https://www.arcgis.com/apps/dashboards/10fc360faf7e421183a566605a530cb4This project combined air-quality monitoring data with census-based urbanization indicators to test whether more urbanized areas of Toronto tend to experience higher nitrogen oxide (NO) concentrations over time (2011–2016).
I sourced NO measurements (ppb) from four monitoring stations (2011, 2014, 2016) and matched them with dissemination-area (DA) urbanization metrics from Toronto’s neighbourhood/census profiles (2006, 2011, 2016), including population/dwelling density and transportation mode shares. Because station data is sparse, I used spatial interpolation (kriging) to estimate NO levels across the city and assign values to each DA, then built a single, analysis-ready spatial dataset by joining and normalizing variables (e.g., converting transport counts into percent shares and combining driver + passenger into one indicator).
From there, I:
Checked relationships among predictors with a correlation matrix to reduce multicollinearity (e.g., density variables).
Mapped patterns and change through time (dominant year, NO change, mobility contrasts).
Measured spatial co-variation using bivariate spatial association (Lee’s L) to identify local clusters and significance.
Modeled local relationships using GWR (and tested a global regression, which had low explanatory power).
The key takeaway: air pollution patterns didn’t align cleanly with urbanization alone, suggesting other drivers (e.g., local emission sources, road networks, industry, meteorology) likely play a major role.
