We present a survey of AIS data analytics techniques for shipping business decision-making. Our survey provides an indicative categorization of the areas where AIS data analytics may assist in strategic decision-making, based on the costs that a shipping business needs to cover. These areas include chartering and freight markets, vessel operation and environmental footprint. Our survey is useful both as a catalogue of existing research, as well as a critical evaluation of the field. The use of AIS data has facilitated the state-of-practice in shipping business decision-making. Furthermore, enriching AIS data with other data sources is necessary more often than not.
and destination, and dynamic messages, including the Maritime Mobile Service Identity Number (MMSI), rate More than 80% of global trade by volume and over 70% of turn, speed over ground, position coordinates, course by value is carried by sea [1]. The shipping business is over ground, heading and navigational status. The use a global, highly-competitive and cyclical industry that of AIS data supports a wide range of applications in the involves heavily leveraged assets, exposing shipowners shipping industry. These include, among others, collision to various business risks that require timely decision- avoidance, fishing fleet monitoring, maritime security, making [2]. For these reasons, exploiting the benefits of infrastructure protection, trade analysis, as well as ship maritime informatics [3], primarily through the analy- and port performance. Lee et al [6] reviewed the historisis of Automatic Identification System (AIS) data, forms cal developments of AIS applications in the management an important tool for shipping companies striving to of waterways, natural resources, freight and ports. In a place themselves in front of the competition and sur- similar spirit, Svanberg et al [7] provided a structured vive the shipping business cycle [4]. For instance, mar- overview of various AIS applications, including interacitime data analytics can support: a) optimal steaming for tions with natural resources (e.g., species, fishing and ice), just-in-time arrivals, b) reduction of unnecessary waiting collision avoidance, oil spills’ investigation, as well as times by enhancing coordination, c) eficient utilization trafic and logistics analysis. Yang et al [ 8] reviewed some of human resources, d) service providers and service con- applications of AIS data analytics, including navigation sumers while establishing market-based business deals, safety, trade analysis, fishing activities, environmental e) predictive maintenance based on digital twins of criti- evaluation, oil spill risk analysis, and ship and port perforcal assets and their components, and f) optimized cargo mance. Emmens et al [9] and Bereta et al [10] examined planning [5]. the promises and perils of AIS data, including unrealistic
The International Maritime Organization (IMO), under tracks, vulnerability to external conditions, inaccuracies the Safety of Life at Sea (SOLAS), has adopted AIS across by human input, attacks, as well as intentional commuseveral other reporting systems associated with tracking nication gaps . vessels. There are 64 diferent types of AIS messages We present survey of AIS data analytics for decisiondivided into two main categories: static messages, in- making in the shipping business. Due to space limitacluding e.g. the IMO number, name of the vessel, type of tions, we chose to focus mostly on the literature of the the vessel, dimensions, estimated time of arrival, draught maritime domain for the benefit of the Data Scientist. Our survey provides an indicative categorization of the areas where AIS data analytics may assist in strategic decision-making, based on the costs that a shipping business needs to cover. We discuss how AIS data may be used for vessel chartering decisions, the assessment of vessel operation, maritime trade, and environmental imProceedings of the Workshop on Big Mobility Data Analytics (BMDA) co-located with EDBT/ICDT 2023 Joint Conference (March 28-31, 2023), Ioannina, Greece * Corresponding author. $ a.kouvaras@unipi.gr (A. Kouvaras); dtsouknidis@aueb.gr (D. Tsouknidis); a.artikis@unipi.gr (A. Artikis) © 2022 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
pact calculation. For each area, we provide an overview, rectification of incidents. and a synthesis of the main contributions and limitations.
In other words, our survey is useful both as a catalogue of existing research, as well as a critical evaluation of the ifeld. 2. Survey Structure In this section, we discuss the use of AIS data for supporting the chartering decisions of shipowners and charterers.
