The below claims and disclaimers are based on comparisons between the AirBeam, a Thermo Scientific pDR-1500 with a PM2.5 cut-point inlet, and teflon filter samples subjected to gravimetric analysis. The pDR-1500 is a $5,000, 2.5 lb air quality monitor frequently used by government and academic researchers to evaluate personal exposure to fine particulate matter or PM2.5. Teflon filter samples were taken with a Leland Legacy 10L pump and PM2.5 cut-point inlet and weighed at the NYU School of Medicine’s filter weighing room, which meets EPA guidelines for filter conditioning, storage, and gravimetric measurement of PM2.5 and PM10 filters. Filters subjected to gravimetric analysis are the “gold standard” for measuring PM2.5. Additional research is required to fully characterize the performance of the AirBeam and we look forward to working with the AirCasting community to “fill in the gaps”.
When presenting our performance data on the AirBeam below, we include R2 or R-squared values to indicate how the AirBeam compares with other methods for measuring PM2.5. R2 is a statistical measure that indicates how well data fit a statistical model, in this case, the prediction of the Y-axis (AirBeam) from the X-axis (pDR-1500) using a linear (straight) or nonlinear (curved) line. The R2 value is a fraction that ranges from 0.0 to 1.0 with higher values indicating that the regression came more closely to the points. An R2 value of 1.0 means that the predictive power of the model is perfect, that all the points lie along the line or curve with no scatter.
Below 100 micrograms per cubic meter (µg/m³), samples collected in ambient air in Manhattan (samples were collected on 11 different occasions and averaged over 12 hour periods) and while burning cardboard indoors (samples were collected over a 1 hour period and averaged every minute) both showed a strong linear relationship between the AirBeam and pDR-1500 measurements. As illustrated in Figure 1, the R2 values below 24 µg/m³ for two AirBeams in ambient air in Manhattan were .98 or better.
As illustrated in Figure 2, the R2 values below 100 µg/m³ for four AirBeams while burning cardboard indoors were .94 or better. Also shown in Figure 2, “out-of-the-box” variability between AirBeams is more pronounced as the measurements climb above 30 µg/m³. Meaning that measurements recorded by two AirBeams exposed to identical air samples may begin to drift apart as PM2.5 concentrations increase. Out-of-the-box variability can be substantially reduced by using the AirCasting app calibration feature (still in beta) and adjusting the side-facing potentiometer on the Shinyei PPD60PV.
Because the relationship between the AirBeam and pDR-1500 measurements becomes increasingly non-linear above 100 µg/m³, a nonlinear regression curve was used to determine the relationship between the AirBeam and pDR-1500 measurements at higher concentrations, see Figure 3 (samples were collected over a 1 hour period and averaged every minute). During separate sampling runs, we calculated R2 values for the nonlinear regression curve ranging from 0.60 to 0.80. The decrease in R2 values as compared to the linear regression is likely attributed to higher variability near and above the AirBeam’s maximum limit of detection, which we estimate to be approximately 400 µg/m³.
Additional research is required to see how the maximum limit of detection is impacted by the reflectivity of the aerosol being sampled. The relative reflectivity of aerosols impacts the AirBeam measurements. Highly reflective aerosols, like wood smoke, bias the AirBeam measurements upwards, whereas less reflective aerosols, like diesel exhaust, bias the AirBeam measurements downwards.
During ambient air sampling in Lower Manhattan during the summer months, measurements from a pDR-1500 and two Airbeams were compared against a teflon filter subjected to gravimetric analysis, see Figure 4. Sampling was done in 12-hour averages each day for 11 days and averaged to compare the real time instruments against the gravimetric filters. When compared against the gravimetric filters, the R2 value of AirBeams was found to be 0.70 compared to 0.76 for the pDR-1500. Time weighted averages of the gravimetric filter data showed consistently higher values as compared to the pDR-1500 at ambient levels. We assume this downward bias is also in effect with the AirBeam, since both are light scattering particle counters. Further, we assume part of this this bias can be attributed to the relative reflectivity of the aerosol being measured. The R2 value of the pDR-1500 measured against the AirBeams during these 12-hour day averages was found to be 0.98.
Research conducted by others on light scattering particle counters indicates that high relative humidity (>80%) is likely to have a negative impact on the accuracy of the AirBeam. When relative humidity is high, aerosols take on water becoming more reflective. Additional research is required to better characterize this effect as it applies to the AirBeam.
AirBeam performance data collection, analysis, and findings are the work of Alex Besser and Michael Heimbinder. Alex is a graduate student in Environmental Toxicology at New York University. Michael is the Founder and Executive Director of HabitatMap and AirBeam Lead Developer. Dr. George Thurston, Alex’s academic adviser and professor of Environmental Medicine at New York University School of Medicine, provided the material resources and guidance that made this research possible.
PUBLISHED: October 16, 2014