Introduction
Nitrogen oxides (NOx) are a key component of tropospheric
chemistry due to their impact on oxidant chemical cycles. This paper focuses
on the NOx budget of the marine boundary layer (MBL). The
traditional view has been that relatively low concentrations of
NOx in the MBL are maintained by a balance between the
transport of NOx to the MBL (either by direct transport of
NOx from continental source regions or by transport and
subsequent decomposition of NOx reservoirs such as organic
nitrates) and chemical destruction by reaction with hydroxyl radicals (OH)
(e.g., Moxim et al., 1996). NOx emissions from ships
(Lawrence and Crutzen, 1999; Kasibhatla et al., 2000; Davis et al., 2001; Beirle et al., 2004),
downward transport of lightning-produced NOx from the free
troposphere (Levy II et al., 1996, 1999), and halogen-mediated chemistry (von
Glassow et al., 2004; Keene et al., 2009; Long et al., 2014; Schmidt et al.,
2016; Sherwen et al., 2016a, b) also modulate NOx levels in
the MBL.
Past laboratory studies have suggested that recycling of NOx
by photochemical reactions of adsorbed nitrate on surfaces might also serve
as an in situ NOx source (e.g., Zhou et al., 2003; Handley
et al., 2007; Chen et al., 2011; Laufs and Kleffmann, 2016). More recently,
field studies have suggested that photolysis of particulate nitrate
(p-NO3) associated with MBL sea-salt aerosols (SSA) can be a
significant source of NOx. In particular, Ye et al. (2016)
analyzed trace gas measurements from two aircraft flights over the western
subtropical North Atlantic Ocean during the summer 2013 Nitrogen, Oxidants,
Mercury and Aerosol Distributions, Sources and Sinks (NOMADSS) field campaign
and hypothesized that measured daytime nitrous acid (HONO) concentrations in
the MBL were indicative of a significant in situ source from
p-NO3 photolysis (since HONO photolyzes rapidly to yield
NOx, p-NO3 photolysis is effectively a
NOx source). Their box model analysis suggested that the
peak p-NO3 photolysis rate coefficient was on the order of 2×10-4 s-1, about 300 times larger than the corresponding
photolysis coefficient for gas-phase nitric acid (HNO3).
Manipulation experiments involving irradiation of samples of ambient (Ye et
al., 2016) and synthetic marine
aerosol (Zhou et al., 2008) provide additional support for the hypothesized
importance of p-NO3 photolysis in the MBL. It is also worth noting
that photolysis of nitrate in seawater has been measured at naturally
occurring concentrations, albeit at much lower rates with photolysis rate
coefficients averaging 2×10-8 s-1 (Zafiriou and True,
1979).
Measurements of NOx and HONO at the Cape Verde
Atmospheric Observatory in the tropical eastern Atlantic Ocean (CVAO, 16.864∘ N,
24.868∘ W) have provided additional evidence for in situ
NOx production in the MBL (Reed et al., 2017). Long-term
NOx measurements at CVAO show a consistent diurnal cycle,
with peak daytime mixing ratios 10–15 pptv higher than nighttime levels in
all seasons except in winter (Reed et al., 2017). Shorter-term HONO
measurements from CVAO also show a strong diurnal cycle, peaking at a
few pptv around midday and dropping to near zero at night (Reed et al.,
2017). The observed levels of HONO again indicate an in situ source, while
the diurnal cycle suggests that this source is photolytic. Box model analysis
of the CVAO measurements indeed indicates that rapid recycling of
NOx via p-NO3 photolysis is an important source of
NOx in the tropical MBL, though estimated p-NO3
photolysis rate coefficients were only 10–15 times larger than
HNO3 photolysis rate coefficients (Reed et al., 2017; Ye et al.,
2017). While some of the variability in estimated p-NO3 photolysis
rates between the NOMADSS and CVAO analyses may be due to differences in SSA
composition at the two locations (Ye et al., 2017), the limited measurements
preclude from definitive conclusions in this regard.
In this paper, we explore the global-scale significance of this
newly recognized p-NO3 photolysis mechanism for tropospheric
oxidant chemistry using the GEOS-Chem global chemical transport model. We
perform a series of simulations to assess the sensitivity of
NOx to the rate of p-NO3 photolysis in SSA in the
MBL and attempt to constrain the magnitude of this in situ
NOx source using NOx and HONO measurements
from CVAO. We then estimate the impact of p-NO3 photolysis on the
large-scale distribution of NOx, O3, and OH in the
MBL. Finally, we perform an initial analysis of the impact of photolysis of
nitrate on other aerosol types and consider its implications for continental
boundary layer oxidant chemistry.
Models and measurements
GEOS-Chem model configuration
We use the GEOS-Chem chemical transport model (v10-01, Bey et al., 2001;
www.geos-chem.org, last access: 20 July 2017), as
updated by Sherwen et al. (2017). The model includes a comprehensive
tropospheric
NOx–VOC–HOx–Ox oxidant
chemistry scheme (Mao et al., 2013), coupled to an aerosol module that
includes inorganic sulfate, ammonium, and nitrate aerosols (Park et al.,
2004), as well as primary black carbon, organic carbon, dust, and sea-salt
aerosols (Park et al., 2003; Fairlie et al., 2007; Alexander et al., 2005).
