1
Supplemental Information
Observations of Volatile Organic Compounds in the
Los Angeles Basin during COVID-19
Paul Van Rooy,
†
Afsara Tasnia,
†,‡
Barbara Barletta,
§
Reina Buenconsejo,
∥
John D. Crounse,
∥,b
Christopher Kenseth,
∥
Simone Meinardi,
§
Sara Murphy,
a
Harrison Parker,
a
Benjamin Schulze,
a
John H. Seinfeld,
a,b
Paul O. Wennberg,
a,b
Donald R. Blake,
§
and Kelley C. Barsanti*
,†,‡
†
Bourns College of Engineering, Center for Environmental Research and Technology, Riverside,
CA, 92507, USA
‡
Department of Chemical and Environmental Engineering, University of California, Riverside,
CA, 92507, USA
§
Department of Chemistry, University of California, Irvine, CA, 92697, USA
∥
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena,
CA, 91125, USA
a
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA,
91125, USA
b
Division of Engineering and Applied Science, California Institute of Technology, Pasadena, CA,
91125, USA
2
Index of Figures
Figure S.1
Boxplot showing the range of % NO
x
decreases in the LA Basin between 2010 and
2020, measured across 15 stations during 15 April to 15 July. ...................................................... 4
Figure S.2
Wind direction and speed recorded at 05:30-06:30 in April, May, Jun, and July during
the LAAQC-2020 campaign.. ......................................................................................................... 5
Figure S.3.
Wind direction and speed recorded at 14:00-15:00 in April, May, Jun, and July during
the LAAQC-2020 campaign.. ......................................................................................................... 6
Figure S.4.
Wind direction and speed recorded at 05:30-06:30 and 14:00-15:00 in April-July
during 2020 and 2010 ..................................................................................................................... 7
Figure S.5.
Afternoon airmass back trajectories for each week of LAAQC-2020 calculated using
Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT). ......................................... 8
Figure S.6.
Percent difference between UCI and UCR mixing ratios plotted against the percentage
of paired VOC measurements that fall within the percent difference ............................................ 9
Figure S.7.
Measured UCI vs. UCR mixing ratios for decane. .................................................... 10
Figure S.8.
Measured UCI vs. UCR mixing ratios for octane. .................................................... 11
Figure S.9.
Measured UCI vs. UCR mixing ratios for nonane. ................................................... 12
Figure S.10.
Measured UCI vs. UCR mixing ratios for 2-methylpentane. .................................. 13
Figure S.11.
Measured UCI vs. UCR mixing ratios for 2,3-dimethylpentane. ............................ 14
Figure S.12.
Measured UCI vs. UCR mixing ratios for ethylbenzene. ........................................ 15
Figure S.13.
Measured UCI vs. UCR mixing ratios for o-xylene. ............................................... 16
Figure S.14.
Measured UCI vs. UCR mixing ratios for m- and p-xylene. .................................. 17
Figure S.15.
Measured UCI vs. UCR mixing ratios for toluene. ................................................. 18
Figure S.16.
Calculated CO intercepts for 59 individual VOCs plotted against their OH rate
constant
3
for each specific VOC. .................................................................................................. 20
Figure S.17.
Hourly median (blue dots) and daily median (orange line) CO mixing ratios measured
on the Caltech campus (CITAQS) before, during, and after LAAQC. ......................................... 21
Figure S.18.
OH-exposure calculated using the ratio of benzene to 11 more-reactive VOCs. .... 24
Figure S.19.
Methods to determine the constant emission ratio of benzene:1,2,4-trimethylbenzene:
time series of the ratio of measured benzene to measured 1,2,4-trimethylbenzene (top); and
scatterplot of benzene and 1,2,4-trimethylbenzene mixing ratios (bottom). ................................ 25
Figure S.20.
Observed VOC versus CO mixing ratios for acetylene (blue), m+p-xylene (yellow),
and α-pinene (green). .................................................................................................................... 26
Figure S.21.
Nighttime, morning, and afternoon VOC/ΔCO ratios versus calculated OH exposure
for acetylene, m+p xylene, and α-pinene. .................................................................................... 27
Figure S.22.
OH- and O
3
-corrected emission ratios plotted against nighttime emission ratios to
CO (top) and acetylene (bottom) .................................................................................................. 29
Figure S.23.
Terpene mixing ratios plotted against temperature. ................................................ 30
Figure S.24.
The ratio of average June-July mixing ratio to average April-May mixing ratio for
the 40 compounds listed in Table S.3. .......................................................................................... 32
3
Figure S.25.