We classify the AIS data analytics approaches into five ar- Adland et al [25] provided a theoretical exposition and eas. Table 1 summarizes, for each approach, the employed empirical analysis of the micro- and macro-economic dedata types, the geographical and temporal coverage of terminants of vessel capacity utilization in bulk shipping the empirical analysis, as well as the vessel types for markets. One step further, Sugrue et al [20] suggested a which the analysis was performed. (All aproaches have linear model to predict vessel capacity based on water employed AIS data, and thus this is omitted from the surface elevation. In an efort to gain further understandtable to save space.) We start the survey by discussing ing of the freight market, it is important to derive the the area of chartering and freight markets (see Section 3). actual demand and supply balance. Towards this, ProcFreight rate revenue is the principal source of repayment hazka et al [18] provided a prediction of the demand and in connection with ship financing, impacting liquidity supply balance in the freight market, based on historical and profitability ratios that financiers use to monitor the and online AIS data. In the same vein, Regli et al [16] performance of a shipping loan. In this context, ship- proposed a method for calculating the short-term capacping business decisions are influenced by the economic ity in the voyage charter market based on the ratio of cycles and the volatility in demand, impacting freight available to active vessels. They investigated the percentrates and ship values for diferent segments of maritime age of vessels available for orders by using AIS draught transport, law regulations, risk profiles and ownership measurements. requirements. Bai et al [56] explored the efectiveness of risk man
In Section 4, we discuss the vessels’ operation. Ship- agement strategies for mitigating risk exposure to freight ping companies use key performance indicators (KPIs) rate and bunker fuel prices, using vessel and voyageto monitor and analyze the performance of each vessel, related data. In this context, Prochazka et al [15] invessuch as the number of overdue planned maintenance tigated the factors afecting the preferable fixture locatasks. Due to the high market competition, this process tion, such as market conditions, vessel characteristics, should be completed quickly. For this reason, the tradi- and charterer’s preferences. Bai and Lam [12] explored tional daily noon reports are insuficient. Instead, AIS the impacts of selected attributes (i.e., freight rate, comdata can contribute to minimizing the time required for modity price arbitrage, bunker price and the number of assessing vessel operation. ships in a specific area) on the charterer’s destination
Maritime trade analysis, discussed in Section 5, can choice. Jia et al [13] used machine learning techniques identify commodity flows and ship trading patterns rep- to predict the destination for crude oil exports. They resenting the diferent shipowners’ and charterers’ be- investigated the micro- and macro-level determinants of haviors. Events like the COVID-19 pandemic, Oil Glut the preferable destination. In a similar study, Zhang et al (2014-2016), and the financial crisis of 2008-2009, can be [19] suggested a data-driven model for vessel destination used to study freight flows as examples to avoid future prediction based on the similarity between the current crisis situations. trajectory and historical trajectories. Regli et al [17] iden
Finally, shipping environmental impact analysis, as tified the vessel specifications that afect the charterer’s presented in Section 6, demonstrates the increasing inter- decision to exploit storage arbitrage opportunities using est in sustainability. Shipping companies evaluate ESG historical AIS data. (Environmental, Social, Corporate Governance)-related KPIs, mainly concerning the investment required to de- 3.2. Contributions carbonize and operate in terms of reduction of CO2. Sustainability can potentially impact cash flows, the collateral value of a ship, and, therefore, the value-to-loan ratio.
