259

1 INTRODUCTION

Assessing safety and ease of navigation is a critical task

for ensuring safe and efficient ship traffic on inland

waterways. It is particularly relevant considering the

increasing occurrence of extreme environmental events

due to climate change, such as prolonged periods of

extreme shallow water, as well as the ongoing

development of autonomous ships. In this context, ship

handling simulators are an established tool to perform

navigability analyses in restricted waters with the

support of professional pilots. Typically, an expert

rating is conducted to evaluate safety and ease, in

which the pilots state their perception on the difficulty

of a particular navigation situation [1].

The investigations involve a comparative variant

analysis, aimed at identifying differences in navigation

between two or more hydraulic design variants. As

stated by the Permanent International Association of

Navigation Congresses (PIANC) [1], a systematic

approach using quantitative criteria is necessary for

such a detailed comparative study. The pilots’

judgement, by nature, is highly affected by their

personal condition and expertise, making it only

partially suitable for conducting an accurate analytical

study.

To overcome this issue, PIANC recommends to

quantify safety and ease of navigation with measurable

metrics such as ship kinematics and waterway

conditions, as they influence safety and ease

significantly. Nevertheless, the subjective assessment

of the participating pilots should be incorporated in the

evaluation process.

A Data-driven Method for Assessing Safety and Ease

of Inland Waterway Navigation

L. Spielberger, L. Zentari, G. Göbel & N. Maedel

Federal Waterways Engineering and Research Institute, Karlsruhe, Germany

ABSTRACT: The quantification of safety and ease is an essential aspect of risk assessment in the context of inland

waterway navigation using ship handling simulators. This task involves deriving and analysing quantifiable

parameters from inherently complex manoeuvres, including ship kinematics and waterway-specific measures.

The main challenge lies in establishing a systematic and standardized approach that can be applied to a broad

variety of manoeuvres. In this paper, we introduce a data-driven method for assessing the safety and ease of

inland waterway navigation, including ship and waterway related parameters extracted from simulation data.

What is novel about this method is the spatially-resolved evaluation of the parameters, allowing for precise

identification of local critical situations along the waterway. The method is extended by averaging the parameters

across multiple simulations, which reduces the influence of outliers and highlights critical situations that persist

across simulations. We demonstrate our method's universal applicability with three representative case studies

on German inland waterways: a lock entry in Schwabenheim on the River Neckar, a sharp river bend near

Reibersdorf on the River Danube and an entry into a planned rest harbour in Niedermörmter on the River Rhine.

http://www.transnav.eu

the International Journal

on Marine Navigation

and Safety of Sea Transportation

Volume 20

Number 2

June 2026

DOI: 10.12716/1001.20.02.02

260

Such approach has been adopted by numerous

scholars in recent years. Gronarz et. al [3] carried out

real-time simulations in order to assess the safety and

ease of navigating a section on the River Danube near

Straubing in Germany. Mansuy et. al [4] studied the

safety of turning manoeuvres in a turning basin on the

River Meuse in France. In two separated studies,

Mansuy et. al [5], [6] investigated the safety of

navigation on a section of the River Seine in Paris.

Despite different hydraulic conditions of the

investigations, the approach to quantifying safety and

ease in all of the mentioned studies follows the same

principle. First, a set of dynamic ship-related

parameters, such as rudder angle, propeller rotational

speed and ship position, is recorded across a series of

ship-handling simulations. Second, non-dimensional

safety reserves are defined for each parameter. A

reserve is considered to exist as long as the parameter

remains within its prescribed maximum value. The

higher the reserve, the safer and easier a manoeuvre is

considered to be. Third, the simulation data is analysed

to determine whether each parameter retains an

adequate reserve throughout a given manoeuvre.

Finally, a safety assessment is conducted on the basis

of a mean reserve associated with each parameter in a

simulation. In the studies considered, the authors

defined three safety levels corresponding to a colour

code: green designates a simulation with no

constraints; yellow indicates an acceptable simulation

and red marks an inacceptable reserve. Should any

parameter in a simulation fall below a prescribed safety

level, the manoeuvre is classified as inacceptable.

The authors in [3], [4], [5], [6] considered two

categories of parameters for assessing safety and ease

of navigation. The first category refers to ship-related

parameters, including rudder angle, as well as the

rotational speeds of the main engine and thrusters. The

second category consists of parameters related to the

ship’s surroundings and includes various waterway-

specific spatial measures, such as the distances to

bridge pillars and buoys, under-keel clearance and the

distance to the fairway boundary.

