The Bruker Optics EM27/SUN solar-viewing Fourier transform spectrometer, owned by Los Alamos National Laboratory (LANL), collected continuous daytime column measurements in Farmersville, California (36.31, -119.19), from 8 September until 17 October 2020, for a total of 40 d of observations. The EM27/SUN Xgas values were retrieved from unaveraged, double-sided interferograms using the interferograms to spectra (I2S) and GFIT (GGG2014 version; https://tccon-wiki.caltech.edu/, last access: 15 September 2022) retrieval algorithms automated by the EM27/SUN GGG interferogram (EGI) processing suite (Hedelius et al., 2016). Surface pressure is required to retrieve dry air columns in GGG, and we used a Coastal Environmental Systems ZENO Weather Station to record surface pressure at our field site for retrievals. Retrievals also require atmospheric profiles of temperature, pressure, altitude, and water, and these profiles were extracted from the NCEP/NCAR (National Centers for Environmental Prediction/National Center for Atmospheric Research) reanalysis product (Kalnay et al., 1996). We calibrated the EM27/SUN via co-located measurements alongside the IFS125, a high spectral resolution FTIR operated by TCCON at the California Institute of Technology (CIT), both before and after the collection periods, to determine calibration factors (Rgas), assuming a linear model forced through the origin for each gas (e.g., XTCCON=XEM27Rgas; Chen et al., 2016; Hedelius et al., 2016). The TCCON network sets the standard as the current state-of-the-art, ground-based validation system for remote sensing and satellite-based observations of greenhouse gases (Wunch et al., 2011), and TCCON observations are tied to the World Meteorological Organization (WMO) standard greenhouse gas scale. Co-locating the EM27/SUN and TCCON instruments ensures system stability of the EM27/SUN after transportation to field sites. Co-located measurements were performed on 2–3 September 2020 and 30 October–1 November 2020. Results of the correction factors from the co-located measurements are shown in Table A1. The TCCON instrument also uses the GFIT retrieval algorithm with the same a priori profiles; however, due to different instrument spectral resolutions and averaging kernels, we correct for the differences between the EM27/SUN and TCCON instrument, following Hedelius et al. (2016; Eq. A4), to adjust the EM27/SUN retrievals before comparing with TCCON and deriving calibration factors.

The EM27/SUN solar spectrometer has been previously used to study emissions from urban and agriculture CH4 and CO2 sources (Chen et al., 2016; Viatte et al., 2017; Dietrich et al., 2021; Heerah et al., 2021; Makarova et al., 2021; Alberti et al., 2022a). The recent addition of a CO detector in Bruker's EM27/SUN FTIR spectrometer increases the instrument's utility for measuring combustion sources and as a validation tool for TROPOMI column XCO, as it covers the same spectral region (Hase et al., 2016). The EM27/SUN uses the Sun as the light source which allows it to derive AOD, as demonstrated by Barreto et al. (2020) at the TCCON FTIR and AERONET site at the Izaña Atmospheric Observatory, Spain. In their study, TCCON spectra were degraded to the same resolution as the EM27/SUN (0.5 cm-1), and it was concluded that EM27/SUN spectra would be able to effectively derive AOD. Following their approach, we derive AOD for the wildfire period from our measurements. Further details of the AOD calculation are found in Appendix B.

Prior to measurements in California, the EM27/SUN was stationed in Fairbanks, Alaska, for several months. Given the different settings used with the CAMTRACKER, the solar disk was not centered on the camera, and this misalignment was found on 7 September. Based on co-located measurements with the CIT TCCON on 2 and 3 September, it was determined that the observations within the second detector of XCO were affected on the days prior when the camera was misaligned (2, 3, 6, and 7 September). For this reason, we report measurements of XCO, XCO2, and XCH4 beginning on 8 September and use the 30 October–1 November co-located measurements to calculate correction factors for all gases. AOD was derived from micro-windows within the first detector; thus, calculations of AOD were not affected.

