Observations of Isocyanate, Amide, Nitrate, and Nitro Compounds From an Anthropogenic Biomass Burning Event Using a ToF-CIMS

Anthropogenic biomass burning is poorly represented in models due to a lack of observational data but represents a signi ﬁ cant source of short-lived toxic gases. Guy Fawkes Night (bon ﬁ re night) is a regular UK-wide event where open ﬁ res are lit and ﬁ reworks are set off on 5 November. Previous gas phase studies of bon ﬁ re night focus on persistent organic pollutants primarily using off-line techniques. Here the ﬁ rst simultaneous online gas phase measurements of several classes of compounds including isocyanates, amides, nitrates, and nitro-organics are made during bon ﬁ re night (2014) in Manchester, UK, using a time-of- ﬂ ight chemical ionization mass spectrometer (ToF-CIMS) using iodide reagent ions. A shallow boundary layer and low wind speeds favor pollutant buildup with typical HCN, HNCO, and CH 3 NCO concentrations of tens of parts per thousand increasing by a factor of 13 to potentially harmful levels > 1 ppb. Normalized excess mixing ratios relative to CO for a range of isocyanates and amides are reported for the ﬁ rst time. Using a HNCO:CO ratio of 0.1%, we distinguish emissions from ﬂ aming and smoldering combustion and report more accurate normalized excess mixing ratios for the distinct burning phases. While bon ﬁ re night is a highly polluting event, NO 2 concentrations measured at this location are higher at other times, highlighting the importance of traf ﬁ c as an NO 2 emission source at this location. A risk communication methodology is used to equate enhancements in hourly averaged black carbon and NO 2 concentrations caused by bon ﬁ re night as an equivalent of 26.1 passively smoked cigarettes.


Introduction
Biomass burning (BB) is a major source of gas and carbonaceous aerosol emission to the atmosphere (Andreae & Merlet, 2001), both of which act to reduce air quality worldwide (Molina et al., 2007). Solid biofuel burning makes up part of the anthropogenic contribution to BB, being the primary source of heating and cooking for 3 billion people worldwide (World Health Organization, 2015). Emissions from BB affect large population centers across the globe. For example, in New Delhi, India, a mega city with a population of 26,454,000 (United Nations, 2016), approximately 99% of inhabitants are exposed to more PM 2.5 than the 10 μg m À3 WHO air quality guideline (World Health Organization, 2006). An estimated 20% of the PM 2.5 , which contributes to this exposure originates from open BB (Amann et al., 2017). While this is not necessarily as important in the UK, solid biofuel burning is becoming more popular for financial, esthetic, and environmental reasons (Caird et al., 2008) and is significant in its contribution to reducing air quality (Fuller et al., 2014). One example of a regular, nationwide BB event in the UK is Guy Fawkes Night, or bonfire night, which is celebrated annually on and around 5 November by lighting open fires and fireworks as part of community events and at individual households. These bonfires are lit at roughly the same time during the evening and are designed to have a strong flaming phase that lasts for 1-2 h. After flaming, the fires are not refueled and so there is an extended period of smoldering as the fires are left to die away. The UK Environment Agency Isocyanic acid (HNCO) is another highly toxic, long-lived gas (lifetime of days to decades; Borduas et al., 2016) emitted from BB with similar anthropogenic and biogenic sources as HCN. Urban sources of HNCO are attributed to primary activity such as automotive emission (Jathar et al., 2017), residential heating (BB) (Woodward-Massey et al., 2014), and industrial processes, for example, from brick kiln emissions (Sarkar et al., 2016). A secondary source of HNCO is amide oxidation (e.g., Borduas et al., 2015), which has been observed at a suburban site in Mohali, India (Chandra & Sinha, 2016), and in an urban environment in Pasadena, California (Roberts et al., 2014). Mean urban concentrations of HNCO are variable having previously been measured to be on the order of 10-100 ppt by acetate CIMS in Pasadena (Roberts et al., 2014) and~1 ppb using a proton transfer reaction-mass spectrometer (MS) in Mohali (Chandra & Sinha, 2016). This higher level is attributed to a strong regional background potentially caused by the oxidation of another set of precursors, alkyl amines, originating from agricultural BB (Sarkar et al., 2016(Sarkar et al., , 2017. In addition to its toxicological importance, HNCO measurements are useful as the criteria of [HNCO]/ [CO] has been used to define the separation of flaming from smoldering combustion phases (Roberts et al., 2011). Methyl isocyanate (MIC) is a homologue of HNCO and a known secondary pollutant with precursors originating from both biogenic and anthropogenic sources (Lu et al., 2014;Woodrow et al., 2014) MIC has been measured as a direct emission from industrial processes and the burning of common building materials (Blomqvist et al., 2003;Henriks-Eckerman et al., 2002).
Amides are known products of BB (e.g., Stockwell et al., 2015) and from the burning of household materials (Kim et al., 2015). These compounds are known toxins (Gescher, 1990), which, as previously mentioned, are also precursors to isocyanic compounds (Borduas et al., 2015). In Shanghai, Yao et al. (2016) found that amides with greater than three carbon atoms were prevalent over lower mass amides, with concentrations on the order of parts per billion (ppbs) suggesting both secondary and industrial primary sources.
Organic nitrates are another class of toxic compounds with wide implications for atmospheric chemistry. They are a reservoir of NO 2 , thus enabling the transport of NO x and subsequent ozone formation at remote locations, and are typically semivolatile, contributing to secondary organic aerosol formation and thus particulate matter (PM) with further associated air quality issues. Other nitrates such as peroxynitric acid have previously been detected by iodide CIMS  and peroxyacetyl nitrate (PAN), which has previously Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316 been detected by PAN iodide CIMS . Nitro-organic compounds such as nitrophenol and its degradation products, which have been shown to be genotoxic (Sekler et al., 2004), have many sources including vehicular emission, degradation of pesticides (Lüttke et al., 1997, and references therein), secondary gas phase reactions (Berndt & Böge, 2006), and aqueous aerosol phase reactions (Yuan et al., 2015) and have vapor pressures low enough to readily partition into the condensed phase . Nitrophenols, methyl nitrophenols, and nitrocatechols have been detected in laboratory BB smoke using ToF-MS (Iinuma et al., 2010) and also in urban plumes (Mohr et al., 2013).
Measurements were carried out from 29 October 2014 to 11 November 2014 at the University of Manchester to assess the impact of bonfire night activity on local air quality and to probe the complex composition of an urban BB plume using novel identification methods to identify toxic species that currently do not feature in air quality health assessments.

