Interactive comment on “ The impact of monoaromatic hydrocarbons on OH reactivity in the North Sea boundary layer and free troposphere ”

Introduction Conclusions References

tant sub-class. A method based on comprehensive two-dimensional gas chromatography coupled to time of flight mass spectrometry (GC × GC-TOFMS) has been developed that extends the degree with which larger VOCs can be individually speciated from whole air samples (WAS). The technique showed excellent sensitivity, resolution and good agreement with an established GC-FID method, for compounds amenable to 10 analysis on both instruments. Measurements have been made of VOCs within the UK east coast marine boundary layer and free troposphere, using samples collected from five aircraft flights in winter 2011. Ten monoaromatic compounds with an array of different alkyl ring substituents have been quantified, in addition to the simple aromatics, benzene, toluene, ethyl benzene and Σmand p-xylene. These additional compounds 15 were then included into constrained box model simulations of atmospheric chemistry occurring at two UK rural and suburban field sites in order to assess the potential impact of these larger monoaromatics species on OH reactivity; they have been calculated to contribute an additional 2-6 % to the overall modelled OH loss rate, providing a maximum additional OH sink of ∼ 0.9 s −1 .

Introduction
It is well known that the OH radical controls the daylight oxidising capacity of the atmosphere. In the presence of NO x and VOCs, reactions involving OH can contribute to the formation of a range of important secondary pollutants including tropospheric ozone, NO 2 and secondary organic aerosol (SOA). The [OH] is controlled partly by species a way to improve the identification and quantification of VOCs in the atmosphere. Comprehensive two-dimensional gas chromatography (GC × GC) involves the coupling of two columns with different separation mechanisms via a mid-point modulator. The technique was first performed by Philips and Liu in 1991 (Phillips and Liu, 1991) on petrochemical mixtures and has since enabled the separation of high complexity samples in 10 a range of fields, giving separations that were previously unattainable by conventional capillary GC and GC-MS methods. The first GC × GC analysis of VOCs in air was performed by Lewis et al. (2000) with a flame ionisation detector. This identified that much of the hydrocarbon loading in an urban atmosphere was unaccounted for using conventional one dimensional GC- 15 FID techniques. Hamilton and Lewis (2003) reported the monoaromatic composition of gasoline, gasoline vapours and urban air samples using GC × GC-FID with a simple cooled loop injection. They found many of the larger aromatic species present in gasoline vapour were also present in urban air. A total of 147 monoaromatic isomers were isolated from a polluted urban air sample and were calculated to be a potentially signif- 20 icant source of tropospheric ozone. GC × GC has subsequently been used in a range of studies of atmospheric composition at locations with different sources and degrees of ageing (Xu et al., 2003a, b;Lee et al., 2006;Bartenbach et al., 2007;Saxton et al., 2007).
The complex nature of the chromatograms and the huge quantity of data produced Introduction trace gases including carbon monoxide, ozone, oxides of nitrogen, non-methane hydrocarbons (NMHC), oxygenated volatile organic compounds (OVOC), N 2 O 5 and radicals intermediates (NO 3 , OH, and HO 2 ). In addition, aircraft positional data was recorded, along with temperature, pressure, humidity and turbulence.