TROPOMI is an instrument launched in late 2017 on board the European Space Agency's (ESA) Sentinel-5 Precursor (S5P). The instrument measures Earth radiance spectra in the ultraviolet, near-infrared, and shortwave infrared, allowing for measurements of a wide range of atmospheric trace gases and aerosol properties (Veefkind et al., 2012). The satellite has a sun-synchronous orbit, with daily global coverage and a spatial resolution of 5.5×7 km2 for CH4 and CO operational level 2 (L2) products. The offline (OFFL) CO total column L2 data product filtered for quality assurance values >0.5 are used in this work, as recommended in the product readme file (https://sentinel.esa.int/documents/247904/3541451/ Sentinel-5P-Carbon-Monoxide-Level-2-Product-Readme-File, last access: 4 August 2022). This selection filters out high solar zenith angles, any corrupted retrievals, and influences from high clouds. The majority of the TROPOMI XCH4 product was flagged out near the observational site during our measurement period and hence was not included in this analysis. Following Sha et al. (2021), the TROPOMI CO column densities were converted to XCO (parts per billion – ppb) by using the modeled surface pressure and total column of water to calculate the column of dry air.

There is growing interest in using the TROPOMI XCO product for understanding global wildfire fluxes; however, few studies focus on evaluating those observations (e.g., Jacobs, 2021; Rowe et al., 2021). We measured a range of XCO levels of mixed smoke plumes with our EM27/SUN and were able to isolate a concentrated smoke plume from a nearby fire. This allowed for a ground-based evaluation of the TROPOMI sensor under various wildfire conditions, including high XCO and aerosol loading in the atmosphere. A correction factor was calculated for the EM27/SUN to account for differences in the a priori profile used in the retrieval of XCO in both instruments. We follow the a priori substitution method described in Jacobs (2021) and Sha et al. (2021) to calculate an additive factor for the EM27/SUN. Due to the possibility of measuring narrow smoke plumes on subgrid spatiotemporal scales, we perform a sensitivity study to determine the best co-location criteria for the EM27/SUN to TROPOMI comparison by varying the maximum radius (5–50 km) from the observational site and averaging time (5–30 min) for the EM27/SUN measurements around the TROPOMI overpass time. We required a minimum threshold of at least three 1 min averages of EM27/SUN retrievals within the averaging time aggregations.

AERONET (https://aeronet.gsfc.nasa.gov/, last access: 15 June 2022) is a global network of Sun/sky radiometers, with over 600 sites operated around the globe. AERONET observations include measurements of AOD, microphysical, and radiative properties. The stations are frequently calibrated, and they set the standard for aerosol measurements and validation for satellite products (Giles et al., 2019). AERONET measures AOD at several spectral windows from 340, 380, 440, 500, 675, 870, 940, 1020, and 1640 nm. The Ångström exponent (AE), describing the wavelength dependence of aerosol optical thickness, is calculated from the spectral AOD. We used the AERONET Level 2.0 Version 3 AOD and AE data from the Fresno_2 site (36.78; -119.77) that has been operating in the same location since 2012. This site is located about 90 km away from our EM27/SUN measurement site. Further quality control information can be found in Giles et al. (2019).

We demonstrate the capability of ground-based solar column measurements to calculate important variables for fire research, including EFs and MCE, for determining fire emissions and understanding different combustion phases of wildfires. As a case study, 12 September observations were selected, as this day had the highest observed XCO and a dominant influence from the SQF Complex (Fig. 1b). We estimate emission ratios of CH4 and CO (ERi) by calculating the slope from a York linear regression of CO and CH4 excess mole fractions (ΔXi) relative to excess mole fraction of CO2 (ΔCO2), as shown in Eq. (1). The York linear regression considers the instrument errors along both axes. 1ERi=ΔXiΔXCO2=Xi,Fire-Xi,BkgdXCO2,Fire-XCO2,Bkgd

