of 21
Ozone production chemistry in the
presence of urban plumes
†
W. H. Brune,
*
a
B. C. Baier,
a
J. Thomas,
ab
X. Ren,
c
R. C. Cohen,
d
S. E. Pusede,
de
E. C. Browne,
df
A. H. Goldstein,
g
D. R. Gentner,
h
F. N. Keutsch,
i
J. A. Thornton,
j
S. Harrold,
jk
F. D. Lopez-Hil
fi
ker
jl
and P. O. Wennberg
m
Received 1st December 2015, Accepted 23rd December 2015
DOI: 10.1039/c5fd00204d
Ozone pollution a
ff
ects human health, especially in urban areas on hot sunny days. Its
basic photochemistry has been known for decades and yet it is still not possible to
correctly predict the high ozone levels that are the greatest threat. The CalNex_SJV
study in Bakers
fi
eld CA in May/June 2010 provided an opportunity to examine ozone
photochemistry in an urban area surrounded by agriculture. The measurement suite
included hydroxyl (OH), hydroperoxyl (HO
2
), and OH reactivity, which are compared
with the output of a photochemical box model. While the agreement is generally within
combined uncertainties, measured HO
2
far exceeds modeled HO
2
in NO
x
-rich plumes.
OH production and loss do not balance as they should in the morning, and the ozone
production calculated with measured HO
2
is a decade greater than that calculated with
modeled HO
2
when NO levels are high. Calculated ozone production using measured
HO
2
is twice that using modeled HO
2
, but this di
ff
erence in calculated ozone
production has minimal impact on the assessment of NO
x
-sensitivity or VOC-sensitivity
for midday ozone production. Evidence from this study indicates that this important
discrepancy is not due to the HO
2
measurement or to the sampling of transported
a
Department of Meteorology, Pennsylvania State University, University Park, PA 16802, USA. E-mail: whb2@
psu.edu
b
Department of Earth and Planetary Sciences, Johns Hopkins University, Baltimore, MD 21218, USA
c
NOAA Air Resources Laboratory, College Park, MD, 20740, USA
d
Departments of Chemistry and Earth and Planetary Science, University of California, Berkeley, CA 94720, USA
e
Department of Environmental Sciences, University of Virginia, Charlottesville, VA 22904, USA
f
Department of Chemistry and Biochemistry, University of Colorado, Boulder, CO 80309, USA
g
Department of Environmental Science, Policy, and Management, University of California, Berkeley, CA 94720,
USA
h
Department of Chemical and Environmental Engineering, Yale University, New Haven, CT 06511, USA
i
Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA 02138, USA
j
Department of Atmospheric Sciences, University of Washington, Seattle, WA 98195, USA
k
Puget Sound Clean Air Agency, Seattle, WA 98101, USA
l
Paul Scherrer Institut, Villigen, PSI, Switzerland
m
The Linde Center for Global Environmental Science, Caltech, Pasadena, CA 91125, USA
†
Electronic supplementary information (ESI) available. See DOI: 10.1039/c5fd00204d
This journal is © The Royal Society of Chemistry 2016
Faraday Discuss.
,2016,
189
,169
–
189 |
169
Faraday Discussions
Cite this:
Faraday Discuss.
,2016,
189
,169
PAPER
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
View Journal
| View Issue
plumes but instead to either emissions of unknown organic species that accompany the
NO emissions or unknown photochemistry involving nitrogen oxides and hydrogen
oxides, possibly the hypothesized reaction OH + NO + O
2
/
HO
2
+NO
2
.
Introduction
Ground-level ozone (O
3
) is a serious health hazard,
1,2
with no known safe limit,
and is estimated to cause millions of deaths per year globally. In the United States,
urban ozone levels have decreased dramatically in the past two to three decades
due to the signi
cant investments made in understanding the cause of ozone
pollution and to air quality regulations.
3
These regulations have encouraged lower
emissions of nitrogen dioxide (NO
2
) and nitric oxide (NO) (collectively termed
NO
x
) and volatile organic compounds (VOCs), which are the raw ingredients
needed to produce ozone. Although ozone reductions in some parts of the United
States appear to have leveled o
ff
, ozone reduction in the United States is a success
story
–
one that should be emulated globally. However, with limited resources,
much of the world cannot a
ff
ord the investment the United States has made. The
solution is an optimized approach that can target the speci
c emissions to which
ozone production is most sensitive.
An optimized approach to ozone pollution reduction requires a solid scienti
c
understanding of the causes of ozone pollution. The basic chemistry has been
known for decades.
4
–
6
Ozone production begins with the early morning produc-
tion of the hydroxyl (OH) and hydroperoxy (HO
2
) radicals in the presence of NO
x
(NO
2
+ NO) and VOC emissions. Hydroxyl reacts with the VOCs, causing a cascade
of reactive organic compounds, including organic peroxy radicals (RO
2
) and HO
2
.
The peroxy radicals react with NO to form peroxy radicals and NO
2
. Nitrogen
dioxide is decomposed by ultraviolet sunlight to form NO and atomic oxygen,
which immediately reacts with molecular oxygen (O
2
) to form ozone. This process
is the dominant production pathway for tropospheric ozone.
However, NO also reacts with O
3
to make NO
2
, NO and NO
2
come into a steady-
state balance within tens of seconds, and NO
2
then reacts with OH to terminate
NO
x
and HO
x
(OH + HO
2
) cycling. By this theory, the instantaneous ozone
production,
P
(O
3
), initially increases as NO increases, but then decreases with
continued NO increase. The peak
P
(O
3
) is a sensitive function of HO
x
produc-
tion.
7
–
9
One simple test of the non-linear dependence of ozone production on NO
x
is the so-called weekend-weekday e
ff
ect, for which ozone is greatest on weekends
when NO
x
is less.
10
Despite this well-accepted theory, there are discrepancies between measured
and ozone calculated by regional air quality models. The models agree with the
observations on average, but tend to be too high for values below 50 ppbv and too
low for values approaching 80
–
90 ppbv and beyond. While models have had better
agreement with observations for individual cities and times of the year, this
discrepancy between modeled and measured ozone is found for several di
ff
erent
models and urban areas.
11
–
13
The causes of the disagreements between measured
and modeled ozone are usually attributed to errors in the emissions inventories
for nitrogen oxides (NO
x
) and volatile organic compounds (VOCs), but there is
some evidence that the well-tested theory may need some modi
cation.
The budget equation for ozone is
Faraday Discussions
Paper
170
|
Faraday Discuss.
,2016,
189
,169
–
189
This journal is © The Royal Society of Chemistry 2016
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
v
½
O
3

v
t
¼
P
chem

L
chem
þ
w
e
D
O
3

u
d
½
O
3

H

V
ð
v
½
O
3
Þ
;
(1.1)
where
P
chem
is the chemical O
3
production rate,
L
chem
is the loss rate,
w
e
D
O
3
/
H
is
the ozone entrainment rate between the mixing layer and the free troposphere,

u
d
[O
3
]/
H
is the ozone deposition rate,
H
is the mixing layer height, and

V
(
v
[O
3
]) is the ozone advection rate by the mean wind,
v
. Net ozone production can be
calculated from 1.2
–
1.4,
P
(O
3
)
¼
P
chem

