A220-0010
First measurements of PM2.5 in Kinshasa, Democratic Republic of Congo Using Field-Calibrated Low Cost Sensors

Wednesday, 16 December 2020
Poster
Celeste McFarlane, Columbia University of New York, Chemical Engineering, Palisades, United States, Paulson Kasereka, L’Ecole Régionale Postuniversitaire d’Aménagement et de Gestion intégrés des Forêts et Territoires tropicaux, Kinshasa, Congo, Raymond Sinsi Lumbuenamo, World Bank Group, Kinshasa, Congo, Ranil Dhammapala, Washington State, Department of Ecology, Pullman, WA, United States, V Faye McNeill, Columbia University of New York, Chemical Engineering, Palisades, NY, United States, Carl Malings, Carnegie Mellon University, Mechanical Engineering, Pittsburgh, PA, United States, R Subramanian, Carnegie Mellon University, Pittsburgh, PA, United States and Daniel Westervelt, Columbia University of New York, Palisades, United States
Abstract:
Estimates of air pollution mortality in sub-Saharan Africa are severely limited by a lack of surface observations. Kinshasa, DRC is a megacity with over 11 million inhabitants and no reference grade air pollution monitors. Recently, a few reference grade monitors have been deployed in other parts of sub-Saharan Africa, including Kampala, Uganda. Reference grade air quality monitors can serve as an important basis for the correction and calibration of low cost sensors (LCS) for air quality monitoring. LCS may offer a path to remedy the lack of air pollution data in Africa, but they are affected by environmental conditions such as temperature and humidity and therefore require careful calibration and correction to establish high quality data. We present the first ever in-field calibration of LCS against reference grade monitors in Kampala, Uganda. A Purple Air PM2.5 monitor was colocated next to the Kampala US Embassy BAM-1020 starting in August 2019. Raw Purple Air data is strongly correlated with BAM (r = 0.94), but has a mean absolute error of approximately 12 μg/m3. Two calibration models, multiple linear regression and a random forest machine learning approach, can decrease mean bias to less than 0.01 from 12 μg/m3 and improve the correlation coefficient to 0.96. We then apply the colocated in-field correction factors to four Purple Air sensors in Kinshasa, DRC and one in neighboring Brazzaville, deployed beginning April 2018. Data in Kinshasa shows peaks from June-September, when rain is infrequent, with a trend of decreasing peak size every calendar year (approx. 80 μg/m3 in 2018, 55 μg/m3 in 2019 and 38 μg/m3 in 2020). Regionally averaged diurnal trends show a peak of approximately 48 μg/m3 at 6 pm GMT (7 PM WAT). Comparing data from the COVID lockdown period of 2020 to that of 2019 shows a reduction in PM2.5 concentrations from 45 μg/m3 in April of 2019 to approximately 10 μg/m3 in April of 2020. Daily mean PM2.5 in Kinshasa is nearly always above the WHO Tier 1 guideline for PM2.5 of 25 μg/m3 , suggesting a need for clean air solution implementation in Kinshasa.
<< Previous Abstract | Next Abstract >>