Effects of global ship emissions on European air pollution levels

Abstract. Ship emissions constitute a large,
and so far poorly regulated,
source of air pollution. Emissions are mainly clustered along
major ship routes both in open seas and close to densely
populated shorelines. Major air pollutants emitted include sulfur
dioxide, NOx, and primary particles. Sulfur and
NOx are both major contributors to the formation
of secondary fine particles (PM2.5) and to
acidification and eutrophication. In addition, NOx is a
major precursor for ground-level ozone.
In this paper, we quantify the contributions from international shipping
to European air pollution levels and depositions. This study is based on global and regional model calculations. The
model runs are made with meteorology and emission data
representative of the year 2017 after the tightening of the
SECA (sulfur emission control area) regulations in 2015 but
before the global sulfur cap that came into force in 2020.
The ship emissions have been derived using ship positioning data.
We have also
made model runs reducing sulfur emissions by 80 % corresponding
to the 2020 requirements. This study is based on model
sensitivity studies perturbing emissions from different
sea areas: the northern European SECA in the North Sea
and the Baltic Sea, the Mediterranean Sea and the Black Sea, the Atlantic
Ocean close to Europe, shipping in the rest of the world, and
finally all global ship emissions together. Sensitivity studies
have also been made setting lower bounds on the effects of ship plumes
on ozone formation. Both global- and regional-scale calculations show that for
PM2.5 and depositions of oxidised nitrogen and
sulfur, the effects of ship emissions are much larger when emissions occur
close to the shore than at open seas. In many coastal countries,
calculations show that shipping is responsible for 10 % or more of the
controllable PM2.5 concentrations and depositions of oxidised
nitrogen and sulfur. With few exceptions, the results from the global and
regional
calculations are similar.
Our calculations show that substantial reductions
in the contributions from ship emissions to PM2.5 concentrations
and to depositions of sulfur can be expected in European coastal regions
as a result of the implementation of a 0.5 % worldwide limit of the
sulfur
content in marine fuels from 2020. For countries bordering the
North Sea and Baltic Sea SECA, low sulfur emissions have
already resulted in marked reductions in PM2.5 from shipping
before 2020. For ozone, the lifetime in the atmosphere is much
longer than for PM2.5, and the potential for ozone formation
is much larger in otherwise pristine environments. We calculate
considerable
contributions from open sea shipping. As a result, we find that
the largest contributions to ozone in several regions and countries
in Europe are from sea areas well outside European waters.


farmtype model in CAPRI (Common Agricultural Policy Regional Impact). ECLIPSEv6a emissions are available in 5-year intervals from 2005 onward. In this study the emissions are interpolated to 125 2017.
The land-based emissions used in the regional model calculations are described in Gaisbauer et al. (2019) and are mainly based on the officially reported data from the countries. In Table 1 these officially reported emissions are listed aggregated for the EU27 countries compared to the ECLIPSEv6a emissions. Differences are of similar magnitude for the individual EU countries. The 130 most significant difference is for sulfur, where the ECLIPSEv6a emissions are of the order of 15% higher than those reported to EMEP.
Ship emission data sets used in both the global and regional model calculations are originally from the Finish Meteorological Institute, based on AIS data processed in the STEAM model (Johansson et al., 2017) and downloaded from the ECCAD database (https://eccad.aeris-data.fr/). Ship emissions 135 of various species, based on the global data set, are listed in Table 1 separately for the Baltic Sea, the North Sea (including the English Channel), the Mediterranean Sea and the Black Sea. In addition emissions are listed for the remaining Atlantic area outside Europe, but bounded by 30 -82 degrees north and 30 degrees west to 90 degrees east corresponding to the "Northeast Atlantic Ocean" also included in the regional calculations. These three sea areas are depicted in Figure 1. Finally 140 emissions are also listed for the total global sea area. Annual ship emissions used in the regional model calculations are based on the same source (Gaisbauer et al., 2019). Even so, ship emissions used in the global calculations are somewhat higher than in the regional calculations (see EMEP Status Report 1/2019 (2019), appendix B).
In the FMI emission data all PM emissions are assumed to be emitted as PM 2.5 . Emissions from 145 leisure boats are not included. In a separate study Johansson et al. (2020) have quantified the emissions from leisure boats in the Baltic Sea only. Compared to emissions from the commercial fleet these emissions were insignificant for NO x and PMs. However, in regard to emissions of NMVOC the study concluded that these can be significantly larger from leisure boats than from registered vessels in the Baltic Sea, especially during summer (about 500% larger). However, as shown in Table 1, 150 land-based NMVOC and NO x emissions are of similar magnitude, the NMVOX to NO x ratio is very small for ship emissions.

