PM2.5 Sensor Intercomparisons and Regional Trend Assessments from Low-Cost Sensor Networks in Accra, Ghana and Lomé, Togo

GARIMA RAHEJA, Emmanuel Appoh, Ebenezer Appah-Sampong, Maxwell S. Sunu, John K. Nyante, Allison Felix Hughes, Celeste McFarlane, Rob Pinder, Stefani Penn, R. Subramanian, Michael Giordano, Levi Stanton, Daniel Westervelt, Columbia University

     Abstract Number: 407
     Working Group: Urban Aerosols

Abstract
Metropolises in sub-Saharan Africa experience high levels of ambient air pollution, but remain scarcely measured by reference-grade monitors. We combine insights from three years of data collected by Purple Air and Clarity sensor networks, with newly developed correction factors for low-cost sensors in this region based on collocation with reference monitors.

For Accra, Ghana, we utilize a network of 18 Clarity (low-cost) sensors deployed since August 2018 to analyze regional trends, and assess the reductions in PM2.5 caused by clean air interventions as well as the COVID19 pandemic. We find that calibrated daily averaged PM2.5 is 26.4 µg m-3 (within the EPA Ghana standard for 24-hour mean of 35 µg m-3). On average 94.8% of days exceed the WHO daily standard.

In Lomé, Togo, our 5-node PurpleAir network deployed in 2019 represents the first ever long-term ambient PM2.5 measurements in the city to our knowledge. We find the daily PM2.5 average is 23.5 µg m-3 (1.5 times the WHO daily average guideline of 15 µg m-3). In Accra and Lomé, sensor sites show diurnal patterns with morning peaks 2.5 times stronger than evening peaks (with notable exceptions linked to areas with intense rush hour traffic jams), as well as annual patterns with mean PM2.5 concentrations 1.5 times higher during the Harmattan period. At all sites, more than 87% of measured days surpass WHO PM2.5 Daily Guidelines.

Finally, we present the results from one year of intercomparison between various low-cost (PurpleAir, Clarity, MODULAIR) monitors and reference-grade (BAM-1020, Teledyne T-640) monitors. We find that Clarity monitors show the best correlation with reference-grade monitors, with R2 = 0.76 and Mean Absolute Error = 3.36 µg m-3. This intercomparison research contributes to our development of globally-applicable models for calibrating and correcting low-cost sensors.

[1] https://pubs.acs.org/doi/full/10.1021/acsearthspacechem.1c00217
[2] https://pubs.acs.org/doi/10.1021/acsearthspacechem.1c00391