Isocyanic acid (HNCO), an acidic gas found in tobacco smoke, urban
environments, and biomass-burning-affected regions, has been linked to
adverse health outcomes. Gasoline- and diesel-powered engines and biomass
burning are known to emit HNCO and hypothesized to emit precursors such as
amides that can photochemically react to produce HNCO in the atmosphere.
Increasingly, diesel engines in developed countries like the United States
are required to use selective catalytic reduction (SCR) systems to reduce
tailpipe emissions of oxides of nitrogen. SCR chemistry is known to produce
HNCO as an intermediate product, and SCR systems have been implicated as an
atmospheric source of HNCO. In this work, we measure HNCO emissions from an
SCR system-equipped diesel engine and, in combination with earlier data, use
a three-dimensional chemical transport model (CTM) to simulate the ambient
concentrations and source/pathway contributions to HNCO in an urban
environment. Engine tests were conducted at three different engine loads,
using two different fuels and at multiple operating points. HNCO was measured
using an acetate chemical ionization mass spectrometer. The diesel engine was
found to emit primary HNCO (3–90 mg kg fuel
Isocyanic acid (HNCO) is a mildly acidic gas, which is highly soluble at physiologic pH and can participate in carbamylation reactions in the human body (Wang et al., 2007) and lead to adverse health outcomes such as cataracts, atherosclerosis, and rheumatoid arthritis (Fullerton et al., 2008; Scott et al., 2010). Isocyanates, the family to which HNCO belongs, are extremely hazardous. The accidental release of methyl isocyanate from a pesticide plant in Bhopal, India, in 1984 resulted in thousands of deaths and hundreds of thousands in injuries within weeks of the release (Broughton, 2005). Although isocyanates are understood to be toxic and regulated through best practices in indoor and occupational environments (Alexeeff et al., 2000; SWEA, 2005), it is unclear whether ambient concentrations of HNCO (and isocyanates in general) are high enough for it be of concern as an outdoor air pollutant. Roberts et al. (2011) proposed that exposure to HNCO concentrations exceeding 1 ppbv could be harmful to humans.
Only a few studies have measured ambient concentrations of HNCO. Roberts and coworkers (Roberts et al., 2011, 2014) used a chemical ionization mass spectrometer (CIMS) to measure HNCO in urban areas (Pasadena, CA) and biomass-burning-affected regions (suburban and rural Colorado). While concentrations in urban areas were consistently below 100 pptv, ambient HNCO concentrations exceeded 100 pptv in regions affected by wildfire plumes and agricultural burning; in case of the later source, HNCO concentrations reached as high as 1.2 ppbv. In urban areas, they found evidence that 60 % of the HNCO came from primary (i.e., directly emitted) sources while 40 % came from secondary (i.e., photochemically produced) sources. Wentzell et al. (2013) used a CIMS to measure ambient concentrations of HNCO in urban Toronto and found that the measured HNCO (20–140 pptv) correlated strongly with benzene, which the authors interpreted as a fossil-fuel-based source. Zhao et al. (2014) measured HNCO using a CIMS at an elevated site near La Jolla, CA, and found evidence for photochemical production as well as significant uptake of HNCO by clouds. And finally, Chandra and Sinha (2016) measured ambient HNCO in Mohali, India, using a proton transfer reaction mass spectrometer (PTR-MS) and found that the ambient concentrations regularly exceeded 1 ppbv during post-harvest agricultural burning. Despite only a handful of observations in a few locations, it is clear that atmospheric HNCO has both natural (e.g., wildfires) and anthropogenic (e.g., agricultural burning) sources. With these sparse observations, we are just beginning to understand the spatiotemporal distribution and the contribution of natural and anthropogenic sources to ambient concentrations of HNCO.
Similar to the ambient observations, there have only been a handful of
studies that have investigated HNCO emissions from anthropogenic sources,
most of which have focused on gasoline- and diesel-powered sources. Previous
HNCO studies on natural, biomass burning sources have been performed by
Roberts et al. (2011) and Coggon et al. (2016). Brady et al. (2014) measured tailpipe emissions of HNCO from eight light-duty gasoline
vehicles (LDGVs) and suggested that HNCO emissions from LDGVs were not a
result of in-cylinder combustion but rather a result of CO- and NO
Increasingly, new diesel engines sold in developed economies (e.g., United
States, Canada, European Union) need to be equipped with selective catalytic reduction (SCR) systems to reduce tailpipe emissions of NO
To date, there has only been a single study that has used a large-scale
model to simulate ambient concentrations of HNCO from biomass burning and
biofuel combustion. Leveraging the measurements of Roberts et al. (2011), Young et al. (2012) simulated ambient concentrations of HNCO
using a global model. They found that surface HNCO concentrations might only
be of human health concern (> 1 ppbv for more than 7 days of the
year) in tropical regions dominated by biomass burning (Southeast Asia) and
in developing countries (northern India and eastern China) dominated by
biofuel combustion. Although Young et al. (2012) acknowledged that
anthropogenic sources such as gasoline and diesel engines and secondary
processes in the atmosphere might be important contributors to atmospheric
HNCO, they did not include these sources/pathways in their study.
