Introduction
With an increasing proportion of the world's population living in
cities, rising from only 3 % in the 1800's to over
47 % by the end of the 20th Century
, the impact of urban air pollution has become
a significant factor in global health . The
costs of air pollution are high even in those locations that have seen
considerable improvements in air quality over the past decades
; in the UK exposure to particulate matter (PM) alone is
estimated to reduce life expectancy on average by around 7–8 months,
with a cost to society estimated at up to GBP 20 billion per
year .
Primary urban air pollution emissions are dominated by PM, nitrogen
oxides (NOx), carbon monoxide (CO) and volatile organic
compounds (VOCs). Many of these species can react in the atmosphere to
create secondary pollutants, such as ozone (O3), oxygenated
VOCs (OVOCs), peroxy acyl nitrates (PANs) and condensed materials in
the form of secondary organic aerosol (SOA), which add to the overall
PM load .
Air quality in London has been controlled and monitored for over
60 years, making it in theory one of the better understood
atmospheres of the world's megacities. Current measurements in London
focus on assessing national compliance with legally prescribed air
quality standards, and this includes the hydrocarbons 1,3-butadiene
and benzene. However, compliance measurements in themselves are
insufficient to fully describe the chemical and physical processes
occurring in the urban atmosphere , and
a particular weakness lies in speciating the many different classes of
carbon compounds in urban air.
The past 2 decades have seen declining concentrations of most
smaller hydrocarbons in European and US cities, a result of tighter
regulation of sources such as vehicle exhaust, evaporation and
solvents , better control of
natural gas leakage and an overall switch from gasoline to diesel
powered vehicles. Current national emissions estimates suggest that
the bulk of organic emissions to air are associated with smaller
hydrocarbons, and this has driven policy, regulation and
observation strategies for
compliance. Figure shows
Government-estimated emissions for the UK (left) and US (right),
categorised into the dominant emission sources. It is clear that based
on current emission inventories, gaseous organic emissions from diesel
appear to represent a negligible fraction of reactive carbon released
into the atmosphere. The demand for diesel fuel is expected to
increase by 75 % between 2010 and 2040 and by 2020 it is
expected to overtake gasoline as the number one transport fuel used
worldwide . The environmental impacts of this
change are evaluated in part based on the national emission
inventories that underpin Fig. .
Total mass by carbon number and functionality from UK 2012 (left)
and US 2011 (right) emission inventories. The carbon number and functionality
of emissions have been estimated by applying the speciated inventory of
emission sources of to the most recent estimates of
non-methane hydrocarbon source apportionment for each country (full details
can be found in the Supplement).
![]()
The efficiency with which O3 and SOA can be formed from diesel
or gasoline emissions is dependent on the mass of available organic
carbon, and the reactivity and volatility of that material
. To quantify this requires individual speciation
of VOCs in order that each property can be properly estimated. The key
urban sources of organic compounds include combustion products,
unburnt fuels and evaporative emissions of fuels and solvents, all of
which are highly complex, often propagating the original complexity of
fossil fuels into the air. Whilst each VOC has a unique set of
reaction mechanisms, in general terms, as the carbon number increases,
the relative complexity of reactions and yields of SOA and O3
also increase . The organic mixture in air is
complicated further by the presence of secondary oxygenated products. This requires a combined approach to investigate VOC composition, such as using two different gas chromatography systems.
Several recent field studies have investigated the relative importance
of gasoline, diesel and biogenic emissions in generating SOA
. These
have been carried out predominately in the US, particularly in
California, where current diesel usage is rather low by global
standards. The US diesel fleet is dominated by heavy-duty vehicles,
leading to a difference in the source strength of diesel and gasoline
engines between weekdays and the weekend. In contrast, this trend is
not observed in London (see Supplement and Fig. S1), which has
a different vehicle fleet composition, with 60 % diesel fuel
use, and a large number of diesel buses operating at similar times to
domestic vehicles. It is also important to consider that London and
Los Angeles (arguably the world's most well-studied city for air
pollution) have significant differences in terms of urban geography,
population density, commuting patterns, amount of green space/trees
and upwind sources (expanded in the Supplement and Table S1). Specifically, the
high population density in central London, and the fact that many
vehicle journeys both begin and end in central London (rather than
radial vehicle commuting); resulting in the central London atmosphere
experiencing both cold and warm start vehicle emissions. Also worth
noting is that there is no large upwind source of BVOCs in London,
rather the natural emissions are distributed rather homogeneously
across the city (Fig. S2).
