Professor Julian Marshall

CSENG Civil, Envrn & Geo- Eng
College of Science & Engineering
Twin Cities
Project Title: 
Geographic Attribute Regression Pollution Prediction

Outdoor air pollution kills approximately three million people per year. Better estimates of how air pollutant concentrations vary in space and time will allow the predicting the number of deaths with greater precision and accuracy. Many existing concentration estimates that are used to predict air pollution health effects are based on the linear regression of land use attributes (e.g., roads, impervious surfaces) against measured pollution concentrations to predict concentrations in locations where there are no measurements. Based on the hypothesis that geographical attributes (e.g., restaurants, power plants, factories) are a better predictor of air pollution than land use types are, these researchers propose to create new concentration estimates based on a regression of geographical attributes from the OpenStreetMap database against measured concentrations. They propose to use advanced machine-learning techniques and Google’s TensorFlow machine learning library to handle the large number of feature categories that are available in the OpenStreetMap data.

Research by this group was featured on the MSI website in July 2017: A New Computer Model for Air Pollution, February 2015: Effects of Alternative Fuels on Air Quality, and September 2014: Air Pollution and Socioeconomic Status.

Project Investigators

Richard Barnes
Matthew Bechle
Adam Both
Cassidy Buckley
Sean DeBruzzi
Andrew Goodkind
Steve Hankey
Kathryn Lundquist
Professor Julian Marshall
Dev Millstein
Nam Nguyen
Eric Svingen
Sumil Thakrar
Maninder Pal Si Thind
Kristina Wagstrom
Are you a member of this group? Log in to see more information.