The evolution of shipping emissions and the costs of regulation changes in the northern EU area

An extensive inventory of marine exhaust emissions is presented in the northern European emission control area (ECA) in 2009 and 2011. The emissions of SO x, NOx, CO2, CO and PM2.5 were evaluated using the Ship Traffic Emission Assessment Model (STEAM). We have combined the information on individual vessel characteristics and position reports generated by the automatic identification system (AIS). The emission limitations from 2009 to 2011 have had a significant impact on reducing the emissions of both SOx and PM2.5. The predicted emissions of SO x originated from IMO (International Maritime Organization)-registered marine traffic have been reduced by 29 %, from 320 kt to 231 kt, in the ECA from 2009 to 2011. The corresponding predicted reduction of PM2.5 emissions was 17 %, from 72 kt to 61 kt. The highest CO 2 and PM2.5 emissions in 2011 were located in the vicinity of the coast of the Netherlands, in the English Channel, near the south-eastern UK and along the busiest shipping lines in the Danish Straits and the Baltic Sea. The changes of emissions and the financial costs caused by various regulative actions since 2005 were also evaluated, based on the increased direct fuel costs. We also simulated the effects and direct costs associated with the forthcoming switch to low-sulfur distillate fuels in 2015. According to the projections for the future, there will be a reduction of 87 % in SOx emissions and a reduction of 48 % in PM 2.5 emissions in 2015, compared with the corresponding shipping emissions in 2011 in the ECA. The corresponding relative increase in fuel costs for all IMO-registered shipping varied between 13 % and 69 %, depending on the development of the prices of fuels and the use of the sulfur scrubber equipment.


Introduction
It has been estimated in the recent literature that the upcoming Marpol Annex VI agreement will be costly for the shipping industry.The financial costs will increase from 25 % to 40 % within short sea-shipping lanes inside the northern European Sulfur Emission Control Area, due to the shift to marine gas oil (MGO) (0.1 %) fuel in 2015 (Notteboom et al., 2010).This cost increase will probably lead to changes in the modes of transportation.Possible consequences may be the reduction of capacity for short sea services and an increased cargo transfer by trucks; these changes may undermine the planned benefits associated with reduced marine emissions.However, the estimates of these consequences have up to date taken into account neither (i) the increases of fuel costs for individual ships or ship categories nor (ii) spatially and temporally accurate activity data of ships.
Emission abatement strategies that specify reduced fuel sulfur content will result in lower emissions of both fine particulate matter and SO 2 from ships.This in turn tends to decrease adverse health effects in human populations, especially within the riparian states and in coastal cities.Also, greenhouse gas emissions from shipping are an increasing concern.Various cost effective mitigation plans have therefore been suggested for CO 2 originated from shipping, using various policies and technological improvements.Corbett et al. (2009) estimated that fuel savings of up to 70 % per route could be achieved by halving the cruising speed of container ships, which would cause an equally dramatic decrease in CO 2 emissions from these vessels.However, the loading capacity and overall fleet size would probably need to be correspondingly increased (Corbett et al., 2009).

L. Johansson et al.: The evolution of shipping emissions
The auxiliary engines are responsible for a significant portion of the total fuel consumption, and any reduction in cruising speed will inevitably result in an increase in auxiliary fuel consumption.Further, the engine load affects emission factors and engine efficiency.Ultimately, in order to evaluate the overall feasibility of slow-steaming scenarios, the increase in total operational time for ships needs to be accounted and reflected on fuel consumption savings and the need for additional ships.
This study addresses the shipping emissions of the northern European emission control area (ECA), which includes the North Sea, the Baltic Sea and the English Channel, from 2011 to 2015.In the following, we refer to the northern European ECA simply as "the ECA".The first aim of this paper is to present an extensive inventory of shipping emissions in the ECA in 2009 and 2011.We have presented the predicted emissions of CO, CO 2 , SO x , NO x and PM 2.5 among different flag states and ship types.The high-resolution geographical distribution of CO 2 and PM 2.5 emissions has also been presented.The second aim of this paper is to present the results of model simulations for selected scenarios, assuming different regulations for the fuel sulfur limits, the reductions of the cruising speeds, and the installations of sulfur scrubbers.For each of these scenarios, we have evaluated the respective impacts on shipping emissions and fuel costs.In particular, the direct fuel costs and emission reductions have been evaluated for the forthcoming Marpol Annex VI requirement, according to which there will be a shift to 0.1 % MGO fuel in 2015.

Methods
The emissions presented in this paper were evaluated using the Ship Traffic Emission Assessment Model (STEAM).A brief overview of this model is presented in the following; for a more detailed description, the reader is referred to Jalkanen et al. (2009Jalkanen et al. ( , 2012Jalkanen et al. ( , 2013)).

The STEAM model and its input values
This modelling approach uses as input values the position reports generated by the automatic identification system (AIS); this system is globally on-board every vessel that weighs more than 300 t.The AIS system provides automatic updates of the positions and instantaneous speeds of ships at intervals of a few seconds.For this paper, archived AIS messages provided by the North Sea and the Baltic Sea riparian states in 2009 were combined, covering the entire ECA.In order to avoid the processing of an excessive amount of data, the AIS message set used in this study has been down-sampled; the temporal separation between messages is commonly 6 min.The combined data set for 2009 however, still contains more than 552 million archived AIS messages.For the ECA in 2011, AIS messages were extracted from a data set given by the European Maritime Safety Agency (EMSA).This extracted data set contains 607 million archived AIS messages.
The model requires as input also the detailed technical specifications of all fuel consuming systems on-board and other relevant technical details of the ships for all the ships considered.Such technical specifications were therefore collected and archived for over 50 000 ships from various sources of information; the data from IHS Fairplay (IHS, 2012) was the most significant source.
The STEAM model is then used to combine the AISbased information with the detailed technical knowledge of the ships.The model predicts as output both the instantaneous fuel consumption and the emissions of selected pollutants.The fuel consumption and emissions are computed separately for each vessel; by using archived regionalscale AIS data results in a regional emission inventory.The STEAM emission model allows for the influences of the high-resolution travel routes and ship speeds, engine load, fuel sulfur content, multiengine set-ups, abatement methods and waves (Jalkanen et al., 2012).

