Ship emissions measurement in the Arctic from plume intercepts of the Canadian Coast Guard Amundsen icebreaker from the Polar 6 aircraft platform

Decreasing sea ice and increasing marine navigability in northern latitudes have changed Arctic ship traffic patterns in recent years and are predicted to increase annual ship traffic in the Arctic in the future. Development of effective regulations to manage environmental impacts of shipping requires an understanding of ship emissions and atmospheric processing in the Arctic environment. As part of the summer 2014 NETCARE (Network on Climate and Aerosols) campaign, the plume 5 dispersion and gas and particle emission factors of emissions originating from the Canadian Coast Guard Amundsen icebreaker operating near Resolute Bay, NU, Canada have been investigated. The Amundsen burnt distillate fuel with 1.5 wt % sulfur. Emissions were studied via plume intercepts using aircraft measurements, an analytical plume dispersion model, and using the FLEXPART-WRF Lagrangian particle dispersion model. The first plume intercepts by research aircraft were carried 10 out on 19 July 2014 during the operation of the Amundsen in the open water. The second and third plume intercept measurements were carried out on 20 and 21 July 2014 when the Amundsen had reached the ice edge and operated under icebreaking conditions. Typical of Arctic marine navigation, the engine load was low compared to cruising conditions for all of the plume intercepts. The measured species included mixing ratios of CO2, NOx, CO, SO2, particle number concentration (CN), 15 refractory Black Carbon (rBC), and Cloud Condensation Nuclei (CCN). The results were compared to similar experimental studies in mid latitudes. 1 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-1032, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 25 January 2016 c © Author(s) 2016. CC-BY 3.0 License.


Introduction
International shipping is responsible for approximately 3.3 % of global CO 2 emissions, 5 to 8 % of global anthropogenic SO 2 emissions, and 2 % of global Black Carbon (BC) emissions (Lack and Corbett, 2012).The regulations for air pollutants released by ships are set by the International Convention for the Prevention of Pollution from Ships (MARPOL) within the International Maritime Or-35 ganization (IMO) accessible at http://www.imo.org/en/OurWork/Environment/PollutionPrevention/Pages/Default.aspx.In addition, specific sensitive regions are subject to more stringent limits for Emissions Control Areas (ECAs), such as those in effect for the Baltic Sea, the Mediterranean Sea, and the Caribbean Region.In the high Arctic, including Canadian waters, there is currently no ECA established, despite the very sensitive nature of the Arctic environment and ecosystems.At the same 40 time, the decreasing sea ice and increasing marine navigability in the shipping season have already increased annual traffic in the Canadian Arctic in the recent decades (Pizzolato et al., 2014).Future projections in Arctic ship traffic also suggest increased emissions by mid century (Corbett et al., 2010a;Winther et al., 2014).Development of effective regulations require an understanding of observed ship emissions and processing in the Arctic environment.

Research objectives
In this study we use measurements from airborne plume intercepts to estimate emissions factors for the Amundsen ship, while operating in the Arctic and burning low sulfur fuel, for gaseous and particle pollutants.In addition, we study the geometrical evolution of the Amundsen's plume in 100 the Arctic marine boundary layer.We compare these observations to other similar studies in mid latitudes.The first plume measurement was carried out on 19 July 2014 during the operation of the CCGS Amundsen in the Lancaster Sound of the Northwest Passage (74 • ,18 N, 83 • ,54 W).The second and third plume measurements were carried out on 20 and 21 July 2014 after the CCGS Amundsen reached the ice edge and operated under ice conditions, north of Somerset Island, less 105 than 50 km from Resolute Bay.These measurements provide differences in plume characteristics between operation under open water conditions as well as sea ice conditions in the Arctic.

Specifications of Amundsen icebreaker
The Amundsen (IMO: 7510846) belongs to the Canadian Coast Guard fleet with full specifications 110 available at http://www.ccg-gcc.gc.ca/Fleet/Vessel?vessel_id=3.It is an Arctic Class 3 vessel, 98.2 m long, with gross tonnage of 5911.0 t, and maximum speed of 16.0 kts.The propulsion is provided by a diesel electric AC/DC system with 6 main Alco M251F engines of total power 13200 kW.It has 3 Alco MLW251F generators and a Caterpillar 398 emergency generator.During the campaign, Amundsen burned marine distillate fuel that contained 1.5 wt % sulfur content (ISO 8217 2010 DMA 115 Fuel Standard).