The oxidant chemistry scheme also includes a detailed treatment of
tropospheric halogen chemistry (Parrella et al., 2012; Schmidt et al., 2016;
Sherwen et al., 2016a, b, 2017). Trace gas and aerosol emissions are
specified as described in Ford and Heald (2012).
The model transports two size classes of SSA, representing accumulation-mode
(0.01–0.5 µm dry radius) and coarse-mode (0.5–8 µm dry
radius) SSA. Emissions of SSA are calculated online in the model using the
formulation of Gong (2003), as described by Jaeglé et al. (2011). The
model explicitly tracks particulate sulfate and nitrate associated with
coarse-mode SSA (identified as SO4s and NITs, respectively,
in the model), with SO4s production and gas-phase HNO3
uptake (to form NITs) calculated as described by Alexander et al. (2005). In
this scheme, SO4s and NITs formation at each time step is
determined by the mass flux of gas-phase sulfur dioxide and HNO3 to
coarse-mode SSA, but limited by the local alkalinity emission flux in that
time step. In Sect. 3.1, we show that this imposed alkalinity limitation
unrealistically limits NITs production in the MBL. We therefore update the
model calculation of HNO3 uptake on coarse-mode SSA as described
below.
Laboratory measurements (e.g., Guimbaud et al., 2002) and field studies
(e.g., Keene et al., 2009) indicate that freshly produced SSA are rapidly acidified
via incorporation of HNO3 and other acids leading to the
volatilization of hydrochloric acid in accordance with expectations based on
respective thermodynamic properties (Henry's law and dissociation constants)
and associated meteorological conditions (relative humidity and temperature).
We therefore updated the model to calculate HNO3 uptake (and
concomitant NITs production) by coarse-mode SSA using the mass transfer
equation appropriate for the transition regime as in Feng and Penner (2007):
-ddtHNO3=ddtNITs=4πDgRcSSANcSSAfKn(HNO3-[HNO3]eq),
where [.] represents the mixing ratio, Dg is the gas-phase
diffusivity of HNO3, RcSSA is the radius of coarse-mode
SSA, NcSSA is the number concentration of coarse-mode SSA, and
Kn is the Knudsen number. [HNO3]eq is the
equilibrium concentration of gas-phase HNO3 associated with
coarse-mode SSA, determined using ISORROPIA II (Fountoukis and Nenes, 2007).
Since the model does not include aerosol microphysics, RcSSA and
NcSSA are estimated from effective dry radius and mass
concentration of coarse-mode SSA, assuming a dry SSA density of
2200 kg m-3 and using relative-humidity-dependent SSA growth factors
(Lewis and Schwartz, 2006). The function fKn is the ratio
of gas-phase flux to the particle in the transition region to that in the
continuum regime and is specified using the Fuchs and Sutugin (1971)
formulation:
f(Kn)=0.75α(1+Kn)Kn2+Kn+0.283Knα+0.75α,
where α is the HNO3 mass accommodation coefficient,
specified to be 0.2 (Jacob, 2000).
We evaluate the effect of p-NO3 photolysis in SSA in the MBL by
including the photolysis of NITs (yielding HONO and NO2) in our
updated model. We note that NITs only corresponds to p-NO3
associated with coarse-mode SSA. The current version of the GEOS-Chem model
also tracks p-NO3 associated with accumulation-mode aerosols
(identified as “NIT” in the model), but does not track the fraction that is
specific to accumulation-mode SSA. We therefore do not include photolysis of
accumulation-mode p-NO3 in our analysis of MBL SSA nitrate
photolysis. We discuss the implication of this model limitation in Sect. 3.1.
We also present results of a sensitivity simulation designed to explore the
effect of accumulation-mode NIT photolysis on continental boundary layer
composition in Sect. 3.4.
Following the empirical analysis of Ye et al. (2016), the NITs photolysis
rate coefficient (JNITs) is scaled to the photolysis rate of
HNO3 (JHNO3). We refer to this scaling coefficient
as Jscale in the rest of this paper. Box model analyses of
measurements from NOMADSS, the spring 2007 Reactive Halogens in the Marine
Boundary Layer (RHaMBLe) project, and measurements made at Cape Verde suggest
Jscale values of between 10 and 300 (Ye et al., 2016, 2017; Reed et
al., 2017), with the variability possibly related to differences in SSA
composition at different locations in the MBL (Ye et al., 2017). In light of
this uncertainty, we show model results for a range of Jscale
values and use measurements of NOx and HONO from CVAO to
estimate reasonable bounds for this scaling factor. In the subsequent
discussion, we use the notation Jscalen, where the superscript
n indicates the value of Jscale. Note that in this notation
Jscale0 refers to the “no NITs photolysis” case. NITs
photolysis is initially assumed to yield HONO and NO2 in the molar
ratio of 0.67:0.33 (Ye et al., 2016; Reed et al., 2017), but we also
present results from a subset of sensitivity simulations with the
HONO:NO2 molar yield set to 0.93:0.07 (Ye et al., 2017).
The current version of the model does not include a dry deposition term for
HONO. While neglecting this term might have a small effect on simulated
nighttime HONO concentrations in the MBL, the impact should be negligible in
terms of simulated daytime HONO concentrations because the lifetime of HONO
during the day is controlled by photolysis and not by dry deposition. Dry
deposition velocities for species with Henry's law coefficients similar to
HONO are typically 0.5–1 cm s-1 over oceans. For a daytime MBL depth
of 1 km, this corresponds to a lifetime against dry deposition of about
1–2 days. By contrast, HONO has a lifetime of minutes against photolysis
during the day at a location like CVAO.