The FAAM ARA has the facility to accommodate a total of 64 WAS canisters used for 5 collection of air samples during flight. The canisters are 3 L in volume and the interior has been coated with a layer of deactivated silica (silcosteel). The canisters, once fitted into the hold of the aircraft, are filled at various points during flights using an all stainless steel assembly and double headed metal bellows pump. Canisters are typically filled to their maximum fill pressure of 3 atm giving a sample volume of approximately 9 L. The 10 fill pressure is reduced at higher altitudes, but typically always remains above 2 atm. The individual WAS canisters are housed in metal flight cases of between 8 and 15 bottles. Within these cases, bottles are connected together via a common inlet and outlet port. Individual bottles are filled via electronic activation of pneumatic solenoid valves that are controlled by a PC within the aircraft cabin. WAS bottles can be filled 15 at a desired point in the flight or set to automatically fill at regular time intervals using the WAS filling software. All fittings, connectors and gauges used on the WAS case and aircraft assembly are 316 stainless steel to ensure minimal losses and eliminate contamination. 20 A schematic of the instrument setup is shown in Fig. 1 200 • C and all other flow paths were heated to 150 • C. A post-sampling line purge was performed for 3 min using carrier gas to ensure the lines were conditioned and cleaned before analysis of the next sample. To prevent excessive trapping of water, a 200 mL cold finger submerged in an ethylene glycol water chiller set at −30 • C was used during all of the atmospheric gas analysis. To ensure each sample was effectively flushed 5 though the cold finger before sampling onto the sorbent trap, a pre-sampling line purge was performed for 5 min at 100 mL min −1 . Activation of the correct WAS bottle was achieved using an in house control box, programmed via LabVIEW, which activated the electronic solenoid valve of the desired WAS bottle. A contact closure was sent using a relay connected to the sampling port 10 valve on the CIA8. During the Unity's prepurge step the airserver sample valve was opened, triggering the opening of the desired canister. The length of time the WAS canister remained open was also controlled by software. The software was set to leave the WAS bottle open slightly longer than the Unity, to ensure the software did not trigger the opening of another WAS bottle prematurely. Connections between the CIA8, the 15 cold finger and the WAS bottles were made using 1/16 instrument grade stainless steel (Thames Restek UK Ltd) which were heated and insulated to approximately 70 • C. Narrow bore tubing was used to reduce dead volume, ensuring the line was thoroughly flushed during the pre-purge step of the next sample. 20 The GC × GC-TOFMS consisted of an Agilent 6890 (Agilent Technologies, Palo Alto, CA USA) gas chromatograph and a Pegasus III TOF-MS (Leco, St. Joseph, MI, USA). Integration and data processing was performed using ChromaTOF 3.32 software (Leco, St. Joseph, MI, USA) optimised for Pegasus data files. The primary column was a nonpolar BPX-5 (50 m × 0.32 mm × 1.0 µm d.f film thickness) (SGE Analytical Science) and Introduction lium carrier gas was used in constant pressure mode set at 40 psi (C.P grade, BOC) and was set using the Unity 2. The GC oven was configured with the following parameters. An initial temperature of 40 • C held for 4 min, the oven was then ramped at 3 • C min −1 to 160 • C and then further ramped at 50 • C min −1 to 200 • C and held for 1.2 min. The secondary oven used the same parameters, however was offset by 5 +25 • C. The modulator temperature offset was set at + 45 • C and a modulation time of 2 s was used. The TOFMS was set to an acquisition rate of 100 Hz using a detector voltage of 1950 V, a filament bias of −70 eV and acquired for a mass range of m/z 35 to 300. The ion source and transfer line were both set to 250 • C and a 2 min solvent delay was used.

Calibration
Calibrations and quantification were performed using a series of parts-per-billion (ppbv) gas standards; NPL30 ozone precursor mix and terpene standard (National Physical Laboratory, Teddington, UK), and AR54 and AR74 Hydrocarbon standards (Apel-Reimer, Boulder, CO, USA). Various volumes and replicates were sampled for method 15 validation. Calibration of the TD-GC × GC-TOFMS system during the RONOCO flying campaign was performed using a combination of the AR54 and the AR74 gas standards. The standards were run as frequently as possible to ensure accurate quantification and to minimise uncertainty due to MS detector drift. The AR54 hydrocarbon standard was analysed every time a WAS case was connected to the instrument (ev-20 ery 8 to 16 samples). The AR74 hydrocarbon standard was run less frequently (once every 24 h) due to limited standard availability.

GC-FID
The samples were also analysed using a single column dual channel FID system (DC-GC-FID), which is well established and operated by the NCAS Facility for Ground Atmospheric Measurements (FGAM). The system uses two GC columns running in parallel. After sampling and desorption, the flow is split 50 : 50 and passes down a aluminium oxide (Al 2 O 3 ) Porous Layer Open Tubular (PLOT) column (50 m, 0.53 mm id) for analysis of NMHCs and two LOWOX columns (10 m, 0.53 mm id) in series for analysis of OVOCs. The instrument setup and operation is described in detail in Hopkins et al. (2006Hopkins et al. ( , 2011. Calibrations were performed using a 30 component ppbv level ozone 5 precursor standard, NPL30 (National Physics Laboratory).

