Seasonal Variability in Local Carbon Dioxide Combustion Sources over the Central and Eastern US using Airborne In-Situ Enhancement Ratios

We present observations of local enhancements in carbon dioxide (CO2) from local emissions sources over three eastern US regions during four deployments of the Atmospheric Carbon Transport-America (ACT-America) campaign between summer 2016 and spring 2018. Local CO2 emissions were characterized by carbon monoxide (CO) to CO2 enhancement ratios (i.e. ΔCO/ΔCO2) in airmass mixing observed during aircraft transects within the atmospheric boundary layer. By analyzing regional-scale variability of CO2 enhancements as a function ofΔCO/ΔCO2 enhancement ratios, observed relative contributions to CO2 emissions were contrasted between different combustion regimes across regions and seasons. Ninety percent of observed summer combustion in all regions was attributed to high efficiency fossil fuel (FF) combustion (ΔCO/ΔCO2 < 0.5%). In other seasons, regional contributions increased from less efficient forms of FF combustion (ΔCO/ΔCO2 0.5-2%) to as much as 60% of observed combustion. CO2 emission contributions attributed to biomass burning (BB) (ΔCO/ΔCO2 > 4%) were negligible during summer and fall in all regions, but climbed to 10-12% of observed combustion in the South during winter and spring. Vulcan v3 CO2 2015 emission analysis showed increases in residential and commercial sectors seasonally matching increases in less efficient FF combustion, but could not explain regional trends. WRF-Chem modeling, driven by CarbonTracker CO2 fire emissions, matched observed winter and spring BB contributions, but conflictingly predicted similar levels of BB during fall. Satellite fire data from MODIS and VIIRS suggested higher spatial resolution fire data might improve modeled BB emissions.


Introduction 47
Carbon dioxide (CO2) is a direct product of fossil fuel combustion and a relatively inert compound 48 in the atmosphere making it a good tracer of anthropogenic emissions, which collectively have a 49 strong influence on regional air quality and global climate. Accurately quantifying the 50 accumulation of atmospheric CO2 from its broad variety of sources is critical to predicting future 51 trends in global temperature and climate. Models utilize emissions inventories of CO2, combined 52 with ocean and land biosphere models, to predict its transport and accumulation in the atmosphere. 53 Thus, the proper apportionment and quantification of emission sources is important in order for 54 models to predict future behavior. For CO2, combustion is one of the primary anthropogenic 55 sources, but sources range widely in terms of both spatial distribution and emission type (Gurney  Colored boxes denote the regions defined in this study. The border between the Midwest and South regions was 37°N, while the border between the Mid-Atlantic region and the other two was a line drawn between 45°N, 89°W and 32°N, 82°W.

