Mortality Estimation Tool

About the Data

Wildfire Smoke Concentrations

The map shows daily 24-hour-average concentrations of one component of wildfire smoke, small particulate matter called PM_2.5.

Daily wildfire-smoke PM_2.5 concentrations were calculated by combining measurements of total PM_2.5 from air quality monitors on the ground with satellite observations of smoke plumes, following existing research methods.

First, provisional 24-hour-average all-source PM_2.5 concentrations for all air quality monitors in North America were downloaded for each day between January 1, 2014 to August 31, 2023 from the Environmental Protection Agency AirNow Tech website. Daily smoke-plume maps were also downloaded from the National Ocean and Atmospheric Administration Hazard Mapping System.

Then, for each monitor, the smoke plume maps were used to designate days as smoke days (smoke plume over the monitor) and non-smoke days (no smoke plume over the monitor). All-source PM_2.5 data was included from 1,019 monitors in U.S., Canada, and Mexico that had at least 90 non-smoke days with PM_2.5 data.

Next, baseline seasonal PM_2.5 was defined as the 99^th percentile of 24-hour-average PM_2.5 from non-smoke days for each season (December to February, March to May, June to August, September to November). The 99^th percentile was selected to reduce likelihood of overestimating wildfire smoke.

Finally, wildfire smoke PM_2.5 concentrations were calculated for each day for each monitor by subtracting the baseline seasonal non-WFS PM_2.5 from the observed 24-hour-average PM_2.5. Concentrations were interpolated between monitors, and the county-average value was extracted.

Mortality Estimates

The map shows estimated same-day mortality attributable to the wildfire smoke concentrations estimated above.

Daily wildfire-smoke-attributable mortality was calculated using a health-impact function, which is a common policy method for attributing deaths to pollution². The method combines, for each county, the population, the baseline mortality rate without the smoke, and a measure of how deadly smoke is on average. This map uses an estimate aggregated from numerous other studies³. These estimates are made without knowledge of the actual deaths occurring in that county on that day because that data is typically not available for months or years.

For more detailed methods including the actual calculations, see our publications (links coming soon).

Uncertainty

The data in this tool should all be considered provisional estimates, with inherent uncertainty. A recent study evaluated uncertainty in statewide mortality from different methods, and found greater uncertainty between health-impact functions than from smoke-estimation methods. Across ten estimates using five smoke-estimation methods and two health-impact functions, they found a lowest value of 4 deaths and a highest value of 30 deaths statewide during a wildfire smoke wave. Though their study didn’t evaluate our methods specifically, we anticipate similar uncertainty in our work and hypothesize that is an acceptable level for emergency managers to take action.

Expected data gaps

Because existing methods average over 24 hours and counties, we expect to miss the impacts of smoke events that are smaller in space or time. For example, a daytime prescribed burn is unlikely to appear in the data because it is both small in scale and also lasts a short period.