Boxplots showing the CO mixing ratio measured as a part of SARP (2015-2019, left)
and during LAAQC (2020, right). ................................................................................................ 33
Figure S.26.
Boxplots showing the acetylene mixing ratio measured as a part of SARP (2015-
2019, left) and during LAAQC (2020, right). ............................................................................... 34
Figure S.27.
Boxplots showing the i-pentane mixing ratio measured as a part of SARP (2015-
2019, left) and during LAAQC (2020, right). ............................................................................... 35
Figure S.28.
Boxplots showing the total non-methane hydrocarbon (NMHC) mixing ratio
measured as a part of SARP (2015-2019, left) and during LAAQC (2020, right). ...................... 36
S.1 Chemical and Meteorological Comparisons between 2010 and 2020
NO
x
data presented in Parker et al.,
1
obtained from 15 South Coast Air Quality
Management District air monitoring stations across the LA Basin, were used to calculate the
change in NO
x
levels between 2010 and 2020 during the period of 15 April to 15 July. The data
are presented in Figure S.1. On average, NO
x
levels decreased by 49% between 2010 and 2020.
In addition, NO
x
data from the California Air Resources Board monitoring station in Pasadena
(South Wilson) were used to calculate the change in NO
x
levels between 2010 and 2020; an
approximate 50% decrease in NO
x
was calculated using these data.
Wind speed and direction during daytime and nighttime sampling periods in April, May,
June, and July 2020 and daytime sampling periods during April-July in 2010 and 2020 are shown
in Figures S.2-S.4. Morning wind was generally stagnant (0-3 m/s), while afternoon wind was 3-
5 m/s from the southwest. Wind speed in 2020 is qualitatively similar to 2010, particularly
during the afternoon sampling periods when wind speeds were between 3-5 m/s, suggesting
transport times and trajectories were similar. Afternoon airmass back trajectories (Figure S.5)
starting at 14:00 local time were calculated using the Hybrid Single-Particle Lagrangian
Integrated Trajectory (HYSPLIT) model
2
at an altitude of 250 m on Wednesday for each week of
the campaign. Meteorological data at a resolution of 12 km x 12 km were obtained from the
4
North American Mesoscale Forecast System (NAMS) archive
(https://ready.arl.noaa.gov/archives.php).
Figure S.1.
Boxplot showing the range of % NO
x
decreases in the LA Basin between 2010 and
2020, measured across 15 stations during 15 April to 15 July. The median is represented by the
red line, the 25th and 75th percentiles by the edges of the blue box, and the mean by a black '
'.
5
Figure S.2.
Wind direction and speed recorded at 05:30-06
:
30 in April, May, Jun, and July during
the LAAQC-2020 campaign. Wind data
we
re from the NOAA integrated surface database.
6
Figure S.3.
Wind direction and speed recorded at 14:00-15:00 in April, May, Jun, and July
during the LAAQC-2020 campaign. Wind data
we
re from the NOAA integrated surface
database.
7
Figure S.4.
Wind direction and speed recorded at 05:30-06:30 and 14:00-15:00 in April-July
during 2020 and 2010. Wind data were from the NOAA integrated surface database.
8
Figure S.5.
Afternoon airmass back trajectories for each week of LAAQC-2020 calculated using
Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT). Symbols are used to
approximate sampling area and source regions: Pasadena (red square), Los Angeles (yellow
triangle), Torrance (green triangle), and Long Beach (blue triangle).
S.2 Comparison of VOC mixing ratios quantified by UCI and UCR
During LAAQC nine compounds (decane, octane, nonane, 2-methylpentane, 2,3-
dimethylpentane, ethylbenzene, o-xylene, m+p-xylene, and toluene) were quantified by both
9
UCR and UCI. A total of 1170 paired (9 × # of samples collected at the same date and time by
UCI and UCR) VOC measurements were collected. Figure S.6 shows the extent of agreement,
represented as % difference, between UCI and UCR paired VOC measurements, where
% difference=100%×
∣
େ୍ିୈ
∣
ቚ
ిశి
మ
ቚ
. Good agreement was observed for a majority of the paired
VOC measurements: over half, 58%, agreed within 25% and 84% agreed within 50%. Potential
reasons for disagreement outside the uncertainty limits of the methods include influence of
highly localized sources: samplers were co-located ~15 feet apart, but had different orientations
with one inlet on the south side of the building and one inlet on the east side of the building; and
oxidation: UCR uses a pre-filter to scrub O
3
whereas UCI does not. Comparisons of individual
compounds are shown in Figures S.7-S.15.
Figure S.6.
