Impacts of Biomass Burning on Ozone, Particulate Matter, and Carbon Dioxide in the Northwest U.S.

Abstract

Wildfires (or Biomass Burning events) in the northwest U.S. have been increasing in size and frequency throughout the last few decades. This significantly affects vulnerable populations through hazardous fine particulate matter (diameter < 2.5 µm [PM2.5]) and/or ozone (O3) exposure, loss of habitat and homes, and causes disruption of outdoor activities throughout the summer season. While the U.S. EPA regulates many hazardous pollutants through the National Ambient Air Quality Standards (NAAQS) for the protection of human health, during wildfire events, these pollutants can reach dangerous levels. In most cases, these exceedances of the NAAQS during wildfire events cannot be controlled. While we cannot completely control wildfires, we can endeavor to understand how the emissions from these events affects our air quality. In order to accomplish this, we must first determine how to trace and attribute different pollution events to wildfire emissions. This goal is highlighted by the probability of wildfire occurrence, which is predicted to increase through the end of the century. This dissertation provides results from three studies which investigate the impact of wildfires in the northwest U.S. by tracing their emissions. First, I investigate the role of carbon dioxide (CO2) and its use as a tracer for biomass burning. Secondly, I use historical PM2.5 data to determine which areas in the U.S. are seeing significant increases in PM due to wildfires. Lastly, I investigate the attribution of wildfire-influence to high O3 days seen in urban areas using observations and modelling. To investigate the role of CO2 as a tracer for biomass burning, I used data from the Mount Bachelor Observatory (MBO) in central Oregon for 2012-2014. First I investigated the variations in CO2 in the free troposphere (FT) and boundary layer (BL) to determine variations due to wildfires, different transport patterns, and seasonal variations. For all seasons, I found that FT air had a higher average CO2 mixing ratio than BL air. BL air was most often observed during the afternoon and evening (12-20 PST), while FT air was most often observed during the night and early morning (20-8 PST) due to the role of up-slope and down-slope flow, respectively. Fall and summer showed the lowest mixing ratios of CO2, while winter and spring showed the highest due to uptake and respiration of vegetation, respectively. The maximum daily change in CO2 was found during spring and summer. Using HYSPLIT back-trajectories and cluster analyses during spring months, I determined that the highest CO2 mixing ratios were also associated with the highest O3 and lowest water vapor mixing ratios. I also looked at four case studies (one long-range transport and three wildfire events) that showed significant variations in CO2. During one of the wildfire events, I saw the expected profile, large enhancements of CO2 that were well correlated with carbon monoxide (CO). However, in a different wildfire case, CO2 decreased during the event. This was likely caused by uptake of CO2 during BL transport, which counteracted the enhancements from wildfire emissions. This event also provided insight into the variations of a typically used wildfire tracer, CO2, and showed that it is not always reliable. Significant variations in downwind CO2 mixing ratios may also influence metrics that use CO2 in the calculations, such as modified combustion efficiency (MCE). Assessing historical data from rural monitoring sites (IMPROVE sites) across the contiguous U.S., I evaluated PM2.5 trends during 1988-2016. To calculate trends in the policy-relevant 98th quantile of PM2.5¬, I use Quantile Regression (QR). This methodology allows me to estimate changes, not only in the mean, but in the full distribution. I also use Kriging and Gaussian Geostatistical Simulations to interpolate changes in PM2.5 trends between monitoring sites. Overall, the 98th quantile PM2.5 showed a positive trend in the northwest U.S. (average = 0.24 ± 0.15 µg/m3/yr [± 95% CI]). This was in contrast with overall negative trends in the 98th quantile PM2.5 observed in the rest of the country (average -0.56 ± 0.10 µg/m3/yr). The decrease in the rest of the country is likely due to the reduction in anthropogenic emissions seen in recent years. I also evaluated 98th quantile trends in total carbon (TC) and sulfate across the country to examine the effects of wildfire and anthropogenic emissions on the PM2.5 trends, respectively. This analysis showed a positive trend in TC and no trend in sulfate across the northwest U.S., which confirmed the influence of wildfire activity on positive PM2.5 trends. I also evaluated daily MODIS Aqua aerosol optical depth (AOD) data for 2002-2017 to compare with ground-based trends. Comparing the IMPROVE and MODIS data, I found positive 98th quantile PM2.5 trends in the Northwest (1.94 ± 1.00 & 2.12 ± 0.81 %/yr, respectively) and negative trends throughout the rest of the country (-2.81 ± 0.43 & -1.20 ± 0.51 %/yr, respectively) through 2016. If I include 2017 MODIS data, the trend in the Northwest is even greater due to 2017 being