Observations of atmospheric oxidation and ozone production in South Korea

Atmospheric Environment 269 (2022) 118854

Available online 20 November 2021

(

http://creativecommons.org/licenses/by-nc-nd/4.0/

).

Observations

of atmospheric

oxidation

and ozone production

in

South Korea

William

H. Brune

a

,

*

, David O. Miller

a

, Alexander

B. Thames

a

, Alexandra

L. Brosius

a

,

Barbara

Barletta

b

, Donald

R. Blake

b

, Nicola J. Blake

b

, Gao Chen

c

, Yonghoon

Choi

c

,

James H. Crawford

c

, Joshua P. Digangi

c

, Glenn Diskin

c

, Alan Fried

d

, Samuel

R. Hall

e

,

Thomas

F. Hanisco

f

, Greg L. Huey

g

, Stacey C. Hughes

b

, Michelle

Kim

h

, Simone

Meinardi

b

,

Denise D. Montzka

e

, Sally E. Pusede

i

, Jason R. Schroeder

c

,

1

, Alex Teng

h

, David J. Tanner

g

,

Kirk Ullmann

e

, James Walega

d

, Andrew

Weinheimer

e

, Armin Wisthaler

j

,

k

, Paul O. Wennberg

h

a

Department

of Meteorology

and Atmospheric

Science,

Pennsylvania

State University,

University

Park,

PA, USA

b

Department

of Chemistry,

University

of California,

Irvine,

CA, USA

c

NASA

Langley

Research

Center,

Hampton,

VA, USA

d

Institute

of Arctic

and Alpine

Research,

University

of Colorado,

Boulder,

CO, USA

e

National

Center

for Atmospheric

Research,

Boulder,

CO, USA

f

NASA

Goddard

Space

Flight

Center,

Greenbelt,

MD, USA

g

School

of Earth

and Atmospheric

Sciences,

Georgia

Institute

of Technology,

Atlanta,

GA, USA

h

Division

of Engineering

and Applied

Sciences,

California

Institute

of Technology,

Pasadena,

CA, USA

i

Department

of Environmental

Sciences,

University

of Virginia,

Charlottesville,

VA, USA

j

Institute

for Ion Physics

and Applied

Physics,

University

of Innsbruck,

AT, USA

k

Department

of Chemistry,

University

of Oslo, Oslo, Norway

HIGHLIGHTS

•

For South Korea, observed

and modeled

OH and HO

2

agree to within uncertainties.

•

Modeled

aerosol

uptake of hydroperoxyl

is inconsistent

with observed

hydroperoxyl.

•

Missing

OH reactivity

came from Korea and increased

from spring to summer.

•

Observed

ozone changes

are consistent

with calculated

ozone production.

ARTICLE

INFO

Keywords:

Air quality

Hydroxyl

Hydroperoxyl

Aerosol

uptake of hydroperoxyl

Missing

OH reactivity

Ozone production

rate

ABSTRACT

South Korea routinely

experiences

poor air quality with ozone and small particles

exceeding

air quality stan

-

dards. To build a better understanding

of this problem,

in 2016, the KORea-United

States cooperative

Air Quality

(KORUS-AQ)

study collected

surface

and airborne

measurements

of many chemical

species,

including

the

reactive

gases hydroxyl

(OH) and hydroperpoxyl

(HO

2

). Several different

results are reported

here. First, OH and

HO

2

measured

on the NASA DC-8 agree to within uncertainties

with values calculated

by two different

box

models,

both in statistical

comparisons

and as a function

of altitude

from the surface to 8 km. These comparisons

show substantial

scatter,

likely due to both variability

in instrument

performance

and the difficulty

in interpo

-

lating measurements

made with frequencies

different

from those of the model time step. Second,

OH and HO

2

calculated

by a model including

HO

2

uptake on aerosol

particles

in the chemical

mechanism

are inconsistent

with observations.

Third, in the planetary

boundary

layer over both ocean and land, measured

and model-

calculated

OH reactivity

are sometimes

different,

and this missing

OH reactivity,

which is as much as ~4 s

1

,

increased

from April to June and originated

primarily

from the Korean peninsula.

Fourth,

repeated

missed ap

-

proaches

at the Seoul Air Base during several days show that the changes

in the sum of ozone and nitrogen

* Corresponding

author.

E-mail

address:

whb2@psu.edu

(W.H. Brune).

1

Now at California

Air Resources

Board, Sacramento,

CA, USA.

Contents

lists available

at ScienceDirect

Atmospheric

Environment

journal

homepag

e:

www.else

vier.com/loc

ate/atmos

env

https://doi.org/10.1016/j.atmosenv.2021.118854

Received

9 July 2021; Received

in revised form 9 November

2021; Accepted

13 November

2021

Atmospheric Environment 269 (2022) 118854

2

dioxide

are consistent

with ozone production

rates calculated

from HO

2

either observed

or modeled

by the

Langley

Research

Center model.

1. Introduction

South Korea has an air quality problem

(Susaya et al., 2013; Jung

et al., 2018; Kim and Lee, 2018; Schroeder

et al., 2020). While the

harmful

small particle

pollution

is easy to see, the harmful

ozone

pollution

lurks unseen.

Air quality standards

are routinely

exceeded

in

ozone, nitrogen

dioxide,

and small particles

(PM

2.5

), particularly

for the

Seoul metropolitan

area (SMA) (IQAir, 2021). Of the ~52 million South

Koreans,

about half live in the SMA, which encompasses

substantial

industrial

activity

and vehicular

traffic. In 2016, over half of freight

trucks and almost half of cars used diesel fuel, contributing

to the poor

air quality in terms of particles

and ozone (Reuters,

2021). Since then

the proportion

of diesel vehicles

is beginning

to decrease,

but air

pollution

problems

remain.

Models

used to diagnose

air quality contain

ozone chemical

mech

-

anisms that have been developed

over decades

and tested against

ob

-

servations

in environmental

chambers

and the atmosphere.

The well-

known chemistry

contained

in these models

starts with emissions

of

nitrogen

oxides (NO

x

=

NO

+

NO

2

) from combustion

and volatile

organic

compounds

(VOCs)

from hundreds

of sources.

The VOCs are initially

oxidized

by the hydroxyl

radical (OH), creating

organic

peroxyl

radicals

(RO

2

).