Airborne measurements
The airborne instrument platform was the Polar 6 aircraft, a DC-3 converted to a Basler BT-67, owned and operated by the German Alfred Wegener Institute -Helmholtz Center for Polar and Marine Research (Fig. 1) (Leaitch et al., 2015).Below, experimental methodologies for the measure-120 ments of state parameters and meteorology, gas phase, and particle phase pollutants are presented.The sampling frequency is greater than 40 Hz, but in this study a sampling frequency of 1 Hz is used.

Gas phase measurements
Trace gas CO 2 measurement was based on infrared absorption using a LI-7200 enclosed CO 2 /H 2 O 135 Analyzer from LI-COR Biosciences GmbH.In-situ calibrations during the flight were performed on a regular time interval of 15 to 30 min using a NIST traceable calibration gas with a known CO 2 concentration at atmospheric levels.The uncertainty for the measurement of CO 2 is 0.3 ppmv relative to the standard.Trace gas CO was measured with an Aerolaser ultra fast carbon monoxide (CO) monitor model AL 5002 based on VUV-fluorimetry.The same in-situ calibrations during 140 inflight were performed.The calibrations and zero measurements allowed for corrections of instrument drifts increasing the stability and accuracy of the instrument, thus leading to an uncertainty of ±2.3 ppbv relative to the standard.
Trace gas NO x measurement was based on chemiluminescence using a Thermo Scientific 42i NO−NO 2 −NO x analyzer with a time resolution of 1 s and an uncertainty of 0.4 ppbv.Trace gas 145 SO 2 measurement was based on UV Fluorescence light-scattering using a Thermo Scientific Model 43i-TLE Enhanced Trace Level SO 2 analyzer with a time resolution of 1 s and an uncertainty of 1 % of reading or 0.2 ppbv, whichever is greater.Trace gas O 3 measurement was based on UV photometry using a Thermo Scientific 49i analyzer with a time resolution of 10 s and an uncertainty of 0.2 ppbv.For simplicity trace gas mixing ratio units of [ppbv] is presented as [ppb] hereafter.(Cai et al., 2008).The time resolution was 1 s and the measurements are referred to as UHSAS hereafter.
Cloud Condensation Nuclei (CCN) concentrations were measured by a DMT CCN Model 100 counter operating behind a DMT low pressure inlet at a reduced pressure of approximately 650 hPa 160 and a nominal water supersaturation of 1 %.The effective supersaturation at 650 hPa was determined to be approximately 0.6 % and was held constant throughout the study to allow for more stability of measurements, improved response, and to examine the hygroscopicity of smaller particles.The time resolution was 1 s and the measurements are referred to as CCN hereafter.
Extensive calibrations and evaluations for CPC, UHSAS, and CCN measurements were performed 165 in the laboratory prior to integration of the instruments on the aircraft and again with instrumentation in the aircraft at Resolute Bay.Full discussions can be found in the study by Leaitch et al. (2015).
Particle size distribution for particle diameters greater than 0.25 µm was measured using a Sky Optical Particle Counter (OPC model 1.129).Measurements were based on 90 • scattering light and a time resolution of 6 s.The accuracy is ±3 % at 1 sigma confidence.These measurements are 170 referred to as OPC hereafter.
The refractory black Carbon (rBC) was measured using a single particle soot photometer (SP2) from DMT Boulder.The SP2 (Schwarz et al., 2010) is an instrument able to evaluate individual aerosol particles for the rBC mass content, size and mixing state based on the laser-induced incandescence method and can gather information on the scattering part of the aerosol ensemble.The time 175 resolution was 1 s and the measurements are referred to as SP2 hereafter.
Particle sampling is described in full detail by Leaitch et al. (2015) and was performed so that the efficiency of particle transmission to instruments would be close to 100 % for particles from 20 nm to 1 µm in diameter.

