ACPAtmospheric Chemistry and PhysicsACPAtmos. Chem. Phys.1680-7324Copernicus PublicationsGöttingen, Germany10.5194/acp-17-13669-2017A new diagnostic for tropospheric ozone productionEdwardsPeter M.pete.edwards@york.ac.ukhttps://orcid.org/0000-0002-1076-6793EvansMathew J.https://orcid.org/0000-0003-4775-032XWolfson Atmospheric Chemistry Laboratories, Department of Chemistry, University of York, Heslington, York, YO10 5DD, UKNational Centre for Atmospheric Science, Department of Chemistry, University of York, Heslington, York, YO10 5DD, UKPeter M. Edwards (pete.edwards@york.ac.uk)17November20171722136691368025April20171June201718September201713October2017This work is licensed under the Creative Commons Attribution 3.0 Unported License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/3.0/This article is available from https://acp.copernicus.org/articles/17/13669/2017/acp-17-13669-2017.htmlThe full text article is available as a PDF file from https://acp.copernicus.org/articles/17/13669/2017/acp-17-13669-2017.pdf
Tropospheric ozone is important for
the Earth's climate and air quality. It is produced during the oxidation of
organics in the presence of nitrogen oxides. Due to the range of organic
species emitted and the chain-like nature of their oxidation, this chemistry
is complex and understanding the role of different processes (emission,
deposition, chemistry) is difficult. We demonstrate a new methodology for
diagnosing ozone production based on the processing of bonds contained within
emitted molecules, the fate of which is determined by the conservation of
spin of the bonding electrons. Using this methodology to diagnose ozone
production in the GEOS-Chem chemical transport model, we demonstrate its
advantages over the standard diagnostic. We show that the number of bonds
emitted, their chemistry and lifetime, and feedbacks on OH are all important
in determining the ozone production within the model and its sensitivity to
changes. This insight may allow future model–model comparisons to better
identify the root causes of model differences.
Introduction
The chemistry of the troposphere is one of oxidation (Levy, 1973; Kroll et
al., 2011). Organic compounds together with nitrogen- and sulfur-containing
molecules are emitted into the troposphere where they are oxidised into
compounds which can either be absorbed by the biosphere, be involatile enough
to form aerosols, deposit to the surface, or be taken up by clouds and rained
out. The oxidation of these compounds is significantly slower than might be
expected based on the atmospheric composition of 20 % molecular oxygen
(O2).
The inefficiency of ground-state O2 as an atmospheric oxidant are due to
its electronic structure. In quantum mechanics, all atomic particles have an
intrinsic angular momentum known as spin (Atkins and De Paula, 2014). The
spin of an electron is described by the spin quantum number,
s, and can have values of either +1/2 or -1/2 for a single electron.
The Pauli exclusion principle states that if two electrons occupy the same
orbital, then their spins must be paired and thus cancel. With two unpaired
electrons, ground-state O2 is a spin triplet with a total spin quantum
number S=1/2+1/2=1 (giving a term symbol of 3Σg-).
In contrast, virtually all trace chemicals emitted into the atmosphere
contain only paired electrons and are thus spin singlets (S=0). The
quantum mechanical spin selection rule ΔS=0 means that allowed
electronic transitions must not result in a change in electron spin. From a
simplistic perspective (i.e. ignoring nuclear spin interactions, inter-system
crossings, nuclear dipole effects, etc.) this spin selection rule means that
the reaction of ground-state O2 with most emitted compounds is
effectively spin forbidden. Electronically excited O2 (1Δg or 1Σg+) is a spin singlet and is more reactive in
the atmosphere, but low concentrations limit its role (Larson and Marley,
1999). Instead, atmospheric oxidation proceeds predominantly via reactions
with spin-doublet oxygen-derived species (S=1/2), notably the hydroxyl
(OH) and peroxy radicals (RO2= HO2, CH3O2,
C2H5O2, etc.) or spin-singlet species (e.g. ozone (O3)).
One of the few spin-triplet species in the atmosphere other than O2 is
the ground state of atomic oxygen (O(3P)), which readily undergoes a
spin-allowed reaction with O2 to produce the spin-singlet O3
molecule. This spin-allowed reaction is responsible for the creation of
O3 in both the stratosphere, where it forms the protective O3
layer, and the troposphere. The ability of O3 to oxidise other
spin-singlet species makes it a powerful oxidant, and it is thus considered a
pollutant with negative health effects. Sources of O(3P) within the
troposphere are limited because solar photons at sufficiently short
wavelengths to directly photolyse O2 to O(3P) are essentially
unavailable.
Aside from the photolysis of O3 itself, the only other significant
source of tropospheric O(3P) is the photolysis of nitrogen dioxide
(NO2) (Crutzen, 1971). Nitrogen oxides are emitted into the troposphere
as nitrogen oxide (NO), which can be oxidised to NO2 by O3 and
other oxidants. A large thermodynamic energy barrier prevents the oxidation
of NO to NO2 by the OH radical (Nguyen et al., 1998), and therefore NO
oxidation occurs through reaction with either O3 or RO2. In terms
of O3 production, the oxidation of NO by O3 forms a null cycle.
Thus, only the reaction of NO with RO2 leads to a net production of
O3.
