2.1.1 Information from Jingtang Port: sampling sites and ships

The ambient sampling site was located in Jingtang Port, Tangshan City, Hebei Province, China. Jingtang Port is located in Bohai Bay and belongs to the Port of Tangshan, which is one of the core ports in the domestic emission control area. According to the China Port Yearbook 2015, the annual ship traffic in the Port of Tangshan reached 15 084 and the total throughput exceeded 500 million tons, ranking it 5th among global port throughputs. Jingtang Port area is surrounded by the “Port Economic Development Area”, which has a population of 78 300. Tangshan is a typical industrial city with an average PM_2.5 concentration of 117 µg m⁻³ in winter (Zhang et al., 2017). The current PM_2.5 source apportionment studies in Tangshan have not included shipping emissions due to a lack of basic information and research. However, background information indicates the significance and urgency of studying the impact of shipping emissions and the effect of the fuel switching policy.

As is shown in Fig. 1a, the population center is mainly concentrated in a residential area, located to the north, about 2 km away from the port. Approximately 2.5 km to the west of the port area there is a thermal power plant with after-treatment facilities, which operates according to the strict emission control standards of power plants in China. Between the port and the other zones are two main roads that trucks use to carry containers in and out of the port; the roads are about 1 km away from the sampling site. Besides trucks and the power plants, there are no further emission sources near the port area.

Figure 1(a) The location and surroundings of Jingtang Port shown at both a large scale (map inset) and smaller scale. The yellow marker represents the sampling site inside the port area. The road, the residential distribution and their distance from the port can be observed from this map. (b) The location of the sampling site at a further amplified scale. The distribution of the berths, pools and the surroundings of the sampling site are illustrated on this map.

Download

The site where ambient particle collection and instrumental analysis was carried out is surrounded by four pools and a channel, and is located on an open, flat corner close to the Number 26 and Number 27 berths as well as the container yard inside the port (Fig. 1b). No tall buildings exist around the sampling instrument. The distribution of the berths, the pools and the sampling site guarantees that plumes from ships at berth are prone to reach the sampling instrument.

The information for the 20 container ships included in this study was collected via on-board inquiry and is listed in Table 2.

Table 1Abbreviations used in this text.

Download Print Version | Download XLSX

Table 2Brief information regarding the 20 ships sampled.

Download Print Version | Download XLSX

2.1.2 VOC sampling and analysis

VOCs from ships at berth were sampled using Entech SUMMA canisters with a standard volume of 3.2 L. When auxiliary engines were in operation, 1 m Teflon tubing was stretched into the exhaust pipe and the other end was linked to the SUMMA canister. The flow rate was kept constant by an Entech CS1200ES passive canister sampler, which also filtered impurities like particles and ashes. The first batch of samples was regarded as preliminary tests in order to determine proper dilution factors, sampling time and flow rate. Sample dilution was conducted using an Entech 4600 dynamic diluter with various dilution factors ranging from 10 to 80, depending on the original sample concentration. An Agilent 5975C-7890A GC-MS was calibrated with standard gas and used for analyzing diluted VOC samples. A total of 93 VOC species were detected and the mass percentage of single compounds could be calculated according to the sample inlet volume, the dilution factor and the calibration curve. Owing to the limit of the chromatographic column property equipped on the GC-MS, alkanes and olefins with a carbon number smaller than four were not detected. VOCs with a carbon number larger than six are more relevant to the yield of secondary organic aerosols (Gentner et al., 2012).

In order to complement the quantification of lighter hydrocarbon compounds with two to three carbons, selected ion flow tube mass spectrometry (SIFT-MS) was applied, which has been recognized as a real-time analytical technique for combustion gases and components in exhaled breath (Smith and Spanel, 2010). This part of the analysis was done without pretreatments and took place immediately after the gaseous samples were taken from auxiliary engines. VOC samples were transferred from the canister to a Teflon bag using an Agilent headspace syringe with a standard volume of 10 mL and then diluted with nitrogen. The species that SIFT-MS quantified were not totally consistent with the GC-MS. According to the Photochemical Assessment Monitoring Stations (PAMS) and Toxic Organics-15 (TO-15) standard gas applied in GC-MS calibration, 51 species were selected and normalized by mass in SIFT-MS data. Among the 51 species, acetylene, ethane, ethene, propane and propene were the most significant low-carbon compounds that were not quantified by GC-MS. The mass proportion of these species was calculated according to the quantification results obtained by SIFT-MS analysis, and the impact on ozone forming potentials could then be evaluated.