Site Description
Manchester is located in the center of Greater Manchester Metropolitan County, an administrative area encompassing 10 different metropolitan boroughs of mostly urban districts, with a collective area of 1,276 km 3 and a population of 2.6 million inhabitants. The Whitworth Observatory is an urban rooftop measurement site approximately 15 m above street level and 100 m from the nearest road, located in the Simon Building at the University of Manchester's south campus, approximately 1.5 km south of Manchester City Centre (53.467°N, 2.232°W). For a map of the measurement location and sites of large-scale public bonfires, see Reyes-Villegas et al. (2017).

Instrument Description 2.2.1. Time-of-Flight Chemical Ionization Mass Spectrometer
A high-resolution ToF-CIMS as described by Lee et al. (2014) was deployed to measure a vast suite of atmospheric species using the iodide reagent ion. It consists of a reduced pressure ion molecule reaction (IMR) region coupled to a Tofwerk atmospheric pressure interface high-resolution time-of-flight mass spectrometer (Junninen et al., 2010). The instrument has been described in detail elsewhere (Lee et al., 2014); here we describe the specific details related to its setup for this campaign.
The IMR was held at a constant pressure of 100 mbar by a scroll pump (Agilent SH-112) controlled with a servo control valve placed between the scroll pump and IMR. Ambient air was drawn into the IMR via a critical orifice at 2.2 standard liters per minute (slm). The reagent I À ions were created by flowing 10 standard cubic centimeters per minute of methyl iodide (CH 3 I) emitted from a permeation tube held at 40°C through 1/8″ perfluoro alkoxy (PFA), which is carried by a 2 slm nitrogen (N 2 ) flow through a Po-210, 10 mCi, alpha emitting reactive ion source (NRD Inc Static Solutions Limited) orthogonal to the ambient flow into the IMR. Once in the IMR, the I À ions then ionize species of interest. The ionizer and ambient air flows mixed for approximately 30 ms until a fraction of the flow was sampled through an orifice into the first of four differentially pumped chambers in the ToF-CIMS. The first chamber was held at 150 mbar by a scroll pump (Triscroll 600), and the second stage was pumped by a split flow turbo molecular drag pump and held at 1.50 mbar. Quadrupole ion guides transmit the ions through these stages while simultaneously providing extracollisional cooling and energetic homogenization of the ions as they enter the extractor region. The electric field strengths in the axial direction (<2 V/cm) were set to optimize the total ion signal and transmission of the iodized species of interest (E/N ratio of 65 Td). Optimization of the I·H 2 O À cluster signal is considered essential to optimize system sensitivity since the detection of particular compounds, for example, formic acid, is dependent on the number of I·H 2 O À adducts. The ions were subsequently pulsed into the drift region of the ToF-CIMS at 22.22 kHz where the arrival time is detected with a pair of microchannel plate detectors. The mass calibration was performed for seven known masses: NO 3 À , I À , I À ·H 2 O, I À ·HCOOH, I À ·HNO 3 , I 2 À , and I 3 À , which covers a range of 62 to 381 m/z. The mass calibration was fitted to a third-order polynomial and was accurate to within 2 ppm; ensuring that peak identification was accurate below 20 ppm. The resolution was 3,560 at 127 m/z and 3,845 at 381 m/z. All species reported are observed as adducts with I À . 2.2.1.1. Sampling, Calibration, and Backgrounds To minimize sampling losses through the inlet tubing, a fast inlet pump is implemented to sample at 15 slm through 1 m long ¾″ PFA tubing. This translates to an inlet residence time of 1.2 s. Approximately 2.2 slm of Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316 the centerline flow is subsampled into the IMR of the ToF-CIMS through a conical shaped critical orifice, while the remaining flow is exhausted by the scroll pump. Backgrounds were taken every 6 h for 20 min by overflowing the inlet to the IMR with dry N 2 .
