GIS-based analysis of the spatial distribution of CO₂ emissions and slow steaming effectiveness in coastal shipping

Abstract

The sustainability of maritime activities is increasingly gaining interest, with the shipping sector actively focusing on decarbonization efforts. Throughout the years, researchers have considered slow steaming for improving the environmental footprint of maritime networks. In order to assess such strategies’ effectiveness on existing emissions, research also focuses on the accurate estimation of emission inventories. However, there is a significant gap concerning both fields when considering short-sea shipping, especially passenger shipping. Furthermore, while emissions are characterized by spatial aspects in several studies, there is an apparent gap in considering such aspects for detailed analysis purposes rather than only for visualization purposes. In this study, the Greek Coastal Shipping Network (GCSN) is considered, with its emissions estimated using a top-down method, creating a spatial emission inventory used for further spatial analysis for accurate identification of highly polluted areas. Results indicate that ship emissions do not spread homogeneously throughout the GCSN and that targeted interventions are necessary in several areas of the network. The effectiveness of spatially related slow steaming implementations is evaluated and compared with their implementation on the whole network. The study highlights the need for additional future emission mitigation strategies, such as service optimization, network restructuring, continuous emission monitoring, and fleet renewal with more environmentally efficient ships. The study’s aim is to fill the research gap regarding the environmental assessment of passenger shipping and the effects of slow steaming on such networks while presenting an adaptable GIS-based decision support system for enhanced decision-making regarding the environmental efficiency of maritime networks.

Introduction

The planning of maritime services focuses on efficient fleet utilization and the provision of an acceptable level of service so that profitability is achieved. However, the externalities of maritime operations, especially environmental impacts, are often overlooked by planners and operators. Greenhouse gas (GHG) emissions from transportation have doubled since 1970, while in comparison to other modes of transport, GHG emissions in the shipping industry are projected to rise by 50–250% until 2050 due to the sector’s continuous expansion (Smith et al. 2014; Zisi et al. 2021). Heightened environmental concerns have prompted the International Maritime Organization (IMO), other marine institutions, and the research community to recommend various approaches for the gradual phasing out of fossil fuels in the shipping sector. Several studies have investigated the environmental footprint of maritime operations and proposed mitigation actions, such as the use of cleaner fuels and slow steaming (Crist 2009; Eide et al. 2009; Franc and Sutto 2014; Rehmatulla et al. 2017; Xing et al. 2020; Zisi et al. 2021; Ju and Hargreaves 2021). Slow steaming in particular is the reduction of a vessel’s operating speed, which, apart from cutting operating costs, can considerably reduce emissions (Cariou 2011; Woo and Moon 2014; Yin et al. 2014; Pastra et al. 2021; Degiuli et al. 2021), especially for high-speed vessels (Psaraftis et al. 2009).

The Greek Coastal Shipping Network (GCSN) offers connections between Greek mainland ports and fifty-five inhabited Aegean islands (Lekakou and Remoundos 2015), carrying about 12 million travelers and 2.5 million vehicles annually in 2019 (Hellenic Statistical Authority - EL.STAT 2019). The GCSN consists of about 115 ports, 89 of which are exclusively connected by sea, with no active airline connections at the port’s island, and is operated by a total of 97 vessels (S.E.E.N. 2022). As the Aegean islands are popular tourist destinations, the GCSN demand exhibits high seasonality: the summer peak period (June–September) accounts for about 45% of the total annual demand for ferry transportation (Hellenic Statistical Authority - EL.STAT 2019). Both conventional and high-speed catamaran ferries are operating in the GCSN, with the latter servicing routes mainly during the summer peak period. The performance and operations of the GCSN have been widely investigated in the past (Goulielmos 1998; Sambracos 2001; Kapros and Panou 2007; Lekakou 2007; Schinas 2009; Lekakou and Vitsounis 2011; Mitropoulos et al. 2022), with the literature indicating the paramount importance of the GCSN to the economic growth and social cohesion of the Aegean islands with the mainland and between them, therefore retaining the country’s overall cohesion and accessibility to all areas at an adequate level of service. The latter practically translates to acceptable travel times during both peak and off-peak seasons. Especially during summer peak periods, fast access to the Aegean islands is crucial for accommodating high tourist flows and providing supplies to tourist facilities while also improving the islands’ attractiveness as tourist destinations and, as a result, their economic viability and growth. This is mostly achieved by increasing the speeds of conventional ferries and by routing high-speed (catamaran) ferries, while also increasing the overall frequency of trips during high-traffic periods in the summer. Although high-speed ships provide an increased level of service and limit travel times as much as possible in order to meet passengers’ demands, they unavoidably worsen the GCSN’s environmental footprint as greenhouse gas (GHG) emissions and pollutants increase in an otherwise environmentally sensitive archipelago (Bigano and Sheehan 2011; Iliopoulou et al. 2018). GHG emissions have also been characterized in several cases by their spatial aspect and distribution in specific areas. Spatial data visualization, mapping techniques, and the overall use of GIS software have become popular in the last decades in cases where it is highly important to analyze and visualize specific spatial characteristics, as is the case in environmental externalities. Several studies have assessed the impact of certain techniques for mapping data clusters, such as hot spot analyses (Grubesic and Murray 2001; Prasannakumar et al. 2011; Said et al. 2017), and how different cluster mapping techniques work (Jana and Sar 2016; Sánchez-Martín et al. 2019). Such methods are also widely utilized in the transportation literature (Iliopoulou et al. 2020), especially in the analysis of traffic accidents and other risk-related data sets (Prasannakumar et al. 2011; Truong and Somenahalli 2011; Aghajani et al. 2017; Colak et al. 2018). However, to the authors’ knowledge, the application of such methods to maritime emissions’ analysis has remained at a superficial level, limited to visualization, without a clear link to decision-making.

Τhis study aims to address the relevant literature gap by providing concise and easily adaptable methodological frameworks for the assessment of the spatial distribution of CO₂ emissions and the implementation of slow steaming strategies for emission mitigation, considering the resulted distinct areas where targeted interventions are necessary. Additionally, the study also considers slow steaming, even in passenger shipping networks, due to the fact that the structure of several such networks, like the GCSN, leaves little room for changes. This is mostly apparent in the case of the network considered due to policies applied by the Greek state, decreased flexibility from operators, and accessibility needs (Katarelos and Koufodontis 2011; Karampela et al. 2014), which can exist in several other passenger shipping networks as well.

The remainder of the paper is structured as follows: the next section presents the literature review on emission monitoring studies through spatial analysis techniques and slow steaming strategies. Next, the proposed methodological framework is presented for the identification of high-polluted GHG areas, followed by the implementation of slow steaming strategies in different scenarios. The study concludes with the results of all implemented scenarios, while the discussion section highlights the main findings for the case of the GCSN, with paths for future studies presented and discussed.