Emission factors of CH4 and CO (EFi) were then calculated, as shown in Eq. (2), by multiplying the ERi by the molar mass of either CO or CH4 (MMi), divided by the molar mass of carbon (MMC) and total carbon emitted (CT), while assuming 500±50 g of carbon is emitted per kilogram of dry biomass consumed (MBiomass; Burling et al., 2010; Akagi et al., 2011). CT is given by Eq. (3), where n is the number of carbon-containing species measured, Nj is the number of carbon atoms in species j, and ΔXj is the excess mixing ratio of species j (Yokelson et al., 1999). 2EFi=ERiCT⋅MMiMMC⋅MBiomass3CT=∑j=1nNj×ΔXjΔXCO2 The MCE is commonly used as a relative measure between the smoldering and flaming combustion phases. Smoldering emissions have an MCE from 0.65–0.85, pure flaming emissions have an MCE of 0.99, and emissions near 0.9 have roughly equal amounts of flaming and smoldering combustion (Akagi et al., 2011). MCE was calculated by dividing the excess mole fraction of CO2 by the sum of the excess mole fractions of CO and CO2, as follows: 4MCE=ΔXCO2ΔXCO+ΔXCO2.

Due the difference in averaging kernels across the trace gases, an averaging kernel correction is applied to Eqs. (1) and (4) (see Appendix C). The enhancement over background mixing ratios (ΔXi) for each measurement day was calculated by subtracting the background (Xi,Bkgd) determined as the 2nd percentile of the daily measured mixing ratios (Xi). A sensitivity test showed that emission ratios did not change significantly if the background was calculated using the 1st–5th percentiles. Leveraging the comparison between our ground-based instrument and TROPOMI, we compared the spatial background to show that the 2nd percentile was appropriate (Fig. S2 in the Supplement). The monthly background in September was 411.3 ppm (parts per million) for XCO2, 99.4 ppb for XCO, and 1905.3 ppb for XCH4. The monthly average mixing ratios measured in situ at Mauna Loa for CO2 were 411.5±0.2 ppm and CH4 1884.7±1 ppb during September 2020 (https://gml.noaa.gov/obop/mlo/, last access: 15 April 2022). Data collected during this period from TCCON sites located in southern California (CIT and NASA Armstrong) were explored as background sites; however, during this period, XCO was elevated due to local wildfires in those areas and thus not appropriate to use during this time.

The first week of trace gas measurements is shown in Fig. 2, in addition to the daytime fire radiative power (FRP), an indicator of fire intensity measured by the Visible Infrared Imaging Radiometer Suite (VIIRS) active fire and thermal anomalies product from NOAA-20. Fire-emitted CO can be observed in the time series, and XCO is exceptionally high on 12 September, reaching mixing ratios 10 times higher than the previous days. A regional smoke plume was captured by the NOAA-20 VIIRS satellite on 12 September that originated from the SQF Complex and traveled west directly over the measurement site, as seen in Fig. 1b. The measurement on 12 September also corresponds to the highest FRP during this record. On the next day, 13 September, both fires remained active; however, their smoke plumes were transported northward, as reflected by a lower XCO in our observations relative to 12 September.

XCO2 and XCH4 were also enhanced on the 12 September smoke event and followed the same trend as XCO over the course of the day. Over 30 dairy farms are located northwest of the measurement site, and they are expected to influence observed XCH4 and XCO2; moreover, dairy influence is notable on days with predominantly westerly winds (e.g., 8 and 11 September). XCO, XCO2, and XCH4 averaged at 154±78 ppb, 413±1 ppm, and 1938±27 ppb from 8 September to 17 October. XCO and XCO2 peaked on 12 September at 1012.8 ppb and 421.6 ppm, while XCH4 peaked on 28 September at 2050.1 ppb due to the dairy farms in the area. The measured XCO on 12 September is the highest reported XCO value in the EM27/SUN literature. Retrievals of Xgas, using the EM27/SUN in such dense smoke plumes, has not been reported in previous studies. Using this date as a case study, we calculate total column EFs and MCE (further described in Sect. 3.4). We isolate the 12 September fire smoke plume by taking the XCO mixing ratios that exceeded the 98th percentile (>335.1 ppb) from all observations over our measurement period. This period corresponded to mixing ratios recorded after 12:00 PDT, when XCO and XCO2 began to increase considerably.