Measurements used for model evaluation
MBL NITs and HNO3 measurements in the
eastern Atlantic Ocean
During the RHaMBLe field campaign in May and June 2007, volatile
water-soluble nitrate (dominated by and hereafter referred to as
HNO3) was sampled from the top of the CVAO tower over 2 h
intervals at nominal flow rates of 20 L min-1 using tandem mist
chambers (MCs), each of which contained 20 mL deionized water (Keene et al.,
2009). To minimize artifact phase changes caused by mixing chemically
distinct aerosol size fractions on bulk prefilters, air was sampled through a
size-fractionating inlet that inertially removed super-µm-diameter
aerosols from the sample stream. Sub-µm aerosol was removed
downstream by an in-line 47 mm Teflon filter (Zefluor 2 µm pore
diameter) that was changed daily. Samples were analyzed on-site by ion
chromatography (IC) usually within a few hours after recovery. Data were
corrected based on dynamic handling blanks that were loaded, briefly (few
seconds) exposed to ambient air flow, recovered, processed, and analyzed
using procedures identical to those for samples. Collection efficiencies were
greater than 95 % and, consequently, corrections for inefficient sampling
were not necessary. Relative precisions based on paired measurements varied
as functions of concentration and typically averaged ±10 % to ±25 %. The average detection limit (DL; estimated following Keene et al.,
1989) was 12 pptv. Longer-term aerosol nitrate measurements were also made
at CVAO, as part of a larger suite of aerosol measurements. Samples were
collected on quartz-fiber filters using a high-volume sampler with a PM10
inlet and were subsequently analyzed by ion chromatography. Here, we use the
measurement summaries for summer marine air masses from 2007 to 2011 reported
by Fomba et al. (2014).
Comparison of modeled gas-phase HNO3 and accumulation-
plus coarse-mode particulate nitrate (p-NO3) from the
standard (a) and updated (b) models with measurements. CVAO
HNO3 measurements are for open ocean air masses from the spring
2007 RHaMBLe campaign (Lawler et al., 2009), while p-NO3
measurements are from the summer marine aerosol subset of samples from 2007
to 2011 (Fomba et al., 2014). N-AFR, ITCZ, and S-ATL HNO3 and
p-NO3 shipboard measurements are for the North Africa, ITCZ, and
South Atlantic transport regimes in the eastern Atlantic during November 2003
(Keene et al., 2009). Model results are for May (CVAO HNO3),
June–August (CVAO p-NO3), and November (N-AFR, ITCZ, and S-ATL
HNO3 and p-NO3) from the 2014–2015 simulations. Symbols
and lines indicate (i) means and ranges and/or 2 standard deviations for
measurements and (ii) medians and 5th and 95th percentiles of daily means for
models.
During October and November 2003, the German research ship
Polarstern cruised along a latitudinal transect through the eastern
Atlantic Ocean from Germany to South Africa. HNO3 was measured from
the ship at approximately 23 m above the water line over 2 h intervals
using the same MC–IC technique described above. The average DL was 2 pptv.
During this cruise, ambient aerosols were also sampled in parallel over
discrete daytime and nighttime intervals using a modified Graseby-Anderson
Model 235 cascade impactor configured with a Liu–Pui type inlet, precleaned
polycarbonate substrates, and quartz-fiber backup filters (Pallflex 2500
QAT-UP) (Keene et al., 2009). At an average sampling rate of
0.78 m3 min-1, the average 50 % aerodynamic cut diameters for
the impaction stages were 20, 11, 4.3, 2.5, 1.2, and 0.65 µm. Bulk
aerosol was sampled in parallel on quartz-fiber filters at an average flow
rate of 1.1 m3 min-1. Impactors and bulk-filter cassettes were
cleaned, dried, loaded, and unloaded in a Class 100 clean bench configured
with impregnated filters on the inlet to remove alkaline- and acidic-reactive
trace gases. Exposed substrates and filters were halved, transferred to
polypropylene tubes, sealed in glass mason jars, frozen, and express shipped
from Cape Town to the Mount Washington Observatory, NH, for analysis. Half-sections
of each substrate were extracted in 13 mL deionized water using a
mini vortexer and sonication; half-sections of exposed backup and bulk
filters were similarly extracted in 40 mL deionized water. Substrate and
filter extracts were analyzed by IC for major ionic species including
NO3- and Na+. The corresponding average DLs for bulk
samples were 0.04 and 2.6 nmol m-3, respectively; those for the geometric mean diameter (GMD)
0.46 µm size fraction were 0.05 and 3.5 nmol m-3,
respectively; and those for the larger individual size fractions were 0.01
and 0.1 nmol m-3, respectively. Internal losses of super-µm
aerosols within slotted cascade impactors of this type average about 25 %
to 30 % (e.g., Willeke, 1975); other sources of bias for size-resolved
particulate analytes based on the above procedures are generally unimportant
(Keene et al., 1990).
Simulated 2014–2015 MBL-average (average over bottom 1 km)
HNO3 and NITs (p-NO3 associated with coarse-mode SSA) for
the standard and updated models.
Long-term measurements of NOx and
O3 from CVAO
Ambient air for NOx and O3 measurements is drawn
from a 40 mm glass manifold (QVF) with a hooded inlet positioned 10 m above
ground level at the CVAO site. NO and NO2 are measured by NO
chemiluminescence (Drummond et al., 1985) with NO2 converted
photolytically to NO by illumination at 385–395 nm (Lee et al., 2009). The
limits of detection for NO and NO2 were 0.30 and 0.35 pptv when
averaged over an hour, with an accuracy of 5 % and 5.9 %, respectively.
