Access the Data

CanOSSEM (version 4): Estimates of total (all-source) PM2.5 and wildfire smoke days for 2003-2024

CanOSSEM (version 4) includes estimates of daily mean total (all-source) PM2.5 and wildfire smoke days for populated regions of Canada for 2003-2024. The data are available as .Rdata files at the following geographic resolutions:

  • CanOSSEM raster cell (5 km x 5 km)
  • Census dissemination block (DB)
  • Census dissemination area (DA)
  • Postal code (PC)
  • Forward sortation area (FSA)

All datasets also include estimates of daily mean, minimum, and maximum temperature and daily mean relative humidity, obtained from MERRA-2. The raster cell dataset additionally includes variables for observed PM2.5, daily smoke plume presence, and fire radiative power in the wildland and wildland-urban interface.

Crosswalk files are provided to facilitate mapping of raster cells to the other geographic resolutions. READMEs are available for each dataset and crosswalk and provide detailed information on the included variables, dataset generation methods, and data sources.

Requesting Access

The CanOSSEM dataset is publicly available. To access the version 4 data and help us track its use, please complete the request form at the link below. Once the form is completed, you will automatically receive a link to a Dropbox folder where you can download the .Rdata files. The data can also be provided as .csv files upon request.

Request Data

Citing CanOSSEM

If you use the CanOSSEM estimates in your research or other public-facing work, we request that you cite CanOSSEM in all published or public-facing materials as follows:

Paul, N.; Yao, J.; McLean, K. E.; Stieb, D. M.; Henderson, S. B. The Canadian Optimized Statistical Smoke Exposure Model (CanOSSEM): A Machine Learning Approach to Estimate National Daily Fine Particulate Matter (PM2.5) Exposure. Sci. Total Environ. 2022, 850, 157956. https://doi.org/10.1016/j.scitotenv.2022.157956

CanOSSEM Performance & Code

CanOSSEM (version 4) has strong predictive performance. For the 2003-2024 total PM2.5 prediction set estimates, the root mean square error was 2.60 μg/m3, the R2 was 0.96, and 98.2% of estimates were within 5 μg/m3 of the observed values.

The code used to generate the total PM2.5 and wildfire smoke day estimates for CanOSSEM (version 3) can be found at: https://github.com/BCCDC-PHSA/CanOSSEM. The code for CanOSSEM (version 4) will be made available in the coming months.

CanOSSEM (version 3)

CanOSSEM (version 3) provided estimates of daily mean PM2.5 and wildfire smoke days for 2010–2023. We recommend using CanOSSEM (version 4), which includes extended spatial and temporal coverage (2003–2024) and additional predictor variables. However, if you require access to version 3 for replication, continuity with previous analyses, or other purposes, please complete the request form below. Once completed, you will automatically receive a link to a Dropbox folder containing the .Rdata files.

Request CanOSSEM (version 3) Data