In-Situ Airborne Measurements 116
The two aircraft contained identical payloads for measuring in situ gas phase carbon species. CO2, 117 CO, and methane (CH4) dry mole fractions were measured using a commercial cavity ringdown 118 spectrometer (G2401-m, PICARRO, Inc.) with a custom gas sampling & calibration system (Fig.  119   S1). Air was sampled through a commercial stainless steel total air temperature probe modified 120 for gas sampling (Buck Research Instruments, LLC). The probe extended 12" outboard from the 121 fuselage to avoid aircraft boundary layer contamination. Sampled air was dried using a commercial 122 Confidential manuscript submitted to Journal of Geophysical Research -Atmospheres Nafion dryer tube (PD-200T-24, Permapure, Inc.) then passed through a flow controller 123 maintaining a constant 1.5 standard liter per minute (SLM) total flow. The sample was then 124 compressed to a constant ~1050 mbar, from which the spectrometer sampled at 0.5 SLM, with the 125 remaining flow exhausted to the cabin using an absolute proportional pressure relief valve set at 126 ~1065 mbar. With the exception of the Nafion dryer, all inlet materials were stainless steel to 127 minimized sample contamination through gas permeability. The spectrometer cycled between 128 measurements of each species, along with a measurement of the remaining water in the gas sample, 129 sequentially every 2.5 s. The instrument temporal through the gas system and instrument was 130 measured to be typically ~2-3 s. Calibration gas was introduced through a solenoid valve upstream 131 of the dryer at 2 SLM, with the excess flow exhausted through the inlet to ensure that only calibrant 132 was sampled. By introducing the calibrant before the Nafion dryer, the dry calibration gas was 133 humidified to the same level as the dried ambient air (typically 0.03-0.05%), thus avoiding water 134 vapor-dependent calibration discrepancies (Reum et  varied from 100-120 m/s at these altitudes, the spatial extent of the bin windows varied between 159 3-14 km. Figure S2 shows line histograms of the distribution of ΔCO/ΔCO2 observed during the 160 winter campaign. In Fig. 2a To examine the relationship between CO2 and CE, an extension of the technique is required. Thus, 168 each observed slope was binned by both ΔCO/ΔCO2 and the total ΔCO2 in the bin, the latter used 169 as a metric for the CO2 intensity of the emission. The result is a 2D heat map representing the 170 enhancement in CO2 as a function of ΔCO/ΔCO2 enhancement ratio (Fig. 2a). To calculate the 171 ΔCO2-weighted distribution with respect to ΔCO/ΔCO2, the data were summed with respect to 172 ΔCO2 for each ΔCO/ΔCO2 bin: 173 where NWF is the normalized ΔCO2-weighted bin frequency as a function of ΔCO/ΔCO2 while 175 ni,j is the number of points in the ith bin of ΔCO2 and the jth bin of ΔCO/ΔCO2.  ΔCO/ΔCO2 values less than 0% are associated with mixing with air influenced by CO2 uptake, as 235 non-photochemical CO sinks are not known to be common. In particular, these negative enhance 236 ment ratios have been hypothesized to be associated with ecosystem uptake (  Confidential manuscript submitted to Journal of Geophysical Research -Atmospheres between 0-0.5% were attributed to very high efficiency fossil fuel (FF0-0.5%) combustion, between 239 0.5-1% were attributed to high efficiency fossil fuel (FF0.5-1%) combustion, and between 1-2% were 240 attributed to low efficiency fossil fuel (FF1-2%) combustion. ΔCO/ΔCO2 values between 2-4% were 241 attributed to a very low efficiency combustion regime (FF2-4%), as they are too low to typically be 242 considered BB but are fairly high for typical FF combustion in the United States. Finally, values 243 greater than 4% were attributed to biomass burning (BB>4%) combustion. 244 CO2 uptake contributions could contribute to the positive regimes if the CO2 uptake source is also 245 a CO sink, and these cases would be indistinguishable from combustion plumes with this 246 technique. However, to the best of the authors' knowledge, such a CO sink does not exist on any 247 scale as to significantly alter our findings. Biological uptake also generally occurs on scales larger 248  Table S1 shows the numerical NWF contributions in each regime. In order to focus 259 on the relative enhancement ratios of local sources in the PBL, all data above 1 km AGL were 260 rejected in order to focus on the 300 m level altitude flight legs. NWF contributions from negative 261 ΔCO/ΔCO2 slopes peaked during the summer campaign, exhibited lesser contributions in spring 262 and fall respectively, and were almost negligible in winter. This behavior is consistent with the 263 expected levels of CO2 biogenic uptake in each season. As a result, this negative ΔCO/ΔCO2 slope 264 regime will be neglected for the reminder of the analysis. The inset in Fig. 3