These radicals

react with nitric oxide (NO) to form nitrogen

dioxide

(NO

2

) and another

organic

peroxy radical,

RO

2

. RO

2

can react with NO

to form NO

2

and RO, from which O

2

extracts

a hydrogen

to form the

hydroperoxyl

radical (HO

2

). HO

2

then also reacts with NO to produce

NO

2

. This NO

2

produced

by RO

2

and HO

2

adds to the NO

2

produced

in

the reaction

of NO with O

3

. The rapid daytime

photo-destruction

of NO

2

results in NO and O

3

, and within minutes,

these reactions

set up a bal

-

ance among O

3

, NO

2

, and NO (with a small contribution

HO

2

and RO

2

in

the balance).

The new O

3

that is produced

by these reactions

rapidly

partitions

between

O

3

and NO

2

. In addition,

any NO

2

produced

directly

by sources,

such as diesel engines,

also quickly

partitions

into NO

2

and

O

3

. Thus, examining

the chemical

processes

that produce

O

3

requires

analyzing

the changes

not in just O

3

but in the sum of O

3

and NO

2

, called

O

x

.

Chemical

transport

models

combine

the effects of chemistry

and

weather

in an attempt

to simulate

observed

O

3

in urban and surrounding

areas. Many studies

show that they can be excellent

at simulating

average

O

3

levels, but they tend to calculate

less O

3

at O

3

levels that

exceed air quality standards,

precisely

where their ability to simulate

O

3

is needed the most (Appel et al., 2007; Im et al., 2015). Several

studies

focused

on Southeast

Asia also find that the models tend to underpredict

hazardous

O

3

levels (Kang et al., 2016; Oak et al., 2019; Le et al., 2020,

Fig. 4; Park et al., 2021). Usually

this model failure is attributed

to some

problem

in the model meteorology,

emissions

inventory,

or methodol

-

ogy. Counter

to these studies are those involving

the weekend-weekday

effect, wherein

O

3

increases

on the weekends

when NO

x

levels are lower

than on weekdays

(Harley et al., 2005; Pollack

et al., 2013). However,

when O

x

is considered

as opposed

to O

3

, the recent weekend

effect is

minimized

to just the partitioning

between

O

3

and NO

2

with NO

x

less on

weekends

so that O

3

is greater (de Foy et al., 2020).

There is a possibility

that the well-known

O

3

chemistry

is incom

-

plete. Most urban measurements

of HO

2

show that, for increasing

NO,

the observed

HO

2

becomes

increasingly

greater than that modeled

with

a photochemical

box model (Brune et al., 2016). Such an increase

in HO

2

radicals

would lead to more O

3

production

than expected

at higher NO

and could resolve

the discrepancy

noted in the previous

paragraph.

Brune et al. suggested

the reaction

OH

+

NO

+

O

2

→

HO

2

+

NO

2

, but,

despite

being energetically

feasible,

it has transition-state

energy bar

-

riers that prevent

it from occurring

(Fittschen

et al., 2017). Thus, the

cause of this widely observed

greater-than-expected

HO

2

at higher NO

amounts

remains

unresolved.

Aerosol

particles

can interact

with gas-phase

reactive

gases and solar

ultraviolet

radiation

to alter O

3

production.

A recent rise in O

3

pollution

in China has been attributed

to a decrease

in small particles

(Li et al.,

2018). The chemistry

involves

the uptake of HO

2

on the particle

sur

-

faces, which, depending

on the metal content

of the particles,

will lead

to the production

of water vapor or hydrogen

peroxide

(H

2

O

2

) (Mao

et al., 2013). An HO

2

uptake with an effective

uptake coefficient

of 0.2 in

the GEOS-Chem

chemical

transport

model is sufficient

to explain

the

observed

O

3

increase

over the past four years in China as the aerosol

particle

amounts

have decreased.

It is therefore

possible

that HO

2

uptake

on aerosol particles

over South Korea is influencing

HO

2

and OH just as

it appears

to be doing over China.

The rich variety of VOCs and high levels of NO and NO

2

emitted

in

South Korea are combined

with the aged pollution

coming

from China,

providing

a challenge

to model calculations

of OH and HO

2

. The focus of

this paper tests two of model mechanisms

using airborne

measurements

of OH, HO

2

, NO, O

3

, and ~100 other chemical

species during flights

over South Korea and its environs

in May to June 2016 during the

KORea-US

cooperative

Air Quality

(KORUS-AQ)

field study (Crawford

et al., 2021). Also measured

was the OH reactivity,

which is the fre

-

quency at which the sum of all atmospheric

chemical

species react with

OH. These measurements

enable, for the first time, a solid test of model

calculations

of atmospheric

oxidation,

including

calculated

ozone pro

-

duction

rates, in South Korea

’

s chemically

complex

environment.

2. Methods

2.1. KORUS-AQ

KORUS-AQ

focused

on the air quality

of South Korea, with an

emphasis

on the SMA. The study included

enhanced

ground

sites that

complemented

the existing

National

Institute

of Environmental

Research

(NIER) air quality

network

and four aircraft,

including

the

NASA DC-8. The DC-8 was based in Osan, 47 km south of the center of

Seoul. The results of this paper come primarily

from measurements

made on the NASA DC-8, which conducted

20 flights over, upwind,

and

downwind

of the South Korean peninsula.

The DC-8 carried a suite of instruments

that could measure

hundreds

of chemical

species,

aerosol

properties,

photolysis

frequencies,

and

meteorological

conditions

(Schroeder

et al., 2020, Table 2). These

measured

chemical

species

include

O

3

, NO, NO

2

, CO, formaldehyde

(CH

2

O), peroxyacetyl

nitrate,

and many VOCs, including

some C

7

and

larger aromatics,

and oxygenated

VOCs. In addition,

the DC-8 also had

measurements

of the hydroxyl

radical (OH), the hydroperoxyl

radical

(HO

2

), and the OH reactivity,

which is the inverse

of the OH lifetime.

This extensive

measurement

suite enabled

comparisons

between

modeled

and measured

OH, HO

2

, OH reactivity,

and calculated

net O

3

production

rate (PO

3

).

A common

flight pattern emerged

in which the DC-8 would take off

from Osan in the morning,

and then do an early missed approach

at the

Seoul Air Base, thus performing

an altitude

profile through

the mixed

layer. The Seoul Air Base is located 7 km southeast

of the city center. The

DC-8 would then execute

the rest of the flight pattern

for the day,

whether

it would be upwind

or down through

flight corridors

over South

Korea. Often, the DC-8 would execute

another

missed approach

at the

Seoul Air Base near midday

and then again in the late afternoon

near the

end of the flight. We call them early, midday,

and late throughout

the

rest of the paper. In this paper, we will use these missed approaches

to

compare

the observed

O

x

changes

to that calculated

from the ozone

W.H. Brune

et al.