Model performance and uncertainty considerations
The model has been able to predict aggregate annual fuel consumption of a collection of large marine ships with a mean prediction error of 9 % (Jalkanen et al., 2012).Largescale comparisons to ship owner fuel reports have been constrained by the availability of vessel fuel reports, but have so far been done for a data set of 20 vessels.The capability of the model for estimating instantaneous power consumption has been evaluated to be moderately less accurate, compared with the corresponding accuracy for predicting the fuel consumption, with a mean prediction error of 15 % in a thorough case study (Jalkanen et al., 2012).The evaluated emissions agree fairly well with the results of several measurement campaigns presented in literature, for various engines, engine loads and pollutants.A more detailed description of the model evaluation studies have been presented in Jalkanen et al. (2009Jalkanen et al. ( , 2012)).Model uncertainties have been previously assessed in Jalkanen et al. (2013).
Accurate modelling of emissions with the presented method requires that (i) the vessel routes and shipping activities are evaluated correctly, (ii) the instantaneous power requirements of ships are successfully evaluated and (iii) the resulting fuel consumption and emissions are accurately predicted.Considering each of these three consecutive steps, the following sources of uncertainty can be identified.These uncertainties correspond to regional scale emission inventories, as compiled in this study.

Ship routes and harbour activities
High geographic accuracy (tens of metres) of shipping routes can be expected, due to the GPS based location signaling.The temporal and spatial coverage of archived AIS messages was good in the ECA.Therefore there is only a very small fraction of route segments that cross land masses, such as peninsulas or islands.
Accurate modelling of maneuvering activities in harbour areas would require a data set with more frequent (several times per minute) dynamic updates, as the speed of vessels can change frequently and rapidly.We applied in this study down-sampled AIS messages on 6 min intervals.Furthermore, the use of auxiliary engines for ships at berth is difficult to predict as, in contrast to main engines, detailed engine specifications of auxiliary engines are not commonly available.In some, cases however, auxiliary engine information has been augmented with data from classification societies.We estimate that moderate to high uncertainty can be associated with harbour emissions within regional emission inventories.

The characteristics of vessels and fuels
The ship characteristics database includes detailed information for more than 50 000 ships with a unique IMO (International Maritime Organization) identification number.However, the number of unidentified ships without an IMO number has been increasing steadily.For instance, the unidentified ships were the second largest ship type category in terms of the number of ships in the ECA in 2011.All unidentified ships are presumed to be small vessels, and we have treated those in the modelling by assuming only generic specifications (weighting 500 t with a single 1000 kW four-stroke engine).The emissions originated from unidentified vessels are therefore known with a significantly lower accuracy.
The fuel type and especially the fuel sulfur content (FSC), affects significantly the SO x and PM 2.5 emissions.We assume that all ships conform to ECA sulfur limits.Considering that ship owners have economic incentive to use fuel grades, which have the maximum allowed FSC, we can estimate that the uncertainty arising from fuel type evaluation is fairly small.However, some engines may use fuel with even lower FSC than the allowed maximum, for technical reasons.This causes additional uncertainties in the evaluation of the emissions, especially for the estimation of fuel type used in auxiliary engines.

The emissions of various species
We evaluate that the estimated CO 2 emissions have the lowest margin of error, compared with those of the other modelled species, as the amount of CO 2 per fuel burned can be estimated fairly accurately.Also the NO x emission factor, which is almost unaffected by engine load and fuel type, can be estimated with a relatively good accuracy.We use Tier I and II NO x limits for vessels, depending on the year they were built.There may therefore be some underestimation of NO x for old ships that are not obliged to conform with Tier I requirements.
The conversion rate of fuel sulfur to SO 4 , the main component of PM 2.5 emissions, has been assumed to be independent of engine load.However, some recent studies suggest that this conversion rate may be affected by engine load (Petzold et al., 2010).Numerical computations with the model have indicated that conversion rates for SO 4 as presented by Petzold et al. (2010) would significantly reduce the estimated emissions of SO 4 (up to 50 % in mass).Furthermore, the emissions of organic and elemental carbon, as well as ash particles, have been assumed to be unaffected by the fuel type; this assumption may prove to be inaccurate.The highest margin of error is expected with estimated CO emissions, as the emission factor has been observed to be highly sensitive to engine load and its rapid changes.

