Airborne hydrogen cyanide measurements using a chemical ionisation mass spectrometer for the plume identification of biomass burning forest fires

. A chemical ionisation mass spectrometer (CIMS) was developed for measuring hydrogen cyanide (HCN) from biomass burning events in Canada using I − reagent ions on board the FAAM BAe-146 research aircraft during the BORTAS campaign in 2011. The ionisation scheme enabled highly sensitive measurements at 1 Hz frequency through biomass burning plumes in the troposphere. A strong correlation between the HCN, carbon monoxide (CO) and acetonitrile (CH 3 CN) was observed, indicating the potential of HCN as a biomass burning (BB) marker. A plume was deﬁned as being 6 standard deviations above background for the ﬂights. This method was compared with a number of alternative plume-deﬁning techniques employing CO and CH 3 CN measurements. The 6-sigma technique produced the highest R 2 values for correlations with CO. A normalised excess mixing ratio (NEMR) of 3.68 ± 0.149 pptv ppbv − 1 was calculated, which is within the range quoted in previous research (Hornbrook et al., 2011). The global tropospheric model STOCHEM-CRI in-corporated both the observed ratio and extreme ratios derived from other studies to generate global emission totals of HCN via biomass burning. Using the ratio derived from this work, the emission total for HCN from BB was 0.92 Tg (N) yr − 1 .


Introduction
Biomass burning (BB) is considered to be a major source of trace gases in the atmosphere (Li et al., , 2009Shim et al., 2007) and at levels significant enough to perturb regional and global atmospheric chemistry and composition (Levine, 2000). For example, large boreal forest fires in Russia from 2002 to 2003 were responsible for global growth rates of many trace gases including carbon dioxide and methane (Kasischke et al., 2005;Yurganov et al., 2005;, but also to contribute to climate change (Damoah et al., 2004;Vivchar et al., 2010;Tilmes et al., 2011).
BB is considered to be the major source of HCN in the atmosphere (Li et al., , 2009Liang et al., 2007;Shim et al., 2007) via the pyrolysis of N-containing species within the fuel (Johnson and Kang, 1971;Glarborg et al., 2003). Cooking fire emissions of HCN have also been observed in Mexico and Africa (Christian et al., 2010), although concentrations fell below Fourier transform IR (FTIR) detection limits. Singh et al. (2003) observed enhancements of HCN in China which correlated with CH 3 Cl indicating a source from hard coal burning for cooking. It must also be noted that biofuel is widely used in China (Streets et al., 2003) although data from Africa suggest emissions of nitriles are negligible Yokelson et al., 2003). HCN is also known to be emitted from motor exhausts, but is thought to be at negligible levels Lobert et al., 1991).
HCN has previously been observed from field biomass fires (Hurst et al., 1994a, b;Goode et al., 2000a;Yokelson et al., 2007b;Crounse et al., 2009) and using laboratory biomass combustion systems (Lobert et al., 1991;Holzinger et al., 1999;Christian et al., 2004;Becidan et al., 2007). Column measurements of HCN were measured from the International Scientific Station of Jungfraujoch (ISSJ) by solar infrared (IR) spectroscopy in 1998 during an intense period of biomass burning in the tropics . Akagi et al. (2011Akagi et al. ( , 2013 note the HCN emission ratios between similar fires can vary up to 60 % , from undetectable in wood burning to 3 % in peat fires. Singh et al. (2003) observed a HCN contribution from automobiles and from aircraft over the United States, although sources from automobile exhaust and industrial processes are thought to be negligible in comparison with BB (Lobert et al., 1991;Bange and Williams, 2000;Holzinger et al., 2001). A field experiment also indicated no detectable emissions of HCN from domestic biofuels . Thermodynamic calculations carried out by Boldi (1993) predict that an air parcel associated with lightning strikes could have a chemical composition such that the HCN / CO ratio would be around 10 −4 . Stribling and Miller (1987) showed that simulated lightning in a laboratory could produce HCN on planets such as Jupiter, strengthening the case that lightning-produced HCN has been observed on this planet (e.g. Podolak and Barnum, 1988). There have been examples where HCN has been observed in lightning perturbed air in the Earth's troposphere Liang et al., 2007), making lightning a possible additional source of HCN, although how much is still to be determined.