L
chem
(1.2)
P
chem
¼
k
NO
þ
HO
2
½
NO
½
HO
2
þ
X
i
k
NO
þ
RO
2i
½
NO
½
RO
2i

(1.3)
L
chem
¼
f
H
2
O
J
O
3
[O
3
]+
k
O
3
+OH
[O
3
][OH] +
k
O
3
+HO
2
[O
3
][HO
2
]+
k
NO
2
+OH
[NO
2
][OH]
+
L
(O
3
+ alkenes) +
L
(O
3
+ halogens) +
P
(RONO
2
)
(1.4)
where
k
's are rate coe
ffi
cients; NO is nitric oxide; HO
2
is the hydroperoxyl radical;
RO
2
is the organic peroxyl radical;
J
O
3
is the photolysis frequency of O
3
;
f
H
2
O
is the
fraction of the excited state O(
1
D) atoms from O
3
photolysis that react with H
2
O;
OH is the hydroxyl radical; and RONO
2
represents organic nitrates.
In several studies, the measured HO
2
o
en is less than modeled HO
2
at low
NO, equals modeled HO
2
when NO is about 1 ppbv, and increasingly exceeds
modeled HO
2
at increasingly higher NO abundances.
9,14
–
18
This measured-to-
modeled di
ff
erence is greatest when HO
x
production is lowest, such as during
morning rush hour, where the measured-to-modeled HO
2
ratio can be higher
than ten. It is least when HO
x
production is greatest, such as during the a
er-
noon. This greater-than-expected HO
2
at higher NO is also inferred from perox-
ynitric acid (HO
2
NO
2
) measurements in Mexico City in 2006.
19
Therefore, ozone
production (
P
O
3
) calculated using the measured HO
2
can exceed
P
(O
3
) calculated
using the modeled HO
2
. In addition, a direct measurement of the ozone
production rate shows that measured
P
(O
3
) is twice
P
(O
3
) calculated using
modeled HO
2
and RO
2
, although the timing of the
P
(O
3
) peak agrees better with
the
P
(O
3
) calculated with modeled HO
2
than with
P
(O
3
) calculated with measured
HO
2
.
20,21
These observations are inconsistent with the current understanding of
ozone production.
This discrepancy can be explained in several di
ff
erent ways. A
rst hypothesis
is that the HO
2
measurement is being a
ff
ected by the atmospheric NO levels. A
second hypothesis is that calculation of
P
(O
3
) is being distorted by the averaging
over plumes of NO
x
-rich and HO
2
-rich air so that the product of averaged
measured [HO
2
] and [NO] is larger than the average of the products of the plume-
scale NO and HO
2
:
½
HO
2

½
NO

.
½
HO
2
½
NO

:
(1.5)
A third hypothesis is that HO
x
measurements made when NO
x
is not in pho-
tostationary state (PSS) are being compared to models that calculate HO
x
while
assuming an NO
x
photostationary state. Because NO
x
is o
en emitted from
combustion as NO, the HO
x
loss due to OH + NO
2
would be less than in the model
and measured HO
2
would appear to be greater than modeled. The fourth
Paper
Faraday Discussions
This journal is © The Royal Society of Chemistry 2016
Faraday Discuss.
,2016,
189
,169
–
189 |
171
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
hypothesis suggests that unknown HO
x
sources are co-emitted with the NO
x
. The
h hypothesis is that unknown chemistry is missing from the current under-
standing of the chemistry between nitrogen oxides and hydrogen oxides.
The multi-agency California Research at the Nexus of Air Quality and Climate
Change-San Joaquin Valley (CalNex-SJV) study in Bakers
eld CA during June 2010
meets the criteria needed to test these hypotheses. Bakers
eld is situated in the
southern San Joaquin Valley 180 km north-northwest of Los Angeles and experi-
ences the second worst ozone pollution in California.
22
It is surrounded by agri-
cultural land, but at the same time has active oil and gas
elds, an oil re
nery and
signi
cant tra
ffi
c on highways that has a higher fraction of diesel powered vehi-
cles than other urban areas in the US.
23
Its numerous point and distributed NO
x
and VOC emission sources create a heterogeneous mix of plumes of di
ff
erent sizes
and chemical composition. This combination of high VOC and NO
x
emissions
and the typically cloudless, hot weather causes high ozone abundances. Bakers-
eld CA typically has 90 exceedances of ozone air quality standards each year and,
while the number has decreased during the past decade, it is still not in
compliance with EPA air quality standards.
22
The calculated ozone production has been examined for the San Joaquin Valley
(SJV), including Bakers
eld. In a study using midday (10:00
–
14:00 local time)
measurements of temperature as a surrogate for OH reactivity from VOCs, NO
x
,and
frequency of O
3
exceedances, Pusede and Cohen showed that the calculated midday
P
(O
3
) is consistent with a transition from VOC-sensitive to NO
x
-sensitive regimes for
much of the SJV, starting
rst with higher maximum daytime temperatures
(34
–
45

C) followed by the beginning of a transition at more moderate temperatures
(28
–
33

C).
24
In a second study focused on the CalNex-SJV site, Pusede
et al.
25
show
that
P
(O
3
) calculated with an analog model is NO
x
-sensitive for moderate-to-high
temperatures, except for weekdays at moderate temperatures, for which it is VOC-
sensitive. These conclusions appear to be inconsistent with the
P
(O
3
)calculated
from the measurements of HO
2
and RO
2
as a function of NO described above.
However, high NO and greater-than-expected
P
(O
3
) occur mainly in the morning
before 10:00 PST and it is not clear how this di
ff
erence will a
ff
ect the averaged
P
(O
3
)
sensitivity to NO
x
and VOCs calculated for midday.
In this paper, we will test these hypotheses using primarily the HO
2
and NO
abundances measured during CalNex-SJV. First we will compare the measured
OH, HO
2
, and OH reactivities to those calculated by a near-explicit box model that
is constrained by all other simultaneous
eld measurements in Bakers
eld.
Measured and modeled OH production and loss rates and ozone production from
HO
2
are also compared and discussed. Finally we add the evidence learned from
this study to that from previous studies to assess the likelihood of the hypotheses
for the greater-than-expected ozone production at greater NO abundances.
Methods
Measurement site
The CalNex 2010
eld campaign in California consisted of aircra
and ship
measurements, as well as two ground sites: one in Pasadena CA and a second in
Bakers
eld CA. This work focuses on the CalNex-SJV Bakers
eld site (Fig. 1). A
sca
ff
olding tower 18 m high was erected in a
eld just to the east of a parking lot at
the California Agricultural Experiment Station, which is located in south Bakers
eld
Faraday Discussions
Paper
172
|
Faraday Discuss.
,2016,
189
,169
–
189
This journal is © The Royal Society of Chemistry 2016
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
(35.346153 N, 118.965519 W). The land to the south is
elds and settling ponds.
Light businesses were located to the north, along with highway 58, a highly traf-
cked corridor, about 760 m away. A school bus facility was also located 600 m to
the west.
Measurements were made from 15 May to 28 June 2010. For that time of year,
the weather in Bakers
eld is typically hot and dry, with temperatures routinely
exceeding 37

