Authors: Oluwatobi Oke*, Department of Civil and Environmental Engineering, Colorado State University, Ellison Carter, Department of Civil and Environmental Engineering, Colorado State University, Sheryl Magzamen, Department of Environmental and Radiological Health Sciences, Colorado State University, Shantanu Jathar, Department of Mechanical Engineering, Colorado State University, Ander Wilson, Department of Statistics, Colorado State University, Bruce Draper, Department of Computer Science, Colorado State University, Jeremy Auerbach, Department of Environmental and Radiological Health Sciences, Colorado State University, Charles He, Department of Mechanical Engineering, Colorado State University
Topics: Human-Environment Geography, Environment, Land Use
Keywords: Built Environment, Chemical Exposures, Physical activity, Deep Learning
Session Type: Paper
Start / End Time: 9:55 AM / 11:35 AM
Room: Wilson A, Marriott, Mezzanine Level
Presentation File: No File Uploaded
There is an increasing recognition that health is determined by numerous factors including the “Built Environment (BE)”- where you live, play, work, poor access to healthy food, abundance of fast food chains, lack of recreational facilities etc. The main challenge with exposure assessment is exposure misclassification. This is also true with assessing “exposure” to features of the built environment, which are compounded when many different features are used to generate a single indexed exposure metric. Exposure misclassification within the field of the built environment is often tied to geographical contexts and differences in the audit tools used. These inconsistencies could be due to variations in measures or measurement tools across study settings, reducing outcome reliability, validity, and comparability. Here, we used deep learning technique to extract data representing features of the built environment from high-resolution satellite images for the cities of Oakland, Pittsburgh, and Milwaukee. The output feature maps were converted to vectors for linear regression models to quantify the association between features of built environment and chemical exposures, including air pollution exposure. We hypothesize that the prediction of these exposures from the extracted features of the BE is considered to explain how the BE influences the prevalence of different chemical pollutants in all census tracts. A good understanding of the association between the specific features of the BE and chemical pollutants can lead to how to effect structural changes in the design of the BE in order to reduce the amounts of pollutants causing adverse effects on human health.