Data-Driven Analysis of Regional Ship Carbon Emission Reduction: The Bohai Bay Area Case Study

Abstract

With the tightening of marine carbon emission reduction policies, the sustainable development of the shipping industry has attracted much attention, and it is of great significance to use Automatic Identification System (AIS) big data to study the carbon emissions of marine ships. Taking ships around Bohai Bay as the research object, this paper constructs a calculation method of ship carbon emissions driven by the ship AIS trajectory. The AIS information of ships is extracted, and the sailing status is determined. The carbon emission calculation model is built based on the AIS data, the carbon emission in 2023 is empirically measured, and the characteristics are analyzed. At the same time, a speed simulation model was built to evaluate the impact of speed reduction on carbon emissions and put forward emission reduction measures. The results show that the carbon emission of ships around Bohai Bay in 2023 was 8.8072 million tons, with cargo ships contributing the most, and the carbon emissions of the cruise state was significant. A 10% reduction in speed would reduce annual carbon emissions by about 6%. This study provides a reference for understanding the impact of speed on carbon emissions and formulating emission reduction measures, which can be used to compare historical and future data to support the emission reduction in ports and shipping enterprises.

1. Introduction

The issue of global climate change is becoming increasingly severe, and the pressure from the international community, especially from the Paris Agreement, has urged governments worldwide to take more proactive actions in emission reductions. With ships as the primary mode of maritime transport, their carbon emissions have gradually attracted the attention of the international community due to their growing share in global greenhouse gas emissions [1]. According to the International Maritime Organization (IMO), the carbon dioxide emissions from the global shipping industry account for about 3% of the total global emissions [2]. In 2018, the IMO adopted global shipping emission reduction targets, aiming to reduce carbon dioxide emissions from ships by 50% by 2050 [3]. This decision has also led China, an important part of the global shipping industry, to place greater emphasis on the management and control of ship carbon emissions. In recent years, China has successively issued documents such as the “Implementation Plan for the Special Action on Ship and Port Pollution Prevention (2015–2020)” and the “Implementation Plan for Ship Emission Control Areas in the Pearl River Delta, Yangtze River Delta, and Bohai Rim (Beijing–Tianjin–Hebei) Waters”, all of which call for stronger control over ship carbon dioxide emissions [4].

The Bohai Rim Economic Region, as one of China’s five major port clusters and one of the country’s first three ship emission control areas, is a key maritime trade and logistics hub. This region includes several major ports in cities and provinces such as Beijing, Tianjin, Hebei, and Shandong, including Tianjin Port, Qingdao Port, Tangshan Port, Dalian Port, among others. The region experiences high shipping volumes and frequent vessel traffic, which places significant pressure on local air quality and marine ecological environments due to ship-related air pollution emissions. Compared to global trends in carbon reduction, ship carbon emissions in the Bohai Bay region remain relatively severe, which is one of the reasons that the area has become a key focus for air pollution control in China. Therefore, establishing a localized ship emission inventory and assessing the feasibility of carbon emission control measures for ships in ports within the Bohai Bay Economic Region is of great significance for effectively controlling ship carbon emissions in the area.

The current research methods can be mainly divided into two categories: one is the traditional “top-down” research method represented by the “fuel consumption method”. The fuel consumption method estimates ship carbon emissions by measuring fuel consumption and using known fuel carbon emission factors. This method was first applied in the 1990s, when foreign scholars such as Corbett [5] used international shipping fuel data, combined with emission factors for different types of engines and fuels, to estimate the global ship emission inventory. Endresen [6] and others used the fuel consumption method to estimate the ship emission inventory for the period 1925–2002. The fuel consumption method boasts simplicity and operational ease, given that it merely requires fuel consumption data that are usually readily accessible. This made it a popular method among early scholars [7,8,9,10,11] for estimating maritime emissions on regional or global scales. However, the fuel consumption method does not account for various factors such as ship speed, sea conditions, and sailing status during navigation, which may result in significant discrepancies from actual emissions. As a result, emission inventories estimated using the fuel consumption method are often underestimated [12].

The alternative category is the “bottom-up” research methodology represented by the “power-based method”. The power-based method is an estimation approach based on ship activity, which calculates carbon emissions by analyzing a ship’s power demands and its actual operational conditions, combined with real navigation data [13,14,15,16,17]. Compared to the fuel consumption method, the power-based method accounts for the ship’s actual operating conditions and energy consumption, and, by incorporating load factors and emission factors for different engine types under various conditions, it provides a more accurate estimate of carbon emissions. The accuracy of the power-based method depends solely on the caliber of the ship’s dynamic data. In recent years, using ship trajectory data obtained from the Automatic Identification System (AIS) for scientific research has become a growing trend both domestically and internationally [18,19,20,21,22,23]. The AIS provides high-resolution dynamic data for ships, including speed, course, latitude, longitude, and operational status; it also offers high-resolution static data, such as ship classification, dimensions, and deadweight tonnage. In 2007, foreign scholar Corbett [24] initially proposed application AIS ship trajectory data to estimate ship carbon emissions, using voyage speed to determine carbon emissions under a distinct operational scene. In 2009, Jalkanen et al. [25] used AIS data to establish the first ship emission inventory for the Baltic Sea.