Exploring the distribution, source and sinks of tropospheric O3 is a
central theme of atmospheric science. Chemical transport models (online and
offline) are essential tools enabling this understanding, but their validity
needs to be continually assessed. Model–model comparison exercises are
commonly performed to assess performance, and comparisons of modelled O3
budgets traditionally form part of this assessment (Stevenson et al., 2006;
Wu et al., 2007; Wild, 2007; Young et al., 2013). Ozone production is
diagnosed from the flux of NO to NO2 via reaction with each of the
speciated RO2 in the model's chemical schemes. This approach provides
information on the relative importance of the different RO2 in the fast
NO + RO2 reactions within the model but gives very little detail on
how the longer-timescale model processes (emissions, chemistry, deposition)
influence O3 production. Thus, exploring the reasons that models differ
in their O3 production is difficult and progress has been slow.
A new diagnostic framework that links large-scale model drivers such as
emission, chemistry and deposition to O3 production would allow an
improved assessment of why model ozone budgets differ. We attempt to provide
such a framework here.
A new diagnostic framework
The rate of production of tropospheric O3 is limited by the rate of
oxidation of NO to NO2, which is in turn limited by the rate of
production of peroxy radicals (RO2). Peroxy radicals form through
association reactions of hydrogen (H) atoms or alkyl radicals (both spin
doublets, S=1/2) with O2, forming a highly reactive spin-doublet
radical on an oxygen atom. This spin-allowed reaction converts spin-triplet
O2 that cannot react with spin-singlet pollutants into a spin-doublet
O2-containing species that can. As such the formation of RO2 is
central to the atmosphere's oxidation capacity, and its production is limited
by the rate of production of H atoms or alkyl radicals. Thus, the maximum
potential rate of tropospheric O3 production is equal to the rate at
which H atoms and alkyl radicals are produced.
Hydrogen atoms and alkyl radicals are predominantly produced via the
spin-allowed breaking of the spin pairing between the two electrons in a C-
or H-containing covalent bond (S=0), such as those in hydrocarbons. These
spin pairings can be broken in the atmosphere either chemically or
photolytically, with the products necessarily conserving spin. The breaking
of a covalent bond by a photon (S=1) can result in two products with S=1/2 or two products with S=0. Likewise, oxidation by a radical (S=1/2) will result in one product with S=0 and one with S=1/2 because
the unpaired electron on the radical reactant pairs with one of the
covalent-bond electrons to produce a spin singlet.
Although the majority of RO2 is formed from emitted C- or H-containing
covalent bonds, there are a few notable exceptions. Hydrogen atoms can also
be produced through the oxidation of CO to CO2 by OH. During this
reaction the coordinate bond between the C and O atom is broken and the H
atom is produced via the breaking of the O–H bond. The other notable
exception is the oxidation of an SO2 lone pair of electrons to SO3
by OH, where again the H atom produced comes from the OH. In both of these
exceptions a spin-singlet electron pairing (CO coordinate bond or SO2
lone pair) is broken during the production of the H atom, and we can
therefore consider these reactions similar to the breaking of a C- or
H-containing covalent bond. For simplicity these spin-singlet electron
pairings that can be broken in the troposphere to produce either a H atom or
an alkyl radical will be referred to as “oxidisable bonds” (C–C, C–H,
C=C, CO coordinate bond, S:).
Peroxy radical production during the tropospheric oxidation of
CH4. Moving from left to right, the oxidisable bonds (emitted: red;
produced: blue) present in CH4 are removed via a range of tropospheric
processes, indicated by the coloured arrows. The large numbers across the top
of the figure indicate the number of oxidisable bonds at each stage of this
oxidation. The production of RO2 is indicated by the +1/+2 numbers
with the associated process arrows for producing one or two RO2
respectively.
Tropospheric O3 production occurs through the oxidation of NO by
RO2. Following the above rationale, these RO2 are produced during
the spin-allowed breaking of oxidisable bonds predominantly contained within
emitted volatile organic compounds (VOCs). This perspective allows us to build a new metric for the
production of tropospheric O3 based around the spin-conserving
properties of oxidisable bond breaking. In the extreme case, all oxidisable
bonds are photolysed to produce two spin-doublet RO2 products, which
then react exclusively with NO to generate O3. Thus, at steady state,
the maximum rate of O3 production is equal to the rate of production of
RO2, which is equal to twice the rate of destruction of the number of
oxidisable bonds. This in turn is equal to twice the rate of emission of
oxidisable bonds. Deviation from this maximum is determined by
the relative importance of processes that produce spin-singlet vs.
spin-doublet products during oxidisable bond breaking;
the fraction of spin-doublet products from oxidisable bond breaking which
form RO2;
the fraction of RO2 that go on to oxidise NO to NO2.
To illustrate this, Fig. 1 shows the tropospheric oxidation of a methane
(CH4) molecule through various steps to either a carbon dioxide
(CO2) molecule or a species that is deposited (CH3OOH, CH2O,
CH3NO3). Methane contains four times C–H oxidisable bonds
(eight paired bonding electrons) and as the oxidation proceeds, the number of
oxidisable bonds decays to 0. Figure 1 highlights the steps in the
tropospheric CH4 oxidation mechanism that form spin-doublet products,
with between one and 5 RO2 produced depending on the oxidation pathway.
This compares with the theoretical maximum of eight if all the original C–H
bonds were photolysed to yield two spin-doublet products.