Reed et al. (2016) showed that in regions that typically have low
NOx concentrations and are cold, in situ measurements of
NO2 may be instrumentally biased high due to thermal decomposition
of NOy species such as PAN and HO2NO2,
leading to apparent perturbations in the Leighton relationship (Leighton,
1961). These effects are however considered to be minimal at CVAO (Reed et
al., 2016). Ozone (O3) is measured by a Thermo Scientific model 49i
UV photometer which quantifies O3 by single wavelength absorption
at 254 nm along a 38 cm path length whilst simultaneously determining a
zero-background measurement in a second absorption cell. The lower detectable
limit is 1 ppbv.
Observed and modeled diurnal cycles of NO at CVAO during different
seasons (MAM: March, April, May; JJA: June, July, August; SON: September,
October, November; DJF: December, January, February) in 2014–2015. Solid
lines represent median values, and shaded regions show first and third quartiles
of model values and standard error of observations. Model simulations assume
a HONO:NO2 molar yield of 0.67:0.33 from NITs photolysis.
Observed and modeled diurnal cycles of NO2 at CVAO during
different seasons (MAM: March, April, May; JJA: June, July, August; SON:
September, October, November; DJF: December, January, February) in
2014–2015. Solid lines represent median values, and shaded regions show
first and third quartiles of model values and standard error of observations. Model
simulations assume a HONO:NO2 molar yield of 0.67:0.33 from NITs
photolysis.
Observed and modeled diurnal cycles of NOx at CVAO
during different seasons (MAM: March, April, May; JJA: June, July, August;
SON: September, October, November; DJF: December, January, February) in
2014–2015. Solid lines represent median values, and shaded regions show
first and third quartiles of model values and standard error of observations. Model
simulations assume a HONO:NO2 molar yield of 0.67:0.33 from NITs
photolysis.
Short-term measurement of HONO from CVAO and RHaMBLe
Measurements of the HONO at CVAO were performed between 24 November and
3 December 2015 using a Long Path Absorption Photometer (LOPAP-03 QUMA
Elektronik & Analytik GmbH), described in Heland et al. (2001). Briefly,
the LOPAP is a wet chemical technique where gas-phase HONO is sampled within
a stripping coil into an acidic solution and is derivatized into an azo dye,
with the light absorption of the dye at 550 nm measured with a spectrometer.
The LOPAP was operated and calibrated according to the standard procedures
described in Kleffmann and Wiesen (2008) and was set up at CVAO in order to
maximize the sensitivity of the LOPAP, as described in Reed et al. (2017).
Regular baseline measurements (8 h) using an overflow of high-purity
N2 were performed to account for instrument drift. The detection
limit (2σ) of the LOPAP was calculated to be 0.2 pptv, with the
relative error conservatively set to 10 % of the measured concentration.
Kleffmann et al. (2006) have explicitly shown that particle-bound nitrite
does not give rise to significant interference in the LOPAP. If such a
positive bias were to occur, we would expect this to also be apparent at
night. The observed low HONO concentration (<1 pptv) at night puts a
(low) upper limit on any such contribution.
During RHaMBLe (see Lee et al., 2010; Sander et al., 2013) HONO was measured
in parallel with HNO3 using the same MC–IC technique described
above. The average detection limit was 2 pptv. These data are considered
semi-quantitative because performance of the MC–IC technique for HONO has not
been rigorously evaluated. However, intercomparisons of opportunity with HONO
measured in parallel by MC–IC and long-path DOAS (differential optical
absorption spectroscopy) during summer 2004 in
coastal New England as part of the International Consortium for Atmospheric Research on Transport and Transformation (ICARTT; ranging from <45 to 730 pptv;
Keene et al., 2006, 2007) and by MC–IC and NI-PT-CIMS (negative-ion proton-transfer chemical-ionization mass spectrometry) during winter 2011 in
Colorado as part of the Nitrogen, Aerosol Composition, and Halogens on a Tall Tower campaign (NACHTT; ranging from <32 to 200 pptv; VandenBoer et
al., 2013) indicate generally good agreement within ±20 %.
Model simulations
We performed a series of 2-year model simulations (each with a 6-month
spin-up from a standard GEOS-Chem benchmark simulation) covering the 2014–2015
period, so as to compare with long-term measurements of NOx
and O3 at CVAO during this period. Given our focus on the MBL, we
used the 4∘×5∘ horizontal-resolution version of the
model for computational efficiency. The vertical grid is a 47-level hybrid
sigma pressure grid, extending from the surface to 80 km. The vertical grid
resolution in the lowermost kilometer is about 125–150 m. All simulations used
meteorological fields from the Goddard Earth Observation System Forward
Processing (GEOS-FP) assimilation system produced by the NASA Global Modeling
and Assimilation Office (Luchesi, 2013). Data processing was performed using
the open-source python packages Numpy (van der Walt et al., 2011) and Pandas
(McKinney, 2010), and figures were prepared using the open-source python
plotting packages matplotlib (Hunter, 2007) and Seaborn (Waskom et al.,
2017).
In the rest of this paper, we refer to the model with the imposed alkalinity
limitation to acid uptake as the “standard” model and the model with
additional nitric acid uptake due to hydrochloric acid displacement
(described in Sect. 2.1) as the “updated” model.
