Public health applications of historical smoke forecasts: An evaluation of archived BlueSky data for the coterminous United States, 2015-2018.

BACKGROUND

Wildfires are increasing in magnitude, frequency, and severity. Populations in the wildland-urban interface and in downwind communities are at increased risk of exposure to elevated concentrations of fine particulate matter (PM) and other harmful components of wildfire smoke. We conducted this analysis to evaluate the use of modeled predictions of wildfire smoke to create county-level measures of smoke exposure for public health research and surveillance.

METHODS

We evaluated four years (2015-2018) of grid-based North American Mesoscale (NAM)-derived PM forecasts from the U.S. Forest Service BlueSky modeling framework with monitoring data from the Environmental Protection Agency Air Quality System (AQS), the Interagency Monitoring of Protected Visual Environments (IMPROVE), the Western Regional Climate Center (WRCC), and the Interagency Real Time Smoke Monitoring (AIRSIS) programs. To assess relationships between model-derived estimates and monitor-based observations, we assessed Spearman's correlations by spatial (i.e., county, level of urbanization, states in the western United States impacted by major wildfires, and climate regions) and temporal (i.e., month and wildfire activity periods) characteristics. We then generated county-level smoke estimates and examined spatial and temporal patterns in total and person-days of smoke exposure.

RESULTS

Across all counties in the coterminous United States and for all days, the correlation between county-level model- and monitor-derived PM estimates was 0.14 (p < 0.001). Correlations were stronger using data from temporary monitors and for areas and days impacted by high wildfire smoke, especially in the western United States. Correlations between county-level model- and monitor-derived estimates in non-metropolitan counties, and at higher concentrations ranged from 0.25 to 0.54 (p < 0.001).

CONCLUSIONS

In general, public health practitioners and health researchers need to consider the pros and cons associated with modeled data products for conducting health analyses. Our results support the use of model-derived smoke estimates to identify communities impacted by heavy smoke events, especially during emergency response and for communities located near wildfire episodes.