VOC samples from nine different ships were collected in this study, including two coastal cargo ships, three ocean-going vessels, and four inland cargo ships in the Yangtze River. The detailed technical parameters of the sampling ships are shown in Table . Different types of cargo ships had different technical parameters in China. For example, the engine power of coastal cargo ships varied largely, with about 57 % being equipped with engines of more than 500 kW. Of the other remaining coastal cargo ships, 17 % of which range from 150 to 250 kW. Therefore, one large coastal cargo ship with the main engine power of 1470 kW and another small coastal cargo ship with main engine power of 178 kW were selected here. Coastal cargo ships typically transport cargo among different coastal ports with 1 to several days per voyage. The main operating modes are cruise (∼ 75 % engine load), maneuvering (low and variable engine loads), and idling.

Ocean-going vessels usually have large tonnages with large power main engines. Statistical automatic identification system (AIS) data show that engines with power from 4 to 10 kW account for the largest proportion (∼ 25 %) of the total OGVs in China, followed by 2 to 4 kW (∼ 23 %) and 10 to 20 kW (∼ 20 %). Besides, newly built OGVs have a tendency to have greater and greater engine power. Hence, three ocean-going vessels with different engine power levels ranging from 13.5 to 15.7 kW were tested in this study. They are designed for transporting goods across borders, usually with several months per voyage. The main operating mode is cruising in the open ocean. While during the processes of coming in and out of the port, the engines of OGVs are typically active in the maneuvering mode, with relative lower and variable engine loads, which could have a great influence on the nearshore environment due to higher emission levels of pollutants.

Most inland cargo vessels are generally equipped with high-speed, small main engines powered by up to 1000 kW (∼ 70 %). Among them, the vast majority are below 500 kw. Therefore, four typical inland cargo ships with engine power between 138 and 300 kW were chosen in this study. The inland cargo vessels are typically active among different inland ports or coastal ports near inland rivers with several hours to several days per voyage. Affected by the complicated water conditions of inland rivers, cruise and maneuvering are the most important operating modes for inland cargo ships.

In brief, the measured ships in this study could represent the typical cargo ships in China to a certain extent. It is worth noting that the ocean-going vessels were newly constructed ships, while the inland cargo ships had older engines (6 to 14 years) compared with other types of ships (younger than 10 years).

Besides, most large cargo ships are equipped with both a main engine and an auxiliary engine. The main engine provides navigation power, and the engine loads vary greatly with the different operating modes. While the auxiliary engine mainly provides domestic electricity or heating on board, and the engine load is relatively stable with about 75 % load. Small cargo ships are equipped only with main engines, such as the tested inland cargo ships and small coastal cargo ships in this study.

Characteristics of the HFO and diesel oil used for the test ships in this study are shown in Table S1 in the Supplement. In order to meet the requirements of diesel engines for non-road mobile machinery in China, regular diesel (0#) was used for all inland cargo ships here. Results showed that the sulfur contents of all the fuels were no more than 0.5 % (m/m), which were within current ship emission control standards of both China and the IMO. As typical tracers of high-sulfur-content HFO, nickel and vanadium content levels and their ratios were still higher but not distinguishable enough in low-sulfur-content HFO compared with diesel oil, which provides further evidence that it is necessary to be cautious when using them as tracers of ship emissions under the current low-sulfur regulation. It should be noted that much higher levels of calcium and zinc were detected in lubricating oils of OGVs.

A portable dilution sampling system was used in this campaign, and the components and principles have been described elsewhere (Zhang et al., 2018). Briefly, two separate sampling pipes were placed into the exhaust stacks (about 1.5 m deep of the exhaust outlet) to route emissions from the main engine and auxiliary engine, respectively, to the sampling system on the highest deck of the ship. Then, the probe of a flue gas analyzer (testo 350, Testo, Germany) was placed into the sampling pipe to test the gaseous matters directly to obtain online data (CO2, O2, CO, NO, NO2, and SO2). Another probe was used to extract the flue gas for the diluted system. The dilution ratios ranged between 1–10 in this study. VOC samples were collected by summa canisters from both the main engines and auxiliary engines of all the ships listed in Table . The sampling time was 20–30 min for each sample, according to actual operating conditions.

A total of 48 VOC samples was obtained for the test ships, involving different engine types with different fuels under different operating modes (see Table S2 for detailed information). For the coastal/inland cargo ships, all samples were collected based on actual operating modes (about 1 to several days from one trip). For ocean-going vessels, samples from many more operating modes could be obtained thanks to the testing of the newly constructed ships (about 1 week from one trip).

