 5040 LOQ, handheld sensors can “aid exposure assessment interpretations” Ogura et al (2013) • Applied handheld sensors in a CNT manufacturing setting to measure environmental concentrations Handheld sensors for personal exposure assessment to nanomaterials (Asbach et al 2017) • Reviewed several next generation nonmaterial measurement devices • “Robust and ready for field use” • Accuracy is lower than conventional equipment, but “more than compensated” by breathing zone placement Ogura et al 2013
- 19. Gurung et al 2012 Drivers and near-roadway workers Bus, taxi, & Uber drivers; police officers; subway conductors • Substantial air pollutant exposures • Traffic police: 128 µg/m3 PMresp, 3.5 mg/m3 CO, 11.5 µg/m3 Benzene in Italy (Cattaneo 2010), • Drivers and Metro Riders in Belgium: ~ 5000 ng/m3 of Black Carbon (Dons et al 2012) • Smaller devices & longer battery life reduces worker strain during measurement Vice News, HBO WNYC One parameter; bulky backpack Multiple parameters; unobtrusive vest Gurung et al 2012
- 20. DPM and other source differentiation Diesel particulate matter (DPM) • Typically < 1 µm in diameter & composed of BC • Comprised of > 40 known carcinogens • Sources: ships, trains, trucks, buses MultiWavelength Measurement & Analysis • AethLabs MA350: • 375 nm, 470 nm, 528 nm, 625 nm, 880 nm • Black carbon = 880 nm • Indicative of diesel emissions • UVPM = 375 nm • Indicative of woodsmoke, tobacco, biomass burning • PM smaller than most optical monitors can detect MA350; coin images added for approximate scale
- 21. Carcinogen of great concern to mine and construction workers Traditionally measured via Fourier transform Infrared spectroscopy (FTIR) — offsite lab analysis with large bench top devices. Quantum cascade laser FTIR • QCL-IR — portable RCS measurement (Wei et al 2017) Can be used to determine exposures during and after shifts on premises Advances in FTIR sensor technologies may allow for realtime, personal exposure assessments USGS Wei et al 2017 Traditional FTIR Respirable crystalline silica (RCS) QCL-IR (a) performing similarly to traditional FTIR (b) Insurance Journal
- 22. Data revolution
- 23. Decreases in computer size and cost, increases in performance Phalanx of data from consumer devices, government surveys, employer databases Open source data science movement • The internet (geeks around the world can congregate!) • Python and R (free, transparent, and very powerful statistical analysis software) • GitHub (host code) Major advances in software and applied statistical methods emerging from Silicon Valley and Academic Research • Artificial Intelligence field may have started in 1956 at Dartmouth College, advanced at universities in the heart of silicon valley like UC Berkeley and Stanford and elsewhere (UW, MIT, etc.) • Google’s “TensorFlow” free, advanced machine learning library in Python and other languages • Netflix recently released its “Metaflow” Python-based data science management tool to the public • Companies like Apple, Google, and others are beginning to publish public blogs about their artificial intelligence work Data revolution Franki Chamaki
- 24. Combination of sensor data and non-monitoring data to predictively model PM2.5 exposure models in cooks, Lao PDR (Hill et al 2019) • Measuring environmental concentrations is much (much!) easier than convincing a person to wear a device for extended periods • Area Concentration ≠ Personal Exposure • 48 hr avg in kitchen: 462 µg/m3 • 48 hr avg personal: 123 µg/m3 • Surveys are common (e.g., Demographic Health Survey is administered in > 90 countries) • Can “easier” environmental measurements be combined with existing datasets and analyzed using ML to predict actual personal exposures? Data revolution Hill et al 2019 Leverage hidden relationships to improve exposure estimation
- 25. Data revolution Increased use of Machine Learning (ML) to improve sensor calibrations
- 26. Hill et al 2018 Data revolution Existing and emerging techniques in Machine Learning combined with the throngs of publicly available data produced in near-realtime may also be used to improve the accuracy and utility of low-cost sensors.
