3.1.1 Benzene

In PTR-MS, protonated benzene is detected at m∕z 79. The comparisons presented in Table 4 indicate a significant difference at the 95 % confidence limit between the mean values measured by the PTR-MS at m∕z 79 and the benzene reported by the AT-VOC method.

Reduced major axis regression analysis between the PTR-MS data for m∕z 79 and the AT-VOC benzene data yielded a slope of 1.47±0.04, an intercept of 0.02±0.00 ppbv and an R² of 0.96 (Fig. 1a). The high R² value and small RMSD of 0.04 ppbv (RMSD / median =8 %) (Table 5, Fig. 1a) indicates the AT-VOC and PTR-MS were both responding to benzene.

It is possible the slope of ∼1.5 was a result of contributions to the PTR-MS signal at m∕z 79 from compounds other than benzene, such as fragment ions from ethylbenzene, propyl- and isopropyl-benzene, and butyl- and isobutyl-benzene which can potentially contribute to the signal at m∕z 79 (Warneke et al., 2003; Gueneron et al., 2015). In addition, an unknown CH₂O₄H⁺ ion signal was detected at m∕z 79 by high-resolution proton-transfer-reaction time-of-flight mass spectrometry (PTR-ToF-MS) in a rural atmosphere (Park et al., 2013).

Ethylbenzene was measured in the AT-VOC samples; however, propyl- and isopropyl benzene were not, and their contribution to the PTR-MS ion signal at m∕z 79 could not be assessed. Using the AT-VOC data for ethylbenzene and literature values of the ethylbenzene and benzene PTR-MS response variables – branching ratios (Gueneron et al., 2015) and ionization reaction rates (Cappellin et al., 2012) – a correction was applied to the PTR-MS m∕z 79 data to subtract interference due to the presence of fragment ion signals from ethylbenzene. This correction procedure is described in detail in the Supplement Sect. S.2.

The slope of the RMA regression between the corrected PTR-MS data and the AT-VOC data improved slightly to 1.36±0.03 (intercept $=\mathrm{0.03}\pm \mathrm{0.00}$, R²=0.96), indicating ethylbenzene made a minor but measurable contribution to the PTR-MS signal at m∕z 79 in this study.

The quantitative agreement between the measurement of benzene by PTR-MS and AT-VOC in this study was poorer than those reported in similar real-world inter-comparisons, most of which have reported slopes between 0.8 and 1.2 shown graphically in Fig. 2 (Warneke et al., 2001; de Gouw et al., 2003; Kato et al., 2004; Kuster et al., 2004; Jobson et al., 2010; Rogers et al., 2006; de Gouw and Warneke, 2007; Kaser et al., 2013; Wang et al., 2014; Kajos et al., 2015; Cui et al., 2016).

The degree of interference will vary with the relative concentrations of higher aromatics to benzene in the atmosphere being studied. As the higher aromatics have shorter atmospheric lifetimes than benzene, the interference will vary with ageing of an air mass. Thus, when measuring aged air masses, PTR-MS reported values should show better agreement with more selective GC techniques. In this study, within a large city, fresh emissions would be present, containing on average a greater fraction of higher aromatics. Thus, we would expect a larger contribution to the m∕z 79 signal from fragment ions of higher aromatics.

In summary, a comparison between the measurements of benzene by PTR-MS and the AT-VOC technique indicates a significant difference in the measured concentrations which is unresolved but is likely to vary according to the relative contribution of higher aromatics in different atmospheres.

3.1.2 Toluene

In PTR-MS, toluene undergoes non-dissociative proton transfer from H₃O⁺ producing a single ion signal at m∕z 93 (Gueneron et al., 2015). The comparisons presented in Table 4 indicate no difference between mean values measured by PTR-MS and AT-VOC methods at the 95 % confidence limit.

