By Jillian Kunze
There are a few methodologies for creating maps of pollutants on the scale of neighborhoods. Various modeling strategies are available, which depend on empirical measurements for training and evaluation. It requires a lot of effort to obtain measurements at a high spatial detail across a city, and such an undertaking is often accomplished through a network of stationary monitors.
A popular approach for mapping pollution is to measure particulate matter that is less than 2.5 microns in width (PM2.5), which does not require very high resolution to characterize in space. However, PM2.5 is not a good target if researchers want to investigate the highly localized pollution patterns that traffic might cause through primary combustion from vehicles — this can create variation on the 100-meter scale.
At the 2022 SIAM Conference on Mathematics of Planet Earth, which took place last week concurrently with the 2022 SIAM Annual Meeting, Sarah Chambliss of the University of Texas at Austin gave a minisymposium presentation about an effort to use mobile monitors to measure several different pollutants across neighborhoods in the San Francisco Bay Area at a very fine scale. “In mobile monitoring, you don’t need to establish a network on the ground of instruments,” Chambliss said. “What you do is you put those instruments in a vehicle, and then you drive those vehicles around and take measurements.”
Figure 1. Box and whisker plots showing the range of exposure to different pollutants (nitric oxide, nitrogen dioxide, black carbon, and ultrafine particles) in various neighborhoods in the San Francisco Bay Area. The white dots represent the mean concentrations. Figure courtesy of Sarah Chambliss.
Through sponsorships from Google and the Environmental Defense Fund, Chambliss and her collaborators were able to put pollution monitors in the form of the Aclima mobile platform on Google Street View vehicles that drove around the city. These cars revisited the same areas multiple times, so the data was sufficient to estimate temporal variation and find a decent uncertainty on the measurements. Chambliss found that if the cars visited the same location 15 times over a year, this was able to give a decently bounded estimate of the median concentration of pollutants in the area.
Figure 2. The exposure to pollutants among different racial-ethnic groups in the San Francisco Bay Area. Figure courtesy of Sarah Chambliss.
The monitors were looking for several pollutants: nitric oxide (NO), nitrogen dioxide (NO
2), black carbon, and ultrafine particles, which are particulate matter that can pass deeper into the lungs. Chambliss was interested in the levels of pollution that people experience right outside their homes, which she called their exposure. She wanted to sample a variety of neighborhoods with different characteristics that would have different sources of pollution.
The resulting data exhibited the relevance of looking at fine-scale details on pollution maps. Figure 1 provides box and whisker plots with the range of concentrations for the four pollutants within 13 neighborhoods. There are noticeable differences in the mean concentrations between neighborhoods, as well as a range of concentrations detected within each neighborhood. These differences could have relevance for the health outcomes in a population, such as for asthma.
“So, one question that we investigated from this is if you know that within a neighborhood, two people down the block from each other might have different exposures, does this have implications for people of color being exposed to higher levels of pollution than people identifying as white?” Chambliss asked. Figure 2 illustrates that there is indeed a small difference in the mean exposure to pollutants between different racial-ethnic groups in these neighborhoods, though the differences might be too small to consider health-relevant except in the case of ultrafine particles. Black and Hispanic/Latino residents experience the highest average levels of exposure, while the Asian population has the greatest range of exposure. The lower tails of the distributions also stand out, as they show that white residents tend to have the privilege of living in the cleanest areas.
Another question is which spatial scale drives the disparity in exposure — do people of color within a neighborhood tend to live closer to highways or industrial areas, for example? While the pollution might vary a lot within one neighborhood, it turns out that the racial and ethnic makeup of individual neighborhoods tends to be fairly homogeneous. Since the Bay Area is somewhat segregated in this way, it is more important to compare between neighborhoods when considering exposure disparities. Figure 3 illustrates the pollution exposure for different racial-ethnic groups according to neighborhood of residence. Just one noticeable inequitable outcome is that since some pollutants have much higher concentrations in Oakland, and many residents of Oakland are Black, there is systematically more exposure for the Black population.
Figure 3. Pollution exposure for different racial-ethnic groups in the San Francisco Bay Area, color-coded by neighborhood. Figure courtesy of Sarah Chambliss.
These results have implications for how to choose a model to answer questions related to pollution and environmental injustice. There is important information to find at a scale of 100 meters, but if one is more interested in drawing general patterns across a city, then looking at the neighborhood-scale provides a decent picture. Land-use regression does not do well at capturing the full range of concentrations, so Chambliss and her collaborators are currently investigating other ways to predict extreme concentrations in different areas.
Figure 4. Maps of the spatial patterns of nitric oxide, nitrogen dioxide, black carbon, and ultrafine particles in the San Francisco Bay Area. Figure courtesy of Sarah Chambliss.
Chambliss also noted that there was only moderate correlation between the concentrations of pollutants in different census blocks, indicating that different sources drive concentrations between neighborhoods. All four of the pollutants in this study are generally considered traffic-related, but using just one of the four pollutants to extrapolate about traffic emissions could lead to missed details.
Figure 4 provides a map of the spatial patterns for the four pollutants, exhibiting interesting characteristics. The patterns of NO are fairly easy to index to the location of highways or traffic. NO2 has a similar distribution to NO—as the two pollutants are highly correlated—but NO2 is more diffuse since it forms secondarily in the air further away from sources. Some pollution hotspots also appear in locations that one might not expect based on the location of roadways alone. In the Richmond District, for example, an area with many restaurants is a hotspot for NO. The financial district also has spots of high pollution that are likely related to construction, and high concentrations of black carbon and ultrafine particles can be caused by industrial sources such as a cast metals recycling plant. Overall, pollutants that are directly emitted by sources tend to reach low levels in areas that are far from sources, while pollutants that form secondarily in the atmosphere (including ultrafine particles) have higher concentrations across wider areas.
“A conclusion from all of this is that local and regional solutions are needed to address exposure inequity,” Chambliss said. It is important to both look at pollution that may be blowing into a neighborhood from outside sources as well as causes of localized pollution — especially in American cities, which tend to be very segregated. “Also, there’s a lot of value to creating purely empirical maps,” Chambliss added. “It shows patterns that you would not expect just from information about where highways are located within an area or even more detailed land use — but it is a very high-effort data collection method.”
 |
Jillian Kunze is the associate editor of SIAM News. |