Ocean Engineering 255 (2022) 111427

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

A comprehensive ship weather routing system using CMEMS products and

A* algorithm
Manel Grifoll a, *, Clara Borén b, Marcella Castells-Sanabra b
a
Civil and Environmental Engineering Department, Universitat Politècnica de Catalunya (UPC-BarcelonaTech), Barcelona, 08028, Spain
b
Nautical Science and Engineering Department, Universitat Politècnica de Catalunya (UPC-BarcelonaTech), Barcelona, 08003, Spain

A R T I C L E I N F O A B S T R A C T

Keywords: We describe the implementation of a comprehensive software for Ship Weather Routing referred to as SIM­
Weather ship routing ROUTE. The A* pathfinding algorithm is used to optimize the sailing route as a function of the wave action. The
A-star algorithm aim of the software is to provide a comprehensive, open and easy tool including pre- and post-processing for ship
CMEMS
weather routing simulations. The software is constructed considering the Copernicus Marine Environment
Ship emissions
SIMROUTE
Monitoring Service (CMEMS) wave predictions systems which are available for free use. The code provides the
optimized route and the minimum distance route together with additional modules to compute ship emission and
safety on navigation monitoring. SIMROUTE has been tested in several cases using different CMEMS products
over short and long distances. The comprehensive structure of the code enables it to be easily modified to include
additional ship wave resistance models and the effect of the water currents and winds on navigation. SIMROUTE
is also used for academic purposes, providing skills for ship routing optimization in the framework of standards
of training, certification and watchkeeping (STCW) for competence-based maritime education and training. Due
to the simplicity of its use, SIMROUTE is a good candidate for benchmarking strategies and inter-comparison
exercises with advanced methods for ship weather routing. This contribution highlights the technical aspects,
code organization and structure behind SIMROUTE, demonstrating its capabilities through examples of route
optimization.

1. Introduction for ship routes. The emerging field of the autonomous ships also shows
an increasing interest for optimization tools in the framework of per­
There is a growing concern to reduce emissions in order to mitigate formance management systems that includes analysis and real-time
climate change. The environmental impact of transport is significant remote monitoring, hindcast/forecast wave conditions and make deci­
because it is the major user of energy. Shipping is the prevalent transport sion support (e.g. Ruth and Thompson, 2022; Zhao et al., 2022).
mode for overseas freight and is frequently recognized as a sustainable, From the shipping industry point of view, the minimization of
energy-efficient and relatively environmentally friendly means of operating costs is a multi-faceted problem, which involves fleet man­
transport (DfT, 2004). However, shipping is still a substantial source of agement, vehicle routing problems such as scheduling or ship weather
greenhouse gas emissions (Chapman, 2007). In this respect, the shipping routing, among other strategies. Ship Weather Routing (SWR) is defined
industry is not exempt from the need to reduce emissions, and several as the development of an optimum sailing course and speed for ocean
initiatives have been taken by stakeholders and academics to promote voyages based on nautical charts, the forecasted sea conditions, the
sustainable growth (Goldsworthy and Goldsworthy, 2015; Kanellos captain’s experience and the individual characteristics of a ship for a
et al., 2014; Wang et al., 2016; Zakaria et al., 2022; Zis et al., 2014a, particular route (Simonsen et al., 2015). Academic research has focused
2014b). In the context, global decarbonisation target specific shipping on ship routing optimization through pathfinding algorithms (Hinnen­
sectors and the academia addresses several pertinent questions on in­ thal and Clauss, 2010; Mannarini et al., 2016; Shin et al., 2020;
formation collection and planning routes (Wu et al., 2021; Zakaria et al., Simonsen et al., 2015; Szłapczynska and Smierzchalski, 2009; Taka­
2022; Zis et al., 2020b). In addition, a major factor of competitiveness in shima and Mezaoui, 2009; Zhao et al., 2022; Zis et al., 2020a), which
the maritime industry is the minimization of time and fuel consumption take into account meteo-oceanographic forecasts (i.e. wind, waves or

* Corresponding author.
E-mail addresses: [email protected] (M. Grifoll), [email protected] (C. Borén), [email protected] (M. Castells-Sanabra).

https://doi.org/10.1016/j.oceaneng.2022.111427
Received 12 October 2021; Received in revised form 17 March 2022; Accepted 25 April 2022
Available online 19 May 2022
0029-8018/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

currents predictions). Some of these contributions have been tested 2. Materials and methods
through a “proof-of-concept” based on oceanic distances (e.g. Hinnen­
thal and Clauss, 2010; Lin et al., 2013; Simonsen et al., 2015), but The software structure consists in a sequential execution of Python
benchmarking tests have not been found in the literature covering scripts. A flow chart to guide SIMROUTE users is shown in Fig. 1,
different regional areas. A noticeable effort was made by (Mannarini including the folder organization (i.e. in/, out/ and /storeWaves) and the
et al., 2016), using an optimization algorithm based on the graph-search post-processing tools. Input variables are included in a unique .py file:
method with time-dependent edge weights in the Mediterranean Sea. In params.py. This set-up file establishes the characteristics of the mesh,
this respect, SIMROUTE presents a flexible structure, permitting all the period of simulation, wave effect on the navigation model and ship ve­
CMEMS products (which cover all the world’s seas) in a comprehensive locity, among other parameters (see Fig. 1 for a short description). The
design, which creates a fast-executing tool from the user perspective. code execution is designed following three steps of .py scripts. The script
SIMROUTE may enrich the ship weather routing field of investigation, get_waves_CMEMS.py is a pre-processing file that downloads sea surface
which has been receiving increasing interest from academics in recent waves variables in daily files configuration using motuclient from
years (see review in Zis et al., 2020a). The code also provides a set of CMEMS repository. Secondly, make_waves.py interpolates the wave in­
visualization tools for easy and direct inter-comparison of new SWR formation into a specific mesh generated as a function of the boundaries
methods, which may include artificial intelligence and machine learning and the mesh increment size. Finally, main.py executes the optimization
in the framework of autonomous ship development alternatives (e.g. algorithm, resulting in two alternatives: an optimized route and a min­
Kuhlemann and Tierney, 2020). imum distance route.
The above-mentioned contributions have established an important
source of knowledge for seeking efficient ship routes. With a comparable 2.1. Wave information source
motivation in mind, we present SIMROUTE, an open-source, versatile
and computationally efficient software for modelling optimal weather The wave information files are downloaded from the European
ship routes. SIMROUTE targets one of the aspirations of ship weather Union’s Earth observation programme Copernicus. The Copernicus
routing by minimizing time of navigation and, in consequence fuel Marine Environment Monitoring Service (CMEMS) provides full, free
consumption and emissions. SIMROUTE has been under active devel­ and open access data and information related to the physical state of the
opment since 2014 (Grifoll et al., 2018; Grifoll and Martínez de Osés, global ocean. Several ocean wave products are provided in the CMEMS
2016) and offers modular functionalities including economic assess­ catalogue (European Commission, 2021) covering different geograph­
ment, safety in navigation, and emissions estimations oriented to ical regions. Fig. 2 shows the regional coverage of the different products
determine the benefit produced by the optimized route, including pre- excluding the Arctic and Global products also used in SIMROUTE. The
and post-processing tools for direct and easy use. The Python pro­ get_waves_CMEMS.py script downloads the wave files in netcdf format
gramming language was chosen as the basis for SIMROUTE due to its provided by different CMEMS products (forecasting data sets) as a
flexible, free, and cross-platform-compatible nature. In addition, a function of a specific flag (wave_prod). This script also uses the param.py
MATLAB code version is available that is more oriented to academic to establish and trim the geographical area of the wave fields. Down­
purposes. The open-source code (GPL Licenses), together with support loaded wave files are stored in storeWaves/in daily format. The products
materials, is available in GitHub.com/ManelGrifoll/SIMROUTE. Code has available in the SIMROUTE structure are summarized in Table 1.
been tested using Python 3 and the specific imported modules are Motuclient is used to extract and download data through a Python
included in the headers of the code. This package also provides plotting command line. This client enables the handling and transformation of
tools and oriented modules for ship emission or the assessment of safety huge volumes of oceanographic data without performance collapse and
conditions during navigation. The objective of this contribution is to is available at the GitHub platform.
present the conceptualization, software structure, formulation and the
main features of the SIMROUTE, illustrated by examples, as an 2.2. Mesh and wave effect on navigation
open-source software for the scientific and teaching community.
The main methodologies used in SWR include the isochrone method The computational mesh is established as a function of the grid
(Hagiwara, 1982), dynamic programming (Shao et al., 2012), resolution (inc variable) and the boundary limits defined in params.py.
path-finding algorithms in grid-based approach (Mannarini et al., 2016) Once the mesh is obtained, the nodal connections possibilities are
and artificial intelligence (Maki et al., 2011) among others (Walther increased to enable smooth destinations composed from a sequence of
et al., 2016). suggest that grid-based approaches are only suitable for edges. SIMROUTE nodes are connected by 48 edges, allowing 48 distinct
short routes (i.e. coastal shipping) because of its computational effi­ directions per node (see Fig. 3) and enabling a smoothing of the sailing
ciency. The improbable smoothness of the solutions is one of the po­ routes alternatives (Cheung, 2018) This enables angular courses of a
tential drawbacks. However, SIMROUTE uses grid-based approaches A* range of 3.2◦ ‒ 14.0◦ resolution to be obtained. Singular points on the
as a trade-off between accuracy and computational time assuming mesh boundaries and corners are treated particularly to avoid
standard computational resources. Other, conceptual assumptions have non-defined mesh points when node searching. The initial and final
been considered in the development of the SIMROUTE software. SIM­ nodes are defined also in params.py (see input variables in Fig. 1). These
ROUTE assumes that the weather effect increases the total resistance nodes are referred to the mesh computed previously; so a script is pro­
acting in a vessel. Therefore, avoiding bad weather conditions will vided (i.e. find_ports.py) to convert the coordinates of the initial and final
reduce the sailing time. Also, it means that reducing the sailing time will points, including checking to determine if the node is sea or land. Also,
reduce the fuel consumption and the economic cost per voyage. In additional information is provided by this script in a command prompt
consequence, SIMROUTE optimizes the sailing time, taking into account for an iterative process on searching the initial and final nodes on the
eventual speed reduction due to the waves. In order to evaluate the sea. The discrimination between either sea or land is given by the wave
benefit of the optimization, the minimum distance route is also provided fields interpolated. Fig. 4 shows the mesh generated which the routes (i.
by SIMROUTE. e. minimum and optimal) are computed and an example of the inter­
This paper is organized as follows: after the introduction (Section 1), polated significant wave height.
a SIMROUTE description, methods and accuracy test are presented in Because the CMEMS meshes and SIMROUTE mesh differ, the wave
Section 2. Section 3 shows test cases also included in the repository. parameters (wave direction, wave period and significant wave height)
Finally, Section 4 discusses the results and contextualize the work and from the CMEMS product are linearly interpolated in the computational
Section 5 includes concluding remarks, identifying future areas of mesh (see Fig. 4). Time resolution data are provided from CMEMS
research. products (i.e. hourly or 3-hourly). The wave direction requires a