The use of a time-averaged value for each reserve in

a simulation suggests a shortcoming in the above

outlined studies. In consequence, local critical areas

may be difficult to identify. This could lead to a

situation where, even if a manoeuvre is deemed safe,

these critical areas may still impair navigation and

potentially endangering the entire manoeuvre.

In the present study, we overcome this limitation by

introducing a systematic, spatially resolved approach

to quantifying safety and ease of navigation. Our

method is subdivided into three main phases:

1. Identification of ship- and waterway-related

navigational parameters.

2. Definition of a non-dimensional, spatially-resolved

reserve for each parameter.

3. Establishment of safety criteria for each reserve and

subsequent assessment

Furthermore, we extend our method to include an

averaging of reserves across multiple simulations for

each hydraulic scenario. This further refines our safety

assessment by minimizing the effect of outliers and to

highlight critical areas that persist across simulations.

We validate our approach through three

representative ship handling simulation exercises, all

conducted on our in-house simulator, the Advanced

Inland Nautical Simulator 6000 (ANS6000): a lock entry

on the River Neckar in Schwabenheim; traveling a

sharp river bend on the River Danube near

Reibersdorf; entering a planned rest harbour on the

River Rhine near Niedermörmter. We apply our

method to each of these test cases to gain a spatially-

resolved safety assessment, which highlights local risks

for navigation. The simulations were previously

executed by professional pilots with varying degrees of

experience.

Our paper is structured as follows:

− Section 2 details the simulation set up as well as the

test cases that were selected.

− Section 3 presents our methodology including the

definition of key metrics and associated reserves, as

well as the assessment procedure.

− Section 4 presents the results of our methodology

and discusses its advantages and shortcomings.

− Section 5 provides a conclusion summarizing our

main findings and briefly details possible paths for

future research.

2 SIMULATION SET UP

2.1 Ship Handling Simulator

The simulations were conducted using the ANS6000,

developed by Rheinmetall Electronics, as pictured in

Figure 1. It comprises a fully operational inland

waterway ship bridge, including modern ship controls

for rudder, bow thruster, main engine control, auto-

pilot and RADAR. RADAR and autopilot were

developed by Argonics, an Alphatron Marine

company, specializing and guidance and control

devices for inland waterway ships. The environment is

rendered in real time through a 3D graphical interface.

The simulation core has been specifically tailored to

inland waterway operations, using state-of-the-art in-

house ship–ship and ship–waterway interaction

models, including lock-entry forces, channel effects,

and bank effects, among others. Equations of motions

are solved in real time, updating the ship’s position and

orientation within the simulation environment. The

graphical interface, which communicates with the

simulation core, translates these updated ship

parameters in real time.

Figure 1. Ship Handling Simulator at BAW in Karlsruhe

(Courtesy of BAW).

Environmental conditions, such as wind speed,

current, and initial ship position, are set by the operator

prior to each simulation. River current is evaluated

beforehand with a flow solver and integrated in the

simulation environment. The pilot then starts the

261

exercise according to a pre-defined scenario. During

the exercise, communication with the pilot is restricted

to ensure realistic simulation conditions. Afterwards,

the pilot provides subjective feedback on the perceived

ease of navigation. In the event of a severe collision, the

exercise is aborted. For each test case, a series of

exercises is performed under varying conditions such

as ship type, water level and water flow rate. Pilots are

selected based on their knowledge of the investigated

ship class, as well as their experience and familiarity

with the exercise area. Depending on the objective of

each study, at least two pilots are selected with varying

degrees of experience and local knowledge.

During simulations, ship kinematic parameters and

commands are recorded in real time to enable

subsequent analysis and assessment. Table 1 lists a

selection of these parameters recorded with a

frequency of 1 Hz, equivalent to the simulation time

step. These include rudder angle



, engine rotational

speed RPME, percentual thruster rotational speed

RPMT and heading



. Furthermore, the ship’s position

in latitude and longitude is recorded, as well as the

waterway-kilometres (ww-km), which divide the

entire German waterway network.

Table 1. Selection of parameters recorded during ship

handling simulations at BAW with a frequency of 1 Hz.