2.1.3 Particle sampling and analysis

Ship exhaust particles were collected directly from the exhaust pipes of the auxiliary engines on ships utilizing Tedlar bags and metal tubing designed specifically for particle sampling. The whole sampling process was achieved using a non-contact sampling box and air pump. Samples were then sent to the single particle aerosol mass spectrometer (Hexin Analytical CO., Ltd) (Li et al., 2011) as soon as possible to be analyzed. SPAMS shared a common principle and mechanism with the aerosol time-of-flight mass spectrometer (ATOFMS), in that it is frequently applied in the online measurement and analysis of single particle aerosol from heavy diesel vehicle exhausts (Shields et al., 2007) and biomass burning (Bi et al., 2011; Xu et al., 2017).

Ambient particle sampling was conducted from 27 December 2016 to 15 January 2017, spanning about 20 days. Ambient particles were sampled and analyzed by SPAMS, with the inlet fixed at a height of 3.6 m from the ground level.

3.3.1 Average ion mass spectra

Samples collected by Tedlar bags and glass bottles from 20 ships at berth were analyzed using SPAMS. Excluding some invalid samples with very few particles, the average ion mass spectra of both positive and negative ions expressed in relative area were obtained after SPAMS analysis and classification (Fig. 5).

Figure 5Average ion mass spectra derived from SPAMS analysis of 20 samples of ship exhaust. Standard deviations are given in the figure and typical ion peaks are also marked in both the positive and negative ion mass spectra.

Download

The average ion mass spectra presented the overall characteristics of the total 20 ships. For positive ion spectrum, the Na⁺ signal was observed to be the highest peak. Other abundant ions were EC and OC, occurring in the following forms: C⁺, ${\mathrm{C}}_{\mathrm{2}}^{+}$, ${\mathrm{C}}_{\mathrm{3}}^{+}$ and so on for EC; and C₃H⁺, ${\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{3}}^{+}$ and so on for OC. It was observed that the relative signal area of EC ions was higher and more widely distributed than for OC ions. This was consistent with previous research that found that fuel combustion produced more EC than OC particles. Although not as obvious as the abundant ions listed above, the percentages of V⁺ and VO⁺ were relatively low but not negligible.

For the negative ion spectrum, the components were not as complex as for the positive ion spectrum. The highest peak was ${\mathrm{HSO}}_{\mathrm{4}}^{-}$ with m∕z = −97, which exceeded 50 % of the overall negative ion relative area. ${\mathrm{HSO}}_{\mathrm{4}}^{-}$ is a typical marker of sulfate particles, which are formed via secondary reaction in the atmosphere. This will be discussed further in following sections. EC signals such as ${\mathrm{C}}_{\mathrm{2}}^{-}$, ${\mathrm{C}}_{\mathrm{3}}^{-}$ and ${\mathrm{C}}_{\mathrm{4}}^{-}$ were also observed in the negative spectrum. In addition, markers of nitrates with m∕z = 46 and 62 were ${\mathrm{NO}}_{\mathrm{2}}^{-}$ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$. Nitrates were regarded as secondary particles that might have come from the aging of primary particles emitted directly by fuel combustion or the transformation of gas-phase NO₂. The relative area of nitrate signals was much lower than ${\mathrm{HSO}}_{\mathrm{4}}^{-}$, illustrating that nitrate particles were not predominant in ship exhausts and not related to the main research goal in this study.

3.3.2 Manual grouping results

Based on single particle ion spectrum similarity and the ART-2a neural network algorithm, particles from ship exhaust were further grouped into clusters of 175, among which the top 86 clusters covered 95 % of total particle numbers, with the rest defined as others. Clusters were set as OC, EC, ECOC, Na (Na-Rich), K (K-rich), V (V-rich), Cu (Cu-rich), Fe (Fe-rich), Mn (Mn-rich) and others, depending on the similarity and manual grouping. The method and criteria of the manual grouping mainly focused on positive ions. Particle grouping is fundamental for accurate source apportionment in a target region.

