About CanOSSEM

What is CanOSSEM?

The Canadian Optimized Statistical Smoke Exposure Model (CanOSSEM) is random forest machine learning model that estimates 24-hour mean total (all-source) fine particulate matter (PM2.5) concentrations at a 5 km x 5 km spatial resolution across all populated regions of Canada (Figure 1). CanOSSEM also provides an indicator for wildfire smoke-impacted days. CanOSSEM is optimized for estimating PM2.5 related to biomass smoke by integrating regulatory air quality monitoring network observations with remotely sensed data on wildfire activity, atmospheric aerosols, smoke plumes, meteorological conditions, and elevation (Figure 2). Estimates of total PM2.5 concentrations and wildfire smoke days from CanOSSEM (version 4) are currently available for 2003-2024. Details on the original implementation of CanOSSEM can be found in Paul et al., 2022.

CanOSSEM was initially developed under the first round of Health Canada’s Addressing Air Pollution Horizontal Initiative (AAPHI) funding. CanOSSEM (version 1) estimated daily total PM2.5 across populated regions of Canada for 2010-2019.  Several improvements have been made to CanOSSEM since the initial funding investment, including the expansion of the spatial grid to cover more populated raster cells, extension of the time series to include additional years, and inclusion of additional predictor variables. With ongoing funding from Health Canada, work is underway to further improve CanOSSEM.

Figure 1. Example of CanOSSEM daily mean total PM2.5 estimates on June 25, 2023. Estimates are available at a 5 km x 5 km spatial resolution for all populated regions of Canada.
Figure 2. The workflow used to generate the total PM2.5 estimates in the current version of CanOSSEM (version 4), which provides estimates for 2003-2024.