Literature review

Over the past several years, a variety of strategies have been employed within the shipping sector to adhere to the stringent environmental regulations imposed by the International Maritime Organization (IMO). These methods, inclusive of slow steaming, route optimization, and hull fouling management, reflect an industry-wide attempt to diminish the environmental footprint associated with maritime transport. Nevertheless, it is important to note that these strategies have only resulted in a relatively modest decrease in emissions, approximating 10–15%, as reported by Ammar (2018) and Cullinane and Cullinane (2013).

The IMO continues to pledge its commitment to reducing greenhouse gas (GHG) emissions in accordance with the preliminary actions instigated during the 72nd Marine Environment Protection Committee (MEPC 72). This commitment is projected to culminate in a gradual limitation on the usage of fossil marine fuels by the year 2050, as anticipated by Raucci et al. (2017). Reasonably, this has led to increasing interest among researchers, with several studies investigating emission mitigation in the maritime section. This section discusses emission inventory studies and studies investigating slow steaming strategies and subsequently outlines the contribution of this study.

Emission inventories

The issue of air pollution stemming from maritime transport has been subject to increasing scholarly interest over the past few years, with researchers such as Aksoyoglu et al. (2016), Russo et al. (2018), and Zhu et al. (2022) making notable contributions in the field. Additionally, the relevance of spatial data analytics-based emission inventories has also been recognized as instrumental in efficiently overseeing the externalities arising from maritime transport (Johansson et al. 2017; Ding et al. 2018; Okada 2019; Topic et al. 2021).

Considering emission studies incorporating spatial analysis methods, the importance of spatial data analytics and their use can provide new approaches to the implementation of emission mitigation strategies. However, a main issue highlighted by our study is the scarcity of studies in the existing literature that focus explicitly on the emissions resulting from domestic shipping through the integration of spatial data analysis methods, with only a handful of recent studies, most notably those of Buber et al. (2020) and Toz et al. (2021), providing an in-depth examination of this particular area.

Emission inventory studies have deployed a variety of techniques for estimating emissions, with the two main approaches being the fuel-based approach (top-down) and the activity-based approach (bottom-up). In the activity-based approach, it is possible to quantify emissions produced by each vessel and aggregate results for the calculation of emissions generated by different fleets, albeit at the cost of needing several detailed data sets on the technical and operational characteristics of each vessel under study (Nunes et al. 2017), and it is generally widely adopted in the maritime sector. On the other hand, top-down approaches generally rely on macro-level statistical data for the calculation of emissions, through the fuel-based approach, by considering fuel consumption data and their relevant emission factors (Eide et al. 2009; Tzannatos 2010). However, such approaches largely rely on available statistical data and their overall validity, which in many cases is characterized by a certain level of uncertainty. This approach is used in our study as well, as it is highly useful in cases where ship movement activity data are not available or do not constitute the main focus of the study. In this particular study, as emissions are only utilized for benchmarking reasons, actual reported data for fuel consumption is used based on the available data from the EMSA/MRV-THETIS database in order to estimate shipping emissions.

The validity of the data used was also thoroughly discussed in a recent study by Doundoulakis and Papaefthimiou (2022). In their study, they employed reported real data for actual fuel consumption from the EMSA/MRV-THETIS database to evaluate the validity of estimated results through a comparative analysis between estimated and actual fuel consumption and CO₂ emissions. Due to the requirements of EU-MRV regulation, and as ship owners report data including fuel consumption and CO₂ emissions, a bottom-up methodology approach was discussed, where the fuel-energy consumption and air emissions of ships were calculated, with results compared to the reported emission data from the EMSA/MRV-THETIS database. The study concluded by highlighting the significance of the EMSA/MRV-THETIS database in providing actual fuel consumption and CO₂ emission data from ships, which is crucial for researchers involved in the calculation of air emissions in shipping. Specifically, comparing results regarding fuel consumption and CO₂ emissions of ships for the year 2020 showed a difference in total fuel consumption of about 6.14–12.07% and a difference of 5.66–11.85% regarding total CO₂ emissions. While these differences are directly related to the nature of each study and the necessary accuracy in emission calculations, in this study they are considered acceptable due to the fact that they serve the distinct purpose of providing a benchmark for the evaluation of the results of different slow steaming strategies and whether these could be sufficient in themselves to reduce the environmental footprint of the network under study.

Russo et al. (2018) have reviewed and assessed the validity of available emission inventories and the accuracy of their spatial representations. In their study, they have highlighted that while overall emission values did not show significant differences between the analyzed inventories, the spatial representations of each one showed highly different emission estimations and visualizations. As a result, while such emission inventories could be considered in terms of an overall assessment and quantification of emissions, it is apparent that policymakers would not benefit from the use of available spatial data and visualizations due to inaccuracies, which would lead to poor decision-making. As the spatial dimension is crucial to effective environmental management, relevant spatial approaches in decision-making in terms of policy regulations are also important. This was also highlighted by Aksoyoglu et al. (2016) through a critical examination of the impact of marine transport emissions, particularly sulfur, on air pollution levels. Their study showed that even with recent regulatory changes, emissions, especially outside the designated sulfur emission control areas (SECAs) and from other non-regulated components, have been on an upward trend over the past two decades. Key findings include significant increases in particulate sulfate in the Mediterranean, the English Channel, and the North Sea and increases in particulate nitrate in areas with high land-based NH₃ emissions like the Benelux region. The study highlighted the need for better regulation of maritime emissions and spatially related policy regulations through NOx emission control areas (NECAs), as predicted increases in maritime emissions could offset the benefits gained from reductions in land-based emissions.

The study of Czermański et al. (2021) provides an in-depth examination of container shipping emissions, their reduction, and their overall impact on the environment. Container shipping, being the most significant emission producer within the maritime shipping industry, has necessitated various measures to curtail its emission levels, as highlighted in the researchers’ comprehensive study. Importantly, their approach estimates ongoing emission reductions continuously, effectively filling data gaps, as the most recent global container shipping emission records were from 2015. One of the main contributions of the study was the fact that the spatial dimension of emission inventory assessment was discussed, with potential areas for significant emission reduction being highlighted. In addition, the study suggests a detailed consideration of ship speed in emission estimates, as emission levels are highest at sea. The energy consumption approach presented offers a way to accurately estimate current emission levels and potential reduction areas on a more spatially associated basis, supplemented with early-stage detection and continuous monitoring of environmental impacts. Policymakers in the shipping industry could also greatly benefit in terms of decision-making through targeted, spatially related interventions, considering that spatially associated environmental monitoring has been thoroughly discussed by researchers in other sectors (Froemelt et al. 2021). In the broad field of maritime transport, spatial analysis methods, and specifically spatial pattern identification techniques, have also been successfully implemented to assess CO₂ emissions considering port container distribution facilities, yielding valuable results (Wang et al. 2020). Additionally, the performance of spatial analysis methods for the estimation of CO₂ emissions in a wider geographic region has also been highlighted for effective, spatially related emission inventories and environmental management (Uddin and Czajkowski 2022).