2
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

Fig. 1. Flow chart of SIMROUTE and list of input variables in params.py.

Fig. 2. CMEMS domains used in SIMROUTE. The legend shows the identification established in SIMROUTE software. The GLOBAL and ARCTIC domains are
excluded in this figure.

3
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

Table 1 particular interpolation based on Cartesian decomposition. The inter­

CMEMS products included in SIMROUTE system. * One-hour resolution for all polated wave fields are saved on an intermediate file (waves.npz by
domains except GLOBAL (3 h). default defined at params.py) in the in/ folder. Route possibilities during
CMEMS Geographical Spatial wave_prod Reference ship routing computations are linked to the availability of wave
Id. covering Resolution Flag information.
GLOBAL Global Ocean 0.083◦ × 1 Ardhuin et al. The wave, or sea state, parameters are essential in studies related to
0.083◦ (2010) evaluations of the safety and fuel efficiency of marine designs and op­
MEDSEA Mediterranean 0.042◦ × 2 Korres et al. erations (Nielsen, 2021). In this contribution, the parametrization of the
Sea 0.042◦ (2021)
wave effect on navigation is based upon three methods, namely, Man­
IBI Iberia-Biscay-Irish 0.05◦ × 3 European
Regional Seas 0.05◦ Commission narini (suggested by Bowditch), Aertssen and Khokhlov (suggested by
(2021) Lubkovsky) (Aertssen, 1975; Lubkovsky, 2009; Mannarini et al., 2013).
AENWS Atlantic European 0.03◦ × 4 European All three methods depend on the significant wave height as well as the
North West Shelf 0.014◦ Commission wave direction (see Appendix I). However, Aertssen’s method also takes
Seas (2021)
BLACK Black Sea 0.037◦ × 5 (Staneva, J.,
into account the ship’s length dimension, whereas Khoklov takes into
SEA 0.028◦ Behrens, A., account the deadweight of the ship (Borén et al., 2019) showed the
Ricker, M., & discrepancies when using different calculation methodologies on ship
Gayer, 2020) routing in terms of the sailing time during storm episodes (large sig­
BALTIC Baltic Sea 2 km × 2 6 European
nificant wave height). The Aertssen and Khokhlov methods give very
SEA km Commission
(2021) similar results whereas the first method showed considerably higher
ARCTIC Arctic Ocean 3 km × 3 7 European values of ship velocity penalization due to waves. The ship-wave
OCEAN km Commission encounter direction is computed using an edge-based spherical
(2021) method to preserve the coordinate system of the mesh.

2.3. Optimization algorithm

The pathfinding algorithm used is the A* algorithm due to its

simplicity and efficiency in computational time. This algorithm is
applied to a gridded scheme where each grid point (node) is connected
to a set of adjacent points. To each connection (edge), a weight related to
the distance is assigned. The great circle (orthodromic) track is used for
the spherical coordinates of the grid nodes. A* solves problems by
searching among all possible paths to the solution (goal) for the one that
incurs the smallest cost (least distance travelled, shortest time, etc.), and
among these paths, it first considers the ones that appear to lead most
quickly to the solution. A* is formulated in terms of a weighted mesh:
starting from a specific node of the mesh, it constructs a tree of paths
starting from that node, expanding the paths one step at a time, until one
of its paths ends at the predetermined goal node. At each iteration of its
main loop, the A* algorithm needs to determine which of its partial
paths to expand into one or more longer paths. It does this based on an
estimate of the cost (in our case, the travel time) to reach the goal node.
Specifically, A* selects the path that minimizes the total cost function
f(Nn ):
Fig. 3. Nodal connections and neighbouring scheme from departure mode ∑
n
(in grey). f (Nn ) = h(Nn ) + g(Ni ) (1)
i=1

Fig. 4. Grid mesh generation and wave interpolation.

The left‒hand figure shows the mesh generated
discerning between sea and land nodes (blue and
orange respectively). The right‒hand figure includes
the significant wave height (Hs), in meters, interpo­
lated on the mesh to ensure nodal location on a sea
point. An eventual departure point is also shown in
magenta. The region corresponds to Balearic Islands
(NW Mediterranean Sea). (For interpretation of the
references to colour in this figure legend, the reader is
referred to the Web version of this article.)