HCN is lost in the troposphere via the reaction with the hydroxyl radical (OH), creating a lifetime of a few years, although the reaction with singlet oxygen (O 1 D) is suggested to be important in the lower stratosphere. Uptake into the ocean is currently thought to be the dominant sink with an inferred global HCN biomass burning source of 1.4-2.9 Tg (N) yr −1 and an oceanic saturation ratio of 0.83 . This oceanic loss produces a lifetime of 2-5 months Singh et al., 2003). Although HCN may play an insignificant role in atmospheric chemistry (Cicerone and Zellner, 1983), it is thought to be an important source of nitrogen in remote oceanic environments . HCN is currently thought to be a useful tracer of BB as a consequence of its limited sources and sufficiently long atmospheric lifetime (Lobert et al., 1990;Holzinger et al., 1999).
Carbon monoxide (CO), acetonitrile (CH 3 CN) and HCN are all currently used as a BB tracer, but a standard approach has not been well defined. Thresholds of CO are used, but CO has many other strong sources (e.g. industrial activity). There can be difficulties in filtering out plumes which do not originate from BB. Recent studies implementing these various methods of identifying a BB plume have resulted in an uncertainty in the ratio of HCN to CO due to the variability observed in fires and potential mixing from other sources (Simpson et al., 2011;Vay et al., 2011;Hornbrook et al., 2011;Yokelson et al., 2009 ;Sinha et al., 2003;Andreae and Merlet., 2001). Using measurements of HCN and CO, BB plumes can be uniquely identified; enabling emission factors to be calculated from aircraft measurements.
Previous atmospheric measurements of HCN have been made using IR spectroscopy (Coffey et al., 1981;Zhao et al., 2002;Kleinböhl et al., 2006;Rinsland et al., 2007;Li et al., 2000). In situ measurements were first made in the stratosphere using NI-CIMS (negative ion-chemical ionisation mass spectrometer) (Schneider et al., 1997). Tropospheric measurements were then made by long-path Fourier transform IR (FTIR) spectroscopy within BB plumes (Goode et al., 2000a, b;Yokelson et al., 2007b), with a gas chromatography (GC) system, using a reduction gas detector (RGD) , by NI-CIMS using CF 3 O − as the reagent ion (Crounse et al., 2006Yokelson et al., 2007a), and by PTR-MS (proton transfer-mass spectrometry) (Knighton et al., 2009). Crounse et al. (2006) have shown that CIMS can be used to detect HCN selectively in the lower atmosphere, with a low detection limit and at a high frequency. Hornbrook et al. (2011) recently reported HCN to CO ratios from BB plumes using HCN measurements from a chemical ionisation mass spectrometer.
The overall goal of "quantifying the impact of BOReal forest fires on Tropospheric oxidants using Aircraft and Satellites (BORTAS)" campaign was to investigate the connection between the composition and the distribution of biomass burning outflow, ozone production and loss within the outflow, and the resulting perturbation to oxidant chemistry in the troposphere . Airborne measurements were taken on board the BAe-146 large atmospheric research aircraft (ARA) over eastern Canada between 12 July and 3 August 2011. The evolution and composition of these BB plumes were studied using the airborne instruments, ground instruments and satellites. In order to study

CIMS
A chemical ionisation mass spectrometer (CIMS) was used for real-time detection of HCN. The CIMS instrument employed here was built by the Georgia Institute of Technology as previously described by Nowak et al. (2007) and has been previously described for formic acid measurements (Le Breton et al., 2012). Subsequently various adjustments have been made to the inlet, and these are described in the following section. The schematic in Fig. 1 shows the set-up used and operating conditions of the CIMS on board the airborne platform FAAM BAe-146 research aircraft. The inlet consisted of 3/8 OD diameter PFA tubing of length 580 mm, which was heated to 50 • C to reduce surface losses. An orifice of diameter 0.9 mm was positioned at the front of the inlet to restrict the flow to 5.8 SLM. The pressure in the ionisation region was maintained at 19 Torr (133.322 Pa) throughout the flight by controlling the flow of nitrogen into the ionisation region using a mass flow meter.