Slow steaming in passenger shipping

Considering slow steaming, the maritime sector has been increasingly recognizing its potential for effectively mitigating GHG emissions. While such strategies have been widely studied in terms of liner shipping, especially in the case of containerships (Cariou 2011; Woo and Moon 2014), they still gain attention as a potential long-term solution for the decarbonization of shipping. However, the effectiveness of slow steaming approaches is inherently tied to the volatility of bunker prices, as it is mostly considered a strategy to minimize fuel consumption, thus resulting in substantial cost savings in addition to reductions in GHG emissions. As highlighted by Cariou (2011), in order for such strategies to be sustainable in the long term, bunker prices should remain within a break-even threshold of $350–$400. In the context of this study, it is important to note that the bunker prices for most conventional fuels still employed in the shipping sector remain above the aforementioned break-even price point, as reported by World Bunker Prices (Ship and Bunker 2023).

While the literature is rich in terms of slow steaming applications on liner shipping, with a focus on containerships dominating the interest of researchers, there is an escalating interest to broaden such applications and to evaluate their effectiveness on more localized or regional scales, as opposed to just the global level typical of containership trade. However, there is still an apparent limitation when considering the literature concerning the evaluation of such strategies in the context of domestic shipping and fast ships. For instance, the adoption and compliance of RoRo and RoPax shipping companies with respect to slow steaming measures have been studied (Raza et al. 2019), while studies have also focused on the overall impact of slow steaming at a more regional level, such as in the Mediterranean (Degiuli et al. 2021).

In cases of specifically fast ships operating domestically, mostly servicing passenger shipping, the works of Psaraftis and Kontovas (Psaraftis et al. 2009; Psaraftis and Kontovas 2009) are the most notable ones. In their studies, they highlighted that while high-speed vessels yield higher emissions per tonne-kilometer, reducing their speed could have significant implications for the shipping industry, such as the need for more ships and increased cargo inventory costs. Therefore, while speed reduction appears to be a straightforward method for decreasing emissions, its side effects can lead to non-trivial costs, potentially making it less cost-effective. In addition, it is also emphasized that speed reduction’s effectiveness strongly depends on the possibility of reducing port time, highlighting the critical role ports play within the intermodal supply chain. Psaraftis and Kontovas (2009) also provided an in-depth analysis of CO₂ emissions from the world’s commercial fleet, utilizing data from the 2007 Lloyds-Fairplay world ship database. Different ship types, such as bulk carriers, crude oil tankers, container vessels, and others, were considered, and emissions were estimated both in terms of grams of CO₂ per tonne-kilometer and total CO₂ produced annually, with the study providing a comprehensive emission inventory, which is crucial for researchers and policymakers in implementing relevant environmental regulations. However, as is the case in our study, the authors recognized that the study’s findings are influenced by the quality and availability of the data used, with more precise results requiring additional data, including worldwide ship movements and exact bunker consumption figures for the global fleet. In addition, the authors highlight the need for understanding the relative impacts of different ship types and sizes on CO₂ emissions, which is valuable for prioritizing measures to reduce greenhouse gas emissions in shipping, as addressing GHG emissions from shipping is a complex task with substantial differences in opinions on how to proceed, indicating the need for collaborative approaches.

Taking these into consideration, it is becoming clear why, throughout the years, there has been increasing interest in containerships and overall liner shipping but also a significant gap considering the slow steaming of passenger shipping. In addition, as clearly stated in the most recent study of Vakili et al. (2023), slow steaming could prove practically unfeasible for small passenger ships or mixed passenger/cargo ships, such as RoRo and RoPax ferries, as these are the only viable transportation options for the accessibility needs and overall connectivity of small island regions, such as the ones in the Aegean Sea. To the authors’ knowledge, at least up until this point, only the studies of Psaraftis et al. (2009) and Zis et al. (2020) have thoroughly assessed the adoption of slow steaming in the case of ferries, mainly due to the inelastic nature of the short-sea shipping sector, with speed reduction strategies not considered an option by RoRo and RoPax shipping companies (Raza et al. 2019).

Contribution of the study

As the review of relevant studies demonstrates, while the literature regarding the environmental impact of the maritime industry is extensive, the environmental externalities of domestic passenger shipping have not been as thoroughly examined. While there are several studies assessing emission inventories and monitoring shipping environmental externalities, the study identifies two distinct gaps in the literature. First, considering emission inventory studies, there are few that have delved deeper into their spatial context by addressing the spatial distribution of GHG emissions and the areas where these originate from. While these studies have discussed the importance of the spatial aspect of emissions, their use was limited to data visualization purposes. In addition, only a few studies have incorporated spatial analysis methods in their studies for the assessment of emissions’ specific spatial patterns, although without further exploiting their results in proposing relevant spatially related mitigation strategies based on their outputs. Secondly, while slow steaming strategies have been widely discussed in the maritime sector, there is a scarcity of studies assessing such strategies’ potential regarding passenger and domestic shipping.

To better highlight the contribution of our study, Table 1 shows the most relevant studies addressing the spatial dimension of emissions.

Table 1 Relevant studies highlighting the spatial aspect of emission inventories

As shown in Table 1, although several studies have included the spatial dimension of emissions, predominantly for visualization purposes, only a few have exploited spatial data analyses in their methods to facilitate decision-making, not only to the extent of assessing calculated emissions and their spatial distribution but also to extend these results in the context of implementing emission mitigation strategies through effective spatially related decision-making. Moreover, while slow steaming has also been thoroughly examined in the existing literature, only a handful of studies have assessed its effectiveness when applied to passenger shipping. In addition to this, and to the authors’ knowledge, up until this point, no study has proposed the implementation of spatially related slow steaming when considering highly polluted areas, apart from cases where emission control areas (ECAs) exist. Recognizing these research gaps, this study intends to address these unexplored domains in the existing literature. The study’s first goal is to highlight the fact that emission inventories can enhance decision-making and policy regulations by considering the spatial dimension of maritime transport, not only to pinpoint areas of high-GHG emissions but also to provide policymakers with specific options to limit shipping’s environmental footprint in areas where this is most needed. Although more interventive emission mitigation strategies are feasible in coastal shipping with the exploitation of such networks’ topological characteristics and their restructuring (Karountzos et al. 2023), their implementation still requires significant policy-related changes. Therefore, considering the versatility of slow steaming for immediate application in addition to the scarcity of studies in terms of its implementation in passenger shipping, the study aims to assess both its feasibility and effectiveness in a specific sector where it has not been thoroughly discussed while accounting for the results of continuous emission monitoring through a spatial decision support system.

