Evidence for a larger contribution of smoldering combustion to boreal forest fire emissions from tower observations in Alaska

Abstract. With recent increases in burned area within boreal forests that have been linked to climate warming, there is a need to better understand the composition of emissions and their impact on atmospheric composition. Most previous studies have estimated boreal fire emission factors from daytime samples collected via aircraft near fire plumes or at the surface near actively burning fires. Here we quantified emission factors for CO and CH4 from a massive regional fire complex in interior Alaska during the summer of 2015 using continuous high-resolution trace gas observations from the CRV tower (Fox, AK). Averaged over the 2015 fire season, the CO/CO2 emission ratio was 0.138±0.048 and the CO emission factor was 145±50 g CO per kg of dry biomass consumed. The CO/CO2 emission ratio was about 35 % higher and more variable than most previous aircraft-based estimates for fresh wildfire emissions. The mean CH4/CO2 emission ratio was 0.010±0.003 and the CH4 emission factor was 6.05±2.09 g CH4 per kg of dry biomass consumed, with means similar to previous reports. CO and CH4 emission factors varied in synchrony, with higher CH4 emission factors observed during periods with lower modified combustion efficiency (MCE). By coupling a fire emissions inventory with an atmospheric model, we identified that at least 35 individual fires contributed to trace gas variations measured at the CRV tower, representing a significant increase in sampling compared to the number of boreal fires measured in all previous boreal forest fire work. The model also indicated that typical mean transit times between trace gas emission and tower measurement were 1–3 days, indicating that the time series sampled combustion across day and night burning phases. The high and variable CO emission factor estimates reported here provide evidence for a more prominent role of smoldering combustion, highlighting the importance of continuously sampling of fires across time-varying environmental conditions that are representative of typical burning conditions.


the upper troposphere and lower stratosphere where it can be transported across the Northern Hemisphere [Forster et al., 2001;Turquety et al., 2007;Peterson et al., 2018].
Emissions from boreal fires can significantly influence atmospheric composition throughout the Northern Hemisphere.
Fire plumes from regional fire complexes in Alaska and western Canada, for example, have been shown to influence air quality over Nova Scotia [Duck et al., 2007] and across the south-central US [Wotawa et al., 2001;Kasischke et al., 2005] and Europe 5 [Forster et al., 2001]. Similarly, fire plumes from Siberia have caused deadly air quality in Moscow [Konovalov et al.., 2011] and have affected ozone and other trace gases concentrations across the western US [Jaffe et al., 2004]. Over the past few decades, annual burned area in several regions in boreal North America has increased [Gillett et al., 2004;Kasischke and Turetsky, 2006;Veraverbeke et al., 2017], and future projections suggest further increases may occur in response to changes in fire weather and a lengthening of the fire season [Flannigan et al., 2001;de Groot et al., 2013;Young et al., 2017]. As a consequence, fires are likely 10 to play an increasingly important role in regulating air quality and climate feedbacks during the remainder of the 21 st century.
To understand boreal fire impacts on the atmosphere, high quality observations characterizing the composition of emissions are necessary [Andreae and Merlet, 2001;Akagi et al., 2011]. Here we used trace gas observations from the CARVE (CRV) tower located in interior Alaska (Fox, AK) to derive emission factors of CO and CH4 from boreal forest fires that burned over a period of a month or more during the extreme 2015 fire season. In past work, the most common approach for measuring 15 emission factors from boreal fires is to fly aircraft near or within plumes, measuring trace gases using infrared gas analyzers mounted in the aircraft or by collecting flasks of air that are measured later in the laboratory. A summary of past studies using this technique to measure CO and CH4 emission factors is shown in Table 1. Over a period of more than 25 years, in-situ CO and CH4 emission factors have been measured and reported from a total of 15 boreal fires sampled by aircraft, including 11 wildfires and 4 prescribed fires (Table 1). Additional airborne measurements of trace gas emissions from boreal forest fires are present in the 20 literature, but do not include published CO and/or CH4 emission factors for individual boreal fires. Aircraft sampling is a highly effective approach for sampling large and remote wildfires, especially for characterizing reactive trace gas emissions or aerosols that have lifetimes of hours to days. It also important to recognize potential limits associated with sampling fires in this way.
Aircraft observations are mostly confined to periods with good visibility, often sampling well-developed fire plumes during midday and during periods with relatively low cloud cover. These environmental conditions represent a subset of the variability that a 25 large wildland fire may experience as it burns over a period of weeks to months. An alternative approach to measuring in-situ emission factors involves using a tower that continuously measures trace gas concentrations located in an area downwind of fires.
This has been done in a previous boreal forest fire study during a moderate fire season in Alaska [Wiggins et al., 2016].
Environmental conditions, including weather, vegetation, and edaphic conditions are known to influence the composition of emissions, in part by regulating the importance of flaming and smoldering combustion phases [Ward and Radke, 1993;Yokelson 30 et al., 1997;Akagi et al., 2011;Urbanski, 2014]. Flaming combustion is more efficient at oxidizing organic matter directly to CO2 2007]. Over the lifetime of a large fire, smoldering combustion is more likely to occur during periods with lower temperatures and higher atmospheric humidity that increases the moisture content of fine fuels [Stocks et al., 2001;Ryan, 2002].
Here we used trace gas observations of CO, CH4, and CO2 from the CRV tower to estimate emission factors from boreal forest fires that burned during the near-record high Alaska fire season of 2015. The summer of 2015 was the second largest fire season in terms of burned area since records began in 1940 with about 2.1 million hectares burned [Hayasaka et al., 2016;Partain 5 et al., 2016]. An unseasonably warm spring and earlier snowmelt allowed fuels to dry early in the season [Partain et al., 2016]. In mid-June, thunderstorms caused an unprecedented number of lightning strikes (over 65,000) that ignited over 270 individual fires on anomalously dry fuel beds over the course of a week [Hayasaka et al., 2016;Veraverbeke et al., 2017]. Fires expanded dramatically under favorable weather conditions through mid-July, until multiple precipitation events and cool, damp weather minimized fire growth for the rest of the summer fire season.