Furthermore, the grid resolution of the model used by Young et al. (2012) was too coarse (2.8
In this work, we performed laboratory experiments to measure HNCO emissions
from an SCR-equipped, modern-day, non-road diesel engine to test whether SCR
systems were a potential source of HNCO. To quantify HNCO emissions under
different operating conditions, we performed these tests under varying urea
injection rates (stoichiometric ratios of 0 to
The HNCO experiments were conducted on an engine dynamometer-mounted
(Midwest Inductor Dynamometer 1014A) 4-cylinder, turbocharged and
intercooled, 4.5 L, 175 hp, John Deere 4045 PowerTech Plus diesel engine;
this engine platform has been part of several earlier research studies
(Jathar et al., 2017; Drenth et al., 2014). The engine
consisted of a variable geometry turbocharger, exhaust–gas recirculation,
and electronically controlled high-pressure common rail fuel injection and
met non-road Tier 3 emission standards. A diesel oxidation catalyst (DOC,
John Deere RE568883) and diesel particulate filter (DPF, John Deere
RE567056) were retrofitted on the exhaust system to meet non-road interim
Tier 4 emission standards. Recently, Jathar et al. (2017) found that
the DOC
We built and installed a custom SCR system in the exhaust line that allowed
us to control and explore HNCO emissions as a function of varying urea
injection rates; the SCR system was installed downstream of the DOC
We performed a total of nine engine tests at three different engine loads (idle-like, 50 % load at intermittent speed, and 50 % load at rated speed) and with two different fuels (diesel and fatty acid methyl ester-based biodiesel). The three engine loads were (i) 45 Nm at 2400 RPM and 11 kW, (ii) 284 Nm at 1500 RPM and 45 kW, and (iii) 226 Nm at 2400 RPM and 57 kW, which corresponded to modes 4, 7, and 3 on the ISO 8178-4 C1 duty cycle, respectively. The ISO 8178 duty cycle is an international standard used for emission certification for non-road diesel engines. The diesel fuel was commercial, non-road, ultra-low-sulfur diesel (ULSD) and sourced locally while the biodiesel fuel (B100) was sourced from AG Processing Inc. (Sergeant Bluff, IA) and produced from soy; we have included the fuel certificate in the Supplement. Each test included a sweep across three to four urea injection rates for each engine load–fuel combination. We used commonly available diesel exhaust fluid – a 32.5 : 67.5 mixture of urea and water – as our urea source. After changing the urea injection rate, engine load, or fuel, the emissions were allowed to stabilize for approximately 10 min before values were recorded.
Raw exhaust was transferred to a Siemens five-gas analyzer using a
110
Background-corrected emission factors (EF) for CO, NO, NO
Emissions and emission ratios for gasoline, diesel, and biomass burning sources for the state of California for an average summer day in 2005.
The UCD/CIT is a regional chemical transport model that has been extensively
used to simulate the emissions, transport, chemistry, deposition and source
contribution of pollutants in the lower troposphere (Kleeman and Cass, 2001)
and evaluated against meteorological and gas- and particle-phase measurements
(Hu et al., 2012, 2015; Jathar et al., 2015, 2016). HNCO simulations were
performed for the state of California at a grid resolution of 24 km followed
by a nested simulation over the South Coast Air Basin (SoCAB) domain at a
grid resolution of 8 km from 15 July to 2 August 2005. Simulations were
performed for California since the state is home to the five most polluted
cities in the United States for ozone and particulate matter (American Lung
Association, 2016). We used the (i) CRPAQS
(California Regional PM
Primary emissions of HNCO were calculated by first determining a source-specific HNCO : CO ratio (see Table 1) and then combining them with source-specific, spatiotemporally resolved CO emissions to build an inventory for HNCO emissions. We used a ratio-based approach rather than an emission-factor-based approach for the following reasons. First, to our knowledge, there were no available HNCO emission factors for biomass burning. Second, there was large variability in the measured HNCO emission factors for both gasoline and diesel engines across different studies (this is illustrated in Fig. 3 and discussed in Sect. 3.2), which presumably arose from differences in engine sizes and technology. And finally, only a handful of sources have been characterized for HNCO emissions in previous studies and, in our view, the data may not be representative enough to develop an HNCO inventory using emission factors and fuel activity data. For the same reason, we assumed equivalence between on- and non-road engine sources in developing a source-specific HNCO : CO ratio.