This work uses high-resolution VOC measurements to investigate the
abundance and trends of diesel-related hydrocarbons in the atmosphere
at a typical urban background site in London. By comparison to the
emission inventories, we highlight a severe underestimation in the
impact of gaseous VOC emissions from diesel on urban air quality that
is likely replicated across Europe and other cities globally where
diesel vehicle use is high.
Experimental
Clean air for London campaign
Two 5-week campaigns were conducted as part of the National
Environment Research Council (NERC) funded Clean Air for London
(ClearfLo) project in January/February and July/August 2012 at an
urban background site (see Fig. S3) based at Sion Manning School in
North Kensington, London. For more information about this site and the
ClearfLo project refer to and
respectively.
Gas chromatography measurements
Two gas chromatography (GC) instruments were used during the ClearfLo
campaign, a dual channel GC-flame ionisation detector (DC-GC-FID) and
a comprehensive two dimensional GC (GC × GC-FID). Outside air was
sampled from a manifold at 4 m from the ground through
a condensation finger in an ethylene glycol bath held at
-30 ∘C, to remove any moisture from the sample. See
Supplement for details about calibrations.
The DC-GC-FID was operated by the National Centre for Atmospheric
Science (NCAS) Facility for Ground Atmospheric Measurements (FGAM)
with the instrument set up and calibration described in
. In brief, the system has three GC columns
that operate in parallel, where after sampling and desorption the flow
is split 50:50; one column is an aluminium oxide (Al2O3)
Porous Layer Open Tubular (PLOT, 50 m, 0.53 mm id) for
non-methane hydrocarbons (NMHCs) analysis; and the other column is
actually two LOWOX columns in series (10 m, 0.53 mm id) for OVOC analysis.
The GC × GC-FID is comprised of a Markes TT24-7 thermal
desorption (TD) unit with an air server attachment (Markes
International, Llantrisant, UK) and an Agilent 7890 GC (Agilent
Technologies, Wilmington, DE, USA) equipped with an FID operating at
200 Hz. The TD unit sampled at a rate of
100 mLmin-1 for 55 min, giving a total sample
volume of 5.5 L. The trap temperature was set to
-10 ∘C, held for 3 min, then on injection
heated at 100 ∘Cmin-1 to 200 ∘C
to ensure all analytes of interest were desorbed.
The GC × GC-FID system first dimension column was a BPX-5
(25m×0.15mm, 0.4 µm df), at 50 psi,
combined with a second dimension column of a BP-20 (5m×0.25mm, 0.25 µm df), at 23 psi (SGE, Australia),
with column pressures controlled using the Agilent 7890 EPC. The total
transfer flow valve modulator
incorporated a 6-port, 2-way diaphragm valve (Valco Instruments, Houston, TX, USA), with actuation achieved
using a solenoid valve, controlled by software written “in house”. The
modulator was held at 120 ∘C throughout the run and
had a modulation period of 5 s, with 4.7 s sample and
0.3 s injection times. The chromatographic and modulation
configuration of the GC × GC-FID system is detailed in
. During the injection of sample, liquid
carbon dioxide (CO2) was sprayed onto the first 2 cm
portion at the head of the first dimension column for 60 s to
re-focus the sample.
An oven temperature programme was developed which optimised separation
and resolution of compounds of interest. The initial oven temperature
was 30 ∘C, held for 1 min, ramped at
2.5 ∘Cmin-1 to 130 ∘C, held for
1 min then ramped at 10 ∘Cmin-1 to
200 ∘C, and held for 1 min; giving a total run
time of 50 min. This, combined with the TD run time, gave
a total analysis time of 55 min.