10
The CRV tower captured an integrated signal of trace gas emissions from fires across interior Alaska during the 2015 fire season . The data stream was comprised of continuous sampling from June 15 -August 15 with more than 65,000 30s averaged samples. The CRV tower experienced enhanced and highly correlated CO, CH4, and CO2 trace gas signals from fires for about 7% of the duration of the 2015 fire season. We identified events when fire emissions had a dominant influence on trace gas variability at CRV tower and used these events to derive emission factors. This data stream represents an order of 15 magnitude increase in sampling density compared with the sum of observations from past work.
We coupled a fire emissions inventory, the Alaska Fire Emissions Database (AKFED) [Veraverbeke et al., 2015] with an atmospheric transport model, the Polar Weather Research and Forecasting Stochastic-Time Integrated Lagrangian Transport (PWRF-STILT) model [Henderson et al., 2015], to quantify the spatial and temporal variability of individual fires and their influence on CO, CH4, and CO2 at the CRV tower. Our tower-based approach allows for an integration of emission factors through 20 the day-night fire cycle and over the full duration of many individual fires. Our results suggest the smoldering phase of boreal fires may have a higher contribution to emissions than previous estimates.

CARVE (CRV) Tower Observations
Atmospheric CO, CH4, and CO2 mole fractions were measured using a cavity ring-down spectrometer (CRDS, Picarro 25 models 2401 and 2401m)  at the CRV tower in Fox, Alaska (64.986°N, 147.598°W, ground elevation 611m above sea level). The tower is located about 20 km northeast of Fairbanks Alaska (Figure 1), within the interior forested lowlands and uplands ecoregion in interior Alaska [Cooper et al., 2006]. There are three separate inlets on CRV tower at different heights above ground level from which the spectrometer draws sample air. The spectrometer samples air from the highest level for 50 minutes out of every hour, and then draws air from the other levels for 5 minutes at each level . The data 30 stream from this spectrometer has gaps every 50 minutes as the spectrometer cycles to the lower inlets. We used observations from air drawn from the top intake height at a height of 32 m above ground level because this level had the highest measurement density and the smallest sensitivity to local ecosystem CO2 fluxes near the tower . All raw 30 s average measurements were calibrated according to Karion et al. [2016]. Each 30 s average measurement served as an individual point in our calculation of emission factors described below. We isolated intervals when fire had a dominant influence on trace gas variability observed at CRV tower to calculate emission factors. An interval with dominant fire influence was defined as a continuous period that had: 1) a minimum of at least thirty trace gas measurements (with each measurement representing a mean over 30 seconds), 2) a mean CO over the entire interval exceeding 0.5 ppm, and 3) significant correlations between CO:CO2 and CH4:CO2 exceeding an r 2 of 0.80.
We used the gaps in the data stream when the spectrometer sampled air from the lower levels to separate the dataset into 5 a set of continuous intervals of trace gas observations with less than 15 s between each new 30 s averaged measurement and by applying a minimum duration criterion of at least 30 measurements. We calculated the mean CO mole fraction for each interval and removed all intervals with a mean CO less than 0.5 ppm. For each interval with high levels of CO, we then extracted CO, CH4, and CO2 mole fractions and calculated correlation coefficients between all three gases. Only periods with high and significant correlations between CO:CO2 and CH4:CO2 (r 2 > 0.80; p<0.01, n > 30) were retained, because covariance among these co-emitted 10 species is a typical signature of fire emissions [Urbanski, 2014].