RMA regression analysis between the PTR-MS data at m∕z 93 and the AT-VOC data for toluene yielded a slope of 1.25±0.02, an intercept of $-\mathrm{0.03}\pm \mathrm{0.00}$ ppbv and an R² of 0.98 (Table 5, Fig. 1b). The RMSD was 0.11 ppb, which was only 5 % of the median PTR-MS value (Table 5). The high R² value and small RMSD indicates the PTR-MS signal at m∕z 93 was dominated by toluene.

The slope>1 may be a result of contributions to the PTR-MS signal at m∕z 93 from compounds other than toluene. These include α- and β-pinene, p-cymene and several C₉ aromatics (ethyltoluenes, 1,2,3-trimethylbenzene), all of which are known to produce fragment ions at m∕z 93 in PTR-MS (Warneke et al., 2003; Maleknia et al., 2007; Ambrose et al., 2010; Gueneron et al., 2015).

These potential interferent compounds, with the exception of p-ethyltoluene and 1,2,3-trimethylbenzene, were measured in the AT-VOC samples. Using the AT-VOC data for α- and β-pinene, p-cymene, and m- and o-ethyltoluene, as well as literature values of their PTR-MS response variables – branching ratios (Gueneron et al., 2015) and ionization reaction rates (Cappellin et al., 2012) – a correction was applied to the PTR-MS m∕z 93 data to subtract interference due to the presence of fragment ion signals from these interferents. This correction procedure is described in detail in the Supplement Sect. S.2.

This correction had a minor impact on the slope of the RMA regression ($\text{slope}=\mathrm{1.21}\pm \mathrm{0.02}$, intercept $=-\mathrm{0.03}\pm \mathrm{0.00}$, R²=0.98) and the reason for the remaining discrepancy was unresolved.

With the exception of two of studies (Kato et al., 2004; Kajos et al., 2015) previous inter-comparisons between toluene measurements by PTR-MS and GC techniques have reported slopes of 0.8–1.2 and generally good correlations (R²>0.75) (Fig. 2.) (Warneke et al., 2001; de Gouw et al., 2003; Kuster et al., 2004; de Gouw and Warneke, 2007; Kaser et al., 2013; Wang et al., 2014; Kajos et al., 2015; Cui et al., 2016).

In summary, a comparison between the measurements of toluene by PTR-MS and the AT-VOC technique indicates that there was not a significant difference in the measured concentrations at the 95 % confidence limit. There may be some residual unquantified interference with the PTR-MS toluene measurement which may vary due to contributions from the many additional monoterpene species commonly present in the atmosphere but not accounted for here (Geron et al., 2000; Maleknia et al., 2007).

3.1.3 C₈ aromatics

In PTR-MS, the signal at m∕z 107 is commonly regarded as a measure of the sum of the C₈ aromatic isomers ( m-, p-, o-xylenes and ethylbenzene) (de Gouw and Warneke, 2007). The comparisons presented in Table 4 indicate no difference between mean values of the sum of the C₈ aromatic isomers measured by the PTR-MS and AT-VOC techniques at the 95 % confidence limit; however, as observed for benzene and toluene, there appears to be a systematic difference between the two methods.

Table 5The slope (m), intercepts (b) and correlation coefficients (R²) from the RMA regression analysis between the PTR-MS, AT-VOC and DNPH–HPLC measurements. Also included are the estimates of random measurement uncertainty expressed as RMSD for each species and the ratio of the RMSD to the median PTR-MS value expressed as a percentage (%).

Download Print Version | Download XLSX

RMA regression analysis between the PTR-MS signal at m∕z 107 (ppb) and the AT-VOC data for the sum of the C₈ aromatics yielded a slope of 1.16±0.02, an intercept of $-\mathrm{0.01}\pm \mathrm{0.01}$ ppbv and an R² of 0.98 (Table 5, Fig. 1c). The RMSD of 0.09 ppbv was only 7 % of the median PTR-MS value (Table 5), which when combined with the high R² value indicates both techniques were responding to the C₈ aromatics.

