Emissions factors for gaseous and particulate pollutants 1 from offshore diesel engine vessels in China

23 Shipping emissions have significant influence on atmospheric environment as well as 24 human health, especially in coastal areas and the harbor districts. However, the 25 contribution of shipping emissions on the environment in China still need to be 26 clarified especially based on measurement data, with the large number ownership of 27 vessels and the rapid developments of ports, international trade and shipbuilding 28 2 industry. Pollutants in the gaseous phase (carbon monoxide, sulfur dioxide, nitrogen 1 oxides, total volatile organic compounds) and particle phase (particulate matter, 2 organic carbon, elemental carbon, sulfates, nitrate, ammonia, metals) in the exhaust 3 from three different diesel engine power offshore vessels in China (350kW, 600kW 4 and 1600kW) were measured in this study. Concentrations, fuel-based and 5 power-based emissions factors for various operating modes as well as the impact of 6 engine speed on emissions were determined. Observed concentrations and emissions 7 factors for carbon monoxide, nitrogen oxides, total volatile organic compounds, and 8 particulate matter were higher for the low engine power vessel (HH) than for the two 9 higher engine power vessels (XYH and DFH), for instance, HH had NOx EF of 25.8 10 g·kWh -1 compared to 7.14 and 6.97 g·kWh -1 of DFH and XYH, and PM EF of 2.09 11 g·kWh -1 compared to 0.14 and 0.04 g·kWh -1 of DFH and XYH. Average emissions 12 factors for all pollutants except sulfur dioxide in the low engine power engineering 13 vessel (HH) were significantly higher than that of the previous studies (such as 30.2 14 g·kg -1 fuel of CO EF compared to 2.17 to 19.5 g·kg -1 fuel in previous studies, 115 15 g·kg -1 fuel of NOx EF compared to 22.3 to 87 g·kg -1 fuel in previous studies and 9.40 16 g·kg -1 fuel of PM EF compared to 1.2 to 7.6 g·kg -1 fuel in previous studies), while for 17 the two higher engine power vessels (DFH and XYH), most of the average emissions 18 factors for pollutants were comparable to the results of the previous studies, engine 19 type was one of the most important influence factors for the differences. Emissions 20 factors for all three vessels were significantly different during different operating 21 modes. Organic carbon and elemental carbon were the main components of 22 particulate matter, while water-soluble ions and elements were present in trace 23 amounts. The test inland ships and some test offshore vessels in China always had 24 higher EFs for CO, NOx and PM than previous studies. Besides, due to the significant 25 influence of engine type on shipping emissions and no accurate local EFs could be 26 used in inventory calculation, much more measurement data for different vessels in 27 China are still in urgent need. Best-fit engine speeds during actual operation should be 28 based on both emissions factors and economic costs. 29

Abstract.Shipping emissions have significant influence on atmospheric environment as well as human health, especially in coastal areas and the harbour districts.However, the contribution of shipping emissions on the environment in China still need to be clarified especially based on measurement data, with the large number ownership of vessels and the rapid developments of ports, international trade and shipbuilding industry.Pollutants in the gaseous phase (carbon monoxide, sulfur dioxide, nitrogen oxides, total volatile organic compounds) and particle phase (particulate matter, organic carbon, elemental carbon, sulfates, nitrate, ammonia, metals) in the exhaust from three different dieselengine-powered offshore vessels in China (350, 600 and 1600 kW) were measured in this study.Concentrations, fuelbased and power-based emission factors for various operating modes as well as the impact of engine speed on emissions were determined.Observed concentrations and emission factors for carbon monoxide, nitrogen oxides, total volatile organic compounds, and particulate matter were higher for the low-engine-power vessel (HH) than for the two higherengine-power vessels (XYH and DFH); for instance, HH had NO x EF (emission factor) of 25.8 g kWh −1 compared to 7.14 and 6.97 g kWh −1 of DFH, and XYH, and PM EF of 2.09 g kWh −1 compared to 0.14 and 0.04 g kWh −1 of DFH, and XYH.Average emission factors for all pollutants except sulfur dioxide in the low-engine-power engineering vessel (HH) were significantly higher than that of the previous studies (such as 30.2 g kg −1 fuel of CO EF compared to 2.17 to 19.5 g kg −1 fuel in previous studies, 115 g kg −1 fuel of NO x EF compared to 22.3 to 87 g kg −1 fuel in previous studies and 9.40 g kg −1 fuel of PM EF compared to 1.2 to 7.6 g kg −1 fuel in previous studies), while for the two higher-engine-power vessels (DFH and XYH), most of the average emission factors for pollutants were comparable to the results of the previous studies, engine type was one of the most important influence factors for the differences.Emission factors for all three vessels were significantly different during different operating modes.Organic carbon and elemental carbon were the main components of particulate matter, while water-soluble ions and elements were present in trace amounts.The test inland ships and some test offshore vessels in China always had higher EFs for CO, NO x , and PM than previous studies.Besides, due to the significant influence of engine type on shipping emissions and that no accurate local EFs could be Published by Copernicus Publications on behalf of the European Geosciences Union.
F. Zhang et al.: Emission factors for gaseous and particulate pollutants used in inventory calculation, much more measurement data for different vessels in China are still in urgent need.Bestfit engine speeds during actual operation should be based on both emission factors and economic costs.

