Wildfires have been increasing in frequency in the western United States (US) with the 2017 and 2018 fire seasons experiencing some of the worst wildfires in terms of suppression costs and air ...pollution that the western US has seen. Although growing evidence suggests respiratory exacerbations from elevated fine particulate matter (PM2.5) during wildfires, significantly less is known about the impacts on human health of ozone (O3) that may also be increased due to wildfires. Using machine learning, we created daily surface concentration maps for PM2.5 and O3 during an intense wildfire in California in 2008. We then linked these daily exposures to counts of respiratory hospitalizations and emergency department visits at the ZIP code level. We calculated relative risks of respiratory health outcomes using Poisson generalized estimating equations models for each exposure in separate and mutually-adjusted models, additionally adjusted for pertinent covariates. During the active fire periods, PM2.5 was significantly associated with exacerbations of asthma and chronic obstructive pulmonary disease (COPD) and these effects remained after controlling for O3. Effect estimates of O3 during the fire period were non-significant for respiratory hospitalizations but were significant for ED visits for asthma (RR = 1.05 and 95% CI = (1.022, 1.078) for a 10 ppb increase in O3). In mutually-adjusted models, the significant findings for PM2.5 remained whereas the associations with O3 were confounded. Adjusted for O3, the RR for asthma ED visits associated with a 10 μg/m3 increase in PM2.5 was 1.112 and 95% CI = (1.087, 1.138). The significant findings for PM2.5 but not for O3 in mutually-adjusted models is likely due to the fact that PM2.5 levels during these fires exceeded the 24-hour National Ambient Air Quality Standard (NAAQS) of 35 μg/m3 for 4976 ZIP-code days and reached levels up to 6.073 times the NAAQS, whereas our estimated O3 levels during the fire period only occasionally exceeded the NAAQS of 70 ppb with low exceedance levels. Future studies should continue to investigate the combined role of O3 and PM2.5 during wildfires to get a more comprehensive assessment of the cumulative burden on health from wildfire smoke.
•Ozone was associated with respiratory ED visits during a wildfire, but associations disappeared when adjusted for PM2.5•PM2.5 was associated with respiratory ED visits and hospitalizations during a wildfire period even when adjusted for ozone.•Our findings may be context specific; other wildfire-health studies should examine other air pollutants besides PM2.5
Epidemiologists use prediction models to downscale (i.e., interpolate) air pollution exposure where monitoring data is insufficient. This study compares machine learning prediction models for ...ground-level ozone during wildfires, evaluating the predictive accuracy of ten algorithms on the daily 8-hour maximum average ozone during a 2008 wildfire event in northern California. Models were evaluated using a leave-one-location-out cross-validation (LOLO CV) procedure to account for the spatial and temporal dependence of the data and produce more realistic estimates of prediction error. LOLO CV avoids both the well-known overly optimistic bias of k-fold cross-validation on dependent data and the conservative bias of evaluating prediction error over a coarser spatial resolution via leave-k-locations-out CV. Gradient boosting was the most accurate of the ten machine learning algorithms with the lowest LOLO CV estimated root mean square error (0.228) and the highest LOLO CV Rˆ2 (0.677). Random forest was the second best performing algorithm with an LOLO CV Rˆ2 of 0.661. The LOLO CV estimates of predictive accuracy were less optimistic than 10-fold CV estimates for all ten models. The difference in estimated accuracy between the 10-fold CV and LOLO CV was greater for more flexible models like gradient boosting and random forest. The order of estimated model accuracy depended on the choice of evaluation metric, indicating that 10-fold CV and LOLO CV may select different models or sets of covariates as optimal, which calls into question the reliability of 10-fold CV for model (or variable) selection. These prediction models are designed for interpolating ozone exposure, and are not suited to inferring the effect of wildfires on ozone or extrapolating to predict ozone in other spatial or temporal domains. This is demonstrated by the inability of the best performing models to accurately predict ozone during 2007 southern California wildfires.
Display omitted
•Machine learning methods can model ozone reasonably well during a wildfire.•Leave-one-location-out CV more accurately estimates prediction error than 10-fold CV.•Gradient boosting and random forest predicted ozone more accurately than other models.
Flexible machine learning methods model ozone well during a wildfire. LOLO CV more accurately estimates prediction error than 10-fold CV.