Research methodology

The methodology adopted involves the identification of high-polluted CO₂ areas and the implementation of slow steaming in such areas based on the results of the spatial analysis methods. For the implementation of the methodological framework, the Greek domestic shipping network in the Aegean Sea is considered, where slow steaming can be implemented to improve its environmental footprint, with a focus on CO₂ emissions. Based on that identification of high-polluted areas, alternative scenarios of fleet slow steaming are implemented and compared using spatial analysis tools in a GIS-based environment.

As mentioned previously, the operating model and structure of the GCSN leave little room for operational changes. In addition, a main issue with the implementation of such strategies is the fact that speed reductions could potentially lead to a deterioration of service, especially considering passenger shipping. Nevertheless, slow steaming of the fleet could be considered a transitional solution, at least until new, more environmentally friendly, or even zero-emission ship technologies are introduced to the network, ultimately leading to a much-needed renewal of the operating fleet. As this study highlights whether slow steaming could be implemented on passenger shipping networks, where accessibility and travel times are highly important, it is acknowledged that reducing the fleet’s speed on the whole network would not pose a viable solution, as the network’s offered level of service would highly deteriorate. Therefore, our study aims to provide a different methodological approach to limiting the environmental footprint of such networks by identifying the most sensitive areas in terms of environmental deterioration for more targeted interventions. This approach not only provides a better understanding of the spatial dynamics of a network in terms of its environmental impact on specific areas by increasing the identification accuracy of high-polluted areas, but also helps policymakers in decision-making processes regarding environmental policy management and emission mitigation strategies. In addition, implementing slow steaming strategies through targeted interventions in environmentally deteriorated areas will not only limit total emissions but also affect travel times less, as such strategies will not be applied on the entirety of an examined ship route, thus avoiding a significant compromise regarding the offered level of service. For these purposes, spatial analysis techniques are applied in a GIS environment for the identification of high-GHG areas and the implementation of slow steaming strategies. Figure 1 shows the proposed workflow for the study.

Fig. 1

Workflow of the proposed methodological framework of the study

Emission inventory methodology

Reported maritime CO₂ emissions from the European Union Monitoring, Reporting, and Verification System (EU-MRV) (European Maritime Safety Agency 2019) are used for the purposes of this study. In accordance with EU Regulation 2015/757, reported CO₂ emissions are used in the case of our study in order to avoid statistical variances in model-based estimations. However, one of the major drawbacks of EU-MRV data is the fact that, based on EU Regulation 2015/757, CO₂ emission reports only apply to vessels with a gross tonnage above 5000. In the case of this study, not all vessels operating in the Aegean Sea are above the aforementioned limit of 5000 GT in size, and their respective operators and shipping companies are not obliged to report either CO₂ emissions or fuel consumption. In such a case and for the purposes of this study, when a route is operated by a vessel with not-reported CO₂ emissions, this route is assumed to be operated by a vessel of the same type with similar capacity and engine characteristics, for which such data exists. All available emission data are then added to a GIS database and incorporated into their respective main routes of the GCSN as linear spatial features, as shown in Fig. 2. Both ship and route data are analyzed up until 2019, in order to avoid operational changes affected by several disruptions due to the COVID-19 pandemic. Consequently, for the purposes of the applied methodology, 2019 routes are used as the base scenario of the GCSN under normal operations, as shown below.

Fig. 2

Routes of the Greek Coastal Shipping Network as of 2019 (data collection from shipping companies and ferry planning services as of 2019)

At this point, it is worth noting that the goal of this study is not to highlight the estimation of fuel consumption and emissions or propose related model-based estimations, but rather to propose a methodological framework that facilitates the better implementation of emission mitigation strategies through targeted interventions, thus contributing to the decision-making process through spatially related assessments. As a result, the reported fuel consumption and emission data used in this study are only utilized as benchmarks for the evaluation of different implementation levels of the proposed strategies. By taking this into consideration, the proposed methodology can therefore be applied to any given scenario and any network under analysis while also considering the variability of inputs given by the users, either as a result of model-based top-down or bottom-up approaches or through reported and statistical data. Therefore, the proposed workflow is also characterized by its wide applicability and adaptability to different case studies.

Through the use of the actual reported EU-MRV data, which concern fuel consumption, time at sea, and CO₂ emissions, it is possible to calculate a ship’s average operating speed on an annual basis, thus taking into consideration a vessel’s actual speed rather than its maximum or design speed, resulting in better accuracy of the output results. This is also important as updated fuel consumption and CO₂ emissions can be estimated in cases where slow steaming is proposed. Nevertheless, even though speed reduction on all routes of the GCSN would be an option, it would also cause scheduling and travel time issues. As a result, through the use of GIS software and spatial analysis methods, it is crucial to identify routes where there are high concentrations of CO₂ emissions. For the calculation of CO₂ emissions, the average fuel consumption of each ship per nautical mile traveled is utilized, with an emission factor of 3.114 considering heavy fuel oil used by the ships of the study, which is in conjunction with reported CO₂ data from the THETIS EU-MRV database. As a result, CO₂ emissions for each trip and per distance are calculated as follows:

$${CO}_2= FC\ast {F}_{CO_2}$$

(1)

where CO₂ denotes CO₂ emissions in metric tonnes, FC the ship’s fuel consumption, and ${F}_{CO_2}$ the CO₂ emission factor for heavy fuel oil, which is equal to 3.114.

Fuel consumption and emission calculations are crucial in creating the relative emission inventory of the study and associating the relative data with each route for the creation of the needed geodatabases used in the study. By utilizing the created geodatabases and their respective spatial features through the proposed methodological framework, an initial assessment of whether there is a certain spatial pattern of CO₂ emissions in the Aegean Sea is conducted.