Power law model for plume growth 180
The methodology of von Glasow et al. (2003) describes plume dispersion with a power law which models plume dimensions in horizontal (w pl ) and vertical (h pl ) directions.
with w 0 and h 0 being plume dimensions at reference time (t 0 = 1 s) and α and β being plume expansion rates in the horizontal and vertical directions.Fitted values for expansion rates are provided 185 in the literature for mid latitude marine boundary layers (von Glasow et al., 2003;Petzold et al., 2008); however, it remains to be verified if expansion rates are similar or different over the Arctic marine boundary layer.The power law describes plume cross-section with a semi-elliptic shape with Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-1032, 2016 Manuscript under review for journal Atmos.Chem.Phys.Published: 25 January 2016 c Author(s) 2016.CC-BY 3.0 License.area A pl = π 8 w pl h pl .It is assumed that plume expansion in the vertical direction is inhibited when it reaches the top of the marine boundary layer, where subsequent expansion only continues in the 190 horizontal direction.
A convenient and practical way to fit for plume expansion rates is to intercept a portion of the plume and measure the mixing ratio of a chemically inert species in the plume such as CO 2 .Assuming uniform dilution of such species in the plume, it is possible to derive a relationship between the species mixing ratio in the plume (c pl ) and expansion rate coefficients (α and β), where c bgd is the background mixing ratio of the species and γ is either α+β for plumes not reaching marine boundary layer or α for plumes that evolve after reaching the top of the marine boundary layer.Then γ is the expansion rate and m = −γ is the slope of the linear relationship.The reference time for this calculation is independent from the reference time introduced earlier.Since mixing in 200 real plumes is not uniform, time or cross sectional-averaging of the airborne-measured mixing ratio and multiple measurements at various distances from the source are necessary to arrive at a better estimate for the plume expansion rate.

Estimation of plume age
Plume age can be estimated by the aircraft measurements.For this, plume intercepts are first mapped 205 on a latitude/longitude plot.This provides a scatter plot to which a plume center line is fitted with a high order polynomial.The wind measurements on board of the aircraft closest to each point on the center line are then used to estimate wind velocity along the fitted plume center line.This methodology enables plume age estimation at each intercept by calculating the time it takes for the plume center line to reach the nearest location to the intercept using the following formula 210 which is a line integral starting from the plume center line origin (l = 0) to the nearest plume intercept on the center line (l = L).U (l) is the estimated horizontal wind speed along the plume center line.

215
A common method to calculate emission factors (EF) in [g kg − diesel −1 ] is the net peak area method (Alföldy et al., 2013) using the CO 2 balance concept (Hobbs et al., 2000).For a pollutant measurement in units of [ppb], the molecular weights of carbon and a gaseous pollutant species of interest are considered.Given the carbon mass percent in diesel fuel (87 ± 1.5 %; (Cooper, 2005)), the emissions factor for species X can be expressed as, where C() represents the mixing ratio of species above background levels integrated over time for an entire peak and M W stands for molecular weight, which for carbon is 12. EF can be estimated at a reference customary plume age or as an average for all plume encounters.
For pollutant measurement in units of mass concentration (e.g.[µg m −3 ]), EF can be estimated us-

225
ing the same methodology, however, the molecular weight of the pollutant is not necessarily needed since the measurement in units of mass per volume is already available (Lack et al., 2009), where the constant 1620 [g µg −1 m 3 ppb kg − diesel −1 ] accounts for the same carbon mass percent in diesel fuel.For particle emissions in units of [cm −3 ], the emissions factor can be calculated using 230 (Lack et al., 2009), where the constant 1.62 × 10 15 [cm 3 ppb kg − diesel −1 ] accounts for the same carbon mass percent in diesel fuel.
If a modal emissions factor with units of [g kWh −1 ] is reported, which applies to both gaseous 235 and particle phases, it is possible to convert it to units of [g kg − diesel −1 ] if emissions factor for Similarly, if a modal emissions factor with units of [kWh −1 ] is reported, which applies to number of particles, it is possible to convert it to units of [kg The calculated EF for conserved pollutants, such as CO 2 , is constant and not a function of plume age.
However, for other pollutants it may increase (production) or decrease (consumption) as a function of plume age.Due to limited number of plume intercepts in this study, we compute average emissions 245 factors for all plume intercepts.

