=Paper=
{{Paper
|id=Vol-3651/BMDA_paper4
|storemode=property
|title=Detecting the Spatiotemporal Characteristics of the Supply-chain Disruption and Estimating its Short Term Effects
|pdfUrl=https://ceur-ws.org/Vol-3651/BMDA-4.pdf
|volume=Vol-3651
|authors=Giannis Spiliopoulos,Cesar Ducruet,Leonardo Millefiori,Paolo Braca,Dimitris Zissis
|dblpUrl=https://dblp.org/rec/conf/edbt/SpiliopoulosDMB24
}}
==Detecting the Spatiotemporal Characteristics of the Supply-chain Disruption and Estimating its Short Term Effects==
https://ceur-ws.org/Vol-3651/BMDA-4.pdf
Detecting the spatiotemporal characteristics of the supply-chain
disruption and estimating its short term effects
Giannis Spiliopoulos∗ César Ducruet∗ Leonardo M. Millefiori∗
Intelligent Transportation Systems French National Center for Scientific NATO STO Centre for Maritime
Lab, University of the Aegean Research (CNRS), UMR 7235 Research and Experimentation
Ermoupolis, Syros, Greece EconomiX, University of Paris (CMRE)
gspiliopoulos@aegean.gr Nanterre, France La Spezia, Italy
cesar.ducruet@economix.fr leonardo.millefiori@cmre.nato.int
Paolo Braca∗ Dimitris Zissis†∗
NATO STO Centre for Maritime Intelligent Transportation Systems
Research and Experimentation Lab, University of the Aegean
(CMRE) Ermoupolis, Syros, Greece
La Spezia, Italy dzissis@aegean.gr
paolo.braca@cmre.nato.int
Figure 1: Visual representation of the data transformation flow.
ABSTRACT potential congestion buildup and where cargoes should be rerouted
Ships are often considered as the backbone of the global economy. to. Herein, we analyze a large-scale high-resolution mobility data
A fundamental unresolved problem is how to best operate fleets, set of more than 7,000 container ships, collected over an extended
given a sudden increase in demand, such as that reported following period of 36 months, covering the entire globe, so as to measure
the first months of the COVID19 pandemic. Advancing our knowl- quantitatively the effects of the pandemic post-hoc on the supply
edge of the supply chain’s delicate equilibrium between demand chain. To further understand these fine-grained mobility patterns,
and supply, requires analyzing huge amounts of ship-related posi- we introduce a mobility model for calculating ship presence times
tional data, thus revealing which areas should be avoided due to (or waiting times) at a global scale. We then reveal the congestion
points, which strongly correlate with port waiting areas and anchor-
∗ These authors contributed equally to this work
† Corresponding author ages. We analyse the data to reveal how times at port areas were
affected by the rising number of ships waiting to load or unload
Copyright © 2024 for this paper by its authors. Use permitted under Creative Commons cargo. Following this, we transform this data into a ‘port to port’
License Attribution 4.0 International (CC BY 4.0). Published in the Proceedings of the
Workshops of the EDBT/ICDT 2024 Joint Conference (March 25-28, 2024), Paestum,
graph mapping the international flow of containerised trade. We
Italy.
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
apply methods of graph theory, complex networks, and multivari- the hidden patterns is intrinsic to trajectory data mining, and the
ate statistics to unravel the hidden relationships between global seminal work in [28] covers the field in detail. This growth has been
maritime structure and ship time distribution. This analysis is novel boosted also by the availability of data from large sensor networks,
in respect to the size of the data analyzed, the algorithmic approach such as the AIS, which has provided researchers with enormous vol-
and the impact of the results which reveal some affinities between umes of information for the study of maritime transportation and
pre-COVID and post-COVID shipping patterns. the maritime industry in general. According to the authors of [10],
one of the main scientific topics discussed in the maritime literature
is indeed the applications of big data techniques to AIS. Recently,
1 INTRODUCTION
mostly due to the COVID-19 pandemic, but also to other events,
The supply chain is a complex network involving suppliers, trucks, such as the Suez canal blockage in 2021 and the Red Sea crisis in
distribution centers, warehouses, logistics centers and vectors, 2024, the maritime transportation has been often disrupted [19, 23]
working together in concert to deliver goods to the customers. and several studies were conducted to measure effects of this disrup-
Today, this complex network is distributed all around the world. tion [15, 17]. The strong academic interest to study large mobility
With more than 80% of global trade by volume and up to 70% of its data at scale combined with necessity for fast decision making dur-
value being carried by the shipping industry, vessels can be seen ing the pandemic has accelerated advancements in the field [29].