The concentration of C₈ aromatics detected by the PTR-MS at m∕z 107 was quantified using a calibration factor of 19.78 ncps ppbv⁻¹ that was determined from measurements of a certified gas standard containing m-xylene. Unlike m-, p- and o-xylene, ethylbenzene undergoes fragmentation in the PTR-MS, and at the operating conditions used in this study ∼90 % of the ethylbenzene ion signal occurs at m∕z 107 (Gueneron et al., 2015). Consequently, using a calibration factor based on m-xylene alone will lead to an underestimation by PTR-MS when quantifying the sum of the C₈ aromatic isomers from the signal at m∕z 107. Using the AT-VOC data to determine the relative abundance of the C₈ aromatic isomers in the atmosphere as well as literature values of their PTR-MS response variables – branching ratios (Gueneron et al., 2015) and ionization reaction rates (Cappellin et al., 2012) – a corrected calibration factor of 19.61 ncps ppbv⁻¹ was applied to the PTR-MS m∕z 107 data to correct for the presence of ethylbenzene. The correction procedure is described in detail in the Supplement Sect. S.2.

This correction resulted in a minor increase in the slope to 1.19±0.02, with an intercept of $-\mathrm{0.02}\pm \mathrm{0.01}$ ppbv and an R² of 0.98.

A minor contribution to the PTR-MS signal at m∕z 107 may occur due to the presence of protonated benzaldehyde (de Gouw and Warneke, 2007), which was measured by the DNPH method in this study and comprised 2 % on average (range 0–5 %) of the sum of the C₈ aromatics reported by AT-VOC. Subtracting the concentration of benzaldehyde from the PTR-MS signal at m∕z 107 had no effect on the slope of the RMA regression (slope =1.16, R²=0.98) and resulted in a minor increase in the negative offset (intercept $-\mathrm{0.03}\pm \mathrm{0.01}$ ppbv) (see Supplement S.2).

The results reported here are similar to many previous intercomparison studies that have reported good quantitative agreement, within ±20 % (R²>0.85), between PTR-MS and GC techniques for the measurement of the sum of the C₈ aromatics (Warneke et al., 2001; Kuster et al., 2004; Jobson et al., 2010; Rogers et al., 2006; de Gouw and Warneke, 2007; Wang et al., 2014; Cui et al., 2016). However, slopes as low as 0.6 (Kato et al., 2004) and as high as 3.2 (de Gouw et al., 2003) have been reported with the discrepancy in both cases, which is attributable to calibration inaccuracies.

In summary, a comparison between the measurements of C₈ aromatics by PTR-MS and the AT-VOC technique indicates that there was not a significant difference in the measured concentrations. There may be some residual unquantified interference with the PTR-MS C₈ aromatic measurement.

3.1.4 Isoprene

In measurements of the atmosphere the PTR-MS signal at m∕z 69 is attributed to isoprene. The RMA regression analysis between the PTR-MS and AT-VOC data for isoprene yielded a slope of 1.23±0.07, an intercept of 0.31±0.10 ppbv and an R² of 0.75 (Fig. 1d). The lower R² and higher RMSD of 0.13 ppbv (Fig. 1d, Table 5) observed for isoprene indicate the two instruments may not have been responding entirely to the same compounds (Fig. 1d). The comparisons presented in Table 4 indicate a significant difference at the 95 % confidence limit between the mean values measured by each instrument.

Isoprene emissions are dominated by biogenic sources and are strongly light and temperature dependent with maxima in the afternoon. For SPS 2, when only the afternoon data were considered, closer agreement was observed between the PTR-MS and AT-VOC data for isoprene attributed to a 0.2 ppb lower intercept (0.11±0.10 ppb) and significantly higher R² of 0.93 ($\text{slope}=\mathrm{1.18}\pm \mathrm{0.06}$, RMSD =0.12 ppbv) (Table 5, Fig. 1e). There is no significant correlation between AT-VOC isoprene and the PTR-MS signal at m∕z 69 for the period 05:00–10:00 (R²=0.34), rendering the RMA slope and intercept essentially meaningless. There is a slope of 1.18 and offset of 0.41 ppb in the RMA regression for the period 19:00–05:00 (R2 =0.83), indicating that other compounds may be contributing to the PTR-MS signal at m∕z 69 and their relative contribution is largest at night and in the early hours of the morning when isoprene concentrations are lowest.