4
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

where Ni denotes the ith nodes along the path candidate, g(Ni ) the cost of waves fields and routes are available using the Cartopy python library.
going from Ni to the parent Ni − 1, h(Nn ) is a heuristic that estimates the Examples of post-processing tools based on test cases are displayed in
cost of the cheapest path from n to the goal (see Cheung, 2018 for a Figs. 6–8.
complete description). The heuristic allows longer paths to be elimi­
nated progressively during the search. For the algorithm, to find the 2.5. SIMROUTE for academic purposes
actual shortest path, the heuristic function must be admissible, meaning
that it never overestimates the actual cost to get to the nearest goal node. From the academic point of view, ship weather routing and marine
The heuristic function used in SIMROUTE is the travelling time associ­ environment protection are specific topics in all Maritime Academies
ated with the minimum distance between the origin and destination. and Universities and related to the Seafarers’ Training, Certification and
When h(Nn ) = 0, A* reduces to the Dijkstra algorithm. Watchkeeping (STCW) Code (International Maritime Organization,
The accuracy of A* implementation is tested by comparing with the 2010), as a part of mandatory training. Castells-Sanabra et al., 2019
orthodromic (or great-circle) distances. Different tests have been carried identified which STCW competences the learner will achieve using
out in order to evaluate the error of the recovered path versus the SIMROUTE software, providing skills of ship routing optimization, to
analytical formula of the orthodromic distance in terms of the distance assess the impact of the meteo-oceanographic variables on ship navi­
travelled. Four tests have been defined considering different positions gation and to highlight the relevance of ship routing in terms of the
(latitude and longitude) for the start and end point (see Table 2). Fig. 5 sailing time, safety, fuel consumption and harmful emissions for the
shows the shortest path estimated by A* on the surface of a sphere for environment. Then, two academic modules were developed: i) safety
each of the test cases. The mesh grid is equal to 15 miles, substantially restrictions and dangerous motions, and ii) ship emissions assessment.
larger than the test cases shown in the next section. The distance trav­ The SIMROUTE software is available for the academic community,
elled computed by A* and the differences obtained compared with the together with comprehensive documentation and test cases oriented to
orthodromic distance are shown in Table 2, in which differences less guide teachers.
than 0.4% are obtained. The results guarantee a proper application of
the shortest path algorithm used (i.e. A*) embedded in SIMROUTE and 2.5.1. Safety restrictions module
its application for SWR problems. The susceptibility of a vessel to dangerous phenomena will depend
on the stability parameters, ship speed, hull shape and ship size. This
means that the vulnerability to dangerous responses, including
2.4. Results files and post-processing capsizing, and their probability of occurrence in a singular sea state may
differ for each vessel. During navigation, unstable motions (surf-riding
The SIMROUTE results information consists of two sets of files and parametric rolling) may be encountered, which may lead to a cargo
including the final and intermediate results and appropriate information or equipment damage and the unsafety of the persons on board. Safety
to ensure an eventual simulation replication (see Fig. 1). On the one restrictions to avoid surf-riding and parametric rolling are implemented
hand, a set of ASCII files (.txt) includes information of the cost function into SIMROUTE according to the guidelines of the International Mari­
optimized (sailing time and distance including wave effect on naviga­ time Organization (International Maritime Organization, 2007) in order
tion), minimum distance route, great circle distance and information to to know and avoid all the dangerous nodes from the route (see formu­
the start and end point, including departure time and mesh information lation in Appendix II). safety_restrictions.py is a post-processing tool to
(Res_route.txt). Metadata_route.txt includes information to replicate the identify both dangerous unstable motions in the optimized route (see
simulation (e.g. wave fields, wave effect on navigation formulation, example in Fig. 9).
etc.). Finally, the file Route_route.txt includes the nodal information of
the optimized route (longitude, latitude and wave conditions per each 2.5.2. Ship emission module
node visited by the optimized route). On the other hand, the .npz file (a A ship emission assessment module is included in SIMROUTE as a
file format by numpy Python library that provides storage of array data post-processing file. As an example, the emissions calculation method­
using gzip compression) is provided to allocate all the input/output ology implemented in the SIMROUTE software is STEAM2 (Jalkanen
variables of the simulation, including wave interpolated information. et al., 2012), which takes into account the influence of the ship speed,
The simulation name, which is linked with the output file names, is engine load, fuel sulphur, multi-engine setups, abatement method and
provided also in params.py (see Fig. 1). waves (see Appendix III for a short description of the formulation
Post-processing tools are based on the .npz result file to ensure a implemented in the ship emission module of SIMROUTE). (Borén et al.,
decoupling of the simulation and post-processing analysis. Different 2018) compared different emission assessment methodologies in the
graphical tools are oriented to provide comprehensive results including framework of route optimization and concluded that the STEAM2
Lambert and Plate-Carrée projections. Also, the synchronous plotting or methodology was the least factor-dependent methodology, since it de­
pends on the type of fuel, specific fuel oil consumption and engine load.
Table 2 The STEAM2 model is combined with the power increase (ΔP) needed to
Numerical results of the orthodromic distance and shortest path distance (in overcome the speed penalization caused by the effect of waves on nav­
nautical miles, nmi) comparison exercise shown in Fig. 5. The shortest path igation (Δv) using the formula suggested by (Molland et al., 2017) as a
distance is estimated by A* implemented in SIMROUTE code. function of the Ptransient :
Case Initial Final Orthodromic Shortest Difference in
point Point Distance path miles (Mean ΔP 1
=( )3 − 1 (2)
(lon,lat) (lon, lat) Distance Absolute Ptransient 1 − Δv
v
Percentage
Error) The new power (Pnew ) needed to overcome the adverse wave condi­
1 0◦ , 10N◦ 120◦ E, 7023.01 7038.25 15.24 (0.21%) tions is computed using the power transient conditions in each trip in­
10◦ N terval provided by the route simulator module.
2 0◦ , 20N◦ 120◦ E, 6536.24 6549.36 13.12 (0.20%)
20◦ N Pnew = PNSR + ΔP (3)
3 0◦ , 30N◦ 120◦ E, 5830.85 5845.51 14.66 (0.25%)
30◦ N where the PNSR is then estimated from the engine load and power
4 0◦ , 40N◦ 120◦ E, 4987.29 5005.56 18.27 (0.37%) installed following the characteristics of the ship engine. Then, the total
40◦ N
fuel consumption (FC) is calculated taking into account the sailing time

5
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

Table 3
Travel times (in hours) and distances (in nautical miles) and additional information of the CMEMS product test cases included in the repository. Min.: minimum
distance route, Opt.: optimized route. Hs is the significant wave height.
Test CMEMS Min. Dist. Opt. Min. Opt. Max(Hs) Opt/ Emissions Savings Period of File in SIMROUTE
case product time time distance distance Min (in m) (Min vs Opt) analysis repository

R1 MEDSEA 33.29 27.33 386.30 390.63 5.26/5.34 28.69% 13th-14th/01/ params_MEDSEA_2.py

2021
R2 AENWS 38.64 37.85 539.12 544.38 4.23/4.17 4.47% 18th-19th/02/ param_AENWS.py
2020
R3 GLOBAL 66.48 62.25 984.93 986.47 2.54/5.28 13.25% 18th–20th/04/ param_GLOBAL.py
2020
R4 IBI 26.10 24.13 249.07 287.42 5.68/7.04 15.05% 16th-17th/02/ param_IBI.py
2020
R5 GLOBAL 61.40 60.12 912.16 917.77 8.70/10.13 5.02% 8th-10th/08/ param_GLOBAL_2.py
2019
R6 MEDSEA 16.93 14.97 151.38 161.56 6.95/6.85 16.14% 20th-21st/01/ params_MEDSEA.py
2020
R7 BALTIC 17.47 17.43 266.64 266.88 2.73/2.86 0.61% 5th-6th/02/ param_BALTIC.py
2020
R8 GLOBAL 187.04 183.60 2687.2 2725.9 6.00/6.44 4.42% 18th–25th/01/ param_GLOBAL_3.py
2020