As the ToF-CIMS is able to detect 10 2 -10 3 species, calibration becomes difficult. One methodology to simplify calibration is to apply a uniform calibration factor using a maximum sensitivity based on adduct formation at the collision limit and assuming maximum transmission efficiency (Ehn et al., 2014;Lopez-Hilfiker et al., 2015). Another methodology makes use of an average organic acid calibration factor (Chhabra et al., 2015), while others aim to derive sensitivities from binding enthalpies of the reagent ion (Iyer et al., 2016). Here we apply the laboratory-determined calibration factors for a series of C 1 to C 5 aliphatic organic acids to species with an equivalent number of carbon atoms, assuming that the carbon number dictates the collision limit of adduct formation.
Above a carbon number of 2 within this particular series, the calibration factor is independent of m/z ( Figure 1). We use this assumption to apply a uniform calibration factor of 6.20 Hz ppt À1 , derived from formic acid, for C 1 compounds and a calibration factor of 1.28 Hz ppt À1 , derived from acetic acid, for C n compounds (n > 1).
Formic acid was both measured and calibrated for throughout the campaign by adding known concentrations of HCOOH as previously described in Le . The ToF-CIMS HCN signal was cross calibrated with the calibrated quad-CIMS. Known concentrations of HCN were produced by flowing from a HCN calibration cylinder (BW Technologies) that was diluted from 10 ppm mix with an accuracy of ±10% as described by Le Breton et al. (2013) (supporting information S1). The sensitivity of the quad-CIMS to HNCO is assumed to be the same as HCN. The sensitivity of the quad-CIMS and ToF-CIMS is shown in Table 1.
Sensitivity changes due to reagent ions are minimal. Mean average I À + I·H 2 O À counts were high (3.52 × 10 6 ± 5.2 × 10 5 (1σ)) and well above the threshold where we observe a much reduced dependency of sensitivity on reagent ion count (Le . We normalize to I À , I·H 2 O À , or the sum of both depending on which has the best correlation with the signal of the species of interest. If there is no discernable correlation, no normalization takes place. None of the signals for the species reported in this manuscript were normalized to I·H 2 O À or the sum of I À + I·H 2 O À . The calibration standard stock solutions for each acid were made with 95-97% reagent grade organic acids (Sigma-Aldrich) to produce 1% volume per volume solutions in water. These stock solutions were subsampled to make calibration standards ranging from 0.1 to 1.1 ppm. About 100 μL of calibration standard was injected into an evacuated Pyrex impinger connected to the evacuated 118 L Extreme Range Reaction Chamber (Leatheret al., 2010). The calibration standard volatilizes when exposed to the evacuated system, and partitioning to the gas phase is aided by passing N 2 carrier gas through the impinger until the chamber is filled to~760 torr (1.74% error). Calibration mixes of 500, 1,000, and 2,000 ppt are subsampled by the ToF-CIMS through 70 cm ¼" PFA tubing. Blank water backgrounds were performed before every calibration by injecting 100 μL of water and the lines cleaned with N 2 between sampling. These calibrations were performed postcampaign and compared with in-house made gas mixtures as described in Bannan et al. (2015) showing the same results to within 5% error.
A literature survey (Brophy & Farmer, 2015;Iyer et al., 2016;Lee et al., 2014; of formic acid calibration factors (normalized to 1 × 10 6 Hz reagent ion) provides a range between 2.9 and 13 Hz ppt À1 (with reported errors of 0.6-5.0 Hz ppt À1 ) highlighting  the variation in sensitivity that is associated with the instruments' condition and circumstance for this particular compound. Our formic acid calibration factor of 6.20 ± 1.28 Hz ppt À1 (4.13 ± 0.85 Hz ppt À1 normalized to 1 × 10 6 Hz reagent ion) is well within this range. Of 61 compounds surveyed in the literature, a range of calibration factors from 7.6 × 10 À5 to 22 Hz ppt À1 with a mean sensitivity and error of 5.20 ± 0.97 Hz ppt À1 and 1σ = 6.61 Hz ppt À1 further demonstrate the variability of the detection capabilities of the ToF-CIMS.