Park et al. (2013) observed three peaks at m∕z 69 in high-resolution PTR-ToF-MS spectra in a rural area: C₃H₂O₂H⁺ (∼10 %), C₄H₄OH⁺ (∼14 %) and C₅H₈H⁺ (∼75 %). PTR-ToF-MS measurements also identified dimethylcyclohexane and cyclopentene at m∕z 69 in air impacted by evaporative fuel emissions (Yuan et al., 2014). GC-PTR-MS analysis has also shown multiple other species can contribute to m∕z 69, specifically 2- and 3-methylbutanal, and 1-penten-3-ol in urban air (de Gouw et al., 2003); furan in air masses impacted by biomass burning (Christian et al., 2004); and 2-methyl-3-buten-2-ol in air masses impacted by emissions from pine trees (Karl et al., 2012). Unfortunately independent measurements of these interferent compounds are not available for this study and their contributions to the PTR-MS signal m∕z 69 cannot be estimated.

The results reported here are consistent with previous inter-comparisons studies between PTR-MS and GC techniques which have reported slopes of 0.79–2.15 often with significant (up to 0.39 ppb) offsets, shown graphically in Fig. 2 (de Gouw et al., 2003; Kato et al., 2004; Kuster et al., 2004; Wang et al., 2004; de Gouw and Warneke, 2007; Kaser et al., 2013).

In summary, a comparison between the measurements of isoprene by PTR-MS and the AT-VOC technique indicates a significant difference in the measured concentrations which may vary according to the relative contribution of other species that contribute to the PTR-MS signal at m∕z 69 particularly at night. The influence of these compounds on measurements of isoprene by PTR-MS, while well known, are not quantified in the bottom-up measurement uncertainty analysis of the PTR-MS technique.

3.2.1 Formaldehyde

In PTR-MS, protonated formaldehyde is detected at m∕z 31 (Hansel et al., 1997). The measurement of formaldehyde with PTR-MS is complex as its proton transfer chemical ionization reaction with H₃O⁺ is close to endothermic and loss via back reaction in humid air is non-negligible (Hansel et al., 1997; Inomata et al., 2008). In order to account for the water vapour dependence of the PTR-MS response to formaldehyde, daily instrument background and calibration measurements were made using zero air that had the same mole fractions of H₂O as the ambient air being sampled. The linear relationship observed between the formaldehyde calibration factors measured daily and the respective water vapour density (g m⁻³) was determined, and a corrected calibration factor was applied to the ambient hourly data based on the ambient water vapour density measured hourly.

The comparisons presented in Table 4 indicate that the mean values reported by PTR-MS and DNPH agree within 95 % confidence limits. RMA regression analysis between the PTR-MS signal at m∕z 31 and the formaldehyde in the DNPH–HPLC samples yielded a slope of 1.30±0.04, an intercept of $-\mathrm{0.07}\pm \mathrm{0.01}$ ppbv, an R² of 0.92 and an RMSD of 0.14 ppbv (N=77). The high R² value gives confidence that both the PTR-MS and the DNPH technique were both responding to formaldehyde.

To examine any possible effect of liquid water, the analysis was repeated excluding the data of 16–24 April. The results yielded a slope of 1.25±0.05, an intercept of 0.04±0.02 ppbv, an R² of 0.90 and an RMSD of 0.15 ppbv (N=53) (Table 5 and Fig. 1f), indicating a minor but significant effect of liquid water.

The slope of 1.25 may be a result of contributions to the PTR-MS signal at m∕z 31 from compounds other than formaldehyde, including methanol, ethanol, and methyl hydroperoxide (Inomata et al., 2008) and glyoxal (Stonner et al., 2016). The protonated molecular ion signal of ethanol and methyl hydroperoxide cannot be unequivocally identified in the PTR-MS spectra and their concentrations were not determined independently by either the AT-VOC or DNPH method, and consequently their contribution to the m∕z 31 signal cannot be determined in this study.

