The DSI Distinguished Speaker Series highlights senior researchers who are applying data science to a broader scientific or academic expertise.

Guest Speaker

Dan Westervelt, Lamont Assistant Research Professor at Columbia University’s Lamont Doherty Earth Observatory in Palisades, NY

Moderated by: Abraham Liddell, DSI Postdoctoral Research Scientist

Details & Recording

Monday, October 24, 2022 (11:00 AM – 12:00 PM ET) – Hybrid

In-Person Location: 414 CEPSR – 4th Floor (Campus Level) – 530 W 120th St, New York, NY 10027

Virtual Location: Zoom link to be sent upon registration

Talk Information

Data Science and Air Pollution in Africa: From Low Cost Sensors to Supercomputers

Abstract: In Africa, air pollution exposure has been linked to 1 million premature deaths annually and, without intervention, these numbers are likely to climb. Sparse pollutant monitoring across the continent makes these estimates uncertain and also hinders the development of mitigation policies and regulations. Additionally, some of the most severe climate impacts are also felt in Africa, despite the continent’s relatively small contribution to global greenhouse gas emissions. This talk will highlight some recent efforts to close the air pollution data gap in Africa using a variety of methods spanning from traditional reference monitors, to consumer-grade low-cost sensors, to satellite retrievals, and to air quality and climate models. In particular we demonstrate the effectiveness of well-calibrated low cost particulate matter sensors in several previously-unmonitored megacities including Kinshasa (DRC), Lomé (Togo), Accra (Ghana), Nairobi (Kenya), and more. We find that many consumer-grade Plantower-based low-cost PM2.5 monitoring devices, such as PurpleAir, Clarity, and QuantAQ, perform well (r-squared ~ 0.6, MAE ~ 7 µg m-3) compared to locally available reference monitors, but can be improved dramatically (r-squared ~ 0.8, MAE ~ 2) using a variety of statistical methods, including linear regression, random forest regression, and Gaussian mixture regression. These well-calibrated sensors form the basis of dense urban networks of PM2.5 monitors in several African megacities, for example in Kinshasa (DRC), where the annual mean PM2.5 in 2019 was approximately 45 µg m-3, or ~8 times the WHO annual guideline. This talk will also present some research that leverages well-calibrated low-cost sensors and reference monitors to evaluate and improve the Africa region GEOS-Chem nested high-resolution model (25km x 25km). We also demonstrate the potential of fusing satellite data with ground-based observations using machine learning to develop high spatiotemporal resolution PM2.5 datasets, which may be useful in policy and health settings in addition to model evaluation. Finally, I will present climate modeling results that quantify how both local and remote aerosol emissions changes can have a substantial impact on African climate, in particular rainfall in the drought-stricken Sahel region of Africa. 

Bio: Dr. Daniel M. Westervelt is an Assistant Research Professor at Lamont-Doherty Earth Observatory (LDEO). Dr. Westervelt is also an affiliate faculty member of the Columbia University Data Science Institute, an affiliated scientist with NASA Goddard Institute for Space Studies, and an air pollution advisor to the US State Department. He is also a Columbia University Climate and Life Fellow. His current research spans from air quality and climate modeling to deployment and calibration of low cost sensors for air quality. Prior to his faculty position at Lamont-Doherty Earth Observatory, he worked as an Associate Research Scientist at LDEO, and as a Science, Technology, and Environmental Policy (STEP) postdoctoral research associate at Princeton University. He completed his PhD degree in May 2013 in Civil and Environmental Engineering from Carnegie Mellon University.