Fluxes of gaseous compounds and nanoparticles were
studied using micrometeorological methods at Harmaja in the Baltic Sea. The
measurement site was situated beside the ship route to and from the city of
Helsinki. The gradient (GR) method was used to measure fluxes of SO
The Baltic Sea, owing to its nature as a relatively small inland sea that
experiences heavy ship traffic and is surrounded by populated areas, is very sensitive
to pollutants. Due to the very narrow and shallow strait of Kattegat in
Denmark, the exchange of seawater between the North Sea and the Baltic Sea
is limited. The load of phosphorus and nitrogen in the Baltic Sea mainly
comes from rivers. The rivers bring in fresh water, maintaining the
salinity of the seawater in the Baltic Sea at
The goal of this study was to (i) measure the gas and nanoparticle exchange
between the sea–air interface in a marine coastal environment close to
ship routes, (ii) study the transport and dispersion of a ship plume to
the footprint area, (iii) define the FSC of ship emission plumes,
and (iv) characterize the uncertainty sources of the measurement results.
The measurements took place in the Baltic Sea on the small island of Harmaja
in the vicinity of the city of Helsinki during the summers of 2011 and 2012.
The ship routes between the city of Helsinki and the cities of Tallinn,
Stockholm, and St. Petersburg pass by the measurement site. The exchange of
gaseous NO
Micrometeorological methods are used to measure gas exchange across
the surface layer (Kaimal and Finnigan, 1994); of these, the eddy-covariance (EC) method and the gradient (GR) method are commonly used. The use of micrometeorological methods requires certain criteria to be met with respect to the atmospheric conditions: the homogeneity of the turbulence flow field on the footprint area, the stationarity of the measuring processes, and the absence of swell (e.g., Foken and Wichura, 1996; Miller et al., 2010; Drennan et al., 2003). The eddy-covariance method is a direct flux measurement method,
whereas the gradient method is an indirect measurement method. In the eddy-covariance method, the flux of a gas compound is measured using fast
sensors (response better than 10 Hz) to measure the fluctuation in wind
velocities and the concentration of chemical compounds. In contrast, the gradient method
overcomes the problem of the fast analysis of chemical compounds. However, this method requires that the atmospheric conditions are stationary,
needs very accurate measurements of the parameters that it uses
(Businger, 1986), and assumes a constant layer flux (Dyer and
Hicks, 1970). In this study, both the GR and the EC methods were used: the
gradient method was used for gas compounds (oxides of nitrogen, ozone,
sulfur dioxide, and carbon dioxide) and nanoparticles, and the eddy-covariance method was used to measure the flux of carbon dioxide. In the GR
method, the wind speed should be measured at different heights, usually with
conventional cup anemometers, whereas in the EC method, the fluctuation in the
three-dimensional wind speed is measured by a sonic anemometer. Short
descriptions of both methods are given below, with more emphasis on the
gradient method.
In Eq. (5), the concentration difference at heights
On the other hand, the sonic anemometer measures the wind velocity with the
help of acoustic pulses that propagate along the path between the sound
emitter and the receiver. The three-dimensional wind components, i.e.,
horizontal (
Similarly to the momentum flux in Eq. (6), one can express the vertical flux
of a gas compound (e.g., CO
The chemical interconversion of the NO–O
The chemical cycle in the NO–O
The maximum allowed sulfur content in marine fuel used on the oceans is defined
in Annex VI of the MARPOL agreement (IMO, 2008). The agreement also
defines the sea areas where a lower marine fuel sulfur content
must be used. These restricted sea areas, called “SO
A method that can be used to determine the FSC during a cruise is to
measure the ratio of
Equation (10) yields lower limits for the FSC (Williams et al., 2009), as a small
part of the sulfur in the fuel, less than 6 % (Alföldy et al., 2013)
or 0.7 % (Moldanová et al., 2013), might be emitted as SO
The uncertainty in the measurements is one of the most important issues to solve when analyzing the results. The measurements are influenced by a number of error sources that need to be identified and quantified, including the performance characteristics of gas and particle analyzers and the different probes and sensors used for the measurements. Besides the uncertainties associated with the instruments, the EC and GR methods are very sensitive to the topography and the atmospheric conditions. In particular, the stochastic nature of turbulence (Lenschow et al., 1994; Rannik et al., 2006) and the noise present in the measured signals cause random errors (Lenschow and Kristensen, 1985; Rannik et al., 2016), which are difficult to estimate. However, a number of error sources have been identified that have an influence on the results of flux measurements from the EC method (Businger, 1986; Rinne et al., 2000). The statistical error of an EC estimate is usually quite large.