Ionisation scheme
The ion-molecule chemistry using iodide ions (I − ) for trace gas detection has been described by Slusher et al. (2004) and was utilised here to detect HCN. A gas mixture of methyl iodide, CH 3 I, and H 2 O in N 2 is used to obtain reagent ions I − and water clusters I − ·(H 2 O), of which the latter is important for the ionisation of HCN, forming the adduct observed in & ) )

HCN calibrations, sensitivity and limit of detection
HCN was calibrated relative to that of formic acid, which was measured and calibrated in-flight throughout the campaign. The sensitivity of HCN relative to formic acid was determined from laboratory calibrations performed with lab air with a relative humidity (RH) ∼ 55 % and dry air by passing the lab air through a Drierite dryer. Known concentrations of HCN ( Fig. 2) and HCOOH were flowed into the CIMS under these conditions, and sensitivities for both gases were calculated. The BW Technologies HCN calibration cylinder was diluted from a 10 ppm mix with an accuracy of ±10 %, and the formic acid calibration standard was made as previously described in Le Breton et al. (2012). An average sensitivity ratio of 33 : 1 was observed. The HCN sensitivity was found to be independent of water cluster counts. The ion count signal throughout the flights was normalised to the formic acid sensitivity, which was determined by calibrations before, after and during the flight. The average sensitivity (±1σ ) for each flight was determined by taking the normalised sensitivity and multiplying by the reagent ion count rate to account for reagent ion variability from flight to flight. The average sensitivity for HCN was 4 ± 0.9 Hz pptv −1 for 1 MHz of reagent ion signal. The 0.8 Hz data were then averaged over 3 s for the analysis here. The limit of detection for HCN averaged to 3 s was 5 pptv.

STOCHEM-CRI modelling
The STOCHEM-CRI global chemistry-transport model has been described in detail in several recent papers (Archibald et al., 2010;Cooke et al., 2010a, b, c;Utembe et al., 2009Utembe et al., , 2011 and will only be briefly described here. STOCHEM-CRI is a global three-dimensional model, which uses a Lagrangian approach to advect 50 000 air parcels using a fourth-order Runge-Kutta scheme with advection time steps of 3 h (Collins et al., 1997). The transport and radiation models are driven by archived meteorological data, generated by the Met Office numerical weather prediction models as analysis fields with a resolution of 1.25 • longitude and 0.83 • latitude and on 12 vertical levels extending to 100 hPa (Derwent et al., 2008). The CRI (Common Representative Intermediates) chemical mechanism (CRIv2-R5; Watson et al., 2008;Utembe et al., 2009) has been incorporated into STOCHEM. CRIv2-R5 emits methane and 22 non-methane hydrocarbons. Each air parcel contains the concentrations of 219 species involved in 618 photolytic, gasphase and heterogeneous chemical reactions, using a 5 min time step. Formation of secondary organic aerosol (SOA) was derived from the oxidation of aromatic hydrocarbons, monoterpenes and isoprene (Utembe et al., 2009(Utembe et al., , 2011. Surface emissions for CO, NO x and non-methane hydrocarbons (NMHCs), distributed over five emission types (anthropogenic, biomass burning, vegetation, ocean and soil), are taken from the POET (Precursors of Ozone and their Effects in the Troposphere) inventory (Granier et al., 2012). The distributions for lightning emissions are parameterized based on the work of Price and Rind (1992) with the emissions being distributed evenly between the convective cloud top height and the surface. The emissions are scaled so that the global total NO x emission from lightning is 5 Tg (N) yr −1 .
The NO x emissions from civil and military aircraft are taken from NASA inventories for 1992 (Penner et al., 1999). The implementation of the emissions from aircraft is the same as for lightning with an annual total of 0.85 Tg(N) yr −1 . The model dynamical scheme and depositional schemes have been tested extensively through comparisons with 222 Rn and other models (e.g. Stevenson et al., 1998) and were part of a major model inter-comparison study of the CO budget using 26 global chemistry transport models (Shindell et al., 2006). This inter-comparison showed that model transport schemes compared favourably with measurements and other models. The model stratosphere-troposphere exchange (Collins et al., 2003) and its ability to transport pollutants over a range of scales effectively (e.g. Derwent et al., 2004) has also been demonstrated. In addition, convection within the model has been extensively tested and validated (e.g. Collins et al., 1999Collins et al., , 2002. Therefore, in terms of transport and depositional loss (loss via OH for HCN is slow), the model is more than adequate for the intended study.