As shown in Table S3, a total of 106 volatile organic compounds was detected in this study, according to United States Environmental Protection Agency (U.S. EPA) TO-15 1999, including 11 oxygenated volatile organic compounds (OVOCs), 17 aromatics, 29 alkanes, 11 alkenes, 35 halohydrocarbons, and 4 other species. These measured VOC species were typical concerning VOCs, could be considered the main VOC components referring to relative studies (Huang et al., 2018a; Wu et al., 2020; Araizaga et al., 2013), and could reflect the emission conditions of ship exhaust fumes. As shown in Eqs. () and (), the carbon balance method was used to calculate the EFVOCs, which was also introduced in our previous study (F. Zhang et al., 2016). 1EFx=ΔXΔCO2⋅MXMCO2⋅EFCO2, where EFx is the EF for VOC species X (g kg-1 fuel), ΔX and ΔCO2 represent the concentrations of X and CO2 with the background concentrations subtracted (molm-3), MX represents the molecular weight of species X (gmol-1), MCO2 is the molecular weight of CO2 (44 gmol-1), and EFCO2 is the EF for CO2 (g kg-1 fuel). 2EFCO2=CFc(CCO)+c(CCO2)+c(CPM)+c(CHC)×c∗(CO2)⋅MCO2, where CF represents the mass of carbon in 1 kg diesel fuel (gC(kgfuel)-1); c(CCO), c(CCO2), c(CPM), and c(CHC) represent the mass concentrations of carbon as CO, CO2, PM, and HC (gCm-3), respectively, in the flue gas; and c∗(CO2) is the molar concentration of CO2 (molm-3).

Detailed calculation processes of normalized ozone reactivity (RO3; gO3g-1 VOCs), OFP (gO3kg-1 fuel), normalized secondary organic aerosol reactivity (RSOA; mgSOAg-1 VOCs), and SOA formation potential (SOAFP; mgSOAkg-1 fuel) are given in the following.

Normalized ozone reactivity (RO3; gO3g-1 VOCs) and OFP (gO3kg-1 fuel) were calculated using the maximum incremental reactivity (MIR) coefficient method (Carter, 2010a) which represents the maximum contribution of VOC species to the near-surface O3 concentration under optimal conditions. The equations are as follows: 3RO3=∑i(ωi×MIRi), where ωi is the mass percentage of the total VOC emissions for species i, and MIRi is the MIR coefficient for VOC species i, which was referenced from Carter (2010b) (see Table S3 for details). 4OFP=∑i(MIRi×[VOC]i), where OFP is the ozone formation potential (gkg-1 fuel), and [VOC]i is the emission factor for VOC species i (gkg-1 fuel).

As for O3, the normalized secondary organic aerosol reactivity (RSOA; mgSOAg-1 VOCs) and SOA formation potential (SOAFP; mgSOAkg-1 fuel) were also calculated. Their equations are as follows: 5RSOA=∑i(ωi×Yi),6SOAFP=∑i(EFi×Yi), where Yi is the SOA yield for VOC species i (see Table S4 for details). The SOAFP of VOCs under high-NOx and low-NOx conditions was calculated.

Rigorous quality assurance and quality controls were conducted during the whole experiment. Ambient air blanks were analyzed in the same way as mentioned above to determine background concentration. The VOC concentrations of each sample were obtained by subtracting ambient air blank results. Duplicate samples, as well as standard gas, were examined after analyzing a batch of 10 samples to ensure that the error was within 5 %.

EFVOCs for the test ships are shown in Fig.  and Table S5. In order to calculate the EFVOCs and investigate their influence factors, EFs of other gaseous pollutants such as CO2, CO, NO, and NO2 were also given and discussed briefly. For CO2, the emission factors ranged from 2622 to 3185 gkg-1 fuel that was influenced by both fuel type and operating mode. CO showed the opposite trend with CO2, varying from 0.62 to 180 gkg-1 fuel and reflecting the condition of the combustion efficiency. The EFNOx ranged from 6.26 to 92.8 gkg-1 fuel, with 60 % to 99 % NO, which inferred the condition of the combustion temperature in the cylinder.

