Evidence for multi-decadal fuel buildup in a large California wildfire from smoke radiocarbon measurements

In recent decades, there has been a significant increase in annual area burned in California’s Sierra Nevada mountains. This rise in fire activity has prompted the need to understand how historical forest management practices affect fuel composition and emissions. Here we examined the total carbon (TC) concentration and radiocarbon abundance (Δ14C) of particulate matter (PM) emitted by the KNP Complex Fire, which occurred during California’s 2021 wildfire season and affected several groves of giant sequoia trees in the southern Sierra Nevada. During a 26 h sampling period, we measured concentrations of fine airborne PM (PM2.5), as well as dry air mole fractions of carbon monoxide (CO) and methane (CH4), using a ground-based mobile laboratory. We also collected filter samples of PM2.5 for analysis of TC concentration and Δ14C. High correlation among PM2.5, CO, and CH4 time series confirmed that our PM2.5 measurements captured variability in wildfire emissions. Using a Keeling plot approach, we determined that the mean Δ14C of PM2.5 was 111.6 ± 7.7‰ (n = 12), which was considerably enriched relative to atmospheric carbon dioxide in the northern hemisphere in 2021 (−3.2 ± 1.4‰). Combining these Δ14C data with a steady-state one-box ecosystem model, we estimated that the mean age of fuels combusted in the KNP Complex Fire was 40 years, with a range of 29–57 years. These results provide evidence for emissions originating from woody biomass, larger-diameter fine fuels, and coarse woody debris that have accumulated over multiple decades. This is consistent with independent field observations that indicate high fire intensity contributed to widespread giant sequoia mortality. With the expanded use of prescribed fires planned over the next decade in California to mitigate wildfire impacts, our measurement approach has the potential to provide regionally-integrated estimates of the effectiveness of fuel treatment programs.


Introduction
Wildfire activity in California, including area burned and occurrence of large fires, has increased over the last several decades, with intensifying impacts on society, the economy, and ecosystems (Dennison et al 2014, Williams et al 2019, Safford et al 2022, Wang et al 2021. During high fire years, mandatory evacuation orders force tens of thousands of people to flee their homes (Safford et al 2022). Direct costs of wildfires, including fire suppression and property losses, have more than doubled in the last decade (California Department of Forestry and Wildfire Protection, 2022a). Wildfires also incur significant indirect costs (Wang et al 2021) like environmental cleanup, lost business revenue, infrastructure repair, and health impacts. Wildfires account for up to half of the exposure to PM 2.5 (airborne particulate matter (PM) with diameter <2.5 µm) in the western U.S. (Burke et al 2021), and smoke exposure has been linked to increased respiratory-related hospitalizations and adverse health outcomes related to restricted activity and days of work lost (Kochi et al 2010, Reid et al 2016). California's ecosystems are also negatively affected by intensifying wildfires, which can weaken landscape-scale carbon storage, shift vegetation composition, reduce biodiversity, and threaten water supplies and other ecosystem services (Wu et al 2011, Stevens 2017a, Foster et al 2020. These worsening impacts motivate an urgent need for research aimed at informing and evaluating wildfire management strategies in California and the western U.S.