The time since the emission of the observed smoke plumes was estimated to be ∼1.5 h. This was calculated by dividing the distance away from the SQF Complex fire (∼60 km) by the average wind speed (11.2±0.8 m s-1) at the height of the smoke plume (4.1±1.2 km). The height of the plume was determined by taking a mean of the available pixels within the smoke plume of aerosol layer height product from TROPOMI (http://www.tropomi.eu/data-products/aerosol-layer-height, last access: 15 July 2022). The mean wind speed measured at 4.1±1.2 km came from a 915 MHz wind profiler located in Visalia, CA, about 20 km west of the observational site (data available at ftp://ftp1.psl.noaa.gov/psd2/data/realtime/Radar915/, last access: 15 July 2022).

We show a time series of AOD at 500 nm derived for the first week of measurements in Fig. 2e (8–15 September), plotted with AOD at 500 nm from an AERONET station in Fresno, located about 90 km north of the measurement site (Fig. 1). Similar to observations of XCO, enhancements of AOD are observed throughout the week, with the highest recorded AOD on 12 September. The observational sites were relatively far from each other (∼90 km), and although smoke reaching the two sites varied over these spatial scales, the FTIR AOD follows the same inter-day trend as the AOD measured by the AERONET, with a peak in AOD on 12 September. Intraday variability between the sites does not seem to follow the same trend. This suggests that the EM27/SUN AOD estimate was also able to qualitatively capture the increase in aerosols in the SJV as fires burned more intensely and smoke from fires moved into the valley due to synoptic conditions. A comparison between the FTIR and AERONET hourly AOD can be found in Fig. S3, where we find a slope of 1.4±0.3 and R2 of 0.39. Differences are observed in the AOD time series, as these two sites were downwind of two different fires in the Sierra Nevada. The Creek Fire was located directly east of Fresno, and the SQF Complex was located directly east of the EM27/SUN measurement site. This may be the reason why the peaks observed at the FTIR site are not seen in the Fresno AERONET data. Ahangar et al. (2022) determined that the SJV air quality was mainly impacted during the 8–15 September period, with the Creek and SQF Complex fires responsible for the majority of the smoke within SJV. Although the Creek Fire began on 5 September, the air quality began to deteriorate a few days after, possibly due to the westerly downslope winds that pushed the smoke east of the Sierra Nevada at the beginning of the fire (Cho et al., 2022). Low AOD from AERONET was observed prior to 8 September, with values of 0.50±0.28, illustrating the air quality was cleaner and deteriorated after the activity from the Creek and SQF Complex fires increased (Ahangar et al., 2022).

Figure 3 shows XCO plotted against simultaneously collected AOD at 500 nm for 1 min intervals. The points are colored to distinguish the different measurement days from 8 to 15 September. The error bars are the uncertainty in AOD (further described in Appendix B), and for XCO, it is 1 standard deviation (SD) on the mean. Due to the rapidly changing XCO as the fire plume traversed over the instrument, we use the standard deviation from the 1 min XCO mean as the natural variability or uncertainty, which is larger than the XCO instrument error. Table 1 shows the slope and intercepts, with standard errors from a York linear regression fit that considers errors in x and y. We find strong relationships (R2>0.61) between the EM27/SUN XCO and AOD at 500 nm, with slopes ranging from 29.01 to 92.41 ppb/AOD (Table 1 and Fig. 3). Several studies have also found a strong correlation between CO and AOD at 500 and 550 nm from fire events and downwind of polluted sources (Lobert, 2002; Paton-Walsh et al., 2005; Kampe and Sokolik, 2007; McMillan et al., 2008). McMillan et al. (2008) found mean slopes of fire plume observations from the Atmospheric Infrared Radiation Sounder (AIRS) CO and Moderate Resolution Imaging Spectroradiometer (MODIS) AOD that ranged from 40 to 74 ppb/AOD and over clean regions slopes averaged at 35 ppb/AOD. Most of the days in our observations have slopes that fall within these ranges, and the days with lower slopes (10 and 11 September) follow a similar linear trend (gray line in Fig. 3), as in McMillan et al. (2008), over a clean region in Alaska and Canada. Kampe et al. (2007) found that AOD to CO ratios varied strongly, and this variation may depend on age of smoke plume, distance from source, combustion efficiency, and local meteorological factors. Our measurements were sensitive to nearby smoke plumes in addition to mixed smoke from distant fires. The intercepts of the fitted lines reflect different local backgrounds of CO during measurement periods, with 10–12 September having the largest backgrounds of XCO. We find that the AOD on 12 September reached values above 15, indicating extremely high aerosol loading from the smoke plume event transported from the SQF Complex in the Sierra Nevada.