Meteorology and Air Quality
Meteorological measurements (pressure, temperature, relative humidity, wind speed and direction, precipitation, visibility, and actinic flux) are made approximately 100 m away on the nearby George Kenyon Building approximately 40 m above ground level in the University's south campus (approximately 53.466°N, 2.232°W). For a map of the measurement site and large-scale public bonfires, see Reyes-Villegas et al. (2017). Temperature, humidity, pressure, and wind speed and direction instruments are located approximately 3-5 m above the surface of the station to reduce the impact of the building below on the measurements being made. A summary of meteorological measurements, instruments, ranges, resolutions, accuracies, and data collection frequencies is summarized in Table 2. Automated high-frequency (0.1 Hz) long-term trace gas measurements of NO x , O 3 , CO, and SO 2 are made at the same location as the ToF-CIMS measurements. It is noted that the NO x measurement technique uses a molybdenum catalyst that is known to cause an overestimation of NO 2 concentrations by the nonselective conversion of NO y species as well as NO 2 to NO, which is actively detected by chemiluminescence. However, as the UK automatic urban rural network also uses this technique to measure NO x , the results presented here should be comparable. A lack of CO 2 measurements prevents the calculation of modified combustion efficiencies (MCEs) (Yokelson, Griffith, & Ward, 1996).

Kendrick Mass Defect
Kendrick mass defect (KMD) analysis is a mass spectrometric data analysis technique used to identify individual species in complex mixtures. While KMD analysis has been used to study the chemical composition of secondary organic aerosol (SOA) (Walser et al., 2008), it relies on a high-resolution (m/Δm > 50,000) typical of magnetic sector instruments with high accuracies, for example, <1 ppm (Bristow & Webb, 2003).
The average resolution of the ToF-CIMS was 3,800, and accuracy was <20 ppm. While these metrics indicate that the instrument is not able to unequivocally distinguish individual species, it is possible to use KMD analysis to suggest potential chemical formulae for unknown species if we account for the error in exact mass peak assignment and the selectivity of the I À reagent ion.
From the ToF-CIMS high-resolution mass spectra recorded over the campaign, a total of 652 peaks was manually identified; 75 peaks were first identified by a priori knowledge and by using peak fitting software, leaving 577 unknown assignments.
The peak list, containing the 75 known assignments, was then run through a program that matches known and unknown species that share the same KMD (within error) when normalized to different moieties (e.g., CH 2 , O, and CHO) as well as 12 C. The procedure is as follows: Where the m/z of the unknown assignment is an integer n times the m/z of the known assignment, it is interpreted that the assignments are related in KMD space and the formula of the unknown assignment can be determined from the known assignment (±n moieties). These potential assignments are assessed for their validity in terms of realistic structures and detectability by the ionization method. Figure 2 shows the total number of peak assignments as well as the known assignment of formic acid and the ability of the program to identify four "unknowns" in the same series, that is, the C 2 -C 5 organic acids.
As this technique relies heavily on high mass accuracy, accurate mass calibration and peak shape are vitally important. Peak assignments are treated as indicative and cannot account for isomeric or isobaric compounds. A sample of the chemical formula and suggested chemical species identified by this technique in conjunction with time series correlation analysis is discussed below, and isocyanate, nitrate, nitro, and amide compounds are tentatively assigned.

Pollution Events
Diurnal profiles of CO and NO x concentrations show a prominent daily bimodal cycle associated with traffic rush hours (07:00-10:00 UTC and 16:30-18:00 UTC). At the peak of the morning rush hour (approximately 08:30) average CO and NO x concentrations peak at 225 ppb and 31 ppb and again at 250 ppb and 30 ppb during the evening rush hour at 16:45. Minima for both species are observed during the early morning (04:35), with average values of 146 ppb and 5 ppb. Periodic stagnation events occur when wind speeds are low (<2.0 ms À1 ) with the highest concentrations of NO x and CO measured when these events coincide with a rush hour period. At these times, CO and NO x concentrations reach, at their maximum, 400 ppb and 508 ppb, respectively. Figure 3 summarizes the measurement period. All measurements are reported as 1 min averages.