Following this, a composition analysis was conducted, the result of which are shown in Fig. 6. It was shown that EC and ECOC particles dominated with respective proportions of 35.74 and 33.95 %, while Na-rich particles ranked 3rd with a percentage of 21.12 %. Data regarding EC and ECOC indicated that carbon emitted by fuel combustion tended to produce more EC and ECOC than pure OC particles. The amount and fraction of OC particles were lower than those of EC and ECOC particles by a factor of ∼ 10. Meanwhile, OC tended to attach to EC particles, forming ECOC particles and increasing the fraction of ECOC particles. In view of the composition of fuel combusted by marine engines, Na does exist with a mass concentration of around 13–22 mg kg⁻¹ in several samples of heavy fuel oil (HFO) and at a lower concentration in marine diesel oil (Moldanová et al., 2013; Lack et al., 2009). Moldanová et al. (2009) analyzed particles from ship exhausts using two-step laser mass spectrometry, and Na⁺ was observed to have an obvious peak in the mass spectra. According to Sippula et al.'s (2014) research on marine diesel engines, Na can also exist in lubrication oil and be exhausted along with the lubrication and grinding process. It is widely recognized that the metal content existent in exhausted particles has a strong correlation with that in fuel (Celo et al., 2015; Moldanová et al., 2013). Although the content of Na is obviously lower than that of vanadium (55–133.8 mg kg⁻¹), the peaks of Na⁺ in ion mass spectra are mostly higher than the peaks of the V⁺ ∕ VO⁺ signal. This phenomenon indicates that there are external sources of Na and that one of the most possible of these is sea salt, where the typical element is Na. In conclusion, the Na-rich particles come from both fuel composition and sea salt perturbation. Only particles with a relative area of Na⁺ obviously higher than other positive ions were counted as Na-rich particles. The same rule was also adopted for K-rich particles. Owing to the fact that the content of K in most marine fuels is lower than that of Na, V and other typical metallic elements, one of the possible sources of K-rich (as with Na-rich particles) particles is sea salt (O'Dowd et al., 1997). Na⁺ and K⁺ were both widely spread among all particles. V-rich particles were likely to contain mixtures of ammonia sulfate or sulfuric acids (Divita et al., 1996), which could be proven from their ion mass spectra. Particles with relatively high V⁺ ∕ VO⁺ signals occupied just 1.1 % of the total according to the classification criterion, and were only observed from a few ships. This will be discussed in following sections.

Figure 6Manual classification of 10 types of particles from 20 samples and their proportion by numbers.

Download

3.3.3 Size distribution

Particles with an aerodynamic diameter ranging from 0.2 to 2.5 µm, covering the accumulation mode (0.1 to 1 µm) and coarse mode (over 1 µm) (Frick and Hoppel, 2010), were analyzed by SPAMS. This size range has a significant correlation with PM_2.5 pollution, which should be seriously evaluated. Size information regarding 257 911 particles from 20 ships was derived from SPAMS and 19 342 of them were hit. This result showed a skewed distribution with a peak that appeared between 0.38 and 0.44 µm. This was at odds with the measurements conducted in Ireland using SMPS (Healy et al., 2009), in which particles measured were generally around 100 nm. Moreover, a bimodal distribution was not observed due to the limited measurement of particles smaller than 2.5 µm, while the coarse mode usually occurred above that. This divergence might result from the differences in measurement instruments and fuels. Ninety-five percent of all particles with size information lay between 0.2 and 1.46 µm, which implied that particles from ship auxiliary engine exhaust could be classified as fine particles. This would mean that they would contribute to the PM_2.5 in ambient air. It is remarkable that the magnitude of the size distribution in this study was consistent with some of the test results from engines using different fuels. Most of the samples in this study were collected after 1 January, when new policy forcing ships to switch to clean diesel fuel came into effect. This could have resulted in a decline in particle aerodynamic diameter.

In order to further analyze the size distribution information of different types of grouped particles, Fig. 7 was developed to reveal the variation in the proportion of different particles by size range. As OC particles only accounted for less than 5 % of overall particles, they show no apparent trends and were dispersedly distributed in each section. It can be observed that in sections larger than 1.30 µm  OC particles occasionally “disappeared” in some bins, while they remained continuous between 0.2 and 1.30 µm. This might indicate that OC particles tend to concentrate in a relatively tiny size range. The EC particle distribution was slightly similar to a bimodal distribution, with the first peak found in the 0 to 0.6 µm section. In the 1.75 to 2.5 µm section, it appeared to present another peak but it was not as obvious as the first one. Accordingly, in the 0.2 to 0.5 µm section, the proportion of ECOC particles was lower than in the over 0.5 µm section, implying that ECOC tends to form larger particles than EC. Another obvious fact was that K-rich particles were more likely to present a normal distribution in the 0.25 to 1.0 µm range, which was discrepant with the size distribution of Na-rich particles. Na-rich particles were distributed almost equally among bins in the 0.2 to 1.25 µm range, and the proportion increased starting from 1.25 µm. Metal containing particles were observed between 0.25 and 1.25 µm, occupying a much lower proportion than other massive sorts of particles. Metal contained in particles emitted by ship engines mainly comes from the combustion of heavy fuel oil and are known to catalyze the oxidation of SO₂ and the subsequent formation of sulfuric acid or sulfate particles.