Introduction
Gaseous and particulate pollutants emitted from vessels operating in the open ocean as well as in coastal areas and inland waterways have significant adverse impacts on human health, air quality, and climate change (Cappa et al., 2014;Righi et al., 2011;Marmer and Langmann, 2005;Winebrake et al., 2009).It has been estimated that 87 000 premature deaths occurred in 2012 due to burning of marine fuels with high sulfur content.Shipping-related particulate matter (PM) emissions have been reported to be responsible for approximately 60 000 cardiopulmonary and lung cancer deaths annually, with most cases occurring near coastlines in Europe (Viana et al., 2014), East Asia, and South Asia (Corbett et al., 2007).Approximately 9200 and 5200 t year −1 of PM are emitted from oceangoing and coastal ships, respectively, in the USA (Corbett and Fischbeck, 2000), most of which are fine or even ultrafine aerosols (Viana et al., 2009;Saxe and Larsen, 2004).Globally, about 15 % of nitrogen oxides (NO x ) and 5-8 % of sulfur oxides (SO x ) emissions are attributable to oceangoing ships (Corbett and Fischbeck, 2000).Shipping emissions affect acid deposition and ozone concentrations, contributing more than 200 mg S m −2 year −1 over the southwestern British Isles and Brittany as well as additional 6 ppb surface ozone during the summer over Ireland (Derwent et al., 2005).Moreover, aerosol emissions from international shipping also greatly impact the Earth's radiation budget, directly by scattering and absorbing solar radiation and indirectly by altering cloud properties (Righi et al., 2011).Besides, according to estimates from IMO (2014), total shipping emissions were approximately 938 million tonnes CO 2 and 961 million tonnes CO 2e (CO 2 equivalent) for GHGs combining CO 2 , CH 4 and N 2 O for the year 2012.International shipping emission accounts for approximately 2.2 and 2.1 % of global CO 2 and GHG emissions on a CO 2 equivalent (CO 2e ) basis, respectively.Because nearly 70 % of ship emissions are estimated to occur within 400 km of land (Endresen, 2003), ships have the potential to contribute significantly to air quality degradation in coastal areas.In addition, ports are always the most concentrated areas for ships to berth at, emission reduction measures such as switching from heavy fuels to cleaner fuels are required when ships are close to ports or offshore areas, but not all of them can obey the regulations (De Meyer et al., 2008), which results in significant influence on atmospheric environment of port cities and regions.
Rapid developments of ports, international trade, and the shipbuilding industry in China have negatively affected the ambient air quality of the coastal zone due to shipping emissions.It was estimated that 8.4 % of SO 2 and 11.3 % of NO x were emitted from ships in China in 2013 with port cities being the worst effect areas (http://news.xinhuanet.com/politics/2015-06/08/c_127890195.htm).In 2013, there were 0.18 million water transport vessels (Ministry of Transportation, 2013) active in Chinese waters, 8 ports in China were listed among the world's largest 10 ports, and 11 container ports were listed among the world's largest 20 container ports.The number of ports with cargo handling capacity of more than 200 million t year −1 grew to 16 (Ministry of Transportation, 2010).Rapid development of ports in China has resulted in increasingly serious pollution of ambient air, particularly in coastal zones and near ports.Only a few studies have focused on pollution from shipping emissions in China.Rough estimates of the influence of shipping emissions on ambient air in the port of Shanghai, the largest port in China (Zhao et al., 2013) and in the Bohai Rim (Zhang et al., 2014); these estimates have been generated using empirical formulas.One case study of real-world emissions of inland vessels on the Grand Canal of China has been conducted (Fu et al., 2013), and another study focused on inland ships and several offshore vessels resulted in some rough emission data (Song, 2015).Other studies also have developed to the inventories in large ports or delta regions (Zheng et al., 2011;Zheng et al., 2009) by using EFs (emission factors) obtained from other countries or areas.However, there are no systematic studies of vessel emissions in the coastal zone or in ports, nor accurate estimates of shipping emissions to ambient air based on measured emission factors.Conditions in China differ substantially from those in other countries, such as in vessel types (more small motor vessels, and the type composition of offshore vessels is shown in Table S1 in the Supplement, more light-tonnage vessels, e. g. with 79.3 % less than 3000 t in offshore area of Yangtze River Delta that is shown in Table S2), different fuel standards compared with other countries (the GB/T 17411-2012 standard with sulfur content of less than 3.5 % m / m; however, the ISO 8217-2010 international standard has the maximum sulfur content according to the relevant statutory requirements that always have lower values, such as less than 0.1 % in emission control areas; besides, a large percentage of diesel fuel was used in China, especially in offshore areas, seen in Table S3).The age of vessels is also important (Chinese commercial vessels have an average age of 19.2 years compared with 8.0 years and 8.9 years for Japan and Germany, respectively).However, a large part of previous studies focused on emission of largetonnage vessels such as cargo ships (Moldanova et al., 2009;Celo et al., 2015), large marine ships (Khan et al., 2013;Sippula et al., 2014), tankers (Agrawal et al., 2008b;Winnes and Fridell, 2010) and so on, whose fuel types were typically of the heavy oil variety, and most of them had engines less than 10 years.Thus, experimentally determined EFs for vessels in other countries cannot be used directly to estimate shipping emissions and their contribution to ambient air quality in China, especially in offshore areas.Systematical measurement EFs for different kinds of vessels in China is essential.
The IMO has set the emission limits for NO x and SO x in the revised MARPOL (Maritime Agreement Regarding Oil Pollution) Annex VI rules (IMO, 1998).Ships operating in the emission control areas (ECAs) (the Baltic Sea, the North Sea, the North America and the Caribbean) should use fuels with a sulfur content of less than 0.1 % m / m since January 2015.Even more stringent limits have been laid down in some national or regional regulations.For example, in some EU ports, seagoing ships at berth are required to switch into fuels of under 0.1 % m / m sulfur since 2010 (The Council of the European Union, 1999); both marine gas oil and marine diesel oil used in water area within 24 nautical miles of coastline in California should have a sulfur content of less than 0.1 % m / m since 2014 (California Code of Regulation Titles 13 and 17, 2016).Emission standard of Tier II for NO x set by MARPOL VI has been executed since January 2011 in ECAs, and more stringent rules of Tier III will be executed from January 2016.However, in China, no specific policy or limit for shipping emissions has been implemented except in Hong Kong, which is making legislation about the limit of 0.5 % sulfur content fuel used when berth in the port from 2015.But because of the serious air pollution currently in China, emission limits for the main sources such as vehicle exhaust, coal combustion, biomass combustion and fugitive dust have become more and more stringent  China, 2015).All the regulations were set mostly based on other directives and regulations.Detailed measurement data will assist with further policy, making it more relevant to current situations of vessels.
Average EFs are often used for shipping emissions inventories on large scales or in regional areas (Tzannatos, 2010;Eyring et al., 2005).However, to evaluate the effects of shipping emissions on air pollution in local areas such as near ports, various ship speeds and operating modes should be considered, including docking, berthing, and departing from ports etc.Previous studies have confirmed that EFs are significantly different under various load conditions (Petzold et al., 2010) or in different operating modes (Fu et al., 2013;Winnes and Fridell, 2010) for individual vessels.Therefore, more detailed measurements of EFs in different operating modes are necessary to better estimate the impacts of shipping emissions on the environment.
In this study, experimental data for three different dieselengine-powered vessels were collected.All pollutants were measured directly in the stack.Gaseous emissions and PM from the diesel engines were the main targets, including CO, carbon dioxide (CO 2 ), SO 2 , NO x , total volatile organic compounds (TVOCs), and total suspended particulates (TSPs).Fuel-based EFs for the three vessels were calculated using the carbon balance method under different operating conditions.In addition, fuel-based average EFs as well as powerbased average EFs to values reported in other studies and for other vessels were compared.Finally, the impacts of engine speed on the EFs of NO x were evaluated.