C. However, 2010 was not a typical year. The temperature did not
exceed 35

C until the last three days, and during one week, the weather was
cloudy with some light rain with temperatures ranging from 20

C and to 25

C.
Winds were generally from the north at 4 ms

1
, although they were o
en calm
and from the east in the hours around sunrise (Fig. 2).
HO
x
measurements
OH and HO
2
were measured with Penn State's Ground-based Tropospheric
Hydrogen-Oxides Sensor (GTHOS),
26
which uses laser-induced
uorescence (LIF).
The hydroxyl radical (OH) is detected as the sampled air is pulled through a 1 mm
inlet into low pressure (

6 hPa) and passes through the path of a laser tuned to
the Q
1
(2) OH absorption line at 308 nm, and then
uoresces. This
uorescence is
measured in a
rst detection axis by a gated microchannel plate detector posi-
tioned perpendicular to the sample
ow and the laser beam. The sampled air
ows through the
rst detection axis, and has NO injected into it to react with
HO
2
to form OH, which is then detected by LIF in a second detection axis in order
to detect HO
2
.
A tunable dye laser pumped by a 532 nm Nd:YAG laser produces the 308 nm
light used to detect OH. This laser wavelength is alternately tuned by an etalon to
Fig. 1
Map of Bakers
fi
eld CA and location of the CalNex-SJV site (black dot) (Google,
2015). The site was at 35.346153

N and 118.965519

W. The heavily traveled four-lane
highway 58 was located 760 m to the north of the site. The nearest local road, East Belle
Terrace, was 50 m to the north and had light tra
ffi
c during the study period.
Paper
Faraday Discussions
This journal is © The Royal Society of Chemistry 2016
Faraday Discuss.
,2016,
189
,169
–
189 |
173
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
a wavelength where OH absorbs and then
uoresces or to the background on
alternate sides of the OH absorption line, all within a 30 second cycle. The
di
ff
erence between these two signals is proportional to OH. The proportionality
constant is determined by laboratory and
eld calibrations.
26
This method of
measuring OH is referred to as OH
wave
. Wavelength modulation has been the
most common method for measuring atmospheric OH by LIF.
A second OH measurement method involves injecting an OH reactant into the
air to scavenge the OH before it is sampled through the instrument inlet. Reactant
amounts are chosen to maximize the fraction of OH removed in the 10 ms
between injection and entering the inlet and to simultaneously minimize the OH
removed inside the instrument. Hexa
uoropropylene (C
3
F
6
) was used as the OH
scavenger. By turning C
3
F
6
injection on and o
ff
, the OH signal is found by sub-
tracting the signal when injection is on from the signal when injection is o
ff
. This
method is called OH
chem
. The di
ff
erence between OH
wave
and OH
chem
is the OH
from an interference, called OH
int
. The OH interference has the spectral signature
of OH, but all studies show that it is not laser-generated. To test the functionality
of the OH
chem
system, a UV lamp was a
ffi
xed to the instrument near the inlet. The
lamp, which photolyzed water vapor to make a large OH signal, was turned on for
a few minutes three times a day to ensure that the C
3
F
6
injection was scavenging
OH properly. For this paper, the OH measurements were all made with the
chemical removal method.
Two years a
er CalNex-SJV, it was reported that some alkene-based and
aromatic-based organic peroxides were also detected along with HO
2
in most
instruments that used NO to convert HO
2
to OH.
27
Methods have been found to
minimize this interference, but these methods were not used in the CalNex-SJV
study. This sum of HO
2
and a subset of RO
2
has been called HO
2
*
, which is then
compared to modeled HO
2
*
.
28
In this paper, our approach is to bound the
possible HO
2
values by subtracting all the modeled RO
2
from HO
2
*
and calling
this value HO
2
, which may be an overcorrection. This HO
2
is actually the lower
Fig. 2
Behavior of measured NO (blue solid line),
J
(NO
2
) (red dashed line), and wind speed
(dot-dashed black line) for day-of-the-year (day) 160, 9 June. This behavior is typical for
most days during the study period, although the NO bump on day 160 was one of the
largest. The widths of the NO spikes range from seconds to hours.
Faraday Discussions
Paper
174
|
Faraday Discuss.
,2016,
189
,169
–
189
This journal is © The Royal Society of Chemistry 2016
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
bound of HO
2
and HO
2
*
is the upper bound. As will be seen, the main conclusions
are not sensitive to the choice of HO
2
between these two bounds.
In addition to OH and HO
2
measurements, measurements were also made of
OH reactivity, which is the inverse of the OH lifetime.
29,30
Approximately 150 LPM
of ambient air is drawn into the instrument and
ows through the aluminum
ow
tube (7.5 cm dia.). At the far end of the
ow tube is a sampling inlet and an OH
measurement system nearly identical to the one used in the main GTHOS system.
Before the air
ow reaches the sampling inlet, it
ows past a movable source of OH
called the wand. Inside the wand, 5 LPM of moist nitrogen
ows past a mercury
lamp, which photolyzes the water vapor to produce OH and HO
2
and then jets out
the detection axis. As the wand is moved farther away from the sampling inlet, the
OH has more time to react with trace gases in the ambient air
owing through the
tube and the OH signal decreases exponentially. Moving 10 cm is equivalent to
a decay time of 140 ms and the wand completes a cycle in 30 seconds. The OH
reactivity is the slope of the change in the log of the OH signal divided by the
reaction time.
The OH reactivity decay is a
ff
ected by atmospheric NO because HO
2
+NO
/
OH + NO
2
recycles OH that has decayed, thus causing curvature in the OH decay.
A least-squares linear
t to this curved decay produces a decay slope that is less
than the real slope and thus produces an OH reactivity value that is too low. The
e
ff
ects of atmospheric NO on the OH reactivity measurement are minimized by
applying the correction algorithm described in Shirley
et al.
31
The suite of measurements at the CalNex-SJV site was extensive as documented
in the CalNex overview paper.
32
It included meteorological parameters, inorganic
species, volatile organic compounds (VOCs) and oxygenated VOCs, and many
aerosol abundances and properties. Data used in this study were drawn primarily
from measurements taken at or near the top of the measurement tower.
Photochemical box modeling
The simultaneous measurements of all available inorganic and organic species
and meteorological parameters were used to calculate HO
x
using the near-explicit
Master Chemical Mechanism, Version 3.2 (MCMv3.2)
33
in a box model framework
developed by G. Wolfe.
34
MCMv3.2 contains approximately 6700 unique chemical
species and 17 000 reactions. Measured VOCs that are included in the model are
treated explicitly; those that are not represented in the model are aggregated into
appropriate MCMv3.2 species based on OH-reactivity or their molecular structure.
It should be noted that isoprene chemistry in MCMv3.2 is replaced with explicit
reactions detailed in Mao
et al.
35
Chemical species were assumed to have a life-
time of one day to prevent buildup in the model. Data were then averaged into
ten-minute time intervals for the modeling and the comparisons to
measurements.
Photolysis frequencies were not measured during CalNex-SJV, so instead they
were calculated using the NCAR Tropospheric Ultraviolet and Visible Radiation
Model.
36
These calculations assume clear overhead skies with the overhead ozone
column of 300 D.U. taken from satellite measurements; to correct for the e
ff
ects of
overhead cloud cover, photolysis frequencies were scaled by the ratio of the
calculated JNO
2
to JNO
2
measured from a UV radiometer on the measurement
tower. This same cloud correction factor was applied to the other calculated
Paper
Faraday Discussions
This journal is © The Royal Society of Chemistry 2016
Faraday Discuss.
,2016,
189
,169
–
189 |
175
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
photolysis frequencies and has been shown to give accurate photolysis
frequencies.
37
Model results include days between 23 May and 28 June 2010 to allow for the
greatest number of simultaneously measured chemical species to constrain the
model. Data were taken from three periods when HO
x
, OH reactivity, and other
chemical species important for this analysis were being measured: day 145
–
150;
day 156
–
164; and day 166
–
174. They were averaged or interpolated into 10 minute
time intervals for the model runs. For analysis of HO
x
and
P
(O
3
) as a function of
NO, the 10 minute data were limited to hours between 7:00 and 17:00, a time
period chosen to capture the portion of morning rush hour in which the
photolysis and transport were well de
ned. All times are Paci
c Standard Time
even though the o
ffi
cial time was Paci
c Daylight Time during the
eld study.
CalNex-SJV measurement and model uncertainties
The absolute uncertainties for OH and HO
2
are approximately