Since the mid-1990s a new generation of Earth-observing satellites has been able to detect tropospheric air pollution at increasingly high spatial and temporal resolution. Most primary emitted ...species can be measured by one or more of the instruments. This review article addresses the question of how well we can relate the satellite measurements to quantification of primary emissions and what advances are needed to improve the usability of the measurements by U.S. air quality managers. Built on a comprehensive literature review and comprising input by both satellite experts and emission inventory specialists, the review identifies several targets that seem promising: large point sources of NOx and SO2, species that are difficult to measure by other means (NH3 and CH4, for example), area sources that cannot easily be quantified by traditional bottom-up methods (such as unconventional oil and gas extraction, shipping, biomass burning, and biogenic sources), and the temporal variation of emissions (seasonal, diurnal, episodic). Techniques that enhance the usefulness of current retrievals (data assimilation, oversampling, multi-species retrievals, improved vertical profiles, etc.) are discussed. Finally, we point out the value of having new geostationary satellites like GEO-CAPE and TEMPO over North America that could provide measurements at high spatial (few km) and temporal (hourly) resolution.
•Comprehensive review of studies of satellite data applied to emissions estimation.•Overview of retrievals for eight major tropospheric air pollutants.•Techniques to enhance the usefulness of satellite retrievals.•Identification of target source categories for satellite data application.•Recommendations on ways to improve the usability of satellite retrievals.
Estimating population exposure to particulate matter during wildfires can be difficult because of insufficient monitoring data to capture the spatiotemporal variability of smoke plumes. Chemical ...transport models (CTMs) and satellite retrievals provide spatiotemporal data that may be useful in predicting PM2.5 during wildfires. We estimated PM2.5 concentrations during the 2008 northern California wildfires using 10-fold cross-validation (CV) to select an optimal prediction model from a set of 11 statistical algorithms and 29 predictor variables. The variables included CTM output, three measures of satellite aerosol optical depth, distance to the nearest fires, meteorological data, and land use, traffic, spatial location, and temporal characteristics. The generalized boosting model (GBM) with 29 predictor variables had the lowest CV root mean squared error and a CV-R2 of 0.803. The most important predictor variable was the Geostationary Operational Environmental Satellite Aerosol/Smoke Product (GASP) Aerosol Optical Depth (AOD), followed by the CTM output and distance to the nearest fire cluster. Parsimonious models with various combinations of fewer variables also predicted PM2.5 well. Using machine learning algorithms to combine spatiotemporal data from satellites and CTMs can reliably predict PM2.5 concentrations during a major wildfire event.
Wildfires generate substantial emissions of nitrogen oxides (NO x ) and volatile organic compounds (VOCs). As such, wildfires contribute to elevated ozone (O3) in the atmosphere. However, there is a ...large amount of variability in the emissions of O3 precursors and the amount of O3 produced between fires. There is also significant interannual variability as seen in median O3, organic carbon and satellite derived carbon monoxide mixing ratios in the western U.S. To better understand O3 produced from wildfires, we developed a statistical model that estimates the maximum daily 8 h average (MDA8) O3 as a function of several meteorological and temporal variables for three urban areas in the western U.S.: Salt Lake City, UT; Boise, ID; and Reno, NV. The model is developed using data from June-September 2000–2012. For these three locations, the statistical model can explain 60, 52, and 27% of the variability in daily MDA8. The Statistical Model Residual (SMR) can give information on additional sources of O3 that are not explained by the usual meteorological pattern. Several possible O3 sources can explain high SMR values on any given day. We examine several cases with high SMR that are due to wildfire influence. The first case considered is for Reno in June 2008 when the MDA8 reached 82 ppbv. The wildfire influence for this episode is supported by PM concentrations, the known location of wildfires at the time and simulations with the Weather and Research Forecasting Model with Chemistry (WRF-Chem) which indicates transport to Reno from large fires burning in California. The contribution to the MDA8 in Reno from the California wildfires is estimated to be 26 ppbv, based on the SMR, and 60 ppbv, based on WRF-Chem. The WRF-Chem model also indicates an important role for peroxyacetyl nitrate (PAN) in producing O3 during transport from the California wildfires. We hypothesize that enhancements in PAN due to wildfire emissions may lead to regional enhancements in O3 during high fire years. The second case is for the Salt Lake City (SLC) region for August 2012. During this period the MDA8 reached 83 ppbv and the SMR suggests a wildfire contribution of 19 ppbv to the MDA8. The wildfire influence is supported by PM2.5 data, the known location of wildfires at the time, HYSPLIT dispersion modeling that indicates transport from fires in Idaho, and results from the CMAQ model that confirm the fire impacts. Concentrations of PM2.5 and O3 are enhanced during this period, but overall there is a poor relationship between them, which is consistent with the complexities in the secondary production of O3. A third case looks at high MDA8 in Boise, ID, during July 2012 and reaches similar conclusions. These results support the use of statistical modeling as a tool to quantify the influence from wildfires on urban O3 concentrations.