Atmospheric Environment 269 (2022) 118854

3

production

rates using the measured

and modeled

HO

2

and the modeled

RO

2

.

2.2. OH and HO

2

measurements

OH and HO

2

were both measured

using the Penn State Airborne

Tropospheric

Hydrogen

Oxides Sensor (ATHOS)

instrument,

which uses

laser induced

fluorescence

in low-pressure

(3

–

13 hPa) detection

cells to

detect OH in one cell and, in a second cell connected

to the first, HO

2

,

which is converted

to OH by reaction

with NO added between

the two

cells. ATHOS

is discussed

extensively

in Faloona

et al. (2004) and as

configured

and operated

during KORUS

in Brune et al. (2020). For

KORUS-AQ,

OH and HO

2

mixing ratios are reported

each 30 s.

The OH signal was separated

from background

signals by two ways:

subtracting

the signal when the laser wavelength

was not resonant

with

an OH absorption

line from the signal when it is and scrubbing

the OH

by reacting

it with perfluoropropylene

in a special inlet attached

to the

low-pressure

detection

cell pinhole

inlet. For HO

2

, the potential

inter

-

ference

from some organic

peroxy radicals

was mitigated

by using NO

reagent

amounts

and reactions

times so that only ~28% of the HO

2

was

converted

to OH, a proven method

(Fuchs et al., 2011; Feiner et al.,

2016). A detailed

description

of these signal interference

reduction

strategies

is given in Brune et al. (2020).

ATHOS

was calibrated

in the laboratory

and monitors

of laser power

were used to maintain

that calibration

during flight. The estimated

ab

-

solute uncertainty

for both OH and HO

2

is

±

35%, 95% confidence.

The

limit-of-detection

(LOD) for OH is set by counting

statistics

and for HO

2

by small impurities

that remained

in the reagent

NO. For 1-min averages

measured

in the dark on flights during the NASA Atmospheric

Tomog

-

raphy (ATom)

mission

(Thompson

et al., 2021), the standard

deviation

of the background

mixing ratios is 0.018 pptv for OH and 0.2 pptv for

HO

2

. This standard

deviation

is taken to be the LOD. In concentration

units, the OH LOD is 4.5x10

5

cm

3

at 0 km altitude

and 1.5x10

5

cm

3

at

10 km for 1-min averages.

2.3. Measuring

atmospheric

OH free from interferences

Of the research

groups that use LIF to detect OH, we were not the first

group to attempt

to implement

a scrubbing

method

for OH on an

airborne

instrument

but we were the first to demonstrate

this method

in

a publication

(Brune et al., 2020). Some OH interference

was measured

during KORUS-AQ,

but 96% of the statistically

significant

OH interfer

-

ence (i.e., greater than 5x10

5

cm

3

) was located below 2 km altitude

in

the planetary

boundary

layer (PBL). This result is consistent

with the

results from ATom, which showed

no evidence

for OH interference

above the planetary

boundary

layer, and for ATom, very little in the PBL.

The PBL in KORUS-AQ

was more polluted

than in ATom, even over the

East China Sea, so some OH interference

might be expected.

On average,

only 28% of observed

OH below 2 km altitude

had statistically

signifi

-

cant OH interference

and the mean interferences

was only 13% of the

ambient

OH. The OH interference

was not observed

everywhere,

but it

was seen in the PBL both over land and ocean.

2.4. OH reactivity

measurements

The OH reactivity

concept

and the basic instrument

(Kovacs

and

Brune, 2001) was redesigned

for aircraft

operation

(Mao et al., 2009),

and then further modified

for use in KORUS

and ATom (Thames

et al.,

2020). During operation,

ambient

air flows into a flow tube (15 cm

diameter)

and past the pinhole

inlet to an OH low-pressure

detection

cell. Clean moist air flows past a mercury

lamp in a moveable

tube (1.2

cm diameter),

producing

OH (~10

–

100 pptv) that is then sprayed

into

the ambient

flow. The OH detection

system measures

the decay in the

OH signal as the moveable

tube gets farther from the OH detection

system,

which increases

the reaction

time.

The OH reactivity

is just the slope of the linear least squares

fit to the

logarithm

of the OH signal versus reaction

time when the measured

OH

loss due to the walls or impurities

in clean moist air. A decay was

measured

every 30 s. The absolute

uncertainty

due to uncertainties

in

fitting the linear least squares

line to the measured

OH signals and in the

offset measurement

due to wall loss and impurities

is estimated

to be

±

0.64 s

1

, 90% confidence.

This OH reactivity

instrument

measures

the

“

instantaneous

”

OH

reactivity,

which is only the reactions

occurring

within the measurement

time, which is 0.4 s. All subsequent

OH reactions

with the products

of

the first OH reaction

are typically

not measured.

However,

in environ

-

ments where NO is greater than a few ppbv, the reaction

HO

2

+

NO

→

OH

+

NO

2

is fast enough

to convert

HO

2

to OH, thereby

altering

the

observed

OH decay. For the airborne

OH reactivity

instrument

used in

KORUS,

the analysis

program

can correct for this recycling

from HO

2

to

OH for NO amounts

only below about ~5

–

10 ppbv, whereas

for the

ground-based

instrument

it is possible

to make the correction

up for NO

as high as 50

–

100 ppbv. Less than 10% of all OH reactivity

measure

-

ments were lost by this restriction

the data to times when NO was less

than 5 ppbv.

2.5. Models

The analysis

in the paper uses two different

well-established

photo

-

chemical

box models:

the NASA Langley

Research

Center photochemical

box model, here called LaRC (Crawford

et al., 1999; Olson et al., 2004,

2006; Schroeder

et al., 2020), and the F0AM modeling

system (Wolfe

et al., 2016; Wolfe, 2017) with the Master Chemical

Mechanism

v3.3.1,

here called MCM (Jenkin et al., 2003; Saunders

et al., 2003).

Briefly,

the LaRC model chemical

mechanism

has explicit

inorganic

and reaction

rate coefficient

for some grouped

VOC types. For each 1-s

time step along the DC-8 flight path, the model is constrained

by the

observed

chemical

species,

photolysis

frequencies,

and meteorological

variables

and calculates

OH, HO

2

, some organic

peroxyl

radicals,

and

some oxygenated

VOCs. It was run so that it achieved

diurnal

steady-

state, that is, the values for these chemical

species did not change for

the same conditions

24 h later. The model results are archived

on a

NASA website

(Aknan and Chen, 2017). We use the

“

constrained

”

runs

to better match the constrained

input chemical

species that we use with

the F0AM model framework

and MCM.