as the backbone of the global supply chain. Thus, shipping can be In this work, we use big data analytics and graph analysis to bet-
viewed as a barometer for the global economic climate, and any ter understand the disruption in the supply chain. The utilisation of
reduction in activity is expected to have a cascading effect on the big data in AIS analysis is an area of research that has received a lot
global economy. On the other hand, ports are located at the nexus of attention [3, 24, 30] recently. For instance, in [22], authors evalu-
of the supply chain and connect global and regional actors. ate the performance of clustering algorithms for route modelling
The COVID-19 pandemic and containment measures had a sig- on a full year global AIS dataset, and in [16], the KDE technique is
nificant impact on global maritime traffic in the short term. Several adapted to map-reduce paradigm to compute seaports’ extended ar-
studies have attempted to quantify these effects and assess the size eas of operations from AIS data. Other examples are [20], where an
of this reduction. However, after the initial shock of the pandemic, image analysis on density maps to detect traffic flows is introduced,
when customer demand rebounded, vessels stacked up high with as well as [27], in which authors introduce trip semantic objects and
containers could be seen anchored outside many of the world’s the use of density based clustering to identify clusters of way-points
major ports. and stops. In a complementary to this work approach, authors in [4]
Starting from the idea that shipping traffic data can be used build voyage graph feature time-series (VGT) to study their evolu-
to assess the effects of restrictive measures on the global supply tion from a time-series perspective. In this work, we quantify and
chain, in this work we focus on revealing the major delay points allocate to ports the effects of each vessel slow-down in range by
in the maritime global supply network, as well as attempting to introducing the waiting and approaching time indicators. We study
answer the question if these correlate with major port locations. their evolution over time to understand which ports are affected
We also explore if congestion at ports is simply related to a sudden the most.
increase in ship traffic and evidently quantify the increase in delay Graph theory, and its widely known extension known as com-
times. Along these lines, we attempt to reveal how the pandemic plex networks, can be employed to characterize (port) nodes by
has affected ports over an extended period of time and attempt to providing a hierarchy of centrality/accessibility in the container
reveal the trajectory of recovery. shipping network. Early applications provided global-network mea-
The main technical challenge is that of a big-data mining task of sures of connectivity [11, 12, 26], as well as a cartography of degree
transforming huge amounts of geospatial data–as collected from or betweenness centrality [9]. Research on the relationship between
vessels using the Automatic Identification System (AIS)–into a de- centrality and other port performance indicators remains relatively
scriptive and compact data model, that can be used for identifying scarce in the literature [13, 14, 25], usually confirming the strong
the underlying relationships or patterns. In our case, the patterns correlation between degree centrality (i.e., number of connections
are those of normal port to port traffic connections. Our approach to other ports) and weighted degree (i.e., total traffic in twenty-foot
relies on data transformations and distributed raster-based analyt- equivalent units [TEUs]). Thus, in the present study, we innovate
ics as a first step to reduce the size of the data, followed by graph by applying a statistical analysis of traffic, centrality, and time in-
analysis to reveal the hidden patterns in the data. dicators. It complements the work of Ducruet and Itoh [8] on the
The main methodological contribution of this work is showing statistical relationships at stake between ship time, port centrality,
that mobility data can be processed to shed light on the temporal and port traffic by focusing on a specific event and its supply chain
and spatial characteristics of the supply chain as a network at the consequences. A recent review of the field [7] also showed that
global scale. To the best of our knowledge, no major study has within a corpus of 212 papers about shipping networks published
attempted to analyze such a large dataset with this aim before. between 2007 and 2022, nearly 20% concerned the topics of crisis
and vulnerability, i.e., the second largest category after "network
1.1 Related work and contribution structure".
Over the last years, there has been an exponential growth of scien-
tific publications related to maritime traffic analysis involving big
data analytics and/or novel Artificial Intelligence (AI) techniques.