The principal atmospheric source of oxidisable bonds is the emission of C–H,
C–C and C=C bonds in hydrocarbons, with the only other significant sources
being the emission of CO and the chemical production of CO and H2 during
hydrocarbon oxidation. Over a long enough timescale, the global atmosphere
can be considered to be in a chemical steady state, where the rate of loss of
oxidisable bonds is balanced by the rate of production or emission. Thus, the
O3 production rate can be described by Eq. (1), where the O3
production metric PsO3 is equal to the number of spin-paired
electrons in oxidisable bonds (i.e. twice the sum of the number of oxidisable
bonds emitted, Ebonds, and chemically produced,
Pbonds), multiplied by the number of spin-doublet radicals
produced per oxidisable bond break divided by the maximum of 2
(FRadicals), in turn multiplied by the fraction of the radicals
produced which are RO2 (FRO2), multiplied by the fraction
of RO2 that goes on to react with an NO to produce an O3 molecule
(FNO). A small correction (I) for the production of RO2
via reactions of spin-doublet radicals other than those that result in the
breaking of oxidisable spin pairings (e.g. O3+ OH → HO2+ O2) is included.
PsO3=2×Ebonds+Pbonds×Fradicals×FRO2+I×FNO
Implementation
We use the GEOS-Chem model to evaluate this new O3 production
diagnostic. GEOS-Chem is a global chemical transport model of tropospheric
chemistry, aerosol and transport (http://www.geos-chem.org version
9-02). The model is forced by assimilated meteorological and surface fields
(GEOS-5) from NASA's Global Modelling and Assimilation Office and was run at
4∘× 5∘ spatial resolution. The model chemistry
scheme includes Ox, HOx, NOx, BrOx and VOC chemistry as
described in Mao et al. (2013) as are the emissions. The new
PsO3 diagnostic has been implemented via the tracking of
reactions by type in the GEOS-Chem chemical mechanism file (further details
given in the Supplement). This tracking of reactions enables the fate of all
oxidisable bonds as well as the production and loss of all RO2 within
the model to be determined using the standard GEOS-Chem production and loss
diagnostic tools. Model simulations were run for 2 years (1 July 2005–1 July
2007) with the first year used as a spin-up and the diagnostics performed on
the second year.
Flow of oxidisable bonds to O3 production in the GEOS-Chem base
simulation. Arrows are coloured according to process, and the arrow thickness
is proportional to the flux through that channel. Spin-paired electrons are
input as oxidisable bonds into the model (left arrow), with the potential to
create 778 T mol yr-1 of radicals. The actual fate of these bonds is
shown in the central arrow, producing 280 T mol yr-1 of RO2, of
which 112 T mol yr-1 reacts with NO to produce O3 (right
arrow).
The standard GEOS-Chem diagnostic for O3 production (PO3) is
shown on the left side of Table 1. This emphasises the very fast cycling
between NO and NO2, but provides little in terms of higher process-level
information. The right side of Table 1 shows the new budget for
PsO3, which tracks the processing of oxidisable bonds within
the model. Both diagnostic methods give the same final answer, but our new
methodology provides more process-level detail. Figure 2 illustrates this new
process-based approach, showing the flow of emitted oxidisable spin-paired
electrons (bonds) to O3 and the magnitude of the various mechanisms that
contribute to and compete with O3 production. The annual oxidisable bond
emission of 389 T mol yr-1 has the potential to create
778 T mol yr-1 of radicals. If all oxidisable bonds were broken by
photons to produce two radical products, the RO2 production would be
778 T mol yr-1. If the oxidisable bonds were instead broken via
radical reaction (e.g. OH), then RO2 production would be
389 T mol yr-1. The various oxidisable bond-breaking/removal pathways
within the model result in the production of 280 T mol yr-1 of
RO2, with the remainder largely producing stable spin-singlet products.
Pie charts showing hydrocarbon emissions in the base GEOS-Chem
simulation. Emissions split by carbon mass (left), number of oxidisable bonds
(centre), and bond type (right).
Of the 280 T mol yr-1 RO2 produced, 112 T mol yr-1
reacts with NO to produce O3. The remainder is lost through the reaction
or deposition of RO2 reservoir species
(RO2y= RO2+ peroxides + peroxy-acetyl nitrates). For
example the production of methylperoxide
(CH3O2+ HO2= CH3OOH) results in the loss of two
RO2s. However, the reaction of methylperoxide with OH can re-release
CH3O2 (CH3OOH + OH = CH3O2+ H2O).
Thus, the production of methylperoxide represents the loss of a HO2 and
the movement of a CH3O2 into a peroxide RO2y reservoir
species. The deposition of a peroxide molecule is thus the loss of a
RO2y reservoir species. Notable in Fig. 2 is that the role of PAN and
nitrate removal of global RO2y is negligible, instead being dominated by
peroxide production and loss and the reaction of RO2 with O3.
Comparison of ozone production diagnostics for GEOS-Chem base
simulation. Standard model PO3 diagnostics (left column) show
reactions responsible for NO to NO2 conversions but provide little
process-level information. The new PsO3 (right) provides
increased information on the processes controlling O3 production within
the model.
