Wildfire Risk in Canada

1 · The fire regime

Lightning ignitions account for most of Canada's burned area

Canada records about 5,717 wildfires a year. About 500 of those fires reach at least 100 hectares, yet they account for more than 99 percent of the total area burned. Lightning causes 48 percent of ignitions and accounts for 87 percent of burned area. These fires often start far from roads and towns, where they last longer. Figure 1 shows national burned area from 2016 to 2025. The median annual burned area was 2.4 million hectares, and the mean was 4.0 million hectares. In 2023, 14.8 million hectares burned, more than in any other year on record.

Burned area also varies substantially across the country. The Prairie provinces and northern territories account for about 60 percent of national burned area, British Columbia and Yukon for 22 percent, Ontario and Quebec for 18 percent, and Atlantic Canada for about 0.5 percent. Table 1 presents the data by province and territory.

2 · How SWIM Models an Individual Fire

Testing the 90th-Percentile Fire-Size Estimate

SWIM simulates fire spread using the Canadian Forest Fire Behavior Prediction System, the 2026 national fuel type map, 20 m hexagons, and the day's weather and fire danger codes. The model was developed using 257,192 Canadian fire records from 1972 to 2025. These include fires reported by provincial, territorial, and national fire agencies and burned areas mapped by satellite in the National Burned Area Composite. Because final fire size is uncertain, SWIM estimates a range of possible outcomes rather than a single value. For each fire, the model reports the median and 75th-, 90th-, and 95th-percentile fire-size estimates and calculates the associated loss.

A well-calibrated percentile should contain approximately the proportion of observed outcomes that it represents. The model was evaluated on 9,922 recorded fires of at least 30 hectares using five-fold geographic cross-validation. Canada was divided into 100 km blocks, and each fire was tested using a model that had not been trained on fires from the same geographic block. Overall, 90.0 percent of observed fires finished at or below the model's 90th-percentile fire-size estimate. Table 2 presents the results by province.

3 · Historical Fire Reconstructions

Six Historical Fires Reconstructed from Their First Satellite Detection

Each historical fire was simulated from the place and day of its first satellite detection using the weather observed that day. The 90th-percentile fire-size estimate was used, or the 95th percentile when the observed fire exceeded that range. The six examples were selected to represent different regions of Canada. In each map, the blue area shows locations burned in at least half of the simulations, the white line shows the observed perimeter, and the dot marks the first detection over satellite imagery (Esri, Maxar, Earthstar Geographics). Table 3 presents the areas.

1. aStarted by lightning; first seen July 5, 2013. SWIM's median size 778 ha, growing over about 10 days at the 90th percentile.
2. bStarted by lightning; first seen July 26, 2012. SWIM's median size 834 ha, growing over about 42 days at the 90th percentile.
3. cStarted by lightning; first seen June 16, 2019. SWIM's median size 188 ha, growing over about 8 days at the 90th percentile.
4. dStarted by lightning; first seen July 31, 2018. SWIM's median size 125 ha, growing over about 9 days at the 90th percentile.
5. eStarted by people; first seen May 12, 2019. SWIM's median size 198 ha, growing over about 3 days at the 90th percentile.
6. fStarted by people; first seen May 8, 2014. SWIM's median size 96 ha, growing within about a day at the 90th percentile.

4 · The risk map

Canada's Modeled Expected Annual Wildfire Loss Is CA$7.82 Billion

The Canada wildfire risk map evaluates about 11.4 million locations nationwide. For each location, SWIM estimates the likelihood of an ignition, the potential fire size under local fire-weather conditions, and the consequences of the fire. Together, these estimates produce CA$7.82 billion in modeled expected annual loss: CA$5.94 billion for land, CA$1.77 billion for fire suppression, and CA$105 million for structures and people. Figure 2 presents modeled expected annual loss, and Figure 3 presents modeled expected annual burned area.

Expected loss in a hexagonBelow the medianThe median or more3.4 times the median or more11 times the median or moreThe top 3%

Hectares expected to burn in a hexagonBelow the medianThe median or more3.8 times the median or more14 times the median or moreThe top 3%

Table 4 compares provinces and territories by expected annual loss, expected annual burned area, and dwellings near burnable land. Northern and Prairie jurisdictions rank highest in burned area, while expected loss also reflects population and timber value. British Columbia has the highest modeled expected annual loss, while Ontario and Quebec have the largest numbers of dwellings near burnable land.

Higher-severity fire years produce substantially greater burned area and modeled loss. Table 5 presents the national 50th-, 75th-, 90th-, and 95th-percentile annual scenarios based on the historical distribution of annual burned area. The map presents each scenario.

The map's burn probability was evaluated in geographic areas that had not been used to fit the model. It assigned a higher burn probability to a burned location than to an unburned location 93 times out of 100.

5 · The fire season

Most Annual Burned Area Comes from Fires That Start Between May and July

Fires that start in May, June, and July account for about 91 percent of Canada's annual burned area (Figure 4). The season typically begins in western Canada and the Prairie provinces during spring, when the previous year's dead grass remains dry before green-up. Large boreal fires in northern Canada, Ontario, and Quebec become more prominent in early summer.

6 · Sources

The map is built on Canadian public data

Table 6 lists the principal public datasets used to construct the Canada model and this report.