The uncertainty sources that contribute to the uncertainty of the flux
results by the GR method are systematic and random in nature. Calibration of
the response of all instruments, correction of the humidity for CO
The combined standard uncertainty of the flux of a compound
The combined standard uncertainty of the fuel sulfur content
The first measurement campaign at Harmaja (60
International ship routes in the Baltic Sea, the ship routes to
the harbors of Helsinki, and modeled NO
The measurement station (Fig. 2) was set up in an old military fire control tower made of steel and concrete. All of the measurement instruments were installed inside the tower, and measurement probes and sampling inlets were installed at different heights on a mast beside the tower. The height of the mast was 9 m, and it was located on a breakwater 3.5 m above the mean sea level (a.m.s.l.). The measurement probes were installed at different heights to get an extensive view of the meteorological parameters of interest. Cup anemometers (WAA15 cup anemometers, Vaisala, Finland) measured the wind speed, while the turbulence parameters were measured using an ultrasonic anemometer (ultrasonic wind sensor, uSonic-3, METEK GmbH, Germany). Pt100 sensors measured the ambient temperature. The cup anemometers were installed at three different heights, 12.2, 10.9, and 9.9 m a.m.s.l. The sonic anemometer was installed at the top of the mast at a height of 12.9 m, and the temperature probes were installed at heights of 12.3, 11.0, and 10.0 m. The sampling intakes for the gaseous compounds were installed at two different heights, 12.58 and 9.98 m. The inlets for particle measurements were installed at 8.0 and 10.0 m. The official weather mast of the Finnish Meteorological Institute (FMI) was located next to this mast (Fig. 2) and was equipped with a cup anemometer (WAA15 wind vane, Vaisala, Finland) and a wind direction vane (WAV15, Vaisala, Finland) at a height of 16.6 m. The current sea level was measured as a 30 min average with reference to the mean sea level. The measurement heights of the probes used in the calculations are from the current sea level.
The atmospheric concentrations of ozone, oxides of nitrogen, and sulfur
dioxide were measured simultaneously by conventional gas analyzers intended
for ambient air quality measurement. Two identical analyzers of each gas
were used to detect the concentration at the two measurement heights (12.58 and 9.98 m). The sampling tubes at each altitude were made equal in
length, and PTFE (polytetrafluoroethylene) was used as the tube material, as it is
an inert material for each of the gaseous pollutants. The measurement technique used
for ozone was the UV-photometric method (EN 14625:2012), and ozone measurements were
performed with APOA-360 analyzers (HORIBA, Japan). For nitrogen oxides
(NO
For particle sampling stainless-steel tubes with an outer diameter of 12 mm
were used. Particle number concentration and size distribution were measured
by two ELPIs (electrical low-pressure impactors, Dekati Ltd., Finland) (Keskinen et
al., 1992); ELPI1 measured at a height of 10.0 m, and ELPI2 measured at 8.0 m. The
measurement principle of both ELPIs was the same: particles were first
charged and then classified into 12 stages according to their aerodynamic
diameter, in the size range from 7 nm to 10
The strategy for the air quality measurements in 2012 was different: it
included the measurement of concentrations of NO, SO
The temperature probes (Pt100) were calibrated in the
calibration laboratory at the FMI for meteorological quantities. The wind speed
anemometers were serviced (cleaned and the ball bearings changed) in the
same laboratory. The gas analyzers were calibrated in the reference
calibration laboratory at the FMI before and after the field campaign. The
calibration laboratory is responsible for the tasks of the national
reference laboratory on air quality, and it conducts the calibrations of the
air quality analyzers and calibration facilities in Finland. The laboratory
maintains the traceability of the calibration to the SI units, and it is
accredited according to ISO 17025:2005 for all measured gas compounds except
CO
The analyzers were also calibrated at the measurement site during the
campaign using a field calibration unit similar to that in the laboratory.