Bonfire Night
Bonfire night (5 November) is defined as the period between 16:30 on 5 November and 07:30 on 6 November. During this period, the mean pressure was 998 hPa, mean temperature was 5.4°C (lowest during the campaign), and mean wind speed was 1.89 m s À1 . This combination of high pressure, low temperature, and low wind speed is indicative of a stable, shallow boundary layer and poor dispersion, increasing the buildup of pollutants. No one bonfire was directly sampled; instead, a mixture from multiple sources (both private and public) was accumulated and mixed together forming a homogenous air mass dominating the conventional background. NO x measurements from the UK automatic urban rural network monitoring network (Manchester Piccadilly, Manchester South, Salford Eccles, Glazebury) recorded maxima with varying wind directions and show little correlation with nearby large-scale events further suggesting that over a wider geographical area, the air mass composition cannot be attributed to any one single event (S2). 3.2.1. Identifying the Bonfire Burn Period Using BB Markers CIMS has previously been used to identify BB plumes using a statistical approach (Le Breton et al., 2013) where the concentration of a BB marker (e.g., HCN) above 6 times the standard deviation of its median background concentration is used to define a BB plume. As previously stated, no one plume during bonfire night could be identified individually. However, as bonfire night burning broadly follows a regimented schedule of prescribed activity (with flaming phase combustion occurring for 1-2 h during the early evening and smoldering thereafter lasting well into the next morning), this approach to defining in-plume sampling is applicable in identifying the temporal bounds of bonfire night burning activity, from here termed the bonfire burn period (BBP). HCN and isocyanic acid (HNCO) were used as BB markers to which the 6 sigma methodology was applied. The BBP started at approximately 16:00 and ended between 04:00 and 05:00. As HCN is more Figure 2. Kendrick mass defect plot of the identified peaks in this data set (gray). (inset) Exact masses (red) and measured masses (blue) of the C 1 -C 5 organic acids in CH 2 Kendrick mass defect space. Error bars show propagated 20 ppm error. All measured peaks are well within error. The exact masses (red) show where the measured masses should be found on this plot, connected by a straight line. Where the assignment is measured (picked, blue), the error in the exact mass is visible as a deviation from the straight line. The transition from formic to acetic acid shows a deviation in the straight line, which is a manifestation of the systematic error in the mass axis. All subsequent measured peaks sit on a straight line, indicating that there is no more deviation, and so the error in the mass axis is propagated but not compounded.   Table 3. Table 4 summarizes mean and maximum concentrations of nitrogen-containing compounds during the measurement period and the maximum concentration during the BBP. The limit of detection (LOD) reported is 3 times the standard deviation of the background measured at 1 min time resolution. Generally, concentrations of all species are highest during the BBP than at any other point. Two exceptions are PAN and trinitrocylcohexane. LODs range from one to tens of ppt.

10.1002/2017JD027316
Not including the BBP, 19 ± 13% (1σ) of the 0.32 μg m À3 mean total detected mass (including all unknown peaks) can be attributed to the 75 identified peaks. This percentage is conserved at the highest mass loading (1.00 μg m À3 ) recorded for this period. However, during the BBP the fraction of identified material increases to 24 ± 21%, and the mean total detected mass increases to 0.70 μg m À3 . At the highest mass loading of 2.0 μg m À3 during the BBP, the fraction of identified material further increases to 41%. The five largest organic concentration enhancements between the BBP and non-BBP are identified as C 6 H 6 O 2 , C 6 H 6 O 3 , CH 2 O 2 , C 3 H 6 O 3 , and C 7 H 8 O 2 .

Identification and Behavior of Key Species
Excluding the BBP, the mean ambient HCN concentration of 35 ppt and maximum concentration of 132 ppt is comparable with other ground-based HCN measurements, both rural and urban (Ambrose et al., 2012;Knighton et al., 2009). The species with the highest correlations with HCN with the BBP removed are propanoic acid (R 2 = 0.79), HONO (R 2 = 0.76), MIC (R 2 = 0.75), formamide (R 2 = 0.71), and N-methylformamide (R 2 = 0.68). Figure 4 summarizes diurnal profiles of selected species with the bonfire night period overlaid.
Low HCN concentrations <10 ppt were more commonly measured during weekends and during the nighttime when anthropogenic activity is at a minimum. Low concentrations were also measured when wind speeds were high and the wind direction was southwesterly, associated with cleaner, inflowing air masses.
Typical HCN concentrations of 30-50 ppt were most commonly observed during weekdays, when wind speeds were greater than 3.0 ms À1 (with no dependence on wind direction). Combined with a diurnal Note. BBP = bonfire burn period; LOD = limit of detections.

