355
1 INTRODUCTION
In recent years, the Norwegian Coastal Authority has
prepared and published so-called “Reference Routes
for Navigation”, covering all relevant voyages along
the Norwegian coast. Developed by experienced
navigators, these routes determine the most advisable
track from A to B through the coastline’s complex
labyrinth of islands and fjords, accounting for both
surface and underwater hazards. The route-tracks are
freely available to all and can be downloaded from a
dedicated website [2] in the standardized RTZ format,
which is compatible with onboard ECDIS and other
Electronic Chart Systems. Additionally, an interface
(API) is available, allowing providers of e-Navigation
solutions to integrate the routes into their services
(back-of-bridge systems).
Beyond being part of the European maritime
digitalization strategy – including preparations for
possible route exchange systems between vessels – the
effort is driven by the expectation of enhancing
maritime navigation and safety in three ways:
1. Minimize human errors in route planning and
plotting of courses.
2. Improve the separation of traffic sailing in opposite
directions within the same fairway.
3. Establish more predictable traffic patterns.
The need for traffic separation, as mentioned in
point 2 above, can of course be debated. However,
Kristic et al. [4] have shown that even experienced
navigators often plot courses mid-fairway in
constricted waterways, increasing the risk of head-on
and close-quarters situations. Also, without traffic
routing systems, such as e.g. traffic separation
schemes, traffic patterns tend to be inconsistent, with
route choices varying as much as the number of
navigators [5]. A more predictable traffic pattern, as
noted in point 3, would likely be beneficial both today
Reference Routes for Sea Navigation – “To Follow,
or Not to Follow?”
M. Jacobsen, M.V. Aarset & O.S. Hareide
Norwegian University of Science and Technology, Aalesund, Norway
ABSTRACT: The Norwegian Coastal Authority has introduced Reference Routes for Navigation to enhance
maritime safety and efficiency along the Norwegian coast. These routes provide predefined tracks aimed at
minimizing human errors in route planning, improving traffic separation, and promoting predictable traffic
patterns. This study evaluates the impact of these reference routes by analysing AIS tracking data from before
and after their implementation. A quantitative approach was used, incorporating statistical methods such as T-
tests, linear regression, and K-Means clustering to assess vessel compliance and adaptation. The findings indicate
a statistically significant, yet modest, improvement in adherence to reference routes, with vessel characteristics
playing an important role in compliance levels. These results suggest that while voluntary routing measures
influence traffic patterns, complete compliance remains unlikely. Further research is recommended to validate
and expand upon these findings.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 19
Number 2
June 2025
DOI: 10.12716/1001.19.02.02
356
and in a future where manned and autonomous vessels
share the waterways.
Traditional route planning can be time-consuming.
As a benefit for navigators who adopt this new
approach, reference routes not only provide pre-
plotted courses but also come with an information
package containing all relevant details for the voyage.
This reduces the need to consult nautical publications.
Instead of spending time gathering information,
navigators can focus on validating the plotted route
and information. It is important to note that reference
routes are a decision support tool, not a substitute for
the navigator’s responsibility in route planning.
In contrast to most traditional traffic routing
measures, such as the IMO's Traffic Separation
Schemes (TSS), reference routes provide a voluntary
and easily modifiable means of directing traffic and
could potentially be a valuable addition to existing
routing measures for coastal states.
The routes are designed and quality-checked for
vessels corresponding to certain dimensions, which are
specified on the routes website [2]. Vessels with
different dimensions may still benefit from using the
reference routes, but must carefully assess their
optimal route, with regard to factors such as physical
constraints and hazards along the route.
This paper presents and discusses the statistical
results from the research conducted by M. Jacobsen, as
part of his master's thesis from The Norwegian
University of Science and Technology - NTNU [3].
The study used AIS tracking data to assess whether
the implementation of reference routes had a
measurable impact on ship traffic patterns. Three
research questions shaped its direction:
1. Have vessels adjusted their voyages to more closely
follow the reference routes?
2. What influences the degree of compliance with the
reference routes?
3. Do the voyages cluster into distinct groups? If so,
what are the group and voyage characteristics of
those groups?
