Smart modeling can help scale air pollution mapping

Report published October 2018

EDF and our partners have shown that on-road pollution data collection can produce highly detailed maps of air quality in urban areas. But that comes at a high cost. In order to scale this technique and bring it to additional locations at a lower price, EDF is exploring a combination of data collection and statistical modeling.

In a recent paper published in the journal Environmental Science and Technology, researchers explain it is possible to create reliable maps that show changes in air quality with limited amounts of mobile monitoring data—when that information is fed into a statistical model.

This study shows us that we can produce reliable maps with less data than previously believed.

Steven Hamburg, Chief Scientist

The study, which uses data collected for an earlier air quality mapping project in Oakland, CA, was led by scientists from the Department of Civil, Architectural & Environmental Engineering at the University of Texas at Austin. The research team—which also included EDF scientists and experts from the United States, Canada, and the Netherlands—combined the data collected for the earlier study and incorporated it into a statistical model to determine just how much collected data was required to create an accurate picture of air quality in Oakland.

More collected data is better, not critical

Driving about 30 percent of the roads in an area a minimum of four randomly sampled times was enough to predict key patterns of how air pollution varied across the cities. Highly accurate estimates of an area's air pollution patterns could be made without a model if researchers sampled every road on ten or more occasions.

And while creating maps from repeated on-road sampling can be a simpler and more accurate approach, combining data with statistical modeling could allow researchers to map new cities with fewer resources, in less time and at a lower cost.