Regional ΔCO/ΔCO2 variability 278
Observed enhancement ratios were also segregated into three ACT-America flight domains: 279 Mid Atlantic, Midwest, and South (Fig. 1). To show the relative effects only attributed to 280 combustion, we analyzed only the positive ΔCO/ΔCO2 slope relative NWF contributions for each 281 regime within the PBL (Fig. 4)  In the spring, Mid-Atlantic FF0-0.5% was lower than in the summer and fall, but this regime once 302 again contributed the majority of the NWF, with the F regime contributing roughly half to the 303 NWF as seen during winter. Midwestern combustion contributions exhibited behavior similar to 304 that during winter but with near zero BB>4% influence. The lack of biomass burning is likely due 305 to both warmer weather and the start of the planting season, during which comparatively little 306 agricultural burning would be likely to occur. Southern contributions were very similar to those 307 during the winter, with the biggest difference being a slight shift from FF0.5-1% and FF1-2% to 308 increased BB>4% and FF2-4% contributions. As agricultural burning in the South is typically focused In all regions, total Vulcan emissions in all regions were relatively flat with respect to season, 325 though some seasonal variability in individual regions. In the Mid-Atlantic region, total FF 326 emissions in the fall were found to be ~45% higher than in the spring and summer, while in winter, 327 emissions were 15-20% higher than in spring and summer. Midwestern total FF emissions were 328 lowest in winter and spring, ~10% higher in summer, and ~25% higher in fall. Southern total 329 emissions were at a minimum in the fall and spring, ~35% higher in winter, and ~60% higher in 330 summer. 331 By sector, electricity production and onroad vehicle emissions (Fig. 5b-c)  from the native 27 km 2 resolution pixels and for flight legs below 1 km AGL, then averaged 355 seasonally and by region (Fig. 6b). The largest modeled fire contribution in the Mid-Atlantic 356 region was during summer at ~10 mol CO2/km 2 *hr, with other seasons averaging less than 1/3 the 357 fire emissions of summer. Midwestern modeled fire average contributions were highest in spring 358 at ~7.5 mol CO2/km 2 *hr, with emissions in other seasons weaker by an order of magnitude. The 359 Southern region had the highest overall average fire emissions during the fall, winter, and spring 360 seasons, ranging from 18.5 -20 mol CO2/km 2 *hr, with a strong drop during summer to ~6 mol 361 CO2/km 2 *hr. 362 As the airborne ΔCO/ΔCO2 analysis calculated relative CO2 contributions from BB compared to 363 overall combustion, the magnitude of these emissions cannot be directly compared to the modeled 364 fire contribution. To form a better means of comparison, Fig. 6c shows the same average modeled 365 fire CO2 emissions as in Fig. 5b, but normalized by the average Vulcan modeled total FF CO2 366 emissions in each region and season in order to account for variability in overall FF emission. 367 These FF-normalized modeled fire emission ratios (Fire/FF) can then be compared with a similar 368 ratio from the airborne data (Fig. 6a), which normalizes contributions from BB>4% to those 369 attributed to FF combustion: 370 >4% ∑ = >4% 0−0.5% + 0.5−1% + 1−2% + 2−4% (2) 371 The airborne BB>4%/ΣFF ratio values were near-zero for all regions in the summer, and during fall, 372 the values only exceed 0.5% in the Midwest. Comparably high BB>4%/ΣFF ratios were observed 373 in the South during winter and spring, low ratios were observed in the Mid-Atlantic region in both 374 seasons and the Midwestern spring, whereas Midwestern winter ratios were between the two. 375 The modeled Fire/FF ratios captured the high airborne BB>4%/ΣFF ratios during winter and spring 376 in the South compared to other regions. However, there are three major discrepancies to highlight. 377 The largest discrepancy is the high Fire/FF emission ratio predicted by the model in the South 378 during fall. The modeled fall:winter emission ratio was ~105%, while the airborne fall:winter fire full spatial distribution of the fire counts and FRP. This is a much simplified approach to methods 401 described in the literature used to translate FRP to gas emissions, but the use here is to use this 402 data as tool to provide insight into the model/airborne agreement. 403 Broadly, the MODIS fire products agreed well with the WRF-Chem/CarbonTracker fire product 404 (Fig. 6b). The highest number of fire counts were in the South for all seasons, and there were many 405 fewer counts in the South during summer compared to the other seasons, both matching the 406 modeled fire emissions. One of the biggest discrepancies between MODIS and the modeled fire 407 emissions was during summer. While the modeled fire emissions were highest in the Mid-Atlantic 408 region during summer, the MODIS weighted counts were lowest in the Mid-Atlantic. Additionally, 409 the modeled fire emissions in the Midwest during winter were a factor of ~7 smaller compared to 410 those from the Mid-Atlantic region, and the two regions had comparable MODIS weighted counts. 411 The causes for this may be attributable to differences in the very simple FRP weighting approach 412 used here and the more complex analysis performed by the GFED and CASA modules. 413 Results using the VIIRS weighted counts were significantly different from MODIS. The ratio of 414 Southern spring:winter weighted counts was ~90% from MODIS compared to ~40% from VIIRS, 415 Confidential manuscript submitted to Journal of Geophysical Research -Atmospheres and the ratio of Southern fall:winter weighted counts dropped from ~115% from MODIS to ~45% 416 from VIIRS. Additionally, the ratio of winter:spring weighted counts in the Midwest increased 417 from ~55% with MODIS to ~300% with VIIRS. As two of the largest discrepancies between the 418 modeled and airborne emissions were the modeled high emissions in the South during fall and the 419 ratio of winter:spring emissions in the Midwest, these shifts provide some circumstantial evidence 420 that spatial resolution of either the satellite product or model may be contributing to those 421 discrepancies. 422

Conclusions 423
In this study, we used airborne measurements of CO and CO2 in the PBL to examine the relative 424 frequency of regional and seasonal CO2 contributions with various CE over the central and eastern imply that a combination of factors, such as undetected smaller fires below satellite product 444 resolution or insufficiently constrained biosphere data, may cause significant biases in predictions 445 of BB CO2 emissions in the US. Additionally, as air quality models use similar modules to drive 446 BB VOC and CO emissions, these same biases would likely affect predictions of regional air 447 quality as well. 448

Acknowledgments and Data 449
The Atmospheric Carbon and Transport (ACT)-America project is a NASA Earth Venture 450