Test vessels and fuel types
Initially, it was hoped that the choice of measurement ships would reflect the shipping fleet in general, i.e. in terms of engine type (engine speed and power output), fuel used, engine age and mode of operation, with more than 10 vessels planned to test.However, consideration was given to the practicalities involved with the measurements, i.e. installation of sampling systems, external conditions, etc. Besides, time and economic constraints weighed heavily and only several shipowners willing to participate in the project.Thus, the chosen vessels of different engine powers with diesel used represent a compromise.Three different diesel-engine-powered offshore vessels, including one engineering vessel, Haohai 0007 (HH), with low-power and high-speed engine, one large research vessel, Dongfanghong 2 (DFH), with high-power and mediumspeed engine, and another research vessel, Xiangyanghong 08 (XYH), with medium-power and medium-speed engine were selected for this study; their technical parameters are shown in Table 1.High-speed and medium-speed engines are the predominant engines used in vessels of offshore and inland rivers in China, which always use light diesel as fuel.Two of the test vessels were small motor, light-tonnage vessels (HH and XYH), and another one was a medium-speed engine vessel with an 18 year-old engine.Engineering vessels are designed for construction activities such as building docks in port areas or waterways, dredging, etc.They are common vessels in coastal areas of China because of the heavy demand for oilfield construction and port expansion.The maintenance of engineering vessels is typically poorer than for other types of vessels and as a result, they may have relatively high emissions.On the other hand, research vessels of DFH and XYH from universities and research institutes are generally well maintained and use high-quality diesel fuel but with different engine powers, which might have relatively low emission factors for pollution.Therefore, these research vessels can reflect the impact of engine power on emissions and also can represent the lower end of expected EFs for Chinese vessels.The test vessels in the present study could account for 34.7 % of the total vessels according to the distribution of vessels through gross tonnage in China (seen in Table S2), which could have a certain degree of representation.Everything considered, a general range of EFs for gaseous and PM pollutants emitted from different offshore vessels of China and their influence factors could be given through the on-board measurement.
The fuels used in all test vessels were common diesel fuels obtained from fuelling stations near the ports.According to statistical data, the total oil consumption of vessels in China was 20.99 million tons in 2011, including 10.99 million tons bonded oil and 5.93 million tons domestic trade oil, with light fuel oil accounting for 40 % of the domestic trade oil and 25 % of the total consumption (shown in Table S3) (Zhu, 2013).The test vessels in the present study could reflect the emission condition of diesel vessels in China, especially in offshore areas where diesel oil always been used as fuel.Re- sults of fuel analyses are presented in Table 2.All of these fuels had relatively low sulfur content (≤ 0.13 % m) and low metals concentrations (V, Al, Si, Pb, Zn, Mn, etc.).