In the Sierra Nevada mountains of California, burned area has increased by more than sevenfold since the 1980s, mainly as a consequence of the cumulative effects of widespread fire suppression and climate change (Taylor et al 2016, Williams et al 2019, Gutierrez et al 2021, Hagmann et al 2021. Widespread suppression of low-to moderateintensity fires has allowed overgrowth of shrubs and small trees, which compete with larger trees for resources (e.g. water) and serve as ladder fuels that facilitate high-intensity crown fires (McKelvey and Busse 1996, Stephens et al 2009, Pausas and Keeley 2019. In addition to fuel buildup, the region has experienced warmer and drier conditions that increase the likelihood of more frequent and extensive wildfires (Williams et al 2019, Abatzoglou et al 2021, Gutierrez et al 2021, Higuera and Abatzoglou 2021. Although California's coniferous trees evolved with fire and can withstand low-to moderateintensity fires with a return interval of about 15 years (Swetnam 1993, Swetnam et al 2009, they are vulnerable to today's high-intensity crown fires (Shive et al 2022). This is especially important to consider in oldgrowth forests, such as stands of the endangered giant sequoia (Sequoiadendron giganteum, (LINDL.) J.T. BUCHHOLZ), which have experienced widespread mortality in recent higher-intensity fires (Shive et al 2022  (Stevens et al 2021), motivating greater investments in forest and wildfire management practices, including prescribed fire, to reduce fire severity (Tubbesing et al 2021).
In this study, we measured the radiocarbon abundance (∆ 14 C; Stuvier and Polach 1977) of PM 2.5 emitted by the KNP Complex Fire over a 26 h sampling period, along with in situ trace gas dry air mole fractions and PM 2.5 concentrations, which were used to validate the influence of wildfire emissions on variability in observed PM 2.5 concentrations. We then used this information to constrain the ages (and types) of combusted fuels and explore implications for fire intensity. Radiocarbon ( 14 C) is a radioactive isotope of carbon that is naturally produced in the atmosphere. Still, the ∆ 14 C of atmospheric carbon dioxide (CO 2 ) has changed considerably over the past 70 years because of the Earth system's response to the production of 'bomb' 14 C from aboveground nuclear weapons testing during the late 1950s and early 1960s (Nydal 1963, Levin et al 2010. In past work, ∆ 14 C measurements of fire-emitted PM have been used to identify the depth of burning in organic duff layers in boreal forest ecosystems (Mouteva et al 2015) and the age of combusted peats contributing to severe haze events in Southeast Asia (Wiggins et al 2018). For western U.S. wildfires, ∆ 14 C measurements can provide information about the size classes of combusted fuels because larger-diameter fuels decompose more slowly and, therefore, persist longer within ecosystems (Harmon 2021).
We hypothesized that fire-emitted PM 2.5 would be considerably enriched in 14 C relative to current atmospheric CO 2 because larger-diameter fuels incorporated 14 C via photosynthesis over several decades when the atmosphere had more bomb 14 C. Older fuels are associated with larger fuel classes, which decompose relatively slowly and build up for decades if not removed mechanically or by fire. Several methods are commonly used to assess fire severity, including the composite burn index (Key and Benson 2006) and remotely-sensed indices, like the normalized burn ratio (Miller et al 2009a). Measuring the ∆ 14 C of fire emissions may complement these approaches by constraining the age (and therefore, size distribution) of combusted fuels. By sampling the atmosphere downwind of a large wildfire, our technique also provides a means to obtain a regionally-integrated snapshot across actively burning areas within the entire fire perimeter.

Sampling location
We collected PM 2.5 samples and measured in situ trace gas and PM 2.5 levels with a ground-based mobile laboratory in the town of Three Rivers, CA, USA (36.453 • N, 118.873 • W, 361 m above sea level) during 2-3 October 2021. We chose this sampling location to be as close to the KNP Complex Fire as possible without being inside a mandatory evacuation zone. This was within a residential community approximately 0.5 km from the town's main road. Traffic was minimal because the area was under an evacuation warning. Our sampling location was about 10 km from two actively burning fire fronts that were located to the north and east and approximately 1000 m lower in elevation. The KNP Complex Fire resulted from two lightning fires (Colony and Paradise Fires) that merged into one large fire on 17 September 2021 (Stephenson and Brigham 2021). The fire was 100% contained on 16 December 2021 after reaching a final size of 357 km 2 (88 307 acres) (California Department of Forestry and Fire Protection 2022b). Over the course of the fire, it burned 18 km 2 (4374 acres) of giant sequoia groves (Shive et al 2022).  Figure 1 shows our sampling location and the distribution of Suomi/NPP Visible and Infrared Imaging Radiometer Suite active fires for the 1:30 pm overpass on 2 October. The active fire line and outer fire perimeter for the KNP Complex Fire on 2 October are also mapped using information from the Fire Events Data Suite version 1 (Chen et al 2022). This fire information is superimposed on an Aqua MODIS 250 m true color image displaying the spatial extent of smoke from the KNP Complex Fire on 2 October. The location of giant sequoia groves is also provided using data from the National Park Service (National Park Service and Sequoia and Kings Canyon National Parks 2017). Three lines of evidence confirmed our sampling location was downwind of the fire and its plume: (1) elevated and covarying trace gas and PM 2.5 concentrations (as described in Results), (2) visual inspection of the smoke cloud in figure 1 and (3) backtrajectory analysis conducted with NOAA's HYSPLIT model (data not shown).