Wildfire smoke (WFS) increases the risk of respiratory hospitalizations. We evaluated the association between WFS and asthma healthcare utilization (AHCU) during the 2013 wildfire season in Oregon. ...WFS particulate matter ≤ 2.5 μm in diameter (PM
) was estimated using a blended model of in situ monitoring, chemical transport models, and satellite-based data. Asthma claims and place of service were identified from Oregon All Payer All Claims data from 1 May 2013 to 30 September 2013. The association with WFS PM
was evaluated using time-stratified case-crossover designs. The maximum WFS PM
concentration during the study period was 172 µg/m
. A 10 µg/m
increase in WFS increased risk in asthma diagnosis at emergency departments (odds ratio OR: 1.089, 95% confidence interval CI: 1.043-1.136), office visit (OR: 1.050, 95% CI: 1.038-1.063), and outpatient visits (OR: 1.065, 95% CI: 1.029-1.103); an association was observed with asthma rescue inhaler medication fills (OR: 1.077, 95% CI: 1.065-1.088). WFS increased the risk for asthma morbidity during the 2013 wildfire season in Oregon. Communities impacted by WFS could see increases in AHCU for tertiary, secondary, and primary care.
Exposure to wildland fire smoke is associated with negative effects on human health. However, these effects are poorly quantified. Accurately attributing health endpoints to wildland fire smoke ...requires determining the locations, concentrations, and durations of smoke events. Most current methods for assessing these smoke events (ground-based measurements, satellite observations, and chemical transport modeling) are limited temporally, spatially, and/or by their level of accuracy. In this work, we explore using daily social media posts from Facebook regarding smoke, haze, and air quality to assess population-level exposure for the summer of 2015 in the western US. We compare this de-identified, aggregated Facebook dataset to several other datasets that are commonly used for estimating exposure, such as satellite observations (MODIS aerosol optical depth and Hazard Mapping System smoke plumes), daily (24 h) average surface particulate matter measurements, and model-simulated (WRF-Chem) surface concentrations. After adding population-weighted spatial smoothing to the Facebook data, this dataset is well correlated (R2 generally above 0.5) with the other methods in smoke-impacted regions. The Facebook dataset is better correlated with surface measurements of PM2. 5 at a majority of monitoring sites (163 of 293 sites) than the satellite observations and our model simulation. We also present an example case for Washington state in 2015, for which we combine this Facebook dataset with MODIS observations and WRF-Chem-simulated PM2. 5 in a regression model. We show that the addition of the Facebook data improves the regression model's ability to predict surface concentrations. This high correlation of the Facebook data with surface monitors and our Washington state example suggests that this social-media-based proxy can be used to estimate smoke exposure in locations without direct ground-based particulate matter measurements.
The Multi‐Scale Infrastructure for Chemistry and Aerosols (MUSICA) enables the study of atmospheric chemistry and aerosols at global to local scales. MUSICA version 0 (MUSICAv0) is a configuration of ...the Community Earth System Model (CESM) with regional refinement (RR). We compared the regional scale spatial and temporal variabilities of air pollutants simulated by this newly developed multiscale model and a widely used regional model—WRF‐Chem (the Weather Research and Forecasting model coupled with Chemistry). The O3 and PM2.5 mean biases of the two WRF‐Chem simulations over the U.S. are 5–9 ppb and −3 μg/m3 during August and September 2020. The MUSICAv0 simulation has larger O3 mean bias (11 ppb) but smaller PM2.5 mean bias (−1 μg/m3) compared to the two WRF‐Chem simulations. As indicated by spatial and temporal statistical measures, the variability in meteorological parameters (temperature, precipitation), and ozone and PM2.5 are similar across a range of temporal and spatial scales in MUSICAv0 and WRF‐Chem. We demonstrate the new capability of the global MUSICAv0 at regional scale through two examples—(a) plumes from western U.S. fires that passed through the domain boundary of WRF‐Chem and were transported back into the domain and impacted air quality on the eastern U.S.; and (b) impact of stratospheric intrusion on the troposphere.