GIS-based spatial pattern analysis

Based on the current structure of the GCSN, by utilizing two widely applied methods of analyzing patterns and mapping statistically significant clusters, it is possible to answer both if specific spatial patterns exist and where specific areas of interest are located. In this study, two widely used spatial autocorrelation regression models are used, which consist of Moran’s I Spatial Autocorrelation, followed by a cluster and outlier analysis (Anselin Local Moran’s I), and high/low clustering (Getis-Ord General G), followed by a hot spot analysis (Getis-Ord Gi*). As a result, the proposed methodological framework for emission spatial distribution mapping is based on the utilization of Local Indicators of Spatial Association (LISA) in both methods (Anselin 2010). Global statistics, such as the Spatial Autocorrelation employing Global Moran’s I, are most effective when the spatial pattern remains consistent across a study area, while local statistics, such as the Getis-Ord Gi*, assess each feature within the context of neighboring features and compare the local situations to the global ones. It is worth noting that hot spot analysis utilizing the Getis-Ord Gi* statistic is highly efficient in cases where clustering is expected, either low or high, as is the case in emission monitoring where high-polluted areas are evaluated. In addition, a Local Moran’s I statistic, while identifying high clustering, also identifies cases where outliers may exist with a combination of both high and low values in a dataset. As a result, it is proposed by this study that both approaches should, when possible, be exploited in combination, as they can provide a more holistic understanding of the spatial patterns in a given study area. The mathematical formulation, regarding the calculation of the Moran’s I statistic for spatial autocorrelation, is given as:

$$I=\frac{n}{S_0}\frac{\sum_{i=1}^n\sum_{j=1}^n{w}_{i,j}{z}_i{z}_j}{\sum_{i=1}^n{z}_i^2}$$

(2)

where z_i is the deviation of an attribute for feature i from its mean (x_i - x̄), w_i,j is the spatial weight between features i and j, n is equal to the total number of features, and S₀ is the aggregate of all spatial weights, given by:

$${S}_0=\sum_{i=1}^n\sum_{j=1}^n{w}_{i,j}$$

(3)

The z_I score for the statistic is computed as:

$${z}_I=\frac{I-E\left[I\right]}{\sqrt{V\left[I\right]}}$$

(4)

where E[I] =  − 1/(n − 1) and V[I] = E[I²] − E[I]².

On the other hand, the second method determines the Getis-Ord local statistic, which is calculated as:

$${G}_i^{\ast }=\frac{\sum_{j=1}^n{w}_{i,j}{x}_j-\overline{\textrm{X}}\sum_{j=1}^n{w}_{i,j}}{S\sqrt{\frac{\left[n\sum_{j=1}^n{w}_{i,j}^2-{\left(\sum_{j=1}^n{w}_{i,j}\right)}^2\right]}{n-1}}}$$

(5)

where x_j is the attribute value for feature j, w_i,j is the spatial weight between feature i and j, n is equal to the total number of features, and:

$$\overline{X}=\frac{\sum_{j=1}^n{x}_j}{n}$$

(6)

$$S=\sqrt{\frac{\sum_{j=1}^n{x}_j^2}{n}-{\left(\overline{X}\right)}^2}$$

(7)

The ${G}_i^{\ast }$ statistic is already a z-score, so no further calculation for a z-index is required. On both instances, (2) and (5) are used to calculate Moran’s I and hot spot analysis indices, respectively, while (3), (4), (6), and (7) describe each variable used on both processes to spatially assess the statistical significance of examined features. Spatial autocorrelation models like the two used in this study are crucial in identifying highly polluted areas where high concentrations of CO₂ emissions are observed.

To better identify areas that show higher concentrations, the study area of the GCSN, which is the Aegean Sea, is divided into a 10 km × 10 km grid, covering all routes of the study under analysis. In this case, each route analyzed is divided into distinct parts contained in each 10 × 10 quadrant, which will then be analyzed as a spatial feature regarding CO₂ emissions. As a result, each quadrant, or grid division, acts as a distinct polygon feature, with all features and their respective spatial relationships considered for the calculation of LISA statistics, as described. However, initial emission calculations were conducted on a route basis, based on the distinct characteristics of each route, by taking advantage of the average fuel consumption per nautical mile and average CO₂ emissions per nautical mile included in the geodatabase for each route. In order to calculate CO₂ emissions for each polygon, the routes under study must be split into distinct parts included in each polygon. Then, by using the average fuel consumption and emission data for each route part in addition to the included spatial information (i.e., each part’s length in nautical miles), the respective ship fuel consumption and generated emissions are calculated for each part included in each spatial feature. Summary statistics are then used for each polygon feature by aggregating the calculated fuel consumption and emission data from each distinct route part, thus leading to the total fuel consumption and CO₂ emissions generated for each polygon area. Following the preparation of the relative data, both methods of spatial correlation are used to identify spatial features (i.e., polygon areas) forming high-concentration clusters, with CO₂ emissions used as the main variable.

Slow steaming approach

As this study concentrates on slow steaming to reduce CO₂ emissions from ships operating on said routes, a possible decrease in operating speed ranging from 5 to 20% is considered. A 20% or larger decrease could prove unfeasible due to the fact that it is necessary for the overall quality of service to be retained at acceptable levels, which translates to acceptable travel times for passengers in the case of passenger shipping. However, the only feasible way to maintain travel times and, therefore, the overall level of service offered for passengers and their trips is to reduce port stop times as much as possible, if there are no alterations to network structure. In the case of passenger shipping, especially regarding the GCSN, this would be highly challenging, as port service times between stops amount to just a portion of the time that a passenger ship spends at sea, specifically for boarding and disembarkation of cars and passengers, which in most cases could be between 5 and 25 min, depending mostly on each port’s infrastructure and passenger flows. As a result, a reduction of port stop times could be explored, but it would be possible only in cases where higher delays are shown at port stops or by exploring different operating systems for the network with an overall redesign and optimization of the network. Nevertheless, just for the purposes of this study, as the network’s structure remains the same, the maximum decrease in speed that will be explored will be 20%, either when implemented on the whole network or on selected areas of high CO₂ concentrations, based on the results of the spatial analysis models used.

For the implementation of slow steaming strategies, new fuel consumption and CO₂ emissions are estimated based on the methodological framework presented by Psaraftis et al. (2009). This specific methodological framework is considered by our study, as the case here is the same as the aforementioned study, where speed reductions are applied on high-speed ships, such as RoPax ferries, which are the types of ships servicing the routes of the GCSN. The mathematical formulation for estimating differences in fuel consumption at sea for slow steaming per trip is given in Eq. (8), while differences in CO₂ emissions are calculated in Eq. (9). It is worth noting that in this case, port fuel consumption is not taken into consideration, mainly for two reasons. First, given the overall structure of the network, it is assumed that port times are low, considering that the methodology is applied to a domestic passenger shipping network. Moreover, port fuel consumption is not added to the fuel consumption estimation in this study as added fuel consumption, as the average fuel consumption from the EU-MRV is used, which is the result of reported actual data of the fuel consumption of a ship through different calculating and monitoring methods, thus taking into consideration both fuel consumption at sea and at ports. Moreover, one of the goals of this study is to address how a passenger shipping network could benefit from slow steaming, which in our case is applied during operations at sea. As our study focuses on the emissios mitigation part, which is described as the difference in emissions at sea, port emissions are, for the purposes of the study, considered to remain the same, and therefore their calculation is omitted at this stage. Still, while our study does not focus on port fuel consumption under different ship states, such calculations can improve the accuracy and validity of the generated results and can be considered for future research.