Evaluation of fuel sulfur content in case of fuel conversion and switching, and exhaust gas cleaning systems
Clearly, the fuel sulfur content significantly affects the PM 2.5 and SO x emissions per amount of fuel burned.The emissions of particulate sulfate (SO 4 ) included in the PM 2.5 emissions are assumed to have a linear dependency with FSC.The other modelled components (ash, elemental and organic carbon particles) are unaffected by FSC (Buhaug et al., 2009;Jalkanen et al., 2012).The remaining sulfur in the fuel, which has not been converted to sulfate, contributes to SO x emissions.
In the ECA region, since the beginning of 2010, the maximum allowed FSC in inland waterway vessels and for ships at berth has been restricted to 0.1 %; however, the latter regulation applies only to vessels which are berthing for more than 2 h.Otherwise, the maximum FSC has been limited to 1.0 % since July 2010.Ship operators have several options for complying with FSC requirements, such as (i) fuel conversion, (ii) fuel switching and (iii) exhaust gas cleaning systems (EGCS).In fuel conversion, all fuel storage tanks, piping systems and combustion equipment are converted to be compatible with low sulfur fuel, which is to be used in all situations.In fuel switching, a secondary low sulfur fuel storage and piping system is installed and low-sulfur fuel is switched on when the ship operates inside the ECA area.The switching process, however, may take a considerable amount of time as the switched fuel needs to be warmed (heavy fuel oil, HFO) or cooled (MGO) before use.Hence the requirement for 0.1 % FSC for ships at berth is applied only for the ships that berth longer than 2 h.For ships using EGCS instead of low sulfur fuel, the amount of exhausted SO x and particle matter is not allowed to exceed the amount that would be exhausted by burning fuel with acceptable FSC.
In the STEAM model, FSC is determined separately for main and auxiliary engines, by taking into account engine specifications and region specific limitations such as, e.g. the EU shipping sulfur directive.The process of fuel type modelling in STEAM, including FSC, grade and cost, is illustrated in Fig. 1.All vessels are assumed to use the cheapest accepted fuel available (commonly this is also the heaviest fuel).The fuel sulfur content is therefore assumed to be where FSC C is the maximum FSC that the engine can use and FSC A is the maximum FSC allowed by the regulations in the considered area.However, if the ship has been equipped with EGCS, then FSC A in Eq. ( 1) is evaluated to be equal to the (relatively higher) sulfur content that would, after gas cleaning, result in acceptable emissions of both SO x and PM 2.5 .In such a case, FSC A in Eq. ( 1) is therefore substituted with the fuel sulfur content before exhaust gas cleaning SC A , which is evaluated from where η SO x and η PM 2.5 are the EGCS's efficiencies in reducing the emissions of SO x and PM 2.5 , respectively.These efficiencies are within the interval [0,1].FSC C is estimated by using the engine's power output rating and engine angular velocity, measured as revolutions per minute (RPM), based on manufactured marine engines statistics presented in Kuiken (2008).Based on these statistics we assume that all main engines with a larger power output than 4500 kW (and engine RPM < 1000) can use the heaviest fuel grades; engines smaller than 2000 kW use 0.5 % MDO fuel and otherwise FSC C is estimated to be 1.0 %.However, according to ship specifications in our database, more than 17 000 ships can be assumed to be equipped with a shaft generator which allows auxiliary power to be produced with the main engines at cruising speed.Thus, if a vessel with a shaft generator has a speed greater than 2.5 m s −1 (5 knots), we assume that all auxiliary power will be produced with main engines; clearly, these use FSC that is associated with the main engines.
The maximum allowed FSC is determined based on region, date and speed.Vessels having a speed lower than 0.5 m s −1 (1 knot) continuously for at least 2 h are assumed to be berthing, resulting in a FSC of 0.1 % in the ECA since the beginning of 2010.

Evaluation of fuel prices and exhaust gas cleaning systems
Combining the fuel consumption and FSC modelling allows us to evaluate fuel costs for each ship using the STEAM model.According to marine fuel bunker statistics, at the port of Rotterdam the current low sulfur marine gas oil (LSMGO with 0.1 % FSC) price in January 2013 was USD 960 per metric ton, whereas heavy fuel oil (HFO380/180) costs approximately USD 611 per metric ton (Bunkerworld.com, 2012).The price of intermediate fuel oil with a maximum FSC of 1.0 % (LS180/380) fuel is priced at USD 668 per ton.
The price premium between HFO and LSMGO as well as their overall price development over time has proven to be highly volatile.For instance, the average price premium between HFO380 (max.4.5 % FSC) and LSMGO between 1995 and2009 has varied between 50 % and 140 % in Rotterdam (Notteboom et al., 2010).Three different price developments for MGO with respect HFO were used in the selected scenarios: 50 % price premium over HFO (FC is fuel cost) (FC50 %), 75 % price premium (FC75 %) and 100 % premium (FC100 %).
According to Notteboom et al. (2010) the FSC in the heaviest and cheapest fuels available can be assumed to be no larger than 2.7 % as the world average of sulfur content in HFO fuels is 2.67 %.We assume that vessels use a mixture of fuels which have an arbitrary average FSC between 2.7 % and 0.1 %, so that the evaluated FSC given by Eq. 1 is achieved.The price estimate of this mixture of fuels is then computed as a function of sulfur content, according to regression curves presented in Fig. 2.
The three price functions in Fig. 2 correspond to the current state and two future price development possibilities: the  Fig. 2. Estimated fuel prices (USD/ton) as a function of the sulfur content of fuel, for three different fuel cost (FC) scenarios.The scenarios correspond to the current state (FC50 %) and two future price (FC75 % and FC100 %) scenarios; these have been defined in the text.The numerical equations of the fits have also been reported.FC50 % curve corresponds to prices (HFO380, LS180 and LSMGO) as they were at the time of writing, in Rotterdam, FC75 % and FC100 % are the price estimates in case the price premium between LSMGO and HFO380 increases to 75 % and 100 % respectively.We apply these fuel prices for all past and future scenarios presented in this paper; the derived fuel costs (and thus the direct costs of regulations to ship owners) of each scenario are therefore comparable with each other.
The use of EGCSs offer potential fuel cost savings for ships that operate in the ECA area, as IMO accepts EGCSs as alternatives to the use of low sulfur fuels.With a scrubber on-board, a ship can consume high FSC fuel and still comply with regulations.In Reynolds (2011) it was estimated that any ship, which consumes annually more than 4000 metric tons of fuel in the ECA, should be a potential candidate for an EGCS installation.Assuming a 50 % price premium for LSMGO with respect to HFO and active use within the ECA for at least 6 yr after 2015, the net financial value for EGCS scrubber installment should be positive.
Scrubbers can use wet or dry physical scrubbing or chemical adsorption to remove combustion products.In Corbett et al. (2010) it was concluded that the PM 2.5 removal is likely to be 75 ± 15 % with a scrubber on-board.Other studies have indicated that the resulting reduction in PM mass can be between 25 % and 98 %, depending on particle size distribution, although the removal rates by species are more uncertain (Lack and Corbett, 2012).Also, a significant reduction in SO x output will occur.In Andreasen and Mayer (2007) it was estimated that a sea water scrubber system can reduce 66 % of SO x emissions.