Definition of the model sensitivity tests
In order to calculate the effects of ship emissions on air pollution and depositions in Europe we use a similar approach as in the SR (Source Receptor) calculations within the EMEP programme  Table 2. We have combined the North Sea and the Baltic Sea into one scenario run because they are both designated as SECA areas. Likewise we have combined the Mediterranean Sea and the Black Sea. The sea areas are shown in Figure 1. ROW (Rest Of World) are all sea areas not included in the sea areas listed above. We have also made additional model runs 165 with sulfur emissions from ships reduced to CAP2020 levels.
In the interpretation of the model results below we let the difference between the Base_2017 and the SR_AllAnt model runs (see Table 2) represent 100% of the effects of all anthropogenic, and thus controllable, global emissions. Similarly we calculate the contributions from global shipping as a whole, or from shipping in a specific area, by subtracting the scenario run for shipping as a whole 170 or from a specific sea area from the Base model run. In this way we can relate the effects of ship emissions in different regions to the total anthropogenic contribution. Even though not strictly linear, this is a widely used approach that was also taken in the TF_HTAP phase II modelling exercise (see workplan under http://www.htap.org/). For all depositions and air concentrations except ozone (and ozone metrics) we add up the SR runs for the individual sea areas (SR_BALNOS, SR_MEDBLS, 175 SR_ATL and SR_ROW) and compare with the SR_AllSh emission perturbation providing a measure of the linearity in the calculations.
In the model calculations described above, the ship emissions are instantly diluted throughout the model grid cells in which the emissions occur. Previous studies (Vinken et al., 2011;Huszar et al., 2010) have shown that this can lead to an overestimation of ozone formation, in particular in sea 180 areas where NO x concentrations are otherwise low. The EMEP model has an option for splitting 50% of the NO x emissions from shipping into a pseudo-species "ShipNOx", see Simpson et al. (2015).
ShipNOx deposits as NO 2 , but undergoes simple atmospheric reactions: Reaction R1 proceeds with the same rate as the normal NO 2 + OH reaction, thus proceeding faster in daylight and in high OH areas. Reaction R2 provides a minimum half-life of about 6 hours, loosely based upon results shown in Vinken et al. (2011). We have repeated the calculations for the scenarios listed above with the ShipNOx reactions included. We then assume that the calculations with and without the ShipNOx split represent a lower and an upper limit of the effects of NO x emissions from 190 shipping on the formation of ozone both globally and in the individual sea areas.