Parameter

Symbol

Unit

Rudder Angle

Engine Rotational Speed

1/min

Percentual Thruster Rotational Speed

0/00

Heading

2.2 Ship Models

Three different representative inland waterway ships

were used in the investigations. Table 2, Table 3 and

Table 4 list their principal dimensions including length

and breadth, as well as the engine and bow thruster

power, respectively. The ships’ draught varies between

2,00 and 2,70 m depending on the test case. Both types

– the Conférence Européenne des Ministres des

Transports (CEMT) class Va ships and the CEMT class

Vb ship – are operating on the River Rhine and Danube,

on channels in Northern Germany and in the

Netherlands and Belgium [2]. Pictures of the models in

the simulator environment are displayed in Figure 2,

Figure 3 and Figure 4. The CEMT class Va ships are

equipped with one propulsion unit and a twin rudder,

whereas the CEMT class Vb ship is equipped with two

propulsion units and two twin rudders. In addition to

the single hull ships, a coupled convoy consisting of the

CEMT class Va ship Odeon and a barge of type “Europe

type b” towed to the ship’s port side were used for the

Reibersdorf test case. Table 5 lists the principal

dimensions of the coupled convoy.

Table 2. Principal dimensions of CEMT class Va ship Baden-

Württemberg, including engine and bow thruster power.

Length

L [m]

105.00

Breadth

B [m]

11.00

Engine Power

PEng [kW]

1118.00

Bow Thruster Power

PBow [kW]

357.00

Table 3. Principal dimensions of CEMT class Va ship Odeon,

including engine and bow thruster power.

Length

L [m]

110.00

Breadth

B [m]

11.40

Engine Power

PEng [kW]

1624.00

Bow Thruster Power

PBow [kW]

507.00

Table 4. Principal dimensions of CEMT class Vb ship Vigilia,

including engine and bow thruster power. The ship is

equipped with two propulsion units with total engine

power of 2.100 kW.

Length

L [m]

135.00

Breadth

B [m]

11.45

Engine Power

PEng [kW]

2 x 1050.00

Bow Thruster Power

PBow [kW]

972.00

Table 5. Principal dimensions of the coupled convoy.

Length of barge

LB [m]

75.00

Breadth of barge

BB [m]

11.40

Length overall

L [m]

110.00

Breadth

B [m]

22.80

Figure 2. Image of the modelled CEMT class Va ship Baden-

Württemberg (Courtesy of BAW).

Figure 3. Image of the modelled coupled convoy, consisting

of CEMT class Va ship Odeon with a barge on port side

(Courtesy of BAW).

Figure 4. Image of the modelled CEMT class Vb ship Vigilia

(Courtesy of BAW).

2.3 Test Cases

We selected three case studies carried out at BAW as

test cases for the validation of our method, each

characterized by different structural and hydraulic

conditions and therefore posing different navigational

challenges. All involve a comparative variant analysis,

expressing the need for a quantitative safety

assessment.

262

Schwabenheim lock is located near the River

Neckar in the region of Heidelberg in Germany. Figure

5 provides a view of the lock geometry and its

surroundings on a map. The waterway is characterised

by a narrow bend in the approach channel on the east

side of the lock that transitions into the lock entry. At

the time of investigation, the lock passage was only

permitted for ships up to a length of 105.00 m.

Investigations intended to examine whether safe

passage downstream could also be guaranteed with a

ship of 110.00 m length or whether an expansion of the

waterway is necessary. Thus, a comparative study

including two scenarios was conducted: one using

CEMT class Va ship Baden-Württemberg and the other

using CEMT class Va ship Odeon. The simulations

included lock entries and exits and were carried out by

two experienced pilots.

Figure 5. Illustration of the Schwabenheim lock geometry on

a map. The lock is located on the downstream end of the

Neckar channel, close to where it meets the River Neckar

(GDWS/WSV).

Near Reibersdorf is a sharp bend on the River

Danube in Germany. A newly constructed floodplain

channel branches from the river downstream into the

bend, as pictured on the top right of the map in Figure

6. Shortly after construction, very high flow rates in the

floodplain channel, including flood rates, caused cross

currents at the branching point that restricted the safety

and ease of navigation. Simulations were performed in

order to examine the influence of different flood plain

channel flow rates on safety and ease of ship traffic.

Three different scenarios were set up with varying flow

rate Q = {0, 90, 120} m

⁄s. The simulations were

performed with CEMT class Va ship Odeon and the

coupled convoy going downstream. Exercises were

carried out by three experienced pilots.

Figure 6. Illustration of the sharp river bend near Reibersdorf

on a map. A floodplain channel branches from the river

downstream into the bend, shown on the top-right of the

picture (GDWS/WSV).

The access to the planned Niedermörmter rest-

harbor lies directly in the outflow area of the Rees flood

channel on the Lower River Rhine, as pictured in

Figure 7. The river acts as a cross-current for ships

entering or leaving the access channel, increasing the

ships’ drift. At higher water levels, this effect is

amplified by the increased flow in the flood channel.

