Ship traffic is a major source of aerosol particles, particularly near shipping lanes and harbours. In order to estimate the contribution to exposure downwind of a shipping lane, it is important to be able to measure the ship emission contribution at various distances from the source. We report on measurements of atmospheric particles 7–20 km downwind of a shipping lane in the Baltic Sea Sulfur Emission Control Area (SECA) at a coastal location in southern Sweden during a winter and a summer campaign. Each ship plume was linked to individual ship passages using a novel method based on wind field data and automatic ship identification system data (AIS), where varying wind speeds and directions were applied to calculate a plume trajectory. In a situation where AIS data are not matching measured plumes well or if AIS data are missing, we provide an alternative method with particle number concentration data. The shipping lane contribution to the particle number concentration in Falsterbo was estimated by subtracting background concentrations from the ship plume concentrations, and more than 150 plumes were analysed. We have also extrapolated the contribution to seasonal averages and provide recommendations for future similar measurements. Averaged over a season, the contribution to particle number concentration was about 18 % during the winter and 10 % during the summer, including those periods with wind directions when the shipping lane was not affecting the station. The corresponding contribution to equivalent black carbon was 1.4 %.
Air pollution from anthropogenic activities, such as ship traffic, affects both human health and climate. Airborne particles cause negative health effects such as pulmonary and cardiovascular diseases, resulting in premature deaths and increased societal costs. Air pollution from combustion sources have an effect on climate due to emissions of greenhouse gases as well as particles with different optical properties and cloud-forming capacities.
In order to reduce air pollution there are regulations on sectors with high
emissions, for example the transportation sector. However, despite these
regulations air pollution continues to be a serious problem. One sector that
has gained relatively little attention in terms of emission control in the
past is international shipping. The relative contribution from shipping to
the total air pollution from transport is an increasing problem due to
expected growth in shipping activity (Brandt et al., 2013; Corbett et
al., 2007). One regulatory measure that has been taken to specifically
reduce sulfur emissions from ships is the introduction of so-called Sulfur
Emission Control Areas (SECAs), where the Baltic Sea SECA was one of the
first to become established (Corbett and Fischbeck, 1997). In the
International Convention for the Prevention of Marine Pollution from Ships
(MARPOL) Annex VI, the main exhaust gas emissions of sulfur oxides (
One way to characterize and quantify ship emissions is through measurements in coastal areas, downwind of a shipping lane. This makes it possible to register an increase in particle levels and the exposure to particles in this area when individual ship emission plumes pass the measurement station. With increasing distance from the emission source, the plume becomes more dilute and physically and chemically transformed due to atmospheric ageing. In order to assess physicochemical properties but still capture features of the aged particles, which differ from the freshly emitted, it is therefore desirable to measure at an intermediate distance to the ships. Measurements of ambient aerosol particles are also important for an accurate assessment of the health effects, which depend on the actual exposure. This motivates measurements of the atmospherically aged ship aerosol particles from all types of ships affecting the coastal population. However, there are challenges associated with measuring aerosols from individual plumes further away from a moving point source such as a ship. Dilution will eventually make it harder to distinguish from background levels – there can be an overlap of several plumes that intersect, and varying wind speed and wind direction makes it less obvious which ship is connected to which plume if the traffic is relatively intense.
All ships on international water with gross tonnage above 300 t, cargo ships with gross tonnage above 500 t, and all passenger ships are required to be equipped with a tracking system called Automatic Identification System (AIS). A ship sends out a position signal with individual International Maritime Organisation ID and information about its type, size, country of origin, speed, etc. These data are collected every 6 min. AIS data in the Øresund region were used in this study to tie individual ship plumes to specific ships. AIS can be used as a tool in ship emission studies, commonly as a source for emission inventory used in models. This bottom-up method has been used and developed by many (e.g. Jalkanen et al., 2009, 2012; Liu et al., 2016; Beecken et al., 2015; Chen et al., 2018; Johansson et al., 2017; Marelle et al., 2016; Goldsworthy and Goldsworthy, 2015). AIS has also been used in connection to ambient plume measurements, to identify individual ship emission plumes. Alföldy et al. (2013) performed visual observations of ships at short distances in a port area and could connect these to live updates of ship positions. Ault et al. (2010) measured plumes and connected these to individual ships by using AIS ship positions and assuming transport with constant wind speed and wind direction. Balzani Lööv et al. (2014) used a similar method to locate plumes after emission, e.g. when doing airborne measurements within plumes further downwind of the ships. Diesch et al. (2013) also measured individual plumes and connected plume properties to ship properties, such as weight, using AIS, also at short distances (1–5 min downwind). Hence, AIS information has successfully been used in several applications, but for doing individual ship plume identification at longer distances where the plume might not travel along a straight path between emission and detection, other approaches might be needed that take into account the non-linear wind speed and direction. One example is the method of following ships either by aircraft or with a ship vessel up to a few kilometres behind the ship (Berg et al., 2012; Petzold et al., 2008; Chen et al., 2005; Williams et al., 2009; Lack et al., 2009). An advantage of this method is that the ships can be followed at different downwind distances and can measure plume dilution and aerosol dynamics. However, it is an expensive method, and only a few ships can be followed due to budget and practical restrictions. Hence, this calls for a more feasible and cost-effective solution.