In summary, substantial research efforts have been dedicated to ship carbon emissions in specific regions or individual ports. However, comprehensive studies on regional ship carbon emissions and the optimization of ship speed under different carbon reduction policies are relatively scarce. Therefore, in order to more accurately and effectively measure ship carbon emissions in the Bohai Bay region and assess the impact of different sailing speeds on carbon emissions, this study proposes a methodology based on ship AIS trajectory data to calculate ship carbon emissions, combined with simulation scenarios to evaluate the influence of different speeds on emissions. First, the AIS trajectory data are cleaned for anomalies and interpolated to fill in missing points, constructing a complete time series of the ship’s trajectory. This allows for determining the ship’s operational mode based on the trajectory. Next, the ship’s power, load factor, and emission factors under different operating conditions are determined, and a ship carbon emission estimation model is formulated. Then, empirical analysis is carried out by demarcating the boundaries of the empirical study using port geographic information and other data. The carbon emission estimation model is applied for empirical calculations, and the results are analyzed from multiple dimensions, including spatial and temporal characteristics. Finally, through the construction of time-series data and slice clustering in the simulation model, a sailing speed simulation model is developed to analyze the degree to which different speeds impact carbon emissions. The three main contributions of this study are as follows:

(1) A technical methodology for constructing complete AIS time-series trajectories is constructed, which includes data decoding and processing, spatial region selection, anomaly data cleaning, and trajectory interpolation to fill in missing points.

(2) A novel ship carbon emission estimation is proposed by considering ship power, load factors, and emission factors under different sailing conditions in AIS data. The results are analyzed in multiple dimensions.

(3) A sailing speed simulation model for shipping carbon reduction is developed, which conducts empirical analysis and proffers policy recommendations for the reduction in shipping carbon emissions under diverse scenarios.

2. Research Methodology

2.1. Research Ideas

The research approach of this study is illustrated in Figure 1. First, the AIS trajectory data underwent preprocessing, including anomaly data cleaning and trajectory interpolation, with the aim of constructing a comprehensive and utilizable AIS time-series trajectory. The ship’s operational status was then determined based on the trajectory. Next, the ship’s power, load factor, and emission factors under different operational statuses were identified, and a ship carbon emission calculation model was developed. Following this, the boundaries of the empirical study were defined using port geographic information and other data. The carbon emission estimation model was applied to perform empirical calculations, and the results were analyzed from multiple dimensions, including spatial and temporal characteristics. Finally, through the construction of time-series data and slice clustering in the simulation, a sailing speed simulation model was developed to evaluate the influence of different sailing speeds on carbon emissions. This study concludes by offering carbon reduction policy recommendations for different scenarios.

2.2. Experimental Data

2.2.1. AIS Data Description

The AIS data of ships include both static and dynamic data. The static data include information such as the MMSI number, ship type, vessel dimensions, and draft depth. These static data mainly come from sources like the China Classification Society (CCS) ship registration files, Lloyd’s ship registry, and other maritime regulatory bodies’ registration archives. Dynamic data refer to real-time ship position data tracked by the Automatic Identification System (AIS). These data primarily include the following: the MMSI number of the ship, the current ship position, navigation speed, the longitude and latitude corresponding to the ship’s heading, and other spatial dimension data. The attributes of AIS data are shown in

Table 1. AIS data attribute description.

No.	Type	Field Name	Update Frequency
1	Dynamic Information (Messages 1, 2, 3, 18, 19)	MMSI	Updated every 2 to 3 min based on speed and heading
2		Ship Status
3		Latitude and Longitude
4		Speed
5		Heading
6	Static Information (Messages 5, 19)	IMO Number	Updated every 6 min
7		Call Sign
8		Ship Name
9		Ship and Cargo Type
10		Ship Size
11		Deadweight (DWT)
12		Estimated Time of Arrival	Updated every 6 min
13		Destination Port

. By sequentially stacking each data point, a continuous trajectory of the ship’s movement can be formed. These trajectory data can be used to determine the ship’s operational status, and they serve as the basis for various calculations, analyses, and studies. The AIS data used in this study were collected from the National AIS Data Center.

2.2.2. AIS Data Preprocessing

Due to factors such as equipment limitations, channel slot restrictions, and the improper use of shipboard devices, AIS data may experience anomalies such as drift, point loss, and errors in its logical structure [26]. Therefore, before conducting research, AIS data must undergo preprocessing to improve its quality. The AIS preprocessing flow is shown in Figure 2.

(1) Error Data Cleaning: Using navigational knowledge, records with erroneous data were identified. For example, if the ship’s MMSI number contained more or fewer than nine digits, or if the ship’s speed is negative, such erroneous data records were deleted.

(2) Missing Data Cleaning: Records with missing static information about the ship were deleted directly. For records with missing dynamic information, provided that the proportion of missing data within the entire trajectory data is less than 20%, interpolation and other mathematical methods were applied during the subsequent processing to fill in and correct the missing points.

(3) Anomaly Data Cleaning: If a ship’s trajectory point is located on land, or if a specific trajectory point deviates significantly from the general direction of the entire trajectory, it indicates that the ship’s position data have undergone an aberration during transmission. Such clearly erroneous trajectory points need to be deleted. In a trajectory (P₁, P₂, …, P_a, …, P_n), when the trajectory point P_a satisfies the following formula, the Trajectory point will be retained.

$$\left\{\begin{array}{c} lo{n}_i\in \left(\overline{lon}-2{\sigma }_{lon}, \overline{lon}+2{\sigma }_{lon} \right)\\ la{t}_i\in \left(\overline{lat}-2{\sigma }_{lat}, \overline{lat}+2{\sigma }_{lat} \right)\end{array}\right.$$

(1)

where $\overline{lon}$, $\overline{lat}$ represent the average longitude and latitude of all trajectory points, and $\sigma \mathit{lon}, \sigma lat$ represent the variance of longitude and latitude, respectively. Comparison of AIS abnormal data processing is shown in Figure 3.