Results and discussion
Gas–particle partitioning of nitrate in the MBL
Observed and modeled diurnal cycles of HONO at CVAO. Panel
(a) shows the comparison of model results for May from the 2014–2015
simulations with measurements during the spring 2007 RHaMBLe campaign (Sander
et al., 2013). Panel (b) shows comparison of model results with
measurements (Reed et al., 2017) during November–December 2015. Measurements
are shown in box-and-whisker notation, with the box showing the lower and
upper quartile values (with a horizontal line at the median) and the whiskers
extending to 1.5 times the interquartile range of the data. Solid and dashed
lines represent model median values for simulations with a HONO:NO2
molar yield of 0.67:0.33 and 0.93:0.07 (denoted as “high HONO”),
respectively, from NITs photolysis. Shaded regions show first and third quartiles
of model values.
Figure 1 shows a comparison of modeled gas-phase HNO3 and total
particulate nitrate (i.e., the sum of NITs and NIT) concentrations with
measurements in the tropical eastern North Atlantic for the standard model
and for the updated model. We note that the modeled time period does not
coincide exactly with the different periods over which the HNO3 and
p-NO3 measurements were taken, indicated in Fig. 1. However,
interannual variability in these regions is typically low (Carpenter et al.,
2010), while the changes in HNO3 and total nitrate concentrations
induced by the model update are large. It is clear from Fig. 1 that the
standard model does not capture the relative partitioning of nitrate between
the gas and particle phases in the MBL. In the standard model, most of the
nitrate is in the gas phase, in contrast to the measurements which show that
most nitrate is in the particle phase.
Simulated 2014–2015 MBL-average (average over bottom 1 km)
concentrations for the Jscale0 model run (a, b, c)
and ratio of simulated MBL-average concentrations for the
Jscale25 (d, e, f) and Jscale100
(g, h, i) model runs relative to the Jscale0 model
run for NOx (a, d, g), OH (b, e, h), and
O3 (c, f, i).
Simulated 2014–2015 MBL-average (average over bottom 1 km)
NOx production rate from PAN decomposition for the
Jscale0 model run (a) and from NITs photolysis for
the Jscale25 (b) and Jscale100
(d) model runs. The ratio of the NOx production
rate from NITs photolysis for the Jscale25 and
Jscale100 model runs relative to the NOx
production rate from PAN decomposition for the Jscale0 model
run is shown in panels (c) and (e), respectively.
Observed and modeled seasonal cycle of O3 at CVAO during
2014–2015. Solid lines represent median values, and shaded regions show
first and third quartiles of model values and standard error of observations.
The biased simulation of p-NO3/HNO3 ratios in the standard model
is the result of the imposed alkalinity limitation to acid uptake (see
Sect. 2.1). For example, tropical North Atlantic SSA emission fluxes are on
the order of 25–50 kg km-2 day-1 (Jaeglé et al., 2011) and
the typical alkalinity of freshly emitted SSA is equal to 0.07 equivalents
per kg (Gurciullo et al., 1999). Assuming a 1 km MBL depth, this translates
to maximum NITs production rates in the MBL of only 1.7–3.5 pptv h-1
in the standard model. In fact, actual NITs production rates in this scheme
will be even lower because of competition with gas-phase sulfur dioxide
uptake. These low production rates, coupled with the short lifetime of NITs
against deposition, is the cause of the biased simulation of MBL
p-NO3/HNO3 ratios in the standard model. In the updated model,
HNO3 uptake by coarse-mode SSA continues after alkalinity is
titrated, due in part to hydrochloric acid displacement (as described in
Sect. 2.1), and the simulation of the MBL nitrate partitioning between gas
and particle phases is much improved. This improvement in the simulation of
gas–particle partitioning of nitrate is important in the context of our focus
on assessing the large-scale importance of p-NO3 photolysis on MBL
chemistry. While this change improves the model performance, it is evident
from Fig. 1 that the model underestimates p-NO3 concentrations for
all of the sites evaluated. Given that the dataset of simultaneous remote
oceanic HNO3 and p-NO3 observations is small, it is
unclear at this stage what could be causing this discrepancy.
We also note that simulated SSA concentrations are in agreement with
measurements at CVAO. The Na+ concentration measurements atop a
30 m tower (40 m above sea level) during RHaMBLe (Lawler et al., 2009)
imply a SSA mass concentration of 6–19 µg m-3. Fomba et
al. (2014) reported similar SSA concentrations (11±5.5 µg m-3) at the top of the tower over a 5-year period
from January 2007 to December 2011. In comparison, the simulated
annual-average SSA concentration at the lowest model level (centered
∼60 m above sea level) at this location is 12.5 µg m-3.
Figure 2 shows maps of average 2014–2015 simulated surface mixing ratios of
HNO3 and NITs in the MBL for the standard and updated models.
Consistent with the discussion above, gas-phase HNO3 is the
dominant form of nitrate in the MBL in the standard model, with mixing ratios
exceeding 100 pptv over ocean basins downwind of major continental
NOx source regions. By contrast, maximum simulated NITs
mixing ratios are about 10–20 pptv over ocean basins. The updated acid
uptake scheme changes the simulated partitioning of nitrate in the MBL
globally. Other than in the immediate vicinity of continental
NOx source regions, simulated nitrate is predominantly
associated with the particle phase everywhere in the MBL in the updated model
(see also Fig. S1 in the Supplement). In the rest of this paper, we use the
updated model to assess the effects of NITs photolysis on MBL oxidant
chemistry.