We considered three sources for primary emissions of HNCO: (1) on- and
non-road diesel, (2) on- and non-road gasoline, and (3) biomass burning
(includes residential wood combustion). HNCO : CO ratios for diesel sources
were determined based on the range of measured HNCO : CO ratios found in
this and previous work. Findings from this work (see Sect. 3.1 and
3.2) suggest that none of the aftertreatment systems
deployed on our diesel engine affected HNCO emissions but the DOC
dramatically reduced CO emissions (factor of
The HNCO : CO ratio for the gasoline sources was determined as the ratio of the median HNCO to the median CO measured by Brady et al. (2014) for eight light-duty gasoline vehicles. The HNCO : CO ratio for biomass burning sources, which includes residential wood combustion, was based on an approximate fit to the laboratory and ambient data measured by Veres et al. (2010), Roberts et al. (2011), and Yokelson et al. (2013). We note that the previous study that simulated HNCO in a 3-D model developed global emissions of HNCO by using a source-specific ratio of HNCO with hydrogen cyanide (Young et al., 2012). The HNCO : CO ratios were then combined with source-specific, spatiotemporally resolved CO emissions to build source-resolved emissions for HNCO. HNCO from the three sources were tracked separately in the UCD/CIT model.
Emission/production factors for HNCO as a function of photochemical
age from Link et al. (2016). The fits, which parameterize the emissions of
the HNCO precursor and the reaction rate constant with OH, have been performed
in this work. The photochemical age is calculated assuming an OH
concentration of
Link et al. (2016) observed strong photochemical production of HNCO from a
diesel engine without any aftertreatment. This secondary HNCO source can be
attributed to photooxidation of amides (e.g., formamide, acetamide) and
potentially other reduced organic nitrogen compounds present in the diesel
exhaust, though the full suite of precursors, their reaction mechanisms and
their HNCO yields remains unknown. Hence, we make
simplifying assumptions to parameterize photochemical production of HNCO in our CTM
simulations. We assumed that diesel exhaust contains a single HNCO precursor
(
Fits and the parameters are shown in Fig. 1. The diesel and biodiesel data
were nearly identical and hence data from both fuels were used to determine
the engine-load-specific fits. The physical interpretation of the fit for
idle conditions is that for 1 kg of fuel burned,
Roberts et al. (2011) have argued that the gas-phase reaction of HNCO with
OH, heterogeneous reaction of HNCO on an aerosol surface, and photolysis of
HNCO are too slow to be relevant in the atmosphere and claimed that the only
relevant loss mechanism for HNCO was dry and wet deposition. Young et
al. (2012) investigated the influence of irreversible uptake of HNCO by
clouds in a global model and found that this loss mechanism competed with dry
deposition only when the cloud
Emission factors for
In Fig. 2, we plot emission factors for NO
Similar to earlier work (Wentzell et al., 2013; Heeb et al., 2012, 2011; Suarez-Bertoa and Astorga, 2016), we observed HNCO when no
NH
Very few studies have investigated HNCO emissions from diesel engines and,
before this work, only four studies
have examined HNCO emissions from an SCR-equipped diesel engine. In Fig. 3,
we compare HNCO emission factors for diesel fuel with all earlier work
involving diesel engines: Link et al. (2016), Wentzell et al. (2013), Heeb et
al. (2011, 2012), and Suarez-Bertoa and Astorga (2016). We also compare our
results to the average HNCO emissions from eight light-duty gasoline vehicles
measured by Brady et al. (2014). For this work, the mean and the standard
errors were calculated using the HNCO data across all NH
Emission factors for HNCO from this work compared to literature data. US06, HWFET, FTP75, ISO8178, WLTC, and UC are vehicle/engine drive cycles. S-B et al. refers to the Suarez-Bertoa and Astorga (2016) study.