Test operating modes
EFs are significantly different under differing load conditions and operating modes.Vessel speed is also an important influence factor for emissions; it was reported by Starcrest Consulting Group, LLC (Starcrest Consulting Group, 2012) that 15-20 % of fuel consumption could be reduced by 10 % of the vessel speed.In this study, vessel operating modes were classified according to actual sailing conditions.There were six modes of HH: low speed (4 knots), medium speed (8 knots), high speed (11 knots), acceleration process, moderating process, and idling; four modes of DFH: cruise (10 knots, medium speed for DFH), acceleration process, moderating process, and idling; and five modes of XYH: low speed (3 knots), high speed (10 knots), acceleration process, moderating process, and idling.Three to five sets of replicate samples were collected for each operating mode.

Emissions measurement system and chemical analysis of particulate matter
A combined on-board emissions test system (Fig. 1) was used to measure emissions from the coastal vessels under actual operating conditions.There was no dilution in this test system with all the species measured directly from the exhaust and there were four main components of the system: a flue gas analyser, three particulate samplers, an eightstage particulate sampler, and a TVOCs analyser.(see Sup-  S4).Three particulate samplers are installed to collect PM using different filters at the same time, including a quartz fiber filter, glass filter, and polytetrafluoroethylene filter to analyse different chemical components of PM.And the portable TVOCs analyser is used to monitor the concentration of total VOCs with isobutylene as correction coefficient gas.Besides, a temperature sensor is installed near the smoke outlet to test the flue gas temperature.A total of 33 sets of samples for HH, 20 sets for DFH, and 23 sets for XYH were collected, with 3 to 5 sets for each operating mode.
The OC (organic carbon) and EC (elemental carbon) were measured on a 0.544 cm 2 quartz filter punched from each filter by thermal optical reflectance (TOR) following the IM-PROVE protocol with a DRI Model 2001 Thermal/Optical Carbon Analyzer (Atmoslytic Inc., Calabasas, CA).The measuring range of TOR was from 0.05 to 750 µg C cm −2 with an error of less than 10 %.Concentrations of watersoluble ions in PM 2.5 , such as Na + , NH + 4 , K + , Mg 2+ , Ca 2+ , Cl − , NO − 3 and SO 2− 4 , were determined by ion chromatography (Dionex ICS3000, Dionex Ltd.America) based on the measurement method of Shahsavani et al. (2012).The detection limit was 10 ng mL −1 with an error of less than 5 %, and 1 mL RbBr with concentration of 200 ppm was put in the solution as internal standard before sampling.The concentrations of 33 inorganic elements in PM 2.5 were estimated using a inductively coupled plasma mass spectrometer (ICP-MS of ELAN DRC II type, PerkinElmer Ltd.Hong Kong) following the standard method (Wang et al., 2006).The resolution of ICP-MS ranged from 0.3 to 3.0 amu with a detection limit lower than 0.01 ng mL −1 , and the error was less than 5 %.

Data analysis
Carbon balance formula was used to calculate the EFs for all exhaust gas components.It was assumed that all carbon in the fuel was emitted as carbon-containing gases (CO, CO 2 , and TVOC) and carbon-containing particulate matter.So there was a certain equilibrium relationship between the carbon in the fuel and in the exhaust: The EF for CO 2 was calculated as follows: where EF CO 2 is the EF for CO 2 (g kg −1 fuel), c(CO 2 ) is the molar concentration of CO 2 (mol m −3 ), and M CO 2 is the molecular weight of CO 2 (44 g mol −1 ).
The remaining EFs were calculated as follows: where EF x is the EF for species X (g kg −1 fuel), X and CO 2 represent the concentrations of X and CO 2 with the background concentrations subtracted (mol m −3 ), and M X represents the molecular weight of species X (g mol −1 ).
In addition, average EFs for each vessel were calculated based on actual operating conditions as follows: where EF X,A is the average EF for species X, EF i is the EF for operating mode i for species X, and P i is the percentage of time spent in operating mode i during the shipping cycle.
Power-based emission factors and fuel-based emission factors could be interconverted with the formula as following: where EF X,P is the power-based emission factor for species X (g kW h −1 ), and FCR is fuel consumption rate for each vessel (kg fuel (kW h) −1 ).Nearly all of the concentrations measured in the exhaust of low-engine-power vessel HH were higher than those of the two higher-engine-power vessels.Concentrations of CO, SO 2 , and NO x from HH were 10.7-756, 5.34-33.1,and 87.8-1295 ppm, respectively, and 14.3-59.5 mg m −3 PM.In contrast, concentrations of CO, SO 2 , NO x , and PM were 50.1-141, 5.27-16.9,169-800 ppm, and 7.06-21.8mg m −3 , respectively, for DFH and 36.0-224,0.49-35.9,and 235-578 ppm and 0.56-6.31mg m −3 , respectively, for XYH.
A previous study demonstrated that concentrations of CO primarily depend on engine power, with higher CO emissions resulting from vessel engines with lower power (Sinha et al., 2003).There was a similar trend in this study with generally higher concentrations for HH and lower concentrations for DFH.The CO concentrations in the present study were similar but slightly lower than those of inland vessels (Fu et al., 2013), except in the idling mode of HH.In different operating modes, CO concentrations were significantly different.For example, the maximum value was observed in idling mode and the minimum value in medium-speed mode for HH.All three ships had the lowest CO concentrations at their economic speeds (medium speed for HH, cruise mode for DFH, and high speed for XYH), demonstrating that their engines are optimized for the most common operating mode.
More than 80 % of the NO x was NO in this study, with NO 2 and N 2 O accounting for < 20 % in all operating modes (Fig. S1).Again, nearly all of these concentrations were higher in the exhaust gas of HH than in that of the two vessels.In high-speed modes, all of the vessels had high concentrations of NO x .NO x emissions mainly depend on the combustion temperature of the engines.More powerful combustion systems operate at higher temperatures, thereby producing more NO x (Corbett et al., 1999).However, the NO x emissions were much lower than for the inland vessels studied by Fu et al. (2013), particularly in cruise mode (NO x concentrations of ∼ 1000 ppm).
SO 2 concentrations in the exhaust gas depend on the sulfur content of the fuel and the flow rate of the flue gas.There were significant differences among the three vessels in their flow rates, which could account for the different concentrations of one vessel in different operating modes.But because of the low-sulfur fuels used in these vessels, the SO 2 concentrations were low compared with those in other studies (Williams et al., 2009;Berg et al., 2012).
Much lower concentrations of PM in the exhaust gas were observed in the present study compared to those of inland ships in China (Fu et al., 2013).However, they were similar to those from ships at berth reported by Cooper et al (Cooper, 2003).HH had higher PM concentrations than the two ves-sels in the exhaust gas.There were significant differences among the different operating modes because of changes in the injection point of the engines (Sippula et al., 2014;Li et al., 2014).