Particle number size distributions have been studied in atmospheric
conditions previously, showing some variations in sizes and number of modes.
This can be expected since many factors affect the emissions, such as engine
operations, and the atmospheric transformation processes. For example,
Jonsson et al. (2011) showed that size-resolved particle number
emission factors were largest around particle diameters of 35 nm, with
smaller sizes observed for ships running on gas turbines than on diesel
engines. Out of these particles, 36 %–46 % were non-volatile and could
contain some black carbon (BC). These measurements are from 2010, i.e.
during the 1 % sulfur limit within SECAs. Pirjola et al. (2014)
showed that the number size distribution had two modes for fresh ship
plumes: a dominating mode peaked at 20–30 nm and an accumulation mode at
80–100 nm. About 30 % of these were non-volatile, and it was also shown
that the after-treatment system affected the total particle number emission.
These measurements are from 2010 to 2011. Diesch et al. (2013) observed a
nucleation mode in the 10–20 nm diameter range and a combustion aerosol
mode centred at about 35 nm. No particles with sizes above 1
We present a new revised method to identify individual aerosol ship plumes
based on AIS data and non-linear wind transport of the ship plume to a
stationary coastal field site, which is several kilometres downwind. The method has
been tested on particle number concentration, particle number size
distribution, and black carbon mass. In addition,
We identified and calculated the contribution as well as the particle size
distribution of individual ships by subtraction of background concentrations
from the identified plume particle number concentrations. In addition, we
have developed and described a new method to calculate the contribution of
aerosol properties when the plume cannot be visually distinguished from
background concentrations due to noisy data and relatively weak contribution
at this fairly long distance from the shipping lane. This method has been
tested on equivalent black carbon (eBC) concentrations. eBC is black carbon
mass concentration derived from optical absorption measurements and a mass
absorption cross section (MAC) value (Petzold et al., 2013). In our measurements,
the MAC value for the 880 nm wavelength was 7.77 m
The location of the sampling site was on the Falsterbo peninsula in
south-western Sweden (55.3843
Location of the measurement station (circle with cross) at the
Falsterbo peninsula together with ship traffic density; the colour bar
indicates an approximate number of distinct vessels passing per day per
squared kilometre (
A PM
Measurement setup. The symbol (
During the summer campaign, the aerosol flow for certain instruments (Fig. 2) was dried using either diffusion or membrane (Nafion) driers. The particle losses in the membrane dryers due to diffusion were determined by laboratory measurements. For 100 nm particles the losses were in the range 0 %–10 %, and for 10 nm particles the losses were about 5 %–20 %. Specifications about the dryers and losses can be found in Table 1 and Fig. 3. These losses are used to correct the size-resolved scanning mobility particle sizer (SMPS) data. Table 1 presents the specifications for each dryer used in the summer campaign to dry the aerosol particles before sampling with some of the particle instruments. Letters A–C correspond to the dryers shown in the illustration of the Falsterbo measurement setup in Fig. 2. The flows for which the losses are characterized were the same flows as used in the field measurements. The aerosol used for the characterization was polydisperse ammonium sulfate in lab room air. The resulting losses, as a fraction of the total particle concentration, are shown as function of particle size in Fig. 3. In addition, corrections for particle losses in the sampling line were calculated using the Particle Loss Calculator tool (Von der Weiden et al., 2009) and were applied to the SMPS size distributions but not for the other instruments.
Specifications of dryers used in Falsterbo; letters A–C correspond to the driers in Fig. 2.