There are several interesting findings of note. First, the emission factors for HNCO from this work were nearly identical to those measured by Link et al. (2016). Since both studies were performed on the same engine and used a similar CIMS instrument, the HNCO was mostly likely produced in the engine cylinder and was unaltered by the DOC and DPF. Second, primary emissions of HNCO and its precursors based on the work of Link et al. (2016) were deemed plausible when compared against emissions of total unburned hydrocarbons; i.e., HNCO and its precursor at idle conditions were less than 0.2 and 1 % of the total hydrocarbon emissions while HNCO and its precursor at 50 % load conditions were less than 0.4 and 0.9 % of the total hydrocarbon emissions. Third, the emission factors for HNCO from the engine used in this work (with or without the aftertreatment devices) were much higher (factor of 10–100) than those measured by Wentzell et al. (2013). Wentzell et al. (2013) used a CIMS instrument similar to that used in this study and therefore the differences could not be attributed to the instrumentation. Link et al. (2016) suggested that the large differences between their study and the Wentzell et al. (2013) study could reflect the variability found in emissions between non- and on-road diesel engines and steady and transient drive cycles. However, when compared using the HNCO : CO ratio, there was much less variability in the ratio between this work and two of the drive cycles examined by Wentzell et al. (2013) (see Table S1), which could suggest that our non-road engine, on account of being larger than the Wentzell et al. (2013) engine, simply produced more HNCO and more CO but yielded the same HNCO : CO ratio. This observation led us to assume (in Sect. 3.2) equivalence between non-road and on-road diesel engines as well as to develop emission inventories for HNCO based on the HNCO : CO ratio rather than through the use of emission factors. To test our findings and assumptions, we recommend that future studies focus on testing a diverse suite of diesel engine sizes under a wide range of steady and transient engine loads.
Fourth, the emission factors for HNCO from this study
(31–56 mg kg fuel
Finally, the emissions of HNCO on a fuel-burned basis from this work and Link
et al. (2016) were more than an order of magnitude larger than the average
HNCO emissions from the suite of light-duty gasoline vehicles tested by Brady
et al. (2014) but similar to those measured by Suarez-Bertoa and
Astorga (2016) for a range of light-duty gasoline vehicles. Suarez-Bertoa and
Astorga (2016) have argued that their emission factors for HNCO were higher
than those measured by Brady et al. (2014) because of differences in sampling
tailpipe versus diluted emissions. Assuming the Brady et al. (2014) data are
more atmospherically relevant, diesel engines, regardless of their use of
aftertreatment devices, might be a much larger source of HNCO than
catalytic-converter-equipped gasoline engines despite higher gasoline
consumption in the United States compared to diesel (
Predictions of 14-day averaged, ground-level concentrations of HNCO from the
CTM simulations are mapped for the state of California in Fig. 4a–b. The low
and high results are from two simulations that used two different primary
emissions and photochemical production parameterizations for diesel engines
(refer to Table 1 for details) and capture the uncertainty in modeling HNCO
contributions from diesel-powered sources. Inland concentrations of HNCO
between the low and high simulations varied significantly but never exceeded
110 pptv and were at least an order of magnitude lower than the 1 ppbv
level proposed by Roberts et al. (2011). Roberts et al. (2011) argued that a
1 ppbv HNCO concentration would translate to a 100
We individually tracked the source/process-level contributions of HNCO in
the CTM simulations and found that diesel use was the dominant source of
HNCO in SoCAB. Based on the low and high simulations, diesel sources
accounted for 55–92 % while gasoline sources accounted for 8–41 % of the
HNCO in SoCAB, with a very small contribution (1–4 %) from biomass burning
sources. The signature of a larger contribution of HNCO from biomass burning
sources can be seen in Fig. 4a in more remote locations of California,
e.g., northwest corner of California, north of Sacramento. Despite the
strong photochemical production observed by Link et al. (2016) in
laboratory experiments, secondary production of HNCO from precursors in
diesel exhaust only accounted for 9–11 % of the total HNCO. The most
likely explanation for this small contribution was that the in-basin
exposure of HNCO precursors to OH radicals was too small to produce a lot of
secondary HNCO. In fact, the slow secondary production of HNCO can be
visualized in Fig. 1, where significant enhancements in HNCO were only
observed after
Furthermore, we investigated the sensitivity of model predictions to dry
deposition by using NO as the surrogate to model dry deposition of HNCO; NO
has a much slower dry deposition lifetime (
To evaluate the HNCO predictions from our CTM simulations, we compared model predictions from the 8 km simulation to two datasets of HNCO measurements in urban areas: (i) observations reported by Roberts et al. (2014) at the Pasadena ground site during the California Research at the Nexus of Air Quality and Climate Change (CalNex) study in May–June 2010 and (ii) observations reported by Wentzell et al. (2013) in Toronto in September–October 2012. Since the model simulations were not for the same time period (in the case of CalNex) or the same location (in the case of Toronto), we present the comparisons in Fig. 4c by regressing concentrations of HNCO against those of benzene. We chose benzene because Roberts et al. (2014) had developed an emission ratios for HNCO with respect to benzene and Wentzell et al. (2013) had previously found ambient HNCO concentrations to vary linearly with benzene concentrations.