ESG initiatives and investments are opportunity costs impacting the predictability of capital, operational and voyage costs. Failure to comply with targets and thresholds may result in fines, higher interest rates, additional capital injections, higher fees to port agents, delays and In most cases, the AIS draught measurement has been used for estimating the cargo payload of a commercial vessel — this is a key variable calculating revenue for a particular voyage and estimating global trade flows for commodities. AIS draught measurement was also used to distinguish between laden and ballast voyages. More accurate calculations for cargo payload estimation may be achieved by combining AIS draught measurements
Theme “-" indicates that there are no available details about the field. with additional indicators, such as the number of ships we discuss the research focusing on the evaluating vessel waiting for a contract, the number of days the ships are behavior. An important feature when operating a vessel waiting, the vessels that are on dedicated routes and do is speed selection, due to the costs involved as well as not contribute to the spot market voyages [18, 12, 17]. its relevance to commercial and charter-parties’ terms. The ratio of available vessels to active vessels is a po- Early research by [22, 25] investigated technical, opertentially helpful indicator of shipping economic activity ational, and macro-economic variables concerning the and, as such, may be used more widely as a freight rate vessel’s speed using a regression model. Adland et al forecast indicator and a proxy for trading and physical [23] attempted to prove, with the use of dynamic AIS market activity [17]. data, that the introduction of stricter regulations of an
The analysis of historical AIS data related to opera- Emission Control Area has no efects on vessel speed. tional risk management strategies (e.g., fleet diversity, Prochazka et al [28] investigated how contractual obligalfeet age, relative trip distance, fleet repositioning flexi- tions afect the speed of vessels. Shu et al [ 24] quantified bility and trading diversity), allows shipping companies the influence of weather conditions and vessel encounto draw valuable conclusions [56]. The models proposed ters on vessel speed, course and path within ports and for AIS data-driven destination prediction may be clas- inland waterways. sified into two categories: a) the turning point-based In addition to the study of the normal behavior of ships, destination prediction methods, and b) the trajectory- the maritime community is also interested in the study of based destination prediction methods [19]. For instance, anomalies [57]. In this context, in a recent study, Zhang predicting oil export destinations allows for better fore- et al [31] proposed a dynamic maritime trafic pattern casting of regional and local market balance, improved recognition model that adapts to the changes in the trafic knowledge of inventory levels, and monitoring of the sup- environment. Finally, despite the usefulness of AIS data ply chain. The model proposed by [14] has an accuracy for monitoring ship behavior, data is often missing due to ranging between 70-90%. human negligence or intention. In this context, Zhou et al [26] investigated the impact of wind and sea currents 3.3. Limitations on ship behavior within ports, where vessel trajectories can be observed, using ship maneuvering information provided by dynamic AIS data. Rodger et al [29] suggested a methodology to map ships which do not report their AIS information using SAR ship detection.
The use of AIS data seems to lead to much better accuracy than the use of traditional noon reports, as a) errors due to human input are reduced, and b) information can be obtained in an online fashion. On the other hand, an important limitation of several studies examining chartering 4.2. Contributions decisions concerns the fact that there is no commercial information available through AIS, such as information In addition to the aforementioned contributions, there about about the cargo and charter-parties. Publicly avail- were eforts towards data pre-processing in order to inable fixtures and freight derivatives information covers crease data reliability [26, 25, 22]. Furthermore, to distinonly a tiny fraction of the voyages observed in AIS data guish between the diferent vessel’s operational modes, [21, 18, 15, 17]. Moreover, in many studies the match- an efort was made to distinguish stops at anchor and ing process of AIS data and fixtures data is based on the stops at berth [54]. The eforts to model the behaviour vessel name because the IMO number is not part of the of dark vessels is also noteworthy, since communication ifxture reports. This can be problematic since the ship’s gaps are increasing over time and often associated with name may change, while many ships may bear the same illicit behaviour. name [18]. One way to address this issue is to consider other static vessel attributes, such as the vessel’s type 4.3. Limitations and dimensions.