During the summer campaign, some parameters had to be changed. Ambient
temperatures were higher in comparison to the winter, meaning the oven
temperature programme had to be altered to allow the oven to reach its
minimum temperature. The initial oven temperature was changed to
35 ∘C held for 2 min and the final temperature
of 200 ∘C was held for 2 min. All other oven
parameters were kept the same. Due to a sensitivity drop, the TD
sampling rate was increased to 200 mLmin-1 for
55 min, to give a total sample volume of 11 L.
Supporting measurements
Measurements of NOx were made using a single channel,
chemiluminescence instrument (Air Quality Design Inc., USA), which has
a wide linear range (1 ppt to
500 ppb). Ozone measurements were made
using a UV Absorption TEI 49C and 49i (Thermo Scientific) with
a limit of detection of 1 ppb.
Observations of hydrocarbons in urban air
The two GC instruments individually quantified 78 VOCs (36 aliphatics,
19 monoaromatics, 21 oxygenated and 2 halogenated), as well as many
hundreds more included in a lumped carbon number assessment from 2667
samples (1352 winter and 1315 summer). The DC-GC instrument measured
volatile VOCs, C1–C7 hydrocarbons and a selection of
OVOCs, with effective saturation concentrations
ranging from 3×107 to 1.4×1012 µgm-3. The
GC × GC-FID instrument measured the less volatile VOC
fraction (effective saturation concentration range of 1.8×106 to 2.4×109 µgm-3), with
hydrocarbons from C6 to C13, plus a large group of
OVOCs (from C3 onwards). There was some overlap in species
measured by both instruments, with good agreement seen (e.g. benzene
R2 0.92, slope 1.070±0.013, see Fig. S4).
Typical GC × GC-FID chromatogram from 25 July 2012,
demonstrating the grouping of compounds. The retentions on the first and
second columns are the x and y axis respectively, with the intensity of
the compound shown by the coloured contours. Labelled peaks and groups are
identified as follows, with the dashed and solid lines indicating compounds
that were identified individually and as a group respectively; (1–8)
aliphatic groups from C6 to C13, (9) benzene, (10) toluene,
(11) C2 substituted monoaromatics, (12) C3 substituted
monoaromatics, (13) C4 substituted monoaromatics, (14) naphthalene,
and (15) C10 monoterpenes with * corresponding to α-pinene
which is the start of that group. The remaining compounds, not enclosed in a
box contain hetero-atoms, primarily oxygenates. The grouping of compounds was
accomplished using the lasso technique in Zoex GC image software (Zoex, USA).
This technique allows the software to calculate the area of all peaks
included in the lasso.
![]()
The number of possible structural isomers increases exponentially with
carbon number and beyond around C9
it becomes impossible to accurately identify the structure of every
hydrocarbon present in air. Using the retention behaviour of each
compound on chromatographic columns it is however possible to assign
individual species to particular chemical classes and
functionalities. Here we group according to carbon number and basic
functionality, an example of which can be seen in Fig. . The
C2-C6 volatility range contains both primary hydrocarbon
emissions, with isomers quantified individually, along with several
oxygenated compounds. Between C6 to C13, a wide
range of hydrocarbon and OVOC species, including the
n-alkanes, α-pinene, limonene, monoaromatics with up
to 3 substituents and naphthalene were quantified individually (see
Fig. S5 and Table S2).
Grouping of unresolved complex mixtures
In previous studies using GC-FID, the larger hydrocarbon fraction, where
diesel VOC emissions are predominately found, is part of an unresolved
complex mixture (UCM). One method used to estimate the relative amounts of
VOCs in this region, is to identify the n-alkane (which is often observed
as a well-defined peak above a raised baseline) and then integrate the area
above the blank baseline between two consecutive linear alkanes (using an
FID) or to use the m/z 57 fragment ion to represent primary intermediate
VOC (IVOC) . In reality, this gives an estimate of the
total or alkyl containing IVOC loading within this volatility range and will
not only include the hydrocarbon fraction with that specific carbon number
but other compounds as well (i.e. lower carbon number aromatics, OVOCs). This
study details the improved resolution of VOCs using GC×GC to allow for
a more stringent grouping of the UCM by carbon number and functionality,
rather than by volatility.
