SOCAAR is pleased to announce the next seminar for Dec 2018:

Wednesday, December 5, 2018

3:00 – 4:00 PM
200 College Street, WB 407

Particulate Matter Air Pollution in Charlotte NC: A Land Use Regression Model based on Citizen Science Monitoring
Matthew Adams, PhD

Assistant Professor

Department of Geography

University of Toronto Mississauga

Abstract: This study assessed the use of air pollution data collected through Citizen Science activities for developing a land use regression (LUR) air pollution model. The model was applied to estimate an air pollution surface for Charlotte, NC. Particulate matter air pollution was monitored for this study with air pollution sensors mounted to bicycles. The lower cost sensors are known to demonstrate bias. The observed values were adjusted with a neural network model derived from a collocation study of the low-cost sensor with a research grade instrument. For each air pollution observation location, land use information was calculated within buffers of varying sizes. A linear regression model was developed, to explain the variation in air pollution observations by the surrounding land use conditions, using a manual step-wise approach. The performance of the multivariate linear regression model was evaluated by cross-validation with data excluded during model fitting. The linear model performance was poor with an R^2 of 0.24. An artificial neural network model was developed in an attempt to improve model performance and obtain a higher predictive performance during cross-validation. The neural network based LUR model achieved a prediction R^2 of 0.71, a significant improvement from the linear regression model. The application of neural networks in a land use regression framework has proven useful as it allows for complex non-linear relationships that may be present in dataset. 


If you are unable to join the meeting in person, please join this meeting from your computer, tablet or smartphone.

Please join my meeting from your computer, tablet or smartphone.

https://global.gotomeeting.com/join/725709717

You can also dial in using your phone.

Canada: +1 (647) 497-9391
Access Code: 725-709-717

For supported devices, tap a one-touch number below to join instantly.

One-touch: +16474979391, 725709717

Access Code: 725-709-717

First GoToMeeting? Let's do a quick system check: https://link.gotomeeting.com/system-check