Model results
In this section we show the calculated effects of all global ship emissions, and the effects of emissions from separate sea areas as defined in the separate scenarios in Section 2.2 . For ozone we also include 6 https://doi.org/10.5194/acp-2020-293 Preprint. Discussion started: 24 April 2020 c Author(s) 2020. CC BY 4.0 License. a discussion on the effects of the ShipNOx split and for PM 2.5 we include the effects of the CAP2020 195 regulations.
3.1 PM 2.5  Figure 3. The main sources for particles and particle formation from shipping are NO 2 and sulfur (of which more than 95% is emitted as SO 2 in the gas phase, and the rest as sulphate particles). In addition ash, EC (Elemental Carbon) and OC (Organic Carbon) are assumed emitted as primary particles. The main oxidation paths for SO 2 are the OH reaction in the gas phase and in-cloud oxidation (mainly with H 2 O 2 ). Both these oxidants have a clear summer maximum, contributing to a 210 summer maximum also for sulphate. In sea areas outside the SECAs sulphate makes up 50 to 80% of the PM 2.5 , dry mass (Figure 8a), explaining the summer maximum in PM 2.5 concentrations in most sea areas.
NO 2 is oxidised to gaseous HNO 3 . HNO 3 can then react with sea salt forming particulate sodium nitrate. However, in the presence of ammonia the formation of ammonium nitrate particles can be a 215 lot faster. The latter reaction requires a surplus of NH 3 over sulfate. Ammonia is mainly emitted from agriculture with a seasonal maximum in spring. In the SECAs, where sulfur emissions from ships are very low, we calculate that nitrates make up almost 50% of the PM 2.5 dry mass.
In addition both sulfate and nitrate from shipping results in an increase in ammonium (ammonium nitrate and ammonium sulfate), About 20 to 30% of the PM 2.5 mass over Europe is ammonium.

220
The effects of the emissions from individual sea areas on PM 2.5 discussed below are based on 2017 ship emissions. The effects of the CAP2020 global reductions in sulfur emissions from ships are described in Section 4.  Table 1). In particular the southwestern parts of this sea area are close to some of the highest ammonia emission regions in Europe. The main source of particles from shipping is NO 2 through the formation of nitrate, predominantly ammonium nitrate. The spring maximum in 7 https://doi.org/10.5194/acp-2020-293 Preprint. Discussion started: 24 April 2020 c Author(s) 2020. CC BY 4.0 License. emissions, mainly from agriculture, peaking in spring.

Contributions from the Northeast Atlantic Ocean
The largest contributions to PM 2.5 concentrations in Europe from shipping in the Northeast Atlantic (see Figure 3 e,f,g,h) are calculated for the regions bordering the ship track in and out of the Mediterranean through Gibraltar, extending north to the English Channel. As this region is outside the 235 SECA, sulfur emissions are high, and a major constituent in PM 2.5 from shipping is sulfur, emitted mainly as gaseous SO 2 and then oxidised to sulfate. The summer maximum in the contributions from the Northeast Atlantic is mainly caused by sulfate.

Contributions from the Mediterranean Sea and the Black Sea
The largest contributions to PM 2.5 concentrations from shipping in the Mediterranean and Black Given the large distance to the European continent, contributions to European PM 2.5 levels from ROW shipping are small.

Country attributions
The source receptor relationships for shipping (total and from separate sea areas) are listed in Table 3 for selected countries. Here we also list the corresponding source receptor results as reported in the  Net formation of ozone depend on the ratio between NO x and NMVOC. In regions with high NO x concentration ozone production is limited by the availability of NMVOC, and further enhancements

Ozone
of NO x will lead to increased ozone titration, and thus reductions of ozone, predominantly in the winter months. In summer additional NMVOC emissions from leisure boats may lead to an increase 285 in ozone levels in such areas. In areas limited by the availability of NO x additional NO x will result in increased ozone production, predominantly in the summer months.

North Sea and Baltic Sea
In the North Sea and Baltic Sea regions (Figure 5a,b,c,d), ship emissions contribute to widespread ozone titration in all four seasons. The strongest titration effects are calculated in winter and the least 290 in summer.

Mediterranean Sea and Black Sea
In the Mediterranean Sea and the Black Sea there is widespread ozone titration close to major shipping lanes and ports in winter (Figure 5i,j,k,l). However, in Spring ozone production starts to dominate, reaching a maximum in summer with contributions from shipping of more than 4ppb in the eastern 300 Mediterranean sea and bordering land areas.

Rest of world shipping
Emissions from Rest Of World shipping affects all of Europe, but western and northern Europe more are comparable, and in some regions higher, than contributions from the other sea areas.