$$\Delta (F.C.)=\frac{L}{24\bullet {V}_0}{F}_0\left({a}^2-1\right)$$

(8)

$$\Delta \left({CO}_2\ emissions\right)={F}_{CO_2}\bullet \left[\frac{L}{24\bullet {V}_0}{F}_0\left({a}^2-1\right)\right]$$

(9)

In the above equations, Δ(F.C.) shows the difference in fuel consumption, L the total distance traveled, V₀ the current average speed of the ship, F₀ the ship’s reported average fuel consumption, Δ(CO₂ emissions) denotes the difference in CO₂ emissions, ${F}_{CO_2}$ is the emission factor used for CO₂ emissions, and a is the speed reduction factor used for the purposes of the study, equaling between 0.80 and 0.95 for speed reductions between 20 and 5%, respectively. In this study, the emission factor for heavy fuel oil will be used for the calculation of emissions, equaling 3.114, which is in accordance with the reported data of EMSA\THETIS-MRV (European Maritime Safety Agency 2019) for most of the ships operating on the GCSN.

Based on the identification of areas with increased CO₂ emissions, two slow steaming strategies for the reduction of CO₂ emissions are assessed. The first approach includes implementing slow steaming on the whole network, which is the full implementation scenario of slow steaming. While this approach may seem somewhat aggressive, as it would be the one affecting travel times as much as possible, it is crucial to implement the proposed speed reductions at such a level to provide a benchmark for comparisons in order to better evaluate the effectiveness of the proposed spatially concentrated approaches. The second strategy is a spatially related one. In this approach, slow steaming is implemented specifically in areas where high clusters of CO₂ emissions appear, simulating a scenario as if emission control areas (ECAs) were to be implemented (ECA scenarios). In these cases, there are two options, with the first being implementing slow steaming on the resulted high-high areas from the Anselin Local Moran’s I analysis and the second being on the resulted areas from the Getis-Ord Gi* statistic. The mathematical formulations presented above show that these two statistics are calculated differently, thus leading to different results in terms of high clusters. As a result, their use provides the decision-makers with two different options for the implementation of slow steaming, as the number of high-value clusters in each model differs. Therefore, the spatially related implementation of slow steaming is conducted and assessed regarding the results of each spatial statistic, thus leading to two distinct scenarios: one for the Anselin Local Moran’s I case and another for the Getis-Ord Gi* case. For the implementation of slow steaming on the highly valued spatial clusters, all necessary data are calculated on the route parts that are included inside the boundaries of each grid division (i.e., polygon feature), with fuel reduction being calculated for each part considering all four possible scenarios of 5%, 10%, 15%, and 20% speed reduction. In both approaches, fuel consumption and CO₂ emission reductions are calculated over a 3-month period, while time delays because of speed reduction are calculated on a single-trip basis, consisting of all port calls between the port of origin and the island destinations of the specific route.

Results

Considering a given maritime network’s environmental impacts, it is crucial to initially investigate the source of its environmental footprint. While route scheduling, trip frequency, and generally overall ship traffic affect the overall emissions generated by a network, it is important to consider that in the case of the Greek Shipping Network, the operating fleet is an older one, with older ships being less energy-effective. For instance, on the examined routes, initial results show that 18 out of 23 ships examined for the purposes of this study emit more than 400 kg of CO₂ per nautical mile traveled, with an overall fleet average of 523.35 kg per nautical mile. Regarding the routes analyzed, on average, approximately 180 metric tons of CO₂ emissions are generated per trip, for a total of 13,936.82 metric tons of CO₂ generated from the Aegean fleet on a single-trip basis. The general inefficiency, in terms of environmental performance, of the fleet is better shown in Figs. 3 and 4, showing ship CO₂ emissions per distance traveled (kg/n. mile) and total CO₂ emissions emitted per route round trip, respectively.

Fig. 3

Ship energy efficiency per distance travelled as CO₂ kg emitted per nautical mile

Fig. 4

Route energy efficiency per trip as CO₂ emissions (m. tonnes) generated per round trip on each route of the study

Determining which areas will be utilized as ECAs for the purposes of this study is the first step in the implementation of the decision-making process regarding slow steaming. As such, by utilizing the two different methods mentioned above for identifying areas of higher concentrated CO₂ emissions, corresponding maps of the results are generated. The first one presented in Fig. 5 shows areas of high concentration of CO₂ emissions, while neighboring areas also show high values of CO₂ emissions, by performing a Cluster and Outlier Analysis (Anselin Local Moran’s I) and by calculating Local Moran’s I statistics of spatial association. Results of this analysis are best given by a code representing the cluster type for each statistically significant feature, such as “HH,” denoting high-high areas, as mentioned above. Other statistically significant areas are shown as “LH” and “LL,” denoting low-high areas, which are areas of low CO₂ emissions with neighboring areas of higher concentrations, and low-low areas, which are areas of low CO₂ emissions with neighboring ones of also low CO₂ emissions. The outputs of LH are, to an extent, justified as they are shown near areas of the HH category, thus denoting areas where ship traffic and therefore emissions are low, with neighboring areas being the highly polluted ones. The second map, which is shown in Fig. 6, shows the results of the Getis-Ord Gi* statistic, which is a hot spot analysis. The results of the statistic are more straightforward, as the higher the value of the statistic, the more statistically significant a feature is, consisting of a hot spot with high values of CO₂ emissions and also neighboring features with high values of CO₂ emissions. For statistically significant positive z-scores, the larger the z-score is, the more intense the clustering of high values (denoting a hot spot), while for statistically significant negative z-scores, the smaller the z-score is, the more intense the clustering of low values (denoting a cold spot). The figure shows areas where there are hot spots, categorized based on their statistical significance, with the scales denoting areas of 90%, 95%, and 99% statistical significance, respectively (Fig. 6).

Fig. 5

Anselin Local Moran’s I statistically significant areas of interest denoting highly polluted areas of CO₂ emissions

Fig. 6

Getis-Ord Gi* hot spots of statistically significant areas of interest denoting highly polluted areas of CO₂ emissions

Results from both methods do not show significant variations, considering high-high areas and hot spots at 99% confidence levels. In the case of the calculations of Anselin Local Moran’s I statistic, 61 areas were identified as high-high clusters. In the case of the Getis-Ord Gi* statistic, 69 areas were identified with a z-score higher than 3, denoting areas of high CO₂ emission concentrations with a 99% statistical confidence, while an overall 110 areas were identified with a Gi* statistic score where high clustering statistical confidence is at least 90%. For the purpose of this study, as the ECA scenarios are implemented to provide more flexibility regarding travel time delays, the more lenient case of slow steaming on the 61 areas of the first map is adopted as the first scenario (level 1 scenario). For the second slow steaming scenario, considered an intermediate solution (level 2 scenario), all high cluster areas of the Getis-Ord hot spot analysis are considered ECAs, which also denote highly polluted areas of CO₂ emissions. It is worth noting that all statistically significant areas (hot spots of more than 90% statistical confidence) of the Getis-Ord Gi* hot spot analysis are considered for the intermediate scenario, as statistically significant areas of at least 99% were almost similar to the level 1 scenario outputs of the Local Moran’s I areas.