Fig. 5. Results for the Great Circle comparison exercise. Great-circle recovered by A* is plotted in magenta and great-circle estimated using Cartopy library from
python is plotted in blue. The exact and estimated orthodromic distances are shown in Table 2. (For interpretation of the references to colour in this figure legend, the
reader is referred to the Web version of this article.)

for the route. The script ship_emissions.py estimates the reduction in Fig. 7 (Routes 2 to 7) and Fig. 8 (Route 8), including identification, are:
emissions of NOx, SOx, CO2 and particulate matter (PM) between the (R1) the MEDSEA case which considers both shores of the Mediterra­
minimum distance and the optimum route. The emissions are estimated nean Sea (Tunis/Nice); (R2) the European North-west Shelf (ENWS)
at each node-to-node interval covering the optimum and the minimum product case (Hirtshals/Tórshavn); (R3) GLOBAL case focused on Japan
distance route. The script generates an illustrative plot (see Fig. 10) coast (Hakodate/Kagoshima), (R4); IBI case (Santander/Lorient); (R5)
together with an ASCII file with the ship emissions assessment. the GLOBAL case applied to the North China Sea (Kaohsiung/Busan);
Using SIMROUTE, Borén et al. (2022) investigated the emissions (R6) the MEDSEA case (Palma de Mallorca/Barcelona); (R7) the BALTIC
reduction for several scenarios covering the Western Mediterranean case (Gdynia/Stockholm) and (R8) the GLOBAL case applied to Atlantic
Short Sea Shipping routes (from 24 to 600 nautical miles and using a real routes (Boston/Plymouth). The R6 results and intermediate files are also
Ro-Pax vessel). The ship routing optimization reveals a reduction up to included in the repository. The period analyzed, the distances sailed and
30% of ship emissions during severe storms on longer routes. Borén at al. the sailing time are summarized in Table 3, including the saving pro­
(2022) also pointed out that the expected increase of extreme weather vided by the optimal route in comparison to the minimum distance route
events, in terms of frequency, intensity and duration due to climate when only waves effect on navigation are considered. The emissions
change, suggests a gradual gain of implementing ship weather routing reduction percentage is also shown in this table. These cases pursue to
optimization in all types of routes, regardless of the distance. present optimized routes using different CMEMS products and does not
take into account relevant factors such as water depth, restricted area,
3. Results of the test cases piracy or territorial sea among others. User definition of these factors
may be included as a non-sailing area manipulating the input netcdf files
Different test cases are included in the SIMROUTE repository (with with specific tools such as Command Data Operators.
additional params.py files). These cases consider different regions using The maritime connection between Tunis and Nice (R1) shows the
different CMEMS products. The study cases shown in Fig. 6 (Route 1), benchmark case in the MEDSEA domain, which considers both shores of

6
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

Fig. 6. Temporal sequence of the snapshot of the Tunis – Nice route (see parameters of the simulation in Table 3, R1). The optimal route is plotted in magenta and the
minimum distance route is plotted in yellow. The colour bar represents the Hs (in meters) and the black arrows the direction of the waves synchronous with the ship
routes evolution. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

the Mediterranean Sea with a ship speed of 16.1 knots. The temporal shorter sailing time is evidenced from the results in the optimized route
evolution of the optimal route (in magenta) and the minimum distance in comparison to the minimum distance route (i.e. 27.33 h versus 33.29
route (in yellow) from the MEDSEA case together with the sequence of h). This means that the route optimization shows that longer distance
the waves conditions are shown in Fig. 6. The simulation corresponds to routes (i.e. 390.63 nmi vs 386.30 nmi) may be covered in less time. In
the period of 13–14 January 2021. The optimal route recovered by this example the percentage of time saved is almost 18%. The AENWS-
SIMROUTE sailed via the eastern shore of Sardinia, avoiding the North- CMEMS product considers the route between Hirtshals (Denmark) and
Westerly storm in the western Mediterranean Sea that was occurring in Tórshavn (Faroe Islands) (R2). Fig. 7a shows that the optimal route sails
those days. In particular, the west Sardinia wave conditions suggested a northward around the Shetland Islands in comparison to the minimum
head sea, which is the most penalizing condition of wave resistance in distance route. In this case, no substantial differences are obtained using
navigation, and also higher waves in comparison to the east shore. The SIMROUTE and the time-saving is equal to 2.1% (see Table 3). The

7
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

Fig. 7. Case test solutions (optimized and minimum distance routes) for different arrival/departure ports using different CMEMS products. a) R2: Hirtshals/
Tórshavn, b) R3: Hakodate/Kagoshima, c) R4: Santander/Lorient, d) R5: Kaohsiung/Busan, e) R6: Palma de Mallorca/Barcelona and f) R7: Gdynia/Stockholm. (See
detailed parameters and outputs of the simulation in Table 3, from R2 to R7).

GLOBAL-CMEMS product examples, covering the world seas, also sug­ the time saving is equal to 7.55% and the optimal route sails near the
gest interesting results when SIMROUTE is applied. Fig. 7b shows an coast to avoid large significant wave height values of the Bay of Biscay.
illustrative case for coastal shipping in Japan (i.e. Hakodate/Kagoshima, The shipping connection between Kaohsiung (Taiwan Island) and Busan
R3), for which the optimum route sails round the opposite coast of (Korea) sailing the East China Sea (Fig. 7d, R5) also shows substantial
Honshu Island in comparison to the minimum distance route. In this case differences between the minimum distance and optimal route (time-
the time saving is equal to 6.36%. The IBI-CMEMS product example savings equal to 2.08%). Also, short shipping activities (less than 200
includes a route between Santander and Lorient (Fig. 7c, R4). In this case nmi) may reveal substantial differences, as in the case of the connection

8
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

Fig. 8. Temporal sequence of the case of Boston – Plymouth (see parameters of the simulation in Table 3, R8). The optimal route is plotted in magenta and the
minimum distance route in yellow. The colour bar represents the Hs (in m) and the black arrows the direction of the waves synchronous with the route simulation.
(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

9
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

Fig. 9. Example of safety_restrictions.py: surf-riding and parametric rolling over the optimized route.

Fig. 10. Interval (node-to-node) and accumulated CO2 emissions (in Tn) for minimum (orange) and optimum (magenta) route for the case of Palma de Mallorca –
Barcelona. The optimal and minimum distance routes are shown in Fig. 7e. (For interpretation of the references to colour in this figure legend, the reader is referred
to the Web version of this article.)