Country attributions
For SOMO35 3 the source receptor relationships for shipping (total and from separate sea areas) are listed in Table 4 for selected countries. We also list the corresponding source receptor calculations as reported in the latest EMEP report (EMEP Status Report 1/2019, 2019) and these results are discussed  4 Effects of CAP2020 on European PM 2.5 levels and on sulfur depositions From January 1th 2020 the maximum allowed sulfur content in marine fuels was reduced to 0.5% (CAP2020). Before CAP2020 the global average sulfur content outside SECAs was around 2.5% 345 although a higher percentage sulfur content of 3.5% was allowed. regulations. However, in sea areas outside the SECA, and in land areas bordering these sea areas, sulfate is the major component in PM 2.5 origination from ship emissions (Figure 8a).
To give an estimate of the effects of CAP2020 on European PM 2.5 levels and the depositions of oxidised sulfur we have made calculations reducing sulfur emissions outside the North Sea and the Baltic Sea SECAs by 80%, corresponding to a reduction from 2.5% to 0.5% in the sulfur content 355 in the fuels. This is a crude estimate, as there are low emission ships operating outside the SECAs.
On the other hand CAP2020 compliance may not reach 100%. Furthermore we have assumed 80% reductions in sulfur emissions also in low emissions zones far from European waters. But, as already shown in Figure 3, emissions outside European waters (ROW shipping) has little or no effects on European PM 2.5 levels. Figure 8b shows the calculated effects of CAP2020 on European A similar pattern as PM 2.5 is seen in Figure 7 for oxidised sulfur depositions, with substantial reductions in depositions of anthropogenic origin in countries bordering sea areas that are not NECAs as the Mediterranean Sea and the Northeast Atlantic.
5 Differences between regional and global model calculations 375 The regional model calculations as reported in the annual EMEP reports (exemplified by the latest EMEP report, EMEP Status Report 1/2019 (2019) In general the results from the global and the regional model calculations are in good agreement. Even so there are some systematic differences in the model results. We have tried to trace these to differences listed below in model input and model setup, and to what extent global and regional calculations could give qualitatively and quantitatively different results for the effects of ship emissions. 3. In the global model we reduce the emissions by 15% for all species in the sea areas simulta-390 neously, whereas in the regional calculations emissions of the individual species are reduced separately.
4. The resolution used in the global and regional model calculations differ.
5. In the regional calculations the boundary and initial conditions for all gaseous and aerosol species were given as 5-year monthly average concentrations, derived from EMEP MSC-W 395 global runs.
Bullet points 3 and 4 were a compromise to keep the computational demand of the global calculations within reasonable limits. Below we discuss the effects this makes for different components in detail. We also make statements on the processes behind these difference, which is of relevance also beyond this study. are consistently lower in the global versus the regional calculations (see Table 3). Most countries bordering the Baltic Sea and the North Sea are high emitters of ammonia. SO 4 (either emitted directly or oxidised from SO 2 ) can react with ammonia forming ammonium sulfate. Much of the emitted 405 NO 2 will form HNO 3 . Given ammonia in excess of SO 4 , HNO 3 will react with ammonia forming ammonium nitrate. As shown in Table 1  In several countries PM 2.5 levels from shipping are markedly higher in the global calculation, in particular in small countries such as Cyprus, and also in Portugal where the shipping lanes are very close to the shore. We believe this is caused by the lower resolution in the global calculations, which implies that grid boxes covering partially land and sea extend further inland, thus artificially 415 extending the effect of ship emissions somewhat further into these countries' territories.