2 RESEARCH METHOD
To assess any changes, two quantitative datasets were
used, each representing the period before or after
implementation of the reference routes. These datasets
were then analysed using appropriate statistical
methods. The whole process is described in the
following two-step approach:
2.1 Step 1: Processing AIS Tracking Data: Gathering,
Filtering, and Operationalization
The route of study is a 11 nm stretch through one of the
most complex traffic areas along the Norwegian coast
(see figure 1). Its complexity, combined with the area's
width allowing for different route choices, were the
main reasons for selecting it as the study area. AIS
tracking data were obtained with assistance from the
Norwegian Coastal Administration (NCA), and all
vessels and objects deemed irrelevant to the study’s
objective were removed from the dataset. Table 1
outlines the delimitations applied to this study. The
characteristics of the study periods, especially their
identical lengths and calendar months, have been
specifically chosen to reduce the impact of seasonal
variations, weather and disturbance factors like e.g.
recreational vessels. It was further assumed that any
changes detected between pre- and post-
implementation were solely due to the reference
routes, with all other factors remaining unchanged. In
this context it must be mentioned that it has not come
to the authors' attention that any external factors we
anticipate might have changed the outcome actually
have changed between the study periods.
Table 1. Delimitations to the study
Study periods
Vessel
dimensions
Excluded
vessel
characteristics
Figure 1. Route of study
With assistance from the geospatial software QGIS,
the AIS tracking data were operationalized into a
measurable and comparable format. A corridor-based
approach was used for this purpose, where parallel
sectors to the route were established and combined to
form corridors. The distance sailed within these
corridors per passage was then measured and
expressed as a percentage of the total reference route’s
distance.
If a single standard cross-track limit (XTL) were
used in voyage planning onboard vessels, only one
corridor would be needed to produce meaningful
results in this study. However, in practice, each ship’s
navigator determines their own preferred XTL [4]. To
address this issue, data from multiple corridors were
initially collected. Ultimately, the combination of the
400 m and 800 m corridors was found to be sufficient
for further analysis in terms of vessel characteristic
diversity. One corridor complements the other in cases
where compliance is either 0% or 100%.
Each corridor spans transversely from one side of
the reference route’s track to the defined distance on
the other. However, the corridor name (e.g. 400m)
refers only to the Cartesian distance (chart projection
EPSG:3857) from the corridor’s centreline – the
reference route’s track – to one side. Lastly, a wide
parallel sector on the starboard side of the reference
route was used to determine how much of each voyage
was sailed on either side of the route. The same
357
percentage-based approach was applied here, where a
value of 50% indicates that the vessel sailed equal
distance on both the starboard and port sides of the
reference route.
Figure 2. Operationalisation of AIS track
These attributes, termed “400m ratio”, “800m
ratio”, and “Starboard ratio”, have been used as
outcome/dependent variables in the statistical analysis
to measure how well a vessel has adapted to the
reference routes. All other available variables from the
AIS data were also collected for the statistical analysis.
Figure 3 provides an overview of all the variables used
in this study.
Figure 3. Attribute-overview
2.2 Step 2: Statistical analysis
After filtration, the dataset consisted of 286 voyages
before and 264 voyages after implementation of the
reference routes. All these voyages were further
analysed using appropriate statistical methods as
shown in Table 2 with reference to the research
questions as presented earlier.
Table 2. Statistical Methods used - sorted by research
question (RQ)
RQ
Statistical Method
1
Independent T-test comparing corridor values before and
after implementation.
2
Linear regression including all recorded attributes, followed
by independent T-tests of significant variables and related
variables.
3
Cluster analysis using the K-Means clustering algorithm.
The T-test is an effective statistical method for
comparing expected values between groups. Linear
regression, on the other hand, is better suited for
identifying variables that correlate with the dependent
variable and determining their level of impact. In
contrast to these methods, cluster analysis does not
require parametric assumptions about the dataset. The
goal of cluster analysis is to group objects in a dataset
so that those within the same cluster are very similar,
while those in different clusters are as different as
possible [1].
In the hypothesis testing for the linear regression
and independent T-tests, the significance level was set
to 0.05. This implies that only results with a p-value up
to this threshold are presented in the results section of
this paper. When applying the K-Means clustering
algorithm, the number of clusters must be manually
selected. In this case, three clusters were found to
produce the most meaningful results. All statistical
testing was executed using the IBM SPSS software.