For instance, the analysis of vessel mobility data to understand
2 METHODOLOGY segment. We annotate as idle any segments whose average speed
is less than 2 knots, and finally we sum up the total time of idle
2.1 Automatic Identification System
segments for each cell and month. High values of idle times are
AIS was originally designed as a collision avoidance system for indicators of choke points for shipping traffic, and the computation
ships. Since 2002, the International Maritime Organisation (IMO) of idle time rasters on a monthly basis allows us to characterize
has made compulsory for all vessels with a tonnage including how the distribution changes over the three-year period consid-
and above 300 gross tonnage to be equipped with a class-A AIS ered. Anticipating the results, we observe that high idle time cells
transceiver. At its core, each AIS transceiver sends and receives typically appear near major container ports and canals.
positional reports (i.e., types 1, 2, 3 and 18) every few seconds Connecting waiting areas with ports. To further investigate
via VHF. The messages contain information about each vessel’s this behavior and explicitly connect cells of high cumulative idle
identity, location, course and speed. Since 2006, the lower-power time with ports, we performed a nearest neighbor analysis [2]
(and lower-cost) class-B transceiver was introduced, allowing also to assign each cell to its nearest port. Then, if the cell is located
smaller vessels to use the AIS, even if with lower performance and within a 100 km range of any top-50 ports 1 (in terms of annual
priority than commercial fleets, which operate strictly on class- reported volume [1]) the cell is reassigned to its closest top-50 port.
A transceivers [21]. The transmission rate of AIS ranges from 2 Again, anticipating a bit the results of our analysis, we notice an
seconds, for fast moving vessels or maneuvering vessels equipped increase of cumulative idle time around major ports both in terms
with a class-A transponder, up to 3 minutes for anchored or moored of cumulative values and number of cells where this happens.
vessels. Measuring in-port and approaching time. The previous steps
For this study, we make use of AIS positional reports of container leave us with the congestion epicenters near major ports, where
ships traveling across the globe for the years 2019, 2020 and 2021. vessels wait to enter the port. The epicenters are located, in most
cases, near the ports. However, as congestion increases, the wait-
2.2 Data transformation ing areas expand vastly following different patterns with respect
To unravel the hidden information of global supply chain perfor- to topology and other local characteristics. It is also possible that
mance from raw AIS messages, we employ a multi-step sequential a non-negligible number of vessels sheltered themselves in these
data mining process. Our main goal is understanding if supply chain areas and never entered the closest major port. To confirm or reject
disruption is measurable and correlated with port activities. Then, this hypothesis, we used accurate information about the end of
we also investigate what are the intrinsic characteristics of these each itinerary. The AIS protocol supports messages that include
ports, to understand if they can be possibly used in a predictive information about the destination port, but unfortunately it cannot
fashion as indicators of future disruptions, so that fleets can be be considered as a reliable source of information, as it is manually
rerouted suitably. In our case study, the effects of a disruption in entered by the crew, without following a specific standard, mak-
global supply chain are not known beforehand. We introduce a pro- ing it thus extremely prone to errors. To tackle this problem, we
cess (Fig. 1) to infer a global supply chain network graph from AIS performed a retrospective analysis on the data to identify the ports
mobility data. Then, we apply advanced graph analytics to identify of origin and destination for each trip, and we calculated the exact
port typologies and changes over its connections to measure effects time of approaching 300 km to destination, as well as the exact time
of disruptions over a three-year period. of each vessel entering the port across all itineraries that reach any
Data cleaning & conditioning The first step in the process is of the top-50 ports. Then, we calculate for each itinerary the total
a cleaning task to ensure that records comply with protocol stan- time spent within a 300 km radius and the time spent in the port.
dards and reject records with missing values. Then, we apply a The 300 km radius ensures that we account for any intentional or
geo-fencing technique to select records located within port areas, unintentional delay that may occur for any vessel before it reaches
and exclude them from the identification of waiting areas part of its final destination. This radius is selected so that all waiting areas
the analysis. To facilitate the numerical calculations, all positional of the first part of our analysis are included. The time in port reflects
reports are re-projected into the Web Mercator (EPSG:3857) coordi- the operational time of a vessel calling a port and it captures all
nate system. time required to moor at berth and perform all kinds of loading and
Raster-based analysis. To identify waiting areas from AIS mes- unloading operations and exit the port.