Figure 7Specific size distribution of 10 classes of particles in the 0.2–2.5 µm range.

Download

3.4.1 Ion mass spectra differences between ships using different fuels

Owing to the fact that all samples were collected after 1 January 2017, the regulated date for the fuel switch, it could not be determined by date whether ships had changed to diesel fuel. However, it was observed from the ion mass spectra of individual ships that some ships had a high V intensity while others did not (Ni is also an element mostly found in heavy fuel, but SPAMS seemed to be less sensitive to Ni detection (Agrawal et al., 2009)). There was also a considerable difference in the ${\mathrm{HSO}}_{\mathrm{4}}^{-}$ intensity between ships. In order to better distinguish ships using HFO and MDO, an analysis was carried out in MATLAB. Signals that satisfied the intersection of conditions, where the intensity of peak m∕z = 51 and m∕z = 67 were higher than 100 while m∕z = 37 was lower than 100, could be recognized as vanadium-related signals (excluding the possibility of counting OC signals for V signals). Ship 13 and Ship 17 shared a similarity in both high V and ${\mathrm{HSO}}_{\mathrm{4}}^{-}$ relative intensity, as shown in Fig. 8. Owing to the fact that vanadium is a typical metallic element mostly found in heavy fuel oil with a higher sulfur content than distillate fuels (Moldanová et al., 2009; Celo et al., 2015), when obvious vanadium signals occurred in the ion mass spectra of particles from Ship 13 and Ship 17 they were very likely due to the use of heavy fuel oils at berth. Accordingly, the sulfur intensity also had the potential be very high. However, for ships with high sulfur but low vanadium intensity, the actual contents of sulfur and vanadium in the fuels used are unknown. This phenomenon cannot be well explained in this study due to instrumental limits and the quantity of particles analyzed in each sample. The main instrument used in this study was SPAMS, which is considered to be semi-quantitative and unable to give accurate emission factors of sulfur and vanadium. Moreover, similar studies using the same methods on shipping emissions are very rare. Therefore, this issue demands further exploration.

Figure 8Vanadium and sulfate relative intensity correlation obtained from the ion mass spectra of 20 ships.

Download

The two ships with vanadium signals higher than the others included a total particle number of 30 009, 2633 of which were measured with ion mass spectrometry. In the ion mass spectra of these ships, higher V⁺ ∕ VO⁺ and ${\mathrm{HSO}}_{\mathrm{4}}^{-}$ signals of over 0.8 in relative intensity were found, while the other ships had an average ${\mathrm{HSO}}_{\mathrm{4}}^{-}$ relative intensity of 0.59 and no apparent V⁺ ∕ VO⁺ signals. Due to the lower number of particles from these ships, there might be abnormally low positive EC signals in their ion mass spectra. Nonetheless, major chemical PM characteristics of different fuel types could be observed through a comparison of the ion mass spectra.

3.4.2 Comparison of vanadium intensity between shipping emissions and ambient data

Previous studies have illustrated that, owing to the vanadium contents in heavy fuel, vanadium in the ambient atmosphere could be used as a tracer indicating the heavy fuel combustion related to ship exhausts (Celo and Dabek-Zlotorzynska, 2010). Field measurements in Shanghai Port in 2011 revealed that the V ∕ Ni ratio could be applied to the identification of shipping traffic emission (Zhao et al., 2013). Research, focused on Bohai Rim in 2013, also identified and quantified the contribution of ship PM emission to ambient air quality using V as a tracer (Zhang et al., 2014). Moreover, studies in Europe have focused on metallic elements representing shipping-source particulate matter in the atmosphere, which played a crucial role in the ambient PM source apportionment analysis (Marco et al., 2011; Perez et al., 2016). As the new regulation came into effect and ships were required to use low sulfur diesel oil at berth, whether vanadium could still be used as a tracer of ship emission to accomplish shipping source identification was worthy of exploration. Distinct from previous research, which was mainly based on filter analysis, this study provided an innovative perspective on the abovementioned issue using spectrometric analysis.