In any case, the methods that estimate the cargo pay- The use of AIS data alone is rarely suficient for the asload by using AIS measurements have additional limita- sessment of vessel operation. AIS data must be enriched tions — consider, e.g., the dificulty of measuring ballast with information from additional sources, such as meteowater and fuel during draught measurement [14]. rological [27, 25, 22], commercial [25], maintenance [22], technical [24], management, flag and port data [ 27]. Consequently, the studies based solely on AIS data often have 4. Vessel operation limited significance. Furthermore, it has been dificult to determine the point at which a ship passes from laden 4.1. State-of-the-art to ballast state, thus the semi-laden voyages are ignored [54, 25]. While the approaches discussed in this section ofer valuable insights on the benefits of AIS data anaVessel monitoring is essential for performance estimation as well as for controlling her activities. In this section, lytics for vessel operation assessment, their conclusions fixtures records to AIS data is not often possible , because are mostly drawn from studying specific geographical not every reported fixture is eventually realized [39]. regions (see Table 1). Thus, it is not clear whether they There are diferences from the oficial customs statiscan be generalised to other or larger areas. tics related to imports and exports in countries whose other transport modes (i.e., pipelines) are important too.
Monthly trade statistics are generally available for some 5. Maritime trade countries but not all. Future research could combine AIS data and oil pipeline data to calculate the marine trade 5.1. State-of-the-art volume, and thus improve the accuracy of the oil trade volume calculated with the use of AIS data.
of commodity flows based on AIS data. Adland et al [23] compared the accuracy of AIS-derived trade statistics against oficial customs data in the crude oil market. Kanamoto et al [38] analysed the global trade flow pattern of dry bulk cargo by commodity. They suggested a model to forecast the future shipping demand by vessel type and commodity. Yan et al [37] calculated the oil trade volume by establishing a model for cargo payload calculation based on draught and vessel’s technical information. Fuentes et al [35] proposed a recognition model of anchored vessels waiting for transit from Suez Canal. They identified the access routes from anchorages based on AIS draught measurements. Li et al [39] proposed a machine learning technique to predict the cargo type transported by coated product tankers.
Existing literature has also examined the characteristics of maritime trade during crisis events. For instance, Millefori et al [36] analysed the efects that the COVID-19 pandemic and containment measures had on the shipping industry per type of commercial shipping.
The empirical analyses of historical maritime trade can help forecast future activity — this is particularly helpful during rare crisis events [36]. To scale to large volumes of maritime trade, trajectory reconstruction techniques have been employed and customized [19, 34, 33]. The proposed algorithms maintained only the minimum number of trajectories reflecting current trafic patterns. Moreover, there are techniques that handle routes with missing data and give the best possible estimation from the available input [33]. Finally, cargo type prediction may also help promote data transparency in the maritime industry, because the type of product a vessel carries is typically private information [39].
Several commodities are often handled at the same port (i.e., multi-purpose terminals), or even at the same berth [38]. Moreover, a vessel often transports several commodities. Consequently, estimating the commodities carried by dry bulk carriers with AIS data only has proven dificult [ 38, 37, 32]. On another issue, matching all the
6.1.1. Emissions calculation Early research estimated the ships’ air pollution emissions during diferent operation modes, i.e., berthing, maneuvering and hotelling [40]. In a similar study, Winther et al [42] calculated the past and future emissions by combining dynamic AIS data with ship engine power functions and technologystratified emission factors. In a more recent study, Schwarzkopf et al [46] constructed future scenarios about ship emissions based on a virtual shipping fleet. The way of calculating pollutant emissions in the open sea difers from the way of calculating them in ports. Tran et al [49] investigated the container vessel segment by compiling a comprehensive emission profile by vessel size, port call and carriers.
The approaches for emission calculation are often hindered by the absence of some static AIS data, or vessel technical information. To address these issues, Peng et al [45] calculated ship emission inventories based on sampling statistics, using individual vessels with all the necessary data for estimating the population’s emissions.
In the absence of engine specifications, Zhang et al [44] calculated emissions from vessels through categorical regression based on vessels with similar characteristics.