The biomass burning emissions for HCN are distributed as that of biomass burning emissions for CO with a single ratio used. The distribution is taken from the POET database; although there are several other distributions, we have not integrated them using other methods so that we can make a direct comparison with other model integrations by us. However, we do provide an analysis of uncertainty in the discussion section of this paper.

HCN
Using the biomass burning emission ratios derived in this work for HCN : CO (3.68 × 10 −3 ), the total emission of HCN via biomass burning introduced into the model was 0.92 Tg (N) yr −1 . HCN is removed by reaction with OH (k = 1.2×10 −13 exp (−400/T ) cm 3 molecule −1 s −1 ) and deposited into the ocean at a rate of 3.4×10 −15 g (N) cm −2 s −1 . The model-derived lifetime for HCN is then ca. 3 months, consistent with other studies. Further simulations were carried out using more extreme ratios derived from other studies (e.g. 0.43 × 10 −3 , which yields a total emission of 0.11 Tg (N) yr −1 , and 12.6×10 −3 , which yields a total emission of 3.13 Tg (N) yr −1 ). In addition to these three integrations, a second set of three, using the three HCN BB emission ratios relative to CO, was performed with a lower deposition velocity (halved) leading to an overall lifetime of ca. 6 months.

Aircraft measurements
In addition to HCN data, observations of CO and CH 3 CN are also used in the analysis. CO data are reported at 1 Hz using a fast fluorescence CO analyser with an uncertainty of ±5 % (Gerbig et al., 1999). CH 3 CN was measured by PTR-MS (see Murphy et al., 2010, for experimental details). During the BORTAS flights the PTR-MS measured selected VOCs with a cycle time of around 15 s. CH 3 CN was measured at a m/z value of 42, which corresponds to the CH 3 CNH + ion.
The BORTAS-B campaign was conducted between 12 July and 3 August 2011 based in Halifax, Canada. CIMS data from 5 flights during this campaign are presented here. Palmer et al. (2013) present an overview of the campaign with full descriptions of the operating area, all flights and fire activity maps. Figure 3 shows the flight paths and altitude of the aircraft for the data presented here.

Biomass burning plume identification
HCN is a known BB tracer (Lobert et al., 1990), and CH 3 CN is also an indicator of BB emissions, which is not significantly enhanced in areas of anthropogenic activity (de Gouw et al., 2003(de Gouw et al., , 2006. These tracers and CO are regularly used to identify BB plumes, but a consistent method has not yet been established. Variation in background levels can make it difficult to define exactly when a plume is encountered without the aid of trace gas measurements which are not characteristic of BB fires. Hornbrook et al. (2011) defined a BB plume as having a CO mixing ratio above 175 ppbv, a CH 3 CN mixing ratio of > 200 pptv and an HCN mixing ratio of > 400 pptv. When the background concentrations are low, the plumes selected are generally picked by enhancement above background. This method is likely to cause inaccuracies as there is no definite point at which "plume data" can be determined. Vay et al. (2011) limited the HCN mixing ratio to above 500 pptv, CO to above 160 ppbv and CH 3 CN to above 225 pptv, but again they will experience similar issues with background concentrations. Simpson et al. (2011) state that the plume locations are defined by maximum CO concentrations. Holzinger et al. (2005) define a plume as CH 3 CN concentrations increasing three standard deviations above neighbouring points.
Here we evaluate a statistical approach to plume identification by assuming that the threshold limit to define "inplume" data is 10 times that of the standard deviation above the variation in the background (ICH-Q2B, 2009). The lifetime of HCN is long enough to allow plumes to be identified weeks away from the date of the fire. Dilution during this period will lower the concentrations, but there will still be a strong characteristic enhancement above the background levels, which will be analysed to evaluate if the normalised excess mixing ratio (NEMR) has been changed. The longrange transport and evolution of BB plumes can be evaluated
using this method as the plume is not identified by a general enhancement for any air mass.