To better evaluate the benefits of slow steaming, the same reductions will be implemented in both scenarios, which consist of reductions of 5%, 10%, 15%, and 20%. In addition, a more radical scenario for slow steaming is implemented, which considers the full implementation of slow steaming on the whole GCSN. While this is a scenario that is not as viable in reality, as travel times will be greatly increased, thus deteriorating the offered level of service (LoS) for passengers, it is implemented for comparison in order to provide a benchmark to better evaluate the effectiveness of the two different strategies implemented on level 1 and level 2 scenarios.

The implementation of all different scenarios aims to cover the potential of the GCSN for both economic and environmental benefits in fuel consumption and CO₂ emission reductions, respectively. Although it would be possible to reduce a ship’s speed by more than 20% on smaller areas and, therefore, shorter route parts, the maximum speed reduction cutoff point of 20% is applied on all approaches, as a larger reduction would lead to less realistic outputs. Results of the different slow steaming scenarios are shown in Tables 2, 3, 4, showing the total fuel consumption reduction in metric tons, the CO₂ emission reduction, the total time lost on the GCSN, and passenger travel delays on a trip basis as a result of slower speeds. Additionally, to better compare the different strategies, two more outputs are introduced, showing the overall efficiency of each strategy by providing metrics on both fuel consumption and CO₂ emission reductions relative to the total time lost on the GCSN from each slow steaming scenario. These metrics aim to provide insight into how effectively each strategy reduces emissions while also aiming to reduce travel times as much as possible. As a result, the higher both of these metrics are, the better the scenario implemented, although bear in mind that both of these metrics should be considered along with the absolute passenger delays. For instance, both of these metrics show their highest values in the full implementation scenario of slow steaming on all routes of the GCSN. While this may show that the rate of CO₂ emission mitigation due to passenger delays is better, absolute travel time delay results show a highly deteriorated LoS for the GCSN. Consequently, all generated outputs must be considered before the evaluation of each scenario (Tables 2 to 4).

Table 2 Environmental and travel time impacts of level 1 scenario (slow steaming on Moran’s I high-high areas) for a 3-month period

Table 3 Environmental and travel time impacts of level 2 scenario (slow steaming on Getis-Ord high clusters areas) for a 3-month period

Table 4 Environmental and travel time impacts of level 3 scenario (full implementation of slow steaming on the GCSN) for a 3-month period

While results in all scenarios are promising regarding fuel consumption and, therefore, the reduction of CO₂ emissions, it is important to take into consideration the total passenger time delays per trip on all approaches. Although both the level 1 and level 2 scenarios do not have significant impacts on time delays, especially considering less-affecting speed reduction strategies, relative fuel and CO₂ emission reductions are significantly less than those of the full GCSN approach. In order to reduce fuel consumption and CO₂ emissions on level 1 and level 2 scenarios, it would be necessary to implement higher speed reductions, which, as shown above, would in fact impact passenger times by increasing time delays on a per-trip basis. From the evaluation of the above results, an efficient strategy that could potentially reduce CO₂ emissions while also aiming to provide acceptable travel times would be a 10% speed reduction on the level 2 scenario of the Getis-Ord hot spot ECAs. In this case, more CO₂ emissions are reduced than the 15% speed reduction of the level 1 scenario, while travel times are not affected as much, showing that targeted and concise selections on areas of slow steaming could lead to the much-needed mitigation of CO₂ emissions while still maintaining an acceptable LoS through fewer passenger time delays.

To better assess the validity of our results, we compare them with the existing study of Psaraftis et al. (2009), as our results are also generated through the mathematical formulations for slow steaming proposed in their study. The generated results are reasonable when compared with their results, which also considered the case of a roundtrip voyage on a RoPax ferry, which is in accordance with most ships operating in the Aegean Sea. Figure 7 shows that through the full implementation of slow steaming in the GCSN, even better results are possible, considering all speed reductions between 5 and 20%. It is worth noting that percentage changes remain the same for all three, as emissions and fuel costs are multiples of fuel consumption.

Fig. 7

Percentage reductions in fuel consumption, CO₂ emissions, and fuel costs compared to the related study of Psaraftis et al. (2009)

Comparing increases in voyage time, the average time increases in our study are also in accordance with the ones in the previous study. Specifically, Psaraftis and Kontovas conclude in the results of their study that a one-knot speed reduction, which is approximately a 5% speed reduction in their study, would lead to a 25-min increase in voyage time for the case set while leading to a 10% reduction in emissions, fuel consumption, and fuel costs. In the case of our study, which considers 78 distinct routes, reductions of 11.55% in emissions, fuel consumption, and costs are achieved, with an average increase in travel times of 27 min. While it is also worth noting that the maximum voyage time increase in this case of speed reduction was approximately 103 min, it should be taken into consideration that this occurs on one of the longest routes of the network in 2019, ranging almost 470 nautical miles. Lastly, considering all routes of the network analyzed in this study, a 5% speed reduction on the whole network would lead to less than or equal to 30-min increases in 48 routes of the study, with 37 of them showing time increases less than 25 min, with overall reductions in fuel consumption, fuel costs and CO₂ emissions showing similar results with the aforementioned related study (Fig. 7).

It is becoming clear from the above results that stricter policy measures are necessary for CO₂ emissions to be reduced. As highlighted by our study, although slow steaming can still generate promising results and be implemented as a transitional measure for mitigating CO₂ emissions, there are several additional strategies in the literature that could be implemented along with slow steaming for increased environmental benefits. It is crucial to also consider that there are several economic benefits to be gained from slow steaming strategies, therefore creating more opportunities for operators to increase their costs or limit their losses in periods of seasonality when demand drops. Results from spatially associated and more targeted scenarios in areas of high emission concentrations also show that by achieving certain environmental goals through speed reduction, operating costs can still be limited through fuel consumption reductions, albeit at a much slower rate than in the case of a full network implementation of slow steaming. Regarding the environmental, economic, and level of service-related impacts of slow steaming, there are several options presented to facilitate the decision-making of operators and policymakers, which can still lead to much-needed reductions in CO₂ emissions while also retaining acceptable passenger travel times on a given network such as the GCSN.