Air temperature and relative humidity reported here were measured by a compact weather sensor (METSENS500, Campbell Scientific, Logan, UT, USA) mounted on the mobile laboratory approximately 3 m above ground level. The air temperature was 32.0 • C at the beginning of the sampling period at 3:00 pm on 2 October and declined through the evening and night, reaching a minimum of 17.4 • C at 7:45 am on 3 October. The temperature rose through the morning and afternoon on 3 October, reaching a maximum of 27.2 • C at 2:25 pm. Relative humidity varied from 14% to 26% over the sampling period, within the minimum observed at 3:00 pm on 2 October and the maximum at 26% at 10:30 am on 3 October.

PM 2.5 measurement and collection
We measured in situ PM 2.5 mass concentrations (µg PM 2.5 m −3 air) using a PurpleAir PA-II air quality sensor with laser particle counting (PurpleAir, Draper, UT, USA) at a sampling height of approximately 1.2 m above ground. We analyzed the PurpleAir Channel A data, which is appropriate for outdoor conditions and has a 2 min resolution.
Filter samples of PM 2.5 were collected at an inlet height of 1.2 m above ground level using a low-volume aerosol sampler (MiniVol Tactical Air Sampler, AirMetrics, Springfield, OR, USA) with a PM 2.5 impactor (#202-100) over periods of 0.5-6.0 h. Sampling duration was determined according to concurrent PM 2.5 concentration (i.e., sampling duration was inversely related to concurrent PM 2.5 concentration). To maintain a uniform volume flow, we used ambient temperature and pressure to adjust the sampler's flow rate between 4.50 and 4.75 l min −1 . Samples were collected on 47 mm diameter quartz fiber filters (GE Healthcare Life Sciences, #1851-047). Before sampling, filters were pre-combusted at 500 • C for 3 h, wrapped individually in aluminum foil, and stored in plastic bags. Blank filters were mounted on the inside of the sampler housing with no active airflow and collected concurrently with 2 of the 12 PM 2.5 samples (#253740 and 253744). After collection, filters were wrapped in aluminum foil, sealed in plastic bags, and stored at −20 • C. A total of 12 PM 2.5 samples and 2 blanks were collected.

Elemental and isotopic analyses of PM 2.5 2.3.1. Measurement of TC and radiocarbon
Filter samples of PM 2.5 were analyzed for TC concentration (µg C m −3 ) and ∆ 14 C at the W. M. Keck Carbon Cycle Accelerator Mass Spectrometry (KCCAMS) laboratory at the University of California, Irvine. We focused our analysis on the ∆ 14 C of PM 2.5 because it is a criteria pollutant for the U.S. Environmental Protection Agency (EPA) and is associated with a wide range of particles generated during combustion (Andreae 2019). Examining fuel age, PM 2.5 presents various practical advantages: it is conveniently sampled in the field and in wildfire smoke, and its relatively high signal-to-noise ratio enables a clear differentiation from the background atmosphere.