The MCM model has an explicit

mechanism

that includes

individual

chemical

species,

available

measured

reaction

rate coefficients,

avail

-

able photolysis

frequencies,

and estimated

reaction

rate coefficients

if

measured

ones are not available.

This mechanism

includes

several C

7

+

aromatics,

which are shown to be important

for Seoul (Schroeder

et al.,

2020). The model was run using the 1-min merge for all flights origi

-

nating for Osan and again using the 1-s merge for the missed approaches

at the Seoul Air Base. It was constrained

to all measured

chemical

spe

-

cies and meteorological

conditions

in the merges except OH and HO

2

,

which were calculated.

The model was run repeatedly

with different

integration

and dilution

times from hours to days to ensure that the chosen values did produce

calculated

OH and HO

2

values different

from the median

values by more

than 5%. While OH and HO

2

values were rather immune

to the choices

of integration

and dilution

times, the RO

2

values were much more sen

-

sitive. For the modeled

OH and HO

2

used in this work, the integration

and dilution

times were both 12 h.

For some MCM model runs, the simplified

heterogeneous

uptake

chemistry

for OH and HO

2

described

in Brune et al. (2016) was added.

This chemistry

assumes

that only the diffusion

to the particles

and the

surface uptake are important

in determining

the rate of gas-phase

loss

and does not consider

possible

surface photochemistry.

The accommo

-

dation coefficient,

α

, was set to 0.2 for both OH and HO

2

(Abbatt et al.,

2012), but previous

work shows that only the uptake of HO

2

is important

because

heterogeneous

OH uptake is much slower than the gas-phase

losses.

In separate

model runs, heterogeneous

O

3

uptake added to test the

W.H. Brune

et al.

Atmospheric Environment 269 (2022) 118854

4

possible

influence

of the heavy aerosol amounts

on O

3

. The accommo

-

dation coefficient

was set at 10

4

, but it is known to be highly dependent

on particle

composition

and the degree of surface O

3

saturation

(Abbatt

et al., 2012). The comparisons

of the results from these model runs with

observations

determines

the potential

for heterogeneous

processes

to

substantially

affect fast ozone-producing

chemistry.

2.6. Estimating

the mixed

layer height

(MLH)

The mixed layer, or convectively

stirred planetary

boundary

layer

(PBL), tends to have fairly constant

amounts

of aerosol

particles,

po

-

tential temperature,

and to a lesser extent water vapor mixing ratio as a

function

of height. In this paper, three methods

are used to determine

the MLH: altitude

of the enhanced

backscattering

by aerosol particles

of

a downward

beam from a lidar on the DC-8; altitude

at which potential

temperature

becomes

approximately

constant

to within 10% during a

missed approach

at Seoul Air Base; and altitude

derived

by the same

approach

for water vapor.

A time series of one of the missed approaches

demonstrates

these

methods

(Fig. 1). The dashed lines indicate

the times for which the water

vapor and potential

temperature

profiles

change dramatically.

The red

circles at the intercept

of the dashed lines and the altitude

profile indi

-

cate the MLH. This method

was done subjectively

for all of the 53 missed

approaches

three separate

times independently

for potential

tempera

-

ture and water vapor and then averaged.

The variability

in the multiple

measurements

indicate

a typical uncertainty

of

±

30%, 90% confidence,

in this MLH estimate.

The lidar MLH comes from the NASA merge file

(Aknan and Chen, 2017).

The MLH from the lidar agrees with that estimated

from potential

temperature

and water vapor for the midday

and afternoon

missed ap

-

proaches,

but tends to be substantially

greater for the morning

(Fig. 2).

This result should

not be surprising

because

aerosol

distributed

throughout

the higher afternoon

mixed layer would remain in the re

-

sidual layer over night and would give enhanced

backscatter

starting

at

the top of the residual

layer, not the lower morning

mixed layer. Thus,

the MLH estimated

from potential

temperature

and water vapor was

used in the analysis

of O

3

production.

2.7. Analyzing

ozone

production

(PO

3

)

The budget equation

for O

3

can be written

as

∂

O

3

∂

t

=

FO

3

DO

3

v

d

O

3

H

v

H

⋅

∇

O

3

(eq. 1)

where

∂

O

3

∂

t

is the time rate of change for O

3

during the missed approach

in

ppbv h

1

,

FO

3

is chemical

O

3

production,

DO

3

is chemical

O

3

loss,

v

d

O

3

H

is

surface loss with a deposition

velocity

v

d

and mixed layer height H, and

v

H

⋅

∇

O

3

is the horizontal

O

3

advection.

The net chemical

O

3

produc

-

tion is

PO

3

=

FO

3

DO

3

.

(eq. 2)

And

FO

3

=

k

HO

2

NO

[

HO

2

][

NO

]+

∑

k

RO

2

i

NO

[

RO

2

]

i

[

NO

]

(eq. 3)

where we have followed

the notation

in Schroeder

et al. (2020). Missing

from eq. (1) is the entrainment

term for the top of the well-mixed

planetary

layer, but we will account

for the O

3

in the morning

resid

-

ual layer in the analysis.

Because

some of the produced

O

3

quickly

partitions

into NO

2

by

reaction

with NO, the change in O

x

=

O

3

+

NO

2

needs to be compared

to

Fig. 1.

Method

for determining

mixed layer height (MLH).

Missed approach

time series for altitude

(black line), scaled water vapor mixing ratio (red line),

potential

temperature

(orange

line), and DC-8 lidar determined

MLH (purple

line). Dashed

lines show times when water vapor and potential

temperature

flatten to within ~10%, indicating

a well-mixed

mixing layer. The intersection

of the altitude

and the dashed lines (red circles) are estimates

of the MLH. (For

interpretation

of the references

to color in this figure legend,

the reader is

referred

to the Web version

of this article.)

Fig. 2.

Estimates

of the MLH determined

by the flattening

of water vapor and

potential

temperature

profiles

versus MLH determined

from the DC-8 lidar

aerosol

backscatter.

Dot colors indicate

time of day. Error bars are estimated

uncertainties

(30%, 90% confidence)

for MLH estimates

by the flattening

of

water vapor and potential

temperature.

(For interpretation

of the references

to

color in this figure legend,

the reader is referred

to the Web version

of

this article.)

