Smoke Exposure (PM2.5 v2) Monthly (Differences...
This is version 2 of the Canadian Optimized Statistical Smoke Model (CanOSSEM), developed by the Environmental Health Services of the BC Centre for Disease Control. It is used to produce estimated concentrations of fine particulate matter across all populated regions of Canada. CanOSSEM is a random forest machine learning model that uses potential predictor variables integrated from multiple data sources and estimates annual, seasonal, monthly, and daily PM2.5 concentrations at a 5 km × 5 km spatial resolution. The estimates are optimized for wildfire smoke through use of multiple variables that are specific to this source. Un-indexed grid files are available on request to naman.paul@bccdc.ca. CanOSSEM requires re-training annually after a major wildfire event or new data becomes available. CanOSSEM was re-trained specifically to address two points: 1) Spatial grid expansion to include remote locations that were missed out in version 1, and 2) Time series extension to cover the years 2020, 2021 and 2022. For CanOSSEM version 1, the Root Mean Squared Error (RMSE) between predicted and observed PM2.5 concentration was 2.85 µg/m3 for the entire prediction set, with over 95% of the predictions lying within an absolute difference of 5 µg/m3 from the NAPS PM2.5 measurements. CanOSSEM version 2 saw an improvement of approximately 18% in the RMSE.
There are no views created for this dataset yet.
Additional Information
| Field | Value |
|---|---|
| Data last updated | September 18, 2023 |
| Metadata last updated | April 12, 2024 |
| Created | September 18, 2023 |
| Format | |
| License | License not specified |
| Datastore active | False |
| Datastore contains all records of source file | False |
| Has views | False |
| Id | 0373fb90-6885-4bf1-908e-3168336277dc |
| Package id | da8a305d-ae3d-4cfb-a8df-4e6364d81fd4 |
| Position | 0 |
| State | active |