2.2.3. AIS Data Interpolation and Point Filling

The transmission of AIS dynamic information messages is periodically based on factors such as the vessel’s speed, heading, and status. This means that the time interval between two adjacent trajectory points is not fixed and exhibits some fluctuation. Based on this characteristic, the interpolation for AIS trajectory data requires calculations based on either two adjacent points or multiple points for interpolation purposes [27]. The update time of AIS dynamic information is related to the vessel’s speed, and, typically, the update time does not exceed 3 min [28]. During this period, the vessel’s trajectory can be approximated as a linear path. However, in practical scenarios, the update time often exceeds 3 min, or even longer, during which the vessel’s speed and direction may change significantly. In such cases, linear interpolation is no longer suitable. Based on existing research by scholars, this study adopts an algorithm for vessel dynamics based on the rate of change, employing a weighted calculation between two trajectory points to obtain the interpolated points.

This study uses the AIS big data similar route trajectory fitting algorithm to further verify the accuracy of the interpolation trajectory points generated by the AIS data interpolation algorithm. Typical navigation routes containing interpolation points were selected, and trajectory points from all similar routes were fitted. The continuous trajectory formed was then compared with the trajectory after the interpolation points were added, ultimately confirming that this interpolation method can generally reflect the missing trajectory points accurately.

Select two adjacent trajectory points P₁(lon_₁, lat_₁), P₂(lon_₂, lat_₂), of the same vessel, and use coordinate trigonometric functions to calculate the reference value of the interpolated point P_M.

P_M (lon__m) = lon_₁ + V₁Sin(R₁) (T_M − T₁)
P_M (lat__m) = lat_₁ + V₁Cos(R₁) (T_M − T₁)

(2)

where lon_₁, lat_₁ are the longitude and latitude of the first trajectory point P₁, respectively; V₁, R₁, T₁ are the speed, heading, and time of trajectory point P₁, respectively; and T_M is the message time of the trajectory point P_M.

In the case of sudden changes in speed and heading, deviations still exist. Therefore, the reference value of the second interpolated point P_N is calculated based on the reference value P_M value and the second trajectory point P₂.

P_N (lon__n) = lon_₂ + V₂Sin(R₂) (T_M − T₂)
P_N (lat__n) = lat_₂ + V₂Cos(R₂) (T_M − T₂)

(3)

where lon_₂, lat_₂ are the longitude and latitude of the second trajectory point P₂, respectively; V₂, R₂, T₂ are the speed, heading, and time of trajectory point P₂, respectively; and T_N is the message time of the trajectory point P_N.

Using the time difference between the three points, a weight analysis was performed on the two interpolated points P_M and P_N, and further calculation was performed to obtain the final interpolated point P_T.

G₁ = 1 − (T_M − T₁)/(T₂ − T_M);
G₂ = 1 − (T₂ − T_M)/(T₂ − T₁)

(4)

P_T (lon__t) = G₁ P_M (lon__m) + G₂ P_N (lon__n)
P_T (lat__t) = G₁ P_M (lat__m) + G₂ P_N (lat__n)

(5)

where G₁, G₂ represent the weight factors of the two points, which can be assigned based on the time intervals between the interpolated point and the two end points. The smaller the time difference, the larger the weight factor. Comparison of trajectories before and after interpolation is shown in Figure 4.

2.3. Estimation Method for Ship Carbon Emissions

2.3.1. Estimation Process

This study estimates ship carbon emissions, which is divided into 8 main steps:

(1) Select AIS trajectory points of ships within the study area. (2) Sort the trajectory points by time series, and calculate the spherical distance, time interval, and instantaneous speed for each ship’s trajectory segment. (3) Match the AIS data with ship profiles to obtain information such as engine type, ship dimensions, tonnage, and maximum design speed. (4) Calculate the ship’s load factor and determine its operational status. (5) Determine the auxiliary engine and boiler power based on the ship’s size and operational status. (6) Calculate the fuel consumption for each ship’s trajectory segment. (7) Calculate the carbon emissions for each ship’s trajectory segment based on the carbon emission factor. (8) Summarize the carbon emissions for all ship trajectory segments. The carbon emission calculation process is shown in Figure 5.

2.3.2. Ship Carbon Emission Calculation Model

The carbon emissions generated from fuel consumption on a ship primarily come from three sources: the main engine, auxiliary engines, and boilers. Therefore, the carbon emissions of an individual ship constitute the aggregate of the emissions from these three components. Based on this, a ship carbon emission calculation model was established:

E = E_m + E_a + E_b

(6)

where E is the total CO₂ emissions of the ship (g); E_m is the CO₂ emissions from the ship’s main engine (g); E_a is the CO₂ emissions from the ship’s auxiliary engines (g); and E_b is the CO₂ emissions from the ship boilers (g).

E_m = P_m × LF_m × T × EF_m
LF_m = (AS/MS)³

(7)

where P_m is the power of the ship’s main engine (kw); LF_m is the load factor of the ship’s main engine; T is the operating time of the ship on the current trajectory (h); EF_m is the emission factor of the ship’s main engine (g·kw⁻¹·h⁻¹); AS is the average speed of the ship on the current trajectory (knots); and MS is the maximum speed of the ship (knots).

E_a = P_a × LF_a × T × EF_a

(8)

where P_a is the power of the ship’s auxiliary engine (kw); LF_a is the load factor of the ship’s auxiliary engine; T is the operating time of the ship on the current trajectory segment (h); and EF_a is the emissions factor of the auxiliary engine (g·kw⁻¹·h⁻¹).

E_b = P_b × LF_b × T × EF_b

(9)

where P_b is the power of the ship’s boiler (kw); LF_b is the load factor of the ship’s boiler; and EF_b is the emission factor of the boiler (g·kw⁻¹·h⁻¹).

