Comparison between Spatially Resolved Airborne Flux Measurements and Emission Inventories of Volatile Organic Compounds in Los Angeles

Los Angeles is a major hotspot for ozone and particulate matter air pollution in the United States. Ozone and PM2.5 in this region have not improved substantially for the past decade, despite a reduction in vehicular emissions of their precursors, NOx and volatile organic compounds (VOCs). This reduction in “traditional” sources has made the current emission mixture of air pollutant precursors more uncertain. To map and quantify emissions of a wide range of VOCs in this urban area, we performed airborne eddy covariance measurements with wavelet analysis. VOC fluxes measured include tracers for source categories, such as traffic, vegetation, and volatile chemical products (VCPs). Mass fluxes were dominated by oxygenated VOCs, with ethanol contributing ∼29% of the total. In terms of OH reactivity and aerosol formation potential, terpenoids contributed more than half. Observed fluxes were compared with two commonly used emission inventories: the California Air Resources Board inventory and the combination of the Biogenic Emission Inventory System with the Fuel-based Inventory of Vehicle Emissions combined with Volatile Chemical Products (FIVE-VCP). The comparison shows mismatches regarding the amount, spatial distribution, and weekend effects of observed VOC emissions with the inventories. The agreement was best for typical transportation related VOCs, while discrepancies were larger for biogenic and VCP-related VOCs.


Text S1: Description of FIVE-VCP Inventory
Mobile source emissions (on-road and off-road engines using gasoline or diesel) are taken from the Fuel-based Inventory of Vehicle Emissions or FIVE (1,2).Briefly, gasoline and diesel fuel sales for on-road and off-road engines are reported by state each year.Monthly regional fuel sales and monthly state-level traffic estimates are used to update this with monthly adjustments to fuel sales at a state-level for on-road and a regional level for off-road (3).For on-road engines, fuel sales are downscaled to the roadway level using light and heavy-duty vehicle count data where possible (~70% of gasoline and ~80% of diesel nationally) with the remaining portion of fuel sales being distributed spatially using population density and the result mapped        Figure S8.Fluxes of further VOCs that were reported to increase relative to population density in a previous study (18), and acrolein as a potential frying/cooking marker (19), grouped as a function of population density.Population density is given in people per km².The left y axis corresponds to the measured data, the right y axis to the two inventories (note significant scale differences between the two axes).
Supplementary Table 3. Mean to median ratio (sorted from highest to lowest) for mass fluxes of some important VOCs that are part of at least one of the inventories.RCOOH: Higher organic acids, ARO1: Other aromatics with kOH < 2x10 4 ppm -1 min -1 .VOCs with strong localized point sources are expected to have higher mean to median ratios than VOCs that are emitted consistently throughout the study area.
onto a 4x4 km grid for the contiguous United States(McDonald et al., 2014).With the mapped fuel, co-emitted air pollutant emissions are estimated using fuel-based emission factors (e.g, g/kg fuel) for gasoline and diesel engines from roadside measurements (4, 5, 2).Day-of-week and diurnal adjustments are made separately for light-duty gasoline and heavy-duty diesel vehicles to estimate hourly emissions(1).Similarly, off-road fuel sales are spatially and temporally (with hourly, day-of-week and seasonal adjustments incorporated) allocated according to the NEI 2017(6).Fuel-based emission factors are again used to calculate hourly emissions(2).Volatile Chemical Product (VCP) emissions are estimated following McDonald et al.(7), where a mass balance of the petrochemical chemical industry was performed to estimate per capita use of adhesives, coatings, cleaning agents, inks, pesticides and personal care products.Emission factors are applied to these categories, as described inCoggon et al. (2021), to produce a 4x4km gridded inventory for the US.Oil and gas emissions are taken from the Fuel-based Oil and Gas Inventory (FOG,(8)).Other area and point source emissions make up only a small fraction of anthropogenic VOC emissions(9) and are taken from the NEI 2017(6).Emissions from the NEI 2017 are adjusted monthly using relevant economic and energy datasets for the U.S. (He et al., in preparation).The combination of these inventories is referred to here as the as the FIVE-VCP inventory.

Figure S1 .
Figure S1.Maps showing flux footprints, regions defined for the analysis, and fluxes for four example

Figure S2 .
Figure S2.Comparison of median flux observations from this airborne study (blue) with previous stationary tower urban flux observations.Error bars show the 25 th to 75 th percentiles.C2-benzenes include xylenes

Figure S3 .
Figure S3.Relevance of the discrepancies between measurements and inventories for the total molar flux

Figure S4 .
Figure S4.Comparison of mean values between measured and inventory emissions of individual VOCs for

Figure S6 .
Figure S6.Weekday and weekend fluxes shown for a selection of VOCs in comparison between

Figure S7 .
Figure S7.Fluxes of relevant indoor-to-outdoor emitted VOCs (Arata et al., in prep.)grouped as a function

Table 4 .
Results of the linear regression (median flux vs. median population density) for the VOCs shown in Fig.S7and S8 for the four population bins.The ratio of the regression slope/uncertainty shows that all listed VOC flux medians show a significant increase with population density (within the 1σ uncertainty) except the monoterpenes, whose slope is therefore shown in brackets.The slopes also provide estimates of the VOC emission per person and hour.It must be noted that the original data from which the medians are calculated have such a large scatter that those data do not show a significant increase with population density.