Abstract View
Performance Characterization of Low-Cost Sensor Observations in a Near-Source Environment in Rural Malawi
ASHLEY BITTNER, Eben Cross, David Hagan, Carl Malings, Eric Lipsky, Andrew Grieshop, North Carolina State University
Abstract Number: 374
Working Group: Instrumentation and Methods
Abstract
Low-cost sensor packages are one approach to increase the spatial and temporal resolution of ambient air quality (AQ) monitoring in low-resource settings. The accuracy, precision, long-term (>1 year) stability of low-cost sensors, and the transferability of calibrations built outside of the deployment environment remain some of the key questions for global low-cost monitoring efforts. To investigate these topics, we characterize long-term performance of three ‘ARIsense’ low-cost integrated sensor packages (QUANTAQ, Inc.; Aerodyne Inc.) using 13 months of ambient data collected from three sites in rural Malawi. Observations indicate minimal decay in sensor response, but issues related to intermittent solar power and blocked inlets. To characterize the performance of the ARIsense’s integrated optical particle counter (OPC), we compare 130 hours of collocated particle measurements to an optical nephelometer/integrated filter measurement device (MicroPEM, RTI International) under ambient conditions in rural Malawi. We find that the performance of the OPC compared to the MicroPEM (using the Pearson correlation coefficient, r) varied with time of day (-0.06 < r < 0.93), relative humidity (0.35 < r < 0.7), temperature (0.06 < r < 0.8), wind direction (0.02 < r < 0.83) and the background concentration at the site (-0.1 < r < 0.7). In general, OPC performance decreased with increasing RH and increased with background concentration. To characterize the ARIsense’s suite of integrated electrochemical gas sensors (CO, Ox, NO2, NO), we compare 15 days of observations collocated with EPA reference instruments in Raleigh, NC. We compare the performance of calibration models (HDMR, MLR, kNN and RF hybrid) built using the collocation data and assess their robustness to reliably measure gas concentrations in the deployment environment. Finally, we discuss the ability of integrated sensor packages to inform source attribution and local AQ characterization, despite performance limitations.