Model predictions from the low simulation seemed to agree with the nighttime
observations of Roberts et al. (2014) and validate the primary
parameterizations and deposition scheme used in the low simulation. Roberts
et al. (2014) have argued that the diurnal differences in the observations
imply a daytime photochemical source of HNCO, where in the mid-afternoon
secondary processing accounts for 40 % of the total HNCO. Hence,
agreement between the model predictions from the high simulation and the
daytime observations of Roberts et al. (2014) should not be construed as a
validation of the inputs for that simulation since the high simulation
predicts a small contribution (
We performed laboratory experiments on an SCR-equipped modern day diesel
engine to measure emissions of isocyanic acid as a function of
varying urea injection rates, engine loads, and fuels. We found no evidence
that the SCR or the other aftertreatment devices (diesel oxidation catalyst
and diesel particle filter) were a source of tailpipe HNCO. We argue that
the HNCO from diesel engines was likely produced inside the engine cylinder
during fuel combustion. This finding is not completely new. Chemical
kinetics models (Mansour et al., 2001), model systems with propane
(Nelson and Haynes, 1994), and engine tests without aftertreatment
devices (Heeb et al., 2011) have previously shown that HNCO (and other
reduced nitrogen-containing compounds) can be produced during combustion in
the presence of NO
Amides such as formamide are known precursors of HNCO (Borduas et al., 2014) and might be part of amide emissions from various types of combustion sources that lead to atmospheric production of HNCO. Recent studies have noted that other forms of reduced organic nitrogen compounds can be oxidized to form HNCO, suggesting that molecules other than amides emitted from diesel exhaust may also be HNCO precursors (Borduas et al., 2016a, b). Fits to the data from Link et al. (2016) suggest that the emissions and the rate of photochemical production of HNCO attributed to diesel sources may not be sufficient to contribute significantly to ambient concentrations of HNCO in urban environments. These precursors, however, might be important in controlling HNCO concentrations in remote/rural environments but based on the results from this study might be deemed too low to be of any concern from a health perspective. Our CTM predictions suggest that the daily-averaged precursor concentrations in urban environments are large enough (50–250 pptv in Los Angeles; see Fig. S3 for precursor concentrations in California) to provide impetus for ambient studies to design and deploy instruments to measure these precursors. Finally, it is possible that sources other than diesel engines (e.g., gasoline engines, biomass burning, agricultural burning) also emit precursors of HNCO and hence need to be studied in the future both in terms of identifying and quantifying the precursors of HNCO and measuring their potential to form HNCO.
Using our experimentally determined emission factors, we used a CTM to simulate ground-level concentrations and source (gasoline, diesel, biomass burning) and process (primary, secondary) contributions to HNCO in California. The predicted HNCO concentrations in Southern California were roughly similar to those measured at Pasadena in 2010, Toronto in 2012, and La Jolla in 2012. A detailed comparison at Pasadena highlighted missing precursors/pathways for photochemical production of HNCO during the daytime. The comparisons also implied that diesel engines (and possibly gasoline engines) are large sources of HNCO in urban areas. In the simulations, daily-averaged HNCO concentrations never exceeded 110 pptv and were an order of magnitude below the 1 ppbv level that Roberts et al. (2014) have proposed could result in human health effects. If we assume that the HNCO–benzene regression from our work holds for other parts of the world, benzene concentrations exceeding 7 ppbv would be associated with 1 ppbv levels of HNCO; we expect benzene and HNCO to correlate only in source and/or urban regions and the regression may not be applicable for remote/rural locations since HNCO and benzene may have very different atmospheric lifetimes. We should note that the 1 ppbv threshold is a rough estimate and we see a need for epidemiological and/or toxicological studies that would better inform that estimate. Emissions from biomass burning sources in the winter combined with a strong likelihood for temperature inversions could lead to higher HNCO concentrations in the winter and need to be explored using both measurements and air quality modeling.
Summary data from the laboratory experiments and hourly-
and daily-averaged data from the CTM simulations are archived at
The authors declare that they have no conflict of interest.
We would like to thank Daniel Olsen for inputs on experimental design, Kirk Evans for technical support, and undergraduate researcher Liam Lewane for engine test support during the study. We would also like to thank DCL International Inc. for donating SCR catalysts for our study and Ricardo Suarez-Bertoa and Covadonga Astorga for sharing HNCO emissions data from their published study. Delphine K. Farmer acknowledges an Arnold and Mabel Beckman Young Investigator Award for funding the laboratory HNCO measurements. Edited by: John Liggio Reviewed by: two anonymous referees