Fraction of total particle concentration lost due to diffusion in
three dryers, as a function of particle diameter,
To confirm the contribution of ship plumes to particle and gas concentrations in Falsterbo, the time when each ship plume should influence the Falsterbo site was estimated with the revised method based on automatic ship identification system position data as well as wind direction and wind speed data from the Falsterbo lighthouse Swedish Meteorological and Hydrological Institute weather station (SMHI, 2017).
Only ships passing by in the area limited by a rectangle with geographical
coordinates (55.16
For each interpolated 1 min ship position, wind trajectories were
calculated describing how the wind travelled from the ship at time 0
(
When the wind was not arriving from the sea, the ships did not influence the
measurements. Ship passages were defined to influence the Falsterbo station
only if the minimum distance between the wind path at
There is a significant uncertainty in finding the
Even for periods when AIS data were not matching plume times well, or when AIS data were missing from the AIS database, particle number concentrations could be used to identify ship plumes instead. This required that there were no other interfering particle number concentration sources, or that these could be distinguished from the ship plumes. The number concentration data were then used to identify the plume time period, since particle number concentrations were always above a detection limit for all ship plumes, and the time resolution was large enough to clearly identify the shape of the plume peak. However, all increases in particle number concentration were not a result of ship emissions but rather land-going vehicles passing the measurement site. These could be recognized and excluded. Normally, the land-going vehicles were influencing the particle concentrations for a minute or shorter, while the ship plumes that influenced the particle concentrations could last for several minutes up to about 20 min. Note that the alternative method of identifying plumes with number concentration is not giving information about which ship passed by the measurement site due to lack of AIS data, unless there are other ways of collecting this information.
For an identified ship plume peak, the contribution from this plume was estimated by calculating the area under the peak after subtraction of background concentrations. An example of a measured ship plume and illustrations of these calculations are shown in Fig. 4. In Fig. 4, the particle number concentration is clearly elevated during a few minutes during a period of relatively constant background concentrations. The estimated time of arrival of the plume, based on wind and AIS data (as described previously), is marked with a star and confirms the measurement of a ship plume and could provide further information about the ship, if desired. Due to the frequent appearance of ship plumes in Falsterbo, the background concentration was calculated as the average concentration of two intervals, one just before and one just after the ship plume, as seen in Fig. 4.
Illustration of method of calculating aerosol contribution of individual ship plumes. Particle number concentration measured by a CPC (solid blue) and black carbon measured by an Aethalometer (AE33, dashed orange) during ca. 25 min of ambient sampling, and calculated time of arrival of the aerosol plume based on AIS and wind data (star). Plume duration is estimated by observation, and background concentrations are based on 6 min plus 6 min of adjacent data. The average of the background is subtracted from the plume concentrations to obtain only ship emission contribution.
One alternative way to calculate plume contribution by subtracting the plume from the background is the method used by Kivekäs et al. (2014). The authors extracted particle background concentrations by taking the 25th percentile values of a sliding window of a few hours for the particle number concentration time series. This is an appropriate automatic method to use on large data sets of ship plumes. The ship lane in the Kivekäs study was between 15 and 60 km away from the station. During periods with sharp increases or decreases in background concentrations, this method did not yield acceptable results, and these periods had to be manually controlled for errors and removed from the final data analysis. However, the Kivekäs method was not possible to use in Falsterbo due to the frequent plume events and the relatively high number concentrations in the plumes, which affected the background values for the sliding window method.
If a measured concentration of some aerosol parameters is noisy or the plumes are similar in concentration to the background, it is still possible to use AIS or particle number concentration to identify plumes and calculate their contribution. This could be the case when particle mass concentrations in the ship plumes are generally low. For example, a plume peak is not clearly distinguished, as depicted in Fig. 4 for eBC mass concentrations. However, based on the identification from the AIS and the estimation of the plume duration from particle number concentration data, the effect of the plume on the other aerosol parameters could be investigated. The contribution from a ship to such an aerosol parameter was calculated in the same way as described above, by subtracting the adjacent background concentrations from the concentration during the plume period. The start and end time of the plume was assumed to be the same as measured by the particle counter.
Besides the contribution to aerosol concentrations in each plume, there is
also a possibility to estimate the contribution from ships at a coastal
location during an extended period of time, like a day, a season, or a year.