Fuel-based emission factors
Fuel-based EFs for the gaseous species CO 2 , CO, NO, NO 2 , N 2 O, and TVOCs and for PM based on the carbon balance method were determined.In addition, SO 2 was calculated based on the sulfur content of the fuels.Fuel-based EFs for the typical pollutants such as CO, PM and nitrogen oxides in different operating modes are shown in Fig. 2 (detailed EFs for all the gaseous pollutants are shown in Table S5 and detailed EFs for PM and its chemical composition are shown in Table S6).
CO 2 emissions from vessels primarily depend on the carbon content of the fuel (Carlton et al., 1995).Accordingly, the EFs for CO 2 in the present study should theoretically be 3177, 3168, and 3171 g kg −1 fuel for complete combustion.Under actual conditions, CO 2 emissions were 2940-3106, 3121-3160, and 3102-3162 g kg −1 fuel for HH, DFH, and XYH, respectively, which means they had combustion efficiencies with 92.5-97.8,98.5-99.7,and 97.8-99.7 % in terms of CO 2 for these three vessels.
CO emissions of HH were much higher than of XYH, followed by DFH.The power of their respective engines was 350, 600, and 1600 kW.In addition, there were large differences in CO emissions among different modes.All of these three vessels had relatively high EFs for CO while accelerating compared with other modes, but the highest EFs were during the idling modes of HH and DFH, as well as during the low-speed mode of XYH.Because CO emissions in diesel engines primarily depend on the excess air ratio (which determines the fuel-air mixture), combustion temperature, and uniformity of the fuel-air mixture in the combustion chamber (Doug, 2004), ship engines with lower power generally have higher CO emissions (Carlton et al., 1995).Localized hypoxia and incomplete combustion in the cylinder were the main reasons for CO emission of diesel engine.CO emissions always had positive relationships with the airto-fuel ratio.There was lower air-to-fuel ratio when in low engine load, which resulted in lower CO emission, and vice versa (Ni, 1999).
Much higher NO x EFs were observed for HH than for the other two vessels.These results were inconsistent with those of Sinha et al. (2003), in which emissions of NO x increased with the power of the ship engine.With increasing vessel speed, NO x EFs for HH first increased and then decreased.XYH had lower EFs when operating at high speed than at low speed.Nitrogen oxides included NO, NO 2 , and N 2 O in the present study.More than 70 % of the NO x was in the form of NO for all vessels, because most of the NO x emissions were generated through thermal NO formation (Haglind, 2008).The primary reasons that slow diesel en- gines such as the one in HH have higher NO x emissions include higher peak flame temperatures and the NO formation reactions being closer to their equilibrium state than in other engines (Haglind, 2008).NO x emissions from vessels are temperature-dependent (Sinha et al., 2003) and also are influenced by the oxygen concentration in the engine cylinder (Ni, 1999).In larger engines, the running speed is generally slower and the combustion process more adiabatic, resulting in higher combustion temperatures and more NO x .Besides, with the increasing of air-to-fuel ratios, concentration of NO x showed a tendency first to increase, then to decrease, which always had the maximum value in the operating mode that close to full load of engine because of the high temperature and oxygen in the engine cylinder (Ni, 1999).Furthermore, there were always higher EF values in acceleration process and lower in moderating process in this study.When the engines were in transient operating conditions, such as acceleration process or moderating process, concentrations of NO x always had corresponding changes in the cylinder.Studies about diesel engines showed that when the rotational speed had a sudden increase, there would be a first increasing, then decreasing and last stable tendency for the NO x concentrations, and vice versa (Tan et al., 2012).
TVOCs emissions from HH were much higher than from the other two vessels; the lowest emissions were observed for DFH.Previous studies (Sinha et al., 2003) have reported that hydrocarbon emissions from vessels depend on engine power, with low-power engines emitting more hydrocarbons.The present results were partially consistent with these previous studies.Besides, hydrocarbon emissions also depend on the percentage utilization of engine power (Sinha et al., 2003).As for various operating modes, TVOCs EFs had large differences.For example, HH had the highest TVOCs emissions in accelerating mode, which was almost 3 times the height of the lowest value in medium-speed mode.The EFs for SO 2 depended solely on the sulfur content of the fuels and were 1.6, 0.9, and 2.6 g kg −1 fuel for HH, DFH, and XYH, respectively in this study.Hydrocarbon could be generated because of the incomplete combustion.For example, in diesel cylinders, there will always be air present in wall regions and crevices; this is also the case when scavenging occurred during the aeration, which could cause the uneven mixing of air and fuel (Ni, 1999).
Fuel-based EFs for PM and its chemical components were shown in Table S6.OC and EC were the main components of PM, followed by SO 2− 4 , NH + 4 , and NO − 3 .Metals such as V, Ni, Cr, Fe, As, and Cd made up a proportionately small part of the total PM mass.However, other rare elements such as Tb, Er, Yb, and Lu had higher values than did some of the common elements.PM was an in-process product during the combustion in a cylinder; the forming process included the molecular cracking, decomposition, and polymerization, which resulted in lack of oxygen.High temperature and oxygen deficiency were the main reasons for the formation in diesel engines, which always had high concentration values in high load operating modes (Ni, 1999).HH had much higher PM emission factors than the other two vessels, the engine type was considered to be the most significant influence factor, which had a good agreement with NO x emission factors.
EFs for OC and EC and the ratios between them are shown in Fig. 3. EFs for OC and EC for HH were higher than for the other two vessels.Organic matter (OM) is generally calculated as OC × 1.2 (Petzold et al., 2008) to account for the mass of elements other than carbon in the emitted molecules.OM EFs for individual vessels mainly depend on the engine type and the amount of unburned fuel, i.e. the efficiency of combustion (Moldanová et al., 2013).BC emissions also depend heavily on the engine type (Lack et al., 2009).Therefore, the different types of engines and their levels of maintenance could account for the large differences in OC and EC EFs observed among the three vessels in this study.The ratios of OC-to-EC in the present study were much lower than those for large diesel ships reported previously (OC / EC = 12) (Moldanova et al., 2009) and also lower than that reported for a medium-speed vessel (Petzold et al., 2010).The usage of non-dilution sampling in this study was one possible reason for the lower OC to EC ratio.Besides, TOR was used to measure OC and EC in PM, which always had a lower OC content compared with other methods (such as TOT) because of the different definitions of OC and EC (Khan et al., 2012).Compared with other diesel engines, the ratios of OC to EC in this study were higher than that of automobile diesel soot, in which EC comprises 75-80 wt % of the total PM (Clague et al., 1999), and also higher than heavy heavy-duty diesel trucks (HHDDTs) with OC to EC ratios below unit for cruse and transient modes even though higher in cold-start/idle and creep modes (Shah et al., 2004).
Studies have shown that SO 2− 4 formed from vessel-emitted SO 2 is a major contributor to CCN and ship track formation (Schreier et al., 2006;Lauer et al., 2007).Sulfate is also an important component of PM emitted from vessels.In the present study, EFs for SO 2− 4 were much lower than previously reported (Petzold et al., 2008;Agrawal et al., 2008a), but similar to those detected by a high-resolution time-offlight aerosol mass spectrometer in a previous study (Lack  , 2009).This may be because EFs for SO 2− 4 are mainly related to the sulfur content of the fuel; SO 2− 4 is not generally emitted directly from the engines, but forms after release from the stack (Lack et al., 2009).Because PM was collected directly from engine emissions in the present study, the sulfur-to-sulfate ratios were low (< 0.6 % for vessels).Other ions such as NO − 3 and NH + 4 accounted for a small percentage of the PM emitted from the vessels compared with SO 2− 4 , consistent with previous studies (Lack et al., 2009).SO 2 is more easily oxidized to SO 3 in catalytic reaction cycles with metals commonly present in the exhaust gas (V, Ni), while hydroxyl radicals are additional needed to convert NO x to NO − 3 (Moldanova et al., 2009).Na + and Cl − were considered to originate from marine air.Their concentrations were highly correlated (r 2 = 0.78); the differing air demands of the engines under different conditions might have caused observed variations in the EFs relative to the fuel demand.
The elemental composition of PM in the present study differed from previous studies showing high elemental content of S, Ca, V, Fe, Cu, Ni, and Al (Agrawal et al., 2008a;Moldanova et al., 2009).V and Ni are typically associated with combustion of heavy fuel oil (Almeida et al., 2005).In the present study, the high-quality fuels resulted in low EFs for V and Ni.In our previous study, PM from shipping emissions was estimated to account for 2.94 % of the total PM 2.5 at Tuoji Island in China, using V as a tracer of shipping emissions (Zhang et al., 2014).Reconsidering the former results based on the EFs obtained in the present study, we determined that the contribution of vessels near Tuoji Island had been underestimated, because the estimate should have included both heavy and other types of fuels.However, some rare elements such as Tb, Er, Yb, and Lu had relatively high EFs compared with those of other elements in the present study, which may be related to the source of the fuels.