Johansson et al [43] proposed a route generation algorithm to compare emission calculations with previous inventories. They introduced the “most-similar-vessel" to complete the missing ship technical information. 6.1.2. Fuel consumption measurement and savings Bai et al [58] investigated, for each ship type, the factors that afect the shipowners’ choices regarding diferent feasible schemes for reducing sulfur emissions. Safaei et al [51] suggested a prediction model to estimate fuel consumption based on multiple linear regression. Kim et al [52] used big data techniques to optimize data processing and computing time of the Energy Eficiency Operational Indicator. To estimate fuel consumption, they used static and dynamic AIS data. Watson et al [50] pro- Due to space limitations, we could only present a fragposed a methodology to estimate the carbon savings by ment of our survey here. We chose to focus mostly on assuming that a ship sails at her lowest observed speed. the literature of the maritime domain for the benefit of Stolz et al [54] investigated the time that ships spend at the Data Scientist. We are currently finalizing the comberth by using AIS data, in order to estimate the auxiliary plete survey, that presents a detailed discussion of the power demand at berth. Sundvor et al [55] investigated literature, including work from the fields of Data Science route requirements and energy demands of high-speed and Artificial Intelligence. passenger vessels, aiming to identify the candidates for zero-emission replacement.
The use of AIS data has facilitated emissions and fuel consumption calculation. A noteworthy contribution, in this area, concerns the eforts towards reducing the impact of missing static data, using primarily sampling techniques [45, 59, 43]. Moreover, simplification techniques for dynamic AIS data have been proposed, in order to take advantage of large volumes of data [48]. [1] UNCTAD, Review of Maritime Transport 2021,
2021. [2] G. Alexandridis, M. G. Kavussanos, C. Y. Kim, 6.3. Limitations D. A. Tsouknidis, I. D. Visvikis, A survey of shipping finance research: Setting the future research For the auxiliary engines and boiler, the ship registry agenda, Transportation Research Part E: Logistics database does not always provide the full specifications and Transportation Review 115 (2018) 164–212. of installed power; thus, engine usage in various oper- [3] A. Artikis, D. Zissis (Eds.), Guide to Maritime Inforational modes may not always be calculated [49]. Un- matics, Springer, 2021. certainty about voyage parameters, such as cruising and [4] M. Stopford, Maritime Economics, Routledge, 2009. maneuvering distance, has also afected the impact of the [5] C. L. Forcellati, C. Georgeson, M. Lind, S. Singh, reported empirical results [41]. C. Sjöberger, J. Woxenius, Support for financial
For further work, Safaei et al [51] proposed the use decision-making, in: M. Lind, M. Michaelides, of non-linear regression methods, because there is a R. Ward, R. T. Watson (Eds.), Maritime Informatics, non-linear relation between speed and fuel consump- Springer, 2021, pp. 255–274. tion. Waves, sea currents and ice cover can increase the [6] E. S. Lee, A. J. Mokashi, S. Y. Moon, G. S. Kim, The required main power; therefore their influence should maturity of automatic identification systems (AIS) be examined. Other data sources may also be used for and its implications for innovation, Journal of Maload profiles and distinguishing between loading and un- rine Science and Engineering 7 (2019). loading at berth, to estimate the distribution of power [7] M. Svanberg, V. Santén, A. Hörteborn, H. Holm, demand over time [54]. Peng et al [45] suggested examin- C. Finnsgård, AIS in maritime research, Marine ing the uncertainty induced by AIS communication gaps. Policy 106 (2019) 103520.