Percent difference between UCI and UCR mixing ratios plotted against the percentage
of paired VOC measurements that fall within the percent difference. The grey square indicates the
10
percentage of UCI and UCR paired VOC measurements that agree within 25%; the blue square
indicates the percentage of paired VOC measurements that agree within 50%.
Figure S.7.
Measured UCI vs. UCR mixing ratios for decane.
11
Figure S.8.
Measured UCI vs. UCR mixing ratios for octane.
12
Figure S.9.
Measured UCI vs. UCR mixing ratios for nonane.
13
Figure S.10.
Measured UCI vs. UCR mixing ratios for 2-methylpentane.
14
Figure S.11.
Measured UCI vs. UCR mixing ratios for 2,3-dimethylpentane.
15
Figure S.12.
Measured UCI vs. UCR mixing ratios for ethylbenzene.
16
Figure S.13.
Measured UCI vs. UCR mixing ratios for o-xylene.
17
Figure S.14.
Measured UCI vs. UCR mixing ratios for m- and p-xylene.
18
Figure S.15.
Measured UCI vs. UCR mixing ratios for toluene.
19
S.3 Background CO and VOC Calculations
Background-corrected CO values were used in a subset of the ER calculations, as
described in the main text. Background CO values were calculated following the approach of de
Gouw et al. 2017.
3
First, the CO intercepts of 59 compounds were plotted against their OH
reaction rate coefficient,
k
OH
(Figure S.16).
4
The CO intercept of each compound was found from
the linear regression of the nighttime mixing ratios of each VOC against the CO mixing ratios.
Then, the CO intercepts of m- and p-xylenes (101 ppb), 1,2,4-trimethylbenzene (124-TMB, 84.5
ppb) and 1,3,5-trimethylbenzene (135-TMB, 85.0 ppb) were averaged yielding a background CO
value of ~90 ppb. The majority of the CO intercepts fall within +/- 15 ppb of the calculated
average, which is therefore defined as the uncertainty in the calculated background CO value.
Background CO values calculated at shorter time intervals (monthly) were within the uncertainty
(+/- 15 ppb) of the calculated average and did not monotonically increase or decrease during the
measurement period. Thus, a constant background CO value of 90 ppb was used in the ER
calculations
.
The background CO mixing ratio was subtracted from the measured CO mixing
ratio where noted,
CO, and used in the ER calculations. The CO levels during the measurement
campaign were not impacted by fires (see Figure S.17) and thus no other CO corrections were
applied.
Similarly to CO, VOCs with low OH-reactivity can build up in the atmosphere, affecting
ER calculations and source attribution. Five compounds (acetylene, propane, i-butane, n-butane
and benzene) with low OH reactivity (
k
OH
<
5×10
ିଵଶ
cm
3
molecule
-1
s
-1
) had non-negligible
background levels. For each of the five compounds, the background VOC mixing ratio was
defined as the lowest value. Two benzene and one acetylene values were omitted in
determination of the background values as they were anomalously low (>40% lower than the
20
average of the lowest 2% of mixing ratios). The background mixing ratios for these five
compounds were subtracted from the measured mixing ratios and the background-corrected
mixing ratios were used in the ER calculations.
Figure S.16.
Calculated CO intercepts for 59 individual VOCs plotted against their OH rate
constant
3
for each specific VOC. The solid black line indicates the calculated value for
background CO (90 ppb); the two dashed lines indicate +/- 15 ppb.
135-TMB
(m+p)-xylenes
124-TMB
21
Figure S.17.
Hourly median (blue dots) and daily median (orange line) CO mixing ratios measured
on the Caltech campus (CITAQS) before, during, and after LAAQC. Dates during the campaign
are within the black square.
S.4 Emission Ratio Calculations
Three different ERs relative to CO and acetylene were calculated: nighttime ERs
(ER
night
), OH-corrected ERs, and O
3
-corrected ERs
night
. Nighttime ERs were calculated using
nighttime data only (e.g., Borbon et al.
5
). VOC mixing ratios were plotted against CO and the ER
was defined as the slope of the linear best fit. This approach was also used to calculate ERs
against acetylene. Morning and afternoon sample data cannot be used to calculate ERs by this
simple regression method, because OH exposure decreases the levels of reactive VOCs in these
samples, resulting in artificially low ERs. de Gouw et al.
3
presented a method for OH-corrected
ERs, in which all data can be used following an OH-loss correction. OH exposure can be
calculated using the ratio of a less reactive VOC with OH to a more reactive VOC with OH.