We calculated background mole fractions of CO and CH4 by taking an average of observations prior to any major fire activity in interior Alaska during DOY 170 -172.5. This yielded a CO background of 0.110 ppm and a CH4 background of 1.90 ppm. We modeled hourly CO2 background concentrations to account for the influence of net ecosystem exchange (NEE) using a Bayesian approach multi-variable linear regression model trained on CRV tower observations during 2012, a year with little to no 15 fire influence on trace gas variability. We assumed negligible influence from fossil fuel combustion on background mole fractions. The hourly CO2 model was linearly interpolated to have the same temporal resolution as CRV tower data. The variables used in the CO2 model include hourly observations of temperature, vapor pressure deficit, precipitation, day of year, latent heat flux, and hourly CO2 observations from Barrow, AK ( Figure 2). Meteorological variables were acquired from the National Climatic Data Center Automated Weather Observing System for Fairbanks International Airport (http://www7.ncdc. 20 noaa.gov/CDO/cdopoemain.cmd). This location was chosen due to its proximity to the CRV tower. We obtained 3-hourly latent heat flux from the NOAH2.7.1 GLDAS/NOAH experiment 001 for version 2 of the Global Land Data Assimilation System (GLDAS-2) [Rodell et al., 2015]. Hourly in-situ CO2 observations from a clean air site in Barrow, AK were attained from the Earth System Research Laboratory Global Monitoring Division [Thoning et al., 2007]. In a sensitivity analysis we found that the removal of the background had only a small effect, because the background did not change appreciably during the duration of each ~50-25 minute time interval used to compute an emission factor.
We estimated emission ratios (ERCO/CO2 and ERCH4/CO2) by calculating the slope from a type II linear regression of CO and CH4 excess mole fractions (ΔCO and ΔCH4) relative to CO2 (ΔCO2) (Equation 1). Excess mole fractions refer to observations of trace gas mole fractions during intervals when fire had a dominant influence on tower trace gas variability with background values subtracted. Emission factors (EFCH4/CO2 and EFCH4/CO2) were calculated by multiplying the emission ratio by a scalar (SCO 30 or SCH4) to convert the molar ratio into grams of CO or CH4 emitted per kilogram of dry biomass burned with the assumption that 450 g C is emitted per kilogram of dry biomass consumed (MBiomass) [Yokelson et al., 1997;Akagi et al., 2011] (Equation 2).
We calculated modified combustion efficiency (MCE) for each emission factor period. Modified combustion efficiency 35 is defined as the excess mole fraction of CO2 divided by the sum of the excess mole fractions of CO and CO2 [Ward and Radke, 1993]. MCE was used to separate events into three categories: smoldering, mixed, or flaming. These categories reflect the dominant https://doi.org/10.5194/acp-2019-1067 Preprint. Discussion started: 4 February 2020 c Author(s) 2020. CC BY 4.0 License.
phase of combustion from fires that contributed to trace gas anomalies at the CRV tower during the summer of 2015. Periods with an MCE less than 0.9 were considered to consist of mostly smoldering combustion, periods with a MCE of greater than or equal to 0.9 and less than 0.92 were classified as consisting of a mixture of smoldering and flaming combustion, and period with an MCE greater than 0.92 were classified as flaming [Urbanski, 2014].