2.3.3. Parameter Description

(1) Rated Power: The ship’s engines are primarily divided into the main engine, auxiliary engine, and boiler, and the rated power of these three components is one of the pivotal parameters for model calculations. The rated power of the ship’s main engine can be obtained from the AIS information platform. For missing main engine power data, this study refers to the relevant research literature [29], uses the ship’s dimensional information to estimate its tonnage, and then applies a power function to fit the relationship between ship tonnage and the corresponding main engine power. The fitting equations for the three components are shown in

Table 2. Fitting equations of ship size to gross tonnage to ME power.

	Ship Type	Ship Type
Fitting Function		Cargo Ship	Passenger Ship	Oil Tanker	Work Vessel	Other Ships
Fitting Function (Size vs. Gross Tonnage) (Gt)		0.08s1.44	0.10s1.40	0.12s1.38	0.12s1.38	0.04s1.52
Size vs. Gross Tonnage (R2)		0.91	0.93	0.9	0.91	0.98
Fitting Function (Gross Tonnage vs. Main Engine Power) (ME)		2.33Gt0.76	4.24Gt0.72	2.14Gt0.80	3.96Gt0.73	2.78Gt0.75
Gross Tonnage vs. Main Engine Power (R2)		0.91	0.89	0.92	0.85	0.9

.

The rated power of the auxiliary engine and boiler is typically correlated with the ship’s dimensional, type, and operational condition. In this study, the rated power of the auxiliary engine as estimated based on the ratio of the main engine power to the auxiliary engine power, as found in the IMO Third GHG (Greenhouse Gas) Study [30] and the research of most scholars [31]. The rated power of the boiler was determined based on the relationship between the ship’s maximum design speed and the rated power of the boiler. The specific ratio relationships are shown in

Table 3. Auxiliary and boiler power and maximum design speed reference.

	At Anchor		Anchored		Underway (or Under Engine Power)		Cruising
Ship Size (TEU)	Auxiliary Engine (kw)	Boiler (kw)	Auxiliary Engine (kw)	Boiler (kw)	Auxiliary Engine (kw)	Boiler (kw)	Auxiliary Engine (kw)	Boiler (kw)
0–999	250	370	250	450	240	790	0	410
1000–1999	340	820	340	910	310	1750	0	900
2000–2999	460	910	460	910	430	1900	0	920
3000–4999	480	1100	480	1350	430	2500	0	1400
5000–7999	590	1100	590	1400	550	2800	0	1450
8000–11,999	620	1150	620	1600	540	2900	0	1800
12,000–14,499	630	1300	630	1800	630	3250	0	2050
14,500–19,999	630	1400	630	1950	630	3600	0	2300
20,000+	700	1400	700	1950	700	3600	0	2300

.

(2) Ship Operating Conditions and Load Factor: The load factor of the ship’s main engine is related to the ship’s sailing speed, typically determined by the proportion of the ship’s actual sailing speed to its maximum design speed. The load factors of the auxiliary engine and boiler, however, are independent to sailing speed. This study refers to relevant research results [6] to determine the load factors of the auxiliary engine under different operational conditions. The load factor of the boiler varies under different conditions: when the ship is cruising, the boiler is in an inactive state with a load factor of 0, and, in normal operation, the load factor is 1. The determination of the ship’s operating conditions and auxiliary engine load factors is shown in

Table 4. Criteria for the classification of a ship’s navigational status.

Navigation Status	At Anchor	Anchored	Underway (or Under Engine Power)	Low-Speed Cruising	Cruising
Classification Criteria (or Dividing Standards)	<1 Knot	Speed ∈ [1,2,3] Knots	>3 Knots and Load Factor LF <20%	Load Factor LF ∈ [20%, 65%]	Load Factor > 65%
Auxiliary Engine Load Factor	0.19	0.48	0.25	0.13

.

(3) Emission Factor: The determinants of the emission factor include the ship type, fuel type, and engine speed. The main types of ship fuel are heavy oil (RO), marine diesel (MD), and marine gas oil (MG), while ship engine speeds are classified into high, medium, and low-speed categories. In this study, the emission factors were selected based on a review of the relevant literature [32], and the chosen emission factors are shown in

Table 5. Emission factors of different engine types.

	Main Engine			Auxiliary Engine			Boiler
Engine Speed	Low Speed	Medium Speed	High Speed	Low Speed	Medium Speed	High Speed
Fuel Types	RO	RO	RO	—	MD	—	MD/MG
CO2 Emission Factor (kg/kw·h)	0.622	0.686	0.686	—	0.6907	—	0.922

.

2.3.4. Uncertainty Analysis

This study’s model calculation results are based on various foundational data, and errors in these foundational data introduce a certain level of uncertainty into the results.

(1) Uncertainty Stemming from Errors in Ship’s Fundamental Data: The model calculations in this study involved both dynamic and static ship data. The uncertainty in dynamic data was influenced by the range of the ship trajectory data collection and the quality of the trajectory data. During the collection of ship AIS trajectory data for this study, efforts were made to collect comprehensive data in order to minimize the impact of sample data being smaller than the total dataset within the research area. Additionally, big data processing methodologies were used to preprocess the trajectory data, improving the quality of the raw data and minimizing its impact on the results as much as possible. However, the inherent quality deficiencies of ship AIS trajectory data cannot be completely avoided, and they still affect the study results.