6
Following de Gouw et al.,
3
benzene was chosen as the less reactive compound and 11 other
22
VOCs were evaluated as the more reactive compound. The results of calculated OH exposure
using the ratio of benzene to each of these 11 VOCs are shown in Figure S.18. The point markers
in Fig. S.18 represent the average OH exposure calculated from the 3
/day samples (05:30–
06:30, 09:00–10:00 and 14:00–15:00 PDT) and the lines represent the average OH exposure
calculated from two diurnals (one sample per hour over a 24-hr period) that were collected
during the campaign.
The ratio of benzene to 1,2,4-trimethylbenzene (yellow trace) was used in Equation S.1 to
calculate OH exposure. It represents the average of the 11 pairs and facilitates comparison with
de Gouw et al.
3
in which 1,2,4-trimethylbenzene was also used for calculating OH exposure.
[
OH
]
∆
푡
=
(
푘
ଵଶସାୌ
−
푘
ୠୣ୬ୣ୬ୣାୌ
)
ିଵ
×
൬
ln
ቀ
ୠୣ୬ୣ୬ୣ
ଵଶସ
ቁ
− ln
൫
ER
ୠୣ୬ୣ୬ୣ
/
ଵଶସ
൯൰
(Eq. S.1)
where
ER
ୠୣ୬ୣ୬ୣ
/
ଵଶସ
is the emission ratio of benzene to 1,2,4-trimethylbenzene and is
assumed to be constant across the basin.
3,7
The value of
ER
ୠୣ୬ୣ୬ୣ
/
ଵଶସ
, 1.84, was determined
as described below.
ER
ୠୣ୬ୣ୬ୣ
/
ଵଶସ
was calculated using two approaches. In the first, a time series of the
ratio of the measured mixing ratios was plotted (Figure S.19, top) and the lowest observed ratio
is taken as
ER
ୠୣ୬ୣ୬ୣ
/
ଵଶସ
;
this approach yielded a value of 1.84. In the second, benzene
mixing ratios were plotted against the 1,2,4-trimethylbenzene mixing ratios (Figure S.19,
bottom). The
ER
ୠୣ୬ୣ୬ୣ
/
ଵଶସ
is then defined as the slope of the best linear fit through the least
processed air mass, the nighttime samples; this approach yielded a value of 2.1. The lower of the
two values, 1.84, was used in the OH exposure calculations. A value of 1.75 ± 0.15 was
estimated by de Gouw et al.
3
for CalNex-2010 using the same approach.
23
Analogous to the OH-corrected ERs calculated from estimated OH exposure, nighttime
O
3
-corrected ERs were calculated from estimated O
3
exposure (Equation S.2).
3
[
O3
]
∆
푡
=
(
푘
ୡଶ୳୲ାଷ
−
푘
ୠୣ୬ୣ୬ୣା
)
ିଵ
×
൬
ln
ቀ
ୠୣ୬ୣ୬ୣ
ୡଶ୳୲
ቁ
− ln
൫
ER
ୠୣ୬ୣ୬ୣ
/
ୡଶ୳୲
൯൰
(Equation S.2)
Similarly to ER
benzene/1,2,4-trimethylbenzene
, ER
benzene/c2But
was calculated using two approaches: 1)
plotting the time series of the ratio of the measured mixing ratios; the lowest observed ratio
yielded a value of 6.64; and 2) plotting benzene mixing ratios against cis-2-butene mixing ratios
from nighttime samples only; the best linear fit yielded a slope of 7. The lowest value, 6.64, was
used in Equation S1 for calculating O
3
exposure. de Gouw et al.
3
estimated a value of 5±2 using
the same approach. Using the nighttime data only, VOC/CO ratios were plotted against the
corresponding O
3
-exposure and the O
3
-corrected ERs were then defined as the y-intercept of the
linear fit.
The methods used to calculate ERs are illustrated in Figure S.20 and Figure S.21. Figure
S.20 shows acetylene, (m+p)-xylene and α-pinene mixing ratios against CO mixing ratios;
marker shape is used to differentiate sampling time. The need for OH-correction is well
illustrated by (m+p)-xylenes, for which the afternoon mixing ratios are lower than the morning
and nighttime mixing ratios due to the relatively high reactivity with OH. The slope of the best
linear fit using the nighttime data only is defined as the emission ratio (ER
night
).
4
In Figure S.21,
VOC/
∆
CO for the same three compounds is plotted against OH exposure; the y-intercept of the
best linear fit is defined as the emission ratio (OH-corrected ER). It can be seen that, as expected,
α-pinene is not correlated with CO (Figure S.20) and the VOC/
∆
CO shows no clear trend with
OH exposure (Figure S.21).