FLEXPART-WRF plume dispersion modeling
In order to study, the dispersion of ship emissions in the Polar boundary layer, we use the FLEXPART-WRF model (Brioude et al. (2013), website: flexpart.eu/wiki/FpLimitedareaWrf),a Lagrangian particle dispersion model based on FLEXPART (Stohl et al., 2005).FLEXPART-WRF is driven by wind speeds measured (Figure 2b).This is also shown by the variable wind speeds and directions in the region of the flight in Figure 3b.On 21 July northwesterly winds throughout Lancaster sound resulted in plume sampling to the southeast of the ship, with consistent wind speeds (but lower than on 19 July 2014) during the plume sampling (Figure 2c).

285
We also characterize boundary layer dynamics using balloon soundings launched from the ship at the times of the flights for plumes 2 and 3 (Figure 4).For plume 1, there was no balloon sounding, therefore we show only the WRF model results for comparison.The measurements and the model are in good agreement, noting that the model under predicted wind speeds below 100 m on 21 July 2014 compared to the measurements.This is also seen in the flight track on 21 July 2014 (Figure 290 2c).We also, however, note that WRF model does perform better than the ECMWF analysis (wind speed and wind direction) for this flight.The boundary layer is statically stable and the boundary layer height is calculated from measurements to be 387 m and 177 m for plume 2 and 3 studies, respectively, using the method of bulk Richardson number developed by Mahrt (1981) and later used by Aliabadi et al. (2015a).Vertical gradients in potential temperature and wind speed show that 295 the emissions are predicted to be mixed into a shallow boundary layer on all three days, both during operations in the open water (plume 1) and within sea ice (plumes 2 and 3).

Ship operating conditions
It is known that both ship speed and engine load influence total fuel burned and emission factors.
For the Amundsen, ship speed is not directly correlated with engine load for two reasons.First, the Amundsen operates on a diesel-electric system, which could provide propulsion power using electricity while the engines are off or operating at partial load.Second, because of the specifics of ships operating in the Arctic within sea ice, even during stationary conditions, the engine may be running to power ice breaking operations.The average ship speed during plume 1, 2, and 3 studies were 3.23±0.25kts, 1.31±1.92kts, and 0.09±0.30kts, respectively.The variation in ship speed is 305 calculated using one standard deviation, noting that both plume 2 and 3 studies involved ice breaking.

FLEXPART-WRF ship plume modeling
In order to show the relationship between the emissions from the ship (plumes) on different days and the flight pattern, we use FLEXPART-WRF partial columns and vertical cross sections.Given The predicted vertical distribution of emissions along and across the plumes are shown in Figure 7.The model indicates that the ship emissions are predicted to be below 300 meters when the ship was operating in the open water (plume 1) and are predicted to stay in the lowest 100 meters when the 320 ship was operating in sea ice (plumes 2 and 3).The vertical cross sections across the plumes (Figure 7 panels marked e -across plume) show that the horizontal distribution of emissions is predicted several kilometers across during plume crossing.It is also worth noting that all of the plume crossings used for emissions factor calculation and analysis were below 90 m (Figure 6), corresponding to the most concentrated portion of the ship plume as predicted by FLEXPART-WRF (Figure 7).The exact 325 properties of the ship plumes are determined by the combination of the meteorological conditions, emissions injection location (horizontal and vertical extent), and the ship movements.This analysis also shows that the predicted plumes mix slowly with the background air in the strongly stable Arctic boundary layer, with implications for the fate of emissions and plume processing.

Ship plume pollutant identification 330
Plume intercepts have been identified using the methodology of Petzold et al. (2008) where a statistically significant change in mixing ratio of a non-decaying gaseous pollutant with respect to background has been observed.Figure 8 shows an example time series where pollution peaks in the plume are evident.This time series is used to identify the location and timing of ship plume crossings (shown in Figure 6), which is also referred to as an excess or peak event.To identify plume crossings 335 the NO x mixing ratio with a threshold of 2 ppb has been used, which was preferred over CO 2 due to unpredictable background variations in the CO 2 mixing ratio.In this method, the background for NO x mixing ratio was computed by averaging three consecutive measurements before and after the threshold.Once time stamps for NO x peak events were identified, all other pollutant peaks were identified using these times, without the need for a threshold.A time shift between peak events was 340 expected between the reference instrument (NO x ) and any other instrument since they sampled air at different locations on the sampling line.This shift was identified and corrected by maximizing the coefficient of determination (R 2 ) for the 1-1 mixing ratio plots between a pair of instruments.

