Total life cycle emissions of post-Panamax containerships powered by conventional fuel or natural gas

ABSTRACT

This study proposes an easy-to-apply method, the Total Life Cycle Emission Model (TLCEM), to calculate the total emissions from shipping and help ship management groups assess the impact on emissions caused by their capital investment or operation decisions. Using TLCEM, we present the total emissions of air pollutants and greenhouse gases (GHGs) during the 25-yr life cycle of 10 post-Panamax containerships under slow steaming conditions. The life cycle consists of steel production, shipbuilding, crude oil extraction and transportation, fuel refining, bunkering, and ship operation. We calculate total emissions from containerships and compare the effect of emission reduction by using various fuels. The results can be used to differentiate the emissions from various processes and to assess the effectiveness of various reduction approaches. Critical pollutants and GHGs emitted from each process are calculated. If the containerships use heavy fuel oil (HFO), emissions of CO₂ total 2.79 million tonnes (Mt), accounting for 95.37% of total emissions, followed by NO_x and SO_x emissions,which account for 2.25% and 1.30%, respectively.The most significant emissions are from the operation of the ship and originate from the main engine (ME).When fuel is switched to 100% natural gas (NG), SO_x, PM₁₀, and CO₂ emissions show remarkable reductions of 98.60%, 99.06%, and 21.70%, respectively. Determining the emission factor of each process is critical for estimating the total emissions. The estimated emission factors were compared with the values adopted by the International Maritime Organization (IMO).The proposed TLCEM may contribute to more accurate estimates of total life cycle emissions from global shipping.

Implications: We propose a total life cycle emissions model for 10 post-Panamax container ships. Using heavy fuel oil, emissions of CO₂ total 2.79 Mt, accounting for approximately 95% of emissions, followed by NO_x and SO_x emissions. Using 100% natural gas, SO_x, PM₁₀, and CO₂ emissions reduce by 98.6%, 99.1%, and 21.7%, respectively. NO_x emissions increase by 1.14% when running a dual fuel engine at low load in natural gas mode.

Introduction

The Marine Environment Protection Committee (MEPC) of the International Maritime Organization (IMO) first reported on greenhouse gas (GHG) emissions from ships in 2000 (Bodansky 2016; IMO 2000). In 2009, the MEPC working group updated the global shipping emissions of air pollutants and greenhouse gases (IMO 2009). The “Third IMO GHG Study 2014—Final Report” updated global shipping GHG emissions from 2007 to 2012. For the year 2012, total shipping emissions were approximately 949 million tonnes (Mt) of CO₂, and 972 Mt of CO₂ equivalent (CO₂e, combining CO₂, CH₄, and N₂O), accounting for 2.7% and 2.5% of global emissions, respectively. International shipping emissions for the year 2012 were estimated to be 796 Mt of CO₂ and 816 Mt of CO₂e, accounting for approximately 2.2% and 2.1%, respectively, of global CO₂ and GHG emissions on a CO₂e basis.Based on the average emissions over the 6-year period, shipping accounted for 3.1% of annual global CO₂ and 2.8% (approximately 1.038 billion tons) of annual GHGs on a CO₂e basis (IMO 2011; 2012; 2014).

Ships emit GHGs, including CO₂, CH₄, and N₂O, and other pollutants including CO, NO_x, nonmethane volatile organic compounds (NMVOC), particulate matter (PM), and SO₂. Some of these emissions influence climate change significantly. Previous studies of SO_x and NO_x emissions from ships found that 87 kg of NO_x was emitted per ton of fuel consumed by low-speed diesel engines and 57 kg of SO₂ was emitted by medium-speed engines, with SO₂ emissions related to fuel sulfur content (Corbett and Fischbeck 1997; Corbett and Koehler 2003). Containerships of 8,000 TEU (20-foot equivalent units) and above emitted 39.7 Mt CO₂ in 2012 globally. This contributes to 19.3% of total emissions from ships of all types (IMO 2014). Emission factors (EFs), defined as the amount (g or kg) of pollutant produced by an activity relative to the intensity (kWh) of that activity, are widely adopted to study shipping emissions (Eyring et al. 2005; Hua et al., 2017).

The first study on EFs used a total fuel cycle analysis (TFCA) for automobiles, integrating large amounts of statistical data, fuel production process boundary conditions, and complex calculations (Winebrake, Corbett, and Meyer 2007). The accurate investigation of emissions from marine transportation requires a total fuel life cycle analysis, encompassing energy use and emissions from all related processes, from the extraction of raw fuel (e.g., oil from deep sea or underground) to the use of the processed marine fuel. Each stage in the fuel life cycle involves activities that emit GHGs and pollutants. In 1996, the Center for Transportation Research at Argonne National Laboratory (ANL) developed a spreadsheet-based fuel-cycle model called the Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation (GREET) model (Winebrake, Corbett, and Meyer 2006; 2007). This model has become the “gold standard” for conducting TFCA. GREET was the first model to apply the well-to-pump (WTP) concept to TFCA and was subsequently utilized in many studies (Wang 1996, 1999).

Winebrake et al. (2007) developed the TEAM model by utilizing the core concepts and algorithms in GREET, conducted total life cycle analysis of marine fuels, and computed air pollutant and GHG emissions from upstream crude oil extraction, transportation, refining, and ship operations. The emissions from six different fuels (conventional diesel, CD; heavy fuel oil, HFO; natural gas, NG; low-sulfur diesel, LSD; biodiesel, B20; and Fischer–Tropschdiesel, FTD) during the entire life cycle were compared.The results indicated that, regardless of the fuel type, ship operations accounted for the highest proportion of CO₂, volatile organic carbon (VOC), CO, NO_x, and PM₁₀ emissions during the fuel life cycle. For a ferry ship, operational processes accounted for 84.15%, 91.99%, 86.55%, 82.61%, 94.38%, and 75.93% of CO₂ emissions from these six fuels, respectively. For a 6,000-TEU containership, HFO was the most efficient fuel but emitted the largest amount of pollutants (190 t of SO_x per voyage), followed by CD (5.7 t emitted).Total NO_x emissions from HFO were greater than those from CD and NG, but lower than from LSD. This was the first study to focus on the total life cycle emissions for different fuels, using EFs (g/kWh) as the basic unit for the estimation of pollutants and GHG emissions, and to make a full comparative analysis of total life cycle emissions. The TFCA results for ferries, containerships, and oil tankers provide significant contributions to marine transportation research (Winebrake, Corbett, and Meyer 2006). The ANL later integrated these results along with EFs from marine engines into the GREET platform. Emission estimation using EFs provides a simplified but widely adopted approach.