The dominant ion signal in the PTR-MS spectra of glyoxal is detected at m∕z 31 due to strong fragmentation of the parent ion (Stonner et al., 2016). However, these authors also found that, like formaldehyde, glyoxal also has a low proton affinity and loss via back reaction in humid air is also non-negligible, resulting in a very low PTR-MS sensitivity of ∼0.3–0.8 ncps ppbv⁻¹ compared to a formaldehyde sensitivity of ∼1.4 ncps ppbv⁻¹ for this study.

Using the PTR-MS and DNPH data for methanol and glyoxal respectively, along with laboratory measurements and literature values of the PTR-MS response variables for formaldehyde, methanol and glyoxal – branching ratios (BR_x) (Dunne, 2016; Stonner et al., 2016) and reaction rates (Cappellin et al., 2012) – a correction was applied to the PTR-MS m∕z 31 data to subtract interference in the measurement of formaldehyde due to the presence of fragment ions from methanol and glyoxal. This correction procedure is described in detail in the Supplement Sect. S.2. Applying the correction for methanol and glyoxal interference to the reduced m∕z 31 dataset (N=53) from this study resulted in a slight improvement in the slope to 1.17±0.06, with an intercept of $-\mathrm{0.07}\pm \mathrm{0.04}$ ppbv and an R² of 0.88 (Table 5).

Previous studies have reported PTR-MS values for formaldehyde that were systematically higher than DNPH–HPLC measurements (Wisthaler et al., 2008; Cui et al., 2016) and higher than differential optical absorption spectroscopy (DOAS) and Hantzsch techniques (Wisthaler et al., 2008; Warneke et al., 2011). Other studies report DNPH–HPLC values for formaldehyde that were systematically lower than those reported by other analytical methods (DOAS, FTIR, Hantzsch, TDLAS) (Kleindienst et al., 1988; Lawson et al., 1990; Gilpin et al., 1997; Hak et al., 2005; Wisthaler et al., 2006).

In summary, a comparison between the measurements of formaldehyde by PTR-MS and the DNPH technique indicates there was not a significant difference in the measured concentrations although some discrepancy between the two instruments remains unresolved.

3.2.2 Acetaldehyde

In measurements of the atmosphere the signal at m∕z 45 in PTR-MS spectra is commonly attributed to protonated acetaldehyde (de Gouw and Warneke, 2007). The comparisons between PTR-MS measurements at m∕z 45 and DNPH measurements of acetaldehyde presented in Table 4 indicate a significant difference at the 95 % confidence limit between the mean values measured by each instrument. RMA regression analysis between the PTR-MS data for m∕z 45 and the acetaldehyde values determined from the DNPH–HPLC samples yielded a slope of 1.47±0.09, an intercept of 0.14±0.02 ppbv, an R² of 0.72 and an RMSD of 0.11 ppbv (N=77).

To examine any possible effect of liquid water, the analysis was repeated excluding the data of 16–24 April (see Sect. 3.2.1). The results were a slope of 1.43±0.05, an intercept of 0.08±0.01 ppbv, an R² of 0.92 and an RMSD of 0.05 ppbv (N=54) (Fig. 1g, Table 5). The results indicate an insignificant effect on slope but a substantial increase in the correlation coefficient and reduction in RMSD by excluding the data indicating liquid water. The comparisons presented in Table 4 indicate a significant difference at the 95 % confidence limit between the mean values measured by each instrument.

A positive bias in PTR-MS measurements of acetaldehyde may result from contributions to the m∕z 45 signal from compounds other than acetaldehyde. Due to structural constraints the signal at m∕z 45 can be either C₂H₅O⁺ ions, ${\mathrm{HCO}}_{\mathrm{2}}^{+}$ and/or CH₃NO⁺. The contribution from protonated carbon dioxide (${\mathrm{HCO}}_{\mathrm{2}}^{+}$) is not relevant here as it is removed by the background zero correction.