In order to define the plume, the median background concentration for each flight was calculated. Ten standard deviations were initially implemented as the threshold for "plume data". It was found that decreasing the number of standard deviations incrementally by 1 made no significant change in the NEMR (i.e. the NEMR was within error the same, and R 2 until 6 sigma). This threshold was therefore utilised to create a HCN to CO NEMR as this allowed the maximum number of data points to be implemented into the calculation. The difference between the slopes produced by the 6-and 10-sigma approach is within 2-sigma error and have similar R 2 values 0.72 and 0.67 respectively. The 6-sigma method reports a higher R 2 and utilises more data points in the calculation; therefore the 6-sigma approach is used for the analysis of flights B621, B622, B624, B626 and B628 as shown in Fig. 4. The HCN to CO ratios derived from the BORTAS flights are similar to those reported in the literature (Table 2). In order to evaluate this method, other possible approaches to BB plume identification were implemented using this data set. HCN, CO and CH 3 CN have all been used in previous work Hornbrook et al., 2011) to identify BB plumes. We have used 7 methods to define a plume: (1) 6 sigma above the HCN background, (2) 6 sigma above the CO background, (3) 6 sigma above the CH 3 CN background, (4) above 100 ppb of CO, (5) above 300 ppt of HCN, (6) above 175 ppt of CH 3 CN and 100 of ppb CO, and (7) above 200 ppb of CO. The HCN to CO NEMRs for each of the methods with the corresponding errors and R 2 value are shown in Table 3, and the percentage of data calculated to be in a plume is shown in Table 4.
The 6-sigma HCN method produced the highest average R 2 of 0.72. The methods using CO as a threshold exhibited low correlations on flight B628, as a result of a peak    Fig. 5. This can be attributed to non-BB sources of CO enhancing the CO levels where no HCN sources were present, as CO is known to have other natural and anthropogenic sources (Logan et al., 1981). This highlights a potential problem when using CO as a marker, as other measurements are required in order to determine the source. B622 had low CO concentrations but did show structure which was reciprocated by the HCN measurements. The CH 3 CN data were too close to the detection limit during flight B628 to be able to determine a 6 sigma above background. The 200 ppb CO threshold approach removed all of these data from this flight, which would suggest that none of the flight encountered a BB plume. The method previously used by Hornbrook et al. (2011) produced an R 2 of 0.30 for this flight as a result of the low CO and CH 3 CN concentrations, whereas the 6-sigma HCN approach produced an R 2 of 0.69. Using CO as a BB marker is limited due to the variability in sources of CO. This method can be used for relatively fresh and unmixed plumes, whereas aged plumes may suffer from enhancements of CO from other sources. The methods using CH 3 CN data did produce a high R 2 on many flights, as shown in Table 3. However, the limit of detection (LOD) of the PTR-MS to CH 3 CN during BORTAS was a factor of 2.5 worse than the LOD of the CIMS towards HCN. As a result when sampling aged (and hence diluted) plumes, the S/N ratio is not significant enough to identify a plume, as exemplified by flight B628. Furthermore, as the PTR-MS was used to detect a range of compounds during each flight, the time response was slower than that of the CIMS system. As a result, the CH 3 CN data have a time average from 9 to 20 s, depending on the number of target gases that were being measured. With this particular sampling protocol, it is difficult to measure accurately plumes close to the source as the lower measurement frequency may struggle to pick up small plumes as a result of the speed of the aircraft. Nevertheless, CH 3 CN can be used to detect a BB plume accurately, under most conditions, as exemplified by flight B626.