Agrawal et al. (2010) conducted long-term measurements of NO_x, CO, CO₂, and PM emissions from the ME on a post-Panamax containership under different engine load conditions according to the International Organization for Standardization (ISO) 8178–1 standard (ISO 2007). Measured EFs were 19.77 ± 0.28, 0.29 ± 0.02, 617 ± 11.00, and 11.53 ± 0.20 g/kWh, respectively. The EF for SO₂ is 2.399 ± 0.052 g/kWh, obtainedfrom the sulfur content in fuel oil. These measured EFs are higher than those of GREET or Winebrake et al. (2006; 2007) and Wang (1996; 1999) because the containership was built in 1998 when NO_x emissions were not yet controlled by MARPOL regulations.

Bengtsson, Andersson, and Fridell (2011) explored the differences in CO₂, NO_x, SO_x, and other GHG ship emissions using HFO, marine gas oil (MGO), or liquefied natural gas (LNG) as fuel in a life cycle analysis. Furthermore, some two-stroke marine diesel engines built from 2000 onward use electronic-hydraulic rather than mechanical camshaft and chain controls for fuel oil injection timing. Consequently, combustion efficiency is much improved, and thus NOx, CO, CO₂, PM, and SO_x emissions are reduced. Using LNG fuel can reduce CO₂ emissions; however, the global warming potential of CH₄ is 34 times that of CO₂ (DNV-GL 2015). Therefore, if LNG dissipation, leakage, or slip during combustion processes (methane slip) cannot be efficiently reduced throughout its full life cycle, excessive release of CH₄ to the atmosphere may lead to more serious global warming effects than when HFO is used. Compared with HFO, SO_x emissions can be reduced by approximately 90% by using LNG fuel. The NO_x EF of 1.3 g/kWh was adopted from the emissions data for Wärtsilä four-stroke spark ignition NG engines, but no further discussion was included on the issue of the remarkably low EF.

According to the IMO (2014), containerships were the largest sources of maritime pollution in 2012, consuming the most fuel oil (approximately 66 Mt) and emitting the most CO₂ (approximately 210 Mt). Post-Panamax containerships, the most popular type of ships during the last decade, consume an average of approximately 200 t of fuel oil per ship per day, accounting for approximately 15% of CO₂ emissions across all types of containerships (IMO 2014). These high total emissions, along with the reported emission reduction potential of using NG fuel (Bengtsson, Andersson, and Fridell 2011), provide the motivation for studying the total life cycle emissions from post-Panamax containerships.

In establishing policy regarding emission inventory and abatement it is important to have a more accurate calculation of the emission of various pollutants in addition to CO₂, under various operation modes. This study proposes an easy-to-apply method, the Total Life Cycle Emission Model (TLCEM), to calculate the total emissions from shipping. Focused on the most essential environmental issues facing the industry, the model aims at helping ship management groups assess the impact on emissions due to their choice of fuels, operation modes, and fuel supply chains. Using TLCEM, this study examines the total air pollutant and GHG emissions during the 25-yr life cycle of 10 post-Panamax container sister ships under slow steaming operation. The ships were built in Taiwan between 2006 and 2013. Life cycle analysis not only consists of total emissions from all ship activities but also from all major processes from cradle to grave, such as steel production, shipbuilding, crude oil extraction and transportation, fuel refining, and bunkering processes. The NO_x EFs and specific fuel oil consumption (SFOC) at different engine loads are measured at the test bed and analyzed to calculate emissions arising from ship operations. Other EFs of VOC, CO, PM₁₀, SO_x, CH₄, N₂O, and CO₂ involved in all life cycle processes are obtained from GREET, from TEAM software, and from corporate responsibility reports (CSC 2013; CPC 2013; CSBC 2013).

Dual fuel (DF) engines can combust gas fuel or liquid fuel and switch smoothly at anytime during operation (Burel, Taccani, and Zuliani 2013; Classification Society 2014; Einang 2007; Germanischer, 2013; IMO 2008; Wang and Huang 2000). In the era of cleaner gas fuels, containerships will be fitted with DF main engines (MEs) and auxiliary engines (AEs) with the same power rating as current engines, using 50% and 100% NG instead of HFO. In recent years, a number of studies have assessed the life cycle impact of shipping using different methods (e.g.,Cucinotta, Guglielmino, and Sfravara 2017; Favi et al. 2017; Jalkanen, Johansson, and Kukkonen 2016). This study presents the average total reductions in the emissions of all pollutants and GHGs (except NO_x emissions, which increased) for the 25-yr life cycle of 10 containerships.The differences between the IMO EFs and those adopted in this study, as well as the reasons behind these differences, are subsequently discussed.

Method and procedure

The life cycle processes considered in our study include steel manufacture, ship building, oil refinery, crude oil exploration, and transport and bunkering. It has wider scope and longer (25-yr) duration, comparing to the functional unit of only single navigation typically assumed in most life cycle study. The EFs and ship operation modes form the core of the total emissions calculation using the proposed TLCEM. The following sections explain the sources of EFs, the specified operation modes, and the equations used in the TLCEM.

The calculation of total emissions for various pollutants is simpler and feasible for various ship types if data of NO_x and SFOC from the test bed and the mathematical model TLCEM are integrated. We used test-bed data due to the difficulty in practice in installing instruments in the engine room, and operating a measurement that has very limited space, and is typically under high temperature, humidity, and vibration. We used the EF method and TLCEM to efficiently calculate the total emission for the 25-yr life cycle.