PO3/T mol yr-1PO3/T mol yr-1(except FRadicals, FRO2 and FNO, which are all unitless) NO + HO2→ NO274Ebonds330NO + CH3O2→ NO227Pbonds58Other RO2+ NO → NO210Fradicals0.40Other1FRO20.86Inorganic RO2 source15FNO0.40PO3112PsO3112Emitted oxidisable bonds
The fuel for tropospheric oxidation chemistry is the emission of oxidisable
bonds, predominantly in the form of hydrocarbons. The production of
tropospheric O3 from the spin-paired bonding electrons emitted into the
standard GEOS-Chem model occurs with an efficiency of 14 %
(112 T mol yr-1 molecules of O3 produced/778 T mol yr-1
spin-paired electrons emitted as oxidisable bonds; Fig. 2). These spin-paired
bonding electrons are predominantly emitted in the form of CH4, isoprene
(C5H8) and CO (37, 28 and 9 % respectively). Oxidisable bonds
produced during chemical reactions (Pbonds) account for 15 %
of the net source. Figure 3 shows emissions of CO and hydrocarbons in the
standard GEOS-Chem simulation in terms of mass of carbon per compound, as the
number of oxidisable bonds per compound and as the number of bonds in
different oxidisable bond types. The commonly used carbon mass approach
splits emissions approximately equally between each of the major sources
(CH4, 29 %; isoprene, 32 %; and CO, 30 %). In contrast, the
oxidisable bonds accounting approach apportions hydrocarbon emissions of 44,
33 and 11 % for CH4, isoprene and CO respectively. This highlights
the high number of oxidisable bonds per carbon atom in CH4 (4) compared
to isoprene (2.8) and CO (1). Thus, efforts to consider emissions on a
per-bond basis may provide more insight into chemical processes, as it is
these bonds that ultimately determine the chain-like chemistry rather than
the mass of carbon atoms. This helps to emphasise the relative importance of
CH4 emissions in global tropospheric chemistry compared with other
emissions such as isoprene or CO. The type of oxidisable bond emitted is
overwhelmingly C–H (71 %).
The total emission and production of oxidisable bonds has the potential to
create 778 T mol yr-1 of radicals. However, only 6 % of the
oxidisable spin pairings are broken to give the maximum two spin-doublet
products (e.g. radical channel of CH2O photolysis). The majority
(68 %) are oxidised via reaction with a spin-doublet species (OH) to
produce one spin-singlet and one spin-doublet product (e.g. OH + VOC).
The remaining 26 % of spin-paired electrons are removed to form two
spin-singlets (e.g. the non-radical channel of CH2O photolysis). Thus,
of the 778 T mol yr-1 spin-paired electrons emitted or produced, only
265 T mol yr-1 (34 %) are converted into RO2, with an
additional 15 T mol yr-1 produced from reactions such as
O3+ OH,→ HO2+ O2 (I). The efficiency of
O3 production from the available oxidisable bonds is further reduced as
only 40 % of the 280 T mol yr-1 of RO2 produced react with
NO to produce NO2. The remainder is lost either through the
self-reaction of RO2 or via loss through deposition or reaction of
RO2y reservoir species (e.g. peroxides). Thus, overall 14 % of the
emitted bonding electrons go on to make O3.
Understanding the effect of NOx and VOC emissions on ozone
production at the process level. Stack plots showing fractional change in
model PO3 compared to base simulation and associated contributions
from the current PO3 (i) and new PsO3 (ii) diagnostic
parameters under changing NOx emissions (a), effective CH4
emission (b) and isoprene emission (c). The
PsO3 diagnostic parameters are derived for each model
simulation using the diagnostic implementation described in Sect. 3 and the
fractional change in each parameter from the base simulation calculated.
The new O3 production diagnostic presented here (PsO3)
shows the impact of processes such as emission, deposition and chemical
mechanism and provides significantly more detail than the standard PO3
diagnostic approach (Table 1). We now explore the sensitivity of model
O3 production to changing emissions of NOx and VOC from the
perspective of the two diagnostic methods.
Model sensitivities
Understanding model response to changing emissions is an important tool for
considering policy interventions. The major controls on O3 production
are emissions of NOx and VOCs. We show in Fig. 2 that from the
perspective of global O3 production, oxidisable bond emissions are
dominated by CH4 and isoprene. Figure 4 shows the impact of changing
emissions of NOx, isoprene and CH4 on O3 production from both
the perspective of this new methodology and the conventional NO+RO2
diagnostic approach. A set of five simulations was performed for each model
sensitivity investigated (NOx, isoprene and CH4), with a common
base simulation, resulting in 13 simulations in total. The following sections
investigate these model responses and use the new diagnostic to provide
insight into the processes driving the observed response in O3
production.
NOx emissions
Figure 4a diagnoses the relative response of GEOS-Chem O3 production to
changing NOx emissions, using simulations where NOx emissions from
anthropogenic, biomass burning, biofuels, soil and lighting sources were
multiplied by factors of 0.5–2. Increasing NOx emissions increases
O3 production. The standard RO2+NO diagnostic (Fig. 4ai) shows that
fractional contributions to the total change in PO3 from HO2
(67 %), methyl-peroxy (CH3O2) (25 %) and other RO2
(8 %) remain approximately constant across the NOx emission range
investigated. This diagnostic provides little detail on the processes driving
the change in O3 production under changing NOx emissions. In
contrast, Fig. 4aii is based on the new PsO3 diagnostic and
shows a range of process-level changes occurring as NOx emissions
change.