Two studies using high-resolution PTR-ToF-MS have observed a single peak at m∕z 45 consisting of C₂H₅O⁺ (Park et al., 2013; Warneke et al., 2015). The C₂H₅O⁺ product ions may result from protonated acetaldehyde; protonated vinyl alcohol; protonated ethylene oxide; or fragment ions from ethylene glycol (Wood et al., 2015), ethanol, (Inomata and Tanimoto, 2009), 2-propanol (Inomata and Tanimoto, 2010), methyl ethyl ketone, methyl glyoxal and methyl isobutyl ketone (Dunne, 2016). None of these compounds were likely to be individually present in sufficient concentrations to account for the discrepancy observed in this study; however, the combined effect of numerous compounds yielding m∕z 45 product ions cannot be dismissed as a possible explanation.

In an atmospheric simulation chamber study three PTR-MS instruments reported acetaldehyde values close to the known injected value, whereas a DNPH method significantly underestimated (∼30 %) the known chamber concentration (Apel et al., 2008). In a recent comparison in urban air between PTR-MS and DNPH–HPLC, Cui et al. (2016) reported a slope of ∼1 between the two methods but a significant positive offset in the PTR-MS data of 0.83 ppbv and an R² of 0.56.

Herrington et al. (2007) reported the collection efficiency of acetaldehyde on DNPH cartridges declined from ∼100 % for a sampling duration of 6 h to ∼60 % for a sampling duration of 12 h, the reasons for which have not been resolved. As 8 and 10 h sampling durations were used for the DNPH sampling in this study, poor collection efficiencies may have resulted in a negative bias in the DNPH–HPLC measurements of acetaldehyde.

As shown in Fig. 2, other real-world inter-comparison studies have reported variable agreement between measurements of acetaldehyde by PTR-MS and GC methods (slopes of 0.87–1.7, intercepts of −0.25–0.22, R² of 0.38–0.86) (de Gouw et al., 2003; Warneke et al., 2011; Kaser et al., 2013; Kajos et al., 2015).

In summary, a comparison between the measurements of acetaldehyde by PTR-MS at m∕z 45 and the DNPH technique indicates there was a significant difference in the measured concentrations. There may be other species that contribute to the PTR-MS signal at m∕z 45, as well as under-reporting in the DNPH measurements. These processes are poorly understood and poorly quantified measurement uncertainties.

3.2.3 Acetone

In PTR-MS measurements, the ion signal at m∕z 59 is regarded as a measure of protonated acetone. The comparisons between the PTR-MS signal at m∕z 59 and DNPH measurements of acetone presented in Table 4 indicate a significant difference at the 95 % confidence limit between the mean values measured by each instrument.

The RMA regression analysis between the PTR-MS signal at m∕z 59 and acetone measured in the DNPH samples for the whole dataset yielded a slope of 1.67±0.13, an intercept of 0.79±0.12 and an R² of 0.51 (N=77). To examine any possible effect of liquid water, the analysis was repeated excluding the data of 16–24 April. The results yielded a slope of 2.01±0.14 ppbv, an intercept of 0.21±0.07 ppbv, an R² of 0.76 and an RMSD of 0.24 ppbv (N=53) (Table 5, Fig. 1h) indicating a significant effect of liquid water.

As discussed previously the compartment housing the DNPH cartridges in the automated sampler was maintained at ∼7 ^∘C and condensation of water from ambient air was observed in some of the chilled DNPH cartridges. As a result higher DNPH cartridge extraction masses occurred at dew point temperatures >7 ^∘C (Fig. 3e), and DNPH reported concentrations of acetone frequently approached zero at dew point temperatures >15 ^∘C (Fig. 3d).