Analytical model of ship plume expansion
Using airborne meteorological and CO 2 mixing ratio measurements, the power law plume expansion 345 model (eq. 1 and e.q. 2), and the estimated plume age (eq.3), the plume geometrical evolution can be explained for all of the plumes.In this calculation, the vertical variations in wind speed and direction are accounted for.Using the methodology in section 2.2.5 a plume age could be assigned to n = 6 data points for plume 1, n = 7 data points for plume 2, and n = 18 data points for plume 3 studies.Figure 9 shows the measured expansion rates for these data points.The expansion rate 350 is the magnitude of slope for the fitted lines (γ = α + β), and it is calculated as γ = 0.75 ± 0.80, 0.93±0.37,and 1.19±0.39for plumes 1, 2, and 3, respectively.The uncertainty is computed for the regression analysis as one standard deviation.This compares reasonably well to values reported in the literature for mid latitudes.3.6 Changes in gas mixing ratios and particle concentrations 3.6.1 Gas Pollutants 360 Figure 10 shows the scatter plot for excess gas pollutants versus excess carbon dioxide.In all three instances, excess carbon monoxide and nitrogen dioxide correlate with excess carbon dioxide, while ozone titration accounts for a negative correlation between excess ozone and excess carbon dioxide.
Given the lower detection limit of the instrument, there was no trace of SO 2 in the plume as measured by the aircraft.It has been verified by SO 2 measurements on-board of the ship that the mixing ratios 365 were below 2 ppb, indicating that the exhaust after treatment on the Amundsen effectively removes this species (Wentzell, 2015, Personal Communication).
Table 1 shows the results for linear regression analysis for excess gas pollutants versus excess carbon dioxide.The only insignificant coefficient of determination belongs to excess carbon monoxide in plume 1.The regression slope (b) for excess oxides of nitrogen in plume 1 is a factor of 2 less than 370 plumes 2 and 3, attributed to ice-breaking conditions, and hence higher engine temperature (but not necessarily engine load), during plume 2 and 3 studies.

Particle Pollutants
Figure 11 shows the scatter plot for excess particle concentrations versus excess carbon dioxide.A correlation is noticeable for all instruments.The SP2 instrument was not functional during plume 375 3 study.Table 2 shows the results for linear regression analysis for excess particle concentrations UHSAS, and CCN concentrations are factors of 5, 4-10, 2-3, 2-5 higher than plumes 2 and 3.This may be related to possible higher engine load (also vessel speed), but lower engine temperature according to section 3.6.1,for this plume.

Emissions factors
Emissions factors (EF) in the literature are reported in different ways.Some studies report EF for one ship or a fleet of ships operating under various engine loading conditions or fuel types (Petzold et al., 2008;Lack et al., 2011;Khan et al., 2012a;Alföldy et al., 2013).Another common approach is to group EF based on vessel gross tonnage in HSD: high speed diesel < 5000 t, MSD: medium 385 speed diesel 5000−30000 t, or SSD: slow speed diesel > 50000 t categories (Lack et al., 2008(Lack et al., , 2009;;Williams et al., 2009;Diesch et al., 2013;Buffaloe et al., 2014).The other approach is to report EF for a single ship operating on specific fuel type as a function of engine load (Agrawal et al., 2008;Petzold et al., 2010Petzold et al., , 2011;;Khan et al., 2012b;Cappa et al., 2014).