To ensure that the uncertainty caused by the ship’s dynamic and static data had minimal impact on the research results, this study comprehensively cleaned duplicate data, incomplete data, incorrect data, and noisy data. Logical interpolation was performed on the missing position data. The total amount of collected data is 4056.3968257 billion pieces. Among them, duplicate data for cleaning accounts for 1%, incomplete and incorrect data accounts for 2.36%, and noisy data accounts for 0.92%. Meanwhile, the interpolation and supplementation of 934,687 trajectory points were completed, forming relatively high-quality research data and reducing the impact of basic ship data on research uncertainty.

(2) The uncertainty resulting from errors in model calculation parameters as addressed by constructing a set of emission factor datasets as the basis for the quantitative uncertainty analysis of emission factors during the research process. This study mainly referred to the theories recommended by relevant guidelines and constructed the emission factor datasets through measured data and related research to obtain the probability distribution model describing the uncertainty of emission factors as well as the uncertainty range and magnitude, as specifically shown in

Table 6. Ship carbon dioxide emission factors and their uncertainties.

Type of Working Engine	Mean Value/(g·kW−1·h−1)	Confidence Interval/(g·kW−1·h−1)	Uncertainty Range
Main Engine	653.32	(633.22, 673.34)	(−3.1%, 3.1%)
Auxiliary Engine	702.01	(687.48, 715.94)	(−2.1%, 2.0%)
Boiler	954.15	(922.00, 970.71)	(−3.4%, 1.7%)

.

In general, the emission of CO₂ from ships mainly originates from the fuel combustion process. Under the conditions of sufficient oxygen and complete combustion, the carbon elements in the fuel are almost completely converted into CO₂. During this process, the amount of CO₂ emission is relatively stable. Therefore, the uncertainty range of the CO₂ emission factor is relatively small. In comparison, it can be seen from the width of the confidence interval that the carbon dioxide emission factor of the auxiliary engine is relatively stable. The boiler has the greatest uncertainty, which is (−3.4%, 1.7%), because the boiler has more variable factors, resulting in more difficult data collection, and, thus, its emission factor may have greater uncertainty. The dataset of the main and auxiliary engine emission factors is larger and has a wider coverage, which means that more data support the estimation of their emission factors, thereby reducing the uncertainty.

3. Empirical Research Analysis

3.1. Research Area and Subjects

The empirical research area in this study is the Bohai Bay region, where the research model and methodologies proposed in the previous chapter are further validated. By reviewing relevant documents approved by the State Council regarding the Bohai Economic Rim, this study defines the research boundary of the Bohai region as a rectangular area delineated by the coordinates 41°1′23.27″ N, 117°32′24.00″ E and 35°0′00.00″ N, 123°1′57.90″ E. The focus is on the Bohai region, with the Beijing–Tianjin–Hebei area serving as its core and the Liaodong and Shandong Peninsulas functioning as the two wings. The location map of the Bohai Bay area for this study is shown in Figure 6.

3.2. Analysis of Total Carbon Emissions from Ships

This study uses the carbon emission calculation model proposed in the previous chapter, based on ship AIS trajectory data, to calculate the total CO₂ emissions from ships in the Bohai Bay region in 2023, which amounts to 8.8072 million tons. A detailed analysis is presented from different dimensions, including ship type, emission source, and navigation status. An example of the ship CO₂ emission calculation is shown in

Table 7. The example of ship carbon emission measurement.

Ship Information
MMSI	414,896,000	Ship Name	NinyuanTJ	Length	197.0	Width	197.0
Main engine power (kW)	18,760	Auxiliary engine power (kW)	750	Boiler power (kW)	1110	Maximum speed (knots)	19
Deadweight (DWT)	56,295.0	TEU	3316
Carbon Emission Calculation
Timestamp (S)	Start coordinate	End coordinate	Status	Speed (knots)	Distance (nM)	Duration(S)	Ship CO2 emissions (g)
1,694,677,037	38°58.223″ N117°54.733″ E	38°58.223″ N117°54.734″ E	Anchorage	0	0.01	2000	6350
1,694,677,037	38°58.223″ N117°54.734″ E	39°1.1436″ N117°46.2745″ E	Low-speed navigation	11	10	3200	11,588
1,694,677,037	39°1.1436″ N117°46.2745″ E	39°1.1437″ N117°46.2745″ E	Mooring	0	0	15,300	3980
……	……	……	……	……	……	……

.

3.2.1. Classification-Based Analysis of Carbon Emissions by Ship Type

The carbon emissions and emission share of each ship type in the Bohai Bay region in 2023 are shown in

Table 8. Carbon emissions of various ship types in the coastal area around the Bohai Sea in 2023.

Carbon Emissions by Ship Type in the Bohai Economic Rim in 2023 (Million Tons)
Total Emissions	Cargo Ship	Tanker Ship	Work Boat	Passenger Ship	Other Ships
8.8072	5.8866	1.0213	0.9135	0.7590	0.2268

. It is evident that the main contributors to carbon emissions are general cargo ships, with a CO₂ emission of 5.8866 million tons, accounting for 66.84% of the total emissions. This proportion exceeds half of the overall emissions. The second largest contributor is oil tankers, with a CO₂ emission of 1.0213 million tons, accounting for 11.6%. The carbon emissions from other vessels, such as workboats and passenger ships, are relatively low. The carbon emissions from cargo ships and oil tankers far exceed those of other ship types, mainly because these two types of ships typically use heavy fuel oils, such as marine fuel oil. These fuels are rich in carbon and sulfur compounds. Compared to lighter fuels like diesel or natural gas, heavy oil has a lower combustion efficiency, resulting in higher CO₂ emissions. Additionally, these two types of ships are usually large in size and travel at higher speeds. To power these large vessels, a strong propulsion system is required, typically involving high-powered engines, which leads to higher fuel consumption and carbon emissions. The carbon emissions classified by ship type are shown in Table 8.