Finally, there is a plethora of dynamic AIS simplification [8] D. Yang, L. Wu, S. Wang, H. Jia, K. X. Li, How techniques that may be taken into consideration [60]. big data enriches maritime research–a critical review of Automatic Identification System (AIS) data 7. Summary and Further Work applications, Transport Reviews 39 (2019) 755–773. [9] T. Emmens, C. Amrit, A. Abdi, M. Ghosh, The We have carried out a survey AIS data analytics tech- promises and perils of Automatic Identification Sysniques for shipping business decision-making. Our sur- tem data, Expert Systems with Applications 178 vey provides a categorization of the areas where AIS data (2021) 114975. may assist in strategic decision-making, based on the [10] K. Bereta, K. Chatzikokolakis, D. Zissis, Maritime costs that a shipping business needs to cover. Moreover, reporting systems, in: A. Artikis, D. Zissis (Eds.), our survey is useful both as a catalogue of existing re- Guide to Maritime Informatics, Springer, 2021, pp. search, as well as a critical evaluation of the field. The use 3–30. of AIS data has facilitated the state-of-practice in ship- [11] R. Adland, H. Jia, S. P. Strandenes, The determinants ping business decision-making. Furthermore, enriching of vessel capacity utilization: The case of Brazilian AIS data with other data sources is necessary more often iron ore exports, Transportation Research Part A: than not. Policy and Practice 110 (2018) 191–201. [12] X. Bai, J. S. L. Lam, A destination choice model for doorn, Influence of external conditions and vessel very large gas carriers (VLGC) loading from the US encounters on vessel behavior in ports and waterGulf, Energy 174 (2019) 1267–1275. ways using Automatic Identification System data, [13] H. Jia, R. Adland, Y. Wang, Latin American Oil Ocean Engineering 131 (2017) 1–14.
Export Destination Choice: A Machine Learning [25] R. Adland, H. Jia, Dynamic speed choice in bulk Approach, IEEE International Conference on In- shipping, Maritime Economics and Logistics 20 dustrial Engineering and Engineering Management (2018) 253–266.
(2019) 345–348. [26] Y. Zhou, W. Daamen, T. Vellinga, S. P. Hoogendoorn, [14] H. Jia, V. Prakash, T. Smith, Estimating vessel pay- Impacts of wind and current on ship behavior in loads in bulk shipping using AIS data, Interna- ports and waterways: A quantitative analysis based tional Journal of Shipping and Transport Logistics on AIS data, Ocean Engineering 213 (2020) 107774. 11 (2019) 25–40. [27] R. Adland, H. Jia, T. Lode, J. Skontorp, The value of [15] V. Prochazka, R. Adland, F. C. Wolf, Contracting meteorological data in marine risk assessment, Redecisions in the crude oil transportation market: Ev- liability Engineering and System Safety 209 (2021) idence from fixtures matched with AIS data, Trans- 0–9. portation Research Part A: Policy and Practice 130 [28] V. Prochazka, R. Adland, Contractual Obligations (2019) 37–53. and Vessel Speed: Empirical Evidence from the [16] F. Regli, R. Adland, Crude oil contango arbitrage Capesize Drybulk Market, 2021 IEEE International and the floating storage decision, Transportation Conference on Industrial Engineering and EngiResearch Part E: Logistics and Transportation Re- neering Management, IEEM 2021 (2021) 1132–1136. view 122 (2019) 100–118. [29] M. Rodger, R. Guida, Mapping Dark Shipping Zones [17] F. Regli, N. K. Nomikos, The eye in the sky – Freight using Multi-Temporal SAR and AIS Data for Marrate efects of tanker supply, Transportation Re- itime Domain Awareness, in: IGARSS 2022 - 2022 search Part E: Logistics and Transportation Review IEEE International Geoscience and Remote Sensing 125 (2019) 402–424. SymposiumAt: Kuala Lumpur, Malaysia, 2022. [18] V. Prochazka, R. Adland, Feature engineering for [30] M. Uney, Y. Wang, Fast Trajectory Forecasting With supply analysis in ocean transportation, IEEE In- Automatic Identification System Broadcasts, in: Deternational Conference on Industrial Engineering partment of Electrical Engineering and Electronics, and Engineering Management 2020-Decem (2020) University of Liverpool, 2022.