Results showed that the EFVOCs for all test ships presented wide differences, which ranged from 0.09 to 3.01 gkg-1 fuel. Ship type, engine type, operating mode, and fuel type could influence the EFVOCs that will be discussed in more detail in Sect. . Briefly, higher VOCs had been observed both in low-load and high-load operating modes such as maneuvering and idling, while in medium-load operating modes, the EFVOCs presented lower levels (a detailed result is also shown in Fig. a). Main engines presented obviously higher EFs levels than auxiliary engines (see Fig. c for details). And CCSs and ICSs had relatively higher EFs compared with OGVs (see Fig. d for details). It was worth noting that when the fuels were switched from HFO to marine diesel oil for OGVs, increasing trends were presented for EFVOCs in this study, while the CCSs showed the opposite trend with a slight decrease for EFVOCs.

Average EFVOCs emitted from ships in this study were also compared with those reported in other studies (Table ). Altogether, the measured EFVOCs varied largely from 0.02 to 23.7 gkg-1 fuel for all test ships. Complex factors could lead to the large uncertainty, such as the different detected VOC species in different studies and different engine types and fuel qualities. This also indicated that the uncertainty should be noticed when EFVOCs were used as basic data to calculate the emission inventory or estimate other environmental influences. The test ships in this study presented a comparable level of EFVOCs with other studies. It seemed that OGVs with large engines typically showed a lower level of EFVOCs no matter what types of fuel was used when compared with river ships and coastal ships. Moreover, compared with on-road vehicles with diesel fuel (H. Zhou et al., 2019), VOCs emitted from non-road engines, such as ships, agricultural machinery, and construction machinery, had much higher levels (Huang et al., 2018a; Hua et al., 2019; Zhou et al., 2022), which should be paid more attention, especially in the case of more and more strict limitations of VOCs being applied to on-road vehicles.

Components and mass fractions of VOCs from the test ships are shown in Figs.  and S1 in the Supplement. OVOCs and aromatics were the main components of the detected VOC species, accounting for 9.38 %–88.0 % and 5.38 %–74.0 % of the total VOCs, respectively. Alkanes also accounted for non-ignorable fractions, which ranged from 0.2 % to 57.2 %. Alkenes, halohydrocarbons, and other quantified species only contributed small fractions of the total VOCs. The results in this study were consistent with that of Huang et al. (2018a) about a large ocean-going bulk carrier but showed different VOC components compared to that of Wu et al. (2019) for a coastal container ship and Wu et al. (2020) for auxiliary engines at berth. The different detected VOC species in different studies played an important role in the differences, while the fuel type and its chemical composition might also have considerable impacts.

The main VOC components of OVOCs, aromatics, alkanes, and alkenes presented different variation patterns under different operating modes, fuel types, and engine types due to their different formation mechanisms (Fig. ). For example, OVOCs from diesel engines are typically from the oxidation of small molecular weight, yet incomplete, combustion hydrocarbons (Hao et al., 2014; Pan, 2008); therefore, the operating mode and engine type could influence the EF levels obviously but not the fuel type. The direct emission of unburned fuel components and pyrosynthesis (formation of aromatics by the regeneration of fragmented radical species) are the two main formation processes of polycyclic aromatic hydrocarbons (PAHs) (Radischat et al., 2015). EFs of aromatics showed relatively higher levels in medium operating modes compared with other modes in this study. One main reason was that the higher temperature in medium operating modes promoted polymerization, resulting in the processes of dehydrogenation and PAH formation (Zhang et al., 2021) which exceeded the direct emission of unburned fuel components (Radischat et al., 2015). Alkanes are mainly from the incomplete combustion of fuels; therefore, alkanes from diesel fuel presented higher EFs than HFO because of the higher aliphatic compounds in diesel fuel (Liu et al., 2022; Sippula et al., 2014). Alkenes emitted from diesel engines are always related to the pyrolysis process of the fuel combustion in the cylinder (Alotaibi et al., 2018; M. Zhang et al., 2022). As a result, in high operating modes of more than 90 % engine loads, it had higher EFalkenes levels in this study due to the pyrolysis process under higher temperature and incomplete combustion because of the smaller air-to-fuel ratios in the cylinder.

It was mentioned above that influence factors such as the operating condition, engine type, ship type, and fuel type could affect the emission level and component of VOCs from ship exhaust fumes. Box-and-whisker plots of VOC emission factors under these different drivers are presented in Fig. . As shown in Fig. a, the engine load could affect the VOC emission significantly. EFVOCs had the lowest level when the engines were operating in medium loads and the highest in low loads. This was consistent with the results of VOC emissions reported by previous studies such as Huang et al. (2018a), Wu et al. (2019), and Radischat et al. (2015), which are also shown in Fig. S2. The combustion condition in the cylinder could be responsible for the variation in the VOC emissions, which means that incomplete combustion was one principal reason for the high VOC emissions.