Plain Language Summary
Atmospheric chemistry is impacted by many scales from local to global, and it also interacts with other components of the Earth system. However, previously models with different infrastructures have been used separately when dealing with different scales. Recently, unified multiscale models have been developed to allow representation of global to regional to local scales. The Multi‐Scale Infrastructure for Chemistry and Aerosols (MUSICA) is an example that enables chemistry and aerosols studies at all relevant scales. To demonstrate the value of MUSICA at regional scale, we compared MUSICA with a widely used regional model and found that MUSICA exhibits skills similar to the regional model in capturing regional scale features. In addition, we show the value of MUSICAv0 at regional scale as a global and multiscale model and unified infrastructure through two examples—transdomain‐boundary transport of fire plume and stratospheric intrusion to the troposphere.
Key Points
The spatial and temporal variabilities of air pollutants simulated by Multi‐Scale Infrastructure for Chemistry and Aerosol version 0 (MUSICAv0) and Weather Research and Forecasting model coupled with Chemistry at regional scale are comparable
MUSICAv0 with variable resolution provides consistent chemistry for plumes, such as from fires, leaving and reentering a regional domain
The MUSICAv0 capability of simulating regional scales within the global context is further demonstrated by a stratospheric intrusion event
W126 is a cumulative ozone exposure index based on sigmoidally weighted daytime ozone concentrations used to evaluate the impacts of ozone on vegetation. We quantify W126 in the U.S. in the absence ...of North American anthropogenic emissions (North American background or “NAB”) using three regional or global chemical transport models for May–July 2010. All models overestimate W126 in the eastern U.S. due to a persistent bias in daytime ozone, while the models are relatively unbiased in California and the Intermountain West. Substantial difference in the magnitude and spatial and temporal variability of the estimates of W126 NAB between models supports the need for a multimodel approach. While the average NAB contribution to daytime ozone in the Intermountain West is 64–78%, the average W126 NAB is only 9–27% of current levels, owing to the weight given to high O3 concentrations in W126. Based on a three‐model mean, NAB explains ~30% of the daily variability in the W126 daily index in the Intermountain West. Adjoint sensitivity analysis shows that nationwide W126 is influenced most by NOx emissions from anthropogenic (58% of the total sensitivity) and natural (25%) sources followed by nonmethane volatile organic compounds (10%) and CO (7%). Most of the influence of anthropogenic NOx comes from the U.S. (80%), followed by Canada (9%), Mexico (4%), and China (3%). Thus, long‐range transport of pollution has a relatively small impact on W126 in the U.S., and domestic emissions control should be effective for reducing W126 levels.
Key Points
Models quantify O3 exposure (W126) in the absence of NA anthropogenic emissions
Models are biased high in the eastern US, relatively unbiased in the west
W126 in the US is most influenced by domestic NOx
Climate change is expected to have many impacts on the environment, including changes in ozone concentrations at the surface level. A key public health concern is the potential increase in ...ozone-related summertime mortality if surface ozone concentrations rise in response to climate change. Although ozone formation depends partly on summertime weather, which exhibits considerable inter-annual variability, previous health impact studies have not incorporated the variability of ozone into their prediction models. A major source of uncertainty in the health impacts is the variability of the modeled ozone concentrations. We propose a Bayesian model and Monte Carlo estimation method for quantifying health effects of future ozone. An advantage of this approach is that we include the uncertainty in both the health effect association and the modeled ozone concentrations. Using our proposed approach, we quantify the expected change in ozone-related summertime mortality in the contiguous United States between 2000 and 2050 under a changing climate. The mortality estimates show regional patterns in the expected degree of impact. We also illustrate the results when using a common technique in previous work that averages ozone to reduce the size of the data, and contrast these findings with our own. Our analysis yields more realistic inferences, providing clearer interpretation for decision making regarding the impacts of climate change.