40% at the 2
s
con
dence level. However, the subtraction of the OH signals with and without
C
3
F
6
scavenging causes the OH limit-of-detection to be about (2
–
3)

10
5
cm

3
.In
addition, with the RO
2
interference in the HO
2
measurement, there is an addi-
tional uncertainty that can be about a factor of two. In studies a
er CalNex, our
measurement strategy was revised to minimize the RO
2
interference in the HO
2
measurement. The uncertainty for the OH reactivity instrument is estimated to be

30% at 2
s
con
dence. The uncertainties in the other measurements are given in
Table 7a of the overview paper.
32
The model uncertainty can be estimated from
a global uncertainty and sensitivity analysis from a previous urban study using the
RACM2 model.
38
These two studies have similar uncertainties in their input
measurements, reaction rate coe
ffi
cients, and products, so the uncertainties in
the modeled OH and HO
2
for comparable urban areas should also be similar.
Thus, the estimated 2
s
uncertainty is approximately 40% for OH and HO
2
. These
uncertainties will be used to assess the signi
cance of the comparisons between
the measured and modeled OH, HO
2
, and OH reactivity in this study.
Results
Comparison of measured and modeled OH and HO
2
Understanding ozone production requires an understanding of OH and HO
2
.A
rst-order test of this chemistry is the comparison of the modeled and measured
OH and HO
2
as a function of the time of day (Fig. 3).
The median measured and modeled OH display the same diel behavior but the
measured OH peaked at (7.3

2.5)

10
6
cm

3
while the modeled OH peaked at
(10.4

4.0)

10
6
cm

3
. This marginal statistical di
ff
erence of

30% persists
during the midday, although at sunrise and sunset, the modeled and measured
OH agree. At night, the median measured OH is 3

10
5
cm

3
, which is near the
instrument detection limit and is consistent with the modeled nighttime OH of
3

10
4
cm

3
.
The OH interference signal peaks at midday at 2

10
6
cm

3
, is about 1

10
6
cm

3
at night, and is only about 25% of the OH determined by the chemical
removal method. Interestingly, this OH interference signal is about the same as
that observed in forests
35,39
and in other cities
17
and may have a common origin,
Faraday Discussions
Paper
176
|
Faraday Discuss.
,2016,
189
,169
–
189
This journal is © The Royal Society of Chemistry 2016
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
but because the OH is about 10 times larger in cities than in forests, the inter-
ference's e
ff
ect of the measured OH is much more signi
cant in forests than in
cities.
Measured HO
2
is on average about 7 pptv at midday and is within the
combined uncertainties of the modeled HO
2
. Measured and modeled HO
2
also
agree at night, but in the morning, median measured HO
2
is more than twice
median modeled HO
2
. The HO
2
with RO
2
interference is about twice as large at
midday and much larger at night. Thus measured and modeled HO
2
agree within
uncertainties for much of the day, if true HO
2
is near the lower bound for
measured HO
2
.
When median measured and modeled OH and HO
2
are plotted against NO for
the daytime, the modeled OH is similar to measured OH for NO up to about 10
ppbv (Fig. 4(a)). Measured HO
2
and HO
2
*
, on the other hand, begin deviating
from modeled HO
2
when NO is 1 ppbv, and are

10 times larger when NO is 10
ppbv (Fig. 4(b)). This result is consistent with the previous reports of higher-than-
expected measured HO
2
for conditions where NO is above 1 ppbv. Note that this
discrepancy does not depend on uncertainties in removing the RO
2
interference
from the HO
2
signal because HO
2
and HO
2
*
behave the same way.
HO
2
increases with the increasing production rate of HO
x
,
P
(HO
x
), for each
value of NO (Fig. 4(c)). The 10 minute data are binned into three
P
(HO
x
) ranges: 1

10
5
cm

3
s

1
<
P
(HO
x
)<5

10
6
cm

3
s

1
,5

10
6
cm

3
s

1
<
P
(HO
x
)<10
7
cm

3
s

1
, and
P
(HO
x
)>10
7
cm

3
s

1
. In Bakers
eld, these ranges correspond to low
P
(HO
x
) for early morning, late a
ernoon, or very cloudy days; medium
P
(HO
x
) for
mid-morning, mida
ernoon, or cloudy days; and high
P
(HO
x
) for midday on
sunny days. Note that both the measured and modeled HO
2
are greater for greater
P
(HO
x
), as would be expected, but modeled HO
2
is more sensitive to
P
(HO
x
) than
measured HO
2
is. Furthermore, in general, the measured-to-modeled HO
2
ratio is
least for the highest
P
(HO
x
) and greatest for the lowest
P
(HO
x
). Thus, regions with
Fig. 3
Median diel variation of HO
x
. (a) OH (10
6
cm