Figure S.22 shows OH-corrected and O
3
-corrected ERs vs. ERs
night
(to both CO and
acetylene). There is good agreement between the ERs calculated using the three different
24
approaches. Generally, the chemistry-corrected ERs are higher than the corresponding ERs
night
.
For highly reactive compounds, the chemistry-corrected ERs are expected to be higher. All ERs
are reported in Table S.1.
Figure S.18.
OH-exposure calculated using the ratio of benzene to 11 more-reactive VOCs.
Point markers represent OH-exposure calculated using the entire dataset (samples collected
3
/day). Lines represent OH-exposure calculated using two diurnal sample collections (one
sample/hour for two 24-hour periods).
25
Figure S.19.
Methods to determine the constant emission ratio of benzene:1,2,4-
trimethylbenzene: time series of the ratio of measured benzene to measured 1,2,4-
trimethylbenzene (top); and scatterplot of benzene and 1,2,4-trimethylbenzene mixing ratios
(bottom). The dashed line in the top plot represents the lowest observed ratio (1.84 ppt/ppt). The
26
colors in the bottom plot are indicative of the time of day the sample was collected with dark
colors representing nighttime and light colors representing afternoon.
Figure S.20.
Observed VOC versus CO mixing ratios for acetylene (blue), m+p-xylene (yellow),
and α-pinene (green). Triangles indicate nighttime samples (05:30-06:30), diamonds indicate
27
morning samples (09:00-10:00), and + indicate afternoon samples (14:00-15:00). Emission ratios
were calculated using the y-intercept of the linear fit for nighttime samples only. The α-pinene
emission ratio was not calculated because of the poor correlation with CO.
Figure S.21.
Nighttime, morning, and afternoon VOC/ΔCO ratios versus calculated OH exposure
for acetylene, m+p xylene, and α-pinene. The y-intercept is defined as the OH-corrected emission
ratio.
28
29
Figure S.22.
OH- and O
3
-corrected emission ratios plotted against nighttime emission ratios to
CO (top) and acetylene (bottom). Slopes are reported for each of the four correlations with the
intercept forced through zero.
S.4 Temperature Dependence of Terpene Emissions
Measured mixing ratios for six terpenes are plotted against temperature in Figure S.23.
Three of the six terpenes, α-pinene, camphene, and eucalyptol appear to be more strongly
correlated with temperature than the other three terpenes, β-myrcene, limonene, and ocimene.
30
Figure S.23.
Terpene mixing ratios plotted against temperature.
S.5 Trends in VOC Mixing ratios
The ratio of the June-July average mixing ratio to the April-May average mixing ratio is
presented in Figure S.24 for 40 compounds; the compounds are ordered as in Table S.5. The
31
ratios for all compounds are close to one, supporting the conclusions that changes in human
activity during LAAQC-2020 did not strongly affect the mixing ratios of this subset of
compounds, which are largely combustion-derived VOCs.
To evaluate whether the 2020 combustion-derived VOC mixing ratios were anomalously
low during the LAAQC sampling period, the average mixing ratios of three individual
compounds (CO, acetylene, and i-pentane) and total non-methane hydrocarbons
(NMHCs) measured during LAAQC-2020 were compared with average mixing ratios measured
over the LA Basin during the NASA Student Airborne Research Program (SARP) between
2015-2019. The SARP data were averaged over a 5 year period to normalize differences in
meteorology between years. The SARP data were collected onboard an aircraft at altitudes < 1
km. It is expected that the mixing ratios of the shorter-lived compounds would be lower in the
samples collected aloft than those collected at the Pasadena ground site. For all three individual
compounds (Figures S.25-S.27) and total NMHCs (Figure S.28), the median LAAQC values fall
within the upper quartile of the SARP values.
32
Figure S.24.
The ratio of average June-July mixing ratio to average April-May mixing ratio for the
40 compounds listed in Table S.3.
33
Figure S.25.
Boxplots showing the CO mixing ratio measured as a part of SARP (2015-2019, left)
and during LAAQC (2020, right). The median is represented by the red line and the 25th and 75th
percentiles by the edges of the blue box.
2011
2020
0
200
400
600
800
1000
1200
1400
2015-2019
2020
34
Figure S.26.
Boxplots showing the acetylene mixing ratio measured as a part of SARP (2015-
2019, left) and during LAAQC (2020, right). The median is represented by the red line and the
25th and 75th percentiles by the edges of the blue box.
2011
2020
0
500
1000
1500
2000
2500
3000
2015-2019
Acetylene,