3 RESULTS AND DISCUSSION
T-tests of the outcome variables “400m ratio” and
“800m ratio” showed an overall improvement of 7–8%
in adaptation to the reference route after
implementation, compared to before. These results are
statistically significant, but the observed adaptations
remain relatively small. Linear regression analysis of
the dataset for the post-implementation period
indicates that different variables correspond to varying
degrees of adaptation to the reference routes. Table 3
shows that some variables contribute positively to
alignment with the reference route, while others have
the opposite effect.
Table 3. Linear regression models
DV
Model
Adj. R
2
800m
Ŷi_800m = 25,5 + 13,2XCargo – 33,4XPassenger + 23,2XTanker +
23,2XOffshore + 0,3XL
0,316
400m
Ŷi_400m = – 3,9 – 42,5XPassenger + 16,0XOffshore + 0,5XL
0,216
When consulting the standardized coefficients of
the variables, one can see from Table 4 that the vessel's
length is the most dominant determinant of
correspondence with the reference routes.
Table 4. Variables ranked according to standardized
coefficients
Order
Variable
800m ratio
400m ratio
1
L (Vessel length)
0,23
0,33
2
Cargo (ship)
0,22
---
3
Offshore (service vessel)
0,19
0,12
4
Tanker
0,18
---
-1
Passenger (ship)
-0,47
-0,58
It is further evident from the linear regression
models that, although all variables from Figure 3 were
included in the regression analysis, only a few
variables stood out as statistically significant at 0.05
level, mainly those concerning vessel types, as
identified in Tables 3 and 4. These, and related
variables, were then isolated from the datasets in order
to conduct T-tests to quantify the changes for each
variable between the period before and after
implementation.
The results of these T-test are summarized in
figure 4.
358
Figure 4. Illustration of T-test results
The sequence in Figure 4 is organized based on the
results from the 400m corridor. It is clear that offshore
vessels, non-European cargo ships (Cargo-RoW), and
tankers are leading in terms of corridor compliance.
The majority of vessels tend to sail on the port side of
the reference route, except for tankers and Cargo-RoW
vessels, which show nearly equal distribution on both
sides.
We also observe that Norwegian cargo vessels,
offshore vessels, and fish carriers have changed their
traffic patterns following the implementation of the
reference routes, showing an increased level of
compliance with both corridors.
It may seem odd that non-European cargo ships
(Cargo-RoW) and tankers have increased their
starboard ratio without a rise in the corridor ratio. This
could be because these vessel types were already
navigating close to the route corridors before, making
changes within corridor limits less detectable. The
reason for their initial proximity to the reference route
could be these vessels frequency of navigating under a
pilot’s guidance. The reference routes are based upon
the routes preferred by pilots.
The T-test confirms the regression analysis results
for passenger ships. No indication of changed
behaviour after implementation was detected,
indicating these vessels have consistently followed the
same tracks as before the introduction of the reference
routes. In this context, it is important to note that the
vast majority of passenger vessels in both datasets
belong to the Norwegian Coastal Express, which has
daily passages through the area.
The K-Means clustering of the dataset post-
implementation utilized the outcome variables: “400m
ratio”, “800m ratio”, and “Starboard ratio”, and
classifying the dataset into three clusters produced the
most interpretative and analyzable results. Cluster 1
includes passages with little to no adaptation to the
reference routes and is the largest cluster. Cluster 2
consists of passages with the best compliance, while
Cluster 3 covers passages that generally follow the
corridors but tend to stay on the port side of the
reference routes.
The results align with findings from Linear
Regression and T-tests while providing further details
as presented in table 5.