As mentioned earlier, the current version of the GEOS-Chem model tracks
accumulation-mode p-NO3, but does not separately track the fraction
of this mode that is associated with SSA. Since coarse-mode NITs is the
dominant form of modeled p-NO3 in the MBL in continental outflow
regions where concentrations are highest (Figs. 2 and S1), this model
limitation is not particularly significant in terms of our analysis of SSA
p-NO3 photolysis. We plan to further evaluate this with a future
version of the model that will explicitly track p-NO3 in
accumulation-mode SSA.
NOx and HONO diurnal cycles at CVAO
The diurnal variation of NOx and HONO in the MBL can
potentially provide useful information on in situ NOx
chemical production and destruction. In the remote MBL, one would expect
NOx concentrations to decrease over the course of the day in
the absence of an in situ chemical source that peaks during the day, owing
to the shorter lifetime against oxidation by OH during daylight hours.
Similarly, HONO has a lifetime on the order of minutes against photolysis
(e.g., Kraus and Hofzumahaus, 1998), and so a diurnal profile that peaks
during the day would provide strong evidence for an in situ photochemical
source for HONO in the MBL. Measurements at CVAO of NOx and
HONO in air masses representative of relatively aged marine boundary layer
air do indeed show HONO and NOx peaking during the day (Reed
et al., 2017). This fact has been used by Reed et al. (2017) and Ye et
al. (2017) to argue for an in situ source of HONO and NOx
from p-NO3 photolysis based on box model analysis of measurements.
Their analyses are, however, limited by the fact that background
NOx levels cannot be explicitly simulated in a box modeling
framework. To overcome this limitation, Reed et al. (2017) modeled the
background source of MBL NOx from PAN decomposition by
prescribing a flux of PAN from the free troposphere to the MBL. Ye et
al. (2017) constrained background NOx levels in their
analysis by specifying NO2 mixing ratios after sunrise. Here, we
reanalyze the CVAO NOx and HONO measurements in the context
of our three-dimensional model that explicitly simulates transport of
NOx and its precursors.
Figures 3, 4, and 5 show comparisons of simulated diurnal profiles of NO,
NO2, and NOx with measurements at CVAO. Model
results are shown for Jscale ranging from 0 (i.e.,
Jscale0, no NITs photolysis) to 100 (i.e.,
Jscale100). To mimic the fact that air masses arriving at CVAO
are representative of aged background MBL air (Read et al., 2008; Carpenter
et al., 2010), model results are shown for the grid box to the northwest of
CVAO to minimize the effect of fresh NOx emissions from ship
traffic off the coast of north Africa on model–measurement comparisons.
Several aspects of the comparisons shown in these figures are worth noting.
In all seasons, except winter, a weak daytime peak is evident in observed
NOx. This peak is the result of a relatively flat
NO2 diurnal cycle, combined with an NO diurnal cycle that peaks
strongly during the day. The model has difficulty reproducing the details of
this diurnal variation. The shape of the NO diurnal cycle is generally well
captured by the model, though peak model NO is generally higher than the
measurements when Jscale is larger than 25. The modeled
NO2 diurnal cycle consistently shows a daytime minimum in the
absence of NITs photolysis, a feature that is not evident in the
measurements. Modeled NO2 agrees better with measurements when
Jscale is specified to be 25–50, though the exact shape of the
NO2 diurnal cycle is not well simulated. This is, in part, due to
the sharp decrease in simulated NO2 at sunrise, owing to the
formation and hydrolysis of halogen nitrates. The absence of this feature in
the measurements has been previously noted by Reed et al. (2017). They
speculate that this could be because of missing halogen chemistry during
nighttime and/or during the night–day transition in the model. While there
are likely shortcomings in our model's representation of NO/NO2
chemistry, the diurnal NOx comparisons (Fig. 5) provide
support for an in situ source of NOx from NITs photolysis,
with NITs photolysis coefficients about 25–50 times larger than
corresponding HNO3 photolysis coefficients. We compare our
estimates of the NITs photolysis coefficient with the corresponding estimates
by Reed et al. (2017) and Ye et al. (2017) at the end of this subsection.
Latitude–altitude plot of average 2014–2015 NOx
(a, b), OH (c, d), and O3 (e, f) for the
Jscale25 and Jscale100 model runs (a,
c, e and b, d, f, respectively)
relative to corresponding fields for the Jscale0 model run at
20∘ W longitude.
Figure 6 shows comparisons of the modeled HONO diurnal cycle with short-term
measurements from the spring 2007 RHaMBLe campaign and with the
November–December 2015 CVAO measurements. The measurements consistently show
a strong daytime peak in HONO, with peak median mixing ratios ranging from
3.5 to 5.5 pptv. The model reproduces the observed shape of the HONO profile
only when NITs photolysis in included. Further, simulated HONO mixing ratios
are less than 0.1 pptv when NITs photolysis is not considered, and the model
underestimates peak HONO even for Jscale100. It is worth
recognizing, however, that simulated HONO mixing ratios at CVAO are roughly
proportional to the assumed molar yield (set at 0.67) of HONO from NITs
photolysis. This is illustrated by the dashed lines in Fig. 6, showing
simulated HONO from Jscale50 and Jscale100
sensitivity runs in which the molar yield of HONO from NITs photolysis is
increased to 0.93 (with a concomitant reduction of the NO2 yield to
0.07). The choice of 0.93 for the HONO yield in the sensitivity runs is based
on the budget analysis of the RHaMBLe measurements by Ye et al. (2017).