Owing to the working principle of SPAMS, the mass fraction of the vanadium element to other symbolic elements could not be derived, and Ni and La showed no apparent signals in SPAMS spectra. The main methodology was to make a comparison between the vanadium intensity of particles from ships to that of particles sampled from the ambient atmosphere. If the vanadium intensity in ship exhaust is commonly lower than that of the corresponding ambient data, then it is no longer a proper tracer to be used to identify ship exhaust sources. The calculation was performed using a similar method to the one previously mentioned. Due to the fact that samples and particle numbers detected by SPAMS differ quite apparently among ships, we defined the ratio of total intensity to SPAMS detected particle number as vanadium-related intensity. The overall results for the 20 ships sampled from 4 to 16 January as well as the ambient vanadium intensity are shown in Fig. 9.

Figure 9Comparison of vanadium intensity of 20 individual ship particle samples to ambient particle samples from 1 to 16 January. The green line is the average vanadium intensity of ambient particles over time. The colored dots represent the average vanadium intensity of different ships distributed by sampling time.

Download

Noticing that the ordinate was logarithmic, considerable distinction existed among ships regarding vanadium intensity. The intensities in the exhausts from 17 ships were well below or comparable to the corresponding ambient intensity, while Ship 2, Ship 13 and Ship 17 were observed to have apparently high vanadium intensity. This phenomenon might revealed the fact that after fuel substitution, vanadium from ship exhausts at berth would not appear to be remarkably higher than that of the ambient atmosphere in most cases. Another crucial piece of evidence is the elemental analysis of three diesel samples collected during field measurements (shown in Table 4). Vanadium was not detected in any of the three diesel samples, which was significantly different from the data from previous studies (38.0–133.8 mg kg⁻¹; Celo et al., 2015) using intermediate fuel oils (IFO) (Table 4). Regarding vanadium “sources”, it can be inferred that if the content of vanadium in fuel keeps declining, the role it plays as a tracer will not be as significant as it has been; furthermore, it will be much harder to detect the existence of vanadium in ambient atmosphere.

Table 4Elemental analysis of three fuel samples and comparison with previous studies.

IFO: intermediate fuel oil; MDO: marine diesel oil. Two of the fuel samples were from Ship 9 and Ship 19. The sample marked “unplanned ship” indicates that this ship was not among the 20 ships included in this study.

Download Print Version | Download XLSX

3.5.1 The impact of wind direction

To better focus on the shipping emissions, we took wind direction data into consideration. The wind directions for the whole sampling period is shown in Fig. 10b. The geographic positions of berths and the wind directions meant that the berths, which were mainly distributed northwest, north and east of the sampling site, had plumes driven to the sampling site by winds from these directions. Moreover, no obvious emission sources other than ships at berth could have interfered with the ambient sampling.

Ambient data with wind directions ranging from northwest to southeast (clockwise) were extracted and separated by the 1 January 2017 cutoff. A total of 10 h with 37 825 particles and a total of 133 h with 682 176 particles were calculated before and after 1 January, respectively. The updated results for the ratio of sulfates identified as shipping emissions to ambient were 35 and 27 % before and after 1 January 2017, respectively, indicating a decrease of the at-berth shipping emission contribution to ambient sulfates.

3.5.2 The impact of atmospheric layer stability

Apart from wind direction, atmospheric layer stability is also a factor that can affect particle diffusion in exhaust plumes. However, owing to the experimental conditions and the obscure sampling location during the measurement period, the actual planetary boundary layer (PBL) height and cloud cover data were not available in this study; these factors can directly reflect the atmospheric stability. Nevertheless, atmospheric layer stability is able to be identified based on indirect parameters. According to a study on the correlation between heavy haze episodes and synoptic meteorological conditions in Beijing, China (Zheng et al., 2015b), polluted periods have been associated with a lower PBL height. This means that the atmospheric layer was found to be more stable during polluted periods than during clean periods. Moreover, according to a study by Petäjä et al. (2016), high particulate concentration decreases the height of the boundary layer via feedback mechanisms. Therefore, it can be inferred that the variation of PM_2.5 and wind speed can indirectly reflect changes in atmospheric stability. Additionally, the temporal variation of CO concentrations can also reflect the local diffusion conditions as well as layer stability (Xu et al., 2011), since it is an unreactive air pollutant with a long atmospheric lifetime (two months) (Novelli et al., 1998) and mainly stems from primary emissions.