Results and discussion
The GC × GC-TOFMS system was calibrated using a series of gas standards and an example chromatogram is shown in Fig. 2, where the retention times in columns 1 and 2 are shown on the x-and y-axes respectively and the peak intensity shown as a coloured 10 contour. The instrument shows excellent separation of the hydrocarbons with >C 5 , with the exception of m/p-xylene as seen in most GC separations using siloxane columns. No signals were seen for hydrocarbons with less than 5 carbons as they were too volatile to be retained on the liquid nitrogen modulator. An advantage of GC × GC is that compounds elute in characteristic patterns depending on their functionality. This 15 pattern can be observed in Fig. 2 with the aromatic species well separated from the aliphatics in the second dimension. Precision, limits of detection (LOD) and linearity of calibration for the species calibrated using gas standards are shown in Table 1. The relative standard deviation was calculated for 5 replicate injections of a 1 L sample of the AR74 hydrocarbon standard at ∼ 100 pptv and the R 2 value given is for a 6 point 20 calibration ranging from approximately 10 pptv to 250 pptv (depending on the standard mixing ratio of each compound). The limit of detection was calculated at S/N = 3. The retention time and quantification ion used for the calibration are also shown. The GC × GC-TOFMS was able to consistently resolve the majority of species in the gas standard with relative standard deviation (%RSD) values less than 5 %, with the exception of 1-pentene, cis-2-pentene and 2-methyl-2-butene that gave a higher %RSD. These compounds co-elute with each other on the column set used and have similar Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | mass spectra leading to poor mass spectral deconvolution. Limits of detection for a 1 L sample were found to be generally sub-pptv and calibrations showed good linearity (R 2 > 0.99). These results show that the GC × GC-TOFMS is both highly sensitive and a high resolution analysis technique. The low detection limits observed are a result of peak amplitude enhancement from the thermal modulator, the improved background 5 inherent to GC × GC when multiple isomers are present and the TOFMS detector which allows characteristic ions to be extracted from the complex air background, removing background interferences.

Inter-comparison with GC-FID
During the RONOCO campaign, the WAS were analysed by both TD-GC × GC-TOFMS and TD-GC-FID. The dual channel GC-FID method is well established and is tailored for the analysis of NMHCs from C 2 to C 8 , with limits of detection typically between 1-5 pptv. Due to the selectivity required for the analysis of very volatile hydrocarbons, the GC-FID system's performance is reduced as the molecular weight of hydrocarbons increases. This results in broadening of the chromatographic peaks and reduced signal 15 to noise for these species. It is important to consider that as carbon number increases, the potential number of isomers for hydrocarbon species increases exponentially, and so the potential for co-elution of species within this heavier region of a GC run becomes more likely with single column techniques (Goldstein and Galbally, 2007). The mixing ratios obtained by both techniques for 191 WAS bottles collected dur- 20 ing RONOCO were plotted against each other in order to compare the performance achieved for real air samples. In total seven compounds were common between both analyses. Good correlations were observed (R 2 = 0.93-0.99) for all species. The higher molecular weight alkanes, heptane and octane, showed a good correlation between the two techniques, as shown in Fig. 3, however there is a deviation from the 25 1 : 1 line. These large VOCs are at the volatility and separation limit of the GC-FID system. For heptane and octane, it should also be noted that the levels detected are typically lower than many other hydrocarbon species detected. There are many sam-32432 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | ples where the octane and heptane levels fall below the limit of detection for the GC-FID instrument and manifest as a space in the GC-FID bottle series shown in the left hand panels of Fig. 3. However for these samples, the GC × GC-TOFMS was able to quantify the species due to its improved LOD. The increased sensitivity of the GC × GC-TOFMS system is once again noticeable 5 for the comparison with octane, where some of the samples are below the LOD of the GC-FID method but are successfully quantified by GC × GC-TOFMS. Very good agreement was seen for the aromatic species, toluene and benzene, between the two techniques as shown in Fig. 3, with both high R 2 values and slopes very close to 1 : 1.