In this section, we compare XCO retrieved from ground-based EM27/SUN observations downwind of the Sierra Nevada wildfires to satellite-based XCO retrievals from coincident TROPOMI overpasses. Previous studies of XCO and XCH4 comparisons between TROPOMI and EM27/SUN have used TROPOMI soundings between 50 and 100 km from the observational site and used EM27/SUN measurements between 40 min and 1 h of the TROPOMI overpass as coincident criteria (Jacobs, 2021; Sha et al., 2021; Alberti et al., 2022b; Sagar et al., 2022). Given the spatial and temporal heterogeneity in smoke plumes from wildfires observed in Figs. 1 and 2, we perform a sensitivity study of different radii (10, 15, 20, 30, 40, and 50 km) from our observational site and time averages (10, 15, 20, and 30 min) to determine adequate criteria for comparison during a wildfire event. An illustration of the sensitivity analysis is shown in Fig. D1 in Appendix D.

We quantify the sensitivity of different TROPOMI radii and averaging times in comparison with our EM27/SUN observations by calculating the mean difference, mean relative difference, and R2 between the linear regression fits for the measurements. We find that all combinations produce a positive mean bias, meaning that TROPOMI overestimates XCO compared to the EM27/SUN measurements. TROPOMI pixels within a radius of 5 km averaged with 30 min aggregations of EM27/SUN give the lowest mean difference of 10.64 ppb, a mean relative difference of 5.5 %, and the highest correlation coefficient of 0.99; however, only four points coincide during the measurement period. To maximize the number of coincidences while maintaining a low bias, we select 15 km as the maximum radius with a 30 min averaging time. This gives a total of 19 coincident data points and a mean difference of 17.2 ppb, a mean relative difference of 9.7±1.3 %, and R2 of 0.97. A time series of the coinciding data pairs from the EM27/SUN 30 min average observation period with TROPOMI overpass with 15 km radii are shown in Fig. 4a, and the comparison is shown in Fig. 4b. Applying these spatial and temporal criteria results in large variance for the largest measured XCO due to heterogeneity in the smoke plume event. The EM27/SUN displays a larger variance than TROPOMI due to capturing the 30 min temporal variability in the plume as it was transported above the instrument. We find a York linear regression fit of y=(1.35±0.01)x-39.30±0.58. The mean relative difference found in this study of 9.7±1.3 % is similar to the systematic difference of 9.22±3.45 % between TROPOMI and all TCCON stations (Sha et al., 2021). These results suggest that the differences found between the TROPOMI and EM27/SUN observations during wildfires are consistent with the systematic differences that exist between the two instruments; however, based on our sensitivity study, biases may exist based on sampling conditions in a spatially and temporally heterogenous source.