Increasing the HONO yield from 0.67 to 0.93 increases peak simulated HONO
mixing ratios by about 35 %, bringing them into better agreement with the
measurements. We note that changing the relative yields of HONO and
NOx from NITs photolysis does not significantly affect
simulated NOx levels because of the short lifetime of HONO
against photolysis to NO.
While our analysis supports the existence of an in situ photolytic source
of HONO and NOx in the MBL, we are unable to draw a
definitive conclusion on the magnitude of this source. The
NOx comparisons presented here suggest that
Jscale ranges from 25 to 50, while the HONO comparisons support a
larger value of Jscale and/or HONO yields from NITs photolysis
being close to unity. In the context of assessing the large-scale impact of
NITs photolysis, we therefore conclude that a range of 25–100 is the most
appropriate range for the parameter Jscale in our model. We note
that our best estimates of Jscale are about a factor of 4–5
higher than the corresponding scaling coefficient estimated by Reed et
al. (2017) and Ye et al. (2017) based on their box model analysis. This is
because simulated NITs mixing ratios at CVAO in our model (median of
∼100 pptv) are about 4 times lower than the p-NO3 mixing
ratios (∼300–400 pptv) prescribed by Reed et al. (2017) and Ye et
al. (2017) in their box models. As noted in Sect. 3.1, further work is needed
to resolve this model discrepancy.
Impact of NITs photolysis on global NOx, OH, and
O3
Simulated 2014–2015 boundary-layer-average (average over bottom
1 km) accumulation-mode nitrate (NIT) and NOx mixing ratios
for the Jscale25 model (that includes photolysis of only
coarse-mode sea-salt nitrate) and the Jscale25∗ model
(that includes photolysis of both coarse-mode sea-salt and accumulation-mode
NIT) runs. NIT and NOx for the Jscale25 model
are shown in panels (a) and (b), respectively. Panels
(c) and (d) show regions where NIT
(NOx) mixing ratios decrease (increase), and panels
(e) and (f) show regions where NIT (NOx)
mixing ratios increase (decrease), due to accumulation-mode NIT photolysis.
Figure 7 shows the simulated MBL-average distribution of
NOx, OH, and O3 from the Jscale0
model run, as well as the relative enhancements (relative to the
Jscale0 case) of simulated mixing ratios for the
Jscale25 and Jscale100 model runs. In the
Jscale0 simulation, highest MBL mixing ratios of both
NOx and O3 occur downwind of midlatitude Northern
Hemisphere continental source regions and, to a lesser spatial extent,
downwind of tropical biomass burning regions for O3. By contrast,
the highest OH concentrations over oceans occur between 30∘ S and
30∘ N, reflecting the latitudinal pattern of primary OH production
from O3 photolysis. As expected, the additional
source of NOx from NITs photolysis in the
Jscale25 and Jscale100 model runs leads to
enhancements in simulated NOx, OH, and O3
concentrations in the MBL. Downwind of continental regions, MBL
NOx mixing ratios increase by 5–50 pptv in the
Jscale25 simulation and by 10–100 pptv in the
Jscale100 simulation, relative to the Jscale0
simulation (Fig. S2). MBL OH mixing ratios increase by
105–106 molecule cm-3, and MBL O3 mixing ratio
increase by 1–5 ppbv over these same regions (Fig. S2). The simulated
relative impact of NITs photolysis peaks over the tropical and subtropical
oceans, with enhancement factors ranging from 5 to 20 for
NOx, 1.2 to 1.6 for OH, and 1.1 to 1.3 for O3
(Fig. 7). The spatial pattern of these relative enhancements due to NITs
photolysis is thus very different from the spatial distribution of NITs
itself (see Fig. 2). The importance of the tropics is due to intense solar
radiation and high temperatures. These conditions lead to an enhanced source
of NOx from nitrate photolysis, together with a reduced
source from organic nitrates which have a short lifetime in the warmer air
and so have lower concentrations over the remote oceans. This can be
understood by comparing the spatial pattern of the chemical production rate
of NOx from NITs photolysis in the Jscale25
and Jscale100 model runs with the spatial pattern of the
chemical production rate of NOx from PAN decomposition in
the Jscale0 model run, as shown in Fig. 8. In the MBL at
Northern Hemisphere midlatitudes, the chemical source of NOx
from PAN decomposition in the Jscale0 simulation is generally
much larger than the source from NITs photolysis in the
Jscale25 simulation. This is also true for the upper-limit
Jscale100 simulation, though to a lesser extent. By contrast,
the NOx source from NITs photolysis is significantly larger
than the NOx source from PAN decomposition over most of the
tropical MBL. Inclusion of NITs photolysis in the model therefore more
significantly impacts NOx chemical production and,
consequently, simulated NOx mixing ratios, in the tropical
MBL.
Simulated 2014–2015 boundary-layer-average (average over bottom
1 km) OH and O3 mixing ratios for the Jscale25
model run (a and b, respectively), and changes in OH and
O3 mixing ratios (c and d, respectively)
when NIT photolysis is included.