Interpolation of shipping routes
In the STEAM model, the travel routes are evaluated in a stepwise manner, by a linear interpolation of the geographical coordinates, for each consecutive AIS message pair.Due to this method of determining routes, it is useful to analyse in addition the validity of each travel segment.The calibration and use of AIS transmitters is also potentially susceptible to human errors.Especially as smaller ships without an IMO number behave erratically in some cases, based on the geographic information included in their AIS messages.Further, in order to ensure a good accuracy of the method, at open sea fairly extensive spatial and temporal gaps can be allowed, whereas at harbours the possible AIS down-time of ships (i.e. the interval between an end of a berthing activity and the start of cruising) needs to be substantially shorter.The methods for the evaluation of route segments were therefore refined for this study.
The validity of each linear route segment has been evaluated based on the average vessel speed v a given by two consecutive AIS messages: the time duration t, which is computed from message time stamps, and the distance s, which is calculated from the two message coordinate pairs.In addition, two other evaluation measures are used: the socalled implied speed, defined as v I = s/ t, and implied distance, defined as s I = v a t.The emission is computed for any route segment, if and only if the following three conditions are satisfied.
-The ship is physically able to travel the distance during the time interval in view of the specified design speed of the vessel.This criterion is confirmed if v a or v I is not significantly greater than the vessel's listed design speed.
-The temporal or spatial separation of a route or berthing segment does not exceed pre-selected maximum values.These maximum values have been specified separately for harbour activities and open sea activities.For each segment in the ECA, we have used the maximum values of 600 km and 24 h for open sea operations and 2 h for berthing activities.
-The vessel would not travel multiple times (or just a fraction of) the distance s within the given v a and t.Thus, s I must be close to s.

Slow-steaming
Required propelling power for any marine vessel increases strongly as a function of its speed, due to the friction against water and the formation of waves.Even a minor reduction of vessel speed can therefore significantly reduce the main engine's fuel consumption.The concept of slow-steaming refers to a situation, in which a marine vessel reduces its speed to achieve significant fuel savings.However, the fuel savings and emission reductions are obviously obtained at the expense of a longer cruising time.
In order to evaluate the net benefits in the selected slowsteaming scenario, the total travel time differential is calculated for each route segment.We assume a fractional speed reduction with a factor of a ∈ [0, 1].The increase in travel time T + , the reduced slow-steaming speed v iR and the increased duration t iR are given by where t i is the duration of the travel of the ship during the ith segment of a route (defined by two consecutive AIS messages) and v i is the average speed in i-th segment of a route before applying speed reduction.t iR is the increased duration of travel with the slow-steaming speed.The reduced speed v iR is used for instantaneous main engine power estimation, which in turn is used for engine load, fuel consumption and subsequently for emission estimation.To account for the fact that engines are being used longer with each segment using the reduced speed, the duration t iR is used instead of t i in emission calculation.Besides the instantaneous speed, the main engine power requirement is affected by various ship attributes, such as hull dimensions and propeller properties.This fairly complicated process was discussed in more detail in Jalkanen et al. (2012).

Auxiliary fuel consumption of non-IMO registered vessels
The number of unidentified vessels in AIS data has steadily increased during recent years.According to AIS data, a substantial fraction of these vessels seem to be inactive; these are mostly berthing.Such vessel behaviour in the model would result in an excessive amount of auxiliary fuel consumption, especially as the number of berthing small vessels increases in time.
We have therefore added to the model a limiting rule for the auxiliary fuel consumption of non-IMO-registered vessels.After 2 h (i.e. a reasonable time required for unloading the vessel) of continuous berthing, the rate of auxiliary fuel consumption is assumed to start to decrease linearly as a function of time.We have assumed that after 8 h of berthing, the rate of auxiliary fuel consumption has been decreased to one fifth (1/5) of the initial auxiliary fuel consumption rate.

Scenarios in the past, since 2005, 2009 and January of 2010
We have evaluated the emissions and fuel costs for three separate scenarios in the past, all of which assume that no abatement of shipping emission had been done.(i) First, we have evaluated the emissions and fuel cost differentials for a scenario in which we assumed that no FSC regulations had been imposed in the ECA after 2005.We have therefore assigned FSC A = 2.7 % in Eq. ( 1), and compared the resulting SO x and PM 2.5 emissions and fuel costs with the status quo emission estimates in 2011.Further, similar simulations are presented for scenarios assuming that (ii) no further regulations had been introduced after 2009, i.e.FSC A = 1.5 %, and (iii) no further regulations had been introduced after January of 2010, i.e.FSC A = 1.5 % and 0.1 % for berthing ships.

Scenarios for the future, in 2015
We have simulated the effects of the upcoming FSC requirements in 2015, by using the archived AIS data for 2011 and assigning FSC A = 0.1 % for all ships and activities.
Another simulation for 2015 was performed, in which EGCS installation candidate vessels were identified (cf.Sect.2.3.2) and were assumed to be equipped with scrubber abatement equipment.Vessels which are equipped with abatement equipment may use cheaper and heavier fuel than LSMGO, provided that the emissions do not exceed those that would be achieved with LSMGO without abatement equipment.

Slow steaming scenario
In the slow steaming scenario, we have evaluated the shipping emissions and statistics, as if each ship would have fared 10 % and 30 % slower while cruising (a = 0.1 and a = 0.3 in Eq. 3c).However, we assume that the speed reduction at slow speeds would not be economically desirable for ship owners.The speed reduction is therefore applied only, if the instantaneous speed exceeds 5.1 m s −1 (10 knots).As the engine power needs to be continuous in time, any reduced speed will not be reduced below this selected threshold value.
The increase in cruising time has been calculated according to Eq. ( 3a)-(3c), and the resulting emissions and fuel consumption with the reduced speed has been compared with the baseline emission estimates and fuel consumption and costs for 2011.Thus, we account for the increase in auxiliary fuel consumption as well as the decrease in main engine loads.We have not taken into account however the potential need for increasing the fleet size, due to the increase in cruising time.

Numerical results
The results were evaluated using the shipping emission model STEAM, with the archived AIS and ship properties data for the ECA region in 2009 and 2011.In the following, we first present an inventory of the emissions in 2009 and 2011 in the ECA, second, we address the spatial concentration distributions of the emissions in 2011, and third, present model predictions for the various assumed scenarios in the past and for the future.