For each sample, a portion of or the entire filter (depending on estimated TC concentration) was sealed with cupric oxide in pre-combusted 6 mm OD quartz tubes and oxidized to CO 2 at 900 • C for 3 h. This TC-derived CO 2 was then extracted and quantified manometrically on a vacuum line, reduced to graphite using a modified sealed-tube zinc reduction method (Walker and Xu 2019), and its ∆ 14 C was measured via accelerator mass spectrometry alongside graphitization standards and blanks. The units of ∆ 14 C are per mille (‰), and the relationship between ∆ 14 C and the widely-used fraction modern (F) measurement is shown in equation (1), where y is the year of 14 C sampling (2021); F is the 14 C/ 12 C ratio of the sample divided by 95% of the 14 C/ 12 C ratio of the oxalic acid I (OX-I) standard measured in 1950 ( 14 C/ 12 C OX1 = 1.176 ± 0.010 × 10 −12 ) corrected for mass-dependent fractionation; 8267 years is the mean lifetime of 14 C; and 1950 is the reference year, The 14 C data are normalized to a common δ 13 C so that differences in ∆ 14 C do not reflect isotopic fractionation processes (Stuvier and Polach 1977).
TC and ∆ 14 C measurements for our PM 2.5 samples are reported in table 1. Error in TC was computed assuming 3% error in sampler flow rate, 0.5% error in sampling duration, and 6% error in our measurement of the filter area. The measured ∆ 14 C of each sample (∆ 14 C meas ) was corrected for contamination by extraneous carbon associated with handling the filters during field and laboratory work using blank filters and a simple isotopic mass balance equation (equation (2)): where 'sample' refers to the sampled filter and 'blank' refers to the blank filter. The mean blank TC concentration was 0.1 µg C cm −2 , and the blank mean ∆ 14 C was −278.6 ± 14.3‰. The error (err) in ∆ 14 C sample was calculated according to equation (3):  (4)) draws upon the linear relationship between ∆ 14 C sample and 1/TC sample and allows for identification of the wildfire ∆ 14 C end member from the regression intercept: The 12 sampled filter measurements of 1/TC and ∆ 14 C were fit using a geometric mean regression, which accounts for errors in both TC and ∆ 14 C measurements. The regression was calculated using the 'lsqfitgm' function in Matlab developed by E. T. Peltzer at the Monterey Bay Aquarium Research Institute 4 . The standard deviation (SD) of wildfire ∆ 14 C is the SD of the y-intercept calculated by the regression function. The Keeling plot approach assumes the composition of the wildfire end member and background atmosphere remain constant over the sampling duration.

Estimation of combusted fuel mean age
The mean age of the combusted fuel was determined using a steady-state one-box ecosystem model forced with ∆ 14 C of the historical atmosphere using observations by Hua et al (2022) and X Xu 5 . Given a user-prescribed mean age of the carbon pool, the model simulates the evolving ∆ 14 C of the pool from 10 000 years before present to 2021 with inputs from photosynthesis and losses from decomposition and radioactive decay each year. We ran the model for a range of mean ages (between 5 and 75 years) to simulate the ∆ 14 C of the terrestrial carbon pool in 2021. We then matched the mean ∆ 14 C of combusted fuel end member from the Keeling plot approach described above (section 2.3.2) to the mean age from the model corresponding to the closest matching ∆ 14 C value. We estimated a 1σ uncertainty range for the fuel mean ages by identifying where the measured wildfire ∆ 14 C end member value minus 1σ intersected the curve of ecosystem model predictions. This approach generated an asymmetric uncertainty range for the fuel age.