between Mallorca Island and Barcelona for MEDSEA-CMEMS (R6). In 187.0 h for the minimum distance route, which is a reduction of 9.1%)
general, the cases shown previously correspond to large significant wave using GLOBAL-CMEMS products. In this sense SIMROUTE provides a
height values entailing substantial differences between the minimum smart route avoiding bad weather conditions (an illustrative snapshot is
and optimal routes. shown for the hour 80, when the optimal route sails southward in
Table 3 also includes information about the maximum significant comparison to the minimum distance route).
wave height faced by both routes. In general, the maximum significant
wave height is reduced for the optimum route in agreement with the 4. Discussion
results shown in Fig. 6. However, in the fetch restricted seas, such as the
Baltic Sea, large significant wave height values may induce the optimal SIMROUTE code execution includes a few sequential steps and a
route to be detached from the minimum distance route. In the BALTIC- unique input file enabling comprehensive use by the practitioner and
CMEMS product case between Gdynia and Stockholm (R7, Fig. 7f), the easy learning in ship route optimization methods. Recently (Zis et al.,
maximum Hs during the optimum trip is equal to 2.73 m; for these wave 2020b), highlighted the increasing attention in similar problems on
conditions the optimal route recovered by SIMROUTE suggests a west­ finding the optimal path and sailing speed considering environmental
ern sailing route of Gotland Island in comparison to the minimum dis­ conditions. In addition, they suggested a standardization of methods to
tance route. During calm periods, the differences are not relevant and facilitate inter-comparison of methodologies. In this respect, SIMROUTE
the minimum optimal route is the same as the minimum distance route. fills the gap, providing an open, comprehensive and efficient software to
Note that the wave information shown in Table 3 reveals substantial compute weather ship routes using CMEMS products, and it seems a
wave conditions for both the minimum distance and optimized routes in good candidate for a benchmarking strategy. For instance, the inclusion
many cases (e.g. Santander/Lorient (R4) or Hakodate/Kagoshima (R3)). of input variables in a unique file enables easy implementation for new
Based on the time cost reduction provided by SIMROUTE, ship emission cases and the modification of input variables testing different scenarios.
reductions are also assessed according to section 2.5.2. Emissions re­ Wave interpolation is still highly time-consuming, so the use of the
ductions range from 0.61% to 28.69%, consistent with the time-saving A* algorithm and the possibility to run the code in unix platforms makes
(see Table 3). the use of SIMROUTE feasible for long distances (e.g. see the GLOBAL-
Finally, Atlantic routes that connect Europe and the USA East coast CMEMS cases shown in the previous section). Taking into account the
also provide interesting insights. These routes have been investigated in different methods presented in the Introduction section, multi-objective
previous publications (e.g. (Hinnenthal and Clauss, 2010; Shao et al., methods suggest robust optimized solutions for a few objectives, such as
2012)), revealing noticeable differences due to the typical adverse the minimum time of arrival, navigation risk and minimum fuel con­
weather conditions in the North Atlantic. Fig. 8 shows the route between sumption etc. (e.g. Hinnenthal and Clauss, 2010; Zhao et al., 2022);
Boston (USA) and Plymouth (UK) obtained by SIMROUTE, where sub­ although the solution is highly dependent on the population size of the
stantial differences are revealed (183.6 h for the optimized route vs. initial generation of solutions requiring large computing resources

10
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

(Simonsen et al., 2015). Accurate objectives parametrization is required description. In this respect, the extension of the forecast horizon will
in multi-objective weather routing systems (Zhao et al., 2022), which permit long-haul routes and multi-port vessel routes to be addressed.
may add complexity and uncertainty on new implementations. Also, As a function of the wave conditions, SIMROUTE may provide a
other methods faced the route optimization in terms of fuel consumption limited extension to the benefits of the optimized route and the mini­
(e.g. (Takashima and Mezaoui, 2009)). In those terms, it should be mum distance route. The results shown in the previous section have
highlighted that the present contribution takes into account the revealed that calm sea conditions or moderate storms may have an
non-linearity in the relation between fuel consumption and engine load. imperceptible impact on the optimized route in comparison to the
This relation is considered to have an approximately parabolic de­ minimum distance. However, the real routes may be far from the min­
pendency and the change on Specific Fuel Oil Consumption (SFOC) is imum distance routes, increasing the relevance of the SWR results. New
assessed over whole route. Therefore, SFOC is calculated for the implementations using SIMROUTE should be designed previously taking
different conditions which the vessel is facing in each node and total fuel into account the forecast horizon of the wave products, the sailing
consumption is obtained through the relation among instantaneous duration and the consequences of grid spacing resolutions that may in­
power, SFOC and the interval time from node to node. As future works, crease substantially the computational cost. As future work, AIS routing
for the sake of fuel consumption estimation accuracy, other de­ inter-comparison will suggest an increase of the importance of SWR,
terminants could be taken into account such as vessel’s draught, trim or with a direct impact on emissions reduction or economic benefits, using
water depth inter alia. frequent commercial shipping routes. Complementary analysis with
However, all methods present benefits and deficiencies as a function SIMROUTE allows the exploration of optimized routes in a wide
of the problem statement (single or multiple), computational resources framework, such as ship emissions evaluation or new routes such as the
and simulation performance. SIMROUTE pursues a trade-off between North Sea passage. Finally, the emergence of autonomous vessels re­
accuracy, user-friendliness and computational time, assuming standard quires specific new route design systems, including safety, fuel effi­
computational resources. In this respect, the current version of the code ciency, emissions and efficient routes (Wu et al., 2021; Zakaria et al.,
uses simple added wave resistance models instead of formulations based 2022; Zis et al., 2020b). As noted by (Zis et al., 2020b), research on SWR
on specific transfer functions, spectrum-based resistance or reduced in world navigation will be a topic deserving ever more attention in the
non-dimensional resistance for calm and added wave resistance (Hin­ coming years as the world pursues greenhouse emissions reduction, the
nenthal and Clauss, 2010; Hu et al., 2014; Mannarini et al., 2016). This blue economy and sustainable development. In this sense, SIMROUTE
allows an easy interpretation of the wave effect on navigation using provides a robust alternative for SWR analysis and development in an
velocity penalization formulations (see inter-comparison exercise in open and collaborative perspective.
(Borén et al., 2019)). Also, the modular structure of SIMROUTE suggests
an easy implementation of more complex wave resistance formulations 5. Conclusions
and other resistance factors such as the effects of wind, water currents or
sea-ice on navigation (see formulations in (Cai and Wen, 2014)). For We describe a comprehensive and free software for Ship Weather
instance, the use of additional CMEMS products (and also ERA5 for wind Routing referred to as SIMROUTE. The code targets one of the aspira­
description), may be a good start point to include these factors on ship tions of ship weather routing by minimizing time of navigation and, in
resistance to navigation. These topics, some of them included in beta consequence fuel consumption and emissions. SIMROUTE uses the A*
versions of SIMROUTE, are considered as future works. Also, the effect pathfinding algorithm and pursues a trade-off between accuracy, user-
of the added resistance formulations may have a direct effect on the fuel friendliness and computational time, assuming standard computa­
consumptions and ship emissions, so future works also include sensi­ tional resources. Several cases using different CMEMS products over
tivity analysis of all these factors including the emissions formulations short and long distances including ship emission assessment have been
itself (STEAM2) which under specific circumstances may differ from real tested. SIMROUTE provides a robust alternative for Ship Weather
values (Berthelsen and Nielsen, 2021). Routing analysis and development in an open and collaborative
The IMO resolution A.528(13) (International Maritime Organization, perspective.
1983) recommends to Governments to advise ships to make use of
weather routing information. In this sense, several commercial products Funding
have been developed in the field of maritime routing but with limited
available information on the underlying methodologies. As pointed out This work was partially supported by MObiLIty sERvices Enhanced
by Zis et al. (2020a), these systems are operating as black boxes with by GALILEO & Blockchain, MOLIERE (H2020 5th call, Contract Num­
suggestions on the optimal route with lack of transparency on optimi­ ber: 101004275).
zations and information processed. In opposite, the open-software phi­
losophy on SIMROUTE could trigger a wide cooperation for the Author contributions
development of weather routing systems, both at the industrial and the
academic level, in compliance with the most recent e-navigation Conceptualization, M.G.; methodology, M.G. and C.B.; software
instances. development: M.G. and C.B.; investigation, C.B. and M.C.; writing-
The software provided and additional analysis may fit in an opera­ original draft preparation, M.G.; writing-review and editing, M.C. and
tional oceanography context (e.g. emissions or safety on navigation) as a C.B.; visualization, M.G. and M.C.; Supervision; M.G. and M.C. All au­
contribution to efficient and safe maritime operations such as route plan thors have read and agreed to the published version of the manuscript.
design based on wave forecast products. SIMROUTE is based on waves
CMEMS products, which provides free and open marine data including Declaration of competing interest
learning services, enabling maritime routes in different domains
including global areas. In this sense, SIMROUTE provides for the first The authors declare that they have no known competing financial
time a model which include the world application area and sub-regional interests or personal relationships that could have appeared to influence
domains (see review paper in Simonsen et al., 2015 and Zis et al., the work reported in this paper.
2020a). Also the portability of netcdf files for wave information makes it
very easy to include new wave products in SIMROUTE from other Acknowledgements
regional and downstream services (e.g. Sotillo et al., 2020). The accu­
racy and impact of SWR will also benefit from the future improvement of The authors acknowledge CMEMS (Copernicus Marine Environment
weather forecasting in terms of accuracy and spatio-temporal Monitoring Service) for the wave predictions provided.