This can be accomplished by multiplying the average plume contribution with the number of ships that have passed during the current period. Further, to account for wind direction, the value is multiplied with the fraction of the time that the wind was passing over the shipping lane towards land. We estimated the daily and seasonal contribution
of ships (
To demonstrate how the ship identification with the AIS method worked, Fig. 5 shows an example of a time series from the CPC for a few hours of sampling during wintertime. Figure 5 also displays the times when the ship plumes were expected to arrive at the measurement station based on AIS and wind data, as described in Sect. 3. The particles from the ship plumes are seen as relatively short and intense peaks, generally matching well with the expected plume passages. The average plume duration was 10 min. All ships identified with the AIS system resulted in an increase in size-dependent particle number concentration when these measurements were available. The method to infer when the ship plume should affect measured concentrations at Falsterbo agreed excellently during winter considering that the wind speed and direction measurements had a 1 h resolution and that these parameters were only measured at Falsterbo and not along the air mass trajectory. In summer, this agreement was reasonable but less certain than in winter, which might be due to more turbulent winds and local meteorological factors such as sea breeze. This shows that the method has a potential to work for many different shipping lanes. All plumes passing the measurement site are observed in the particle counters; that is the fraction of observed plumes predicted by AIS trajectories is in principle 1. We miss some plumes in the individual ship analysis, due to too frequent and overlapping plume passages. We estimate the analysed fraction to ca. 0.4 for the ship traffic near Falsterbo. The analysed fraction depends on the plume duration as well as the frequency of ships. With an average plume duration of about 10 min, it also means that the plume peak maxima should be separated by at least 10 min to be able to correctly calculate plume contributions. For studies which do not require information about individual ships but rather about total ship contribution, the number of missed ships is very low and can be due to temporary AIS malfunction or military vessels passing (they do not transmit AIS). The highest uncertainty of the timing of the plume is introduced through the wind trajectories between the emission and measurement site. Regarding the uncertainty of the attribution of a ship ID to a plume, this is depending mainly on the frequency of ship plumes at the specific location in combination with the wind trajectories. If the plumes from two ships arrive about the same time to the Falsterbo station, we cannot be absolutely sure which ships contributed to which plume concentrations. In that case, we only know that two ships did contribute to elevated concentrations. Also, if these plumes are superimposed on top of each other, we are still not able to calculate the individual ship contribution. We choose only to calculate plume contribution for plumes whose peaks are about at least 10 min apart in order to avoid plume superposition, since average plume duration is about 10 min as stated in the manuscript. In this case, the ship identification is always assigned to the correct plume. We have seen that the timing accuracy of the ship ID with the actual plume contribution is better (lower) than 7 min (95 % CI). Since, we choose only plumes or ship ID data which are at least 10 min apart, this uncertainty has no effect on attributing a ship ID to the correct plume.
Particle number concentration measured with a CPC, and calculated incidents of ship plume passages (stars) determined with AIS and meteorological data, versus time (31 January–1 February 2016), from measurements at the coastline in southern Sweden during an episode with westerly winds blowing from the Øresund strait to the coastal station Falsterbo. The concentrations are those of the total aerosol; i.e. background concentrations are not subtracted.
As an example of what AIS information can be used for, the properties of the ships identified in Falsterbo during the winter campaign are shown in Fig. 6. The distributions of ship weight, length, breadth, and average speed as well as the distance from the emission source to the measurement site (in distance, kilometre; and in transport time, minutes) are shown. The units of the parameters have been adjusted so that all values fit within a similar range in the plot. The linear distance from the ship to the measurement site at the time when the ship contributed to the pollution at the site is denoted “ship to site / km” and given in kilometres, and the transport time of the wind between the ship emissions and the site is denoted “ship to site / min”, and given in minutes. Note that the wind does not necessarily travel along a straight line between the ship and the station if the wind direction is changing, which is considered in the calculation of the “ship to site / min”.
AIS ship information and calculated plume travel data for the 113 plumes evaluated from the winter campaign in Falsterbo. The boxes show the median, 25th percentile, and 75th percentile, and whiskers show the minimum and maximum value. The maximum deadweight of 140 kt is out of range.
No relation was found between emission and ship properties or transport; therefore the data presented are not normalized for weight or transport time but presented as they were measured at the measurement site. A variety of vessels pass the Øresund strait and Falsterbo. The most common ones are cargo ships, tankers, and ro-ro ships (roll-on/roll-off) and others are trawlers, dredgers, reefers, and fishing vessels. The production years of the ships ranged from 1965 to 2015, with a majority from the 1990s and 2000s.