Emission budgets in 2009 and 2011
The predicted emission inventories and shipping statistics are presented in Table 1 for the ECA in 2009.The maximum allowed FSC at the time was 1.5 %.
The corresponding shipping emission inventories according to EMEP have also been included in Table 1.However, there are some methodological differences between the current study and the methods used by EMEP.First, the STEAM model evaluated the PM 2.5 emissions, including the moisture (SO 4 + 6.5H 2 O) for sulfate particles (Jalkanen et al., 2012), whereas EMEP has used the dry weight of SO 4 .Secondly, the EMEP estimates include neither harbour activities nor non-IMO-registered ships, whereas those have been included in the STEAM computations.The accounting of harbour activities is a major methodological difference.According to the predictions using the STEAM model, approximately 22 % of the total fuel was consumed at harbours in the ECA in 2009.Despite this, the total shipping emissions predicted using the STEAM model were 14 % smaller than the corresponding EMEP emissions in case of NO x , while the SO x emissions predicted using the STEAM model were 20 % lower.There were also notable differences between the predictions of these two modelling systems in case of PM 2.5 and CO.
In 2009, approximately 15.5 and 27.5 million tons of CO 2 were emitted at the Baltic Sea and at the North Sea (for simplicity, the latter is here interpreted to include also the English Channel), respectively.The most significant flag states were the Scandinavian countries Norway, Sweden and Denmark, the Netherlands and the United Kingdom.The cargo ships were the single most significant ship type in terms of the CO 2 emissions.The corresponding emission estimates in the ECA in 2011 are presented in Table 2.In contrast to 2009, the maximum allowed FSC for ships at berthing was limited to 0.1 %, and otherwise to a maximum of 1.0 %.The contribution from non-IMO-registered ships in terms of CO 2 has doubled since 2009, but it is still only 5 % of the total estimated CO 2 ; this increase has probably been caused by an increase of the number of small ships that have installed AIS transmitters.The number of non-IMO-registered ships has increased from 8924 (in 2009) to 14754 (in 2011).However, this increase has not necessarily been caused by an increase in fleet size.A larger fraction of smaller ships have installed AIS transmitters, partly as these have become more affordable.The temporal evolution of the emissions of CO 2 has been presented in Fig. 3 for different ship categories and non-IMO-registered vessels both in 2009 and 2011.
The annual IMO-registered marine traffic has significantly increased from 2009 to 2011, in terms of both the CO 2 emissions (+8.9 %) and the cargo payload amounts (+10.6 %), possibly caused by the recovery of the European economy during the study period.There have been significant changes in the distribution of emissions for the various flag states as well.For instance, the number of ships sailing under the flag of Norway has substantially decreased, while the fleet of the Netherlands has significantly increased.A geographical difference map between the CO 2 emissions in 2011 and 2009 reveals a strong increase in the sea regions in the vicinity of the Netherlands, and a distinct decrease near the coasts of Norway (the results not shown here).These changes could be caused either by changes in shipping activities or changes in the use of AIS equipment.The imposed emission limitations up to date have had a significant impact on the emissions of SO x and PM 2.5 .According to results in Tables 1 and 2, the SO x emissions originated from IMO-registered marine traffic have been reduced from 2009 to 2011 from 320 kt to 231 kt.The corresponding predicted reduction for PM 2.5 from 71.6 kt to 60.9 kt.The estimated NO x emissions from IMO-registered traffic are slightly larger in 2011 than in 2009 (+5.1 %).The increase of the emissions of NO x was smaller than the corresponding increase of emissions of CO 2 .The reason for this is that after January 2011, the NO x emission factor was not allowed to exceed the IMO specified Tier II factor, which is slightly lower than the previous Tier I requirement for all engines.We have assumed that ships built after 2008 conform to the new Tier II limitations, as the engine manufactures have been well prepared for those requirements.However, the effect of the implementation of Tier II for the emissions of NO x from 2009 to 2011 seems minuscule, but will certainly increase when the fleet is renewed in time.
Based on the modelled fuel consumption statistics for IMO-registered vessels, 33 % of the total fuel was consumed by auxiliary engines in 2011.However, the ratio of the auxiliary fuel consumption and the total fuel consumption varies significantly between ship types (18 % for passenger ships, 30 % for cargo ships, 35 % for container ships, 31 % for tankers and 64 % for other ships).Approximately 17 000 ships in the ship properties database have been associated with a shaft generator, which allows the main engine to provide power to ship operating systems while cruising.Theoretically, it can be shown by numerical computations that if there would have been no shaft generators available, the predicted fuel consumption of the main and auxiliary engines would have been almost equal in the ECA in 2011.
It has been predicted that the use of HFO significantly outweights the use of distillate fuels.Commonly a ratio, such as 85 % / 15 %, has been used to distinguish the use of distillate fuels and the heavier grades.However, according to results this assumption seems to be biased.Assuming that fuels with a lower FSC than 1 % were distillate fuels (MDO or MGO), the ratio of HFO and distillate fuel consumption of IMOregistered vessels was approximately 76 % / 24 % in 2009.In 2011, this ratio changed to 70 % / 30 %.The high fraction of the distillate fuels is caused by two main factors.First, a major fraction of the fuel consumption originates from auxiliary engines during harbour activities; most of the auxiliary engines cannot use HFO due to engine restrictions (e.g.engine size, RPM and stroke type).Second, distillate fuel consumption for ships at berthing has increased significantly after the introduction of the Marpol Annex VI regulation.