Gas pollutants 390
Figure 12 and Table 3 show emissions factors for NO x in this study in comparison to other studies in the literature.EF NOx is expected to increase for engines operating at higher temperatures (thermal NO x ) (Sinha et al., 2003;Diesch et al., 2013;Cappa et al., 2014).Higher engine loads have been shown to increase EF NOx (Agrawal et al., 2008;Petzold et al., 2011;Khan et al., 2012b;Cappa et al., 2014).Increasing gross tonnage has also been shown to result in higher EF NOx (Williams et al., 2009; 395 Diesch et al., 2013).EF NOx in this study is in good agreement with other studies particularly for low engine loads and HSD-MSD vessel categories.However there is an increase in EF NOx by a factor of 3 for plumes 2 and 3 compared to plume 1.This suggests that icebreaking during these two plumes resulted in higher engine temperatures that correspondingly increased EF NOx .
Figure 13 and Table 4 show emissions factors for CO in this study in comparison to other studies 400 in the literature.Emissions factors for carbon monoxide (EF CO ) are expected to drop with increasing ship engine load (speed) (Agrawal et al., 2008;Moldanová et al., 2009;Agrawal et al., 2010;Petzold et al., 2011;Jalkanen et al., 2012;Khan et al., 2012b;Cappa et al., 2014).EF CO in this study is in good agreement with other studies for which the vessel speed is very slow.

Particle pollutants
405 Figure 14 and Table 5 show emissions factors for rBC in this study in comparison to other studies in the literature.It is important to realize that estimates for BC measurements significantly depend on the methodology used, so caution should be used in interpreting data.For example, refractory derived SP2 measurements of BC underestimates BC emissions by a factor of about 2 relative to other techniques, likely due to methodological limiations, such as the limited range for particle 410 detection (60 nm < d p,V ED < 300 nm) (Buffaloe et al., 2014;Cappa et al., 2014), and so where Figure 15 and Table 6 show emissions factors for total particle count in this study in comparison to other studies in the literature.The caveat in this comparison is the difference in lower size limit for CPC measurements.For this purpose, we have provided lower size limits for other studies.Regardless, EF CPC for plume 1 is higher by a factor of 4 compared to plumes 2 and 3.This suggests 420 that higher engine loading results in higher EF CPC .This is in agreement with studies by Petzold et al. Figure 16 and Table 7 show emissions factors for cloud condensation nuclei in this study in comparison to other studies in the literature.The caveat in this comparison is the difference between 425 supersaturation (SS) for CCN measurements.For this purpose SS is provided for other studies (see Table 7).EF CCN for the Amundsen is comparable to other studies at low engine load conditions and similar SS (Petzold et al., 2010;Cappa et al., 2014).The low EF CCN can be justified by the fact that there was no measurable SO 2 in the plumes, given the lower detection limit of our instrument, due to effective exhaust after treatment to remove this species.This suggests why CCN levels are reduced 430 due to lack of sulphates and that the reduced CCN activity limits the ability of particles to influence clouds.