The contribution of ship traffic to the air pollution at a coastal location was estimated for more than 150 ship plumes. Measurements were carried out with a similar setup during winter (January–March) and summer (May–July) of 2016. All instrument variables were not available for the entire measurement periods, and the wind direction was not always favourable for measuring ship plumes. In total, there were about 3 weeks with optimal data from the winter campaign and 2 weeks from the summer campaign.
For the calculation of how ships contributed to the particle number concentration, plumes were restricted to the following conditions: (1) identified by AIS, (2) clearly distinguishable from the background in the CPC time series, and (3) not overlapping with other plumes. This resulted in 109 (CPC) and 113 (SMPS) plumes from the winter campaign and 61 (CPC) and 8 (SMPS) plumes from the summer campaign used for further calculations. The number of plumes identified by the SMPS in the summer is much lower than identified by the CPC due to a non-functioning SMPS system in periods. Also, periods during which the SMPS was sampling aerosol through a potential aerosol mass oxidation flow reactor were also excluded in this analysis. Finally, there were in general fewer plumes identified in the summer than in winter due to lack of AIS data in summer and since the winds at Falsterbo less frequently arrived from the shipping lane during the summer measurement campaign. A summary of the meteorological conditions during the measurements can be found in Table 2.
Meteorological conditions during measurement campaigns: average, lowest, and highest values.
According to the methods described in Sect. 3.2, we calculated the individual
contributions from the observed ship plumes, both for particle number
concentration and for eBC mass concentration, as well as the estimation of a
daily and seasonal contribution at the specific location according to Eq. (1).
This calculation was based on AIS data, which showed an average of 73 and 63
ships passing per day in winter and summer respectively. Together with the
average plume duration (10 min), this indicates that the Falsterbo site is
affected by ship emissions 51 % of the time in the winter and 44 % in
the summer, when the wind blows from the Øresund strait. Based on historical
wind data from the last 20 years (Swedish Meteorological and Hydrological
Institute), the wind intercepts the shipping lanes in Øresund strait about
70 % of the time in both summer and winter, which was used together with
particle concentration measurements to estimate the seasonal contribution
from ships. Examples of plume contributions – individual, daily, and
seasonal – are shown in Table 3. For each of the
Contribution of particle number concentration and eBC mass concentration to local air quality, from two measurement campaigns at the Falsterbo coastal site.
Regarding the uncertainty in the plume particle number contribution, the
relative statistical error of the CPC count is related to the total count
Despite the fact that the plumes were not clearly visible in the eBC time
series, due to the low contribution to mass, a significant increase in BC
was observed during identified plume events. The seasonal contribution of
ship-emitted eBC is on average only
The ship contribution to the average size distribution of
particles (diameter,
The mean and median particle number size distribution for the ship emission plumes in Falsterbo are shown in Fig. 7. The distributions were calculated by averaging the number concentration in each SMPS size bin for 113 ship plumes for the winter campaign and 8 ship plumes from the summer campaign. A log-normal function (Hussein et al., 2005) with several modes was fitted to the average and median size distribution plumes for the winter and summer seasons. For the log-normal function, only particles with an electrical mobility diameter larger than 15 nm and smaller than 150 nm are considered due to uncertainties and losses for other sizes. The log-normal parameters are listed in Table 4. Four or five modes are used in the log-normal fit of the average size distribution plume since it seems that the typical size distribution contains a smaller- and a larger-sized nucleation mode (mode no. 1 and 2, < 30 nm diameter) and a smaller- and larger-sized Aitken mode (30 to 100 nm diameter). A majority of the ships do not produce the lower-sized nucleation mode, which is why the median size distribution does not contain this first mode. The other modes are often all present at the same time, and the larger particles could arise due to coagulation in an aerosol with a high concentration of smaller particles or due to emissions of relatively large primary soot particles. The uncertainties for the size distribution are large for the particles in the upper Aitken mode (80 to 100 nm diameter) and the accumulation mode (> 100 nm diameter) due to low numbers counted in the SMPS and also due to large variation between individual ships. The Pirjola et al. (2014) study shows that the particle number size distribution has two distinct modes for fresh ambient ship plumes – one in the nucleation mode (< 30 nm diameter) and one in the Aitken mode (30–100 nm diameter). If the number size distribution is remade into a volume size distribution, an accumulation mode also becomes visible (> 100 nm diameter). The current study also contains these modes. In addition, due to the individual variability between ship plumes in the current study, even two Aitken modes are discernible in the log-normal fitted size distributions. A few of the ships have a distinct accumulation mode, and for this reason, the average size distribution also contains this log-normal fitted mode. The data are significantly corrected for particle losses in sampling tubing especially for the nucleation mode sizes (< 30 nm diameter), which makes a second log-normal nucleation mode below 15 nm diameter appear in the log-normal fitted size distributions. Lab engine measurements also show such a mode in the Anderson et al. (2015) study, when higher sulfur content fuel was used, which stimulated new particle formation. Hence, in total, there are three to five log-normal modes fitted to the median and average particle number size distributions (Table 4).