Fuel-based average emission factors
Based on actual operating conditions (Table S7), average EFs for the three vessels in the present study (according to Eq. 4) along with EFs from previous studies are shown in Table 3. EFs for all of the pollutants except SO 2 were significantly higher for HH than for the other two vessels, potentially due to poor combustion conditions.Most of the EFs for DFH and XYH were within the range of emissions for other vessels due to having well maintained engines and the high quality of the fuels used.The EFs for NO x , PM, and SO 2 were much lower than reported in previous studies (other than NO x for ocean-going vessels).All the sulfur of the fuels in the present study were significantly below the emissions limit of 3.50 % established by IMO in the revised MARPOL Annex VI rules, applicable since 2012 (IMO, 1998).The IMO Tier I emissions limit for NO x is 45.0 × n −0.2 g kWh −1 (n, rated speed, 130 < n < 2000 rpm).The rated speed and fuel consumption rates for each vessel are shown in Table 1.Thus, the emissions limits for HH, DFH, and XYH would be 54.5, 57.5, and 56.5 g kg −1 fuel, respectively, calculated combined with Eq. ( 5).The average fuel-based EFs for NO x of ship HH was more than 100 % above the IMO standard, while those of the other two ships were below the IMO standard (Table 3).PM emissions for HH were also higher than previously reported, but those for the two research vessels were much lower (Table 3).Fuel type is one of the most important influence factors on pollutant emissions, for example, sulfur content in the fuel not only influence the SO 2 emission directly, but also had impact on PM formation in the flue gas stack with low sulfur content in fuels reduces PM formation (Lack et al., 2011).Vessels with higher sulfur content always had relatively higher PM emissions, which were also shown in Table 3.In addition, different engines and levels of maintenance have a significant impact on all combustion-dependent emissions.Emission reduction measures have been used in some vessels.For example, NO x emissions can be reduced by measures such as selective catalytic reduction (SCR) and direct water injection (DWI), which had been implemented on some vessels previously studied in a harbour in Finland (Pirjola et al., 2014).The results showed that SCR effectively reduced NO x emissions, while vessels with DWI had high PM emissions.The engine type might be an important cause of the different emissions, such as HH had much higher pollutants emissions with an engine produced in China and yet DFH's engine produced in Germany.Besides, emission tests for a high-speed marine diesel engine with different kinds of diesels showed that diesel type had limited influence on emissions such as NO x , CO and CH, but a significant impact on PM emission (28.9-41.5 %) because of the different sulfur content in fuel (Xu, 2008).