Fractional change in new PsO3 diagnostic parameters
from base run against changing NOx emission (a); effective
CH4 emission (b); and isoprene emission (c).
Impact of changing NOx emission on FNO
Unsurprisingly, as NOx emissions increase the fraction of RO2
reacting with NO to produce NO2 (FNO) increases (red section
in Fig. 4aii). However, this impact only accounts for around 40 % of the
increase in PsO3. Figure 5a shows the fractional change in
all the PsO3 efficiency parameters and the global mean
NOx concentration as a function of the changing NOx emission. As
NOx emissions increase, the increase in NOx concentration in the
model is somewhat dampened. Halving the NOx emission leads to NOx
burdens dropping by ∼ 35 %, and doubling leads to an increase of
95 %. This dampening is due to the impact of NOx emissions on OH
(see Sect. 4.1.2), which is the dominant sink for NOx. Increasing
NOx increases OH concentrations, which in turn shortens the NOx
lifetime, thus dampening the response of concentration to emission.
The response of FNO to changes in NOx emissions is also
dampened relative to the change in NOx emissions. This is due to spatial
variability in FNO, which is not affected uniformly by changing
NOx emissions. Figure 6 shows the probability distribution of
FNO values across all model grid boxes for the base simulation
and the half and doubled NOx emission simulations (black, blue and red
lines respectively). For example, in a grid box in the continental boundary
layer where RO2 reacts overwhelmingly with NO, doubling the NOx
emission may move FNO from 0.90 to 0.95, but it cannot double it.
Similarly, in the remote boundary layer where RO2 reacts overwhelmingly
with other RO2, doubling NOx emissions may move FNO
from 0.3 to 0.4, but again it does not double. Thus, the geographical spread
of NOx chemistry limits the change in FNO caused by changing
NOx emissions. The spatial variability in the new PsO3
diagnostic parameters shows that this approach has significant potential in
the analysis of regional O3 budgets as well as global.
Impact of changing NOx emission on Ebonds
Figure 4aii shows that 60 % of the response in PsO3 to
changing NOx emission is due to factors other than FNO, with
40 % of the increase due to changes in the emissions (Ebonds:
32 %) and chemical production (Pbonds: 8 %) of oxidisable
bonds. This increase in Ebonds is surprising given VOC emissions
are unchanged in these simulations. However, increasing NOx emissions
results in an increased OH concentration in the model, which then leads to an
increase in CH4 oxidation. Methane (CH4) concentrations are fixed
in GEOS-Chem, resulting in an increase in the effective CH4 emission as
OH concentrations increase, causing an increase in the total bond emission
(Ebonds). Figure 7 shows the response of effective CH4 bond
emission to global mean OH concentration as it changes with global mean
NOx concentration. More CH4 oxidation also leads to more CH2O
production and in turn more CO production (PCO), accounting for a
significant fraction of the increase in this term.
Effect of NOx emission on distribution of FNO
values (log scale). FNO values for each model grid box in the
base and NOx emission × 0.5 and × 2 simulations, split
into 50 × 0.02 width bins.
Impact of changing NOx emission on Fradicals, FRO2 and
I
The fraction of radicals produced from bond oxidation (Fradicals)
and the fraction of those radicals which are RO2 (FRO2)
show a slight positive increase with NOx emission, accounting for 9 and
6 % of the change in PsO3 respectively. This reflects
changes in the partitioning of the fate of the oxidisable bonds and is
largely due to the changes in OH. As OH increases with NOx emission, the
rate of chemical oxidation of bonds increases at the expense of other losses,
in particular deposition. The inorganic RO2 source term (I) also
correlates with NOx emission, as it is largely determined by the
concentrations of OH and O3. This change accounts for 5 % of the
observed change in PsO3.
Effective CH4 emissions as a function of global mean OH
concentration for simulations where NOx emissions were changed. Marker
size and colour indicate global NOx concentration.
Thus, with this new diagnostic methodology, it is evident that only 40 %
of the model O3 production response to changing NOx emission is due
to the direct effect of increasing NO concentration on the rate of
RO2+ NO reactions. Another 40 % is due to fixing the
concentration of CH4 within the model, with the final 20 % due to
the increased OH concentration competing for the available oxidisable bonds
and resulting in increased RO2 production.
Changing effective CH4 emissions
As Fig. 2 shows CH4 to be the largest single source of oxidisable
bonds, this section investigates the response of the O3 production
diagnostics to changing CH4 emissions. Figure 4b shows the O3
production diagnostics response to varying the CH4 emission rate within
the model. As the model uses prescribed CH4 concentrations, these were
varied by factors of between 0.5 and 2 from the base simulation and the
CH4 emission diagnosed from the loss rate of CH4 to reaction with
OH, the only CH4 loss in the model. We describe this as the effective
CH4 emission.
As effective CH4 emission increases, O3 production also increases.
The standard diagnostic (Fig.4bi) shows that this increase occurs through an
increased rate of reaction of HO2 and CH3O2 with NO, as would
be expected as these are the RO2s produced during CH4 oxidation.
The rate of other RO2+ NO reactions actually decreases slightly as
CH4 emissions increase, due to lower OH concentrations and increased
competition for NO from HO2 and CH3O2. The new diagnostic
(Fig.4bii), however, shows that the increase in O3 production with
increasing effective CH4 emission is not simply a result of more
HO2 and CH3O2.