Datapoints that coincided with average dew point temperatures >7 ^∘C were omitted from the RMA analysis. The average dew point temperature was <7 ^∘C in only 12 out of 53 DNPH samples, resulting in a significantly reduced dataset. However, omitting datapoints with dew point >7 ^∘C markedly improved the agreement between the DNPH and PTR-MS measurements of acetone, with the results of the RMA analysis changing to a slope of 1.40±0.14, an intercept of 0.22±0.12 and an R² of 0.89 (N=12) when water-affected samples were omitted.

Ho et al. (2014) also identified a significant negative bias in the collection efficiency of acetone on DNPH cartridges that was related to humidity, sample flow rate and sample duration. While Ho et al. (2014) used a similar DNPH cartridge type, these authors reported 35–80 % of acetone was lost under similar conditions as those experienced in this study (RH >70 %, sample flow 1 L min⁻¹, sample duration 8–10 h). These authors proposed a plausible explanation for the observed behaviour: when carbonyls pass through the DNPH sorbent, reactions occur involving the addition of the -NH₂ group to the -C=O group to form a reaction intermediate. The reaction between DNPH and ketones occurs at a slower rate than for aldehydes, resulting in poorer collection efficiencies for ketones. In the second step of the reaction, the intermediate loses a water molecule to form the hydrazone derivative. Therefore, when the water mixing ratio is high (i.e. high absolute humidity), loss via the back reaction may be substantial.

In PTR-MS, the ion signal at m∕z 59 is regarded as a measure of protonated acetone. However, in measurements of the atmosphere the m∕z 59 signal may also contain contributions from propanal and glyoxal (de Gouw and Warneke, 2007; Thalman et al., 2015; Stonner et al., 2016). Acetone, propanal and glyoxal were all measured by the DNPH method in the present study with mean values of 0.74±0.54, 0.03±0.04 and 0.06±0.04 ppbv respectively (N=53).

Using the DNPH data for propanal and glyoxal, along with literature values of the PTR-MS response variables – branching ratios (Spanel et al., 1997; Stonner et al., 2016) and reaction rates (Cappellin et al., 2012) – a correction was applied to the PTR-MS m∕z 59 data (excluding the data of 16–24 April, N=53) to subtract interference in the measurement of acetone due to the presence of propanal and glyoxal. This correction procedure is described in detail in the Supplement Sect. S.2.

Applying the correction for propanal and glyoxal interference to the reduced m∕z 59 dataset (N=53) from this study had a negligible effect on the agreement between the two methods ($\text{slope}=\mathrm{1.98}\pm \mathrm{0.13}$, intercept $=\mathrm{0.19}\pm \mathrm{0.07}$ ppbv, R²=0.76). Similarly when this correction is applied to the m∕z 59 data and compared with DNPH acetone for periods with average dew point temperature <7 ^∘C, only a minor improvement in the agreement between the two methods is observed ($\text{slope}=\mathrm{1.37}\pm \mathrm{0.13}$, $\text{intercept}=\mathrm{0.21}\pm \mathrm{0.12}$ ppbv).

Previous published atmospheric and chamber study measurements reported PTR-MS values for acetone that were ∼30 to >100 % higher than simultaneous DNPH–HPLC measurements (Müller et al., 2006; Apel et al., 2008; Cui et al., 2016). Conversely, generally good agreement has been observed between PTR-MS and GC methods and AP-CIMS (atmospheric pressure chemical ionization mass spectrometry) (slopes = 0.97–1.18, $\text{intercept}=-$0.28–0.06, R²=0.77–0.96) (Fig. 3) (Sprung et al., 2001; de Gouw et al., 2003; Kaser et al., 2013; Wang et al., 2014; Kajos et al., 2015).

Overall, the PTR-MS signal at m∕z 59 was dominated by acetone with minor contributions from propanal and glyoxal. Consistent with previous studies, a significant negative bias was identified in sampling of acetone onto DNPH cartridges and further work is required to determine the performance of DNPH cartridge sampling for quantitative measurements of acetone under real-world conditions. At high humidity, the formation of condensation in DNPH cartridges must be guarded against as stated in TO-11A (USEPA, 1999b).