Discussion

The goal of this study was to present a methodological framework for the determination of high-polluting areas of CO₂ emissions from domestic shipping activities and further utilize such results in decision-making regarding mitigation strategies. The study highlighted three distinct gaps in the existing literature, on which it focused its efforts on providing a more holistic approach in the field of maritime transport where interest is low. First, the study highlights the fact that, while the spatial aspect of maritime transport has been gaining interest, this mostly focuses on visualization efforts rather than actual analysis of spatially related data. Acknowledging the fact that emissions are also characterized by spatial characteristics, the study proposes that emission inventory studies should also focus on the creation of comprehensive geodatabases, which could prove vital for the assessment of emissions and their distribution. By connecting emission data with their spatial aspect, thus providing a spatially related emission inventory, spatial data analysis models can then be utilized to further identify highly polluted areas for more targeted interventions considering mitigation strategies. Utilizing the results of spatial modeling implementation, the study then further extends the existing literature by implementing targeted spatially optimal slow steaming strategies in the most sensitive and polluted areas of the study.

The proposed methodological framework of the study used statistically reported data from the EU-MRV system, which have been proven reliable in other studies, in order to initially create the resulted databases and associate them with the spatial dimension of existing routes of the GCSN. While the resulted data are generated by exploiting actual reported data from the operators of each ship, there is still an apparent limitation in their use. This limitation arises from the fact that not all ships are obliged to report data to the system, as only those with an equal or higher capacity than 5000 GT are obligated to report fuel consumption, hours at sea, and CO₂ emissions. While this may not be an issue for studies considering larger ships or routes operated by large RoPax ferries, an assumption is necessary for routes of lower demand operated by ships not reporting their relative data. This results in certain operating ships in the study being replaced by others of similar characteristics in order for the estimation of emissions to be conducted on all routes considered. While the resulting emission inventory was created just for the purposes of benchmarking in this study, it is still something to be considered for future research and an area for improvement in the proposed methodological framework. Therefore, a suggestion for future studies would be to use more approaches in the emission inventory process for better validation of results, as, for instance, shown by relevant studies that have used a combination of bottom-up and top-down approaches.

In addition, the study has shown that the use of spatial autoregressive models for better identification of emissions’ spatial patterns and highly polluted areas can yield significant results. These results, while they can differ to an extent when considering different spatial analysis models, can still provide policymakers, operators, stakeholders, and decision-makers with the necessary tools to extend their understanding of a given network, leading to more efficient decision-making and targeted interventions to improve operations and limit environmental externalities. However, it is always worth considering that the outputs of spatial analysis methods are directly correlated with the input data prepared and used by the respective user. For instance, this study used a 10 km × 10 km division of the study area, targeting a more precise identification of the spatial distribution of emissions. While the proposed division did generate reasonable results, with highly polluted areas being identified on sections of the network where high traffic of less environmentally efficient ships exists, the results are still dependent on the inputs, the area of implementation and its precision, and the specific spatial analysis model used. A suggestion for future studies would be to extend the grid division to variable precision levels (e.g., 25 km × 25 km) in order to assess the results of different implementations, which could result in larger area extents characterized as highly polluted areas. This would in turn provide additional insights into the decision-making process regarding targeted interventions in such areas and therefore also provide better results in terms of the effectiveness of mitigation strategies.

On that note, the study has also assessed whether a spatial decision support system, as proposed in this study, for emission mitigation strategies could prove useful for policymakers aiming to limit the environmental footprint of a given maritime network. Slow steaming strategies were applied to the areas that resulted in highly polluted areas from the implementation of the spatial autoregressive models. As mentioned before, the effectiveness of spatially associated slow steaming relies heavily on the implemented areas’ extents, which in this case were only a small portion of the total area of the GCSN. As a result, extending their implementation further, with the aforementioned larger extents considering grid division, could lead to even more promising results. The goal of the study was to assess whether slow steaming on its own could limit the environmental footprint of the GCSN, which could potentially be achieved by reducing the speed of the whole fleet, resulting in highly promising results in emission reductions while only fractionally increasing travel times on average. However, an implementation on the whole network would necessarily need a more in-depth analysis, not only on a spatial scale but also on each distinct route of the network. It is clear, as shown by relevant studies, that slow steaming may not always be a feasible solution considering passenger shipping, especially regarding the lengthier routes of the GCSN, spanning several nautical miles and connecting a large number of islands. In addition, a clear suggestion from this study for future ones is the assessment of environmental gains in CO₂ emissions reduced in combination with tradeoffs in voyage time increases for passengers. More specifically, as CO₂ emissions are reduced, so are fuel costs for the operating ships, thus providing operators with the option of introducing new pricing schemes for passengers, which would aim to balance the tradeoffs of increased travel times. In the case of the GCSN, where extreme seasonality characterizes network operations, future studies could also assess the effectiveness of seasonal slow steaming, consisting of more aggressive speed reductions in the winter months where demand is low and more minor reductions in the summer months, as such an approach could also lead to significant CO₂ emission reductions. Still, slow steaming is shown to be effective in the case of passenger shipping as well, confirming the results from the previous study of Psaraftis et al. (2009), which also suggests that a slight speed reduction of 5% could be considered in the long run.

Nevertheless, due to the inelastic nature of the short-sea shipping sector, even minor speed reductions do not seem to be considered, as highlighted by the relevant studies. While our results show that spatially related slow steaming generates reasonable and promising results, the specific measure alone as a way of reducing CO₂ emissions can do so at a much slower rate if applied to specific areas. Overall, through the implementation of spatially related mitigation strategies applied in this study, it is becoming evident that the GCSN is in need of significant improvements in terms of its environmental footprint and the efficiency of its fleet, with some ships operating for more than 25 years. While slow steaming in selected areas can potentially have several positive environmental and economic impacts, the decision-making process could also be supplemented by additional variables aiming at an overall restructuring of the network and optimizing operations in order to achieve better environmental efficiency. Future studies should also focus on the development of new alternative routes while considering continuous and effective emission monitoring and the introduction of more environmentally friendly or even zero-emission ships, which have proven to be feasible on a network with complex topological characteristics such as the GCSN. While slow steaming can provide a viable option, even for passenger shipping networks, for the transition to more environmentally efficient maritime transport, overall fleet renewal should be considered the primary goal of future studies in that area, with newer types of ships using alternative fuels being crucial for operations on higher traffic routes. Consequently, future studies should focus on further enhancing such spatial decision support systems for emission monitoring, the implementation of mitigation strategies, and the overall environmental improvement of maritime networks, thus leading to several environmental and economic benefits. Nevertheless, the overall efficiency of such decision support systems can still be improved with more holistic approaches, considering several other variables, options, and available strategies to further facilitate the decision-making process of policymakers, operators, and stakeholders in the industry.