Higher carbon number aliphatic compounds (C6-C13,
predominantly alkanes with some alkenes and cycloalkanes), C4
substituted monoaromatics and C10 monoterpenes have been grouped
together and the combined class abundance estimated using a response ratio to
the corresponding straight-chained n-alkane, 1,3-diethyl benzene and
α-pinene respectively. The group boundaries are shown in
Fig. , where for example, box 7 corresponds to the
C12 aliphatic group and encompasses alkanes, cyclic alkanes and
alkenes. Only the material within the box is integrated within this retention
window. This is a clear improvement over the 1-D case, as there are a
considerable number of peaks, with a higher 2nd dimension retention time, in
Fig. that would co-elute with the aliphatic group if the
entire retention window was co-sampled (i.e. aromatics, oxygenates and other
hetero atom containing species). The number of individual isomer peaks that
could be isolated in the aliphatics grouping increased from 9 for the
C6 group to 40 for C10 (shown in
Fig. , black squares). A full table of the observed
isomer peaks and group mass concentrations is provided in Table S3.
Unfortunately, the separation of the linear alkanes, branched alkanes, cyclic
aliphatic and alkenes on the GC × GC chromatogram is not sufficient
at higher carbon numbers to allow them to be more fully resolved. This is a
direct consequence of the use of the cryogen free and field deployable valve
modulator, which when used in total transfer mode, where the flow in the
first column slows during the modulation pulse, imposes restrictions on the
column dimensions and internal diameters that can be used
. Also, given the temperature constraints on this
instrument, it is likely that the GC × GC not only misses a fraction
of the C13 aliphatic group but may also be under-reporting the
number of isomers in the higher carbon number groups. This would explain why
the number of isomers decreases after C11 aliphatics, rather than
increases as would be expected. The aliphatic groups have diurnal behaviour
(discussed in the next section) that indicate a dominant traffic-related
source. Fuel composition measurements suggest there is unlikely to be
significant quantities of alkenes from traffic related sources; gasoline
contains around 3–4 wt C% of alkenes, and diesel contains negligible
quantities .
Diurnal behaviour
Diurnal profiles of selected urban pollutants in winter (left-hand
side of each plot) and summer (right-hand side of each plot). Winter and
summer are plotted on the same y axes to show seasonal differences, with
insets allowing the profile to be easily seen. (a and b)
Ozone (n=2915 and 2880; winter and summer, respectively), (c
and d) nitrogen oxides (n=2915 and 2880), (e and
f) ethane (n=660 and 681), (g and h) toluene
(n=660 and 680), (i and j) isoprene (n=660 and
681), (k and l) α-pinene (n=691 and 634),
(m and n) C13 aliphatics (n=692 and
632), (o and p) C4 substituted monoaromatics (n=692 and 563). Figure was constructed using the OpenAir
project for R where the solid line represents the mean daily concentration
and the shaded regions shows the 95 % confidence intervals
surrounding the mean .
![]()
The average diurnal behaviour of a selection of VOCs, O3 and
NOx, are shown in Fig. . VOCs with an
anthropogenic source (e.g. ethane, toluene, C4 substituted
monoaromatics and C13 aliphatics) have higher mixing ratios
in winter, consistent with reduced rates of photochemical removal in
the northern European winter and a lower boundary layer
height. Toluene and NOx are strong indicators of traffic
related emissions, and both have diurnal profiles with rush hour
peaks, commonly observed in urban areas
. The profiles
of the higher carbon number species, C4 substituted
monoaromatics and C13 aliphatics, show similar
traffic-related profiles, strongly indicative that this is their major
source. Those species with a dominant biogenic source (e.g. isoprene
and α-pinene) are higher in summer, due to increased
emission. The winter profiles of isoprene and α-pinene show
possible anthropogenic sources, traffic and cleaning products
respectively.
Reactivity and mass calculations of grouped compounds
Seasonal median values for hydrocarbon mixing ratio, mass
concentration and primary hydrocarbon OH reactivity in London air grouped by
carbon number and potential emission source.