Conclusions and future work
In an effort to understand ship emissions and processing in the Arctic environment, the plume dispersion and emission factors from the Canadian Coast Guard Amundsen icebreaker were quantified The calculated analytical expansion rates were γ = 0.75 ± 0.80, 0.93 ± 0.37, and 1.19 ± 0.39 for plumes 1, 2, and 3, respectively.These are lower than observations in mid latitudes.All emission factors were in agreement with other observations at low engine loads in mid latitudes.Icebreaking appeared to increase the NO x emission factor from EF NOx = 22.3 ± 8.0 to 57.8 ± 11.0 and 450 65.8 ± 4.0 g kg − diesel −1 for plumes 1, 2, and 3, possibly due to high engine temperatures.The CO emission factor was EF CO = 6.4 ± 11.7, 6.8 ± 2.2 and 5.0 ± 1.0 g kg − diesel −1 for plumes 1, 2, and 3.The rBC emission factor was EF rBC = 0.20 ± 0.04 and 0.25 ± 0.12 g kg − diesel −1 for plumes 1 and 2. The CN emission factor was reduced while icebreaking from EF CPC = 1.96 ± 0.41 to 0.43 ± 0.11 and 0.47 ± 0.04 × 10 16 kg − diesel −1 for plumes 1, 2, and 3.The CCN emission fac-
The difference in plume expansion rate compared to mid latitude observations is attributed to unique physics of the Arctic boundary layer, which is characterized by reduced turbulent mixing due to the thermally stable boundary layer.In addition, ship operation at partial engine load and 460 icebreaking mode contribute to different emission factors compared to cruising conditions.
One limitation of this study was that the Amundsen plume was not intercepted at higher engine loads near cruising conditions.Future studies should measure the emission factors and plume geometrical evolution under such conditions to provide a more complete understanding of plume chemistry and physics over the Arctic marine boundary layer.
2.2.1 State parameters and meteorological measurements State parameter and meteorological measurements are performed by an AIMMS-20 instrument, manufactured by Aventech Research Inc., Barrie, Ontario, Canada.The instrument consists of three modules.The Air Data Probe (ADP) measures the three-dimensional, aircraft-relative flow vector 125 (true air speed, angle-of-attack, and sideslip).The temperature and relative humidity sensors are located in the aft section of the probe for protection.A three-axis accelerometer pack facilitates direct turbulence measurement.The Inertial Measurement Unit (IMU) consists of three gyros and three accelerometers providing the aircraft angular rate and acceleration.A GPS module provides the aircraft 3D position and inertial velocity.Horizontal and vertical wind speeds are measured with accuracies 130 of 0.50 and 0.75 m s −1 , respectively.The accuracy and resolution for temperature measurement are 0.30 and 0.01 • C. The accuracy and resolution for relative humidity measurement are 2.0 and 0.1 %.
phase measurements Particle number concentrations greater than 5 nm diameter were measured with a TSI 3787 waterbased ultrafine Condensation Particle Counter (CPC), sampling at a flow rate of 0.6 Lmin −1 and a time resolution of 1 s.These measurements are referred to as CPC hereafter.Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2015-1032,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: 25 January 2016 c Author(s) 2016.CC-BY 3.0 License.Aerosol particle size distributions from 70 nm to 1 µm were measured by a Droplet Measurement 155 Technology (DMT) Ultra High Sensitivity Aerosol Spectrometer (UHSAS) that uses scattering of 1054 nm laser light to detect particles Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2015-1032,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: 25 January 2016 c Author(s) 2016.CC-BY 3.0 License.
Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2015-1032,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: 25 January 2016 c Author(s) 2016.CC-BY 3.0 License.(8) 250meteorology from the Weather Research and Forecasting (WRF) Model(Skamarock et al., 2005), with the specifics of the WRF run for NETCARE provided inWentworth et al. (2015).Here we ran FLEXPART-WRF in forward mode to study plume dispersion from the Amundsen.Running FLEXPART-WRF in forward mode is useful for studying the specific plume structure and emissions location for the case of a single moving point source (e.g. a single ship) involving complex move-255 ments (moving ship location with time) within a complex and changing meteorological situation.FLEXPART-WRF was run in forward mode using the known ship location.Particles were released each minute along the ship track using a source extending 100 m vertically and horizontally centered on the ship location, from 17 July 2014 00:00 UTC to 22 July 2014 00:00 UTC.An arbitrary emissions source strength was assumed for the model run (mass of particles emitted) and considered to 260 be constant in time for the duration of the run.FLEXPART-WRF output was saved on a grid approximately 1 km × 1 km (resolution of 0.01 • Latitude × 0.05 • Longitude) in order to obtain results on a similar spatial scale as the plume sampling.3 Results and discussion 3.1 Meteorological context 265 Plume intercepts in the three consecutive days are referred to as plume 1 (19 July 2014), plume 2 (20 July 2014), and plume 3 (21 July 2014) studies.The flights were planned in advance using WRF andFLEXPART-WRF forecasts (not shown) so that the aircraft could efficiently sample ship emissions downwind of the stack.Following the campaign, WRF was run using ECMWF (European Centre for Medium-Range Weather Forecasts) analysis as initial and boundary conditions, (see Table2of 270Wentworth et al., 2015), in order to refine forecast meteorology and to interpret campaign data.The quality of the WRF predicted meteorology has been evaluated using measurements made on-board both the research aircraft and ship, indicating the forecast accurately predicts the meteorological situation during plume sampling (flight tracks shown in Figure2).Surface wind speed and wind Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2015-1032,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: 25 January 2016 c Author(s) 2016.CC-BY 3.0 License.direction predicted by WRF during plume sampling are shown in Figure 3.During the first plume 275 sampling on 19 July 2014, the flight was conducted west of the ship location due to the easterly winds throughout the plume sampling, characterized by high wind speeds above 10 m s −1 (Figure 2a) in Lancaster sound (Figure 3a).For the second plume, on 20 July 2015, the ship was located just north of Somerset Island (Figure 3b) and the flight sampled ship emissions southwest of the ship, between the ship and the Somerset Island.The meteorological situation near the flight was less consistent 280 in the measurement region on 20 July 2014, indicated by the variable wind directions and lower figures indicate that the plume intercepts are in the same locations as the partial columns predicted 315 Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2015-1032,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: 25 January 2016 c Author(s) 2016.CC-BY 3.0 License.
Petzold et al. (2008) find γ = 1.5±0.06 for a ship plume expansion in the English Channel, and von Glasow et al. (2003) find a best guess value of γ = 1.35 for a number 355 of previous studies also in mid latitudes.Our lower expansion rate suggests that ship plumes in the Arctic marine boundary layer mix with the background to a lesser extent compared to mid latitude due to the statically stable conditions.