The geographical distribution of shipping emissions in 2011
For 2011, the geographical distribution of CO 2 and PM 2.5 emissions in the ECA has been presented in Figs. 4 and 5, respectively.The relative geographical distribution of the shipping emissions is similar also for the other modelled compounds, and those results have therefore not been presented here.The highest CO 2 and PM 2.5 emissions originated from shipping are located near the coast of the Netherlands, in the English Channel and along the busiest shipping lines in the Danish Straits and the Baltic Sea.
In particular, in the vicinity of the coast of the Netherlands, the predicted PM 2.5 emissions per unit sea area are from three to five times higher, compared with the corresponding values in the major shipping lanes of the Baltic Sea.Near several major ports (e.g.Antwerp, Rotterdam, Amsterdam, Hamburg, Riga, Tallinn, Helsinki and St Petersburg) there are localized high amounts of PM 2.5 emissions that exceed the corresponding emissions even within the busiest shipping lanes in the ECA.
The geographic distribution of CO 2 emissions varies substantially between ship types, as illustrated in Fig. 6.Passenger ships operate relatively more at short distances, compared with the other presented ship categories.There is especially intensive passenger ship traffic between the ports of France and the UK, and there is busy traffic also between Rostock and Trelleborg, and between Helsinki and Tallinn.The geographical distributions of CO 2 emissions originated from container ships and cargo ships are similar with each other.However, the cargo ships were responsible for approximately 21 % more CO 2 emissions in 2011 than container ships.A substantial fraction of both container and cargo ships are located along the main shipping lanes from south-west (the English Channel) to north-east (St Petersburg).Miscellaneous ships operate intensively near the ports and the oil rigs on the North Sea.Almost 4 % of the fuel consumed on the North Sea is used by service ships that operate between oil rigs and ports.

Results for the selected scenarios of the emissions and fuel costs
Since May of 2006, the maximum allowed FSC in the ECA has been gradually lowered.In 2015, it will be reduced to 0.1 % for all large marine vessels operating within the area.

Results for the scenarios in the past, since 2005, 2009 and January of 2010
The relative SO x and PM 2.5 emissions and fuel costs for the selected scenarios have been summarized in Fig. 7, in relation to modelled emissions and fuel costs in 2011.The simulations for the past assumed that there would have been no regulative actions since 2005, 2009 or January of 2010, and then proceeded to evaluate the emissions and fuel costs for the reference year of 2011.In the following, we call these scenarios for simplicity the 2005, 2009 and 2010 scenarios.
For the 2005 scenario the SO x emissions in 2011 would have been more than double (+127 %), compared with the actual situation in 2011.The emissions of SO x and PM 2.5 for this scenario would have been 525 kt and to 104 kt, respectively.As expected, the direct fuel costs would have been lower that for the actual situation in 2011, about USD 9.8 billion, based on the current Rotterdam bunker fuel prices; this is USD 1.0 billion less than the actual estimated fuel costs in 2011.
In the 2009 scenario, there would be 337 kt and 76 kt of SO x and PM 2.5 emissions, respectively.These estimates are slightly larger than the presented values that were estimated with the actual data set for 2009.The total fuel costs for 32  all ships would be USD 10.4 billion, which is only USD 250 million more than the costs in the 2005 scenario.The reason is that the price of marine fuel with a FSC close to 1.5 % is only slightly higher than the fuel price for 2.7 % HFO, which was accepted before May 2006 in the ECA.
In the 2010 scenario, in which FSC maximum was set to 1.5 % and 0.1 % for ships at berth, ships would exhaust 309 kt of SO x and 72 kt of PM 2.5 , having fuel cost of USD 10.6 billion, which is roughly USD 220 million less than the estimated fuel costs for 2011 and 580 million more than in the 2009 scenario.Thus, we estimate that the requirement to switch to low sulfur distillates while berthing decreased the SO x emissions in harbours only by 28.4 kt and the PM 2.5 emissions by 4.2 kt.The reduction of FSC to a maximum of 1.0 % starting from 1 July 2010, reduced SO x emissions further by 77.9 kt and PM 2.5 emissions by 11.3 kt; the combined direct fuel costs of these reductions is approximately USD 0.8 billion.

Results for the scenarios of the future, in 2015
The 2015 scenario was simulated with the ECA 2011 data sets, i.e. by assuming that the shipping activities and the properties of the ships will be the same in the future, and by setting a maximum allowed FSC to 0.1 % for all activities.Three different fuel price scenarios were included, as the evolution of the relative prices of these fuels is uncertain; these are denoted briefly by FC50 %, FC75 % and FC100 %.These fuel price scenarios correspond to the cases in which the fuel prices remain the same as in 2011, and MGO is 50 %, 75 % or 100 % more expensive than HFO.
The SO x emissions in this scenario will be reduced to a mere 29.2 kt and fine particle emissions will be reduced to 31.4 kt.In comparison with the situation in 2011, the SO x emissions will be reduced by 87 % and the PM 2.5 emissions will be reduced by 46 %.The relative reduction of PM 2.5 emissions is smaller in comparison to those of SO x , as marine engines produce significant amounts of carbon and ash particles, regardless of FSC.The direct fuel costs will increase to USD 13.3, 15.7 or 18.3 billion, depending on the fuel price development, which corresponds to a cost increase of 23-69 %.Reynolds (2011) estimated that ships with an annual fuel consumption of more than 4000 t would gain economic benefit from scrubber installation, instead of using 0.1 % MGO fuel in 2015, provided that MGO will be at least 50 % more expensive than HFO and each ship with an installed scrubber will be active for at least 5 yr after installation.Using the modelled fuel consumption statistics for the year 2011, the possible candidates for EGCS installment suggested by Reynolds were identified; a total of 635 candidate ships were found.While there was more than 30 000 different ships operating at the time, these 635 ships account for 21 % of the total fuel consumption in the ECA.These ships have been listed in Table 3 according to their ship category.Most of these candidate ships are either container ships or RoPax vessels.
Another simulation was performed with the 2015 regulations, in which a typical scrubber abatement method was assumed to be installed to each candidate ship.The fuel costs of this scenario were significantly lower compared with the corresponding scenario without the scrubbers: USD 12.3, 14.2 or 16.1 billion (a cost increase from 13 % to 49 %).Further, most of the economic benefits from the use of scrubbers (and from using cheaper fuel simultaneously) were in the Baltic Sea shipping.A major portion of the identified EGCS candidate ships operates mainly in the Baltic Sea.The estimated PM 2.5 emissions in this scenario were slightly smaller than in the 2015 scenario without scrubbers.The reason for this is that the virtual scrubbers reduced 66 % from SO x emissions and 75 % from PM 2.5 emissions and, thus, FSC A in Eq. ( 2a) and (2b) results in a slightly lower FSC than would be required in terms of a PM 2.5 emission factor in 2015.
The economic benefits from the use of scrubbers in 2015 are clear, based on these computations.However, the cost of an EGCS installment per vessel can be from USD 5 to 9 million (Reynolds, 2011), and there are also maintenance costs.These installment and maintenance costs have not been taken into account in the presented scenarios.Further, for technical reasons not all ships can be equipped with such systems and it might also not be economically viable, if the vessel is reaching the end of its lifespan.