![]()
Figure shows the relative emission source
contributions of compounds to the total hydrocarbon mixing ratio
(top), mass concentration (middle) and primary hydrocarbon hydroxyl
radical (OH) reactivity (bottom), calculated by carbon number, and
split according to emission source. For each sample, the mass
concentration (µgm-3) and primary hydrocarbon OH
reactivity (s-1, see Supplement Sect. S1.7 for explicit details on the calculation of OH reactivity) were calculated and a seasonal median
calculated using the mixing ratios of the individual components and
the summation of all further unidentified species within the ten class
groups, not including the OVOCs. OH reactivity is defined as the total pseudo
first order rate coefficient for loss of OH when reacting with VOCs in the
atmosphere. This is important in urban atmospheres that are VOC limited, such
as London, as the reaction of VOCs with the OH radical is the driving force
for the formation of O3 and other secondary pollutants.
The Passant (2002) speciated
emissions inventory was used to determine the
main emission sources for each compound. For C2 to C5,
the main source is either natural gas usage or leakage, followed by
road transport use. From C6 onwards, the main emission source
was classified as road transport or other fuel usage categories
(i.e. filling of petrol stations). To determine the
percentage contributions from diesel or gasoline fuel usage, the
detailed fuel characterisation of was used
with a value of 60 % diesel use in the UK
.
Although the summer data have been shown throughout the remainder of the
article, care should be taken when interpreting the impacts. It is likely
that the summer observations are made up of “residual” VOCs remaining after
transport from emission source and subsequent photochemical reactions. This
could lead to a small overestimation of the contribution of diesel-related
hydrocarbons and an underestimation of some species, particularly the OVOCs.
The winter has generally higher abundances of hydrocarbons, with
summer showing a marked increase in the biogenic source compounds. In
both winter and summer, in mixing ratio terms, the distribution is
dominated by high-volatility species (over 85 % of the total)
primarily from natural gas and gasoline sources. When viewed in terms
of mass concentration, however, the distribution of combined natural
gas and gasoline vs. diesel is closer to 70:30 %
in winter and 77:23 % in summer, the latter due
to the increased loss rates for reactive species in summer
disproportionately removing larger hydrocarbons.
In order to calculate primary hydrocarbon OH reactivity of diesel
emissions, rate constants have to be estimated as each individual
species is not uniquely identified and because the rate constants (and
subsequent chemistry) are un-measured in the majority of cases for
hydrocarbons larger than C8. The nearest straight-chain alkane
rate constant is applied to all carbon in
that aliphatic grouping. This will lead to a conservative estimate of
reactivity since cycloalkanes and alkenes would be expected to react
faster (i.e. the kOH for the reaction of
n-dodecane is 1.32×10-11 cm3molecule-1s-1, compared to that of
1-dodecene which is 5.03×10-11 cm3molecule-1s-1
, for further details see Supplement Sect. S1.7 and
Table S4). For the C4 substituted monoaromatic and
C10 monoterpene groups, the rate constants of 1,3 diethyl
benzene and α-pinene respectively were used.
Calculation of unmeasured diesel emissions
The GC measurements stop at C13, however diesel typically
has a range of hydrocarbons from C9 to C22 that
peaks with n-hexadecane as the most abundant
compound. Using the observed distribution of C10-C13,
allows for an estimate to be made of the remaining,
unobserved NMHC fraction of gaseous diesel emissions in the
C14-C22 range, using the fuel composition-based emission
factors for the gas phase compounds from .
Assuming no atmospheric loss, a reasonable approximation in winter, it
is estimated that the GC × GC-FID technique observes around
25–30 % of the total gaseous hydrocarbon emissions from
diesel sources. It is then possible to estimate a seasonal average
unmeasured gas phase NMHC mass concentration from diesel sources in
London as 76.1–97.8 µgm-3 in winter and
26.8–34.3 µgm-3 in summer. These values can be
compared to typical primary organic aerosol measurements of
1 µgm-3 ,
indicating that, in ambient air, diesel-related emission of
hydrocarbons are overwhelmingly (a factor of 100) to the gas phase,
consistent with laboratory and tail pipe studies
. In contrast, using gasoline liquid fuel
speciation, the combined GC approach can observe approximately over
98 % of the mass of gasoline.