Differences in nitrogen and sulfur deposition between global and regional model calculations
Above we argued that parts of the lower contributions from ships to PM 2.5 concentrations could be caused by less ammonia available for ammonium nitrate formation in the global calculations, 420 resulting in a higher HNO 3 to ammonium nitrate ratio. As the dry deposition of HNO 3 is faster than for ammonium nitrate, more oxidized nitrogen (mainly ammonium nitrate, HNO 3 , NO 2 ) is deposited in nearby countries where ammonia emissions are high.
In several countries both N an S depositions are higher, in particular in small countries such as Malta and Cyprus, and in Portugal where the shipping lanes are very close to the shore. As for PM 2.5 425 concentrations, we believe this is caused by a lower resolution in the global calculations as grid boxes covering partially land and sea extend further inland.

Differences in SOMO35
In Table 4 the contributions from ship emissions to selected countries are listed, both for the global and regional model calculations. Given the large compensating contributions from ozone titration, 430 13 https://doi.org/10.5194/acp-2020-293 Preprint. Discussion started: 24 April 2020 c Author(s) 2020. CC BY 4.0 License. mainly in winter, and ozone production, mainly in the summer months, SOMO35 calculated with the global and the regional model versions are remarkably similar. However, there are substantial differences, mainly confined to the very high NO x emitting regions bordering the North Sea.
In the global calculations there are substantial contributions from ROW shipping that can not be attributed in the regional calculations. As shown in Table 4 there are substantial contributions from 435 ROW, and in several countries ROW is the largest contributor.
With the ShipNOX parameterization included in the global calculations the contributions to SOMO35 from the sea areas is reduced by about 50% (see Figure 6) and considerably lower than in the regional calculations. ShipNOX is not used in the regional calculations, but the largest effects of ignoring the ship plume chemistry should be in low NO x areas with large gradients between the 440 plumes and ambient air most often found in pristine sea areas.

Conclusions
Emissions from shipping are large sources of air pollution and depositions of oxidised nitrogen and sulfur. In this study we have mainly restricted ourselves to the effects on European pollution levels, but the effects are global. In particular in coastal regions/countries, we attribute a large portion of in winter, and ozone production, mainly in summer. This is also the reason for the different behaviour of annual averaged ozone and the SOMO35 ozone metric. SOMO35 is hardly accumulated in winter when ozone titration events are most frequent as ozone levels in winter are regularly below the 35ppb threshold.
The lifetime of ozone in the atmosphere is considerably longer than for PM 2.5 ranging from hours 470 to a few days in the boundary layer to weeks and even months in the free troposphere (TF HTAP, 2010).
As a result ozone can be transported at intercontinental scales, explaining the large contributions from ROW shipping.
Global model calculation require substantially more computer power than regional calculations, and thus global scale source receptor calculations, even with a half a degree resolution, would not 475 be possible. The source receptor relationships derived from the global and regional calculations are similar. Where there are differences, these can largely be attributed to model setup and input data. Most of species levels, and the resulting surface depositions, highlighted in EMEP regional calculations are relatively short-lived. As a result the effects of emissions originating outside the be included. For several countries/regions we show that for ozone contributions from ROW shipping are comparable, and in some regions higher, than contributions from sea areas bordering Europe. In the regional model source receptor calculations bic (boundary and initial concentrations) only account for ozone 'produced within the regional model domain from NO x (emissions of NMVOC from shipping are very small) transported from outside the regional model domain.

490
The dispersion and chemistry in the shipping plumes represents an uncertainty in the calculations.
Calculations including the "ShipNOX" parameterisation short circuit the NO x chemistry so that only 50% of the emitted NO x enters the ozone cycle, and as a result the effect of shipping on ozone is also reduced by about 50%. Calculations with and without the "ShipNOX" parameterisation gives an upper and lower range for the effects of shipping on ozone. The largest effects of ship plume chemistry are 495 likely to occur where the gradients between ship plume and ambient air NO x concentrations are large.
Such conditions are less common in waters close to Europe.
Acknowledgements. This work has been partially funded by EMEP under UNECE. Computer time for EMEP model runs was supported by the Research Council of Norway through the NOTUR project EMEP (NN2890K) for CPU, and NorStore project European Monitoring and Evaluation Programme (NS9005K) for storage of data.