sages, we use a raster-based analysis. We first define the raster Defining the waiting time network. Maritime flows can be
characteristics, such as the shape and size of its cells. For the analy- modeled as a graphical structure G, where the ports (𝑣) are the
sis performed in this work, the raster consisted of square-shaped nodes (or vertices), which are connected by inter-port connections
cells of a 9.7 km side length, each one of them covering an area of (𝑒) as links (or edges), so that G = (𝑣, 𝑒) [6]. The connections among
approximately 100 km2 on average with respect to the projection ports are in general not known a priori and can change over time,
systems’ distortion. Then, we assign the AIS messages to the grid but in principle they can be learned by inspection of the AIS data by
cells, by splitting each trajectory into segments that match the grid looking at vessels navigating from one port to another. Each vertex
definition (i.e., each segment is allocated to exactly one cell and 𝑣 in the graph G stores static features and summary statistics of the
it is annotated with the cell’s id). Each segment consists of either port’s traffic flow strength that corresponds to the graph weights
two consecutive AIS messages or a grid intersection point and an and they are calculated on quarterly basis. The static features are the
AIS message,where the location and timestamp of the intersection
point are interpolated assuming constant speed. Then, we calculate 1We consider the top-50 ports to be representative of the whole port system in their
the average speed required for a vessel to cover the distance of each proximity.
port identifier and country each port belongs to, while the summary top of the diagram. Interestingly, we can see the fluctuation in the
statistics measure the number of vessels calling the port, as well positions, with ports switching their ranking throughout the year.
as their aggregated maximum capacity. Each link 𝑒 ∈ G consists To understand the characteristics of the ports and areas of delay,
of the pair of ports identifiers it connects, as well the number of we move onto the second part of the analysis, which makes use of
voyages on this connection and their cumulative maximum TEU graph theory and PCA.
capacity and aggregated time indicators. Those statistics are also
calculated in correspondence to nodes on a quarterly basis.
Graph analysis To assess the level of disruption on port con- 3.1 Single and multivariate linkage analysis
nections, we rely on quarterly created summary statistics for nodes We apply PCA to two distinct datasets: 1) based on static port char-
(𝑣) from the previous step, and we calculate the differences for the acteristics in Q4 2019; and 2) based on absolute changes of these
total and the in port time between the last quarter (Q4) of 2019 characteristics between Q4 2019 and Q1 2020. Both datasets include
and the first quarter (Q1) of 2020. We apply linear transformations time evolution as a means of checking its affinity with other vari-
to summary statistics to define supply chain port characteristics ables, which are its potential determinants. The two PCAs provided
such as the number of vessels calls (frequency) and total vessel interesting results, with 72.1% of variance contained in the three
traffic (frequency × vessel capacity in TEU). We complement our first components for static variables (with eigenvalues > 1), and
dataset with graph-theoretical indices calculated for all network 76.0% for the first four components for dynamic variables (with
nodes and both quarters, namely the degree centrality (number of eigenvalues > 1). Figure 3 represents the distribution of variables
shipping links), betweenness centrality (number of occurrences on along the two first components for each dataset (left, static; right,
shortest paths in the graph), and inverse clustering coefficient (local dynamic). Interestingly, worsening time is opposed to traffic and
hub power). The average port time in Q4 2019 will also be used centrality level/growth in the two figures. This is even truer for dy-
as a pre-existing characteristic. We apply a Principal Components namics, where calls (trip_dif) and traffic (teu_dif) are more directly
Analysis (PCA) to all nodes to the (Q4) of 2019 quarter and quar- opposed to port time evolution. Another difference between the two
terly calculated residuals to reveal the hidden trends at stake in the PCAs is the opposition between connectivity changes and traffic
shipping traffic data. PCA is a statistical method serving to unravel changes along the second component (vertical axis) for dynamics.
a limited number of unobserved (latent) variables among a set of It means that although growing ports in general witnessed reduced
observed, correlated variables [5]. Such unobserved variables, often port time, those with growing connectivity tended to increase port
called principal components or “factors”, constitute the basis of a time, contrary to ports increasing traffic. Lastly, the static analysis
clustering analysis that will distribute observations (port nodes) shows that ports with longer times in Q4 2019 were also the ones
among distinct groups. The clusters, i.e. groups of ports of similar increasing port time in Q1 2020 (vertical axis).