Log-normal fit parameters for the average and median size
distribution of the detected ship emission particles, during winter (
The number size distributions in Fig. 7 show that essentially all particles
in the average and median ship plume have an electrical mobility diameter
below 100 nm, most of them around 20–40 nm. Similar results have been
shown in laboratory and on-board measurements (Kasper et al., 2007; Betha
et al., 2017; Isakson et al., 2001; Kivekäs et al., 2014). There have
also been observations of larger particle diameters in the micrometre range,
e.g. Fridell et al. (2008). In our study, the APS instrument did not
show any contribution to micrometre-sized particles from ships at the current
distance from the shipping lane. The APS has a high sensitivity for single
particles but did not measure that ship plumes contained significant
particle number concentrations above background concentrations for particles
larger than 0.5
The size distribution of the average plume shows higher concentrations than that of the median plume, both for the summer and the winter data. This is due to the high contribution of some ships skewing the results. Due to higher and noisier background particle concentrations in the summer (Table 3), and the lack of AIS data, it is possible that plumes with relatively low particle number concentrations were not distinguishable from the background, and hence the selection of plumes in the summer might have been biased towards the more-polluting ship plumes. Also, the difference in sample size should be noted here: 113 good plumes observed during winter and 8 during summer. This difference depends mainly on instrument malfunction, unfavourable wind directions, and lack of AIS data. From the available data, there is however an indication that the number and the size of the particles from ships are somewhat larger in summer. This seasonal difference could possibly be explained by secondary particulate matter formed by atmospheric ageing, which is expected to be more significant in the summer, but more measurements are needed to confirm this.
Due to the distance to the shipping lane, the ship emissions were diluted enough to have concentrations below the detection limit of some instruments. In order to capture the relatively short plume events, the time resolution could not be too long either, making the detection limit of some instruments higher.
The AIS method to identify which ship influenced exposure on land and to identify individual ship plumes from measurements about 10 km downwind of ship lanes proved to be very exact for the winter data and worked relatively well for summer data. We know that the method to observe individual plumes on top of background concentrations does not work for all ships at the distance 25–60 km downwind of a shipping lane (Kivekäs et al., 2014). There, only a fraction of the plumes were distinguishable. In contrary, at our Falsterbo site, there were no such issues with the plume identification method. Hence, the method can be expected to work at least up to 10 km and get worse towards 60 km. This is true for particle number concentration measurements (with a CPC) but not for mass concentration measurements. So, to be able to detect plumes at maximum distances a particle counter is of importance. Considering the wind data were available only with a 1 h resolution, the plume identification worked well. Availability of wind data with better time resolution does not seem to be necessary at this specific site. Although at longer distances between the ship lane and the station, this can potentially be an issue and it would be advantageous to have meteorological data with better time resolution. When AIS data were missing for one reason or the other, the particle number concentration detected with a condensation particle counter also proved to work very well to identify ships, although it could not give the information about which ship it was.
The method to estimate plume contribution from individual ships proved to be
straightforward for the clearly visible ship plumes at the measurement
station. For the eBC concentration, the plume identification was less
straightforward since the plume signal was very low relative to the noise
level. For many plumes, no increase in eBC was observed with the bare eye.
We still used the already identified plumes to calculate the contribution to
eBC. A very low but still significant plume contribution could be
calculated. Even if the proposed method yields non-significant plume
contributions for a specific parameter, this does not mean that the method
does not work. Rather it means that ship emissions do not contribute to
significant exposure inland for this parameter and that the detection
capabilities of the instrument do not allow for detecting this
non-significant contribution. The calculation was done using the precise
time of the plume incidence observed in the particle counter. This was also
a surprisingly robust method without systematic biases due to the noise.