We further infer that the source of MBL NOx from long-range
transport of PAN does not change significantly when NITs photolysis is
included in the model. Above the MBL, maximum increases in PAN concentrations
range from 2 to 5 % and from 5 to 10 % for the Jscale25 and
Jscale100 simulations, respectively, relative to the
Jscale0 simulation. The impact of the additional
NOx source from NITs photolysis on MBL NOx
mixing ratios is modulated by the reduction in NOx lifetime
due to the increase in model OH when NITs photolysis is included (see
Fig. 7). We also note that the inclusion of NITs photolysis improves the
simulation of O3 at CVAO (Fig. 9), offsetting to some extent the
simulated effect of halogen-mediated chemical loss of O3 (Sherwen
et al., 2016b) in the MBL.
While the inclusion of NITs photolysis is important in terms of modeled
NOx, OH, and O3 in the tropical marine atmosphere,
it is important to note that its significance is limited to the boundary
layer, owing to the short lifetime of coarse-mode SSA, and thus of NITs,
against deposition. This is illustrated in Fig. 10, which shows simulated
enhancements in NOx, O3, and OH due to NITs
photolysis for a vertical cross section at 20∘ W. In all cases, the
impact of NITs photolysis on NOx, OH, and O3 is
significant only for altitudes below 1.5–2 km. Very little coarse-mode SSA
gets out of the boundary layer and so the impact of NITs photolysis is
limited to essentially the MBL. This limits the global impact of this new
process. Although Fig. 9 shows some large changes in MBL concentrations, when
averaged over the whole troposphere these changes are relatively small.
Relative to the Jscale0 model run, increases in the
tropospheric masses of NOx, OH, and O3 in the
Jscale25 (Jscale100) model runs are 0.6 %
(1.7 %), 1.1 % (2.9 %), and 0.5 % (1.4 %), respectively.
If photolysis of accumulation-mode SSA nitrate had been included, the impacts
on MBL NOx (in terms of relative changes in concentration)
would be more widespread, especially over remote Southern Hemisphere oceans
where a significant fraction of modeled nitrate is in the accumulation mode
(Fig. S1). The impact in the vertical will likely still be limited, owing to
the relatively short lifetime (∼1 day) of accumulation-mode SSA against
deposition (Jaeglé et al., 2011).
Impact of accumulation-mode p-NO3 photolysis in
continental regions
Although much of the recent research emphasis has been on the photolysis of
nitrate in SSA, there is some evidence that this might be a more general
phenomenon and that p-NO3 photolysis may also occur in other
aerosol types. For example, Ye et al. (2018) found p-NO3 photolysis
to be a significant source of HONO in background terrestrial air masses over
the southern United States. In order to illustrate the potential large-scale
impact of this process, we performed one additional simulation in which we
allow for photolysis of both NITs and accumulation-mode nitrate (identified
as “NIT” in the model). This accumulation-mode aerosol, composed of
sulfate, ammonium, and nitrate, is the dominant form of p-NO3 over
continental regions. In this simulation, we set the photolysis rate scaling
coefficient (i.e., Jscale) to be 25 (chosen for illustrative
purposes) for both NIT and NITs and assume that the HONO:NO2 molar
yield is 0.67:0.33 as in the Jscale25 simulation. We refer
to this simulation by the notation Jscale25∗. The effect
of NIT photolysis is isolated by comparing the Jscale25∗
simulation with the Jscale25 simulation.
Figure 11 shows simulated annual-average NIT and NOx mixing
ratios in the boundary layer from the Jscale25 simulation and
the change in these mixing ratios when NIT photolysis is included in the
Jscale25∗ simulation. Simulated NIT and
NOx mixing ratios are highest (>1 ppbv) in the eastern
United States, western Europe, eastern China, and northern India. Adding NIT
photolysis decreases NIT mixing ratios by around 0.01–0.1 ppbv in these
regions. Surprisingly, the expected increase in NOx from NIT
photolysis is seen in the model results only in northern India. In fact,
adding NIT photolysis decreases simulated NOx in the eastern
United States, western Europe, and most notably in eastern China. There are
also a few model grid boxes, especially in southern India and eastern China,
where NIT mixing ratios increase when NIT photolysis is included. These
counterintuitive features are due to the fact that p-NO3 photolysis
is also a source of OH, and NIT photolysis increases OH in all continental
source regions (Fig. 12). The net impact of NIT photolysis on
NOx in a region is determined by the balance between the
relative importance of the increased NOx source due to NIT
photolysis and the decreased NOx lifetime due to the
increase in OH. The balance between these two counteracting effects on
NOx varies from region to region in the model. In all cases
though, adding NIT photolysis leads to an increase of 1–2 ppbv in
O3 in most regions, with larger increases (>6 ppbv) seen in
northern India and eastern China. From a tropospheric average perspective,
the impact of NIT photolysis is significantly larger than the impact of NITs
photolysis. Relative to the Jscale25 model run, increases in
the tropospheric masses of NOx, OH, and O3 in the
Jscale25∗ model run are 2.8 %, 4.6 %, and 2.6 %,
respectively.
Our results highlight the potential importance of p-NO3 photolysis
in continental high-NOx source regions. We note, however,
that the coarse model resolution used in this study cannot resolve important
chemical gradients in continental source regions. If further field studies
convincingly demonstrate the occurrence of p-NO3 photolysis on
aerosols other than SSA, finer-resolution regional model simulations will be
needed to more accurately assess the effect of p-NO3 photolysis on
continental boundary layer oxidant chemistry.