Impact of changing effective CH4 emission on FNO
The observed change in PsO3 is around one-third smaller than
would be expected from the increase in the oxidisable bond emission
(Ebonds) and bond production (Pbonds) terms alone.
This is due to a countering decrease in the other efficiency parameters with
increasing effective CH4 emission. Figure 5b shows the fractional change
in all the efficiency parameters as a function of the changing effective
CH4 emission. The decrease in the fraction of RO2 reacting with NO
to produce NO2 (FNO) is driven by increasing O3
concentrations, which push the NO / NO2 ratio towards NO2. This
reduces the availability of NO to react with RO2, thereby reducing
O3 production. This shift in the NO / NO2 ratio also increases
NOx loss within the model with increasing CH4 emission, as the
increased CH4 oxidation increases RO2 concentrations resulting in
larger losses of NO2 via compounds such as peroxyacetyl nitrate (PAN)
and peroxynitric acid (PNA).
Impact of changing effective CH4 emission on Ebonds
Increasing the effective CH4 emission results in an increase in
Ebonds. Changing the fraction of total emitted oxidisable bonds
from CH4 does, however, have significant consequences for the loss
mechanisms of these bonds, which influences the other efficiency parameters.
Figure 8 show the split of oxidisable bond loss mechanisms in the base
simulation and those with the CH4 concentration fields multiplied by 0.5
and 2. As the effective CH4 emission increases the fraction of bonds
lost via OH decreases, despite the actual number of oxidisable bonds lost to
OH increasing. A larger fraction of bonds are therefore lost via the other
mechanisms shown in Fig. 8 rather than reaction with OH. As CH4 removal
occurs predominantly in the free troposphere, increasing the effective
CH4 emission also results in a reduction in the fraction of oxidisable
bonds lost via deposition. The largest fractional increase in bond loss
mechanism with increasing effective CH4 emission is for photolysis, with
the increase in the “other” fraction due to the increased loss of bonds to
the stratosphere with increasing CH4.
Oxidisable bond loss mechanism fractions under changing effective
CH4 emissions (0.5 × CH4 concentration field, base
simulation and 2 × CH4 concentration field).
Impact of changing effective CH4 emission on Fradicals,
FRO2 and I
The fraction of oxidisable bonds that goes on to produce radicals
(Fradicals) and the fraction of these that are RO2
(FRO2) also decrease with increasing effective CH4
emissions. This is due to decreasing global OH concentration resulting from
increased loss by reaction with CH4 and a decreasing NO concentration.
This favours bond loss via pathways that produce less RO2 (e.g.
CH2O photolysis). The long lifetime of CH4 compared with the
majority of other sources of oxidisable bonds also results in a decrease in
the fraction of bonds lost to deposition as total bond oxidation increases
fractionally in the free troposphere where deposition is a less significant
loss mechanism than in the boundary layer.
Changing isoprene emission
The species through which the oxidisable bonds are emitted has a significant
impact on O3 production, due to their subsequent removal mechanisms. For
example, in a simulation where the only emission of oxidisable bonds is CO,
Fradicals is 0.5 and FRO2 is 1 as the only CO sink
is reaction with OH to produce one HO2 (OH + CO → HO2+ CO2). The CO coordinate bond, which in theory has the
potential to produce two radicals, only produces one radical, which is an
RO2.
The effect of oxidisable bond parent species on OH, HO2,
O3 and NOx concentrations. Global mean [OH], [HO2], [O3]
and [NOx] for simulations where the effective CH4 emission (solid
lines) and isoprene emission (dashed lines) were changed, against model
Ebonds. The dashed vertical green line indicates
Ebonds in the base simulation (330 T mol yr-1).
Isoprene has the most complex chemistry in the model and is the
second-largest source of bonds for the atmosphere after CH4 (Fig. 3).
Figure 4c shows the response of the two O3 production diagnostics to
varying the isoprene emission within the model. The standard diagnostic
(Fig. 4ci) shows that the most significant increase in PO3 from
increasing isoprene emissions is from NO + HO2 and
non-CH3O2 peroxy radicals, with a smaller increase from
CH3O2. The new PsO3 diagnostic (Fig. 4cii) again
provides more insight, showing significant offsetting of around 0.5 between
the terms.
Impact of changing isoprene emission on FNO
The increased isoprene emission leads to a similar change in the magnitude of
the total number of oxidisable bonds emitted (Ebonds) as the
simulations in which effective CH4 emission were varied. However, the
countering decrease in all of the efficiency parameters is much larger for
isoprene than for CH4. Figure 5c shows the fractional change in the new
PsO3 ozone production diagnostic parameters as a function of
isoprene emissions compared to the base simulation. The change in
FNO is due to both a decrease in global mean NOx
concentrations with increasing isoprene and the spatial distribution of
isoprene emissions. The majority of global isoprene emissions are in regions
with low NOx emissions and thus low values of FNO. Figure 9
shows a decrease in global mean NOx and global mean OH concentrations
with increasing isoprene emissions; however, the effect is less than that
seen when CH4 is responsible for the same increase in oxidisable bond
emission. This is due in a large part to the spatial scales on which the two
compounds impact.