Table 5. Cluster-analysis results
Cluster 1
Cluster 2
Cluster 3
Corridor
compliance
Poor
High
Moderate to
high
Lateral
positioning
Port side
Equal
distribution on
both sides
Mainly
positioned on
the port side
Vessel
characteristics
Passenger
vessels
Tankers
Norwegian
cargo vessels
Fish carriers
Non-European
cargo vessels
(Cargo-RoW)
Shortest ship-
length
No offshore
vessels
No passenger
vessels
Slow speed
passages
Longest vessels*
Medium ship-
length
Occasional
transiting
vessels
High-speed
passages
Single-transit
vessels
Frequent
transiting
vessels
Offshore vessels
European cargo ships
Passages
106
68
90
Cluster metrics**
400m/800m/Stbd
9 / 31 / 4
75 / 92 / 55
45 / 76 / 13
* The dominance of Norwegian Coastal Express vessels may bias
the dataset and affect this outcome. Regression analysis identifies
vessel length as the strongest predictor of route compliance
** Average values of the outcome variables for each cluster
4 CONCLUSION
This study assessed the impact of implementing
Reference Routes for Navigation along the Norwegian
coast by comparing AIS tracking data from before and
after implementation. The results show a statistically
significant, though modest, shift toward greater
compliance. However, vessel characteristics such as
type, length and several other factors significantly
influence the level of compliance.
The length of the vessel is the strongest determinant
for corridor compliance – longer ships show higher
compliance. All vessel categories, except European
cargo ships and Passenger vessels, have altered their
traffic patterns after the implementation of reference
routes. While it is expected that vessels in the category
“Passenger vessels”, primarily dominated by the
Norwegian Coastal Express, maintained their original
routes, it remains unclear why European cargo ships
have not been significantly impacted by the
introduction of the reference routes.
Offshore vessels showed the largest change in
behaviour towards the reference routes, surpassing
tankers and non-European cargo ships in terms of
compliance with the 400m corridor, which serves as the
best indicator of reference route compliance. It is
recommended to investigate the reasons why reference
routes were particularly appealing to offshore vessels
compared to other ship types. The results of such an
investigation could be valuable for determining how to
allocate resources for future work on the reference
routes.
The clustering of the voyages identified three
logical clusters, each characterized by distinct vessel
characteristics. To achieve more uniform traffic
patterns, the most effective approach would be to focus
359
on the vessels in Cluster 1. However, this cluster
primarily includes vessels that frequently transit the
area. It is not unlikely that navigators on such vessels
rely more on their own reference points, hence seeking
less external guidance in form of a reference route.
Therefore, a more impactful strategy could be to focus
on the vessels in Cluster 3, which shows an acceptable
level of corridor compliance but with still room for
improvement.
Overall, this study demonstrates that Reference
routes for Navigation, as a voluntary based traffic
routing measures impacts the traffic patterns, though
vessel flow is still far from uniform. The differences in
route adaptation between vessel types may suggest
that those with a greater need for guidance were the
ones who adopted most to the reference route.
However, achieving complete uniformity may not
even be desirable, as the current solution directs
vessels to the most optimal route from a pilot’s
perspective, while still allowing navigators the
flexibility to plan routes that align with their
operational goals.
The Norwegian Coastal Authority is continuously
working on improvements to ensure that information
about the reference routes and their usability are not
limiting factors. Efforts are in progress to integrate the
routes more seamlessly into Norway’s maritime
reporting and information-sharing platform
(SafeSeaNet).
The results of this study are based on a quantitative
analysis of a limited stretch on one reference route.
Further studies are recommended to validate and
elaborate these results.
REFERENCES
[1] Aarset, M. V., Johannessen, L. K.: On Distributed
Cognition While Designing an AI System for Adapted
Learning. Front. Artif. Intell., vol. 5, Jul. 2022, doi:
10.3389/frai.2022.910630.
[2] Digital route service for navigation. Accessed: Feb. 19,
2025. [Online]. Available: https://routeinfo.no/
[3] Jacobsen, M.: Referanseruter for sjønavigasjon – “To
follow, or not to follow?” En kvantitativ studie av
implementeringens effekter på trafikkmønster, og
hvordan eventuelle endringer og variasjoner kan
forklares. Master thesis, NTNU, 2024. Available:
https://ntnuopen.ntnu.no/ntnu-
xmlui/handle/11250/3179553
[4] Kristić, M. et al.: Zone of Confidence Impact on Cross
Track Limit Determination in ECDIS Passage Planning.
Journal of Marine Science and Engineering, vol. 8, no. 8,
Art. no. 8, Aug. 2020, doi: 10.3390/jmse8080566.
[5] Lušić, Z. et al.: Standardisation of plotting courses and
selecting turn points in maritime navigation. Promet -
Traffic&Transportation, vol. 26, no. 4, pp. 313–322, Aug.
2014, doi: 10.7307/ptt.v26i4.1437.