Contributions of emission source to total mixing ratio, mass and OH
reactivity for winter and summer. Diesel is the summation of measured and
calculated, with error bars indicating the uncertainty of the unobserved
diesel NMHC fraction.
![]()
The impact of the unobserved diesel emissions on primary hydrocarbon
OH reactivity was estimated using the n-dodecane rate
constant as a proxy. The percentage contributions to mixing ratio
(top), mass concentration (middle) and primary hydrocarbon OH
reactivity (bottom), divided by emission source, including the
unmeasured diesel emissions and OVOCs, are shown in
Fig. . It is clear that diesel plays an
important role in the composition of NMHCs and their subsequent
reactivity. The total (measured + calculated) diesel emissions
contribute 5.1 and 1.7 s-1 to OH reactivity in winter and
summer respectively compared to 1.7 and 0.8 s-1 from
gasoline compounds; increasing the contribution of diesel-related
hydrocarbons to calculated VOC OH reactivity from 23 % using
the measured VOCs to 61 % including the unmeasured I/VOCs in
winter and from 8 to 34 % in summer (full details can be
found in Table S5). In summer, the primary emissions have undergone
a degree of loss due to photochemical ageing and so the values are an
underestimate of the fuel sources. These diesel-related hydrocarbons
may be partly responsible for the “missing” OH reactivity observed
between measurements of OH lifetime vs. the value calculated from
observed sinks in many studies
.
Comparison to emissions inventories
Assuming that the winter observations are made “at source” (hence
atmospheric losses and lifetime differences can be neglected), the
measurement location is representative of an urban setting, and the UK
inventory correctly estimates the emission for toluene (based on
direct flux comparisons made in London by ),
there appears to be a significant inventory under-reporting for the
higher carbon number species. When normalising to toluene, the UK
national emissions inventory, which is believed to use best-practice
international reporting methodologies, under-reports by a factor of 4.6
for C9 species, rising to a factor of 74 for C12
compounds (see Fig. ). Given the clear
traffic diurnal profiles of the C10-C13 species (seen in
Fig. m for C13 aliphatics), which
essentially encompasses only diesel fuel, the most likely source of
these species is gaseous emissions from the diesel vehicle fleet,
either evaporative, tailpipe or a combination of the two. These
observations provide the first direct evidence of significant diesel
hydrocarbons in London's ambient air, something that could previously
only be inferred from liquid fuel measurements and exhaust studies.
Winter emissions inventory underestimation (left axis and blue
columns) and the number of isomers included in each grouped set of compounds
(right axis and black squares). Grey line shows a factor of 1, i.e., inventory
emission estimation is consistent with the observations.
![]()
Ozone formation potentials
At present, there is insufficient kinetic and mechanistic data to
allow for the accurate modelling of the impact diesel hydrocarbons
will have on photochemical ozone. Unlike Los Angeles, which is
typically impacted by intense single-day episodic photochemical ozone
events, in London (and NW Europe), higher ozone levels are usually the
result of regional-scale multi-day formation. Therefore, different
control strategies and reactivity scales, which take into account
trans-boundary transport, have been developed and applied in
Europe. Photochemical ozone creation potentials (POCPs,
e.g. ) have been derived using idealised
5-day photochemical trajectory model runs over Europe,
incorporating detailed chemical degradation schemes of the emitted
VOCs . Individual POCP values depend upon
emissions along the trajectory, the reactivity of the VOC and its
propensity to form ozone, i.e. the number of C-C and C-H bonds in the
reactive species. Calvert et al., (2008), using the Master
Chemical Mechanism (MCMv3.1, http://mcm.leeds.ac.uk/MCM) and
speciated emission inventories (), with
little speciation of alkanes above C9, found that alkanes
dominated POCP-weighted emissions (33 %) on the regional
scale in Europe, accounting for slightly more than the aromatics
(29 %) and significantly more than the alkenes
(20 %) and oxygenates (17 %)
. They also note that this is in marked
contrast with that found on the urban scale in Los Angeles, where
alkanes contribute little to the intense episodic ozone formation
observed. The work presented here shows that emission inventories
severely underestimate the amount of alkanes emitted from diesel
sources, hence these have not been included in previous studies on
ozone formation. Based on the POCP results of
for shorter chain alkanes, and
considering the high OH reactivity of larger hydrocarbons,
incorporating the diesel-related aliphatics into future calculations
is likely to have a significant impact on regional ozone formation in
Europe, and likely elsewhere.