Power-based emission factors
Based on the engine power and fuel consumption rates of the vessels, power-based EFs were calculated (according to Eq. 5) and compared to results from previous studies (Table 4).The EFs for HH were much higher than those for the other two vessels, except for SO 2 .HH also had significantly higher EFs for NO x than previously reported values, while EFs for NO x of DFH and XYH were within the range of previously reported results.Engine type was considered to be a significant influence factor for NO x emissions, with lower engine speed having higher NO x emission factors (Celo et al., 2015).In addition, compared to other vessels with a similar engine type and diesel fuel, HH still had relatively higher NO x EF (seen in Table 4), which could reflect the impact of engine condition (engine quality and maintenance level) on shipping emissions.CO EFs for the test vessels in the present study were higher than previous studies, which produced similar results to those of inland ships and other test vessels in China (Fu et al., 2013;Song, 2015).In spite of the influence of engine type on CO emissions that with the higher engine speed having higher CO EFs (Celo et al., 2015), engine condition combined with fuel quality might have significant influence.All of the EFs for SO 2 in the present study were lower than those in previous studies, because of the low sulfur content of the present fuels.Generally, PM emissions from marine diesel fuels are dependent on the fuel (sulfate and metal oxide ash constituents) and on combustion conditions (unburned hydrocarbons and carbon residue constituents) (Cooper, 2003).HH had the highest PM emissions among the test vessels, although there were almost no differences among the fuels (Table S6).Besides, HH had even higher PM EFs than previously reported vessels with HFO fuel, and XYH had much lower PM EFs than all the other vessels with even lower sulfur content fuel.Therefore, combustion conditions were likely the determining factor for the differences.It can be seen from Table 4 that most previous studies focused on the heavy fuel oil of shipping emissions.Compared with diesel fuels, heavy fuel oil always had relatively low CO emission factors and high PM emission factors.And among the heavy-fuel-oil-using vessels, engine type (engine speed and engine power level) always played an important role on emissions such as NO x and CO, which with lower engine speed having higher NO x EFs and lower CO EFs.
Combined with other emission data of test ships in China (Fu et al., 2013;Song, 2015), it could be seen that inland and some test offshore ships in China always had higher NO x , CO, and PM emissions compared with other test vessels in previous studies.And among the test vessels in China, there also were differences for different engine types and ship types.In addition, emission factors that were used for calculation of ship inventories in China always came from other countries and areas.However, there seemed to be significant differences between the reference and test data, such as 10.0 to 13.2 g kW h −1 of NO x EF and 1.1 to 1.7 g kW h −1 of CO EF used for inland ships for ship inventory calculation (Zhu et al., 2015), 10.0 to 18.1 g kW h −1 of NO x EF and 1.1 to 1.5 g kW h −1 of CO EF for harbour ships (Yang et al., 2015), compared to 15 to 17.3 g kW h −1 of NO x EF and 4.6 to 10.3 g kW h −1 of CO EF from test inland ships (convert the fuel-based EF to power-based EF with a factor of 200 g kW h −1 ) (Song, 2015), and 6.97 to 25.8 g kW h −1 of NO x EF and 1.39 to 7.38 g kW h −1 of CO EF in the present study (Yang et al., 2015).Besides, whether there are obvious differences of EFs between other types of vessels in China (such as low-speed engine vessels with heavy fuel oil) and previous studies is still unclear.Therefore, much more measurement data for different vessels in China are still in urgent need for more accurate assessment of shipping emissions.
3.5 Impact of engine speed on NO x emission factors NO x is formed in the combustion chamber by a combination of atmospheric nitrogen and oxygen under high-pressure and high-temperature conditions.Many factors affect NO x formation, including engine temperature, injection point, and fuel quality.The IMO emissions limit for NO x is determined by the rated speed of the engine; however, other factors must also be considered to reduce NO x emissions.
The NO x EFs for the test vessels at various engine speeds are shown in Fig. 4. The rated speeds of the vessels were 1200, 900, and 1000 rpm for HH, DFH, and XYH, respectively.The actual engine speeds of HH were much lower than the rated speed, while the two larger-engine-power vessels operated close to their rated speeds, except during one operating mode of DFH.The NO x EFs for HH differed significantly in different operating modes, ranging from 39.1 to 143 g kg −1 of fuel.The NO x EF was highest when the engine speed reached ∼ 750 rpm (Fig. 4).At lower engine speeds, the NO x EFs had fluctuating but lower values.At higher engine speeds closer to the rated speed of 1200 rpm, the NO x EFs were much lower.The NO x EFs for the two larger-engine-power vessels changed slightly with engine speed, but also had lowest values when their engine speeds approached their rated speeds.Combined with the diesel propulsion characteristic curve, there were large increases in the fuel consumption rate when the engine speed increased.Therefore, a best-fit engine speed should be determined based on both EFs and economic costs.
Engineering approaches for reducing the NO x emissions of marine engines may be applied before, during, or after the combustion process (Verschaeren et al., 2014;Habib et al., 2014).In the present study, the NO x EFs of the two research vessels were below the IMO Tier I emissions limits.However, for EMS, measures should be taken to meet the IMO emissions limit, including increasing the engine speed and applying engineering technologies during or after combustion, such as exhaust gas recirculation (EGR), selective non-catalytic reduction (SNCR), or SCR.