Emission ratios of CO and CH4 on 12 September were calculated with respect to CO2 (Fig. 5). ERCO was 0.1161±0.0005, and the ERCH4 was 0.00730±0.00007, resulting in an EFCO2 of 1632.9±163.3 g CO2 kg-1 biomass combusted, EFCO of 120.5±12.2 g CO kg-1 biomass combusted, and EFCH4 of 4.3±0.8 g CH4 kg-1 biomass combusted. We compared findings from our measurements to the literature values in temperate coniferous forest studies from the Sierra Nevada (Fig. 6) and other locations in North America (summarized in Table 2). All the studies listed in Table 2, except for this study, were based on aircraft measurements for temperate coniferous forests. Due to combustion-phase variability in field studies, we compare our atmospheric-column-based EFs in Fig. 6 with the most relevant studies from the Sierra Nevada, which shows our calculated values are within the expected linear range from in situ aircraft studies. The measurement uncertainties for the EFs were calculated by propagating the error from the ER linear regression standard error, CT, and 10 % error from MBiomass.

The average MCE for the smoke plume on 12 September was 0.89±0.21 (1σ), meaning that observations of the smoke plume consisted of a mixture of flaming and smoldering combustion phases (Fig. 6). During the flaming phase of a fire, CO2 is produced, and convection is created by high flame temperatures, producing the lofting of smoke. High-altitude smoke can be transported large distances, corroborated by observations of ash falling from the sky at a measurement site ∼60 km away from the fire and clearly observable by satellite imagery (Fig. 1b). In contrast to the flaming phase, smoldering fires burn at lower intensity, and incomplete combustion side products like CO, CH4, and organic carbon aerosol are produced. The MCE calculated from total column observations is averaged over the entire vertical plume, as it was being transported over the measurement site. The advantage of a plume-integrated MCE is that vegetation is burned differently throughout the fire, and the atmospheric column observations can represent the fire as a whole by integrating the smoke plume heterogeneity in the vertical atmospheric column.

The EM27/SUN's location enabled us to sample transient fire plumes from local and state wildfires but was also located near a large cluster of dairy farms, which are a large regional source of CH4 emissions (Heerah et al., 2021; Marklein et al., 2021). Dairy farms are known to emit significant amounts of CH4 from the animals' enteric fermentation and on-farm manure management. Because fires also emit CH4, we explored whether dairy and fire sources in this region can be disentangled using the enhancement ratios of the different species measured by the EM27/SUN. Enhancement ratios are also known as the normalized excess mixing ratios. Excess mixing ratios are calculated by subtracting the mixing ratio of a species from a source plume minus a mixing ratio of the same species in background air. To correct for dilution, excess mixing ratios are normalized by a stable tracer such as CO2. When an enhancement ratio does not change with dilution and mixing with background air, then the enhancement ratio is equal to the emission ratio (ER) of a source (Yokelson et al., 2013). Furthermore, our measured ΔXCH4/ΔXCO2 ratios enable us to investigate the contribution of state wildfires to CH4 emissions in 2020. To constrain the observed enhancements, we compared the enhancement ratios of ΔXCH4/ΔXCO2 from September–October 2020 to enhancement ratios collected in September 2018 and 2019 in the same local area that characterize non-fire years. September 2018 and 2019 measurements are further described in the Supplement. We focused on observation days with statistically significant correlations (n=26 d) between CH4 and CO2 enhancements (R2>0.5 and p<0.05) to characterize the enhancement ratios of the SJV non-fire years.

During September–October 2020 observations, ΔXCH4/ΔXCO2 ratios of dairy farm influence were found on several days, in addition to lower slopes indicative of combustion sources (Fig. 7; gray markers). The 12 September smoke plume event is highlighted in Fig. 7 (red markers) and has a smaller emission ratio of 7.3±0.07 (ppb/ppm) compared to larger ΔXCH4/ΔXCO2 enhancement ratios of 38.4±21.7 and 30.5±5.0 (ppb/ppm) observed in September 2018 and 2019. Similar non-wildfire ratios of ΔXCH4/ΔXCO2 were found in Hanford, ∼40 km west of our observation site, from an aircraft study ranging from 35.9–44.4 (ppb/ppm) during a winter campaign (Herrera et al., 2021). Other column-based studies have determined the XCH4/XCO2 for urban sources in the Los Angeles city, finding ratios ranging from 6.65 to 9.96 (ppb/ppm) in 2015 (Chen et al., 2016) and 11±2 ‰ in 2008 (Wunch et al., 2009). Wunch et al. (2009) determined that urban fossil fuel and wildfire XCH4/XCO2 ratios are very similar due to incomplete combustion, and the ratios are not distinct enough to separate. In the vicinity of the measurement site in the SJV, there is a strong influence of dairy farm agriculture and minimal urban emissions away from population centers; thus, we are able to separate ΔXCH4/ΔXCO2 from dairy sources, from fire, or from possible urban emissions. The CH4/CO2 enhancement ratios observed in this area make it evident that dairy farm operations are the dominant source of CH4 during fire and non-fire days. Nevertheless, CH4 enhancements during the strong smoke events greatly exceeded CH4 enhancements from local dairy sources on hourly timescales.