3
), measured (blue circles), modeled
(red squares), and OH interference (green stars), with individual 10 min data for measured
OH (gray dots); (b) HO
2
(pptv), measured (blue circles), modeled (red squares), and
measured HO
2
with the RO
2
interference (green stars), with individual 10 min data for
measured HO
2
(gray dots). Error bars are

1
s
con
fi
dence.
Paper
Faraday Discussions
This journal is © The Royal Society of Chemistry 2016
Faraday Discuss.
,2016,
189
,169
–
189 |
177
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
Fig. 4
Median behavior of HO
x
as a function of NO. (a) OH (10
6
cm

3
), measured (blue
circles), modeled (red squares), and measured averages of 20 seconds (upward pointing
triangles), 1 minute (diamonds), and 1 hour (downward pointing triangles); (b) HO
2
(pptv),
measured (blue circles), modeled (red squares), measured HO
2
with RO
2
interference
(green stars), and measured averages of 20 seconds (upward pointing triangles), 1 minute
(diamonds), and 1 hour (downward pointing triangles); (c) HO
2
(pptv), measured (solid blue
lines) and modeled (dashed red lines), for P(HO
x
)>10
7
cm

3
s

1
(upward triangles), 5

10
6
cm

3
s

1
<
P
(HO
x
)<10
7
cm

3
s

1
(diamonds), and 1

10
5
cm

3
s

1
< P(HO
x
)<5

10
6
cm

3
s

1
(downward triangles). Gray dots are individual 10 min measured data; darker gray
dots are data measured between 12:00 and 15:00. Data are
fi
ltered for daytime hours
between 7:00 and 17:00.
Fig. 5
Median diel variation of OH reactivity. Median measured (blue circles) and calcu-
lated from modeled chemical species (red squares), with individual 10 min data for
measured OH reactivity (gray dots). A total of 339 out of 4392 10 minute data values are
greater than 30 s

1
and are not shown. Error bars are 1s con
fi
dence.
Faraday Discussions
Paper
178
|
Faraday Discuss.
,2016,
189
,169
–
189
This journal is © The Royal Society of Chemistry 2016
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
higher
P
(HO
x
) likely have a measured-to-modeled HO
2
ratio that is closer to 1 even
for higher NO values.
These di
ff
erences between measured and modeled OH and HO
2
are within the
widely varying range of previous studies.
40
Even this behavior as a function of
P
(HO
x
) is the same as in previous studies. The CalNex results are quite similar to
what is observed in an April 2009 SHARP study in Houston TX.
17
An important
di
ff
erence between the Houston and Bakers
eld studies is the higher daytime
levels of NO in the Bakers
eld study, levels that are closer to those observed in
New York City in summer 2001.
15
This di
ff
erence in
uences the
P
(O
3
) sensitivity
to NO
x
and VOCs.
Comparison of measured to modeled OH reactivity
The median measured OH reactivity was greatest at night when the atmospheric
boundary layer height was lowest and was least during the late a
ernoon when
the boundary layer height was greatest (Fig. 5). The median measured OH reac-
tivity was as high at 26 s

1
at 4:00, dropped from 15 s

1
to 11 s

1
between 8:00 and
10:00, then slowly decreased to 9 s

1
at 17:00, a
er which it began a slow increase
to nighttime values. Mean values are

20% higher because of the few large spikes.
From day-to-day, OH reactivity varied from 5 s

1
to more than 50 s

1
at night and
from about 3 to 20 s

1
during the day. The lowest values were on the few cool rainy
days in May. The spikes in OH reactivity generally correlate to NO spikes, sug-
gesting that a large fraction of OH reactivity in Bakers
eld is due to anthropogenic
emissions and combustion,
41
but oxidation products, potentially from biogenic
emissions from natural and agricultural sources, have large contributions to
reactivity with high ambient temperatures.
25
This measured OH reactivity is compared to the OH reactivity calculated from
the measured chemical species and the modeled products of those measured
chemical species. The ratio of model-calculated OH reactivity to measured OH
reactivity is 0.59 at night and 0.53 during the day. In a cooler, rainy period at the
beginning of the study (before day 150, 30 May), measured and calculated OH
reactivity agreed to within 10%. The missing OH reactivity tends to be greater
when the temperature and ozone are higher, but these correlations are weak,
suggesting that other factors such as unmeasured species may be contributing to
the measured OH reactivity.
If only measured OH reactants are included in the calculated OH reactivity
total, the calculated OH reactivity is only slightly less (<10%) than the calculated
OH reactivity that includes modeled chemical species. According to the model,
80% of the calculated OH reactivity is caused by 13 chemical species, with the
most important being NO
2
(22%), carbon monoxide (11%), formaldehyde (6%),
ethanol (8%), methanol (5%), and heptanal + nonanal (5%). The biogenic VOCs
isoprene and limonene each contribute less than 2%, thus indicating the domi-
nance of anthropogenic VOCs in the calculated OH reactivity during this study.
25
From this analysis, when NO is high, NO is a major OH sink, accounting for more
than half the calculated OH reactivity in some plumes.
The OH reactivity is an indicator of the OH lifetime. For CalNex-SJV, the life-
time was 40 ms during the night and at sunrise and

100 ms during the day. With
winds of 0.5 m s

1
during the early morning and 4 m s

1
during the rest of the
day, OH achieved steady state (>3 lifetimes) for air that had travelled less than
Paper
Faraday Discussions
This journal is © The Royal Society of Chemistry 2016
Faraday Discuss.
,2016,
189
,169
–
189 |
179
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
2 meters. However, this view is too simplistic; not only must the lifetime of OH be
considered but also the lifetime of OH sources
42
and sudden changes in OH
sources and sinks.
The e
ff
ects on the OH value due to its sources and sinks can be seen by
suddenly turning o
ff
the photolytic sources of OH and other radicals in a model.
Using CalNex-SJV conditions and chemical species, a photochemical box model
was run until it achieved steady state and then photolysis was terminated and the
run continued for another 100 seconds to follow the decays of OH, HO
2
, NO, and
other short-lived chemical species (Fig. 6). Initially OH drops at a rate about equal
to the OH reactivity until the OH loss rate matches the OH production rate from
recycling, primarily HO
2
+ NO in this case. The OH decay then roughly parallels
the decay of its primary source
–
HO
2
times NO. The half-life of NO is 22 s, HO
2
110 s, and OH 13 s, which is two orders-of-magnitude longer than indicated by the
OH reactivity. In this case, the HO
2
half-life is much longer than the NO half-life,
but in the morning in CalNex-SJV, the half-life of NO is 60 s, HO
2
13 s, and OH 8 s.
Thus, while OH is in steady state with its sources and sinks in less than a second,
its abundance is tied to the half-life of its sources as they respond to changes in
their sources or sinks.
Typically in model-to-measurement HO
x
comparisons, the model is con-
strained to NO, NO
2
,O
3
, and all other measured chemical species except OH and
HO
2
. We have used this method in the analysis for this paper. The model
calculates the steady-state values for OH and HO
2
. However, HO
2
may not be in
steady state because of surface deposition, upwind cloud shadowing, or other
HO
x
production or loss that occurred within the tens of seconds prior to the
measurement. For our example, the typical a
ernoon wind was 4 m s