&)HCN):)CO)correlations)in)plume)determined)by)the)6)sigma)HCN)approach)for)5)flights)during)the)BORTAS)campaign)
)    Figure 7 represents the whole data set from flight B622 exhibiting stratification between atmospheric layers. There is clear evidence of BB plumes as a function of altitude, providing further evidence of the preservation of distinct BB plumes. Further investigation into the variance of NEMRs with altitude shows that all separate plumes' NEMRs lie within 2 standard deviations when plotted against the altitude at which they were intercepted. The biomass-burning-influenced plumes detected throughout the campaign ranged from 1 to 11 days old when the photochemical age is calculated. Mixing with fossil-fuel-influenced air masses would change the HCN : CO NEMR, although a high average HCN : CO R 2 correlation coefficient of 0.86 indicates that these plumes have not been influenced by fossil fuel plumes from North America. 90 80
air mass passed over northwest Ontario, although it may have also been influenced further afield in areas such as northern Alberta and the Northwest Territories, possibly explaining the slight non-linear fit observed in Fig. 8, indicating a degree of mixing during this flight. Acetylene (C 2 H 2 ) data also follow the same structure that HCN, CO and black carbon exhibit.

Emission ratios
The 6-sigma technique is used here to calculate the emission ratio of HCN from 4 flights during the BORTAS campaign 2011. Figure 8 shows all the data points which are used to calculate the mean NEMR: 3.68 ± 0.149 pptv ppbv −1 . The NEMR was calculated using the equation The data from flight B622 were omitted from this calculation due to the possible ageing and mixing of the plume. The NEMRs calculated during BORTAS are similar to those found in previous work, as seen in Table 2. The NEMRs reported in previous work vary from 0.43 to 12.8 pptv ppb −1 . The NEMR calculated here for HCN from Canadian BB plumes is lower than that found by Simpson et al. (2011), 8.2 pptv −1 ppb −1 . Hornbrook (2011) highlights the observed variation in the ratios, and offers a possible explanation for the difference between these NEMRs, but this variation is not seen from flight to flight during the BORTAS campaign, which measured both fresh plumes and aged plumes. Californian fire emission ratios during ARCTAS-CARB were significantly lower than the Canadian and Asian fires (Hornbrook et al., 2011), ranging from 2.4 ± 0.9 to 8.8 ± 3.8 pptv ppbv −1 respectively. Using the NEMRs calculated by Hurst et al. (2001), Hornbrook et al. (2011) report the low ratio of 0.43 pptv ppbv −1 originating from African savannas, tropical forests and extratropical forests, whereas Yokelson et al. (2007b) reported a ratio in the Mexico City region of 12 ± 7 pptv ppbv −1 . The increase in these ratios may be attributed to the high NO x levels found around Mexico City. Although emissions of HCN from motor vehicles are not believed to be important on a global scale, localised emissions may become significant . Boreal forests are the primary source of fires in Canada, whereas Californian fires may be a result of varying fuels, such as coniferous forests and grass and shrubs (Hornbrook et al., 2011).
The 6-sigma HCN method of identifying BB plumes has shown the veracity of HCN as a BB-influenced plume marker. In addition, this method performed better than the others over the BORTAS campaign, as indicated by the statistics presented. Also, the 6-sigma HCN method showed the ability to define BB plumes accurately in air masses which had a low HCN background, enabling the identification of BB plumes in air masses distant from sources that were not constrained by a set threshold concentration. For example, a plume may have dispersed over large distances, lowering the concentration below the limit that defines a plume using previous methods. Nonetheless, this 6-sigma technique is still able to identify these plumes, as they are defined relative to the background. This 6-sigma method also has the same ability to determine VOC to CO ratios with the percentage of data at a high time resolution (3 s). This method is therefore used to determine a HCN to CO ratio for models to calculate a global HCN budget.

Model results
The purpose of the model integrations was to inspect the global HCN levels generated using the extreme HCN biomass ratios (relative to CO) reported in the literature and the value determined in this study, using two ocean deposition velocities that lead to HCN lifetimes of ca. 3 months and ca. 6 months. It should be noted that the variation in emission ratio reported in the literature is not in question here. There are myriad reasons for the variation in terms of vegetation type, temperature of the burn, etc. It is also noted that the limited available field measurements make comparison and constraint of the model somewhat limited. However, as we hope to show, the model results are instructive. The model results are in line with basic expectations: as the emission ratio increases, the global HCN level increases; when the deposition velocity is decreased, the global HCN for all three integrations also increases. Model results are presented in Fig. 9, which shows yearly averaged latitude-altitude profiles. Given the overall uncertainties, it is not justified to present more detailed seasonal results. We have deliberately used one HCN / CO ratio to distribute HCN emissions in these model runs to simplify them. We are not trying to reproduce any field data, but we can compare with measurements and of course compare between the integrations performed in a straightforward manner. If we assume the lower deposition velocity leading to a lifetime of about 6 months, we observe that an emission ratio of 0.4 × 10 −3 leads to a global yearly averaged HCN level of 10-20 ppt. An emission ratio of 12.6 × 10 −3 leads to a global yearly averaged HCN level of 300-600 ppt, and an emission ratio of 3.7 × 10 −3 leads to a global yearly averaged HCN level of 80-180 ppt. In each case the highest levels are observed over the tropical regions, obviously driven by high biomass burning, with little variation in vertical structure, reflecting the surface deposition process dominating loss and leading to a sink in the Southern Hemisphere in the model.