The immense scale of the 2020 wildfires meant that they released a significant amount of CO2 emissions, equivalent to about 36 % of the state's CO2 budget for the year (CARB, 2022a). Our observations of XCH4 suggest that the wildfires may also have had a significant effect on the state's CH4 budget. Given the importance of reducing CH4 emissions for meeting California's climate goals, we calculate the amount of CH4 released from wildfires that burned in the state in 2020 by using estimates of CO2 emissions from the state's wildfire inventory, the ERCH4 calculated from our study, and from the literature values.

The California Air Resources Board (CARB) reported that a total of 106.7 Tg of CO2 was emitted from 2020 wildfires and reported individual CO2 emission estimates for the 20 largest wildfires of 2020. Using the reported CO2 emission estimates from individual wildfires, we derived CH4 emissions by multiplying the CO2 estimates with emission ratios of CH4 (ERCH4) from wildfire smoke and molecular mass ratios as follows: 5ECH4=ERCH4×MCH4MCO2ECO2, where ECH4 is the emissions of CH4 in Gg yr-1, ERCH4 is the emission ratio of CH4 with respect to CO2 (in mol mol-1), MCH4 is the molar mass of CH4, MCO2 is the molar mass of CO2, and ECO2 are the individual wildfire emissions of CO2 in Gg yr-1. Emission ratios from fires are dependent on vegetation type; fires in California fell into temperate forest, shrubland or grassland vegetation types (Xu et al., 2022). Based on the generic vegetation classification from the Fire INventory from NCAR (FINN) model (https://www.acom.ucar.edu/Data/fire/, last access: 15 July 2022), we classify the top 20 California wildfires of 2020 into the three types, based on the dominant vegetation of temperate forest, shrublands, or grasslands. We assign an ERCH4 for each general vegetation type based on the mean EFCH4 found in Xu et al. (2022) that summarized EFs from Prichard et al. (2020). The standard deviation of the EFCH4 was calculated based on Prichard et al. (2020) and taken as the uncertainty that was then propagated in the ERCH4 calculations. For the Sierra Nevada wildfires (Creek Fire, SQF Complex, and North Complex), we derive ERCH4_avg by first calculating an average EFCH4_avg from the Sierra-Nevada-specific EFCH4 in Table 1 (EFCH4_avg=5.6±1.5 g kg-1). We then used Eq. (2) to solve for ERCH4 with CT equal to 1 (ERCH4_avg=0.0084±0.0022). A summary of ERCH4 can be found in Table E1 (Appendix E). Methane emissions for the top 20 wildfires were then calculated, using Eq. (5) with CARB's CO2 estimate from each individual fire, and summed to obtain a total CH4 emitted from these reported wildfires. Figure 8 shows the estimated CH4 emissions from the top 20 wildfires of 2020 compared to CARB's 2020 anthropogenic CH4 inventory emissions (CARB, 2022b). The error bar from this estimate was calculated by propagating the general vegetation ERCH4 error from Table E1 into each individual wildfire CH4 estimate and adding errors in quadrature to obtain a total error. The 20 largest wildfires of 2020 emitted 92 % of the total CO2 emissions released from wildfires in that year and emitted 213.7±49.8 Gg CH4 or 13.7 % of total anthropogenic CH4 emissions.