a very high wildfire year. Overall, these results indicated a decrease in PM2.5 over a majority of the country, while wildfires influence an increase in the 98th quantile PM2.5 trends across the Northwest. In order to investigate wildfire impacts on O3 in an urban area, I use data from the St. Luke’s site in Meridian, ID during a summer intensive campaign in 2017. To identify wildfire influenced periods, I calculate a wildfire criterion based on the NOAA Hazard Mapping System (HMS) smoke product and historical PM2.5 data at St. Luke’s. I also ran a Generalized Additive Model (GAM), which uses multiple prediction variables (i.e., meteorological and back-trajectory data) to model a response variable (i.e., MDA8 O3). When the GAM prediction significantly differs from observed data, I investigate unusual sources of O3 (such as wildfire influences) in this urban area. During the 2017 summer campaign, I found that peroxyacetyl nitrate (PAN), reactive nitrogen (NOy), and maximum daily 8 hour averaged (MDA8) O3 are significantly enhanced during wildfire smoke events compared with non-smoke periods (56%, 38%, and 28%, respectively). I also calculate enhancement ratios (ERs) to assist in the identification of wildfire plumes for ΔPM2.5/ΔCO, ΔNOy/ΔCO, ΔPAN/ΔNOy, and ΔPAN/ΔCO. The 95% confidence intervals for these ERs are found to be 0.123 – 0.133 µg/m3/ppbv, 0.058 – 0.066 ppbv/ppbv, 0.150 – 0.191 ppbv/ppbv, and 3.02 – 3.76 ppbv/ppmv, respectively, for wildfire influenced events. These ERs generally reflect the accepted literature values considering some differences due to mixing with urban air instead of sampling in a non-polluted environment. Using historical data from the St. Luke’s site for 2006-2017, I determine the change in O3 production on smoke vs. non-smoke days. During non-smoke days, I observed a decrease in MDA8 O3 with increasing PM2.5, which was attributed to high NOx mixing ratios and NOx-titration of O3. On smoke days, however, I observed enhancements in O3 production up to high smoke levels (PM2.5 ~ 60-70 µg/m3). At higher PM2.5 concentrations, there is a reduction in average MDA8 O3. Additionally, the GAM shows that while only 4% of days are classified as smoke days, these days show significantly higher residuals than non-smoke days. This suggests that the enhancement in O3 on smoke days is not associated with standard meteorology or transport variables. I also examine four wildfire-influenced high O3 events in this urban area to evaluate the significantly variable conditions that result. In two cases, I investigate smoke days that show significant O3 enhancement and moderate PM2.5 concentrations. These cases suggest that ERs, such as ΔPM2.5/ΔCO and ΔNOy/ΔCO, are less useful in determining wildfire influence in an urban area on moderate smoke days. Another case shows reduced O3 production during a very high, 3-day smoke event (PM2.5 > 70 µg/m3). The day after this high smoke period, a 20 ppbv enhancement in MDA8 O3 is observed in moderate smoke concentrations. The enhancement in MDA8 O3 during moderate smoke immediately following a high smoke periods implies a reduction in O3 production efficiency during high smoke concentrations. These results indicate that wildfire influenced O3 enhancements are highly variable in urban areas but generally increase up to around 60 µg/m3 of PM2.5, after which they decrease at very high smoke concentrations. This analysis also suggests that multiple tracer methodologies, such as GAM results, back-trajectories, and PAN mixing ratios, are effective at characterizing wildfire influence on MDA8 O3 in urban areas. Results from this dissertation have important implications for the identification of wildfire influences in urban and rural areas. They suggest the current suite of measurements taken may be ill-equipped or incomplete to fully investigate wildfire influences. Current CO2 and MCE criteria for wildfire influence may be biased by BL loss of CO2 during transport. ERs typical for wildfires are useful in urban areas when smoke influence is large, but less useful at moderate smoke levels due to high background concentrations of other urban pollutants already within the area. Additionally, highly variable PM2.5 concentrations in wildfire plumes can affect O3 production. While PM2.5 trends are decreasing throughout most of the U.S., policy-relevant 98th quantile PM2.5 shows an increase in the Northwest due to wildfire influence. With a predicted increase in wildfires through the end of the century, it is vitally important to develop methods to assess when wildfire emission are impacting an area. This could be achieved by developing an instrument to measure specific wildfire tracers, such as acetonitrile or a suite of volatile organic compounds (VOCs). Although there are instruments that can already measure these types of compounds, this instrument would need to be specific to wildfires compounds and not require special training or equipment to operate. This would allow cities/states/researchers to positively or negatively trace the influence of wildfires at low operational and personnel cost. This instrument would greatly expand our understanding of the impacts on air pollution due to wildfire emissions.