5
We coupled a fire emission model, the Alaskan Fire Emissions Database (AKFED) [Veraverbeke et al., 2015] with an atmospheric transport model, the Polar Weather Research and Forecasting Stochastic-Time Integrated Lagrangian Transport model (PWRF-STILT) [Henderson et al., 2015] to estimate fire contribution to trace gas variability from CRV tower observations following Wiggins et al. [2016]. AKFED predicts carbon emissions from fires with a temporal resolution of 1 day and a spatial resolution of 450 m. We regridded AKFED to the same spatial resolution as the atmospheric transport model (0.5°) for the model 10 coupling. To account for diurnal variability in emissions, here we imposed a diurnal cycle on daily emissions following Kaiser et al. [2009], where the diurnal cycle is the sum of a constant and a Gaussian function that peaks in early afternoon with 90% of emissions occurring during the day (hours 0600 to 1800) and 10% at night (hours 1800 to 0600). PWRF-STILT calculates the sensitivity of atmospheric trace gas measurements to upwind surface fluxes using an influence function or "footprint" [Henderson et al., 2015]. The footprints are on a 0.5° latitude-longitude grid with a temporal resolution of 1 h during hours 0600 to 1800 (day) Interagency Coordination Center (AICC)) to identify the location of individual fires. AKFED uses the Alaska Large Fire Database for burned area and carbon emissions estimates [Veraverbeke et al., 2015]. We determined an individual fire's contribution to CO at the CRV tower by setting all emissions in AKFED for a particular grid cell to zero and rerunning the model coupling with PWRF-STILT. We confirmed the signals were from a single fire through a complete absence of a modeled enhancement at CRV 25 tower when a specific fire was removed from the simulation. The difference between the original model and the updated coupling is equal to an individual fire's contribution to CO at the CRV tower. Due to the relatively coarse 0.5° grid cell size used for model coupling; more than one fire perimeter existed in some individual grid cells. The contributions for each fire in the same grid cell where determined by weighting the total contribution based on fire size. We also used the influence functions or "footprints" (ppm per μmol/m 2 /s) from the atmospheric transport model to quantify the contribution of day and night emissions and mean transport 30 times between the point of emission and measurement at the CRV tower. We analyzed the footprints for each time period associated with an emission factor period to confirm CRV tower observations represented an integration of emissions from multiple fires and captured variability in emissions across the diurnal fire cycle.

Emission Factors and Modified Combustion Efficiency
During the 2015 Alaska fire season, we observed synchronized enhancements of CO, CH4, and CO2 well above background concentrations in CRV tower observations from DOY 173 -196 ( Figure 3). We identified 53 individual events that span about 50 minutes each to calculate emission factors from the elevated trace gas observations ( Figure 4; Table 2). CO/CO2 emission ratios ranged from 0.025 to 0.272 and CH4/CO2 emission ratios ranged from 0.002 to 0.020. MCE ranged from 0.786 to 0.975 (Table 2). CO emission factors ranged from 26 to 286 g CO per kg biomass combusted, and CH4 emission factors ranged 5 from 1.21 to 12.2 g CH4 per kg biomass combusted. The mean CO/CO2 emission ratio was 0.138 ± 0.044, the mean CO emission factor was 145 ± 46 g CO per kg biomass combusted, and the mean MCE was 0.879 ± 0.068. Concurrently, the mean CH4/CO2 emission ratio was 0.010 ± 0.003 and the mean CH4 emission factor was 6.05 ± 1.95 g CH4 per kg biomass combusted. A strong linear relationship existed between the CH4 emission factor and MCE across the different sampling intervals ( Figure 5).
Each event was used to calculate emission factors was classified as a smoldering, mixed, or flaming emissions event using 10 the MCE. We categorized 39 smoldering events, 9 mixed events, and 5 flaming events throughout the fire season (examples shown in Figure 6 and  (Table 3).