Both ELPIs were factory calibrated and serviced. Zero setting and high-efficiency particulate absorbing filter (HEPA)
filtration tests were performed before and after each measurement period.
Based on the parallel measurements of the ELPIs on 30 August and 2 September 2011
correction factors were inferred for ELPI2, separately for each stage
(
The data acquisition systems consisted of several components. The meteorological measurements were collected and stored by a MILOS 500 system (Vaisala, Finland). The ambient air quality gas analyzers were connected to a EnviDas 2000 (Envimetria, Israel) data collection system, and the sonic anemometer and the LI-7000 were connected to a fast data acquisition system; the Picarro G2301 instrument used the system provided by the manufacturer, and the ELPI software was used for the collection of particle data.
The times, East European Time (EET
The first target for the data analysis was to achieve accurate and good-quality continuous time series for the gaseous compounds and particles at
each of the measuring heights. Secondly, the turbulence parameters (M–O
length, stability parameter, friction velocity) were needed for calculating
the transfer coefficients
In 2012, measurements were also made on the Finnish research vessel
Environmental factors (e.g., the fire control tower) caused challenges with
respect to the measurement signals. Although the measurement probes were
installed at different heights above the top of the tower, the measurement
signals were affected by disturbances in the flow field. Therefore, a
computational fluid dynamic (CFD) program OpenFOAM (version 7;
Figure 3a illustrates the calculated wind field isopleths at a wind speed of 9 m s
As an example, Fig. 4a–b depicts the time series of 1 min averaged
concentrations of the measured gas compounds at 10 m altitude during the
first campaign. The sharp peaks in the concentrations of nitrogen monoxide
are very striking. In a detailed examination, the duration of the emission
peaks from the ships were of the order of a few minutes. Where there was a
peak in the NO concentration, a negative peak was also detected in the ozone
concentration due to the fast reaction producing NO
Time series of 1 min average concentrations of gaseous
The concentration of sulfur dioxide was very low, but clearly
distinguishable peaks, with a maximum of 28 ppb, were observed in the data.
These peaks originated from the passing ships. The ship peaks were also seen
in the CO
Selected air mass trajectories on 28 August at 06:00 UTC
Number
Figure 7 presents the maximum values for the concentrations of each gas
compound and the particle number within 10
Roses of the maximum gaseous concentrations of nitrogen oxides
(NO, NO
Quality control (QC) and quality assurance (QA) procedures are actions that
should be considered in order to improve data quality and make
data comparable with similar data from other studies. Although QA
and QC procedures have slightly different meanings, in this study, the
quality assurance and quality control (QA/QC) procedures are considered
together. The following QA/QC procedures and criteria for flux calculations
were taken into account:
calibration of the analyzers used for gases and particles (Sect. 3.3) – for
the GR and EC methods; criteria for the minimum concentration difference between the measurement
heights (Fig. S2 in the Supplement) – for the GR method; correction of the wind flow field around the measurement mast according to
the CFD calculations (Sect. 4.1) – for the GR method; restriction to open sea, i.e., wind direction in the range of 150–270 analysis of swell to determine the validity of M–O theory with codes 1–3
(Sect. 4.1) – for the GR method; the footprint area was estimated at each of the measurement height under at
neutral, stable, and nonstable conditions (Fig. 8) – for the GR and EC methods; stationarity criteria following the criteria of Foken and Wichura (Foken and
Wichura, 1996) – for the for GR and EC methods; the intermittency was applied according to Mahrt et al. (1998) – for the EC
method; WPL correction due to water vapor and heat flux – for the GR and EC methods; cross-sensitivity of the compounds on the used analyzers – for the GR and EC
methods; preparation of the uncertainty budget for the measurement results – for the
GR and EC methods.