Good correlation of the aromatic species are particularly useful as they validate that 10 the GC × GC-TOFMS is functioning correctly with no leaks.

Analysis of VOCs using GC × GC-TOFMS during the RONOCO campaign
In total, 191 WAS bottles from five flights were analysed by TD-GC × GC-TOFMS over a 3 week period. Samples were collected between 60-3500 m altitude, generally over the North Sea or English Channel. A typical GC × GC chromatogram can be co-elute with the aromatic band. The TOFMS is particularly useful for the identification of analytes against the NIST Mass Spectral library and also allows simplification of the chromatogram as ions characteristic to particular functional groups can be extracted and viewed in isolation (for example m/z 91 for aromatics).
In total 39 hydrocarbons ranging from C 5 to C 11 could be quantified during RONOCO 25 using the GC × GC-TOFMS. The mean, minimum and maximum mixing ratios are shown in Table 2 for all samples analysed. The highly substituted aromatic species (C 9 and greater) were observed to exhibit a strong correlation with toluene as shown in 32433 Introduction  5. This can be rationalised due to their similar emission sources (i.e. petrol vapours and evaporation). The alkyl benzenes generally have the same emission sources and so the gradient of the regression line is a result of both emission ratios and reactivity with atmospheric oxidants encountered by the airmass from emission to sampling, assuming the same dilution rate. The higher MW aromatic species may be atmospher-5 ically relevant (Hamilton and Lewis, 2003) as many of them exhibit an enhanced reaction rate over toluene with OH and NO 3 radicals (Atkinson and Arey, 2003) and thus they could impact tropospheric ozone formation rates. The strong linear relationships observed by the GC × GC-TOFMS presents an opportunity to predict the heavier mono-aromatic VOC loading by measurement of a toluene 10 mixing ratio alone. This would potentially allow techniques that cannot analyse these species (GC-FID, PTR-MS etc) to make a prediction of the additional aromatic content. For the correlation plots presented in Fig. 5, the intercept in each was set to zero. In order to make a meaningful prediction of the additional aromatic content, a conversion that reflects each aromatic species' concentration and rate of reaction with an atmo-15 spheric oxidant is required. This was performed by converting the mixing ratio of each aromatic species into a Toluene Equivalent mixing ratio, [J Tol ]. This technique has been used previously in the literature (as a propylene equivalent) to allow atmospheric modellers to simplify VOC data and express it as one variable (Chameides et al., 1992). The conversion of a mixing ratio into a [J Tol ] value is performed by multiplying the mix- 20 ing ratio of the aromatic [x], by the ratio of its rate constant, k OH (x), to that of toluene, k OH (Toluene) as shown in Eq. (1).
After conversion of the additional aromatic species into a [J Tol ], the relationship between the toluene mixing ratio and the additional aromatic loading as a [J Tol ] value was 25 investigated. For all samples analysed using the GC × GC-TOFMS, the mixing ratios of the ten additional aromatic species ( value with respect to both reaction with OH, ([J Tol ] OH ) and NO 3 ([J Tol ] NO 3 ) since many of the samples were taken at night and aromatic reactivity with O 3 is negligible. As no rate data is currently available in the literature, rate constants for the reactions of 1,3-diethyl benzene and 1,4-diethyl benzene have been calculated using available Structure Activity Relationships (SARs). The SAR approach of Kwok and Atkinson 5 (1995) has been used to estimate total k OH for each diethyl benzene isomer from calculations of partial OH rate constants for hydrogen abstraction from the ethyl groups and OH addition to the aromatic ring. Rate coefficients for OH-addition to the ring are estimated from a correlation of the sum of electrophilic substituent constants (Brown and Okamoto, 1958) with measured OH addition rate constants (Calvert et al., 2002).
Rate constants for reaction with NO 3 (k NO 3 ) have been inferred from measured rate data for m-and p-xylenes (Calvert et al., 2002) using the method applied by Jenkin et al. (2003). As with all estimation methods, there is a reasonable error of uncertainty and as such an error of a factor of 2 should be applied (Kwok and Atkinson, 1995).