Slow steaming
We have investigated the savings in fuel consumption and the reduction of emissions, due to reducing vessel speeds.
In evaluating the financial costs, we have not addressed the additional costs associated with longer cruising times, such as, e.g.increased personnel costs, costs related to the slower delivery of the cargo, and the potential need for increasing the fleet size.
For simplicity, the amount of speed reduction was selected to be proportional to actual speed, viz. 10 % or 30 %.However, such speed reduction was imposed only if vessel speed was higher than 5.1 m s −1 (10 knots), as it would be unlikely to achieve significant economic savings by reducing speeds that are lower than this selected threshold value.The estimated savings in the consumption and costs of fuel, and the reductions in emissions have been presented in Table 4a and b.The results of these slow-steaming scenarios are shown separately for those vessel categories, for which the fuel consumption is > 1.0 % of total fuel consumption in the ECA in 2011.The presented ship types, except for the container ship category, are sub-classes of the vessel categories presented in Tables 1 and 2.
Even a reduction of 10 % in cruising speed will effectively reduce the main fuel consumption of several ship categories.In total, CO 2 , NO x , SO x , and PM 2.5 emissions are reduced by 9.4 %, 11.7 %, 13.2 % and 11.5 % respectively.The reductions of the NO x , SO x and PM 2.5 emissions are larger than those for CO 2 .The reason is that the main engines generally use fuel with a higher FSC and large two-stroke main engines are responsible for higher NO x emissions per provided energy unit, compared with smaller auxiliary engines.
However, the CO emissions per provided energy unit tend to increase for lower engine loads.
Depending on the ship type, the achieved reduction in main fuel consumption ranges from 6.5 % to 18.3 %.The relative change of the operational time (berthing, maneuvering and cruising) is significantly smaller.For instance, the fuel costs of RoPax ships would be reduced by 13.6 %, while the operational time increases by 3.2 %.RoRo and vehicle carriers would achieve the reductions in fuel costs of 14.3 % and 12.5 %, while their operational time would increase by 5.0 %.Together, the categories of RoPax, RoRo and vehicle carriers contribute 22.4 % of the total fuel consumption in the ECA.The container ship category, which is the largest vessel category in the ECA, would gain a more modest 8.6 % reduction in fuel costs, and an increase of operational time of +4.7 %.
For the scenario with a speed reduction of 30 %, the emissions of CO 2 , NO x , SO x and PM 2.5 are reduced by 20.7 %, 26.7 %, 29.6 % and 24.5 %, respectively.Due to the selection of the above mentioned threshold speed (5.1 m s −1 ), only the ships which are cruising faster than 7.4 m s −1 (approximately 14.3 knots) are subject to a full 30 % reduction in speed.Substantial reductions due to a reduced speed would be expected for RoPax ships, vehicle carriers, crude oil tankers and passenger cruisers.
Inter-comparing the results for these two speed reduction scenarios reveals that the savings of fuel costs with respect to the increases of operational times are higher in the scenario with a 10 % speed reduction.This is to be expected, as the slower cruising speed results in a higher fuel consumption of auxiliary engines.A major increase in operational time also results in a need for using additional ships.

Conclusions
The marine exhaust emissions were evaluated using the STEAM model in the ECA in 2009 and 2011.The combined emissions of CO 2 from shipping sources in the ECA were evaluated to have increased from 43 to 48 million tons from 2009 to 2011 (+11 %, using 2009 as the base year), mostly caused by the increase in cargo transport in the ECA region during the study period.Although the number of non-IMOregistered vessels strongly increased, the estimated contribution of these presumably small vessels was only 5 % in terms of CO 2 emissions in 2011.
The predicted SO x emissions originating from IMOregistered marine traffic have been reduced from 320 kt to 231 kt from 2009 to 2011 (−29 %, using 2009 as the base year).The corresponding predicted reduction for PM 2.5 was from 71.6 kt to 60.9 kt (−17 %, using 2009 as the base year).The emission limitations from 2009 to 2011 have obviously had a significant impact on reducing the emissions of both SO x and PM 2.5 .The highest CO 2 and PM 2.5 emissions originating from shipping in 2011 were located in the vicinity of the coast of the Netherlands, in the English Channel, near the southeastern UK and along the busiest shipping lines in the Danish Straits and the Baltic Sea.Near several major ports (e.g.Antwerpen, Rotterdam, Amsterdam, Hamburg, Riga, Tallinn, Helsinki and St Petersburg), there were especially high PM 2.5 emissions per square kilometre, which exceeded the corresponding emission values even within the busiest shipping lanes in the ECA.The geographic distribution of emissions was substantially different for various ship types.Clearly, the emission inventories of this study could be used as input values for evaluating the atmospheric dispersion, population exposure and health impacts caused by shipping.
A number of scenario computations for the past were performed to evaluate more extensively the effects of the grad-ually decreasing maximum allowed FSC.As a result of the restrictions, the SO x and fine particle matter emissions originated from IMO-registered shipping have steadily decreased.A model simulation was performed, in which we assumed that the FSC regulations as they were issued in 2005 would have been in effect until 2011, without any subsequent fuel sulfur content restrictions.The simulation showed that the SO x emissions in the ECA would have been 127 % higher (i.e. more than twice as high), compared with the predicted values in 2011, including all the implemented regulations.The corresponding PM 2.5 emissions would have been 71 % higher.However, the direct fuel costs would have been 10 % lower, according to the predictions.
The potential impacts of the forthcoming reductions regarding the maximum allowed FSC in 2015 were also studied, with simulations using the archived data in 2011.It was estimated that the emissions of SO x will be reduced by 87 % and those of PM 2.5 by 48 %, with respect to the estimated emissions in the ECA in 2011.The direct fuel costs were estimated to increase by 23 % from 2011 to 2015, assuming the contemporary bunker prizes.However, if the price premium of MGO with respect to HFO by that time increases 100 %, due to the increase in demand, then the direct fuel costs would annually be 69 % higher.
Based on the estimated fuel consumption and current fuel prices, it was evaluated that more than 630 IMO-registered ships might benefit from a retrofit scrubber installation.These candidate ships were responsible for approximately 21 % of the total fuel consumption in the ECA in 2011.Assuming that each of these ships would use sulfur scrubbers instead of using 0.1 % sulfur content MGO in 2015, the estimated fuel cost would increase in 2015 either only by 13 % (using the contemporary bunker prizes) or by 49 % (assuming 100 % price premium between HFO and MGO).However, we did not address in these computations the installment costs and running maintenance costs.It is also not technically feasible to retrofit all of the candidate ships with such an EGCS device.
The possibility to achieve emission reductions by decreasing vessel cruising speeds was also investigated.We numerically applied speed reductions of 10 % and 30 % to speeds exceeding 5.1 m s −1 (10 knots).Furthermore, we accounted for the increases in auxiliary engine fuel consumption, decreases in engine loads and computed the resulting fuel savings and emission reductions for each pollutant and ship category individually.The resulting fuel savings were significant even with a 10 % reduction of cruising speed.The relative reduction of NO x , SO x and PM 2.5 emissions was estimated to be higher than the reduction in total fuel consumption.The effectiveness of speed reduction as a way to curb emissions varies substantially between ship types.Especially RoPax, RoRo, tankers and vehicle carrier ships could substantially save in fuel costs, while the increase in operational time would not be significantly increased.The ratio of fuel savings and the increase in operational time was better using the smaller, 10 % speed reduction for all ship types.However, the reduced cruising speeds may result in a need for larger fleet sizes.