Upstream life cycle processes and sources of EFs

Total life cycle emissions data for upstream processes of the fossil fuel industry can vary considerably due to different exploration methods, transportation modes, or locations of oil fields. Previous studies examine the upstream processes for fuels but exclude pollutants and GHGs emitted from steel plants and shipyards, structural material manufacturing, and shipbuilding processes (Bengtsson, Andersson, and Fridell 2011; Brinkman et al. 2005; Kameyama et al. 2004; Wang 1996; 1999; Wang and Huang 2000; Winebrake 2007; Winebrake, Corbett, and Meyer 2006).

In this study, the life cycle for the ships starts at steel manufacture and ends at dismantling. The average lightship weight of 10 ships is approximately 32,785 t. Although a large amount of the weight comes from the ship structure, the machinery, piping, outfitting, and many miscellaneous articles in a ship are also made of steel; therefore, the average lightship weight is considered the total weight of steel. All steel structures for shipbuilding are produced at a steel factory located near the shipyard. In Taiwan, crude oil is imported mainly from the Middle East, transported by very large crude carriers (VLCCs), and refined into various types of fuel oil, of which HFO accounts for 27.72%. These heavy industries conduct emissions inventories and publish annual results to meet ISO and government requirements. Production and emissions data are published in annual corporate social responsibility (CSR) reports (CSC 2013; CPC 2013; CSBC 2013). The EFs of the upstream shipbuilding and oil refinery industries are investigated and listed in Table 1. The crude oil sources, production processes, and pollution EFs involved in the marine fuel oil industries can differ significantly among countries; nevertheless, transportation and bunkering modes for cargo ships are very similar in most international ports. In Taiwan, HFO bunkering for containerships is performed by bunker barges in ports during cargo handling. Specifications of the VLCC, the bunker barge, and the containerships (Figure 1) in this study are listed in Table 2.

Table 1. EFs of upstream shipbuilding and oil refinery industries (kg/t) (Sources: CSC 2013; CPC 2013; CSBC 2013).

	NOx	SOx	PM10	VOC	GHG (CO2e)
Steel factory	0.9787	1.0863	0.3749	0.0724	2,331.2113
Shipbuilding	0.0029	0.0043	0.0209	5.2925	239.6133
Oil refinery	0.2695	0.1983	0.0247	0.1662	420.5758

Note. The unit of EF is weight of emission per tonne (t) of product.

Table 2. Ship specifications (Source: Classification Society 2014).

	VLCC	Bunker barge	Containerships
Dimensions L × B × D (m)	316.0 × 60.0 × 29.7	60.1 × 11.0 × 5.5	333.2 × 42.8 × 24.5
Deadweight (t)	300,000	1,680	8,600 (TEU)
Year built	2008	1999	2006 to 2013
Total ME power (kW)	25,090 × 1 (set)	994 × 2 (sets)	68,666 × 1 (set)
ME (RPM)	78.6	900	94
Rated speed (knots)	17.0	12.0	25.6
Total AE power (kW)	880 × 3 (sets)	400 × 3 (sets)	3,162 × 3 (sets)

Figure 1. Picture of studied containership.

GREET covers emissions information on various upstream raw materials, energy consumption, and WTP transportation in the United States and other countries, and contains life cycle emissions data for NG and shale gas during extraction and production processes. However, GREET lacks detailed and specific information such as the transportation methods and distances (Abrahams et al. 2015; Brinkman et al. 2005; Winebrake 2007; Wu, Wu, and Wang 2006). In contrast, TEAM (Winebrake, Corbett, and Meyer 2006, 2007) provides not only the emissions data for NG production and liquefied processes but also the detailed transportation methods (including pipelines), ship specifications, and specific distances. In this study, comprehensive total life cycle EFs are established by combining the individual functions of GREET and TEAM. The EFs for oil field and LNG WTP are from GREET and TEAM and are summarized in Table 3. The methane leakage from oil and natural gas fields, as well as from the fuel production, was quantified and included in the CH₄ emission factor as noncombustion emissions.

Table 3. EFs of oil extraction and fuel transportation (kg/t).

	VOC	CO	NOx	PM10	SOx	CH4	N2O	CO2
Oil field	0.4152	0.3142	0.7533	0.0863	0.2014	5.7330	0.4910	9.4529E+1
VLCC	2.7260E–2	7.0268E–2	5.7817E–1	5.6792E–2	3.8340E–1	2.3502E–4	1.2388E–3	2.2512E+1
Bunkering	2.3359E–3	6.4238E–3	8.5845E–2	8.4093E–3	6.9377E–2	2.3359E–5	1.8103E–4	4.0817
LNG WTP	7.9000E–2	4.5750E–1	1.5394	3.8227E–2	7.2073E–1	8.4952	1.1785E–2	5.7360E+2

Note. The unit of EF is weight of emission per ton (t) of oil, for fuel extraction or transport.

Ship operation modes and data acquisition

Most large cargo ships are designed for an operational life span of 25 yr. Air pollutant and GHG emissions are mainly from fuel combustion and are related to combustion efficiency, combustion temperature, and the carbon and sulfur content of the fuels (Laursen 2015; MAN Diesel & Turbo 2012; 2014). Different types or sizes of ships, such as containerships and bulk carriers, navigate different trading routes; therefore, operation modes are entirely different. Figure 2 shows possible routes of the studied ships. Evaluating total air pollutant and GHG emissions for an entire ship life cycle based on a single voyage or a single ship cannot reveal exact conditions. The trends toward large-scale ships has increased the propulsive power. Based on the propeller law, a higher ship speed requires a higher propulsive power proportional to the cubic power of the propeller speed, with a corresponding increase in fuel consumption. To reduce fuel costs, almost all large ships adopt slow steaming as the best strategy (Ballou, Chen, and Horner 2008; Bureau of Energy, Ministry of Economic Affairs 2015; Fagerholt, Laporte, and Norstad 2010; MAN Diesel & Turbo 2012; Meyer, Stahlbock, and Voß 2012; Psaraftis, Kontovas, and Kakalis 2009; S&P Global Platts 2015; Wang and Meng 2012).

Figure 2. Possible routes of the studied ships.