operational behaviour (see bottom right legend in Figure 4), are The situation of each port in the observed trends is revealed
then confronted to initial variables to best describe their trends and by means of a hierarchical clustering analysis, which is applied to
characteristics. The next step is to illustrate the typology by means the main components of each PCA, to produce a typology. This
of a single linkage analysis. This method serves to highlight the is combined with a single linkage analysis, to test whether the
main hubs and their “nodal regions” by keeping only the largest obtained types have a specific position in the network’s backbone.
traffic flow link of each port in the graph. Finally, a multiple re- The dynamics-based typology provides the picture of world ports
gression looks at the determinants of time evolution based on port reported in Fig. 4. It considers absolute changes of port character-
characteristics. istics as for the second PCA, and the single linkage analysis is
based on Q1 2020. The most impacted category (yellow) is marked
by drastic traffic decline, slight reduction of centrality, and the
strongest increase of total and in-port time. It includes a vast ma-
3 RESULTS jority of gateway ports (Le Havre, Constanta, Koper, Alexandria,
Areas where ships “wait” evidently depict problematic spots in the Liverpool, Felixstowe, Zeebrugge, Fos, Los Angeles, Long Beach,
supply chain. In order to reveal the spatiotemporal characteristics of Tianjin, Lianyungang, Kobe, Ho Chi Minh, Manila) as well as Hong
delay areas globally, we first define and quantify areas where ships Kong and Port Klang. Except from the latter two ports, these gate-
are idle for long periods, as these can be an indicator of disruptions ways have, in general, a limited role in the architecture of nodal
in the supply and demand balance [17]. Our analysis focused on first regions, due to their specialization in import/export cargoes. An-
understanding if these areas overlapped with specific port areas other category has lost similar amounts of traffic on average (red)
and then to further understand if these had specific characteristics. but such ports slightly increased their centrality. Several of them
As a first result, which may have been expected, we confirmed are large hub ports polarizing their respective nodal region, the
that waiting areas are close to port locations. All top-30 locations largest being, like in the previous figure, Singapore, Busan, and
are within 80 nautical miles range from ports, and we can assume Rotterdam. Like for the other categories (green, blue), these ports
reasonably that ships stationing in these areas are waiting to enter experienced a slight increase of total port time and small decrease
the port. This result is in line with reports and papers reporting the of in-port time. While they also lost traffic, the secondary hubs
increase turn around in port areas [18]. (green) gained enormous centrality in Q1 2020, contrary to what
In Fig.2, we illustrate a Sankey Diagram ranking ports according we can call second-tier hubs (blue), which have the opposite profile.
to the measured total idle time over the last time period of the There is no apparent geographic or functional logic in those two
analysis, where ports with longer waiting times are located at the categories, which are disseminated across regions and contain both
Figure 2: Visual representation in terms of total idle time for the top-10 ports of each month across 2021 year. Line width
represents the value of total idle time. AMB:Ambarli, ANT:Antwerp, BRE:BremerHaven, BUS:Busan, CAP:Cap Town, DAL:Dalian,
DAR:Dar ES Salaam, HAM:Hamburg, HON:Hong Kong, JAK:Jakarta, JEB:JEBEL ALI, LOM:LOME, LOS:Los Angeles, MAN:
Manzanillo, NIG: Nigbo, OAK:Oakland, PIR: Piraeus, POR: Port Lang, Rot: Rotterdam (blue: Waalhaven, orange:Maasvlakte),
SAV: Savannah, SHA: Shanghai, SIN: Singapore, VAL: Valencia.
gateway ports and transshipment ports. The loss of centrality (blue) centrality indicators (cf. Principal Component Analysiss (PCAs)),
is, still, relatively common to European ports while the increase of it expresses a specific dimension of port connectivity, namely the
centrality (green) is better found in Asia. ability to polarize neighboring, or adjacently connected, ports.
A counter-intuitive result is the negative effect of city size (pop-
3.1.1 What determined port time changes in 2020? A multiple re- ulation) on time evolution in both models, as the inclusion of this
gression analysis is applied in two steps, each being a model focus- variable was meant to test the role of potential congestion played
ing on a distinct independent variable: total port time difference by cities on port operations, in terms of lack of space and density.