10.1002/2017JD027316
profile that does not exhibit a local emission signal (Figure 4), this range of concentrations is suggestive of a regional rather than local signal.
For the weekdays of 30 October and 4, 5, and 7 November between the hours of 07:00 and 09:00, HCN concentrations >50 ppt and a positive linear relationship with NO (R 2 = 0.67, 0.94, 0.73, and 0.59, respectively) demonstrates that there is a local vehicular component to the HCN measurement. Maximum HCN concentrations of 82 ppt on 4 November are higher than those associated with the regional signal (30-50 ppt). As the lifetime of NO is much less than HCN, the HCN/NO relationship is most obvious when wind speeds are low (typically <3.0 ms À1 ). As the wind speeds are higher on other weekdays, this is the most likely reason for poor correlations at those times. The example from 4 November indicates that a local enhancement of~20 ppt is possible above the regional background. The maximum concentration associated with the BBP is 1.24 ppb, a factor of 10 higher than ambient.
For HNCO, the nonbonfire mean concentration is 12 ppt with a maximum of 144 ppt. Similarly to HCN, the strongest relationship between HNCO and NO is during rush hour on the low wind speed days of 4 and 5 November (R 2 = 0.71 and 0.72) but, unlike HCN, is poor on the other low wind speed days of 30 October and 7 November. A secondary source component to HNCO is evident as the HNCO/HCN ratio is higher in the presence of the photochemical marker ozone ( Figure 5). One such formation pathway may be the oxidation of formamide (Borduas et al., 2015) with which HNCO has the highest correlation over the data set (R 2 = 0.71) (not including BBP). Further species with high correlations with HNCO are phenol (R 2 = 0.71), toluic acid (R 2 = 0.71), tartaric acid (R 2 = 0.70), and C 5 H 6 O 2 (R 2 = 0.70). The correlation between HCN and HNCO is good R 2 = 0.64. The maximum concentration measured during the bonfire plume of 1.64 ppb is comparable to concentrations recorded in agricultural burning plumes (Chandra & Sinha, 2016;Roberts et al., 2014) and is above the 1.0 ppb threshold, which is considered detrimental to human health (Roberts et al., 2011).
Maximum concentrations of MIC during the BBP of 4.3 ppb and 327 ppt outside of the BBP were observed. An R 2 of 0.75 with HCN and an of 0.53 with HNCO indicates that the behavior of MIC is more akin to HCN and is Additional nitrogen-containing compounds were identified as amides: acrylamide, formamide, Nmethylformamide, N,N-dimethylformamide, and N-formylformamide. Excluding the BBP, mean ambient concentrations are typically approximately tens of parts per thousand. Formamide, N-methylformamide, and N, N-dimethylformamide all exhibit maxima during the early afternoon (13:00-14:00) and minima during the morning rush hour (approximately 07:00-09:00). This is consistent with photochemical formation during the day and loss when NO x concentrations are high. During the BBP, concentrations of all amide species increase to hundreds of parts per thousand very quickly. Unlike the nitrates, no secondary peaks in concentrations are detected later in the morning. Instead, concentrations return to ambient levels indicating that emission is short lived and loss processes are fast. These loss processes are unlikely to be homogeneous gas phase reactions as the lifetimes of these compounds, controlled by reaction with OH, are typically 1-2 days (Borduas et al., 2015;Bunkan et al., 2015).

Distinguishing Combustion Regimes and Emission Ratios.
Laboratory and field studies indicate that the ratio of [HNCOCO] decreases from 0.6-0.1% during flaming combustion to values 5-10 times lower during smoldering combustion (Roberts et al., 2010(Roberts et al., , 2011. By applying a 0.1% threshold, the criteria for flaming combustion is met between 17:30 and 20:00 on bonfire night.