314–318. [31] Y. Zhang, W. Li, Dynamic Maritime Trafic Pat[19] C. Zhang, J. Bin, W. Wang, X. Peng, R. Wang, tern Recognition with Online Cleaning, CompresR. Halldearn, Z. Liu, AIS data driven general vessel sion, Partition, and Clustering of AIS Data, Sensors destination prediction: A random forest based ap- (2022). proach, Transportation Research Part C: Emerging [32] R. Adland, H. Jia, S. P. Strandenes, Are AIS-based Technologies 118 (2020) 102729. trade volume estimates reliable? The case of crude [20] D. Sugrue, P. Adriaens, A data fusion approach to oil exports, Maritime Policy and Management 44 predict shipping eficiency for bulk carriers, Trans- (2017) 657–665. portation Research Part E: Logistics and Transporta- [33] A. Dobrkovic, M. E. Iacob, J. van Hillegersberg, Martion Review 149 (2021) 102326. itime pattern extraction and route reconstruction [21] X. Bai, L. Cheng, Ç. Iris, Data-driven financial and from incomplete AIS data, International Journal of operational risk management: Empirical evidence Data Science and Analytics 5 (2018) 111–136. from the global tramp shipping industry, Trans- [34] S. Zhang, G. you Shi, Z. jiang Liu, Z. wei Zhao, portation Research Part E: Logistics and Transporta- Z. lin Wu, Data-driven based automatic maritime tion Review 158 (2022) 102617. routing from massive AIS trajectories in the face of [22] R. O. Adland, H. Jia, Vessel speed analytics using disparity, Ocean Engineering 155 (2018) 240–250. satellite-based ship position data, IEEE Interna- [35] G. Fuentes, R. Adland, A spatial framework for tional Conference on Industrial Engineering and extracting suez canal transit information from AIS, Engineering Management 2016-Decem (2016) 1299– IEEE International Conference on Industrial Engi1303. neering and Engineering Management 2020-Decem [23] R. Adland, G. Fonnes, H. Jia, O. D. Lampe, S. P. (2020) 586–590.
Strandenes, The impact of regional environmental [36] L. M. Millefiori, P. Braca, D. Zissis, G. Spiliopoulos, regulations on empirical vessel speeds, Transporta- S. Marano, P. K. Willett, S. Carniel, COVID-19 tion Research Part D: Transport and Environment Impact on Global Maritime Mobility (2020) 1–22. 53 (2017) 37–49. arXiv:2009.06960. [24] Y. Shu, W. Daamen, H. Ligteringen, S. P. Hoogen- [37] Z. Yan, Y. Xiao, L. Cheng, S. Chen, X. Zhou, X. Ruan, M. Li, R. He, B. Ran, Analysis of global marine oil Emissions Inventories, Atmoshpere (2022) 1–11. trade based on automatic identification system (AIS) [49] N. K. Tran, J. S. L. Lam, H. Jia, R. Adland, Emissions data, Journal of Transport Geography 83 (2020) from container vessels in the port of Singapore, 102637. Maritime Policy and Management 49 (2022) 306– [38] K. Kanamoto, L. Murong, M. Nakashima, 322.
R. Shibasaki, Can maritime big data be applied [50] R. T. Watson, H. Holm, M. Lind, Green steaming: to shipping industry analysis? Focussing on A methodology for estimating carbon emissions commodities and vessel sizes of dry bulk carriers, avoided, 2015 International Conference on InformaMaritime Economics and Logistics 23 (2021) tion Systems: Exploring the Information Frontier, 211–236. ICIS 2015 (2015). [39] Y. Li, X. Bai, Q. Wang, Z. Ma, A big data approach [51] A. A. Safaei, H. Ghassemi, M. Ghiasi, VLCC’s fuel to cargo type prediction and its implications for consumption prediction modeling based on noon oil trade estimation, Transportation Research Part report and automatic identification system, Cogent E: Logistics and Transportation Review 165 (2022) Engineering 6 (2019).