(model 1) and in-port time difference (model 2), as shown in Ta- This can be explained by our focus on the top of the port hierarchy,
ble 1. As a matter of fact, among the selected dependent variables, where most ports are in fact major metropolitan areas. Another
only two have a statistically significant effect. It is the case of in- commonality between the two models is the negative influence of
port time in model 1 (0.05 significant), which increased total port total port time and the positive influence of in-port time. In-port
time difference between Q4 2019 and Q1 2020. The other case is the time is a component of total port time, but it better represents the
regional dummy Africa in model 2, which increased in-port time dif- core activity of the port, as it is the closest to the length taken by
ference. Despite the low significance of other dependent variables, terminal operations. This crucial component of the whole transport
some of them may be discussed according to the direction of their chain, if prolonged, will inevitably have strong consequences on
effect on time evolution. Among the ones that deserve attention, the rest of the chain, as seen with its positive impact on total time
inverse clustering coefficient stands out as it has the same, negative difference (slowdown, queuing), which includes the water vicinity
effect on time evolution, and is near-to-significant in both models. of the port (e.g. port entrance, access channel). Thus, ports with an
It means that ports ensuring stronger hub functions before the already high in-time (turnaround time) have witnessed worsening
crisis have witnessed lesser congestion and, even, more fluid cargo operations (prolonged times) during the COVID-19 crisis. At the
transfers. Such a result is in line with the single linkage analyses, contrary, total port time had a negative effect on time evolution
showing that pivotal hubs ensure and maintain their domination in both models. While such a result may seem to contradict the
towards other ports, often within a certain geographic radius (nodal former, it should be understood in the light of other port variables
regions). It also confirms the work of Ducruet and Itoh [8] about the in each model. In model 1, the negative effect of total port time on
determinants of ship turnaround times on a long period. Although total port time difference goes along with a negative effect of port
this measure is very much correlated with port traffic and other size (calls, TEUs), meaning that large, busy ports in Q4 2019 (but
(a) (b)
Figure 3: Principal component analyses
Table 1: Determinants of port time evolution.
Model 1 Model 2
total time difference in-port time difference
Estimate Std. err. 𝑡-value Pr(> |𝑡 |) Estimate Std. err. 𝑡-value Pr(> |𝑡 |)
(Intercept) 1.409 00 1.201 00 1.173 0.2440 −0.720 11 0.524 60 −1.373 0.1735
Population −0.024 69 0.035 61 −0.694 0.4899 −0.008 62 0.015 55 −0.554 0.5809
Calls −0.093 89 0.232 51 −0.404 0.6874 0.164 42 0.101 56 1.619 0.1092
TEUs −0.127 66 0.141 70 −0.901 0.3702 0.024 27 0.061 89 0.392 0.6960
Distance −0.037 87 0.088 34 −0.429 0.6692 0.038 61 0.038 59 1.000 0.3200
Betweenness Centrality 0.023 93 0.079 31 0.302 0.7636 0.010 52 0.034 64 0.304 0.7621
Degree Centrality 0.460 94 0.332 03 1.388 0.1687 −0.238 50 0.145 03 −1.645 0.1038
I.C.C.† −0.048 19 0.031 99 −1.506 0.1357 −0.016 58 0.013 97 −1.186 0.2388
Total time −0.167 66 0.135 24 −1.240 0.2185 −0.083 76 0.059 07 −1.418 0.1599
In-port time 0.581 33 0.263 06 2.210 0.0298* 0.084 71 0.114 91 0.737 0.4630
Africa −0.025 91 0.699 50 −0.037 0.9705 0.525 12 0.305 55 1.719 0.0893
Americas 0.180 97 0.664 12 0.273 0.7859 0.147 99 0.290 09 0.510 0.6113
Asia 0.405 55 0.665 50 0.609 0.5439 0.194 31 0.290 69 0.668 0.5057
Europe 0.087 54 0.662 63 0.132 0.8952 0.264 30 0.289 44 0.913 0.3638
† Inverse clustering coefficient
not necessarily the most central, as seen with the positive influence a negative influence of degree centrality (numerous connections).
of betweenness and degree) had more chance to perform better in This means that hub ports with a relatively lower size managed to
times of crisis. In model 2, the same negative effect is associated improve their core operations in the advent of the crisis.
