Mapping Air Pollution with Google Street View Cars: Towards Efficient Mobile Monitoring

Abstract

Air pollution measurements collected through systematic mobile monitoring campaigns can provide exposure data at high spatial resolution. Approaches to minimize the amount of data and driving effort required to successfully map an urban area’s air quality may help improve the scalability of this measurement approach. We explore the data requirements for mapping a city’s air quality using mobile monitors utilizing “data only” and predictive model approaches. We used two Google Street View cars with fast-response (1 Hz) instruments to collect nitric oxide (NO) and black carbon (BC) measurements in Oakland, CA. This rich dataset incorporates ~ 3.6 million measurements collected repeatedly (20-50 ×) on every city street over 2 y. We explore two alternative strategies to efficiently mapping spatial air quality patterns. First, we use a “data-only” approach where we attempt to minimize the number of repeated visits to each road. Second, we combine our data with a land use regression-kriging (LURK) model to predict at unobserved locations, and consider sampling schemes where only a subset of a city’s roads or repeat visits are measured. In all cases, we use a Monte Carlo scheme to systematically subsample the full dataset. LURK models consistently captured general spatial trends in urban air pollution, with cross-validation R2 for log-transformed-NO and BC of 0.65 and 0.5. While LURK models did not capture the full variability of on-road concentrations, models could successfully be trained with minimal data requirements, e.g., with only a small number of repeated observations on ~20% of the roads. Data-only mapping performed poorly for < 4 repeated drives per road segment, but obtained surpassed the performance of LURK approach within 4 to 8 repeated drive days per road segment. Data-only mapping can have surprisingly modest data requirements. LURK models can efficiently produce valid predictions of air quality when it is not possible to sample air quality on every road.