10th International Aerosol Conference September 2 - September 7, 2018 America's Center Convention Complex St. Louis, Missouri, USA
Abstract View
Spatial and Temporal Variability of Air Quality in Pittsburgh, Pennsylvania with a 50-Node RAMP Network
R. SUBRAMANIAN, Carl Malings, Rebecca Tanzer, Aliaksei Hauryliuk, Provat Saha, Aja Ellis, Rose Eilenberg, Sriniwasa P.N. Kumar, Naomi Zimmerman, Allen Robinson, Albert A. Presto, Carnegie Mellon University
Abstract Number: 1501 Working Group: Low-Cost and Portable Sensors
Abstract Low-cost sensors have enabled 24x7 continuous air quality monitoring with high spatial resolution, though the data is often associated with large uncertainties. The Real-time Affordable Multi-Pollutant (RAMP) monitor, developed at Carnegie Mellon University in collaboration with SenSevere (Pittsburgh, PA), can measure up to four gases out of carbon monoxide (CO), sulfur dioxide (SO2), nitric oxide (NO), nitrogen dioxide (NO2), ozone (O3), and volatile organic compounds (VOCs), along with carbon dioxide (CO2), temperature, and relative humidity. Calibration models based on periodic collocations with reference monitors and machine learning algorithms have enabled the RAMP to meet EPA's supplemental monitoring (CO, O3) or hot spot (NO2) data quality. SO2 performance is also promising at tens of ppb or higher. The RAMP uses either a Met-One Neighborhood PM monitor or a PurpleAir PM sensor to measure fine particulate mass (PM2.5). The PM data is corrected for humidity effects using an aerosol hygroscopic growth model and an average aerosol composition for Pittsburgh based on aerosol mass spectrometry data. A final correction is applied based on long-term collocation of one RAMP with a local regulatory monitor to account for Pittsburgh-specific aerosol size distributions that may be different from the factory calibration.
A network of fifty RAMP sensors has been deployed in and around the city of Pittsburgh, Pennsylvania since Summer 2017, representing (to our knowledge) the largest long-term deployment of multipollutant sensor systems. The RAMPs are hosted by schools, residences, public facilities, businesses, and religious establishments. Locations identified by land-use attributes (traffic, restaurant density, and building height) were obtained through direct contact or with the help of local community organizations and interested individuals.
The 50-RAMP network allows us to examine the spatial and temporal variability of air quality across a major urban area at an unprecedented scale. Elevated levels of CO and PM2.5 were observed in the residential East End of Pittsburgh, which is home to several schools and restaurants, with arterial roads running through it and connecting to a major highway, because of which vehicular traffic can exhibit sharp diurnal patterns. CO variability in some of these locations – particularly at some schools – was relatively high, with long-term standard deviation in CO between 300-400 ppb. Sites in downtown Pittsburgh, which has heavy traffic during the day and is bordered by several highways, showed long-term CO standard deviations below 150 ppb. This could be because while average CO during Summer 2017 in the East End locations were comparable to that in downtown Pittsburgh (300-400 ppb), CO was higher in the East End the following Winter.
PM2.5 averages varied from 7.7-11.6 µg/m3 across the RAMP network. While the highest PM2.5 concentrations were recorded in the Monongahela Valley (home to coke plants and a steel mill), similar concentrations (about 2 µg/m3 higher than the nearest regulatory monitor) were observed in the Homewood neighborhood, an environmental justice area within city limits. The lowest PM2.5 concentrations were observed in the suburban neighborhoods of Fox Chapel, Aspinwall, and Mt Lebanon– but also in the near-downtown, poorer Hill District neighborhood. Winter-time PM2.5 is higher than summer concentrations in the river valleys by about 3 µg/m3, but the exact opposite is true for the relatively elevated East End locations; either shift is about 30% of the city-wide average PM2.5. NO2 levels vary by as much as ±3 ppb across the RAMP network, though a pattern is not readily evident and needs further investigation. Additional detailed analysis and insights from the RAMP network with learnings for future low-cost sensor networks will be presented.