1
, so that in
the HO
2
half-life of 110 s, the HO
2
in an air parcel could be a
ff
ected during its
travel over a distance of almost half a kilometer. Thus HO
2
could have values that
are quite di
ff
erent from those calculated by a steady-state model constrained to
the measurements made at the
eld site.
Fig. 6
Normalized modeled decays of OH (solid blue), HO
2
(dashed black), NO (dotted
red), HO
2
times NO (dot-dashed green), and inverse of OH reactivity (leftmost dotted
blue). This case is initialized with O
3
(60 ppbv), NO
2
(2.6 ppbv), NO (1.2 ppbv), and OH
reactivity (10 s

1
).
Faraday Discussions
Paper
180
|
Faraday Discuss.
,2016,
189
,169
–
189
This journal is © The Royal Society of Chemistry 2016
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
OH production and loss
A critical consistency check of the measurements of OH, HO
2
, and OH reactivity is
the balance between production and loss for OH. The loss rate for OH is product
of the measured OH and OH reactivity; the production rate is all the sum of all the
OH sources, including recycling, mostly from HO
2
+ NO, and primary production,
mostly from O
3
photolysis followed by O(
1
D) + H
2
O and HONO photolysis. Most of
these quantities and reaction rate coe
ffi
cients are measured, making this
consistency check very close to being independent of the model. The mean
measured and modeled OH production and loss rates peaked at (5
–
9)

10
7
cm

3
s

1
, (Fig. 7(a)). This OH production is lower by a factor of two compared to other
US cities and by a factor of 6 compared to Mexico City in 2003.
43
As in previous
studies, measured OH production and loss match in the a
ernoon but not in the
morning. From sunrise to almost noon, measured OH production and loss are
both two to three times the modeled production/loss and measured OH
production is about twice measured OH loss. Because of the short OH lifetime,
these two must always match.
To resolve this discrepancy, one possibility is that the OH reactivity is biased
low when ambient NO is high, which can result from two factors. First, the high
OH reactivity in the morning reduces even the initial OH signal by about a decade
compared to midday values, so that the low signal-to-noise of the decay and the
uncertainty in the subtracted OH background signal strongly in
uence the
calculated OH decays. Laboratory experiments indicate that too little background
signal tends to get subtracted when the OH signal is small, causing the calculated
decay to be smaller than it should be. Second, the correction factor to account for
the recycling of OH by the reaction of HO
2
and ambient NO in the OH reactivity
instrument is dependent on NO and the OH reactivity, but it is typically 1.05 to
Fig. 7
Median diel variation of production and loss. (a) OH (10
7
cm

3
s

1
), measured OH
production (blue circles) and loss (red triangles), measured OH production using HO
2
*
instead of HO
2
(green stars), modeled production and loss, which are equal (black
squares), and individual 10 min production data using measured HO
2
(gray dots); (b) ozone
production from HO
2
(pptv),
P
(O
3
)
HO
2
calculated from measured HO
2
(blue circles),
modeled HO
2
(red squares), and measured HO
2
*
(green stars), with individual 10 min data
for
P
(O
3
)
HO
2
calculated from measured HO
2
(gray dots).
Paper
Faraday Discussions
This journal is © The Royal Society of Chemistry 2016
Faraday Discuss.
,2016,
189
,169
–
189 |
181
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
1.1 for 1 ppbv of NO and increases to 6
–
10 for 100 ppbv of NO. The uncertainty in
the correction factor is estimated to be

40% (2
s
con
dence), based on repeated
laboratory decays, although we think that this uncertainty can be cut in half. The
combination of the tendency for insu
ffi
cient background subtraction at low OH
signals and uncertainty in the NO correction factor could be responsible for this
discrepancy in the balance between OH production and loss. The other possibility
is that the HO
2
measurement is in error, which is discussed in a following section.
Ozone production rate
The ozone production from HO
2
,
P
(O
3
)
HO
2
, is only about half the total
P
(O
3
), the
other portion coming from RO
2
. We will con
ne our discussion only to calculated
P
(O
3
)
HO
2
because HO
2
was measured and RO
2
was not. If HO
2
is greater than
expected at higher NO values, then by eqn (1.2),
P
(O
3
)
HO
2
should also continue to
increase for NO greater than

1 ppbv, in contrast to the modeled value, which
peaks when NO is near 1 ppbv and then decreases. This di
ff
erence in measured
and modeled HO
2
translates directly into a di
ff
erence in
P
(O
3
) values that are
calculated from measured and modeled values of OH and HO
2
(Fig. 7(b)).
P
(O
3
)
HO
2
calculated from measured HO
x
is lower than
P
(O
3
)
HO
2
calculated from modeled
HO
x
below

1 ppbv and then becomes more than a decade larger for NO
¼
10
ppbv.
Most of this greater-than-expected
P
(O
3
)
HO
2
occurs in the morning before
10:00. The cumulative median
P
(O
3
)
HO
2
shown in Fig. 7(b) is 55 ppbv for the
model and 97 ppbv for the measurement. This cumulative production can be
compared with the mean diel peak O
3
of 58 ppbv (range