10.1002/2017JD027316
This is the approximate time that most bonfire flaming combustion occurs, as previously stated. These times are indicative rather than exact, as there will be some variation in flaming time between events and air mass mixing will affect the temporal coincidence at the measurement site.
The normalized excess mixing ratios (NEMRs) for a range of identified species relative to CO are calculated during the BBP using equation (1), many of which to the authors' knowledge have not previously been reported ( Figure 6). The background concentrations of the species of interest and CO are taken as the mean value from the diurnal profile of that species and CO at 5 min time steps from the entire measurement period with the BBP removed. "Plume" concentrations of the species of interest and CO are taken Figure 6. Normalized excess mixing ratios of selected compounds to CO. Gray points are smoldering phase emission described by the blue line, nongray points are flaming emission described by the red line, and black lines include all data. Slopes and R 2 are reported in Table 5. Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316 from the BBP. As bonfires are only lit after dark, photochemical production cannot occur during this time and so cannot confound any increases in measured concentrations.
By applying the [HNCOCO] metric to the NEMRs, it is possible to distinguish two distinct burning phases, flaming and smoldering ( Figure 6). Table 5 and Figure 7 summarize the NEMRs for the identified species. Comparing the emission factor (EF) of a species as a function of the MCE (equation (2)) is best used to define the magnitude of emission during smoldering and flaming phase combustion (e.g., Yokelson et al., 1999).
Calculating the MCE here is not possible as CO 2 was not measured; however, by using typical literature values of MCE where flaming and smoldering EF HCN have been measured, it is possible to estimate the concentrations of CO 2 present due to BB. Using MCEs of 0.95 for flaming emission and 0.88 for smoldering emission (Akagi et al., 2013), we estimate that during flaming emission, CO 2 concentrations are 19 times greater than CO concentrations but only 7.3 times greater during smoldering emission. Using CO 2 + CO as a proxy for mass of material burned, 2.4 times more material was burned during the flaming phase than during smoldering. So while the NEMR appears greater for flaming emission than smoldering, it does not take into account the greater mass burned during the flaming emission stage. Normalizing the HCN NEMR to [CO 2 ] + [CO] (as a proxy for mass burned) for the different phases shows that the smoldering NEMR is higher than the flaming NEMR by about a third (32%) (S3).
In all instances, the correlation of the compound and CO during the entire BBP is less good than the correlations observed when the BBP is separated into the two different burn phases. This indicates that separating the BBP produces more accurate NEMRs. Errors during the smoldering emission phase are likely to be overestimated as they include the transition period from flaming to smoldering. Together, this indicates that the NEMR is underestimated, when considering the BBP as a whole, if flaming combustion is occurring and may lead to an artificial reduction in the perceived relationship of a species to a specific combustion phase. For example, the correlation of MIC and CO over the entire BBP is poor (R 2 = 0.21) indicating that there is a poor evidence for MIC emission during the BBP. However, if the burning phase is accounted for, the correlation during flaming emission is much higher (R 2 = 0.71) suggesting that a strong source of MIC emission is occurring during this flaming emission phase. The 2σ error on the MIC BBP regression is 259% of the slope value, yet the flaming phase error is 67%, again indicating that by considering the burn phase, the emission ratio is more accurate. Figure 7 shows that the NEMRs for HCN, HNCO, and MIC are approximately an order of magnitude higher than the amides, which to our knowledge represent the first reported NEMRs for these compounds.
The NEMR for HCN during flaming emission is 1.11 ± 0.62 ppt ppb À1 and is comparable to NEMRs from boreal forest fires in North America and Siberia (Tereszchuk et al., 2013); an agricultural building fire (Brilli et al., 2014); and African savannah, tropical, and extratropical forest fires (Hornbrook et al., 2011) but is low compared with other studies (Hornbrook et al., 2011, and references therein). Higher HCN NEMRs (5-12 ppt ppt À1 ) have been detected in regions containing large cities where deposited NO x contributes to the nitrogen enrichment of the emission from local forest fires (Yokelson et al., 2007), but as the fuels used in this instance are unlikely to have accumulated NO x , this mechanism should not contribute to enhancing the nitrogen content of the fuel being burned.
Emissions of nitrogen-containing VOCs such as HCN and HNCO are known to be highly variable depending on the type and origin of the fuel. Increasing the nitrogen content of a biogenic fuel type by 1% can increase the emission of nitrogen-containing VOCs by 2-6% (Coggon et al., 2016), and the inclusion of a diesel oxidation catalyst to a diesel engine can increase the HNCO NEMR by a factor of 30 (Jathar et al., 2017). Total isocyanate concentrations measured after burning various plastics can vary by 3 orders of magnitude depending on the precursor fuel nitrogen content, and accordingly, plastics containing no nitrogen do not produce any isocyanates (Blomqvist et al., 2003). This variability in fuel type is one potential reason that the HCN NEMR is low. Another reason may be the assumption that the enhancement in CO above the calculated background is entirely due to bonfire burning is not accurate enough, as other sources of CO, for example, fireworks and or enhanced contribution from cars, cause an underestimation in the NEMR. Unfortunately, no discernable tracer of firework activity was found using iodide CIMS. During street to cityscale dispersion experiments over (approximately) 2-5 km, using inert perflurocarbon gas tracers, the ratio of different tracers released at the same point remained constant over all distances measured (Martin et al., 2010(Martin et al., , 2011Wood et al., 2009). Depending on the prevailing wind speed, the tracer ratio did not return to background levels following cessation of tracer release for 30-60 min at the release point itself.
In other experiments where inert tracer was heated to produce a buoyant plume, not only was it detected at much further distances downwind, the ratio was also preserved (Britter et al., 2002). While multiple sources will inevitably have different initial ratios (e.g., [HNCOCO]), those ratios should be preserved over at least kilometer-scale distances. It is impossible of course to categorically define the transition from a flaming phase to a smoldering phase, but given the very long lifetime of HNCO and CO with respect to travel time (based on when fires are likely to have been lit) the two distinct phases identified should be robust.

Air Quality
The highest concentration of CO recorded during the BBP is 1,551 ppb (as an 8 h running mean) which is comparable with summertime pollution events in London . A maximum NO 2 concentration of 43.76 ppb is measured during the BBP. Of the pollutants routinely measured at the Whitworth Observatory, none of the UK National Air Quality Objectives (NAQOs) (Defra, 2012) were breached during the measurement period at this measurement site. The highest maximum recorded concentrations of the pollutants at their NAQO sample periods were between 19:00 and 21:00 on bonfire night except the 24 h mean SO 2 , which occurred 4 days later.
As the Whitworth Observatory is approximately 15 m above ground level and 100 m from the nearest road, it is not considered a roadside/kerbside site, and so these data cannot convey any enhancement in concentration due to local vehicular emission at roadside where pollution levels and human exposure are often highest. However, this analysis does demonstrate that NO 2 and SO 2 concentrations are higher at times in the year other than bonfire night due to other phenomena, typically local traffic emission at times of low wind speed. Although the SO 2 concentrations at their highest are much lower than the NAQOs, NO 2 concentrations are noteworthy (Table 6).