Trace gas measurements
Carbon monoxide (CO) and methane (CH 4 ) dry air mole fractions were measured using a wavelengthscanned cavity ring-down spectrometer (G2401, Picarro, Santa Clara, CA, USA) at an inlet height of approximately 3 m above ground at approximately 1 s resolution. We calibrated measurements before 4 lsqfitgm function by E. T. Peltzer, Monterey Bay Aquarium Research Institute. 5 X. Xu, personal communication 2022. and after the sampling period using two NOAAcertified air standards in compressed gas cylinders with known mole fractions of CO and CH 4 that spanned the range of observed values. Standards were measured for approximately 5 min during the calibration period. Using the linear relationship between known values and values measured during the calibration period, we applied a two-point correction to the CO and CH 4 data obtained during the sampling period (Hopkins et al 2016, Yañez et al 2022). Outliers in the data (values more than three scaled median absolute deviations from the median) were replaced with linear interpolation of neighboring, non-outlier values using the 'filloutliers' function in Matlab 6 . Measurements collected when cavity pressure or temperature in the instrument was unstable (pressure/temperature change between measurements >0) were removed. Calibrated data with outliers removed were then averaged to a 1 min resolution.

Results
Throughout the 26 h sampling period, CO, CH 4 , and PM 2.5 measurements varied synchronously and were elevated relative to expected background levels (figure 2). CO near the beginning of the sampling period at 5:10 pm on 2 October was approximately 1330 ppb. CO increased rapidly at first and then more gradually, reaching a level of 6000 ppb around 4:30 am on 3 October ( figure 2(a)). CO levels varied between 6000 and 5630 ppb from 4:00 am to 11:00 am on 3 October before increasing again in the early afternoon, reaching a maximum of 7890 ppb by 3:10 pm. CH 4 measurements followed a similar temporal pattern with a minimum initial dry air mole fraction of 2090 ppb at 5:10 pm on 2 October and then increasing to a maximum of 2870 ppb at 3:10 pm on 3 October ( figure 2(b)). PM 2.5 concentrations increased during the evening of 2 October with similar timing to CO and CH 4 . PM 2.5 remained relatively constant from midnight to 6:15 am on 3 October (figure 2(c)). A minimum in PM 2.5 concentration of 277 µg m −3 was observed at 7:20 pm on 2 October and a maximum of 1330 µg m −3 was observed at 12:25 pm on 3 October. Compared to CO and CH 4 , PM 2.5 had a much less pronounced rise in the early afternoon on 3 October. The Pearson correlation coefficients for the different trace gas and PM 2.5 time series were relatively high: 0.99 for CO and CH 4 , 0.91 for CO and PM 2.5 , and 0.90 for CH 4 and PM 2.5 . The simultaneous buildup of fire-emitted PM 2.5 and trace gases during the evening and night of 2 October is consistent with a collapsing planetary boundary layer and downslope flow from the fire to our lower elevation sampling location, which The TC concentration of our filter samples closely tracked the in-situ optical estimates of PM 2.5 concentration and varied between 97.5 and 556.4 µg C m −3 (figure 2(c)). The ∆ 14 C of TC was negative near the beginning of the sampling period (figure 2(d)), which is consistent with a substantial contribution from fossil emissions in the background atmosphere (Mouteva et al 2015(Mouteva et al , 2017. During the evening on 2 October, ∆ 14 C increased concurrently with the rise in trace gas mole fractions and PM 2.5 concentrations, reaching a maximum of 81.0 ± 3.1‰ for the sample collected between midnight and 4:00 am on 3 October. From 4:00 am to mid-afternoon on 3 October, ∆ 14 C values were relatively constant, varying between 66.8 ± 2.0‰ and 83.6 ± 5.6‰ before declining to 43.4 ± 2.1‰ during the last sampling interval. The ∆ 14 C of TC increased at higher PM 2.5 concentrations, suggesting that fire-emitted PM 2.5 , which we expect to be enriched in 14 C, drove variability in emissions. Using the Keeling plot approach described in section 2.3.2, we estimated that the mean ∆ 14 C of the wildfire end member was 111.6 ± 7.7‰ (figure 3). The ∆ 14 C of emissions was enriched relative to the northern hemisphere atmospheric CO 2 background Figure 4. Analysis of fuel age using a steady-state one-box ecosystem model. (a) Atmospheric ∆ 14 C (black line) and modeled ∆ 14 C of combusted fuel (colored lines) over time from 1940 to 2021. Combusted fuel ∆ 14 C is modeled here given mean ages ranging from 5 to 75 years, visualized every 1 year. The black filled circle represents the modeled combusted fuel ∆ 14 C for a mean age of 40 years, estimated by the model to correspond to the mean observed ∆ 14 C of 111.6 ± 7.7‰. (b) Expected ∆ 14 C of combusted fuel given mean ages ranging from 5 to 75 years. The red shading corresponds to potential mean ages that are consistent with a 1σ uncertainty on the ∆ 14 C of fire-emitted PM2.5. in 2021 of −3.2 ± 1.4‰, indicating the combusted fuels likely accumulated over a period of many decades, during a time when the atmosphere was more enriched in bomb 14 C ( figure 4(a)).