Figure 1 :Fig. 1 .
Figure 1: Schematic diagram describing the variables used in modelling of FSC, fuel consumption and 690

28Figure 2 :
Figure 2: Estimated fuel prices (USD/ton) as a function of the sulphur content of fuel, for three different fuel cost (FC) scenarios.The scenarios correspond to the current state (FC50%) and two future price (FC75 % and FC100 %) scenarios; these have been defined in the text.The numerical equations of the fits have also been reported.

Figure 3 :
Figure 3: Seasonal variation of the predicted CO2 emissions in the ECA in 2009 and 2011, presented separately for different ship types.Cargo ships include bulk carriers, general cargo vessels and vehicle carriers.Passenger ships include RoPaX ships, ferries and passenger cruisers.

28Figure 2 :Figure 3 :
Figure 2: Estimated fuel prices (USD/ton) as a function of the sulphur content of fuel, for 698 three different fuel cost (FC) scenarios.The scenarios correspond to the current state 699 (FC50%) and two future price (FC75 % and FC100 %) scenarios; these have been defined in 700 the text.The numerical equations of the fits have also been reported.701

Fig. 3 .
Fig. 3. Seasonal variation of the predicted CO 2 emissions in the ECA in 2009 and 2011, presented separately for different ship types.Cargo ships include bulk carriers, general cargo vessels and vehicle carriers.Passenger ships include RoPax ships, ferries and passenger cruisers.

Figure 4 :Fig. 4 .
Figure 4: Predicted geographic distribution of shipping emissions of CO2 in the ECA in 2011.708 The colour code indicates emissions in relative mass units per unit area.709 Fig. 4. Predicted geographic distribution of shipping emissions of CO 2 in the ECA in 2011.The colour code indicates emissions in relative mass units per unit area.

Figure 5 :Fig. 6 .
Figure 5: Predicted geographic distribution of shipping emissions of PM2.5 in the ECA in 2011.712 PM2.5 has been assumed to consist of organic and elemental carbon, ash and moist sulfate particles.713 Fig. 5. Predicted geographic distribution of shipping emissions of PM 2.5 in the ECA in 2011.PM 2.5 has been assumed to consist of organic and elemental carbon, ash and moist sulfate particles.

Figure 7 :
Figure 7: Relative emissions of SOx and PM2.5, and direct fuel costs of IMO-registered marine traffic in the ECA in 2011, for the various selected scenarios.The situation in 2011 has been evaluated also using three different assumed options regarding the regulations of marine emissions in the past (the three sets of columns on the left-hand side).The scenarios for the future have been presented using three fuel cost (FC) options (the two sets of columns on the right-hand side).

Fig. 7 .
Fig. 7.Relative emissions of SO x and PM 2.5 , and direct fuel costs of IMO-registered marine traffic in the ECA in 2011, for the various selected scenarios.The situation in 2011 has been evaluated also using three different assumed options regarding the regulations of marine emissions in the past (the three sets of columns on the lefthand side).The scenarios for the future have been presented using three fuel cost options (the two sets of columns on the right-hand side).

Table 1 .
Buhaug et al. (2009)and shipping statistics for the ECA in 2009.Shipping emission inventories by EMEP have also been presented for comparison purposes.Payload is the amount of transferred freight inside the ECA, which has been estimated based on ship's deadweight and its type-specific fraction of payload reported inBuhaug et al. (2009).

Table 2 .
Predicted emissions and shipping statistics for the ECA in 2011.

Table 3 .
Reynolds (2011)candidate ships for the installment of the EGCS, and their fraction of the total fuel consumption, presented separately for each ship type.The values are based on the estimated fuel consumption in the ECA in 2011.Ships with an annual fuel consumption of at least 4000 t have been qualified as such candidates, according toReynolds (2011).

Table 4 .
The predictions for the slow-steaming scenarios, assuming speed reductions of 30 % (a) and 10 % (b).Speed reductions have been applied only for instantaneous speeds exceeding 10 knots."Share of total FC 2011" refers to the estimated share of total fuel consumption in the ECA in 2011.Operational time is the combined duration of berthing, maneuvering and cruising.