Over a ship’s lifetime, most emissions occur during ship operation; therefore, the mode of operation is important for analyzing total life cycle emissions. We assumed that a VLCC is responsible for the transportation of crude oil from the Middle East to Taiwan. The typical operation mode of a VLCC includes cargo loading, full load departure, navigation under full load conditions, full load arrival, cargo unloading, and then departure, navigation, and arrival under ballast conditions (Table 4). The average data on navigation distance, ship speed, duration, sea factor, and ME and AE load factors for all voyages conducted in a year are detailed by the voyage stage. Sea factor refers to the consumption of additional propulsive power due to hull fouling or rough sea conditions and is assumed to be 3% in this study.

Table 4. VLCC crude oil transportation mode.

Operation mode	Distance (nautical miles)	Speed (knots)	Time (h)	ME load factor	Sea factor	AE sets	AE load factor
Cargo loading	0	0.00	31.00	0.00	0.00	1	0.77
Departure/arrival (full load)	—	4.00	2.00	0.20	1.00	2	0.78
Navigation (full load)	6,100	13.40	455.22	0.50	1.03	1	0.75
Cargo unloading	0	0.00	31.00	0.00	0.00	2	0.70
Departure/arrival (ballast condition)	—	4.00	2.00	0.15	1.00	2	0.78
Navigation (ballast condition)	6,100	13.40	455.22	0.40	1.03	1	0.75

Since a bunker barge navigates between the bunker terminal and cargo ships in the port, the bunkering mode of the operation is simpler (Table 5). The total amount of crude oil transported by the VLCC, or bunker fuel transported by the bunker barge, depends on the HFO demand of the containership activities over the 25-yr life cycle.

Table 5. Barge bunkering operation mode.

Operation mode	Speed (knot)	Time (h)	ME load factor	ME sets	AE sets	AE load factor
Loading	0.00	5.00	0.00	0	1	0.24
Navigation	8.00	1.00	0.70	2	1	0.25
Unloading	0.00	5.00	0.00	0	2	0.70

The 10 containerships of this study navigate between the northern ports of China and international ports of Europe, covering approximately 10,500 nautical miles and visiting 13 ports on a one-way voyage, with an average of five round trips per year. The average loading/unloading hours at the berth, the ship speed during departure/arrival maneuvering inthe port, and the average navigation distance, speed, and time spent at sea, together with ME and AE load factors at each stage, are shown in Table 6. Data on ship operation modes were extracted from a minimum of 12 months of actual activities between 2013 and 2015, where the ME load, power output, RPM, and AE power output were extracted from the electronic control system of the ME and the engine room log book.

Table 6. Containership operation mode.

Operation mode	Distance (nautical miles)	Average speed (knots)	Time (h)	ME load factor	Sea factor	AE sets	AE load factor
Loading/unloading	0	0.00	8.00	0.00	0.00	1	0.73
Departure/arrival	—	6.00	3.00	0.15	1.00	2	0.47
Navigation	10,500	16.13	650.96	0.26	1.03	1	0.72

To reduce fuel expenditure, shipping companies use a strategy of slow steaming, especially for large ships with high power ratings, such as oil tankers, containerships, and bulk carriers. The average ME load is reduced to less than 50% and the ME load for containerships is reduced even further to less than 30%(Ballou, Chen, and Horner 2008; Fagerholt, Laporte, and Norstad 2010; Meyer, Stahlbock, and Voß 2012; S&P GlobalPlatts, 2015; Wang and Meng 2012). However, the NO_x EF of a marine diesel engine in low-load running conditions is higher than that under high loads and has been identified and verified during engine testing (Classification Society 2014). For instance, the NO_x certification value of the two-stroke main (parent) engine of the containerships is 14.0 g/kWh, which was measured using a test bed and calculated using weight factors at different load levels according to the NO_x Technical Code 2008 (IMO 2008). The measured values are 11.94 g/kWh at 100% load, 13.75 g/kWh at 75% load,12.97 g/kWh at 50% load, and 17.70 g/kWh at 25%load. The average load factor of containerships is 0.26 (Table 6); therefore, via linear interpolation, a figure of 17.511 g/kWh (Table 7) is adopted for the NO_x EF in this study to reflect actual operating conditions and increase the accuracy of the results. The same method is applied to four-stroke AEs. To estimate total fuel consumption throughout the life cycle, an SFOC of 182.5 g/kWh of ME at 26% load and 201.6 g/kWh of AE at 72% load are applied, as obtained from the engine tests.

Table 7. EFs of the ME and AE of a containership in HFO mode (g/kWh).

	VOC	CO	NOx	PM10	SOx	CH4	N2O	CO2
ME	0.700	1.800	17.511	1.420	9.450	0.006	0.031	554.790
AE	0.400	1.100	8.940	1.440	11.880	0.004	0.031	698.940

All diesel engines of the VLCC and 10 containerships were tested prior to installation to calculate engine performance, SFOC, and NO_x EFs at rated power quartiles. Engine power was measured using a Fuchino dynamometer (serial no. 91,032, maximum measurement range 80 MW), NO_x emissions were analyzed with a HORIBA instrument (model CLA-155, maximum measurement range 2,000 ppm), and the weighting machine used to measure SFOC was manufactured by CAS (model 1W78). Machine testing, equipment, calibration records, and test results are recorded in detail in well-maintained NO_x Technical Files (Classification Society 2014).