There is no attempt here to reproduce field measurements, but it is instructive to compare field data with the model. We have concentrated on lower and mid-tropospheric measurements and note that there are measurements in the upper troposphere and lower stratosphere. Liang et al. (2007) observed HCN using aircraft during INTEX-A (July-August 2004). This field campaign ranged across the USA and Canada and took in measurements in both the Pacific and Atlantic oceans. Although very high levels were detected in biomass burning plumes (1090 ± 850 ppt), the background levels observed were 290 ± 70 ppt. In their comparison, Liang et al. (2007) reported levels in Asian plumes of 420 ± 60 ppt compared with 270 ± 80 ppt returned by Jacob et al. (2003) during Trace-P. Notholt et al. (2000) conducted vertical column measurements of HCN and other gases between 57 • N and 45 • S across the central Atlantic. HCN was detectable between 30 • N and 30 • S, with column amounts retrieved between 0 and 12 km. The HCN column amounts ranged from 100 to 220 ppt, with the maximum occurring just south of the Equator (10-15 • S). Singh et al. (2003) report HCN levels of around 250 ± 150 pptv for HCN in February to April, and Ambrose et al. (2012) and Rinsland et al. (2007) report mean mixing ratios of 360 ppt and 220 ppt respectively, while Knighton et al. (2009) report a concentration ranging from 100 to 600 ppt and a mean background of 200 ppt. Therefore, based on the available measurements discussed thus far, we would conclude that yearly averaged levels of HCN vary between approximately 100 and 450 ppt in the lower to mid-troposphere. In the upper troposphere lightning may well contribute an additional non-negligible source, and this region will be impacted by continental-scale plumes, evidenced by a variety of measurements (e.g. Liang et al., 2007;Singh et al., 2007;Park et al., 2008;Randel et al., 2010;Wiegele et al., 2012). These plumes will contain a mixture of potential sources of HCN, of which biomass burning may well be the most predominant. It is also recognised that emission ratios will vary for different types of biomass burning, depending on vegetation type, temperature of burn, etc. No one ratio will be representative of the global emission. However, inspection of the model integrations suggests that the extreme ratios returned from field measurements are indeed extreme values: adopting a uniform ratio of 0.4 × 10 −3 returns a globally averaged HCN that is far 9226 M. Le Breton et al.: Airborne hydrogen cyanide measurements (a) ) ) 9a))) ) 9b))) ) (b) ) 9a))) ) 9b))) ) (c) ) 9c))) ) 9d))) (d) ) 9c))) ) 9d))) (e) ) 9e))) ) 9f))) (f) ) 9e))) ) 9f))) Fig. 9. Yearly average HCN derived from model integrations with two ocean deposition velocities leading to a lifetime of 3 months and 6 months. (a) Emission ratio (HCN : CO) of 4 × 10 −4 ppb ppb −1 and a lifetime with respect to deposition of 3 months. (b) Emission ratio of 4 × 10 −4 and a lifetime with respect to deposition of 6 months. (c) Emission ratio of 3.7 × 10 −3 and a lifetime with respect to deposition of 3 months. (d) Emission ratio of 3.7 × 10 −3 and a lifetime with respect to deposition of 6 months. (e) Emission ratio of 12 × 10 −3 and a lifetime with respect to deposition of 3 months. (f) Emission ratio of 12 × 10 −3 and a lifetime with respect to deposition of 6 months. too low, irrespective of whether the lifetime is 3 or 6 months. Similarly, adopting a ratio of 12.6×10 −3 produces HCN levels that have been observed but they are somewhat higher than expected for a yearly average, given the background measurements made. Using the ratio derived in this study as a global value produces HCN levels that are reasonable, compared with available field measurements, but are an underestimate. An underestimate is completely consistent with the fact that more influential biomass burning regions have returned a higher HCN : CO ratio. The satellite-derived measurements of Wiegle et al. (2012), although restricted in altitude to above 5 km, suggest strongly that biomass burning (particularly that located in the Southern Hemisphere) is a dominant source and lends confidence