with a positive influence of port size (worsening in-port time), and
Figure 4: Single linkage analysis and port typology – Q1 2020 and port/time evolution
4 CONCLUSION ACKNOWLEDGMENTS
Based on geospatial data mining approaches, in this paper we pre- This work was partially supported by MarineTraffic. This research
sented a methodology to determine the misbalances in the global is supported by European Union’s Horizon 2020 research and inno-
supply and demand equilibrium as captured through ship move- vation programme under grant agreement No 101092749, project
ments. The presented approach is capable, firstly, of detecting and Critical Action Planning over Extreme-Scale Data (CREXDATA).
defining the areas of potential delays, which in most cases overlap The work of L. M. Millefiori and P. Braca is supported by NATO
with the main port areas, and secondly to determine the specific Allied Command Transformation (ACT) via project “Data Knowl-
characteristics of these ports. Our approach relies on methods of edge Operational Effectiveness” (DKOE). The contribution of César
raster based analysis, graph theory and complex networks analy- Ducruet is supported by the French National Research Agency
sis. Future work will be focused on applying additional methods (ANR) through the research project No. ANR-22-CE22-0002 "Mar-
from the field of graph analysis and complex networks to similar itime Globalization, Network Externalities, and Transport Impacs
datasets. on Cities" (MAGNETICS).
REFERENCES Transportation Review 95 (2016), 326–340.
[1] World Shipping Council. 2020. The Top 50 Container Ports. https://www. [26] Deng Wei-Bing, Guo Long, Li Wei, and Cai Xu. 2009. Worldwide Marine Trans-
worldshipping.org/top-50-ports Accessed: 19-Jan-2024. portation Network: Efficiency and Container Throughput. Chinese Physics Letters
[2] Jon Louis Bentley. 1975. Multidimensional Binary Search Trees Used for Asso- 26, 11 (nov 2009), 118901.
ciative Searching. Commun. ACM 18, 9 (sep 1975), 509–517. [27] Zhaojin Yan, Yijia Xiao, Liang Cheng, Rong He, Xiaoguang Ruan, Xiao Zhou,
[3] Samuele Capobianco, Leonardo M. Millefiori, Nicola Forti, Paolo Braca, and Peter Manchun Li, and Ran Bin. 2020. Exploring AIS data for intelligent maritime
Willett. 2021. Deep Learning Methods for Vessel Trajectory Prediction Based on routes extraction. Applied Ocean Research 101 (2020), 102271.
Recurrent Neural Networks. IEEE Trans. Aerospace Electron. Systems 57, 6 (2021), [28] Yu Zheng. 2015. Trajectory Data Mining: An Overview. ACM Trans. Intell. Syst.
4329–4346. Technol. 6, 3, Article 29 (may 2015), 41 pages.
[4] Emanuele Carlini, Vinicius Monteiro de Lira, Amilcar Soares, Mohammad [29] Chenghu Zhou, Fenzhen Su, Tao Pei, An Zhang, Yunyan Du, Bin Luo, Zhidong
Etemad, Bruno Brandoli, and Stan Matwin. 2021. Understanding evolution Cao, Juanle Wang, Wen Yuan, Yunqiang Zhu, Ci Song, Jie Chen, Jun Xu, Fujia Li,
of maritime networks from automatic identification system data. GeoInformatica Ting Ma, Lili Jiang, Fengqin Yan, Jiawei Yi, Yunfeng Hu, Yilan Liao, and Han Xiao.
26 (2021), 1–25. 2020. COVID-19: Challenges to GIS with Big Data. Geography and Sustainability
[5] Dennis Child. 2006. The essentials of factor analysis. A&C Black. 1, 1 (2020), 77 – 87.
[6] Pasquale Coscia, Paolo Braca, Leonardo M. Millefiori, Francesco A. N. Palmieri, [30] D. Zissis, K. Chatzikokolakis, G. Spiliopoulos, and M. Vodas. 2020. A Distributed
and Peter Willett. 2018. Multiple Ornstein–Uhlenbeck Processes for Maritime Spatial Method for Modeling Maritime Routes. IEEE Access 8 (2020), 47556–47568.
Traffic Graph Representation. IEEE Trans. Aerospace Electron. Systems 54, 5 (2018),
2158–2170.
[7] César Ducruet. 2023. Shipping network analysis: State-of-the-art and application
to the global financial crisis. Routledge. 300 – 333 pages.