Engine type is also one significant influence factor of VOC emissions. The engines were classified into three types in this study according to their engine speed, including low-speed engines (LSEs; rated speed < 100 rpm), medium-speed engines (MSEs; 100 rpm ≤ rated speed < 1000 rpm), and high-speed engines (HSEs; rated speed ≥ 1000 rpm). It can be seen from Fig. b that with the increase in engine speed, the EFVOCs showed an increasing trend. This could be explained by the fact that, when compared with HSEs, LSEs with high engine power usually had higher combustion efficiencies that led to lower levels of VOC emissions (Zhang et al., 2018).

The EFVOCs between the main engines and auxiliary engines also varied obviously. The average EFVOCs from the main engines were 2.3 times that of auxiliary engines in this study (seen in Fig. c). A similar result was also reported by Liu et al. (2022) about the intermediate volatile organic compound (IVOC) emission for the same test OGVs, even though the auxiliary engines were mainly high-speed or medium-speed engines that had higher VOC emissions as mentioned above. Owing to the much lower VOC emissions in medium loads that the auxiliary engines have been using, it could be inferred that the impact of the operating condition exceeded that of the engine type to VOC emissions.

As seen in Fig. d, the EFVOCs varied obviously under different types of ships, with CCSs having the highest levels and OGVs the lowest. This could be explained by the combined influence of the operating condition and engine type, as mentioned above. First, as shown in Fig. b, high-speed and medium-speed engines were equipped for the CCSs, and they could lead to a higher emission factor of VOCs (EFVOCs) compared with low-speed engines that are equipped for OGVs. Besides, the unstable operating conditions of CCSs and ICSs, such as maneuvering and low load, also promoted the emission of VOCs (Radischat et al., 2015). Therefore, it could be indicated that coastal areas with a high population density need get more attention due to the higher VOC emissions from CCSs and ICSs.

As mentioned before, the fuel type could influence the EFVOCs significantly (Wu et al., 2019, 2020), which also would be one of the most important influence factors in the future under the background of increasingly strict ship oil policies. Under the condition of low-sulfur-content fuels in China, the average EFVOCs were 592 and 878 mgkg-1 fuel for diesel and HFO in this study, respectively (seen in Fig. e). In addition to the direct emission of unburned fuel components, VOCs also could be emitted from the pyrosynthesis of the fuel in the cylinder (Radischat et al., 2015). In order to explore the relationship between the chemical composition of low-sulfur-content fuel and VOC emissions, n-alkanes, b-alkanes, and aromatics in the fuels from OGVs were tested (Liu et al., 2022) (see Table S6 for details). Obviously, diesel had higher n-alkanes and b-alkanes content than HFO, and aromatics were the opposite. It can be seen from Fig. S3 that the EFAlkanes, EFAlkenes, and EFhalohydrocarbons from ships with diesel presented higher levels compared with that of HFO. EFAromatics and other components showed the opposite trends, while no obvious difference in EFOVOCs was observed between diesel and HFO. Emission characteristics of the VOC main components were basically consistent with the fuel composition in this study. It could be argued that the composition of fuel did have a significant impact on VOC emissions.

To further explore the impact of the sulfur content of fuel on VOC emissions, EFVOCs of low-sulfur-content fuel (< 0.5 % m/m) and high-sulfur-content fuel (≥ 0.5 % m/m) in this study and previous studies were summarized in Fig. f. The average EFVOCs from low-sulfur-content fuel was significantly higher than that of high-sulfur-content fuel by almost 3.4 times. This indicated that when the fuels were switched from high sulfur to low sulfur there was a dramatic increase in VOC emissions. Low-sulfur-content fuels are usually produced in three ways, including a blending technique that uses light low-sulfur oils mixed with heavy high-sulfur oils; heavy-oil hydrogenation technology that removes sulfur through the hydrogenation of high-sulfur residual oil; and biological desulfurization technology that uses microbial enzymes to catalyze and oxidize the organic sulfur in oil, convert it into water-soluble sulfide, and then remove the sulfur (Kuimov et al., 2016). Among these, blended low-sulfur oils are the most widely used oils (Zhang, 2019; Han et al., 2022). Except for light low-sulfur oils mixed during the production of low-sulfur oils, other non-petroleum refined oils, such as coal tar and chemical waste, are also added. Consequently, emission factors, as well as the composition of VOCs, have changed significantly. Since low-sulfur-content fuels (< 0.5 % m/m) have been used worldwide since 2020 and 0.1 % (m/m) in ECAs since 2015, it would imply that the impact of fuel type on VOC emissions needs to be given sufficient attention.