20
The forward model simulations combining AKFED fire emissions with PWRF-STILT confirmed that the elevated CO signals at the CRV tower can be attributed primarily to boreal forest fire emissions (Figure 7). The AKFED model had a Pearson's correlation coefficient of 0.61 with observed daily mean CO and had a low bias of approximately 7%.
We identified 35 individual fires that contributed to at least 1% of the CO mole fraction time series at CRV tower ( Figure   8; Figure 9; Table 3). On average, these fires were 295 ± 131 km away from CRV tower and located mostly to the west of Fairbanks, 25 in the direction of the prevailing summer surface winds. The total CO emitted from these fires accounted for 75% of the excess CO mole fraction signal observed at CRV tower during DOY 160 -200. The remaining CO signal originated from many smaller fires that were widely distributed across interior Alaska. The Tozitna fire was responsible for the greatest percentage of the total CO anomaly integrated over the 2015 fire season at CRV tower. This fire contributed 11% of the total CO anomaly observed at CRV tower. The fires that significantly contributed the most to the CO anomaly at CRV tower were not necessarily the closest 30 fires to the tower or the largest fires of the 2015 fire season in terms of burned area. Combined, however, this set of 35 fires accounted for 0.97 Mha, or approximately 65% of the total burned area reported during the 2015 fire season [Veraverbeke et al., 2017].
The footprints associated with each emission factor event also were used to determine how much of the signal was coming from burning on previous days and the fraction of emissions emitted during day and night periods. We found that 99% of the fire 35 emissions that influenced CRV tower trace gas concentrations occurred within 3 days of the sampling interval used to derive the emission factor for an individual event at the CRV tower, with 76% occurring within the first 24 hours, 21% during the next 24 hours, and 2% occurring three days prior to the event ( Figure 10). Overall, 64% of the fire emissions that impacted the tower occurred during the day (0900 to 1800 local time) and 36% occurred at night (1900 -0600 local time).

Comparison of tower and aircraft emission factors
Emission factors provide a straightforward way to convert fire emissions of dry biomass into specific trace gas species, such as CO, CH4, and CO2. This technique is commonly used to model emissions of select species and or to compare model results with in-situ or remotely sensed observations. There are limited previous studies on boreal forest fire emission factors, and almost 5 all derived emission factors were determined from aircraft sampling (Table 1) [Cofer et al., 1990;Radke et al., 1991;Nance et al., 1993;Cofer et al., 1998;Goode et al., 2000;Simpson et al., 2011]. In total, all previous aircraft-based studies combined sampled 15 individual prescribed and/or wildfires and derived emission factors that were likely most representative of flaming fires that occurred in the afternoon and were strong enough to generate well-defined plumes. Our emission factors for CO were 35% higher than previous estimates for wildfires derived from the aircraft measurements. We believe a primary contributor to this difference 10 is sampling methodology. Our CRV tower-based sampling was able to integrate over day-night burning cycles, flaming combustion at active fire fronts as well as residual smoldering combustion in soils that persists for days behind the fire line, and emissions associated with a wide range of environmental conditions that occurred during 2015 fire season. This integration was possible because the tower was located several hundred kilometers downwind of the core fire complex located western Alaska. The time delays between emission and detection of trace gas anomalies at CRV allowed for atmospheric mixing of signals from dozens of 15 different fires in different stages of growth and extinction. Collectively, these fires experienced time-varying environmental conditions that were less ideal for flaming combustion than the fire plumes sampled in mid-afternoon by the aircraft.