to the present broad brush model comparisons with measurements. Vegetation has also been suggested as a non-negligible source of HCN (e.g. Fall et al., 2001), and vertical profile data from the Jungfraujoch station in Switzerland   Model derived Taylor et al. (1996) 507 Model comparison Shindell et al. (2006) 494 POET inventory Granier et al. (2005) suggest that, in addition to biomass burning, there may well be a significant direct emission from vegetation. It is clear that the depositional velocity adopted for HCN is crucial in any budget analysis, and for the ones used in this study increasing the lifetime of HCN from ca. 3 months to 6 months increases model HCN levels by a factor of ∼ 1.4, irrespective of the emission ratio used. In this integration the CO biomass burning total used is ∼ 500 Tg yr −1 . There are a range of estimates for this total summarised in Table 5.
There is a wide range of estimates, but the majority lie between 750 Tg yr −1 and 350 Tg yr −1 . Therefore to a first approximation the model-estimated HCN levels will vary by a factor of 1.5 based on the CO emission uncertainty alone. HCN : CO biomass burning emission ratios will vary with type of burn and vegetation and are summarised in Table 5. Note that this ratio will vary with vegetation type and that using one ratio is not physically correct. However, the range reported has allowed us to investigate in a simple way the impact of these ratios on atmospheric levels.
In summary, model integrations suggest that the extreme ratios reported in the literature generate too little or too much HCN and really are extreme values. Using the ratios reported in this study to drive the model emissions produces HCN levels that are an underestimate compared with a range of field measurements, which are consistent with the fact that higher ratios are seen in tropical biomass burning events for example. However, the model integrations highlight that depositional loss is very important to determining HCN atmospheric background levels and that further work is required to constrain this loss process. In addition, more atmospheric measurements are welcome, particularly vertical column and transects.

Conclusions
A CIMS instrument was developed for the airborne measurement of HCN in the lower atmosphere using methyl iodide as the ionisation reagent gas. HCN measurements were successfully attained over Canada in July and August 2011, during the BORTAS-B 2011 campaign on board the FAAM BAe-146 aircraft. The high sensitivity (4 ± 0.9 ion counts s −1 pptv −1 ), low limit of detection (5 pptv) and selectivity of the data acquired and presented here with a time resolution of 3 s illustrate the ability of this instrument to measure HCN with a high precision; it is, therefore, a highly sophisticated instrument for detecting BB-influenced plumes. The mixing ratios measured through the BB-influenced plumes ranged from 0.67 to 5.2 ppb covering the range of previously reported atmospheric levels (Singh et al., , 2012Knighton et al., 2009) and were strongly correlated with CO and CH 3 CN, strengthening the ability of HCN to be a unique marker for biomass burning.
The 6-sigma methodology implemented and tested here for plume definition has been shown to produce the strongest correlation with CO, indicating that it is potentially an excellent method for defining biomass burning plumes. The NEMR (relative to CO) calculated using this plume identification method was 3.68 ± 0.149 pptv ppbv −1 , which is in the range of previously reported values (Andreae et al., 2001;Sinha et al., 2003;Yokelson et al., 2009;Hornbrook et al., 2011) indicating the precision of the HCN measurements. The study-averaged NEMR was then used to estimate the total emission of HCN via biomass burning, which was calculated to be 0.91 Tg (N) yr −1 .
These first results of HCN measurements by CIMS using I − chemistry show the capability of CIMS to attain highfrequency HCN measurements in the lower atmosphere with a high sensitivity and low limit of detection. The data produced also show the accuracy at which HCN measurements can define biomass burning plumes and the reliability of this method.