[8] César Ducruet and Hidekazu Itoh. 2022. Spatial network analysis of container
port operations: the case of ship turnaround times. Networks and Spatial Eco-
nomics (July 2022).
[9] César Ducruet and Theo Notteboom. 2012. The worldwide maritime network of
container shipping: Spatial structure and regional dynamics. Global Networks 12,
3 (2012), 395 – 423.
[10] Ziaul Haque Munim et al. 2020. Big data and artificial intelligence in the maritime
industry: a bibliometric review and future research directions. Maritime Policy
& Management 47, 5 (2020), 577–597.
[11] Yihong Hu and Daoli Zhu. 2009. Empirical analysis of the worldwide maritime
transportation network. Physica A: Statistical Mechanics and its Applications 388,
10 (2009), 2061 – 2071.
[12] Pablo Kaluza, Andrea Kölzsch, Michael T. Gastner, and Bernd Blasius. 2010. The
complex network of global cargo ship movements. Journal of the Royal Society
Interface 7, 48 (2010), 1093 – 1103.
[13] Dong-Joon Kang and Su-Han Woo. 2017. Liner shipping networks, port charac-
teristics and the impact on port performance. Maritime Economics & Logistics 19
(2017), 274–295.
[14] Arom Kim and Jing Lu. 2015. A study on the effects of network centrality and
efficiency on the throughput of Korean and Chinese container ports. In ICTE
2015. 760–769.
[15] David March, Kristian Metcalfe, Joaquin Tintoré, and Brendan J Godley. 2021.
Tracking the global reduction of marine traffic during the COVID-19 pandemic.
Nature communications 12, 1 (2021), 2415.
[16] L. Millefiori, D. Zissis, L. Cazzanti, and G. Arcieri. 2016. A distributed approach
to estimating sea port operational regions from lots of AIS data. In 2016 IEEE
International Conference on Big Data (Big Data). IEEE, 1627–1632.
[17] Leonardo M. Millefiori, Paolo Braca, Dimitris Zissis, Giannis Spiliopoulos, Stefano
Marano, Peter K. Willett, and Sandro Carniel. 2021. COVID-19 impact on global
maritime mobility. Scientific Reports 11 (sep 2021), 18039.
[18] Theo Notteboom and Thanos Pallis. 2020. IAPH-WPSP Port Economic Impact
Barometer. Technical Report. International Association of Ports and Harbors
(IAPH).
[19] Richard Partington. 2024. What is the Red Sea crisis, and what
does it mean for global trade? The Guardian, Jan 3rd 2024
https://www.theguardian.com/world/2024/jan/03/ what-is-the-red-sea-
crisis-and-what-does-it-mean -for-global-trade,. Accessed: 19-Jan-2024.
[20] Roeland Scheepens, Christophe Hurter, Huub Van De Wetering, and Jarke J
Van Wijk. 2015. Visualization, selection, and analysis of traffic flows. IEEE
transactions on visualization and computer graphics 22, 1 (2015), 379–388.
[21] Solas. 2002. Solas Chapter V. Regulation 19 - Carriage requirements for shipborne
navigational systems and equipment. https://assets.publishing.service.gov.uk/
media/5a7f0081ed915d74e33f3c6e/solas_v_on_safety_of_navigation.pdf. [Ac-
cessed: 19-Jan-2024].
[22] Giannis Spiliopoulos, Konstantinos Chatzikokolakis, Dimitrios Zissis, Evmorfia
Biliri, Dimitrios Papaspyros, Giannis Tsapelas, and Spyros Mouzakitis. 2017.
Knowledge extraction from maritime spatiotemporal data: An evaluation of
clustering algorithms on Big Data. In 2017 IEEE International Conference on Big
Data (Big Data). IEEE, 1682–1687.
[23] United Nations Conference on Trade and Development (UNCTAD). 2023. Re-
view of Maritime Transport. https://unctad.org/publication/review-maritime-
transport-2023.
[24] Xiaoguang Wang, Xuan Liu, Bo Liu, Erico N de Souza, and Stan Matwin. 2014.
Vessel route anomaly detection with Hadoop MapReduce. In 2014 IEEE interna-
tional conference on big data (big data). IEEE, 25–30.
[25] Yuhong Wang and Kevin Cullinane. 2016. Determinants of port centrality in
maritime container transportation. Transportation Research Part E: Logistics and