Dispersed background levels of BC were about 0.2
Since the particle counter always yielded visible and smooth plumes at the downwind station, it is recommended to always bring a particle counter when doing these kind of measurements, even if it is not of main interest to estimate particle number contributions. Namely, it might turn out that AIS data are erroneous or missing, and the particle counter is needed to define plume time and background to calculate the plume contributions for instruments with high noise and low time resolution. Since the ship plumes at 10 km downwind or farther from the ship lanes have only a few minutes up to about 20 min duration, it is also recommended that the time resolution of the instruments one brings is not worse than 1 min. For a scanning particle sizer, like the SMPS, one should consider the scan time in comparison to the plume duration, and possibly add a mixing volume to not get rapid changes in the aerosol particle concentration during a scan.
These and other measurements have shown that the number of particles < 30 nm diameter is substantial, even for relatively aged ship plumes. The estimation and correction for particle losses are therefore crucial to be able to assess the true size-dependent particle concentrations, especially when the sampling line to instruments is relatively long. It is our recommendation to place the particle counter (CPC) close to inlet and further to use as short a sampling line as possible with minimum diffusion losses when performing these kinds of studies in general.
The current method of stationary measurements of downwind plumes from a shipping lane has turned out to be very cost-effective compared to aircraft or ship vessel chasing experiments and can fetch a much higher number of ship plumes. Hence, we also urge the use of it for economic and pragmatic considerations when studying relatively aged ship plumes for a high number of ships. In future studies of detailed individual ship plumes and the emission sources, it should be considered whether the particle emissions depend on ship engine power used. It is possible to estimate the engine power required by a ship, using the total power of the ships, their design speed, and actual speed through the propeller law (Moreno-Gutiérrez et al., 2015). This can then be compared to particle number concentration emissions but also particle mass emissions and gaseous emissions. With the method presented in this paper, it is possible to collect information on a very large sample of ships for these kinds of investigations.
Before performing the measurements with the new method, it is important to investigate the meteorological situations at the current measurement site. For example, during sea breezes, local wind measurements could indicate that shipping lane emissions should reach the measurement station, whereas in reality they might not. Care should be taken to account for these periods when the meteorological data will give erroneous results. However, these meteorological phenomena do not take place all the time; hence these specific meteorological conditions will not disqualify any chosen measurement site with the current proposed method. Again, these uncertain wind conditions make it very important to bring a particle counter to register shipping plumes. If the particle counter does not register any ship plumes during a selected time period, this indicates that winds from the ships are not reaching the measurement station, despite the fact that the local wind measurements suggest otherwise.
Beyond providing ambient aerosol data from a SECA from summer and winter measurements, the data from this study can also be used to validate process models simulating ageing processes of particle number size distributions as well as long distance transport along meteorological air mass trajectories in Lagrangian process models. In addition to particle number concentration and eBC, the method was applied in the companion paper, Ausmeel et al. (2019), focusing on other aerosol properties and regional or global scale air quality and climate models could use this kind of data to validate modelled ship contribution in certain grid cells.
The data sets used in this study are available upon request from the authors (S. Ausmeel and A. Kristensson).
AK designed the experiments and all authors carried them out, with AK being the project leader during the winter campaign and SA during the summer campaign. AK developed the model code. SA analysed the plume and aerosol data and prepared the manuscript with contributions from all co-authors.
The authors declare that they have no conflict of interest.
Fredrik Windmark at the Swedish Meteorological and Hydrological Institute (SMHI) is acknowledged for helping to provide AIS ship positioning data. Mårten Spanne, Paul Hansson, Henric Nilsson, and Susanna Gustafsson from the Environment Department at the city of Malmö are acknowledged for helping in preparing and setting up the measurements at Falsterbo. Kirsten Kling of DTU and Antti Joonas Koivisto of NRCWE are acknowledged for helping with the summer campaign and Fredrik Mattsson and Anna Hansson for helping with the winter campaign. Thank you also to Håkan Lindberg and the personnel from the Falsterbo golf course and Lennart Karlsson from the Falsterbo bird watching station, who were willing to prepare a place for our measurement trailer, as well as the County Administrative Board of Skåne and Vellinge municipality for giving permission to measure in the Flommen Nature Reserve.
This research has been supported by the Svenska Forskningsrådet Formas (grant no. 2014-951) and the Crafoord Foundation (project nos. 20140955 and 20161026).
This paper was edited by Folkert Boersma and reviewed by two anonymous referees.