Modeling indoor exposure to PM2.5 and black carbon in densely populated urban slums

Abstract

Mumbai city, India houses ~10 million people in slums, which is half of its total population. Located along the major roadways, most of these smaller sized and poorly constructed slum homes are likely exposed to high levels of indoor air pollutants. Since direct measurement of indoor air pollution levels among the large slum population is infeasible, this study aims to develop predictive regression models for indoor PM2.5 and black carbon (BC) in the densely populated slums in Mumbai. Daily indoor PM2.5 and BC was measured inside the homes of a low-traffic slum (in 44 homes) and a high traffic slum (in 40 homes) during a winter season. Multivariable regression models were developed using measured indoor levels, publicly available ambient PM2.5 and information on local traffic characteristics, building characteristics and occupant activities (collected through questionnaire surveys). Models showed moderate to good performance for BC (adjusted R2 = 0.50–0.64) and PM2.5 (adjusted R2 = 0.53–0.57). Ambient PM2.5 was the most significant predictor for both pollutants, accounting for the temporal variation. BC was positively associated with the road length density of major roads while PM2.5 was associated with the density of all road types. Models for both pollutants were robust with Leave-One-Out-Cross-Validation (LOOCV) R2 ranging 0.43–0.61 (BC) and 0.37–0.51 (PM2.5). The study demonstrates that indoor particulate matter exposures can be reasonably predicted using publicly available air pollution data and information on local traffic and housing characteristics, and also underpins the high exposure to traffic pollution in urban slums.