Health Impacts
It is difficult to translate concentrations of the trace gas species measured with the ToF-CIMS into impacts on human health. Attention has mainly focused on particulate matter, NO 2 , CO, SO 2 , and O 3 . In a comprehensive review of the literature on the human impact of wildland fire smoke, which shares some of its source materials with what is burnt during bonfire night, consistent associations with mortality and respiratory morbidity were observed (Liu et al., 2015), although most of these wildfires would occur for longer periods than the one night evaluated here. Adetona et al. (2016) similarly concluded that for the general (exposed) public there was strong evidence of an association with acute respiratory effects, The maximum 1 h mean NO 2 concentration encountered during the 2014 Bonfire night was 46.76 ppb, which, although below air quality limits, is approximately 13.82 ppb above the estimated regional concentration above the surface layer for 2014 in Manchester . The 1 h mean concentration of black carbon fluctuated between 0 and 5 μg m À3 during the measurement period but increased to its maximum of 21.11 μg m À3 during bonfire night. The average concentrations of these two pollutants from 16:00 to 04:00 over bonfire night are estimated to be the equivalent to 26.1 passively smoked cigarettes.
The methodology described here, however, additionally identified trace amounts of many other chemicals generally not measured, many of which are known irritants and/or toxicants at higher (ppm) exposure levels: for example, in terms of workplace limit values MIC has a threshold limit value of 0.2 ppm (National Institute for Occupational Safety and Health, 2017), the Occupational Safety and Health Administration permissible exposure limit for HCN is 10 ppm (time weighted average) or 5 mg/m 3 averaged over 15 min (U.S. Department of Labor, 2005) and may be carcinogenic or teratogenic. However, the health impact of these chemicals at the low parts per billion level at which they are observed here remains unknown, if any given the likely duration during which human exposure would occur.

Conclusions
Here we present the first simultaneous online gas phase measurements of isocyanates, amides, and nitrates and nitro-organics using the ToF-CIMS during a 2 week period including bonfire night (5 November) and present, to our knowledge, the first reported NEMRs of amides to CO.
Typical HCN concentrations of 0-50 ppt were measured before and after bonfire night with a maximum of 82 ppt measured during rush hours when wind speeds were low. Mean HNCO concentrations were 12 ppt, and a maximum of 144 ppt was measured at the same time as high HCN concentrations. While HNCO concentrations show evidence of a photochemical source, MIC does not and behaves more like HCN as a primary pollutant. Maximum bonfire night concentrations of HCN and HNCO were 1.24 ppb and 1.64 ppb, respectively. Nitrates, nitrated species, and amide enhancements were lower than those for HCN and HNCO at 2-3.5 times higher than ambient levels.
No one bonfire was directly sampled as a shallow inversion layer, and low wind speeds caused a pooling of outflow from many large-and small-scale emission sources. We identify that the bonfire burn period (BBP) lasted between approximately 16:00 and 04:00-05:00 when using HCN as a tracer. No tracer for fireworks was found using iodide CIMS.
In the absence of CO 2 measurements and therefore MCE calculations, we use a HNCO:CO ratio of 0.1% to distinguish flaming and smoldering combustion (Roberts et al., 2011) and report normalized excess mixing ratios (NEMRs) relative to CO for HCN and a range of isocyanates and amides, many of which have not been reported before. The NEMRs separated by combustion phase are more accurate than when treating the BBP as a whole, with flaming phase showing greater enhancements than smoldering. We note that this does not take into account the amount of mass burned. The flaming phase NEMR of HCN = 1.11 ± 0.62 ppt ppb À1 , while low is consistent with other studies of biogenic BB (e.g., Tereszchuk et al., 2013). The uncertainty of the mixed fuel types used in bonfire construction prevents a detailed analysis of whether fuel nitrogen content is consistent with this lower HCN NEMR. Future work should attempt to better understand the composition of the mixed fuel types used in anthropogenic open burning.
The highest concentration of CO at this measurement site within a year and a half period (inclusive of this data) was measured during bonfire night, further confirming that this is a high-pollution event, yet match those measured in London during summer . Conversely, NO 2 concentrations were higher at other times within that year and a half period at this site, indicating that other phenomena can contribute more than bonfire night to increased levels of NO 2 . NO 2 concentrations measured here do not exceed current national air quality objectives (NAQOs) but equally do not represent kerbside measurements where the majority of NO 2 exceedances occur (McLean & Drabble, 2015).

10.1002/2017JD027316
The health impact of the exposures to the many identified chemicals in the parts per thousand range is unknown, although maximum 1 min average concentrations of HNCO during the BBP exceed 1 ppb, which has previously been considered above levels harmful to humans (Roberts et al., 2011). Initial assessment of a combination of black carbon and NO 2 indicates that this exposure, at the measurement location, would be equivalent to an average of 26.1 passively smoked cigarettes between 16:00 and 04:00 during bonfire night. As detection of toxic compounds becomes more routine, their inclusion in health burden quantification methodologies would further contribute to understanding the true impact of BB events on human health.