11
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

Appendix I. Wave effect on navigation formulation

SIMROUTE considers three formulations for the assessment of the wave effect on navigation. The first method implemented into the weather
routing system to analyze the vessel speed reduction due to wave effect according to Mannarini et al., 2013) (inspired by Bowditch (2002)). The final
speed is computed in function of the non-wave affected speed (v0 ) plus a reduction in function of the wave parameters:

v(Hs , θ) = v0 − f (θ)⋅Hs2 (A1.1)

where v0 is the vessel initial speed without wave effect, Hs is the significant wave height and θ is the ship-to-wave relative direction
(clockwise).
Table A1.1
Values of f coefficient

θ f (in kn/ft2)

0◦ ≤ θ ≤45◦ 0.0083
45◦ < θ <135◦ 0.0165
135◦ ≤ θ ≤225◦ 0.0248
225◦ < θ <270◦ 0.0165
270◦ ≤ θ ≤360◦ 0.0083

The second methodology implemented into the SWR is Aertssen’s formula (Aertssen, 1975). Aertssen’s formula also takes into account the ship’s
dimensions, in particular, the ship’s length. For approximating the speed reduction, Aertssen proposes the following equation:

( )
m v0
v = v0 − +n (A1.2)
LBP 100

where LBP is vessel’s length between perpendiculars and m and n are empirical coefficients defined in Table A1.2.
Table A1.2
Values of m and n Aertssen coefficients depending on the wave characteristics (BN= Beaufort number, Hs = significant wave height, Wspeed = wind speed)

θ 150◦ -210◦ 60◦ -150◦ / 210◦ - 300◦ 30◦ -60◦ / 300◦ - 330◦ 0◦ -30◦ / 330◦ - 360◦

BN Hs Wspeed m n m n m n m n

5 2.5 17–21 900 2 700 2 350 1 100 0

6 4.0 22–27 1300 6 1000 5 500 3 200 1
7 5.5 28–33 2100 11 1400 8 700 5 400 2
8 7.5 34–40 3600 18 2300 12 1000 7 700 3

The columns of the table contain estimated values of m and n coefficients for waves hitting a vessel at a particular angle θ in degrees.
Khokhlov formula is suggested by Lubkovsky (2009). This method takes into account the height and direction of the waves and also the ship’s
dimensions, in particular, the deadweight of the ship. According to (Maisiuk and Gribkovskaia, 2014) the standard error for this formula does not
exceed 0.5 knots. Khokhlov method calculates speed reduction as follows:
( )
v = v0 − (0.745 ⋅ Hs − 0.245 ⋅ θ ⋅ Hs )⋅ 1.0 − 1.35 ⋅ 10− 6 ⋅ D ⋅ v0 (A1.3)

where θ (here in radians) and D is vessel’s deadweight (DWT) in tons. Khokhlov method is applicable for vessels with a deadweight range from 4.000
to 20.000 DWT including supply vessels, and design speeds between 9 and 20 knots.

Appendix II. Safety restrictions formulation

The methodology used in SIMROUTE to assess safety restrictions takes into account the recommendations of the International Maritime Orga­
nization (International Maritime Organization, 2007, circular no. 1228) for avoiding dangerous situations in adverse weather and sea conditions
(Krata and Szlapczynska, 2018; Mannarini et al., 2013). Two unstable motions that can cause discomfort to the passage and crew members, generating
dynamic loads to the structure and cargo of the ship are implemented into SIMROUTE: surf-riding and parametric rolling.
When a ship is situated on the steep forefront of a high wave in following or quartering sea conditions, the ship can be accelerated to ride on the
wave. This is known as surf-riding. In this situation the so-called broaching-to phenomenon may occur, which endangers the ship to capsizing as a
result of a sudden change of the ship’s heading and unexpected large heeling. Surf-riding and broaching-to may occur when both of these conditions
are fulfilled:
135◦ < α < 225◦ (A2.1)
and
√̅̅̅̅̅̅̅̅̅
1.8 Lship
v> (A2.2)
cos(180◦ − α)

12
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

where α is the angle of encounter, being the ship-to-wave relative direction (α = 180◦ for following seas), Lship is the length of the ship in meters and v is
the speed of the ship in knots.
On the other hand, parametric roll motions with large and dangerous roll amplitudes in waves are due to the variation of stability between the
position on the wave crest and the position in the wave trough. Parametric rolling may occur in two different situations:
|TE − TR | = ε⋅TR (A2.3)
or
|2 ⋅ TE − TR | = ε⋅TR (A2.4)

where ε is the relative tolerance in frequency matching in %, TE is the encountered period with waves in seconds and TR is the natural roll period in
seconds. The method for calculation of the natural roll period is given in the Intact Stability Code and is based on the initial metacentric height of a ship
(International Maritime Organization, 2008).

Appendix III. Ship emission formulation

The methodology used in SIMROUTE (i.e. STEAM2) to estimate emissions was inspired by a Theoretical Based Method (TBM) with a bottom-up
approach, that obtains their results via modelling with no data recorded on-board (Borén et al., 2018). This TBM, which includes input variables such
as installed power per engine or engine load, was successfully applied in Ro-Pax ships by (Jalkanen et al., 2009) and (Jalkanen et al., 2012). STEAM2
methodology allows the evaluation of the exhaust emissions of marine traffic based on the messages provided by the Automatic Identification System
(AIS). The evaluation of (Jalkanen et al., 2009) and (Jalkanen et al., 2016) in SWR was already discussed in (Borén et al., 2018) concluding that this
method uses ship specific data to obtain more accurate calculations as it is a bottom-up methodology. The required data for emissions assessment and
main assumptions for the ship emission assessment are shown in Table A3.1.

Table A3.1
Required data for emission assessment methodology in SIMROUTE.

Input data Acronym Assumptions

Installed power per engine (in kW) PInstalled From IHS Markit database
Engine Load EL According to (Jalkanen et al., 2012), EL’s values are around 70%–80%
Specific Fuel Oil Consumption (in g/kWh) SFOC From the corresponding manufacturer’s project guide of the engine
Design speed (in knots) Vdesign From IHS Markit database
Sulphur and Carbon Content of fuel (in mass percentage) SC/CC Depends on the fuel burnt
Main Engine Revolutions per Minute (in rpm) rpm If engine data is unavailable, the ship is assumed to use a 500 rpm medium speed diesel
engine by default
Molar mass of Sulphur/Sulphur dioxide/Carbon/Carbon M(S)/M(SO2)/M(C)/M
dioxide (in g/mol) (COa2)

Total emissions for each pollutant (ETp) per ship and per route is the sum of the amount of pollutant (p) emitted into the atmosphere and can be
obtained by applying the following formula, changing the emissions factor related to one pollutant or another:

ET p = P⋅EL⋅EFp ⋅t (A3.1)

where P is the average output power (in kW), EL is the engine load, EF is the emission factor of each pollutant (according to Table A3.1) and t
(in hours) is the total time sailed.
Table A3.2
Summary table of the EF for each pollutant.