However, despite the lack of chemical information available, we can
make an assessment of the effects that the new diesel VOC observations
have on local ozone productivity by calculating the O3
Formation Potential (OFP) of each emission source using a Maximum
Incremental Reactivity (MIR) scale, as determined by
using the SAPRC-07 mechanism. The MIR scale is
based upon 1-day photochemical simulations in a box moving over an
urban basin and subject to ozone precursor emissions. The NOx
concentrations are adjusted so that the final ozone concentration in
a simulation showed the maximum sensitivity to changes in emissions of
organic compounds; these conditions give the MIR. MIRs represent
relatively high NOx conditions, often experienced in US
cities, where control of the emissions of VOCs is the most effective
means of reducing ozone formation .
MIR values for a range of important VOC emission classes (alkanes,
alkenes, aromatics, oxygenates) compare well to POCP values calculated
using detailed MCMv3.1 chemistry in a 1 day US urban photochemical
trajectory model, giving us confidence in the tuned SAPRC-07 chemistry
for predicting photochemical ozone formation under typical 1-day
high NOx, high ozone conditions
. However, caution must be observed when
applying MIR scales to other conditions (i.e. NW Europe), as
previously discussed.
For those species included in one of the carbon number and
functionality VOC groups, a weighted MIR value has been calculated
assuming a composition of 95 % branched alkane and
5 % alkene. The calculated, unobserved diesel emissions were
given an MIR value based on the weighted contributions of the
different compound classes from the diesel fuel gaseous emission
speciation in .
Contribution of emission sources to ozone formation potential, where
diesel is representative of total diesel emissions and the error bars show
the uncertainty of the unobserved diesel fraction calculation.
![]()
The calculated OFP values for summer and winter are shown in
Fig. , with a clear seasonal difference in the
importance of different emission sources. The winter is dominated by
emissions from diesel (over 45±4.6 %). In contrast, the summer
shows a marked increase in contribution from oxygenated species,
rising from 18 % in winter to 40 % in
summer. concluded that gasoline emissions
contributed the majority of potential O3 formation, however
given that gasoline is the primary fuel used in the US (quoted at
73–90 % of total fuel use; ) this
is not surprising. Whereas in the UK, diesel fuel accounts for an
average of 60 % of total fuel use and as such would be
expected to contribute a larger percent towards O3 formation.
Potential SOA mass estimates. Upper: mean VOC mass concentration
[VOC] shown by black columns, and the corresponding SOA yields (Y) for the
VOC precursors in blue circles. Lower: potential SOA mass concentration
[SOA], calculated as the product of mean VOC mass and SOA yields. Winter
shown on left and summer shown on right hand side panels.
![]()
Given that many of the higher carbon number species, likely emitted
from the combustion of diesel, are not currently included in emissions
inventory and as such are missing from many model chemical mechanisms
(e.g. MCMv3.3.1), a more rigorous reactivity analysis is not possible
(e.g. POCPs discussed previously). It is possible to infer from this
analysis however, that as winter diesel emissions contribute nearly
50 % of the OFP and over 60 % of primary hydrocarbon
OH reactivity, it is likely that they would have a large impact on the
overall reactivity and chemistry of the urban atmosphere.
Potential impacts on SOA formation
Over the last few years, there has been robust debate in the
literature over the relative importance of diesel vs. gasoline for SOA
production, arising in part because of the difficulties of measuring
diesel hydrocarbon emissions
. Recent
studies have shown the unspeciated emissions from combustion sources
lead to significantly more SOA production than those that can be
speciated by conventional instrumentation. ,
used simulation chamber data and source-specific SOA yield
parameterisations for these unspeciated emissions to estimate that, in
the US, 90 % of SOA was from biomass burning and gasoline
sources, with 85 % of the SOA coming from unspeciated organic
emissions.