Mathematic model

The following mathematical model, TLCEM, expresses the total emissions of pollutants and GHGs from various processes throughout the ship’s life cycle:

(1) $E{M}_{TL,i}=E{M}_{cm,i}+E{M}_{Ncm,i}$(1)

(2) $E{M}_{cm,i}=\sum _{k}E{F}_{i,k}\times R{P}_k\times L{F}_k\times T_k\times {10}^{-6}$(2)

(3) $E{M}_{Ncm,i}=\sum _{N}E{F}_{i,N}\times T{W}_N\times {10}^{-3}$(3)

where

$E{M}_{TL,i}$= total life cycle emissions of pollutant or GHG$i$ (t)

$E{M}_{cm,i}$= total life cycle emissions of pollutant or GHG$i$,from engine combustion processes $k$(t)

$E{M}_{Ncm,i}$= total life cycle emissions of pollutant or GHG$i$, from non-engine

combustion processes $N$(t)

$E{F}_{i,k}$= EF of pollutant or GHG$i$, in engine combustion processes $k$(g/kWh)

$R{P}_k$= rated power of engine in combustion processes $k$(kW)

$L{F}_k$= load factor of engine in combustion processes $k$

$T_k$= running hours of engine in combustion processes $k$(h)

$E{F}_{i,N}$= EF of pollutant or GHG$i$, in non-engine combustion

processes $N$ (kg/t)

$T{W}_N$  = total weight of product in non-engine combustion processes$N$(t)

The total life cycle processes are divided into engine combustion and non-engine combustion processes in this study. The engine combustion processes, $k$, include the combustion in the main propulsion engine and generator engines, which consume HFO or NG in all ship activities, as well as oil transportation by VLCC and bunkering processes.The non-engine combustion processes, $N$, comprise the production processes in steel factories, shipyards, oil extraction, and refineries. To simplify the total emissions calculations, EFs are expressed in g/kWh or kg/t. The pollutant and GHG EFs of low-speed two-stroke ME and four-stroke AE combustion processes are presented in Table 7 (Winebrake 2007).

Our model of TLCEM follows principles and framework for life cycle assessment (LCA) specified by the international standard ISO 14,040. TLCEM does not fully adopted the guidelines of ISO 14,044, however, as it focuses only on inventory of air pollutants and GHGs, and does not perform a comprehensive inventory of all inputs and outputs. Instead, TCLEM considers in much detail the engine combustion processes, including those from fuel transport and bunkering, and allows for the contemplation of changes in operation modes, choices of fuel chains, or choices of fuels. Non-engine combustion processes are also included, based on the indirect inventory results from steel factories, shipyards, oil extraction, and refineries. TCLEM intends to provide an easy-to-apply method that helps ship management groups assess the long-term environmental impacts due to their capital investment and operation decisions, focused first on the most essential environmental issues facing the industry.

Results

Fuel oil consumption and crude oil demand

Total emissions associated with oil extraction, VLCC transportation, oil refining, and bunkering processes depend on the activities of the containership and fuel oil consumption (FOC) over the 25-yr life span. Average annual FOC per post-Panamax containership is approximately 0.026 Mt. Approximately 2.38 Mt of crude oil must be extracted, transported, and refined into HFO over the 25-yr period.FOC is obtained by multiplying SFOC and total work (kWh) done by each engine. Total work is calculated from the actual ship activities in a year.

Total life cycle emissions in HFO mode

Almost all cargo ships use HFO as fuel at sea to reduce operational costs. Total emissions of pollutants and GHGs from post-Panamax containerships in the HFO mode for each process during the 25-yr life cycle are shown in Table 8. The results are calculated using eqs 1–3 of the TLCEM, and all necessary data are provided in Tables 1–7. Total CO₂ emissions amount to 2.79 Mt (95.37%), which is far greater than those of other pollutants or GHGs, followed by NO_x and SO_x emissions (only 0.066 Mt [2.25%] and 0.038 Mt [1.30%], respectively). The results show that PM₁₀ (5705.33 t) and VOC (3814.75 t) emissions are extremely low, accounting for only 0.19% and 0.13%, respectively. CH₄ and N₂O emissions data for steel production, shipbuilding, and oil refinery processes are not separately reported and are included with other GHGs as CO₂e. EFs from these processes were obtained from the CSR reports of the Taiwanese manufacturers. See Table 1 for details. The GHG emissions from these three processes, though insignificant, were included to provide a complete picture of the life cycle emissions. CH₄ and N₂O emissions from the remaining processes account for only 0.47% and 0.04% of total emissions, respectively, and are primarily emitted during oil extraction. CO is mainly emitted during operation and accounts for only 0.25% of total emissions. Life cycle emissions of CO, CH₄, and N₂O are likely underestimated, given that this study calculates only indirect emissions for steel production, shipbuilding, and oil refinery processes based on EF factors reported by the manufacturers.

Table 8. Containership 25-yr total life cycle emissions in HFO mode (t).

	VOC	CO	NOx	PM10	SOx	CH4	N2O	CO2
Steel factory	2.37	n.a.^a	32.09	12.29	35.62	n.a.	n.a.	76,428.76^1
Shipbuilding	173.51	n.a.^a	0.09	0.68	0.14	n.a.	n.a.	7,855.72^1
Oil field	990.15	749.29	1,796.44	205.80	480.29	13,671.82	1,170.92	225,428.79
VLCC	65.01	167.57	1,378.78	135.44	914.32	0.56	2.95	53,685.87
Oil refinery	109.87	n.a.^a	178.17	16.33	131.08	n.a.	n.a.	278,029.24^1
Bunkering	1.54	4.25	56.75	5.56	45.86	0.02	0.12	2,698.26
Operation	2,472.29	6,393.00	62,334.63	5,329.22	36,613.02	21.48	116.12	2,150,202.97
Total	3,814.75	7,314.11	65,776.95	5,705.33	38,220.34	13,693.87	1,290.11	2,794,329.62
Percentage	0.13%	0.25%	2.25%	0.19%	1.30%	0.47%	0.04%	95.37%

Notes. Emissions from steel factory, shipbuilding, and oil refinery include total GHG emissions, measured in t-CO₂e, and CH₄ and N₂O are not separately reported. EFs from these processes were obtained from the CSR reports of the Taiwanese manufacturers. See Table 1 for details.

^a CO emissions were underestimated due to the fact that EF were not reported in the CSR reports of the steel factory, shipbuilding, and oil refinery companies.