To further constrain the age range of the wildfire PM 2.5 emissions, we used the steady-state one-box ecosystem model described in section 2.3.3, in which the mean transit time of the carbon is equal to its mean age. With the uncertainty range of our observed combusted fuel ∆ 14 C (111.6 ± 7.7‰), we attained a best-fit for fuel age of 40 years, with an asymmetrical uncertainty range of 29-57 years, corresponding to 1σ uncertainty of the mean ∆ 14 C ( figure 4(b)). In 2021, a fuel class with a relatively young mean age (5 years) would have a ∆ 14 C of 9.5‰, which is only slightly enriched relative to the contemporary atmosphere in 2021 ( figure 4(a)). Needles and leaf litter may be examples of materials in this fuel class. Since these materials decompose relatively quickly, very little retain the high degree of bomb labeling from the 1950s and 1960s. As the age of fuels increases from 5 to 40 years, the predicted ∆ 14 C of combusted fuel from the model rises to approximately 110.0‰. Fuels with a 40 year mean age may encompass largerdiameter woody detritus and other woody biomass like live shrubs and small trees.

Discussion
The diameter of woody fuels influences the length of time required for decomposition in forest ecosystems. Fine fuels (i.e., needles, leaf litter, and smalldiameter woody detritus) decompose relatively rapidly over several years. In contrast, fuels with a larger diameter decompose relatively slowly and can remain in the understory and on the forest floor for decades (Harmon 2021). Lower-intensity fires, including prescribed fires, typically consume only fine fuel classes because larger-diameter fuels generally have a higher moisture content and require more energy to ignite and support high-intensity fires (Chuvieco et al 2002, Gorte 2009. High rates of energy release from the consumption of large-diameter fuels, in turn, can contribute to longer flame lengths and increase probability that ground fires will jump into the overstory and develop into crown fires (Stephens et al 2022). High-intensity crown fires are a major concern with respect to preservation of giant sequoias because thick bark near the base of tree makes sequoias nearly impervious to ground fires, yet scorch in the upper canopy can damage foliage and more vulnerable vascular tissues (Shive et al 2022).
Our ∆ 14 C measurements provide evidence that the KNP Complex Fire burned through largerdiameter fuels, likely with a considerably higher intensity than what would be expected for surface fires. It is also possible that the combustion of deeper organic soil (Pellegrini et al 2021), which has a similar mean age to the woody detritus in some forests (Mouteva et al 2015), also contributed to emissions. However, our study does not include depth of burn measurements, which would allow us to apportion the relative contributions from woody detritus and soil. Still, our measurements are broadly consistent with previous studies documenting how the accumulation of fuel in California's coniferous forests (including the buildup of shrubs, small trees, and woody detritus on multi-decadal timescales) contributes to high fire intensity ( Notably, the KNP Complex Fire led to the mortality of about 5900 giant sequoias, meaning that 3%-5% of all giant sequoias in the Sierra Nevada mountains were killed or are expected to die within 5 years (Stephenson and Brigham 2021). This compounds giant sequoia mortality from the 2020 Castle Fire, which killed an estimated 10%-14% of large giant sequoias (Stephenson and Brigham 2021). The native range of giant sequoias does not extend beyond the western slope of the Sierra Nevada, therefore a loss of approximately 15% in such a short period is a major threat to this endangered species.