There is a clear need to improve measurements of larger hydrocarbons
that represent a large part of what is referred to in other studies as
unspeciated chemicals. In this study, the uncertainty in the
unspeciated fraction has been reduced by grouping ambient observations
of VOCs by carbon number and functionality. The
>C9 aliphatic groups are dominated by diesel
emissions at this location and so the SOA source strength can be more
accurately determined. The potential contributions of the higher
hydrocarbons to SOA formation has been estimated by multiplying the
median measured VOC mass concentration (Fig. , top panel,
black columns) by the corresponding SOA yield (Fig. , top
panel, blue circles) . The yields applied were
measured in high NOx chamber studies
(VOC0/ NOx (ppbC ppb-1) of 0.5) with an
organic aerosol mass COA of 10 µgm-3, representative of urban areas
. Here it is assumed that NMHCs
with less than six carbons and aqueous chemistry of water soluble
oxidation products, such as glyoxal, do not contribute to SOA mass . In
both winter and summer, the observed levels of aliphatic compounds
from diesel sources have the potential to form significant quantities
of SOA (Fig. , bottom panel, red columns). If a diesel
SOA yield of 0.15 is applied to the total diesel emissions (as
calculated previously), then gas phase emissions from diesel engines
represent the dominant traffic related precursor source of urban SOA
in a European megacity such as London, where the use of diesel fuel is
prevalent, in line with previous studies in the US
. Each cubic metre of air contains sufficient
gas phase hydrocarbons to potentially produce 14.4–17.6 and
4.9–6.1 µg of SOA in winter and summer following
atmospheric oxidation.
Recent simulation chamber studies indicate that modern engines fitted
with diesel particle filters, such as EURO5 emissions control, have
greatly reduced VOC tailpipe emission and form little SOA under
chamber conditions
. However,
have recently shown that when vehicles are
driven under real-world urban conditions (i.e.
different engines loads cause variable catalyst temperatures which can lead
to limited effectiveness, as opposed to dynamometer tests where the catalyst
is held at optimum operating conditions), the emissions of NOx
from diesel engines have not been reduced as expected given the new
technologies implemented. It is possible to infer that if NOx
emissions are higher than expected, the VOC emissions are also likely
to be higher. In this study we have shown that there is a significant
diesel vehicle source emitting sufficient VOCs to impact ozone and SOA
formation in the urban atmosphere.
Conclusions
From the results presented, it is possible to conclude that current
inventories and emissions estimates do not adequately represent
emissions of gas phase higher carbon number species from the diesel
fleet under real-world conditions and in a developed urban
environment. The calculated impact of these species is significant,
particularly in terms of OH reactivity, ozone formation potential and
SOA production.
In the last decade, there has been a steady shift in fuel use in many
locations. For example, in the UK diesel fuel use as a fraction of
total fuel has risen from 52 % in 2005 to 62 % in
2011 (see Fig. S7 and Table S7) . Although
the UK may be considered typical of Europe (where diesel use varies
between 45–80 %) , the average US value
was around 29 % diesel use in 2013, with the
understanding of other geographical regions currently being poor.
This shift to an increasingly diesel-powered fleet in many developed
cities, as a response to energy efficiency drivers, has therefore
shifted the balance of hydrocarbons in urban air from short to long
chain compounds, and these observations provide direct atmospheric
evidence of this effect in London. Previous air quality assessments of
diesel-related hydrocarbons in the atmosphere are few in number, and as
discussed previously, have been made only in the US where geographic
characteristics and vehicle fleet composition are very different to London,
and Europe more widely. In many cities the impact of diesel hydrocarbons
remains to be determined, but this work demonstrates that it will likely be
significant in locations with substantial diesel fleets. An improvement in
measurement infrastructure appears to be essential if this source is to be
quantified more widely or the impacts of policy evaluated.
Understanding the impact of this change is
significantly hindered however by a lack of appropriate
physico-chemical data for individual longer chain hydrocarbons. There
are already very significant policy challenges for many developed
cities relating to the control of NO2 from modern diesel
vehicles, and this study indicates that there may also be a similar,
but currently un-recognised, policy challenge to control reactive
carbon emissions and their contributions to secondary
pollutants.