3.2.2. Analysis of Carbon Emissions by Ship Emission Source

This study analyzes and compares the carbon emissions from the main engine, auxiliary engine, and boiler to further understand the distribution of carbon emissions from ships. Compared to the auxiliary engine and boiler, the main engine generates more carbon emissions, with its share of carbon emissions reaching a maximum of 66.07%. The auxiliary engine accounts for 10.68%, and the boiler accounts for 23.25%, slightly higher than the auxiliary engine emissions. The main engine is the primary power source for ships, typically using heavy fuel oil. It typically exhibits a large power output and runs continuously during long voyages, operating under high loads for extended periods, which inevitably leads to higher carbon emissions. The auxiliary engine is used to provide auxiliary power to the ship, such as electricity, compressed air, cooling, and pumping. It typically uses diesel as fuel, and, although it does not run under high loads for long periods like the main engine, its start-up and shutdown contribute to fuel consumption and carbon emissions, though to a lesser extent. The boiler provides steam for the ship’s air conditioning and heating systems and typically uses heavy oil, light oil, or liquefied natural gas as fuel. Its carbon emissions are closely related to the type of fuel used and the ship’s operational status, with emissions typically falling between those of the main engine and auxiliary engine. The carbon emissions by ship emission source are shown in

Table 9. Carbon emissions of source classification of ships in the coastal area around the Bohai Sea in 2023.

Carbon Emissions by Emission Sources in the Bohai Economic Rim in 2023 (Million Tons)
Total Emissions	Main Engine	Auxiliary Engine	Boiler
8.8072	5.8724	0.9488	2.0660

.

3.2.3. Analysis of Carbon Emissions by Ship Operational Mode

This study compares and analyzes the carbon emissions from various ship operational conditions, including cruising, maneuvering, slow cruising, anchoring, and mooring. The carbon emissions during cruising account for 44.89% of the total, followed by maneuvering and slow cruising, with emission proportions of 26.89% and 16.10%, respectively. During anchoring and mooring, emissions account for 7.01% and 5.11% of the total carbon emissions, respectively. In the cruising mode, ships typically travel at higher speeds, often their maximum speed during a voyage. In this state, the engine needs to provide maximum power to maintain the high speed, resulting in higher fuel consumption and the most significant carbon emissions. When ships enter or leave ports and waterways, they typically enter the maneuvering state, such as when turning, berthing, or unberthing operations. Although these states last for a relatively short period, they require the ship to quickly change speed or direction and demand higher power output, leading to intense carbon emissions in a short time. During slow cruising, the lower speed reduces the engine power, resulting in lower fuel consumption and less carbon emissions. In anchoring or mooring states, the ship is usually stationary, and the engine provides minimal power to maintain essential systems such as water pumps and ventilation, or the engine may stop entirely, leading to the lowest carbon emissions. The carbon emissions and their share in different operational modes are shown in

Table 10. Carbon emissions of the navigational state of ships in the coastal area around the Bohai Sea in 2023.

Carbon Emissions by Ship Sailing Status in the Bohai Economic Rim in 2023 (Million Tons)
Total Emissions	Cruising	Maneuvering	Slow-Speed Cruising	Anchoring	Mooring
8.8072	3.9532	2.3679	1.4182	0.6175	0.4504

. Carbon emissions classification by different dimensions is shown in Figure 7

3.2.4. Time-Series Analysis of Ship Carbon Emissions

The monthly variation in carbon emissions from ships in the Bohai Bay region in 2023 is shown in Figure 8. From the chart, it is evident that ship carbon emissions in the Bohai Bay economic zone exhibit significant monthly fluctuations. This study finds that the monthly variations in carbon emissions reflect alterations in the region’s economic and trade activities and are also influenced by seasonal factors, holidays, and seasonal port operations. From January to March, carbon emissions are relatively low, mainly due to the cold winter season in northern regions, where some ports or shipping routes may be affected by weather conditions, leading to reduced ship traffic and less frequent maritime transport activities. As a result, carbon emissions tend to be lower during this period. Additionally, during the Chinese New Year, production activities in factories and corresponding goods transportation significantly decrease, making February the month with the lowest carbon emissions of the year. From March to May, as maritime trade and shipping activities resume and demand for cargo transportation increases, carbon emissions from ships gradually rise. During the summer months, ship transport activities become more frequent, especially during peak shipping seasons such as the Dragon Boat Festival and the Mid-Autumn Festival, when logistics transportation reaches a peak. During this time, the number of ships operating, the frequency of voyages, and carbon emissions all reach their maxima levels, resulting in the highest carbon emissions from ships in the July to September period. From October to December, as the season changes and winter approaches, ship carbon emissions gradually decrease.

3.2.5. Spatial Analysis of Ship Carbon Emissions

The study area was partitioned into a grid, and, by leveraging latitude and longitude data, each point source was precisely located within the corresponding grid. The carbon emissions from ships in each grid were then calculated, allowing for a further analysis of the spatial distribution characteristics of ship carbon emissions in the region.