Sulphur dioxide (SO2) SFOCnSC

EF(SO2 ) = M(SO2 )nn(SO2 ) = M(SO2 )nn(S) = M(SO2 )n (g/kWh)
M(S)

Carbon dioxide (CO2) SFOCnCC

EF(CO2 ) = M(CO2 )nn(CO2 ) = M(CO2 )nn(C) = M(CO2 )n (g/kWh)
M(C)
⎧
Nitrogen oxides (NOx) ⎨ 17(rpm < 130)
− 0.2
45rpm (130 < rpm < 2000)
⎩
9.8(rpm > 2000)
Particulate Matter (PM) EF(PM) = SFOCREL (EFSO4 + EFH2O + EFOC OCEL + EFEC + EFASH )
Where:
SFOCREL = 0.455EL2 − 0.71EL + 1.28; SFOC = SFOCREL nSFOCMANUFACTURER

The total amount of each pollutant emission can be assessed by adapting equation (A3.1) to the pollutant analyzed, as shown below:
∑
n
ET p = Pnew new
i ⋅ELi ⋅EF(p)i ⋅\Δti (A3.2)
i=0

Obtaining Pnew and n being the number of intervals i and Δt the time from node-to-node for each interval.
In order to estimate the impact of Engine Load (EL) change on Specific Fuel Oil Consumption (SFOC), the EL for each interval was calculated
(ELnew):
ELnew = Pnew /PInstalled (A3.3)

13
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

These formulas assumed that there is a linear relation between fuel consumption and EL, and SFOC is presumed to be constant. Through manu­
facturers data, it can be seen that SFOC is a non-linear function of EL but that there is an approximately parabolic dependency between them. STEAM2
assumes a parabolic function for all engines getting to following equations for relative SFOC (SFOCnew rel ) and final SFOC (SFOCend) using regression
analysis:
new
SFOCrel = 0.445⋅(ELnew )2 − 0.71⋅ELnew + 1.28 (A3.4)
new
SFOCend = SFOCrel ⋅SFOC (A3.5)

Afterwards, the absolute fuel consumption (FC) was estimated as follows:

∑
n
FC = Pnew new
i ⋅SFOCi ⋅\Δti (A3.6)
i=0

References the Baltic Sea area. Atmos. Chem. Phys. 9, 9209–9223. https://doi.org/10.5194/
acp-9-9209-2009.
Jalkanen, J.P., Johansson, L., Kukkonen, J., Brink, A., Kalli, J., Stipa, T., 2012. Extension
Aertssen, G., 1975. The effect of weather on two classes of container ships in the North
of an assessment model of ship traffic exhaust emissions for particulate matter and
Atlantic. Nav. Archit. 11–13.
carbon monoxide. Atmos. Chem. Phys. 12, 2641–2659. https://doi.org/10.5194/
Ardhuin, F., Magne, R., Filipot, J.-F., Van, A., Westhuysen, D., Roland, A., Queffeulou, P.,
acp-12-2641-2012.
Lefevre, J.-M., Aouf, L., Babanin, A., Collard, F., 2010. Semi-empirical dissipation
Kanellos, F., Prousalidis, J.M., Tsekouras, G.J., 2014. Control system for fuel
source functions for wind-wave models: part I, definition, calibration and validation
consumption minimization–gas emission limitation of full electric propulsion ship
at global scales. J. Phys. Oceanogr. 40.
power systems. Proc. Inst. Mech. Eng. Part M J. Eng. Marit. Environ. 228, 17–28.
Berthelsen, F.H., Nielsen, U.D., 2021. Prediction of ships’ speed-power relationship at
https://doi.org/10.1177/1475090212466523.
speed intervals below the design speed. Transport. Res. Transport Environ. 99,
Korres, G., Ravdas, M., Zacharioudaki, A., Denaxa, D., Sotiropoulou, M., 2021.
102996 https://doi.org/10.1016/j.trd.2021.102996.
Mediterranean Sea Waves Analysis and Forecast (CMEMS MED-Waves, MedWAМ3
Borén, C., Castells-Sanabra, M., Grifoll, M., 2022. Ship emissions reduction using weather
System) (Version 1) [Data Set]. Copernicus Monitoring Environment Marine Service
ship routing optimisation. J. Eng. Maritime Environ. 1–12. https://doi.org/10.1177/
(CMEMS). https://doi.org/10.25423/CMCC/MEDSEA_ANALYSISFORECAST_WAV_
14750902221082901.
006_017_MEDWAM3.
Borén, C., Castells-Sanabra, M., Grifoll, M., 2018. Intercomparison of emissions
Krata, P., Szlapczynska, J., 2018. Ship weather routing optimization with dynamic
assessment methodologies in a short sea shipping framework. In: 19th International
constraints based on reliable synchronous roll prediction. Ocean Eng. 150, 124–137.
Association of Maritime Universities Annual General Assembly. International Center
https://doi.org/10.1016/j.oceaneng.2017.12.049.
for Numerical Methods in Engineering (CIMNE), pp. 416–424.
Kuhlemann, S., Tierney, K., 2020. A genetic algorithm for finding realistic sea routes
Borén, C., Falevitch, L., Castells-Sanabra, M., Grifoll, M., 2019. Added resistance
considering the weather. J. Heuristics 26, 801–825. https://doi.org/10.1007/
parametrizations due to waves in a weather ship routing system. In: International
s10732-020-09449-7.
Conference of Maritime Science & Technology NAŠE MORE, International
Lin, Y.-H., Fang, M.-C., Yeung, R.W., 2013. The optimization of ship weather-routing
Conference of Maritime Science & Technology NAŠE MORE 2019, pp. 50–59.
algorithm based on the composite influence of multi-dynamic elements. Appl. Ocean
Cai, Y., Wen, Y., 2014. Ship route design for avoiding heavy weather and sea conditions.
Res. 43, 184–194. https://doi.org/10.1016/j.apor.2013.07.010.
TransNav, Int. J. Mar. Navig. Saf. Sea Transp. 8, 551–556. https://doi.org/
Lubkovsky, V., 2009. Determination of Wind-Wave Speed Loss of Vessels for Mixed Type
10.12716/1001.08.04.09.
Navigation with Measurement of Wave Parameters by Means of Orthogonally-Linear
Castells-Sanabra, M., Borén, C., Grifoll, M., Martínez de Osés, F.X., 2019. Weather
Wave Meters. Novosibirsk, Russia.
Routing Software for academic purposes A pilot study. In: Proceedings of Teh
Maisiuk, Y., Gribkovskaia, I., 2014. Fleet sizing for offshore supply vessels with
International Association of Maritime Universities (IAMU).
stochastic sailing and service times. In: Procedia Computer Science. Elsevier B.V.,
Chapman, L., 2007. Transport and climate change: a review. J. Transport Geogr. 15,
pp. 939–948. https://doi.org/10.1016/j.procs.2014.05.346
354–367. https://doi.org/10.1016/j.jtrangeo.2006.11.008.
Maki, A., Akimoto, Y., Nagata, Y., Kobayashi, S., Kobayashi, E., Shiotani, S., Ohsawa, T.,
Cheung, J.C.H., 2018. Flight planning: node-based trajectory prediction and turbulence
Umeda, N., 2011. A new weather-routing system that accounts for ship stability
avoidance. Meteorol. Appl. 25, 78–85. https://doi.org/10.1002/met.1671.
based on a real-coded genetic algorithm. J. Mar. Sci. Technol. 16, 311–322. https://
DfT, 2004. The Future of Transport: A Network for 2030, Cm (Series) (Great Britain.
doi.org/10.1007/s00773-011-0128-z.
Parliament). Stationery Office.
Mannarini, G., Coppini, G., Oddo, P., Pinardi, N., 2013. A prototype of ship routing
European Commission, 2021. Copernicus Marine Environmental Monitoring Service
decision support system for an operational oceanographic service. TransNav, Int. J.
[WWW Document]. URL. https://www.copernicus.eu (accessed 4.1.20).
Mar. Navig. Saf. Sea Transp. 7, 53–59. https://doi.org/10.12716/1001.07.01.06.
Goldsworthy, L., Goldsworthy, B., 2015. Modelling of ship engine exhaust emissions in
Mannarini, G., Pinardi, N., Coppini, G., Oddo, P., Iafrati, A., 2016. VISIR-I: small vessels –
ports and extensive coastal waters based on terrestrial AIS data - an Australian case
least-time nautical routes using wave forecasts. Geosci. Model Dev. (GMD) 9,
study. Environ. Model. Software 63, 45–60. https://doi.org/10.1016/j.
1597–1625. https://doi.org/10.5194/gmd-9-1597-2016.
envsoft.2014.09.009.
Molland, A.F., Turnock, S.R., Hudson, D.A., 2017. Ship Resistance and Propulsion, Ship
Grifoll, M., Martínez de Osés, F.X., 2016. A ship routing system Applied at Short Sea
Resistance and Propulsion. Cambridge University Press. https://doi.org/10.1017/
distances. XIII J. Marit. Res. 3–6.
9781316494196.
Grifoll, M., Martínez de Osés, F.X., Castells, M., 2018. Potential economic benefits of
Nielsen, U.D., 2021. Spatio-temporal variation in sea state parameters along virtual ship
using a weather ship routing system at Short Sea Shipping. WMU J. Marit. Aff. 17,
route paths. J. Oper. Oceanogr. 1–18. https://doi.org/10.1080/
195–211. https://doi.org/10.1007/s13437-018-0143-6.
1755876X.2021.1872894.
Hagiwara, H., 1982. Weather Routing of (Sail-assisted) Motor Vessel.
Ruth, E., Thompson, I., 2022. Comparing design assumptions with hindcast wave
Hinnenthal, J., Clauss, G., 2010. Robust Pareto-optimum routing of ships utilising
conditions and measured ship speed and heading. Ocean Eng. 247, 110613 https://
deterministic and ensemble weather forecasts. Ships Offshore Struct. 5, 105–114.
doi.org/10.1016/j.oceaneng.2022.110613.
https://doi.org/10.1080/17445300903210988.
Shao, W., Zhou, P., Thong, S.K., 2012. Development of a novel forward dynamic
Hu, Q., Cai, F., Yang, C., Shi, C., 2014. An algorithm for interpolating ship motion
programming method for weather routing. J. Mar. Sci. Technol. 17, 239–251.
vectors. TransNav, Int. J. Mar. Navig. Saf. Sea Transp. 8, 35–40. https://doi.org/
https://doi.org/10.1007/s00773-011-0152-z.
10.12716/1001.08.01.04.
Shin, Y.W., Abebe, M., Noh, Y., Lee, S., Lee, I., Kim, D., Bae, J., Kim, K.C., 2020. Near-
International Maritime Organization, 2010. International Convention on Standards of
optimal weather routing by using improved A* algorithm. Appl. Sci. 10 https://doi.
Training, Certification and Watchkeeping for Seafarers (STCW). London, UK.
org/10.3390/app10176010.
International Maritime Organization, 2008. International code on Intact stability.
Simonsen, M.H., Larsson, E., Mao, W., Ringsberg, J.W., 2015. State-of-the-Art within
Resolut. MSC 267 (85).
Ship Weather Routing. https://doi.org/10.1115/OMAE2015-41939.
International Maritime Organization, 2007. Annex A IMO Circular MSC.1/Circ. 1228
Sotillo, Marcos G., Cerralbo, P., Lorente, P., Grifoll, M., Espino, M., Sanchez-Arcilla, A.,
Dated 11 January 2007.
Álvarez-Fanjul, E., 2020. Coastal ocean forecasting in Spanish ports: the Samoa
International Maritime Organization, 1983. Resolution A.528(13) Adopted on 17
operational service. J. Operat. Oceanogr. 13 (1), 37–54. https://doi.org/10.1080/
November 1983 Recommendation on Weather Routeing 528.
1755876X.2019.1606765.
Jalkanen, J.-P., Johansson, L., Kukkonen, J., 2016. A comprehensive inventory of ship
Staneva, J., Behrens, A., Ricker, M., Gayer, G., 2020. Black Sea Waves Analysis and
traffic exhaust emissions in the European sea areas in 2011. Atmos. Chem. Phys. 16,
Forecast (CMEMS BS-Waves) (Version 2) Set. Copernicus Monitoring Environment
71–84. https://doi.org/10.5194/acp-16-71-2016.
Marine Service (CMEMS). https://doi.org/10.25423/CMCC/BLKSEA_
Jalkanen, J.P., Brink, A., Kalli, J., Pettersson, H., Kukkonen, J., Stipa, T., 2009.
ANALYSISFORECAST_WAV_007_003.
A modelling system for the exhaust emissions of marine traffic and its application in