Generally, emissions from steel factories, shipbuilding, oil refineries, and bunkering processes are low because the rated power of such engines is very low and short operational hours generate lower emissions. Less than 1 t of NO_x, PM₁₀, and SO_x is emitted from the shipbuilding process, and VOC emissions amount to 173.51 t, most of which is generated by painting activities. The amounts of VOC, NO_x, PM₁₀, and SO_x emitted from the steel production are 2.37 t, 32.09 t, 12.29 t, and 35.62 t, respectively. CO₂ emitted from oil fields and oil refineries accounts for 8% and 10% of total emissions, respectively, and 77% of total emissions when emitted during operation. Most CH₄ and N₂O emissions occur during oil extraction, and other pollutants and GHGs are mainly emitted during ship operation.

EFs and total emissions in NG mode

To investigate the potential pollution reductions associated with cleaner fuels, this study assumed that containerships, VLCCs, and bunker barges are installed with DF MEs and AEs of power ratings equivalent to traditional diesel engines. The spaces on board for LNG fuel systems are sufficient for the required operation and the conditions for each process remain unchanged. The EFs of pollutants and GHGs are listed in Table 9 and total emissions are listed in Table 10 for DF engines of large cargo ships that use LNG fuel. Typically, the NO_x EF is approximately 1 g/kWh less when the main DF engines operate in high-load gas mode compared to the diesel mode; however, the EF increased during the low-load operation, even exceeding the EF in the diesel mode. Therefore, to increase the accuracy of the estimates, the DF engine EFs of CO, NO_x, CH₄, and CO₂ were measured from the test bed and applied in this study, while the other EFs were kept the same as in the TEAM model.

Table 9. EFs of DF engine in NG mode (g/kWh).

	VOC	CO	NOx	PM10	SOx	CH4	N2O	CO2
ME	0.5308	0.2000	19.7000	0.0014	0.0026	0.5000	0.0195	452.4000
AE	0.5308	0.2000	2.0273	0.0014	0.0026	6.0000	0.0195	485.7000

Table 10. Containership 25-yr total life cycle emissions in NG mode (t).

	VOC	CO	NOx	PM10	SOx	CH4	N2O	CO2
Steel factory	2.37	n.a.^a	32.09	12.29	35.62	n.a.	n.a.	76,428.76^1
Shipbuilding	173.51	n.a.^a	0.09	0.68	0.14	n.a.	n.a.	7,855.72^1
Oil field	49.51	37.46	89.82	10.29	24.01	683.59	58.55	11,271.44
VLCC	2.53	0.95	64.07	0.01	0.01	3.94	0.09	2,165.06
Oil refinery	5.49	n.a.^a	8.91	0.82	6.55	n.a.	n.a.	13,901.46^1
Bunkering	0.10	0.04	0.39	0.0003	0.0005	1.16	0.0038	93.75
Operation	1,988.22	749.19	63,351.27	5.36	9.81	4,620.51	73.04	1,711,297.19
LNG WTP	50.26	291.07	979.38	24.32	458.53	5,404.70	7.50	364,932.01
Total	2,272.00	1,078.71	66,526.03	53.78	534.68	10,713.90	139.18	2,187,945.40
Percentage	0.10%	0.05%	2.93%	< 0.003%	0.02%	0.47%	0.01%	96.42%

Notes. Emissions from steel factory, shipbuilding, and oil refinery include total GHG emissions, measured in t-CO₂e, and CH₄ and N₂O are not separately reported. EFs from these processes were obtained from the CSR reports of the Taiwanese manufacturers. See Table 1 for details.

^a CO emissions were underestimated due to the fact that EF were not reported in the CSR reports of the steel factory, shipbuilding, and oil refinery companies.

The comparison of Tables 7 and 9 shows that except for NO_x EFs of ME, the EFs of the DF engine in the NG mode are smaller than those of the traditional diesel engine in the HFO mode, especially the EFs of CO, PM₁₀, and SO_x. When NG is used as fuel (Table 10), CO₂ still accounts for most of the emissions (96.42%), at 2.19 Mt, followed by NO_x (2.93%), at 0.067 Mt. The remaining emissions comprise less than 1% of total emissions. For containerships, the 25-yr operational phase is still the largest source of emissions. Except for pilot fuel, HFO is almost completely replaced by NG, and emissions from crude oil extraction, transportation, and refining processes are, therefore, significantly reduced. Emissions from steel factories and shipbuilding only occur during ship construction; they depend on the amount of steel used and are not affected by the type of fuel used during operation. Emissions of these two processes are the same (Tables 8 and 10).

Compared to the results presented in Table 8, most of the emissions are significantly reduced, but CH₄ emissions increased in VLCC, bunkering, and especially during operation due to the methane slip and higher CH₄ EFs of DF engines (Table 9). The methane slips under various loads were 0.7–3.6% (Alvarez, 2012). The use of LNG does not decrease the global warming potential by more than 8–20%. The reduction amount depends mainly on the magnitude of the methane slip from the gas engine (Brynolf 2014). NG extraction, liquefaction, and transportation are included in the LNG WTP process. A great deal of energy is consumed for NG liquefaction and transportation, resulting in significant emission quantities that are second only to those generated during operation. The emissions of VOC, CO, NO_x, PM₁₀, SO_x, CH₄, N₂O, and CO₂ from ship operation and LNG WTP processes account for 89.72%, 96.43%, 99.71%, 55.20%, 87.59%, 93.57%, 57.86%, and 94.89% of total emissions, respectively. Thus, these two processes account for the majority of emissions during the 25-yr life cycle.

To simplify the calculation, this paper combined non-operational processes (non-engine combustion processes, e.g., steel manufacture, ship building, oil refinery) in the construction of unit emission factors used in TLCEM calculation for total emission. Total emissions from steel manufacturers are shown in Tables 8 and 10. Given that the fuel usage by steel manufacturers does not influence fuel use during ship operation, the total emissions are the same under both fuel modes.