20 ppbv).
P
(O
3
) is about
twice
P
(O
3
)
HO
2
, so the cumulative
P
(O
3
) calculated from both modeled and
measured HO
2
are more than twice the observed ozone. At any given location,
neither the calculated instantaneous nor cumulative ozone production are
necessarily related to the peak ozone, which comes from not only the local
production but also from production throughout the planetary boundary layer
and from transport (eqn (1.1)). The calculated
P
(O
3
)
HO
2
suggests that the Cal-
Nex_SJV site is in an ozone source region that contributes to the ozone production
in the Bakers
eld plume as it moves south during the day.
How does this di
ff
erence in measured and modeled
P
(O
3
)
HO
2
a
ff
ect the
assessment of NO
x
-sensitivity or VOC-sensitivity? In four urban areas, both
measured and modeled
P
(O
3
)
HO
2
were strongly VOC-sensitive during morning
rush hour, but switched to weakly VOC-sensitive or NO
x
-sensitive from mid-
morning to mid-a
ernoon.
43
For CalNex-SJV, mean midday NO was typically
1.5 ppbv, so that measured HO
2
was less than twice modeled HO
2
. As a result, the
greater-than-expected
P
(O
3
)
HO
2
that occurs primarily in the morning does not
signi
cantly change the
P
(O
3
) sensitivity to NO
x
or VOCs during the 10:00 to
14:00 period used by Pusede
et al.
25
in their analysis.
Possible cause of greater-than-expected HO
2
and
P
(O
3
)
HO
2
What could be the cause of the greater-than-expected HO
2
and
P
(O
3
)
HO
2
when NO
exceeds 1 ppbv? We look at the
ve hypotheses presented in the Introduction.
A. HO
2
measurement error.
Measured HO
2
could exceed modeled HO
2
if
there were an error in the absolute HO
x
calibration. However, the HO
2
calibration
is tied to the OH calibration and measured OH is approximately equal to the
Faraday Discussions
Paper
182
|
Faraday Discuss.
,2016,
189
,169
–
189
This journal is © The Royal Society of Chemistry 2016
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
modeled OH. In addition, measured and modeled HO
2
agree within uncertainties
in the a
ernoon, especially if HO
2
is close to the lower limit of possible values.
Second, HO
2
is detected by adding internally a few hundred ppmv of NO to
convert HO
2
to OH, so the 10 ppbv of atmospheric NO is overwhelmed by the
injected NO. Third, a small HO
2
signal o
ff
set is not responsible because the
general di
ff
erences between measured and modeled HO
2
in Fig. 4 persists even
when an o
ff
set of 2 pptv is subtracted from the measured HO
2
. Such an o
ff
set is
not observed and is about a decade higher than the limit-of-detection. Fourth, the
RO
2
interference due to alkene and aromatic peroxyl radicals is also unlikely for
two reasons. The RO
2
is more rapidly converted to HO
2
as NO increases, but the
HO
2
is recycled with OH, so the RO
2
-to-HO
2
ratio decreases as NO increases
(Fig. 4(b)). Next, a previous study showed these unexpectedly high HO
2
values at
higher NO values in the upper troposphere where RO
2
abundances are small.
14
Further evidence is provided by other studies using other instruments and even
another HO
2
measurement technique.
9,15,16,18
So far, no HO
2
measurement error
has been found that would cause the greater-than-expected decrease in
P
(O
3
)
HO
2
at high NO.
B. Measurement averaging.
Does the average of measured HO
2
times the
average of measured NO equal the average of HO
2
times NO? This question can be
answered by using high-resolution HO
2
and NO measurements and averaging
them together. The full width/half maximum (FWHM) of the NO plumes varied
from a few seconds to well over an hour. However, for NO plumes with NO greater
than 5 ppbv, 80% of time was in plumes that were sampled for more than
20 seconds. As a result, we will use 20 seconds as the minimum time resolution
for comparing the products of the averages to the averages of the products.
The averages of HO
2
and NO were calculated for 20 seconds, 1 minute,
10 minutes, and 1 hour. These four di
ff
erent averages were plotted as a function
of NO for HO
2
in Fig. 4 and for calculated
P
(O
3
) in Fig. 8. There are no signi
cant
di
ff
erences among the four di
ff
erent averages. Thus, the greater-than-expected
HO
2
and
P
(O
3
) at higher NO are not due to averaging over NO plume spikes.
In previous urban studies, point measurements have been compared to long-
path absorption measurements to test the assumption that point measurements
can adequately represent the integrated chemistry in urban plumes for
measurement integration times of minutes or more. In a 1999 study in Nashville
TN, long-path and point measurements generally agreed to within 15% or
better.
44
In Mexico City in April 2003, point and long-path measurements of
several VOCs were in good agreement for time scales of 5 minutes or more,
although not for all VOCs.
45,46
While these averages are in space while our aver-
ages are in time, the conclusion is that, if there are no unique emission sources
nearby, point measurements can represent the photochemistry of an urban
region.
C. Conditions when NO
x
is not in photostationary state.
The theoretical
steady-state curve for the HO
2
and
P
(O
3
)
HO
2
versus
NO assumes that HO
x
production is constant and NO
x
is in photostationary state (PSS) as NO changes.
Even though NO emissions are typically out of PSS for at most tens of seconds
during the day, they are continually coming from thousands of sources and these
plumes are continually contributing to ozone production. The analysis presented
here makes no assumptions about NO
x
PSS. As a result, the measured and
modeled curves in Fig. 4 and 8 are averages over di
ff
erent HO
x
production regimes
Paper
Faraday Discussions
This journal is © The Royal Society of Chemistry 2016
Faraday Discuss.
,2016,
189
,169
–
189 |
183
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
and may include measurements during conditions when NO
x
is not in PSS.
However, while NO
x
may be out of PSS if we constrain the model-to-measured NO,
NO
2
, and O
3
, the model does calculate steady-state values for HO
2
and OH. So
a valid comparison between measured and modeled HO
x
requires that HO
x
be in
steady state but not NO
x
.
We tested how close NO
x
was to PSS using 1 minute averaged measurements.
Comparisons were made between the model with NO, NO
2
, and O
3
constrained
(
i.e.
, measurements and no assumption of NO
x
PSS) to calculations of NO
x
PSS.
These comparisons show that NO
x
was generally within 10
–
20% of PSS from 8:00
to 14:00, suggesting that NO had decreased about e

2
¼
0.14 from its initial
emission. The typical NO lifetime is 20
–
60 seconds. When these lifetimes are
multiplied by the wind speed and by 2, the resulting product is the approximate
distance the air must travel from a NO source so that the NO has decreased to 0.14
of its initial value. This distance was in the range of 100
–
150 m, a distance great
enough that several sources could contribute to keeping the NO
x
slightly out of
photostationary state. Because NO
x
was close to steady state, HO
2
was likely to be
in steady state in the morning when its lifetime was 13 s but not in the a
ernoon
when its lifetime was 110 seconds. However, since the greater-than-expected HO
2
occurs primarily in the morning when HO
2
is in steady state, NO
x
being out of PSS
is not the cause of the greater-than-expected HO
2
and calculated
P
(O
3
)
HO
2
.
D. Unknown HO
2
sources accompanying NO emissions.
Any unknown HO
x
sources accompanying NO should show up in the OH reactivity measurement and
in the comparison of the balance between OH production and loss (Fig. 7(a)). If
the accompanying HO
x
source is an HO
2
source, such as an aldehyde, then it
would also likely be an OH loss and appear in the OH reactivity. However, the
agreement between measured and calculated OH reactivity is better when NO is
high than when NO is low; calculated OH reactivity accounts for

