Building a low-cost air quality sensor network
With only one regulatory air quality station for more than 900,000 Denton County residents, how do we really know what's in our air?
Ronney Phillips, Dr. Lu Liang, Dr. Alexandra Ponette-González ----------------------------Department of Geography and the Environment -----------------------------------------------------University of North Texas
Introduction
Air pollution is one of the greatest environmental risks to health (World Health Organization). However, accurately measuring air quality at the neighborhood scale is constrained by the limited number of regulatory stations. In Denton County, a suburb of the Dallas-Fort Worth metroplex and one of the fastest-growing U.S counties, only one regulatory air quality station monitors air quality for more than 900,000 residents. Low-cost sensors offer a feasible solution for continuous and realistic measuring of hyperlocal air quality.
- Create an optimal sensor placement plan to obtain a suitable spatial representation of air quality
- Design a solar panel-powered sensor system for easy installation at important locations in the sampling framework which lack power and WiFi access.
The objectives of this study are:
DATA
The sampling method is based on two indicators that have evident effects on air quality: the percentage of impervious surface and proximity to major roads (Hart et al. 2020). Major roads are strong indicators of urban development. We also used the 2019 impervious surface data from the Multi-Resolution Land Characteristics Consortium (Xian et al. 2011) to determine where development was occurring in Denton County.
METHODS
This study used a stratified sampling design to determine optimal sensor placements in Denton County to successfully quantify intra-urban pollution variability. First, we created 300x300-m grids as our sampling unit and symbolized according to their strata type. By overlaying the distance to major roads and the percentage of impervious surface, we divided the county into six strata (Table 1).
Ideally, an approximately equal number of sites should be selected from each stratum. We counted the distribution of current existing sensors among the six strata to guide us in future placements.
RESULTS
63.24% of Denton County's land area was classified as Rural Low -- that is, 63.24% of the land has less than 20% impervious surface and is located more than 150 meters from a major road. Other land type percentages can be seen below.
Presently, we have 41 sensors placed in Denton County. After examining the distribution of sensors within their respective land classifications, we discovered sensor placement in the Rural Low, Suburban Low and Urban categories were overrepresented. Only seven have been placed within the other three areas. This indicates more air quality monitoring is needed in the Rural High, Suburban High and Urban High areas.
We also conducted random sampling within the six land type classifications to determine the areas' appearances regarding impervious surface and overall development and captured photos of the areas to show their true representation on the ground.
CONCLUSION
New sensors should be placed in areas classified as Rural High, Suburban High and Urban High. Placing additional sensors will allow us to obtain a more representative sample within the county.
Currently, the team is combing through 74 potential sites to determine placement suitability. We anticipate finalization and deployment of new sensors by the end of December 2021.
For more information about the project or hosting a sensor, please click the button below.
REFERENCES
Hart, R., Liang, L. Dong, P., 2020. Monitoring, mapping, and modeling spatial–temporal patterns of PM2.5 for improved understanding of air pollution dynamics using portable sensing technologies. Int. J. Environ. Res. public health, 17(14), 4914.
World Health Organization. (2021). Air pollution. Retrieved from https://www.who.int/health-topics/air-pollution
Xian, G., Homer, C., Dewitz, J., et al. 2011. The change of impervious surface area between 2001 and 2006 in the conterminous United States. Photogramm Eng Remote Sensing, 77(8): 758-762.