The footprint area (i.e., the area upwind where the exchange of gases and particles between the air–sea surface are expected to be a source of the measurement results) was calculated according to Högström et al. (2008). The footprint area was calculated at each of the measured heights under stable, neutral, and unstable conditions. Figure 8a illustrates the relative intensity of the footprint area under neutral conditions as a function of upwind distance from the measurement mast at instrument heights of 4.7, 7.2, and 10.7 m. Stable and unstable conditions are presented in Fig. S3. The cumulative relative contribution (Fig. 8b) indicates that less than 0.3 % of the observed flux at the lowest height (4.7 m) takes place at a distance of 20 m from the mast, reaching 90 % at a distance of 3 km. At a height of 10.7 m, the footprint area starts at 40 m from the mast and reaches 85 % at distance of 3 km. The storage fluxes were not considered in this campaign. The site was by the sea where the turbulent mixing was most likely the main driving force for gas and particle dispersion most of the time.
The stationary requirement was calculated using the method proposed by Foken
and Wichura (1996) for fluxes of CO
Cross-sensitivity of the compounds (e.g., water vapor) on the response of
the analyzers used are included into the uncertainty budget or corrected
directly in the results (see Fig. S1). The influence of NO and NO
The uncertainty sources of the measurement results for fluxes using the
gradient method are presented in more detail in Table S1. To
estimate the uncertainty of the momentum flux and CO
Estimated relative expanded uncertainties for the fluxes of gaseous
CO
The eddy diffusivities
Dispersion of a ship plume is schematically presented in Fig. 10a. The black
curves denote the edges of the plume. When the lower curve reaches the sea
surface, the pollutants would be reflected from the sea surface if there were
no pollutant flux to the sea. Near the surface, the sum of the incident and
reflected concentrations will add up to the same value if the reflection is
total, forming a new boundary layer where the vertical
concentration gradient of the pollutants vanishes. If there is a flux to the
sea surface, this will result in a vertical gradient of pollutants in this
new boundary layer. The fluxes to the sea from the ship can be measured when
the plume is over the flux footprint if the measurement instrument is inside
the new boundary layer. As an example, Fig. 10b illustrates the momentary
plumes at the sea surface for a ship traveling to the city of Helsinki and
passing Harmaja Island at a speed of 21.5 kn (
Before the calculation of fluxes of gases and particles using the GR method, it
is necessary to evaluate the measured concentration differences between the
measurement heights as well as their uncertainty limits. The uncertainty of the
gas and particle analyzers as a function of concentration is presented in
Fig. S1, and the concentration differences between the
measurement heights is shown in Fig. S2. Based on the analysis, the concentration
differences for
Figure 11a illustrates the time series for the CO
Time series of 30 min fluxes for CO
Assuming that there is no inversion below the ships' chimney height,
profile measurements of the compounds indicate that the deposition of
nanoparticles towards the sea surface is most probably caused by ship
emissions (Fig. 11c). More detailed analysis was made to compare the fluxes
of CO
From the campaign in 2012, the CO
The WPL correction performed to both of the data sets in 2011 and 2012 was small
in general (see Fig. S5). In 2011, the sign of the flux was changed after
WPL correction during the period between 28 and 31 August. Generally,
positive CO
Unfortunately, direct comparison between the CO
The FSC was determined in the measurement campaigns in both 2011 and 2012. The two campaigns differed from each other with respect to the measurement strategies as well as with respect to the data collecting frequency, once per minute in 2011 and every 15 s in 2012. In both cases, the data acquisition system calculated the averages over the data collection period. It became very clear that the frequency of once per minute was too low in order to see accurate emission peaks in the ship plumes, as the duration of the plume itself was of the order of a few minutes. This was the reason for shortening the response time of the analyzers by increasing the flow rate from the nominal flow and shortening the integration time, which then made it possible to increase the data collection frequency.
