Mobile-monitoring of Black Carbon and PM2.5 Air Pollution - Data only approach from Bangalore, India

Abstract

OPS 46: Exposure assessment to air pollution in Asia and Africa, Room 315, Floor 3, August 27, 2019, 4:30 PM - 5:30 PM INTRODUCTION Mobile monitoring campaigns that repeatedly cover all roads in an area (“wall-to-wall driving”) have shown they can robustly capture spatial patterns in on-roadway pollution. Yet, this technique has not been tried in a low-income country. India is home to several of the world’s most polluted cities. We explore the potential for mobile monitoring to provide high-resolution exposure data in Indian cities. Methods: Our aim was to over-sample each location, thereby shedding light on how many measurements are needed to obtain robust estimates of concentrations at each location. Our mobile platform was equipped with an aethalometer for black carbon (BC) and an atmos for PM2.5. We sampled each location 45 times during September - December 2018. Monitoring consisted of 2025 vehicle-km (135 h) of measurement, across two neighborhoods: a mixed-use neighborhood (“MUN”; 15.4 km of road) and a residential neighborhood (“RN”; 18.8 km of road). Raw BC and PM2.5 measurements were corrected for common measurement artifacts. We then computed average concentration for each 30m road segment and used Monte Carlo sampling to develop synthetic datasets representing N driving sessions, where N = 1, 2, 4, 6 … 45 driving sessions. We compared average concentrations of the subsampled data-set against the entire 45-day dataset. Results: Mean (median) values (135 h of data) were 33 (18) µg/m3 in MUN, 29 (15) µg/m3 in RN for BC, and 43 (39) µg/m3 [MUN], 31 (39) µg/m3 [RN] for PM2.5. Monte-carlo subsampling for BC revealed R2 = 0.9 for ten rides, and diminishing returns for >10 rides; in contrast, for PM2.5, R2 ~ 0.6 for ten rides, with continued R2 improvement with more data (R2 >0.9 for >30 rides). Conclusion: Data-only approach may be more suitable for BC measurements in the Indian context.