DO

3

=

k

O

1

D

H

2

O

[

O

1

(

D

)

]

[

H

2

O

]+

k

O

3

OH

[

O

3

][

OH

]+

k

O

3

HO

2

[

O

3

][

HO

2

]+

k

OH

NO

2

[

OH

][

NO

2

]+

∑

k

O

3

VOCi

[

O

3

][

VOC

]

i

(eq. 4)

W.H. Brune

et al.

Atmospheric Environment 269 (2022) 118854

5

PO

3

. The missed approaches

at the Seoul Air Base three times a day

allowed

the comparison

of the observed

O

x

change to the calculated

O

3

production

rate using the measured

or modeled

HO

2

and the modeled

RO

2

.

Four estimates

of FO

3

come from LaRC HO

2

+

LaRC RO

2

, MCM HO

2

+

MCM RO

2

, observed

HO

2

+

LaRC RO

2

, and observed

HO

2

+

MCM

RO

2

. The RO

2

values were used as modeled

and were not scaled to the

differences

in observed

and modeled

HO

2

. DO

3

was calculated

using

LaRC, MCM, and observed

OH and HO

2

along with the same other

chemical

species and reaction

rate coefficients.

Four PO

3

estimates

were

found using eq. (2).

Calculated

PO

3

averaged

for the mixed layer and the residual

layer,

which is estimated

to be the altitudes

between

the early and late MLHs.

The mean PO

3

determined

for the column

up to the top of the residual

layer is the sum of the mixed layer and residual

layer PO

3

weighted

by

the fraction

of the column

that the two layers occupy.

The PO

3

for the

late missed approach

is unaffected,

but the PO

3

for the early missed

approach

is reduced

to 20

–

25% of the mixed layer value because

PO

3

was close to 0 ppbv h

1

in the residual

layer.

To compare

these calculated

PO

3

to changes

in O

x

occurring

between

early and midday

and midday

and late, PO

3

was averaged

for early to

midday

and midday

to late. The approach

outlined

below has several

assumptions,

all with uncertainties.

Thus, the best we can expect from

this analysis

is consistency

between

observed

O

x

changes

and calculated

PO

3

to within an estimated

factor of 1.5.

Consider

the surface deposition

and advection

terms in eq (1). At the

times of the missed approaches

the wind was generally

from the west,

with a mean direction

on 260

◦

, and in the afternoon,

the wind speed was

typically

5 m s

1

. However,

O

3

measurements

at NIER sites upwind

of

the Seoul Air Base were generally

the same to within less than 20%,

implying

that advection

is not a substantial

contributor

to

∂

O

3

∂

t

for the

missed approaches

at Seoul Air Base. For the surface deposition

term,

v

d

was assumed

to be 0.5 cm s

1

(Park et al., 2014) and MLH, the mixed

layer height, was determined

as above. Surface

deposition

was about

10% of DO

3

, the chemical

loss.

Two corrections

were applied

to the observed

O

3

. First, the ozone in

the residual

layer became

entrained

into the mixed layer as the mixed

layer grew. Because

the midday

MLH was fairly comparable

with the

height of the morning

’

s residual

layer, the observed

morning

O

x

was

weighted

for the amounts

of ozone in the mixed layer and the residual

layer. Second,

in the times between

early and midday

and midday

and

late missed approaches,

O

3

was lost by chemical

reactions,

deposition,

and, for the test of heterogeneous

O

3

loss, heterogeneous

reactions.

Once

these two corrections

were made, the corrected

O

x

from the early missed

approach

was subtracted

from the O

x

at the midday

missed approach

and the corrected

O

x

from the midday

missed approach

was subtracted

from the O

x

at the late missed approach.

Each difference

was divided

by

the time between

the missed approaches

to get an O

x

rate-of-change

for

morning

and afternoon.

2.8. Significance

of comparisons

between

observed

and modeled

OH and

HO

2

The uncertainties

in the modeled

and observed

OH and HO

2

affect

the assessment

of how well the models are representing

the oxidation

chemistry.

In Section

2.3, the estimated

absolute

uncertainty

for the

observations

is given as

±

35% at 95% confidence,

but, for the purposes

of this analysis,

we will use

±

40% at 95% confidence

because

it is

impossible

to account

correctly

for all contributors

to the uncertainty.

The modeled

uncertainty

is estimated

from Monte Carlo analyses

of a

global chemical

transport

model (Christian

et al., 2017) and a photo

-

chemical

box model (Chen and Brune, 2012; Chen et al., 2012). Typi

-

cally, these uncertainties

ranged from

±

50% to

±

80%, 95% confidence,

in the global chemical

transport

model, although

only roughly

half was

due to photochemistry,

and

±

30

–

±

35%, 95% confidence,

in the

photochemical

box model. As for the observation

uncertainty,

some

modeled

uncertainty

is likely not included,

so we choose a uniform

model uncertainty

of

±

40%, 95% confidence.

The overlap

in the observed

and modeled

uncertainties

is an indi

-

cator of the agreement

between

the observed

and modeled

OH and HO

2

.

If values of the observed

and modeled

OH and HO

2

are separated

by

±

40%, with 20% from observed

uncertainty

and 20% from modeled

un

-

certainty,

then the observed

and modeled

values are different

by a

confidence

interval

of 95%, which corresponds

to a statistical

p-value

of

0.05. Thus, if the comparisons

of observed

and modeled

OH and HO

2

disagree

by a factor of 1.4, or 40%, then this disagreement

is an

indicator

of the transition

between

agreement

and disagreement

between

observed

and modeled

OH and HO

2

. In scatter plots, we use a York fit

(York et al., 2004) because

it takes into account

the uncertainty

of both

the observations

and the models.

3. Results

and discussion

Several results are reported

here. First is the comparison

between

the

observed

OH and HO

2

and that calculated

by the LaRC and MCM

models,

with the MCM model being run with and without

simplified

HO

2

heterogeneous

uptake.

Second

is a discussion

of heterogeneous

uptake of HO

2

. Third is the determination

of the amount

and location

of

missing

OH reactivity

in and around the South Korean peninsula.

Last is

the calculation

of PO

3

using measured

and calculated

HO

2

and OH and

the comparison

of observed

O

3

change

rates with those calculated

O

3

production

rates using observed

and modeled

HO

2

.

3.1. Comparisons

of observed

and modeled

OH and HO

2

Median

altitude

profiles

of OH and HO

2

indicate

whether

the models

are able to simulate

OH and HO

2

over a wide range of pressure,

envi

-

ronmental

factors,

and chemical

composition

(Fig. 3). The two median

modeled

profiles

are in excellent

agreement,

except for HO

2

in the mixed

layer, where LaRC HO

2

is ~30

–

40% greater than MCM HO

2

. Above 3

km, the observed

OH and HO

2

are generally

20

–

50% greater

than

modeled.