Discussion

Emissions from each process

In the HFO mode, crude oil extraction from oil fields accounts for 99.84% and 90.76%, respectively, of CH₄ and N₂O emissions, whereas the emissions of pollutants occur mostly during ship operation (Figure 3). Optimizing the operation is a key to minimizing the environmental impact from shipping without escalating cost significantly (Liu et al. 2018). VOC, CO, NO_x, PM₁₀, SO_x, and CO₂ emissions from ship operation during the 25-yr life cycle represent 64.81%, 87.41%, 94.77%, 93.41%, 95.79%, and 76.95% of total emissions, respectively. These findings are similar to those of Winebrake et al. (2006; 2007). CO₂, NO_x, and SO_x account for 98.9% of total emissions. This explains why the IMO stipulates the ship energy efficiency design index (EEDI), the energy efficiency operational indicator (EEOI), and the ship energy efficiency management plan (SEEMP; IMO 2012) in MARPOL ANNEX VI to reduce CO₂ emissions, SO_x emissions through low-sulfur fuel regulations, and NO_x emissions according to the NTC 2008.

Figure 3. Total emissions proportions for each process in HFO mode.

In this study,VLCCs and bunker barges play the role of oil or fuel transport vehicles and EFs are defined as the weight of emissions per ton of oil or fuel transported.Therefore, total emissions from these two types of ships are lower than those from the target containerships. A similar definition of EFs applies to steel production, shipbuilding, oil extraction, and refining processes.Total emissions of these processes related to the target containerships are presented in Tables 8 and 10. The overall emissions of these industries are excluded.

Emissions reduction by NG

In comparison with HFO, NG is a cleaner fuel of low density and high heating value; it is colorless, odorless, nontoxic, noncorrosive, free of sulfur and PM, and has a low carbon content. It can reduce CO₂ emissions by 21.70 (Figure 4). Using LNG as fuel can help meet the requirements of MARPOL ANNEX VI, which stipulates that sulfur content is to be less than 0.1% in SO_x Emission Control Areas (SECAs). Through appropriate design and control, such as lean-burn control under low pressure, NO_x emissions from DF engines without selective catalytic reduction (SCR) or exhaust gas recirculation (EGR) devices in the NG mode may meet Tier III requirements, except in very low-load conditions or in the diesel mode. Therefore, major global diesel engine manufacturers have invested in DF engine research and development and some engines have been type-approved by independent classification societies. This is a very important development and transition toward the pure gas engine era. Current DF engines still need to be ignited by a small amount of pilot diesel fuel in the gas mode and a higher proportion of pilot fuel is necessary for low-load operations. High-powered containerships require 3–5% pilot fuel to support long-term slow steaming (MAN Diesel & Turbo 2014). Consequently, there is still a small amount of emissions associated with oil extraction, VLCC transportation, oil refining, and bunkering processes.

Figure 4. Emissions reductions due to LNG fuel.

Figure 2 represents the reductions in air pollutant and GHG emissions when HFO is replaced with 50% and 100% NG during the 25-yr life cycle. When 50% of the ship fuel is replaced with NG, emissions of CO, PM₁₀, SO_x, and N₂O are significantly reduced by 43.12%, 50.12%, 49.90%, and 44.63%, respectively. The results in Tables 8 and 10 show that lower HFO consumption significantly reduces the emissions from associated life cycle processes such as crude oil extraction in the oil field, oil tanker transportation, and oil refining processes; however, emissions from NG extraction, liquefaction, and transportation processes increased and are included in the WTP process. NG is a cleaner fuel that is almost free of sulfur. When the fuel is 100% replaced with NG, SO_x and PM₁₀ emissions are reduced by 98.60% and 99.06%, respectively; CO and N₂O are reduced by 85.25% and 89.21%, respectively; and VOC, CH₄, and CO₂ are reduced by 40.44%, 21.76%, and 21.70%, respectively. The CO₂ reduction amounts to approximately 0.61Mt. However, NO_x emissions exceed expectations when DF engines operate at low load in the NG mode. Total NO_x emissions are 1.14% higher than in the HFO mode because of the higher EF. This finding is important, especially for high-powered ships under long-term slow steaming conditions. Special attention should be paid to the finding that CO₂ and NO_x still account for 99.35% of total emissions.

As a comprehensive solution, NG is the most feasible option to replace HFO or marine diesel oil in shipping, showing significant pollutant and GHG emissions reductions. However, the CO₂ emission reduction potential is limited by the carbon content in NG. Nonfossil fuels still represent the best solution for pollution reduction.

Sensitivity analysis

As well as identifying the main scope of all processes included in the TLCEM, establishing each EF is the most critical step in calculating total life cycle emissions. The emissions from ship operation account for the largest proportion of emissions during the 25-yr life cycle and the ME produces the highest emissions during operation. EFs differ between types of diesel or DF engine from the same manufacturer. Various operational environments or conditions may also affect the emissions; however, engine load is a key factor for high-powered two-stroke MEs.

Although the NO_x certification values of the electronically controlled two-stroke MEs of the 10 containerships comply with MARPOL requirements, all the EFs exceed 17.5 g/kWh at 25% load, according to the results of engine tests. This unexpected value is approximately 10% higher than those adopted by GREET, TEAM, and IMO. Following the NO_x emissions profile of the engine load, EF is approximately 19.59 g/kWh when ME is at 15% load (e.g., maneuvering in port). Large-scale containerships always operate at the slowest possible speed to minimize fuel expenditure. To analyze actual NO_x emissions, total NO_x emissions are calculated based on the EFs obtained from engine emission tests rather than certification values.

To understand the impact on our results due to the differences in EF, additional sensitivity analyses were performed considering four scenarios, where IMO EFs baseline were adopted in either (1) diesel ME in HFO mode, (2) dual fuel ME in NG mode, (3) diesel AE in HFO mode, or (4) dual fuel AE in NG mode. In the TLCEM, the total emissions are linearly proportional to the emission factors. Therefore, the differences in total emissions can be derived from the differences of emission factors, as explained in Figures 5 to 8.

Figure 5. Differing EFs of diesel ME in HFO mode.