From the perspective of port areas, ship carbon emissions in the Bohai Bay region are concentrated near the port clusters in Hebei, Shandong, and Liaoning provinces. Among these, the ship emissions in the Shandong port cluster are significantly higher than those in Hebei and Liaoning. This is primarily because Shandong has a relatively large number of ports and more frequent shipping activities. Major ports in Shandong include Qingdao Port, Rizhao Port, and Yantai Port. Qingdao Port, being a key international shipping hub, has carbon emissions not only from container transport but also from large-scale imports and exports of energy and mineral resources. As a result, maritime activities around Qingdao are dense, leading to relatively high emissions in the surrounding sea areas. Next, in the Hebei port cluster, key ports include Tianjin Port, Qinhuangdao Port, and Tangshan Port. Tianjin Port serves as a major shipping hub in the Beijing–Tianjin–Hebei region, and these ports mainly handle coal and steel transportation. With large shipping volumes, carbon emissions in this area are also relatively concentrated. Lastly, the Liaoning port cluster, which includes ports like Dalian Port and Yingkou Port, sees relatively fewer shipping activities compared to Hebei and Shandong, resulting in more stable and lower carbon emissions in this region.

From the perspective of shipping routes, the Bohai Bay has a high route density, especially the main shipping lanes leading to the Beijing–Tianjin–Hebei economic zone. Carbon emissions from ships in these densely trafficked routes are mainly concentrated during regular sailing activities. With high sailing densities in these routes, carbon emissions are relatively large. These routes are often closely linked to industrial zones, energy production areas, and logistics hubs, causing the carbon emissions in these regions to peak more significantly during high-traffic period.

Ship carbon emissions are influenced by multiple factors, including port density, trade activities, and seasonal climate variations. As mentioned earlier, the carbon emissions from the port clusters in the three provinces are lower in January to March, mainly due to the cold weather and the Chinese New Year period, which reduces trade activities. From May to June, carbon emissions gradually rise as trade activities and shipping frequencies increase, with ship emissions reaching their zenith in September. The spatial distribution of carbon emissions is shown in Figure 9.

3.3. Speed Simulation Scenario Modeling

3.3.1. Simulation Data Description

The simulation data used in this study are derived from the dynamic and static AIS data of ships throughout the year 2023. The data focus on the ships operating in the Bohai Sea region, recording their trajectories, processing trajectory points, and analyzing vector paths to generate continuous movement trajectories of the ships. Detailed information is recorded for each trajectory point, including ship status, speed, dynamic draft, and sailing time, which serves as the foundational data for the simulation. Through the analysis and computational mining of the AIS trajectory data, a total of 10,993 ships and 97,628 valid voyages were recorded in the Bohai Sea region in 2023. The specific ship data for the simulation, based on this analysis, are shown in Figure 10.

3.3.2. Preprocessing of Simulation Data

To conduct an analysis and calculation of the carbon emissions of ships under different voyages and speeds, it is necessary to perform clustering and slicing operations of the ordered AIS movement trajectory points of each voyage. These slices are grouped into multiple speed clusters, with each cluster representing a 1 h interval. Carbon emissions are then calculated based on the different speed slices. The specific processing flow is shown in Figure 11.

(1) For the 97,628 valid voyages of ships in the Bohai Sea region in 2023, load the trajectory data for each voyage M, one by one.

(2) Sequentially extract the trajectory points from the voyage, recording information such as the ship’s speed, dynamic draft, and the timestamp of each point to form point P.

(3) Determine whether trajectory point P and the previous trajectory point P-1 are similar enough to be clustered together. The criteria for clustering are primarily that the speed fluctuation does not exceed 2%, and the dynamic draft does not exceed 2%. If both conditions are met, the point is considered a similar-speed point and is added to the current time slice sequence. The speed of this slice is set to the speed of the first point in the slice.

(4) Check if the time difference between trajectory point P and the start point of the slice exceeds 1 h. This helps determine whether a complete slice can be formed. If the time difference is less than 1 h, proceed to analyze the next point in the voyage.

(5) If the time difference between the similar point and the start point of the slice exceeds 1 h, a complete clustering slice for voyage M is retained, with the speed set as the speed of the first point in the slice.

(6) If trajectory point P and the previous point P-1 do not form a similar pair, set point P as the starting point for a new slice and continue the analysis with the next point.

3.3.3. Speed Simulation Calculation and Analysis

After the slicing and clustering of the simulation data, each voyage will result in either 0 or N speed clusters. Valid trajectories need to be selected, where the same voyage contains more than three consecutive speed slices, and the speed deviation between these slices is within approximately 10%. Based on this, a speed simulation calculation model is established:

V_P = MEDIAN (V_n1 + V_n2 + V_n3+ V_nx)

(10)

V_t = ((V_n − V_P)/V_p) × 100%

(11)

V_p−1 = −12% < V_t < −8%

(12)

V_p+1 = 8% < V_t < 12%

(13)

where V_p: The median speed of all speed slices in this voyage, used as the reference speed for carbon emission calculation. V_t: The deviation between the speed of any slice in the voyage and the reference speed (V_p). V_p−1: A speed slice in this voyage that is 8–12% lower than the reference speed (V_p). V_p+1: A speed slice in this voyage that is 8–12% higher than the reference speed (V_p).

According to the model calculations, out of the 97,628 voyages of 10,993 ships in the Bohai Bay area, 8792 ships with 64,753 voyages meet the condition that there are three speed slices within the same voyage, with a speed difference of 8–12%. This study further uses a carbon emission calculation model to assess the carbon emissions of the selected voyages and ship speed slices from the simulation scenarios (

Table 11. Simulation calculation example.