As a result, the HO

2

/OH ratio modeled

by MCM is in excellent

agreement

with the observed

ratio and the LaRC ratio is 10

–

50% larger

at all altitudes.

Below 1.5 km altitude,

which is the altitude

range of

greatest

interest,

the median

observed

values agree to well within un

-

certainties

with MCM OH and HO

2

and LaRC OH but are ~30% lower

than LaRC HO

2

, so that the LaRC HO

2

/OH ratio is also too large by

~30%.

In regions

with sufficient

NO, the cycling

between

OH and HO

2

is

substantially

faster than HO

x

production

and loss. Under these condi

-

tions, it is easier to diagnose

why model OH and HO

2

may be different

from observed.

In cleaner environments,

often HO

x

production

and loss

have rates similar to cycling

between

OH and HO

2

, making

it more

difficult

to diagnose

the model-to-observation

differences.

The differ

-

ences above 3 km suggest that a HO

x

source may be missing

in the MCM

chemistry

because

observed

OH and HO

2

are both greater than modeled

with MCM but the HO

2

/OH ratio is the same. Also, these differences

suggest that the LaRC chemistry

may also have missing

OH loss because

observed

OH is less than modeled

but observed

and modeled

HO

2

agree.

Another

way to compare

observed

and modeled

OH and HO

2

is

scatter plots (Fig. 4). The slopes for observed

versus modeled

OH and

HO

2

are within 13% of the 1:1 line, indicating

that observed

OH and

HO

2

are not statistically

different

from modeled

OH and HO

2

for both

models.

The small percent

differences

of 13% or less confirm

this simi

-

larity between

observations

and models.

Thus, it appears

that both

models can generally

simulate

the observed

OH and HO

2

.

LaRC and MCM models

and their computational

strategies

are

different.

First, MCM has fairly explicit

oxidation

sequences

for most of

the chemical

species measured

by the Whole Air Sampler

on the NASA

DC-8, including

the C

7

+

aromatics

that were lumped

together

in LaRC.

Second,

the approaches

to interpolating

over gaps between

Whole Air

W.H. Brune

et al.

Atmospheric Environment 269 (2022) 118854

6

Sampler

VOC measurements

and missing

data for other measurements,

while similar,

are not the same. Third, MCM is integrated

to a fixed

period of time, but LaRC is integrated

until diurnal

steady-state

is

achieved.

Despite

these differences,

the median

observed

and modeled

OH and HO

2

generally

agree to within combined

uncertainties

for both

LaRC and MCM (Fig. 4). The greater apparent

scattering

for MCM than

Fig. 3.

Altitude

profiles

for median

OH (a), HO

2

(b), and HO

2

/OH (c) for observations

(blue circles),

the LaRC model (light blue stars), and the MCM model (gold

squares).

Absolute

uncertainty

(90% confidence)

given for observations,

and the two models

have uncertainties

similar to those of the observations.

(For inter

-

pretation

of the references

to color in this figure legend,

the reader is referred

to the Web version

of this article.)

Fig. 4.

OH and HO

2

: observed

versus modeled

with

the LaRC and MCM mechanisms.

Linear fits (red

solid lines) and typical combined

measurement

and

model

uncertainties

(68% confidence

level for

each), which is a factor of 1.4 (dashed

lines), are

shown along with the 1:1 line (black dashed line).

Also included

is the linear fit for the MCM model

including

heterogeneous

HO

2

loss (solid orange

line). (For interpretation

of the references

to color

in this figure legend,

the reader is referred

to the

Web version

of this article.)

W.H. Brune

et al.

Atmospheric Environment 269 (2022) 118854

7

LaRC in the HO

2

plot likely results, at least in part, from the sensitivity

that these models have to the treatment

of the constraining

input mea

-

surements,

including

the methods

for data gap filling and interpolation

from the measurement

timestep

to the model timestep.

This good

agreement

among modeled

and measured

OH and HO

2

suggests

that

model details can be unimportant

as long as the chemistry

is represented

as completely

as possible

(Schroeder

et al., 2020).

However,

the observed

scatter and the relatively

low R

2

values,

particularly

for OH, indicate

that there are significant

minute-by-minute

differences

that the median

values and scatter plots gloss over. In Fig. 5,

a time series from one flight provides

an example

of the changes

in the

relationships

of the observed

and modeled

and even between

the two

models on relatively

short time scales. The substantial

differences

for OH

and HO

2

between

the observations

and the model including

heteroge

-

neous chemistry

will be discussed

in Section

3.2. For this flight on day-

of-the-year

(doy) 139 (18 May 2016), the DC-8 took off from Osan, did a

missed approach

at Seoul Air Base and then a high-level

leg over the East

China Sea, followed

by a low-level-leg

return. Then the DC-8 did another

missed approach

at the Seoul Air Base before a high-level

leg over land

to Busan, followed

by a low-level

leg return. Finally the DC-8 did one

more missed approach

at Seoul Airport before returning

to Osan.

The observed

to modeled

agreement

is highly variable,

with sudden

shifts in agreement

with one or both models.

For example,

over the

ocean between

for doy between

139.05 and 139.15,

OH from observa

-

tions, LaRC, and MCM are all substantially

different

even though HO

2

from the three are essentially

the same, with the exception

of one short

period near 139.12.

Interestingly,

the OH agreement

amongst

the three

is consistently

better over land than ocean while the HO

2

agreement

is

worse.

Variability

in the ATHOS

calibration

is not likely the only explana

-

tion for variable

agreement

between

observed

and modeled

OH and HO

2

because

ATHOS

calibration

shifts on these time scales are typically

caused by laser wavelength

drift, which affects OH and HO

2

equally,

whereas

the variations

in agreement

between

models and observations

for OH appear to be independent

from those in HO

2

. In fact, the there is

no correlation

between

the model-observation

percent

differences

for

OH and the model-observation

percent

differences

for HO

2

for LaRC (R

2

=

0.06) and for MCM (R

2

=

0.008). In addition,

there is no correlation

between

the LaRC-MCM

percent

differences

for OH and the LaRC-MCM

percent

differences

for HO

2

(R

2

=

0.08). We suggest that these variations

in relationships

are caused by variations

in instrument

performance

for

measurements

used to constrain

the models and the treatment

of con

-

straining

observations

when interpolating

to either a 1-s or a 1-min time

interval.