435near
Resolute Bay, NU, Canada, during the summer 2014 NETCARE campaign.Three plumes (1, 2, and 3) were studied on consecutive days from 19 to 21 July 2014 by airborne interception using the Polar 6 aircraft, an analytical plume dispersion model, and by the FLEXPART-WRF dispersion model.The first plume measurement was carried out during the operation of Amundsen in the open water while moving at an average speed of 3.23 ± 0.25 kts.The second and third plume mea-440 surements were carried out when Amundsen reached the ice edge and operated under icebreaking conditions with much lower speeds of 1.31 ± 1.92 kts and 0.09 ± 0.30 kts, respectively.The engine load was low compared to cruising conditions during this campaign.The measured species included CO 2 , NO x , CO, SO 2 , particle number concentration using a Condensation Particle Counter (CPC), refractory Black Carbon (rBC), and Cloud Condensation Nuclei (CCN).The results were compared 445 to similar experimental studies in mid latitudes.Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2015-1032,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: 25 January 2016 c Author(s) 2016.CC-BY 3.0 License.

Figure 8 .
Figure 8.An example time series plot for identified pollution peaks in plume 3; sampling time for all instruments is 1 s except for O3 (10 s) and OPC (5 s).

Figure 9 .Figure 10 .Figure 11 .
Figure 9. Calculated plume growth or expansion rate (γ = −m) along the flight tracks using aircraft measurements for plume 1 (a), plume 2 (b), and plume 3 (b).Note: using the methodology in section 2.2.5 a plume age could be assigned to n = 6 data points for plume 1, n = 7 data points for plume 2, and n = 18 data points for plume 3.

Figure 12 .Figure 13 .
Figure 12.Emissions factors for NOx; fuel type (HFO: heavy fuel oil with high sulfur content, and MGO: marine gas oil with low sulfur content), or vessel class based on gross metric tonnage (HSD: high speed diesel < 5000 t, MSD: medium speed diesel 5000 − 30000 t, or SSD: slow speed diesel > 50000 t); plumes 1, 2, and 3 indicated on the plot with numbers 1, 2, and 3

Figure 14 .
Figure 14.Emissions factors for black carbon.

Figure 15 .
Figure 15.Emissions factors for for total particle count.

Table 3 .
Emissions factors for NOx; numbers in brackets indicate engine load (%), fuel type (HFO: heavy fuel oil with high sulfur content, and MGO: marine gas oil with low sulfur content), or vessel class based on gross metric tonnage (HSD: high speed diesel < 5000 t, MSD: medium speed diesel 5000 − 30000 t, or SSD: slow speed diesel > 50000 t).

Table 5 .
Emissions factors for black carbon; a elemental carbon, filter measurement based on a thermal/optical carbon aerosol analyzer according to NIOSH 5040; b black carbon measurement based on weighted average using SP2, SP-AMS, PAS, and PSAP; c black carbon measurement based on weighted average using SP2, SP-AMS, PAS-G, PAS-B, and PSAP; d black carbon measurement based on Multiple Angle Absorption Photometer (MAAP); e light absorbing carbon measurement based on photoacoustic techniques; f black carbon measurement based on PAS; g elemental carbon, filter measurement based on a multi-step combustion method according to VDI guideline 2465-2.Agrawal et al. (2008) a 0.068 (25 %), 0.034 (50 %), 0.021 (75 %)