Conclusions
Three offshore vessels with different engine power sources were chosen in this study to collect measured data of gaseous species and particulate matter, including NO 2 , NO, N 2 O, CO, CO 2 , TVOCs, SO 2 , and the total suspended particulate.Besides, chemical composition of the PM were also analysed to give detailed EFs for OC, EC, water-soluble ions and metal elements.Concentrations, fuel-based EFs, fuel-based average EFs as well as power-based average EFs for species of offshore vessels in China were presented.Furthermore, impact of engine speed on NO x EFs was also discussed.
There were higher concentrations of pollutants for lowengine-power vessel HH than for the other two vessels.CO concentrations for offshore vessels were slightly lower than inland vessels in China, and all the three vessels had the lowest CO concentrations at their economic speeds (the speed of the least vessel operating expenditures during one voyage, they were high-speed mode, cruise mode, and high-speed mode for HH, DFH and XYH, respectively).More than 80 % of the NO x was NO, and all the offshore vessels had higher NO x concentrations in high-speed modes.Because of the low-sulfur fuels used in this study, SO 2 concentrations of these three offshore vessels were lower than that in the literatures.And the PM concentrations were much lower than inland vessels while showing significant differences among different operating modes.
Fuel-based EFs for gaseous species and PM were presented based on the carbon balance method.EFs for CO 2 were 2940-3106, 3121-3160, and 3102-3162 g kg −1 fuel for HH, DFH and XYH.Because of the combustion conditions such as excess air ratio, combustion temperature and uniformity of the fuel-air mixture, EFs for CO showed high values in idling mode, but low values in economic speed.All the offshore vessels had higher NO x EFs in low speed than in high speed, but showed higher values when in acceleration

Figure 1 .
Figure 1.On-board emissions test system and measured analytes.
1) where C F represents the mass of C in per kg diesel fuel (g C kg −1 fuel); R FG represents the flue gas emissions rate (m 3 kg −1 fuel); and c(C CO ), c(C CO 2 ), c(C PM ), and c(C TVOC ) represent the mass concentrations of carbon as CO, CO 2 , PM, and TVOC (g C m −3 ) in the flue gas, respectively.

F.
Zhang et al.:  Emission factors for gaseous and particulate pollutants 3 Results and discussion3.1 Concentrations in shipping emissionsConcentrations of CO, NO x , SO 2 , TVOC, and PM from the three vessels are shown in Fig.S1in the Supplement.

Figure 2 .
Figure 2. EFs for the typical pollutants in different operating modes.

Figure 3 .
Figure 3. EFs for OC and EC and the ratios between them.

Figure 4 .
Figure 4. Emission factors for NO x at different engine speeds.

Table 1 .
Technical parameters of test vessels.

Table 2 .
Results from the fuel analysis (diesels).

Table 3 .
Fuel-based average EFs in the present study and previous studies (g kg −1

Table 4 .
Power-based EFs in the present study and previous studies (g kWh −1).
F. Zhang et al.: Emission factors for gaseous and particulate pollutants