None-the-less,

in terms of percent

difference,

scatter plot slope and

intercept,

and correlation

coefficients,

the observed-to-model

agreement

is about as good as has been obtained

on several other airborne

missions

with ATHOS

(Miller and Brune, 2021), particularly

those that spent a

significant

amount

of time sampling

in the planetary

boundary

layer.

3.2. A test of heterogeneous

uptake

of HO

2

on aerosol

particles

The effects of heterogeneous

chemistry

involving

OH and HO

2

with

an accommodation

coefficient,

α

, of 0.2 can been seen in Fig. 4. When

diffusion

to the surface is considered,

the effective

uptake coefficient

is

~0.17. On the scatter plots, the linear fit to the MCM model with het

-

erogeneous

chemistry

has a slope of 1.4 for OH and HO

2

, just at the edge

of being a statistically

significant

difference.

In Fig. 5, the time series

shows the effects of heterogeneous

chemistry

on OH and HO

2

in detail.

For the low-level

flights over both ocean and land, heterogeneous

chemistry

reduces

modeled

OH to ~60% of observed

and modeled

HO

2

to 35

–

50% of observed,

both substantial

amounts.

Another

approach

to testing whether

heterogeneous

chemistry

is a

significant

HO

x

loss is to plot the ratio of modeled-to-observed

OH and

HO

2

as a function

of aerosol

surface

area (Fig. 6). Although

there is

considerable

scatter for the individual

1-min data, the median

OH and

HO

2

modeled-to-observed

ratios are both well within uncertainties

of

1.0 for more than two decades

of aerosol

surface

area per volume

for

MCM including

only gas-phase

chemistry.

However,

for MCM with

heterogeneous

chemistry,

the modeled-to-observed

ratio becomes

Fig.

5.

Time series of OH, HO

2

, and HO

2

/OH

observed

(navy lines), modeled

with the LaRC

mechanism

(light blue lines), the MCM mechanism

(gold lines), and the MCM mechanism

with het

-

erogeneous

OH and HO

2

loss (green lines). The

altitude

in km (black lines) is also shown. The flight

was over the East China Sea from 139 to 139.1 and

over the Korean

peninsula

from 139.13

to 139.3.

(For interpretation

of the references

to color in this

figure legend,

the reader is referred

to the Web

version

of this article.)

W.H. Brune

et al.

Atmospheric Environment 269 (2022) 118854

8

increasingly

lower as the aerosol

surface

area per volume

becomes

increasingly

larger than 10

6

μ

m

2

m

3

. At a surface area per volume

of

3

×

10

5

μ

m

2

m

3

, the model-to-observed

ratio is 0.7 for OH and less

than 0.5 for HO

2

.

All these different

approaches

demonstrate

that adding HO

2

uptake

to the model is inconsistent

with the observations

for KORUS-AQ.

Thus,

it is unlikely

that O

3

increased

in China between

2013 and 2017 because

aerosol amount

decreased,

resulting

in less HO

2

uptake,

more gas-phase

HO

2

, and therefore

more O

3

production

(Li et al., 2018). It is possible

that aerosol

particles

in China have a different

chemical

composition

that those in South Korea. However,

a large fraction

of aerosol particles

in South Korea originate

in China (Nault et al., 2018), suggesting

that

this difference

in composition

is unlikely

to be responsible

for substan

-

tial differences

in HO

2

uptake.

This result of little HO

2

uptake

is

consistent

with the results from the Deep Convective

Clouds

and

Chemistry

study over the Central United States in summer

2012 (Brune

et al., 2018). We therefore

conclude

that HO

2

uptake on aerosol particles

has little, if any, effect on HO

2

. More work will be needed to reconcile

these results with laboratory

studies.

3.3. Missing

OH reactivity

Over the past 20 years, ever since we reported

the first direct mea

-

surements

of OH reactivity

and then missing

OH reactivity

in a forest,

missing

OH reactivity

has been reported

for a wide range of environ

-

ments by several

research

groups (Di Carlo et al., 2004; Yang et al.,

2016). However,

over time, there has been a general

tendency

for

decreasing

reported

missing

OH reactivity.

This trend was anticipated

even twenty years ago because

the ability to measure

larger, highly

reactive

molecules,

such as sesquiterpenes,

and oxygenated

molecules

has improving

dramatically.

With the fairly complete

instrument

suite

on the DC-8, we would expect that the missing

OH reactivity

might be

within the combined

measurement

and modeled

uncertainty,

which

recently

is typically

~30

–

40%, 90% confidence.

The median

observed

OH reactivity

for land regions

with NO less

than 5 ppbv is 4.6 s

1

in the mixed layer and decreases

to ~1 s

1

above

~3 km (Fig. 7). Over the ocean, the median

value is 3.5 s

1

just above

the surface and decreases

to ~1 s

1

above 2 km. In the lowest 2 km, the

median

OH reactivity

calculated

with MCM is generally

lower than the

median

observed

OH reactivity

over both land and ocean and the dif

-

ference is the missing

OH reactivity.

The median missing

OH reactivity

is

~1.3 s

1

(30%) in the mixed layer over land and ~0.4 s

1

(12%) over

ocean. The median

OH reactivity

over ocean for KORUS

is almost twice

that found in Atmospheric

Tomography

(ATom)

over the tropical

Pacific

Ocean, but the missing

OH reactivity

in ATom was 0.5 s

1

about the

same as for KORUS-AQ

(Thames

et al., 2020). Between

2 and 6 km, the

missing

OH reactivity

becomes

statistically

insignificant.

Thus, missing

OH reactivity

is confined

to the PBL.

A map of the missing

OH reactivity

for all flights at altitudes

below 2

km and for NO less than 5 ppbv shows substantial

variability

(Fig. 8).

Fig. 6.

Modeled-to-observed

OH and HO

2

versus aerosol

surface area. Plotted

are the ratio with the MCM model using only gas-phase

chemistry

(1-min data

points are brown dots; blue linked circles)

and the ratio with the MCM model

including

HO

2

uptake (gray dots; linked golden squares).

(For interpretation

of

the references

to color in this figure legend,

the reader is referred

to the Web

version

of this article.)

Fig. 7.

Measured

and model-calculated

OH reac

-

tivity and missing

OH reactivity

as a function

of

altitude

over land (a) and over ocean (b). Dots are

individual

5-min observed

OH reactivity

over land

(a, green) and ocean (b, blue); Gray dots are indi

-

vidual 5-min missing

OH reactivity

in both (a) and

(b). Error bars are the measured

OH absolute

un

-

certainty

(

±

0.64 s

1

, 95% confidence).