Basic Information of Sampled Ships
Ship ID	MMSI	Ship Name	Length	Width	Deadweight Tonnage	TEU
A1	47,731 ****	HYHX **	399.9	58.6	197,975	21,237
A2	47,717 ****	HYTP **	399.8	58.6	199,924	20,038
B1	47,723 ****	XT **	141	24	12,336	1011
B2	47,753 ****	WH **	142.7	22.6	12,827	1043
……	……	……	……	……	……	……
……	……	……	……	……	……	……
Emission Data of Sampled Ships
Ship ID	Slow-speed Cruising Carbon Emissions (tons)		Medium-speed Sailing Carbon Emissions (tons)		High-speed Sailing Carbon Emissions (tons)
	Speed (knots)	Carbon Emissions	Speed (knots)	Carbon Emissions	Speed (knots)	Carbon Emissions
A1	10.60	0.066	11.65	0.069	13.05	0.073
A2	9.97	0.062	11.08	0.066	12.19	0.070
B1	11.44	0.047	12.85	0.049	14.01	0.051
B2	11.38	0.045	12.64	0.047	14.03	0.050
……	……	……	……	……	……	……

**** is the data desensitization of MMSI. ** is the data desensitization of Ship Name.

). The results show that, when the ship speed in the Bohai Bay area (primarily consisting of container ships, tankers, and bulk carriers) decreases by around 10%, carbon emissions slightly decrease as well, but the reduction is no more than 6%. Although a reduction in speed significantly affects carbon emissions, the Bohai Bay region has dense shipping lanes, is often congested, and ships frequently operate in short-distance voyages. The overall average speed of ships in the region is already relatively slow, usually below the optimal economic speed. Therefore, further reductions in speed will no longer have a significant impact on carbon emissions. The distribution of carbon emissions at different speeds is shown in Figure 12.

4. Conclusions and Discussion

4.1. Conclusions

The mitigation of ship carbon emissions in the Bohai Bay area is a necessary measure to address global climate change, fulfill China’s international environmental commitments, achieve regional sustainable development, and promote green shipping. This study, based on 2023 AIS trajectory data for ships in the Bohai Bay area, utilizes data preprocessing and big data mining techniques to construct ship carbon emission calculation models and speed simulation models. It assesses the influence exerted by ship carbon emissions and speed on the region throughout the year and provides recommendations for measures under different carbon reduction policies.

(1) In 2023, the total carbon emissions from ships in the Bohai Bay area amounted to 8.8072 million tons. By ship type, the main contributors to carbon emissions were cargo ships, with emissions of 5.8866 million tons, accounting for 66.84%. The next largest contributors were tankers and workboats, while passenger ships and other vessels had smaller shares. In terms of emission sources, the largest contributor was the main engine, with emissions of 5.8724 million tons, accounting for 66.07%. Regarding ship operating conditions, the largest share of emissions stemmed from ships in cruising state, which emitted 3.9532 million tons, accounting for 44.89%. This underlines the significant impact of cruising on carbon emissions.

(2) Furthermore, ship carbon emissions are influenced by various factors, including social, climatic, and trade activities. From January to March, ship carbon emissions were relatively subdued, mainly attributable to the seasonal impact of the northern winter, which reduced shipping demand. In the shipping peak season from July to September, ship transport activities were more frequent, and the number of ships and the frequency of voyages reached their peak, leading to the highest carbon emissions in September. During the peak season, carbon emissions also exhibited a relatively concentrated pattern, especially near large ports such as Tianjin, Qingdao, and Dalian, where frequent ship movements due to high trade demand resulted in increased ship activity near ports.

(3) Through speed simulation scenarios, there were 10,993 ships operating in the region in 2023, with 97,628 valid voyages. If all ships reduced their speed by approximately 10%, carbon emissions for the entire region could decrease, but the overall reduction would not exceed 6%. Further reductions in speed would not significantly impact carbon emissions.

4.2. Discussions

The Bohai Bay is an important shipping area in Northern China, covering major ports such as Qingdao, Tianjin, and Dalian, and is an essential part of the global shipping network. Although carbon emissions from ships in the Bohai Bay area may meet some of the IMO’s mid-term requirements, for instance, the Ship Energy Efficiency Design Index (EEDI) and Energy Efficiency Operational Indicator (EEOI), in general, additional measures will be required to fully achieve the IMO’s carbon reduction targets set for 2030 and 2050.

(1) Route and Speed Optimization: Shipping routes in the Bohai Bay area are often restricted by maritime geographical features, port distribution, and weather and current conditions. By optimizing routes, it is possible to reduce the sailing distance and time, minimize congestion delays, and avoid bad weather and adverse currents, thereby diminishing fuel consumption. Although the average speed of ships in the region is already relatively slow, speed optimization can help avoid prolonged periods of cruising, which could further reduce carbon emissions.

(2) Fuel Transition and Technological Innovation: Some ships in the Bohai Bay area still use highly polluting traditional fuels. Shipping companies should be encouraged to use low-carbon or zero-carbon fuels such as liquefied natural gas (LNG), biofuels, and hydrogen. Additionally, green shipping alternatives, including the exploitation of wind energy, solar energy, and battery-driven technologies, should be explored.

(3) Port Facility Enhancement and Management Optimization: Some ports in the Bohai Bay area, such as Tianjin and Qingdao, have already implemented ship carbon emission monitoring and control measures, including the construction of port shore power systems and the application of intelligent scheduling systems. However, the overall progress across the region’s ports remains uneven, and more efforts are needed to enhance equipment upgrades and system development, particularly at smaller and medium-sized ports.

(4) Precision Decision Making with Intelligent Technologies: The Bohai Bay region can also leverage intelligent technologies such as data analysis, real-time monitoring, and optimization algorithms to make precise decisions and provide new solutions for reducing ship carbon emissions. In ship management, the real-time monitoring of ship operating conditions, including the engine and propulsion system, combined with big data analysis and machine learning, can lead to more accurate decisions regarding route and speed optimization. In port management, intelligent decision-making systems can be developed and applied to optimize ship berthing and berth scheduling, reduce waiting times and ballast voyages, and streamline ship operations, thereby reducing carbon emissions.