Fuel treatments, including mechanical thinning and prescribed fire, have historically been used to decrease fuel loads in forests. Legislation was recently passed in California that will expand the use of prescribed fire across the state with a goal of one million acres treated per year by 2025 (California Wildfire and Forest Resilience Task Force 2022). If they are successful, we expect fuel treatments to gradually reduce the mean age of combusted fuels because understory vegetation and larger-diameter woody detritus should be removed (Harmon 2021). Regional PM sampling during periods with prescribed fire might be an effective way to monitor the success of this program. Specifically, after a sustained effort, the ∆ 14 C of combusted fuel should more closely match the contemporary atmosphere because finer fuel classes have younger mean ages. Eventually, wildfires would also be expected to have lower intensities and more closely track the ∆ 14 C of atmospheric CO 2 .
Apart from evaluating the effectiveness of prescribed burning, PM ∆ 14 C monitoring in California may be helpful for understanding the role of wildfire in influencing air quality across the state. For instance, the San Joaquin Valley of California is a serious nonattainment area for PM 2.5 and other pollutants with adverse health effects. It often exceeds both state and national air quality standards for harmful pollutants (Huang et al 2021). Pollution in the San Joaquin Valley is primarily anthropogenic due to both local emissions and those transported from surrounding urban areas, but wildfires play a variable and often significant role in elevated pollutant levels (Schweizer and Cisneros 2017, Burke et al 2021, Frausto-Vicencio et al 2023. Disentangling contributions to observed PM 2.5 levels from prescribed fire, wildfire, and urban sources will be critical for creating effective policy to improve air quality in the San Joaquin Valley and simultaneously meet the State's forest management goals. Future directions for this work include building a longer time series of observations from multiple fires, which will provide information about a broader range of burning conditions. The length of our time series was constrained by rapidly changing conditions and challenges securing a safe sampling location that was both near the smoke and outside of the mandatory evacuation zone. Future analyses should be conducted over longer sampling intervals on a variety of wildland fire types, including grass and shrub fires and prescribed fire, to more broadly understand the effects of climate, fuel, and fuel treatments on the composition of PM emissions across California and other fire-prone forests. Fire emissions contain many carbonaceous components (Olsen et al 2020), so another important research direction is to simultaneously measure the ∆ 14 C of elemental and organic carbon (EC and OC) PM 2.5 fractions and carboncontaining trace gases (CO and CO 2 ) to understand better the linkages between fuel age (and type), flaming and smoldering combustion, and the composition of fire emissions. In past work, for example, fireemitted EC has been shown to have higher levels of ∆ 14 C than OC, likely due to a different mixture of fuels (Mouteva et al 2015).

Conclusions
In this study, we measured the ∆ 14 C of PM 2.5 from the KNP Complex Fire and used these observations to infer that fuel buildup over multiple decades was a dominant contributor to PM 2.5 emissions in smoke. Our analysis is consistent with past work showing that cessation of Indigenous burning practices and implementation of fire suppression in the Sierra Nevada mountains are important contributors to recent increases in fire intensity. We also propose that our measurement techniques can be used to assess the efficacy of prescribed fire and other fuel treatments planned for California in the near future and to identify fire impacts on air quality in remote urban areas. Altogether, fuel management and an enhanced understanding of emissions associated with California's wildfires can help mitigate their social, economic, and ecosystem impacts.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https:// doi.org/10.7280/D1498P (Odwuor 2023).