Integration of emission factor observations across studies and time intervals
Our modeling study confirms that the entire day/night fire cycle was captured by anomalous trace gas observations at CRV tower that was used to calculate emission factors. Wiggins et al. [2016] used a similar tower-based approach to estimate 20 boreal forest emission factors during a moderate fire year, and they found CO and CH4 emission factors that were higher than the compiled mean from previous studies. We found a strong linear relationship between CH4 emission factors and MCE that has also been observed in previous studies [Van Leeuwen and Van Der Werf, 2011;Yokelson et al., 2013;Urbanski, 2014].
Although Table 1

Relative Contributions of Smoldering and Flaming Combustion
Following ignition, boreal forest fires generally begin as stand replacing crown fires followed by smoldering combustion in organic soil layers and coarse woody debris behind the fire front that can continue for weeks after ignition [Bertschi et al., 2003].
This residual smoldering combustion could substantially contribute to trace gas emissions, but is difficult to detect and quantify 35 using remote sensing because of low radiative power associated with this phase of combustion. The relative contributions of https://doi.org/10.5194/acp-2019-1067 Preprint. Discussion started: 4 February 2020 c Author(s) 2020. CC BY 4.0 License.
emissions from flaming and smoldering are uncertain for boreal forest fires, but previous studies have assumed 80% of aboveground carbon is consumed in flaming combustion, 20% is consumed in smoldering combustion, and vice versa for belowground carbon [French et al., 2002;Kasischke and Bruhwiler, 2002]. Our results suggest that the smoldering phase of combustion contributes to more carbon emissions than has been previously estimated.

5
Smoldering combustion produces significantly more CO and PM2.5 than flaming combustion [Bertschi et al., 2003;Chen et al., 2007;Stockwell et al., 2016], and corresponding boreal forest fire emissions of these species are likely higher than previous studies suggest. This conclusion implies changes to the overall impact of boreal forest fires on human health, atmospheric composition, and climate. Emissions from boreal forest fires have the potential to be transported long distances across the Northern Hemisphere [Forster et al., 2001], implying large-scale impacts. CO can lead enhanced tropospheric ozone production downwind 10 of a fire [Lapina et al., 2006], and higher concentrations of CO from fires may indirectly contribute to radiative forcing by consuming hydroxyl radicals and extending the lifetime of CH4 [Levine and Cofer, 2000]. PM2.5 emissions, in contrast, can significantly degrade regional air quality, endanger cardiovascular and respiratory health, and influence the radiative balance of the planet [Reid et al., 2016]. Much of the PM2.5 emitted by smoldering fires is composed of organic carbonaceous aerosol that often leads to climate cooling [Tosca et al., 2010;Jayarathne et al., 2018].

Conclusions
Our tower-based approach to calculate emission factors is a new technique that significantly improves our understanding of trace gas emissions from boreal forest fires. Unlike traditional approaches using aircraft observation, our method represents an integration of trace gas emissions across day and night burning cycles and varying environmental conditions that are both known to considerably influence the composition of fire emissions. We discovered 35 individual fires across interior Alaska significantly 20 influenced trace gas variability at CRV tower from which our emission factors were derived from. This is more than double the number of individual wildfires that have been sampled to calculate boreal fire emission factors in all previous studies combined.
Our results suggest the smoldering phase of boreal forest fires contributes to more trace gas emissions than previously believed, and as a consequence, total CO emissions from boreal forest fires may have been underestimated. The tower-based emission factor method introduced in this study can be applied to other biomes and potentially expand in-situ emission factor observations in 25 regions of interest.        Tables   Table 1. Comparison of CO and CH4 emission ratios, emission factors, and modified combustion efficiency (MCE) from previous studies that measured in-situ boreal fires. Number of fires sampled per study is given in parenthesis following the location. Aircraft based studies are denoted with an (A) following the sampling strategy and ground based studies are denoted with a (G) following 5 the sampling strategy. The studies are organized based on the type of fire they measured, prescribed fires or wildfires. We did not include studies that measured slash and tramp prescribed fires or laboratory fires. The CO emission ratio column has units of ppmv