102831. [52] S. H. Kim, M. I. Roh, M. J. Oh, S. W. Park, I. I. Kim, [40] T. Pitana, E. Kobayashi, N. Wakabayashi, Estima- Estimation of ship operational eficiency from AIS tion of exhaust emissions of marine trafic using data using big data technology, International JourAutomatic Identification System data (Case Study: nal of Naval Architecture and Ocean Engineering Madura Strait area, Indonesia), OCEANS’10 IEEE 12 (2020) 440–454.
Sydney, OCEANSSYD 2010 (2010) 0–5. [53] X. Bai, Y. Hou, D. Yang, Choose clean energy or [41] S. K. Ng, C. Loh, C. Lin, V. Booth, J. W. Chan, A. C. green technology? Empirical evidence from global Yip, Y. Li, A. K. Lau, Policy change driven by an ships, Transportation Research Part E: Logistics AIS-assisted marine emission inventory in Hong and Transportation Review 151 (2021) 102364. Kong and the Pearl River Delta, Atmospheric Envi- [54] B. Stolz, M. Held, G. Georges, K. Boulouchos, The ronment 76 (2013) 102–112. CO2 reduction potential of shore-side electricity in [42] M. Winther, J. H. Christensen, M. S. Plejdrup, E. S. Europe, Applied Energy 285 (2021) 116425.
Ravn, Ó. F. Eriksson, H. O. Kristensen, Emission [55] I. Sundvor, R. J. Thorne, J. Danebergs, F. Aarskog, inventories for ships in the arctic based on satellite C. Weber, Estimating the replacement potential of sampled AIS data, Atmospheric Environment 91 Norwegian high-speed passenger vessels with zero(2014) 1–14. emission solutions, Transportation Research Part [43] L. Johansson, J. P. Jalkanen, J. Kukkonen, Global D: Transport and Environment 99 (2021) 103019. assessment of shipping emissions in 2015 on a high [56] X. Bai, M. Xu, T. Han, D. Yang, Quantifying the imspatial and temporal resolution, Atmospheric Envi- pact of pandemic lockdown policies on global port ronment 167 (2017) 403–415. calls, Transportation Research Part A 164 (2022) [44] L. Zhang, Q. Meng, T. Fang Fwa, Big AIS data 224–241.
based spatial-temporal analyses of ship trafic in [57] A. Jousselme, C. Iphar, G. Pallotta, Uncertainty Singapore port waters, Transportation Research handling for maritime route deviation, in: A. ArPart E: Logistics and Transportation Review 129 tikis, D. Zissis (Eds.), Guide to Maritime Informatics, (2019) 287–304. Springer, 2021, pp. 263–297. [45] X. Peng, Y. Wen, L. Wu, C. Xiao, C. Zhou, D. Han, [58] X. Bai, H. Jia, M. Xu, Port congestion and the ecoA sampling method for calculating regional ship nomics of LPG seaborne transportation, Maritime emission inventories, Transportation Research Part Policy and Management 00 (2021) 1–17.
D: Transport and Environment 89 (2020) 102617. [59] Y. Zhang, J. C. Fung, J. W. Chan, A. K. Lau, The sig[46] D. A. Schwarzkopf, R. Petrik, V. Matthias, nificance of incorporating unidentified vessels into M. Quante, E. Majam, Atmospheric Environment AIS-based ship emission inventory, Atmospheric : X A ship emission modeling system with sce- Environment 203 (2019) 102–113. nario capabilities, Atmoshperic Environment: X 12 [60] K. Patroumpas, Online mobility tracking against (2021). evolving maritime trajectories, in: A. Artikis, D. Zis[47] T. Topic, A. J. Murphy, K. Pazouki, R. Norman, As- sis (Eds.), Guide to Maritime Informatics, Springer, sessment of ship emissions in coastal waters using 2021, pp. 173–201. spatial projections of ship tracks, ship voyage and engine specification data, Cleaner Engineering and
Technology 2 (2021) 100089. [48] Y. Tian, L. Ren, H. Wang, T. Li, Y. Yuan, Y. Zhang,
Impact of AIS Data Thinning on Ship Air Pollutant