Impact of changing isoprene emission on Ebonds
As isoprene is the second-largest source of oxidisable bonds (Fig. 3),
increasing the isoprene emission results in a significant increase in
Ebonds. Differences in both the spatial distribution of emissions
and the oxidation chemistry of isoprene and CH4, however, means that the
impact of the increases in Ebonds on O3 production are
significantly different for the two compounds. This is predominantly because
the fraction of oxidisable bonds that are physically deposited for isoprene
is high compared to those emitted as CH4. This increase is (i) due to
the higher solubility of isoprene oxidation products compared to those of
CH4 and (ii) because the higher reactivity of isoprene means its
oxidation occurs in the boundary layer where both dry and wet deposition are
most effective.
Figure 10 shows the fate of oxidisable bonds in the base simulation and those
with the isoprene emissions multiplied by 0.5 and 2. The complex myriad of
products formed during the isoprene oxidation mechanism also results in the
production of many highly oxygenated multifunctional compounds with high
Henry's law solubility constants, meaning they are more readily lost to
deposition.
Increasing the isoprene emission also has a slight offsetting impact on the
effective CH4 emission, as increased isoprene concentrations decrease OH
concentrations and thus decrease the effective CH4 emission. A doubling
in isoprene emission causes a 6 % reduction in the effective emission of
CH4.
Oxidisable bond loss mechanism fractions under changing isoprene
emissions.
Impact of changing isoprene emission on Fradicals, FRO2 and
I
As shown in Fig. 3cii, increasing the isoprene emission results in a
reduction in all PsO3 efficiency parameters. The reductions
in Fradicals is due to the higher fraction of oxidisable bonds
that are lost via non-radical forming pathways (e.g. deposition) for isoprene
relative to the other main oxidisable bond emission sources CH4 and CO.
The slight decreases in FRO2 and I with increasing isoprene
emission are predominantly due to changes in OH and NOx (Fig. 9).
The complex chemistry of isoprene oxidation combined with the spatial
distribution of isoprene emissions means that the increase in O3
production due to increases in isoprene emissions is roughly half what might
be expected from the increase in oxidisable bond emission alone (i.e. if the
increase was via CO instead of isoprene).
Conclusions
We have shown that this bond-focussed approach to O3 production provides
a significantly more detailed understanding of the processes involved. The
role of modelled VOC emissions and O3 burden has been reported
previously (Wild, 2007; Young et al., 2013). However, previous efforts
extending this to a general process-led approach have not been successful.
This new approach provides a tool with which the processes controlling
O3 production can be investigated and a metric by which different
emissions can be compared. For example, the differing chemistry of isoprene
and CH4 shows that even though their emissions of carbon mass are
comparable, the atmosphere responds in different ways, with the isoprene
bonds being less effective in producing O3 than CH4 bonds. By
quantifying multiple steps in the O3 production process, competing
changes in the system become apparent (as shown in Fig. 4bii and cii) and are
thus testable. This enables the effect of model approximations on O3
production to be quantified (e.g. the effect of NOx on CH4
emissions when using CH4 concentration fields).
This new diagnostic also points towards the importance of observational
datasets for assessing our understanding of tropospheric chemistry. Although
the budget presented in Fig. 2 provides an annually integrated global
estimate, it points towards local comparisons that can be made to assess
model fidelity. Comparisons, both their magnitude and their ratios, between
observed and modelled bond concentration, bond emission and loss fluxes (e.g.
OH reactivity, Yang et al., 2016, or depositional fluxes, Wesely and Hicks,
2000), and O3 production (Cazorla and Brune, 2010) would all provide
comparisons for outputs from the PsO3 diagnostic and help
assess model performance.
Future work is necessary to identify the usefulness of this approach on
smaller spatial and temporal scales. For a regional modelling scale, the
transport flux of bonds into the domain would need to be considered alongside
the emissions of bonds. However, this might help to disentangle O3
production due to local VOC emissions from that due to VOC emissions outside
of the domain. This bond focussed approach may also have usefulness on
shorter timescales. For example, when considering vertical fluxes in and out
of the boundary layer, a bond-centred approach could help. What fraction of
the bonds emitted at the surface are exported to the free troposphere? If a
measurement of reactivity flux could be made, this could be tested
experimentally.
Another potentially important application is in model–model comparisons.
Increases in our understanding of why different models calculate different
O3 production and burdens has been slow (Stevenson et al., 2006; Wu et
al., 2007; Young et al., 2013). Although a complete tagging like that
described here is unlikely to occur for all of the models involved in the
comparison, a small number of additional diagnostics is likely to produce a
significantly better understanding of the models. Diagnosing (1) the total
bond flux (direct emissions plus the flux for those species kept constant),
(2) the rate of production of RO2 and (3) the rate of production of
O3 could help differentiate why certain models produce more or less
O3 than others. The ratios between these fluxes would help identify what
aspect of the emissions of chemistry differs between the models.
In order to enable replication of this work, the mechanism
tagging data and approach are tabulated in the Supplement. Individual model
outputs can be made available upon request.
The Supplement related to this article is available online at https://doi.org/10.5194/acp-17-13669-2017-supplement.
All work presented here was conceived by PME and MJE. The
implementation, model simulations and analysis were carried out by PME,
and the manuscript was written by PME with substantial input from MJE.
The authors declare that they have no conflict of
interest.