- 27. Hill et al 2018 Existing and emerging techniques in Machine Learning combined with the throngs of publicly available data produced in near-realtime may also be used to improve the accuracy and utility of low-cost sensors. Raw outdoor PM2.5 sensor data vs. a large, expensive reference monitor A comparison vs. Reference of the same data run through an ML- enhanced ensemble calibration model produced using open source software (R) and publicly available data (weather, satellite). A much better fit (close to 1:1 for all sensors!) Will this help indoor sensing as well? Data revolution
- 28. Advanced Data Methods & Sensors Leveraged hidden relationships Increased Reliability Reduced Cost Portability Wearability Ubiquity
- 29. Thank You! L. Drew Hill drew.hill@aethlabs.com (415) 529-2355
- 30. Works Cited 1. Greenstone, M.; Qing Fan, C. Introducing the Air Quality Life Index: Twelve Facts about Particulate Air Pollution, Human Health, and Global Policy; University of Chicago, Energy Policy Institute: Chicago. 2. Gurung, A.; Bell, M. L. Exposure to Airborne Particulate Matter in Kathmandu Valley, Nepal. J Expo Sci Environ Epidemiol 2012, 22 (3), 235–242. https://doi.org/10.1038/jes.2012.14. 3. Dye, T. Air Quality Sensor Deployment Rapidly Increasing in California, 2018. 4. Williams, R.; Kilaru, V.; Snyder, E.; Kaufman, A.; Dye, T.; Rutter, A.; Russell, A.; Hafner, H. Air Sensor Guidebook; National Exposure Research Laboratory, Office of Research and Development, 2014; p 73. 5. Alvarado, M.; Gonzalez, F.; Fletcher, A.; Doshi, A. Towards the Development of a Low Cost Airborne Sensing System to Monitor Dust Particles after Blasting at Open-Pit Mine Sites. Sensors (Basel) 2015, 15 (8), 19667–19687. https://doi.org/10.3390/s150819667. 6. Klein, K. How PurpleAir's Founder Put Air Quality Monitoring In The Hands Of The Public https://www.kvpr.org/post/how- purpleairs-founder-put-air-quality-monitoring-hands-public (accessed January 2020). 7. SCAQMD AQ-SPEC. Field Evaluation: Alphasense OPC-N3 Sensor http://www.aqmd.gov/docs/default-source/aq-spec/field-evaluations/ alphasense-opc-n3---field-evaluation.pdf?sfvrsn=12 (accessed January 2020). 8. SCAQMD AQ-SPEC. Summary Gas-Phase http://www.aqmd.gov/aq-spec/evaluations/summary-gas (accessed January 2020). 9. Asbach, C.; Alexander, C.; Clavaguera, S.; Dahmann, D.; Dozol, H.; Faure, B.; Fierz, M.; Fontana, L.; Iavicoli, I.; Kaminski, H.; MacCalman, L.; Meyer-Plath, A.; Simonow, B.; Tongeren, M. van; Todea, A. M. Review of Measurement Techniques and Methods for Assessing Personal Exposure to Airborne Nanomaterials in Workplaces. Science of the Total Environment 2017. https://doi.org/10.1016/ j.scitotenv.2017.03.049. 10.Ogura, I.; Kotake, M.; Hashimoto, N.; Gotoh, K.; Kishimoto, A. Release Characteristics of Single-Wall Carbon Nanotubes during Manufacturing and Handling. J. Phys.: Conf. Ser. 2013, 429, 012057. https://doi.org/10.1088/1742-6596/429/1/012057. 11.Dons, E.; Int Panis, L.; Van Poppel, M.; Theunis, J.; Wets, G. Personal Exposure to Black Carbon in Transport Microenvironments. Atmospheric Environment 2012, 55, 392–398. https://doi.org/10.1016/j.atmosenv.2012.03.020.
- 31. Works Cited 12.Cattaneo, A.; Taronna, M.; Consonni, D.; Angius, S.; Costamagna, P.; Cavallo, D. M. Personal Exposure of Traffic Police Officers to Particulate Matter, Carbon Monoxide, and Benzene in the City of Milan, Italy. J Occup Environ Hyg 2010, 7 (6), 342–351. https://doi.org/ 10.1080/15459621003729966. 13.Horne, A.; O’Shaughnessy, P. An Investigation of Carbon Nanotube Exposure Assessment Methods. MS, University of Iowa, Iowa City, Iowa, USA, 2013. https://doi.org/10.17077/etd.lwyh0s9s. 14.McCann, A. City Cyclists: Here’s How Much Pollution You’re Actually Inhaling. Vice, 2018. 15.What’s in the Air as You Cycle City Streets? | WNYC | New York Public Radio, Podcasts, Live Streaming Radio, News https:// www.wnyc.org/story/bike-way-whats-air-you-cycle-city-streets/ (January 2020). 16.Wei, S.; Kulkarni, P.; Ashley, K.; Zheng, L. Measurement of Crystalline Silica Aerosol Using Quantum Cascade Laser–Based Infrared Spectroscopy. Sci Rep 2017, 7 (1), 13860. https://doi.org/10.1038/s41598-017-14363-3. 17.Technology Planning and Management Corporation. Report on Carcinogens Background Document for Silica, Crystallin (Respirable Size); NTP, 1998. (Note: not specifically cited in text, but it is good background information). 18.Hill, L. D.; Pillarisetti, A.; Delapena, S.; Garland, C.; Pennise, D.; Pelletreau, A.; Koetting, P.; Motmans, T.; Vongnakhone, K.; Khammavong, C.; Boatman, M. R.; Balmes, J.; Hubbard, A.; Smith, K. R. Machine-Learned Modeling of PM2.5 Exposures in Rural Lao PDR. Science of The Total Environment 2019, 676, 811–822. https://doi.org/10.1016/j.scitotenv.2019.04.258. 19.Zimmerman, N.; Presto, A. A.; Kumar, S. P. N.; Gu, J.; Hauryliuk, A.; Robinson, E. S.; Robinson, A. L.; R. Subramanian. A Machine Learning Calibration Model Using Random Forests to Improve Sensor Performance for Lower-Cost Air Quality Monitoring. Atmos. Meas. Tech. 2018, 11 (1), 291–313. https://doi.org/10.5194/amt-11-291-2018. 20.Hill, L. D.; Pillarisetti, A.; Smith, K. R.; Libicki, S. Improving Pollution Source Resolution for Real Time Low Cost Sensors Using Widely Available Data Resources: A Proof of Concept, 2018. https://asic.aqrc.ucdavis.edu/sites/g/files/dgvnsk3466/files/inline-files/ Drew%20Hill_Upload_0.pdf. Accessed January 2020. Note: Many photos borrowed from the internet are credited in the slides — many, but not all, of them came from unsplash.com. When particularly relevant to the presentation topic, full sources are also listed in the above Works Cited.