Based on the discussions above, temporal profiles of PM_2.5 and CO concentrations as well as the variation of sulfate contribution from ships are presented in Fig. 11. The CO and PM_2.5 data is derived from both an online monitoring instrument at the sampling site and a national monitoring site 20 km away. The trends of data from the two different sites show excellent correlation with Pearson's correlation coefficients of 0.91 and 0.88 for PM_2.5 and CO (Fig. 11a and b), respectively; this indicates that our results are representative for the ambient atmospheric environment. It can also be observed that the hourly variation of PM_2.5 concentrations at the sampling site was in good agreement with that of CO (r=0.76), which illustrates that the local air pollution levels were mainly determined by the meteorology conditions, and that the trends of the pollution levels near the sampling site can be applied to indirectly defining the stability of the ambient atmosphere. According to the PM_2.5 concentration threshold suggested by Zheng et al. (2015b), hourly air pollution levels during the whole measurement period were classified into four categories: clean (hourly PM_2.5<35 µg m⁻³); slightly polluted (35–115 µg m⁻³); polluted (115–350 µg m⁻³); and heavily polluted (>350 µg m⁻³). It was found that the hourly averaged PM_2.5 concentrations during our campaign reached 147 µg m⁻³, and over 85 % of the total hours were under polluted conditions. Therefore, the atmospheric layer condition remained stable during most of the measurement period.

Figure 11Temporal profiles of PM_2.5 concentration, CO concentration and sulfate contribution by shipping emissions. (a) PM_2.5 concentration from the two monitoring sites. (b) CO concentration from the two monitoring sites. (c) Sulfate contribution from shipping emissions.

Download

While the contribution of sulfate from ships showed an obvious converse trend to pollution levels ($r=-\mathrm{0.64}$) (Fig. 11c). The average contributions were calculated as 32.9, 27.6, 19.7 and 15.1 % during the clean, slightly polluted, polluted and heavily polluted periods, respectively. That is, when the atmospheric stability went up, the contribution of ships went go down.

The explanation for this phenomenon can be summarized according to two aspects. Firstly, during the winter haze periods in China, stable atmospheric conditions are not solely to blame for increasing the total concentration of PM_2.5; secondary inorganic aerosols are also responsible, especially sulfates, which are rapidly formed via complicated heterogeneous reactions under high relative humidity (RH) (Wang et al., 2016; Cheng et al., 2016; Zheng et al., 2015a). In our study, the RH in port areas was higher than that inland, resulting in the predominance of sulfate formation during polluted periods. The effects of increased ambient sulfate concentrations on reducing the sulfate relative contributions from ships are much more significant than the increased shipping contributions due to stable atmospheric conditions. Moreover, in addition to the local diffusion conditions, the contribution of shipping emissions to the ambient air quality is also highly dependent on the real-time density of ships at berth. The number of ships at berth decreased sharply when heavy haze led to fairly poor visibility. Therefore, less shipping emissions lowered source contributions during the air pollution period. In summary, the effects of atmospheric layer stability on the sulfate contributions from ships was less important during the haze period compared with during clean days. The new Chinese sulfur limit for auxiliary engines corresponds to the global shipping limit that will apply from the year 2020. Studies have been carried out to estimate the effect of the low sulfur fuel limit. On a global level, according to Sofiev et al.'s (2018) estimation, the worldwide implementation of the 0.5 % sulfur content policy could reduce the annual average sulfate concentration by 2–4 µg m⁻³. However, even after the implementation of the 0.5 % sulfur limit for ships at berth, the PM and SO₂ emissions still remain at a level of 770 and 2500 kt, respectively. In Chinese ports, under the scenario of all ships changing over to low sulfur fuel (< 0.5 %) in all of China's emission control areas, the remaining at-berth PM and SO₂ emissions in Jing-Jin-Ji port area could reach up to 1 and 8 kt, respectively (Liu et al., 2018).

According to the source apportionment of sulfate particles in the port area, the number concentration contribution of sulfates from shipping emission at berth was lowered from 35 to 27 % after the switching fuel policy implementation on 1 January 2017. The stricter fuel sulfur limit did reduce the contribution of shipping emissions, but these emissions continue to play an important role in atmospheric pollution; electricity from land will is required to ameliorate this situation.

In general, PM and SO₂ emissions can't be eliminated by merely controlling the sulfur content of fuels, though the stricter sulfur limits is an effective way to reduce emissions. Hence the option of using electricity from land will probably help maximizing the emission reduction of SO₂ and PM.