Acknowledgements
Peter M. Edwards was supported by NERC Grant NE/K004603/1. This work was also
supported by the NERC funded BACCHUS project (NE/L01291X/1). GEOS-Chem
(http://www.geos-chem.org) is a community effort, and we wish to thank
all involved in the development of the model.
Edited by: Paul Monks Reviewed by: two anonymous referees
References
Atkins, P. W. and De Paula, J.: Atkins' Physical chemistry, 10th ed.,
Oxford University Press, Oxford, UK, 2014.Cazorla, M. and Brune, W. H.: Measurement of Ozone Production Sensor, Atmos.
Meas. Tech., 3, 545–555, 10.5194/amt-3-545-2010, 2010.Crutzen, P. J.: Ozone production rates in an oxygen-hydrogen-nitrogen oxide
atmosphere, J. Geophys. Res., 76, 7311–7327, 10.1029/JC076i030p07311,
1971.Kroll, J. H., Donahue, N. M., Jimenez, J. L., Kessler, S. H., Canagaratna, M.
R., Wilson, K. R., Altieri, K. E., Mazzoleni, L. R., Wozniak, A. S., Bluhm,
H., Mysak, E. R., Smith, J. D., Kolb C. E., and Worsnop, D. R.: Carbon
oxidation state as a metric for describing the chemistry of atmospheric
organic aerosol, Nat. Chem., 3, 133–139, 10.1038/nchem.948, 2011.
Larson, R. A. and Marley, K. A.: Singlet oxygen in the environment, Environ.
Photochem., 2, 123–136, 1999.Levy, H.: Photochemistry of minor constituents in the troposphere, Planet.
Space Sci., 21, 575–591, 10.1016/0032-0633(73)90071-8, 1973.Mao, J., Paulot, F., Jacob, D. J., Cohen, R. C., Crounse, J. D., Wennberg, P.
O., Keller, C. A., Hudman, R. C., Barkley, M. P., and Horowitz, L. W.: Ozone
and organic nitrates over the eastern United States: Sensitivity to isoprene
chemistry, J. Geophys. Res.-Atmos., 118, 1–13, 10.1002/jgrd.50817, 2013.Nguyen, M. T., Sumathi, R., Sengupta, D., and Peeters, J.: Theoretical
analysis of reactions related to the HNO2 energy surface: OH + NO
and H + NO2, Chem. Phys., 230, 1–11,
10.1016/S0301-0104(97)00383-2, 1998.Stevenson, D. S., Dentener, F. J., Schultz, M. G., Ellingsen, K., van Noije,
T. P. C., Wild, O., Zeng, G., Amann, M., Atherton, C. S., Bell, N., Bergmann,
D. J., Bey, I., Butler, T., Cofala, J., Collins, W. J., Derwent, R. G.,
Doherty, R. M., Drevet, J., Eskes, H. J., Fiore, A. M., Gauss, M.,
Hauglustaine, D. A., Horowitz, L. W., Isaksen, I. S. A., Krol, M. C.,
Lamarque, J.-F., Lawrence, M. G., Montanaro, V., Müller, J.-F., Pitari,
G., Prather, M. J., Pyle, J. A., Rast, S., Rodriguez, J. M., Sanderson, M.
G., Savage, N. H., Shindell, D. T., Strahan, S. E., Sudo, K., and Szopa, S.:
Multimodel ensemble simulations of present-day and near-future tropospheric
ozone, J. Geophys. Res.-Atmos., 111, D08301, 10.1029/2005JD006338,
2006.
Wesely, M. L. and Hicks, B. B.: A review of the current status of knowledge
in dry deposition, Atmos. Environ., 34, 2261–2282, 2000.Wild, O.: Modelling the global tropospheric ozone budget: exploring the
variability in current models, Atmos. Chem. Phys., 7, 2643–2660,
10.5194/acp-7-2643-2007, 2007.Wu, S., Mickley, L. J., Jacob, D. J., Logan, J. A., Yantosca, R. M., and
Rind, D.: Why are there large differences between models in global budgets of
tropospheric ozone?, J. Geophys. Res.-Atmos., 112, 1–18,
10.1029/2006JD007801, 2007.Yang, Y., Shao, M., Wang, X., Nölscher, A. C., Kessel, S., Guenther, A.,
and Williams, J.: Towards a quantitative understanding of total OH
reactivity: A review, Atmos. Environ., 134, 147–161,
10.1016/j.atmosenv.2016.03.010, 2016.Young, P. J., Archibald, A. T., Bowman, K. W., Lamarque, J.-F., Naik, V.,
Stevenson, D. S., Tilmes, S., Voulgarakis, A., Wild, O., Bergmann, D.,
Cameron-Smith, P., Cionni, I., Collins, W. J., Dalsøren, S. B., Doherty,
R. M., Eyring, V., Faluvegi, G., Horowitz, L. W., Josse, B., Lee, Y. H.,
MacKenzie, I. A., Nagashima, T., Plummer, D. A., Righi, M., Rumbold, S. T.,
Skeie, R. B., Shindell, D. T., Strode, S. A., Sudo, K., Szopa, S., and Zeng,
G.: Pre-industrial to end 21st century projections of tropospheric ozone from
the Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP),
Atmos. Chem. Phys., 13, 2063–2090, 10.5194/acp-13-2063-2013,
2013.