Precision

on

individual

5-min averages

is

±

1.0 s

1

, 95% confi

-

dence. (For interpretation

of the references

to color

in this figure legend,

the reader is referred

to the

Web version

of this article.)

W.H. Brune

et al.

Atmospheric Environment 269 (2022) 118854

9

The greatest

missing

OH reactivity

is in the SMA and along the flight

corridors

from Osan to Busan or Gwangji

(126.7

◦

longitude,

35.2

◦

lati

-

tude) to the south of Osan. Over the East China Sea, the missing

OH

reactivity

varies from 0 to ~2 s

1

, with a few exceptions.

The near-zero

missing

OH reactivity

occurs primarily

over the ocean south of the South

Korean

peninsula,

but can also be seen along flight tracks over the

peninsula,

including

along the flight track from Osan to Gwangju.

Fac

-

tors other than location

that are clearly important.

One important

factor is season. KORUS-AQ

occurred

in the transition

between

winter and summer

and during climatologically

rising tem

-

peratures,

which increased

about 7

◦

C from the beginning

to end of

KORUS.

As a result, missing

OH reactivity

roughly

correlates

with the

day of the year, complicating

the analysis

of the spatial distribution

of

missing

OH reactivity.

For example,

the flight along a straight

line south

through

Gwangju

was in early May and shows little to no missing

OH

reactivity.

A repeat flight down the same flight corridor

on 9 June had to

deviate

around controlled

air space near 35

◦

latitude

and typically

had

2-3 s

1

of missing

OH reactivity.

Another

important

factor is wind direction.

For example,

on 10 June,

the DC-8 flew at 8 km altitude

directly

to Fukuota,

Japan, spiraled

down,

and then did mostly low-level

legs to the west over the ocean before

flying along the Osan-to-Gwangju

flight corridor.

Even though the low-

level legs over this flight corridor

were at the same time of day as the

ones flown on 9 June, the missing

OH reactivity

on 10 June was 0

–

1.5

s

1

, substantially

less than on 9 June. The temperature

was a few de

-

grees lower on 10 June, but the big difference

was the wind direction.

On 9 June, the wind was coming

from the north to east, over the

peninsula,

but on 10 June, it was coming

from south to west, over the

ocean.

Thus, two important

factors in determining

the amount

of missing

OH reactivity

appear to be the time of year and whether

the air sampled

was coming from over the ocean or over the Korean peninsula.

The most

missing

OH reactivity

was in air that was coming

from South Korea it

-

self. It is difficult

to say if the source of missing

OH reactivity

is biogenic

or anthropogenic,

but, for some flights, enhanced

missing

OH reactivity

seems to have spread from the SMA to nearby regions.

The DC-8 in

-

strument

payload

measured

an extensive

suite of both anthropogenic

and biogenic

VOCs, so a likely cause of the missing

OH reactivity

is a

combination

of oxygenated

VOCs, perhaps

ethanol

and oxygenated

products

from C

7

+

aromatics,

which were not quantitatively

measured

during KORUS-AQ

but could be abundant

in the SMA, and oxygenated

biogenic

VOS products.

This combination

can explain

the seasonal

in

-

crease and the land-based

origin.

It may be that the missing

OH reactivity

affected

the OH and HO

2

abundances

as well as the O

3

production.

However,

the OH reactivity

measurement

is instantaneous,

which means that it misses the subse

-

quent reactions

in the oxidation

chemistry

could be recycling

OH and

HO

2

. For some VOCs, a substantial

amount

of HO

x

is recycled,

so that,

without

knowing

the identity

of the missing

OH reactivity

reactant,

it is

not possible

to know if the missing

OH reactivity

is consistent

with the

observed

OH and HO

2

and the calculated

O

3

production.

3.4. Ozone

production

How much O

3

increased

during the day in the SMA was determined

mainly by FO

3

due to the reactions

of HO

2

and RO

2

with NO (Fig. 9),

which, when the chemical

loss DO

3

is subtracted,

gives the net chemical

O

3

production,

PO

3

. The calculated

FO

3

using HO

2

that was observed

and modeled

by LaRC and MCM are shown as a function

of the observed

NO in the top panel and the calculated

FO

3

using only RO

2

modeled

by

LaRC and MCM are shown in the lower panel of Fig. 8.

For HO

2

, observed

and modeled

FO

3

agree up to NO of ~1 ppbv and

increasingly

diverge

at greater

NO (Fig. 9a). For 10 ppbv of NO, FO

3

using observed

HO

2

is 3 times modeled

and FO

3

calculated

with LaRC

HO

2

is 30% greater than that calculated

with MCM. Above 10 ppbv, the

difference

grows much greater,

but the number

of data points is small.

Some of this difference

could be due to sampling

from the DC-8 that, in

1 s, travels 100 m. Thus, it is possible

that the DC-8 instruments

sampled

finer filaments

of NO-rich

air that was separated

from the HO

2

-rich

background

air, but each was recorded

for that second (Olson et al.,

2010). The resulting

product

of the two, averaged

over a second,

would

be greater than the actual product

in the atmosphere,

giving an inflated

calculated

value for FO

3

. The data points with NO above 10 ppbv were

generally

from missed approaches

and so may have suffered

from this

problem.

However,

even if we eliminate

all data from the missed ap

-

proaches,

the deviation

of PO

3

calculated

using observed

HO

2

from that

modeled

is still evident

for NO less than 10 ppbv.

The FO

3

from RO

2

calculated

with LaRC is different

from that

Fig. 8.

Map of missing

OH reactivity

(5-min averages)

at altitudes

less than 2

km over South Korea and vicinity.

The centers of Seoul, Busan, and Fukuoka,

Japan are indicated

by red circles, with Seoul to the north and Fukuoka

to the

south. (For interpretation

of the references

to color in this figure legend,

the

reader is referred

to the Web version

of this article.)

Fig. 9.

FO

3

due to HO

2

(a) and to RO

2

(b) as a function

on NO for measure

-

ments at altitudes

below 2 km. Medians

and 1-s data of PO

3

are shown for

observed

HO

2

(blue circles, green dots), LaRC modeled

HO

2

(ligh blue stars and

gray dots), and MCM modeled

HO

2

(gold squares),

as well as PO

3

from RO

2

modeled

by LaRC (light blue stars and gray dots) and by MCM (gold squares

and

taupe dots). (For interpretation

of the references

to color in this figure legend,

the reader is referred

to the Web version

of this article.)

W.H. Brune

et al.