50% of
Fig. 8
Median behavior of the ozone production rate due to HO
2
,
P
(O
3
)
HO
2
, as a function
of NO. 10 minute averages of measured (blue circles) and modeled (red squares)
P
(O
3
)
HO
2
are plotted along with 20 second (upward pointing triangles), 1 minute (diamonds), and 1
hour (downward pointing triangles) averages. Gray dots are individual 10 min measured
data; darker gray dots are data measured between 12:00 and 15:00 when the wind was
from the north, although the points with greatest NO generally occurred when the winds
were light and out of the east. Data are
fi
ltered for daytime hours between 7:00 and 17:00.
Faraday Discussions
Paper
184
|
Faraday Discuss.
,2016,
189
,169
–
189
This journal is © The Royal Society of Chemistry 2016
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
measured OH reactivity when NO was less than 1 ppbv but as much as 75% when
NO is greater than 10 ppbv, just the opposite of expectations for an unknown HO
2
source that is also an OH loss. Furthermore, an unknown HO
2
source that acts
also as an OH loss would cause the measured OH production to exceed the OH
loss calculated from the known HO
2
sources, but just the opposite is seen. These
arguments are based on OH reactivity measurements that have an uncertain NO
correction, so an unknown HO
2
source cannot be entirely ruled out by these
observations.
E. Missing chemical mechanism
For missing chemistry to cause the greater-than-expected HO
2
dependence on
NO, it must have a few characteristics. First, in order for
P
(O
3
) to increase with NO
as shown in Fig. 8, the HO
2
production must be approximately proportional to NO
because the di
ff
erence between measured and modeled
P
(O
3
)
HO
2
increases about
a decade for each decade of increase in NO (Fig. 8). Second, the HO
2
production
rate from this chemistry would need to be approximately 5

10
7
cm

3
s

1
in
order to balance the HO
2
loss due to the reaction of HO
2
with NO. This value
comes from the di
ff
erence in the OH production with measured and modeled
HO
2
(Fig. 7(a)). We know of no known chemical mechanism that can satisfy these
two constraints. Yet with the uncertainty associated with the products of reactions
such as OH + NO
2
, unknown HO
x
-NO
x
chemistry is a reasonable possibility for the
greater-than-expected HO
2
and calculated
P
(O
3
)
HO
2
.
When we
rst observed this HO
2
discrepancy a decade ago, we hypothesized
that it might be due to the reaction sequence
OH + NO
/
HONO
*
(1.6)
HONO
*
+N
2
/
HONO
(1.7)
HONO
*
+O
2
/
HO
2
+NO
2
(1.8)
The reaction to products HO
2
+ NO is exothermic,
D
H
¼
94 kJ mol

1
, and, to
our knowledge, has never been tested. All studies of OH + NO have been done at
either low pressure or in the absence of molecular oxygen, according to Sander
et al.
and references therein.
47
It is a di
ffi
cult reaction to study because the HO
2
produced recycles immediately to OH by HO
2
+NO
/
OH + NO
2
.Itisdi
ffi
cult to
scavenge HO
2
in laboratory studies, so this recycling would appear as a slower rate
coe
ffi
cient for OH + NO + M, although laboratory studies of HONO formation have
generally taken known recycling into account. This reaction can explain the
observed HO
2
discrepancy only if the formation of HONO
*
is sped up in the
presence of oxygen.
There is some evidence that this reaction could be occurring. Molecular energy
calculations suggest that vibrationally excited HONO (
n
OH
$
3) can react with O
2
to form HO
2
+NO
2
.
48
While this mechanism is a minor HO
2
formation pathway if
HONO
*
is produced by the photo-excitation of HONO, it could be substantially
greater if HONO
*
is produced by the OH + NO reaction. Other calculations suggest
that the H atom can quickly hop from O atom to O atom in sub-nanosecond time
scales, about
ve to ten times faster than the time between collisions with
molecular oxygen and that the H atom could transfer to a colliding O
2
molecule.
49
Paper
Faraday Discussions
This journal is © The Royal Society of Chemistry 2016
Faraday Discuss.
,2016,
189
,169
–
189 |
185
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online
In another study of the OH + acetylene reaction, O
2
reacted with about 25% of the
excited-state adducts before they could be collisionally relaxed.
50
These studies
suggest that it may indeed be possible for this reaction sequence to occur.
The results of our laboratory studies on this possible reaction have been
decidedly mixed. However, the reaction would need to proceed with an e
ff
ective
bimolecular reaction rate of (3
–
15)

10

11
cm
3
molecule

1
s

1
in order to explain
the HO
2
/OH ratio observed in this and previous ground-based studies made with
GTHOS.
Conclusions
CalNex-SJV provides a good opportunity to compare measured and modeled
oxidation chemistry in an environment with plumes of NO
x
and their accompa-
nying VOCs and OVOCs. The diel variation of the median measured and modeled
OH and HO
2
were generally well represented by the box model constrained by
other simultaneous measurements except for HO
2
during morning rush hour. As
in many previous studies, the measured HO
2
and the ozone production rate
calculated from measured HO
2
decreased much more slowly than the modeled
HO
2
and calculated
P
(O
3
)
HO
2
as a function of NO. The amount of missing OH
reactivity was roughly 45% of the total measured OH reactivity. Also as seen in
most previous studies, the OH production and loss rates balance in the a
ernoon
and at night, but in the morning, the production greatly exceeds the loss. These
discrepancies could arise from issues with transport or issues with chemistry.
The presence of frequent NO
x
plumes that lasted from seconds to an hour
suggests that transport timescales were comparable to chemical timescales for
radicals like NO and HO
2
. But even with strong winds (4 m s

1
) from a major
highway and the urban core, NO
x
was close to a photostationary state as it was
sampled at the
eld site. So, despite the heterogeneity of the sampled air masses,
the comparison between the measured and modeled OH and HO
2
was not
signi
cantly a
ff
ected by these plumes in the morning, but may have been a
ff
ected
in the a
ernoon.
When the OH reactivity is between 5 and 25 s

1
, as it was in this study, OH
comes into balance with a change in its sources and sinks in much less than
a second. However, OH will change only as fast as its sources and sinks change. In
this study, the main OH source was HO
2
+ NO, so OH had a half-life of 13 seconds
because HO
2

NO had a half-life of 16 s. The results from this study show that
point measurements can provide a valid test of urban oxidation chemistry if care
is taken to ensure that HO
2
is in steady state. NO
x
does not have to be in steady-
state if NO, NO
2
, and O
3
are constrained in the steady state model along with all
other simultaneous measurements.
Greater-than-expected HO
2
at high NO results in higher ozone production
calculated from measured HO
2
, but it does not strongly in
uence the assessment
of midday ozone production sensitivity to NO
x
or VOCs. In the morning ozone
production is VOC-limited whether modeled or measured HO
2
is used to calculate
ozone production. At midday the di
ff
erence between measured and modeled HO
2
is less, causing at most a small change in the assessment of NO
x
-sensitivity or
VOC-sensitivity for ozone production.
Unknown HO
x
–
NO
x
chemistry emerged as one of the most likely causes for the
greater-than-expected measured HO
2
seen in this and several other previous
Faraday Discussions
Paper
186
|
Faraday Discuss.
,2016,
189
,169
–
189
This journal is © The Royal Society of Chemistry 2016
Published on 23 December 2015. Downloaded by California Institute of Technology on 29/07/2016 18:46:04.
View Article Online