The [J Tol ] values of the 10 individual compounds were then summed for each sample 15 to give a Σ[J Tol ], that represents the additional aromatics measured at each sampling point. The Σ[J Tol ] values were plotted against their respective toluene mixing ratio and are shown in Fig. 6. A reasonable linear relationship is seen with an R 2 of 0.93 and and a gradient of 2.03 and 7.30 for OH and NO 3 respectively. The ten additional monocyclic aromatics may have a considerable effect on reactivity 20 and also potentially ozone formation with a [J Tol ] OH that is on average twice that of the toluene mixing ratio. The average contribution of each aromatic species to the Σ[J Tol ] value can be observed Fig. 7. The tri-substituted benzenes, although at a significantly lower mixing ratio than toluene (average mixing ratio for 1,3,5 trimethylbenzene is 4.7 pptv c.f. 25 toluene which is 136.7 pptv), exhibit a significant rate enhancement for reaction with OH (Fig. 7a)  additional aromatic content still adds a large portion of Σ[J Tol ] NO 3 when compared to the toluene mixing ratio alone, as shown in Fig. 7b. A range of different locations off the East and South coast of the UK were sampled during the 5 RONOCO flights. A summary of the flight tracks and vertical distribution of total [J Tol ] OH from the aircraft WAS bottles can be observed in Fig. 8. This shows 5 an expected reduction in reactive toluene equivalent with altitude. Variation at lower attitudes is the result of differential plume sampling.
The FGAM-DC-GC-FID used in this study exhibits reduced performance when analysing these aromatic species due to poor peak shape and increased co-elution, which contribute to a higher limit of detection. Using the gradients of linear regressions 10 from the measured GC × GC-TOFMS data and the toluene mixing ratio from the FGAM GC-FID, a prediction of the Σ[J Tol ] value was carried out. The predicted additional aromatic Σ[J Tol ] calculated using the FGAM-GC-FID toluene mixing ratio is shown by the blue line in Fig. 6. The additional aromatic content calculated using the proportionality factors shows a good level of agreement with the measured values, with a discrepancy 15 between the two trend lines of approximately < 0.4 % for both radical oxidants. The correlation data for each additional aromatic species can be found in Table 3 along with its corresponding R 2 value for the linear regression. The 2002 UK National Emission Inventory (NEI) ratios to toluene are also presented for comparison. The short daytime lifetimes of these larger molecular weight monoaromatics means 20 the observed daytime ratio to toluene is likely to change depending on the sample's distance from source. The RONOCO flights address this issue as they took place at night (or dawn) resulting in the airmass sampled containing well mixed source material with minimal OH losses. Many of the samples collected in this study were taken off the coast of the UK, (North Sea and the English Channel) providing a well mixed, integrated 25 assessment of the reactive higher VOC to toluene ratio. This is in contrast to surface measurements that can sometimes be affected by localised sources.

Effect of additional aromatic VOCs on ambient OH reactivity
The representation of the additional aromatic loading by means of a [J Tol ] value could potentially be used by atmospheric models as a means of accounting for additional radical reactivity. The OH reactivity (k ) is the total pseudo first order rate coefficient for the loss of OH in the atmosphere and is defined by Eq. (2) where [X ] i and k i are the concentrations and bimolecular rate coefficients for the i th species reacting with OH. Previous field based studies with extensive instrumentation Emmerson et al., 2007;Di Carlo et al., 2004;Sinha et al., 2010;Mao et al., 2009;Sadanaga et al., 2004;Stone et al., 2012) consistently find that a sig-10 nificant portion of measured OH reactivity is unaccounted for when compared to the directly measured OH sinks, which is often attributed to unmeasured VOCs. The potential additional contribution of the newly resolved aromatic species measured in this study to the OH reactivity measured/calculated for two UK ground based summertime field campaigns was investigated using campaign tailored chemical box models incor- 15 porating the Master Chemical Mechanism (MCMv3.1), constrained using the comprehensive range of measurements obtained from each campaign. The TORCH 2 study took place in May 2004 at the Weybourne Atmospheric Observatory (WAO), situated on the North Norfolk coast . During the study, WAO (52 • 57 23 N, 1 • 7 40 E) was subjected to similar air masses to those encountered during RONOCO (predominantly from a north easterly direction). As part of the comprehensive suite of observations carried out during this study, an OH lifetime 5 instrument was used to measure the temporal decay of artificially produced OH radicals with its ambient sinks in an atmospheric flow tube reactor ). The OH reactivity measurement was then compared to the reactivity calculated from the individual species measured at the site using Eq. (2). For much of the measurement period there was a significant difference between the measured and calculated 10 k , with the average value of k meas − k calc = 1.9 s −1 . In addition, the measured species were used as the inputs to a zero dimensional box model using the full MCM chemistry scheme (i.e. the base case TORCH 2 model discussed in this work). The discrepancy between the measured and calculated k' was reduced to 1.27 s −1 , owing to the production of reactive oxygenated intermediates in the model that are not directly measured 15 during the campaign and therefore were not included in the basic calculation. Although the addition of these oxygenated species has reduced the discrepancy between measured and modelled reactivity, there is still a significant degree that remains. To account for this missing reactivity with a surrogate unmeasured species of reactivity similar to o-xylene, would require a mixing ratio of approximately 4.0 ppbv. Similarly, the missing 20 reactivity could be accounted for by a thousand unidentified species with a mixing ratio of about 10 pptv and a rate constant equal to octane (the highest molecular weight species measured during the study). The [J Tol ] OH values derived in this study have been used to investigate the impact of the higher monoaromatic species on OH reactivity during the two TORCH campaigns. 25 The TORCH 1 and 2 models (as described in more details in Emmerson et al. (2007) and Lee et al. (2009) Fig. 6). The second model run (secondary case) was performed by simply increasing the toluene mixing ratio by a factor of 3 (i.e. [Tol] + Σ[J Tol ] OH ). The increased toluene mixing ratio in the secondary case is then allowed 5 to react to form oxidation products resulting in further compounds that can react with OH. The additional reactivity calculated using the modified models for the two TORCH campaigns are shown in Fig. 9. The red traces show the difference between the primary case and the base case and the blue traces show the difference between the secondary case and the base case. Both the TORCH 1 and TORCH 2 data show a similar increase 10 in predicted k values over the base case at 0.5-6.0 %. The TORCH 1 data however shows a significantly larger increase in reactivity over the base rate of between 0.1-1.0 s −1 . This is in contrast to the TORCH 2 data, with a range of between 0.02 and 0.2 s −1 .
The difference in absolute reactivity for the secondary case in the TORCH 1 data, 15 although only a small percentage difference, is still a significant value at a maximum of 1 s −1 . This shows that the addition of only ten, low abundance aromatics to the model can provide a significant extra sink for OH.

Conclusions
Data collected from a series of UK coastal and free tropospheric flights during the win-20 ter 2011 RONOCO flying campaign has shown that GC × GC-TOFMS is a useful tool for the atmospheric analysis of larger monoaromatic compounds. The technique has excellent resolution, sensitivity and is reliable with quantified values in good agreement with the established FGAM DC-GC-FID instrument. The added sensitivity and resolution is particularly useful in the detection and quantification of many species that more Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | that many of the higher molecular weight aromatics exhibited a strong correlation with toluene, indicating similar anthropogenic sources. Assuming this relationship is consistent, the use of the proportionality factors ([J Tol ]) obtained here can be used to predict the mixing ratios of these additional aromatic species without needing to measure them directly. Adding ten previously unaccounted for monoaromatic compounds to model 5 simulations of both polluted and rural chemistry increased the total simulated OH reactivity by up to 6 %, bring the modelled OH reactivity more into line with the measurements. The proportionality factors were determined using VOC measurements taken under a range of different air masses, from localised pollution to aged regional background and so should be typical for the UK. However more work is needed to study 10 whether these scaling factors are consistent at other locations. Quantification of many of the species detected within this study was not possible owing to a lack of appropriate standards and unknown stability within the WAS canisters. Their impact is also uncertain since in many cases kinetic data is not available. GC × GC-TOFMS has the ability to detect and resolve many other functionalities of at-15 mospherically relevant species such as higher MW multifunctional volatile oxygenates, halocarbons and alkyl nitrates shown in Fig. 10. In the future it would be possible to develop an atmospheric sampling method that has the potential to target all these species, within a single analysis, if standards for these species were available.