14
M. Grifoll et al. Ocean Engineering 255 (2022) 111427

Szłapczynska, J., Smierzchalski, R., 2009. Multicriteria optimisation in weather routing. Zhao, W., Wang, H., Geng, J., Hu, W., Zhang, Z., Zhang, G., 2022. Multi-objective
TransNav Int. J. Mar. Navig. Saf. Sea Transp. 3, 393–400. weather routing algorithm for ships based on hybrid particle swarm optimization.
Takashima, K., Mezaoui, B., 2009. TransNav Journal - on the Fuel Saving Operation for J. Ocean Univ. China 21, 28–38. https://doi.org/10.1007/s11802-022-4709-8.
Coastal Merchant Ships Using Weather Routing [WWW Document]. Zis, T., North, R.J., Angeloudis, P., Ochieng, W.Y., Bell, M.G.H., 2014a. Evaluation of
Walther, L., Rizvanolli, A., Wendebourg, M., Jahn, C., 2016. Modeling and optimization cold ironing and speed reduction policies to reduce ship emissions near and at ports.
algorithms in ship weather routing. Int. J. e-Navigat. Marit. Econ. 4, 31–45. https:// Marit. Econ. Logist. 16, 371–398. https://doi.org/10.1057/mel.2014.6.
doi.org/10.1016/j.enavi.2016.06.004. Zis, T., North, R.J., Angeloudis, P., Ochieng, W.Y., Bell, M.G.H., 2014b. Environmental
Wang, K., Yan, X., Yuan, Y., Tang, D., 2016. Optimizing ship energy efficiency: balance of shipping emissions reduction strategies. Transport. Res. Rec. 2479, 25–33.
application of particle swarm optimization algorithm. Proc. Inst. Mech. Eng. Part M https://doi.org/10.3141/2479-04.
J. Eng. Marit. Environ. 232 https://doi.org/10.1177/1475090216638879. Zis, T.P.V., Psaraftis, H.N., Ding, L., 2020a. Ship weather routing: a taxonomy and
Wu, G., Atilla, I., Tahsin, T., Terziev, M., Wang, L., 2021. Long-voyage route planning survey. Ocean Eng. https://doi.org/10.1016/j.oceaneng.2020.107697.
method based on multi-scale visibility graph for autonomous ships. Ocean Eng. 219, Zis, T.P.V., Psaraftis, H.N., Tillig, F., Ringsberg, J.W., 2020b. Decarbonizing maritime
108242 https://doi.org/10.1016/j.oceaneng.2020.108242. transport: a Ro-Pax case study. Res. Transp. Bus. Manag. 37 https://doi.org/
Zakaria, A., Md Arof, A., Khabir, A., 2022. In: Ismail, A., Dahalan, W.M., Öchsner, A. 10.1016/j.rtbm.2020.100565.
(Eds.), Instruments Utilized in Short Sea Shipping Research: A Review BT - Design in
Maritime Engineering. Springer International Publishing, Cham, pp. 83–108.

15