As seen from Fig. S6, major factors influencing the accuracy
of calculating the emission peak area are the difference in the response
time between the SO
The variation between the 2011 and 2012 campaigns with respect to the calculated FSC from
the ships that routinely cruise between Helsinki and Stockholm or Helsinki
and Tallinn is used when estimating the uncertainty in the FSC according to
Eq. (13). In Fig. 13, the relative expanded uncertainty for the FSC,
The expanded uncertainty of FSC,
No violations of the regulations were observed for the calculated FSC
during the campaigns in 2011 and 2012 (Fig. 14). A typical FSC value of 0.4 % was obtained with an uncertainty of 15 %, i.e.,
Peak height concentrations of SO
Direct exchange of gaseous compounds and nanoparticles between the air and sea
interface was studied using micrometeorological methods. The gas compounds
SO
It became quite clear that no direct gas exchange across the air–sea
interface (negative or positive fluxes) could be measured using the GR method.
This was mostly due to the fact that the capability of the analyzers used to measure the gas
concentration differences under clean coastal conditions was not sufficient.
Even though the CO
Both the GR and EC methods were capable of measuring the emissions from the ships. Much effort was invested in studying the transport and dispersion of single ship emissions. Different scenarios depending on the wind speed and wind direction were able to identify the following: (i) pollutants reached the footprint area and the measurement mast or (ii) pollutants bypassed the footprint area but were seen by the measurement mast. When the mixing of the pollutants occurred well before the footprint area for the measurement mast, the measured fluxes were real. When the mixing of the pollutants from the ships was not complete, the M–O theory was violated, and the measurement results described dispersion of the pollutants
The measurements for determining the FSC were in good agreement with the information given by the ship owners. The measurement method used to determine the FSC content of marine fuel from the ambient air in connection with the identification data from AIS gives a clear demonstration of whether the regulations are respected.
The uncertainty analysis of the fluxes and the FSC was conducted according to the well-known law of propagation of errors, and following the recipes from the literature (JCGM, 2008). The uncertainty budget, which defines the sources of uncertainties and their contribution to measurement results, may be conservative; however, it can be made more accurate by selecting a better measurement technique, more capable analyzers, a homogeneous measurement site, and stationary meteorological situations.
To improve the accuracy of the FSC based on measurements of
the ratio of SO
The data used in this study are available from the Zenodo data repository:
The supplement related to this article is available online at:
JW and LP designed the concept of the study. JW, LP, and TW performed the
measurements with help from TL, JH, TT, and MM. HP and KK were responsible
for the
The authors declare that they have no conflict of interest.
This work reflects only the authors' views, and INEA is not responsible for any use that may be made of the information it contains.Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The measurements were carried out during the SNOOP project, financed by the European Regional Development Fund, Cental Baltic INTERREG IV A Programme. The authors are very grateful to Aleksi Malinen at Metropolia University of Applied Sciences for help with the measurements and to Kaisa Lusa and Sisko Laurila at the Finnish Meteorological Institute for help with laboratory calibrations. For revision of the language, the authors are very grateful to Leena Kahma. The editor and anonymous referees are thanked for their critical and insightful but constructive comments that considerably improved the final paper.
This research has been supported by the European Union's Horizon 2020 program (grant no. 814893).
This paper was edited by Drew Gentner and reviewed by three anonymous referees.