The PM₁₀ and N₂O EFs of diesel ME in the HFO mode are the same as the baseline adopted by the IMO (Figure 5). The deviations observed for CO (233%) and CH₄ are rather high, but EF values are relatively small, at 1.8 g/kWh and 0.006 g/kWh, respectively, and represent only a small portion (0.25% and 0.47%) of total pollutant and GHG emissions, similar to the observations of VOC and SO_x. Therefore, the effects of change in the EF of these pollutants on total emissions are limited. CO₂ comprises the majority of total emissions and has an EF that is 8.6% lower than that adopted by the IMO. Total life cycle emissions in CO₂ are thus expected to increase by 6.6% approximately, if IMO EF baseline is used instead in the calculation. CO₂ emissions depend on fuel oil consumption; thus, once total fuel oil consumption data are obtained, the exact CO₂ emissions can be correctly estimated.

The EFs of a two-stroke main DF engine in the NG mode are compared with IMO data in Figure 6. The EFs of VOC, SO_x, N₂O, and CO₂ are very similar to those of the IMO.The IMO adopts EFs for low-pressure NG systems (system pressure < 10 kg/cm²), calculated using Otto cycle DF engine emissions data from Wärtsilä, which differ from the high-horsepower, high-pressure NG systems of the MAN GI series. The other EFs of CO, NO_x, and CH₄ show clear discrepancies; however, the three EFs in this mode are obtained from the test bed.The PM₁₀ results show a large discrepancy; however, the very small EF of 0.0014 g/kWh meansthat it has a negligible effect (less than 0.003%, as shown in Table 10) on total emissions.

Figure 6. Differing EFs of dual fuel ME in NG mode.

When the diesel AE runs in the HFO mode, the EFs of VOC, PM₁₀, SO_x, and CO₂ are consistent with or almost identical to those adopted by the IMO (Figure 7). The deviations in CO, N₂O, and CH₄ EFs have a negligible effect on total emissions because of the very low EFs, which are 1.100, 0.031, and 0.004 g/kWh, respectively (Table 7).The IMO EFs of a four-stroke auxiliary DF engine inthe NG mode are the same as those of a two-stroke diesel ME. Figure 8 shows large deviations for CO, NO_x, PM₁₀, and CH₄, but the lower EFs of these pollutants and total emissions result in an insignificant overall influence.

Figure 7. Differing EFs of diesel AE in HFO mode.

Figure 8. Differing EFs of dual fuel AE in NG mode.

Conclusion

This study developed a mathematical model, TLCEM, to accurately estimate and analyze the 25-yr total life cycle emissions of 10 Taiwan-built post-Panamax sister containerships. The 25-yr life cycle, from cradle to grave, consists of steel production, shipbuilding, crude oil extraction and transportation, fuel refining, bunkering, and ship operation. Total life cycle emissions and percentages of VOC, CO, NO_x, PM₁₀, SO_x, CH₄, N₂O, and CO₂ in each process were calculated and analyzed for ships that use HFO. CO₂ accounts for the majority of total emissions (95.37%), followed by NO_x(2.25%) and SO_x (1.30%), respectively. Emissions of other pollutants account for the remaining 1.08% of total emissions. The majority of emissions occur during ship operation and originate from the ME.

By switching entirely to LNG as a marine fuel, SO_x and PM₁₀ emissions can be significantly reduced by 98.6% and 99.6%, respectively. The emissions of other GHGs/pollutants can also be reduced to a great extent; the reduction in CO₂ emissions amounts to 0.61Mt. In combination, CO₂ and NO_x account for the majority (99.35%) of total emissions during the 25-yr life cycle. The EF of NO_x is higher in the NG mode than in the HFO mode because running the DF engines at low loads causes higher NO_xemissions. This finding should be given special consideration for operating high-powered ships under long-term slow steaming conditions.

Defining the EF of each pollutant and GHG is critical for each emission process that contributes to the total life cycle emissions. For a more accurate estimation of NO_x emissions in this study, the EFs for NO_x as well as CO and CH₄ in the NG mode were obtained from engine tests conducted under different loads rather than from certification values. The EFs for the operation were compared with the values published by the IMO. The results show small differences and negligible impacts on emissions. Consequently, this approach may contribute to more accurate estimates of total life cycle emissions from global shipping. However, to confirm the EFs of every engine, mandatory measurement of emissions of all critical pollutants and GHGs is recommended during the testing of new engines.The results from this paper are important to the analysis for ship operational emissions. The operational emissions contribute the most to total emission, mainly CO₂, followed by NO_x and SO_x. Therefore, the abatement can be started with the operation process. Although using NG can result in major reduction in SO_x, the increases in CH₄ and NO_x need to be analyzed through further investigations.

Limitations and suggestions for further researches

Pollution emissions from ship dismantling and recycling process

The research in ship scrapping industry is immature and has very limited literature. We consider emissions from the process of carbon steel recycling as part of emissions from steel manufacture. The emission factors are shown in Table 1.

Data on pollutant and GHGs emissions from the ship dismantling and recycling industry are incomplete and are, therefore, excluded from the total life cycle analysis. Deshpande et al. (2013) found that India has the largest ship dismantling and recycling industry, scrapping more than 350 ships annually (approximately 27 Mt of steel) and accounting for 47% of the sector worldwide. Emissions from dismantling andrecycling processes are mainly from steel cutting (emitting an average of 21.77 kg CO₂ per 1 km of 1 mm-thick steel plate). Post-Panamax containerships are approximately 330 m long, 43 m wide, and 25 m deep. Assuming the average thickness of steel structures is 25 mm below the main deck and 8 mm in the accommodation deck,and the cutting length is approximately 200 km, then 105 t of CO₂ is emitted during the cutting process. These CO₂ emissions are relatively modest and do not affect the overall results of the total life cycle emissions analysis. However, this is a very coarse estimate of the dismantling process, and processes such as burning paint, oil, or cables that might emit more serious toxic pollutants during the complete recycling processes should also be considered in future studies.

Acknowledgment

The authors thank the classification societies and shipping companies in Taiwan for providing ship navigation and technical data.

Additional information

Notes on contributors

Jian Hua

Jian Hua is associate professor, Department of Marine Engineering, National Taiwan Ocean University.

Chih-Wen Cheng

Chih-Wen Cheng is chief surveyor, CR Classification Society.

Daw-Shang Hwang

Daw-Shang Hwang is associate professor, Department of Marine Engineering, National Taiwan Ocean University.