A comparative study intended to examine safety and

ease of entering or leaving the rest harbour under the

influence of these cross-currents. This test case is

hydrodynamically the most complex as it includes four

scenarios with varying water levels: low, mean low and

two high-water conditions. The CEMT class Vb ship

Vigilia was used for the majority of the simulations,

which were carried out by five experienced pilots.

Figure 7. Illustration of Niedermörmter rest harbour on a

map. The access channel to the rest harbour lies directly in

the outflow area of the Rees flood channel on the Lower River

Rhine (GDWS/WSV).

3 METHODOLOGY

The following ship kinematic parameters are selected

from the recordings of each case study to define ship-

related reserves: the rudder angle δ, the engine’s

rotational speed n and the bow thruster’s rotational

speed nB. For each parameter, we computed its reserve

at every simulation time step. Equation 1 describes the

calculation scheme for the rudder reserve Rδ, the bow

thruster reserve RB and the engine reserve Rn, which

will be further referred to as propeller reserve Rn. In the

equation, Rx,i is the reserve of parameter x at sample i,

xi is the value of parameter x at sample i and xth is the

defined threshold value of x.

=−

(1)

If a ship holds two propulsion units, the resulting

propeller reserve Rn is the mean value of both units, see

equation 2.

(2)

The distance reserve RDist is the only waterway-

related reserve in this study. It indicates the relative

distance of a ship to the nearest waterway boundary

and is computed at each time step. In the event of a

collision, the safety distance is zero.

263

We defined a total reserve RTotal following equation

3. Each reserve is multiplied by an individual

weighting factor Fx and their sum is divided by the

weighting factors’ sum Fx :

B B n n Dist Dist

Total

F R F R F R F R



+ + +



(3)

The weighting factors reflect the importance of each

reserve for a particular scenario and are discussed prior

to each simulation campaign with the pilots. Table 6,

Table 7 and Table 8 list the weighting factors for each

reserve in the scenarios considered in this study.

Table 6. Weighting factors for each reserve in the

Schwabenheim test case

Rδ

RDist

Table 7. Weighting factors for each reserve in the

Reibersdorf test case

Rδ

RDist

Table 8, Weighting factors for each reserve in the

Niedermörmter test case

Rδ

RDist

In all examined manoeuvres, the rudder is the

primary steering unit and is thus assigned the highest

weighting factor. The relevance of the bow thruster for

manoeuvring depends on the case study. In a lock

entry such as in Schwabenheim, the bow thruster is

used in close conjunction with the rudder to achieve

maximum precision, making it indispensable for the

manoeuvre. However, in the Reibersdorf and

Niedermörmter studies, the pilots do not consider the

bow thruster crucial to the manoeuvre, resulting in the

lowest weighting factor being assigned. The use of

propeller also varies significantly across the studies.

Entering a lock requires only the propeller thrust

necessary for manoeuvring, leaving a sufficient

propeller reserve. In the Reibersdorf and

Niedermörmter studies, the propeller is needed

alongside the rudder to cope with the prevailing cross-

currents and thus assigned a medium-high weighting

factor. Finally, the safety distance is a relevant criterion

for identifying bottlenecks and local critical areas.

Cross-current effects in the Reibersdorf and

Niedermörmter studies lead to increased drift toward

the waterway boundary. Depending on the available

navigation space to counteract this drift, the weighting

factor is set to medium or high. A lock entry requires

precise manoeuvring within a confined space, where

brief contact with the lock walls cannot always be

avoided. Since the safety distance may vary depending

on a ship’s manoeuvring ability, a medium weighting

factor is applied.

We established an assessment matrix structured in

three safety levels. The following safety levels are

defined:

Acceptable

Navigation is easy

Tolerable

Navigation is moderately easy

Inacceptable

Navigation is difficult

Gronarz et al. [3] and Mansuy et al. [4], [5], [6]

distinguish the levels by threshold values and assign a

colour code to each. The threshold values are defined

separately for each test case and parameter reserve, as

each requires a different use of control units.

The final safety levels are listed in Table 9, Table 10

and Table 11, for each test case respectively. The rudder

reserve is divided into three equal safety levels due to

the highly dynamic use of rudder. The thresholds for

the bow thruster reserve are either equal to or higher

than those of the rudder reserve, as the bow thruster is

typically used more sparingly but to a greater extent

when engaged. The thresholds for the propeller reserve

are lowest in the Reibersdorf study, as the ship is

sailing en route and must cope with cross-currents

from the floodplain channel. This effect is also reflected

to a lesser extent in the Niedermörmter study, whereas

lower propeller usage is expected in the lock entry,

resulting in higher thresholds. The definition of

thresholds for the safety reserve is described in the

following paragraph. The total reserve is divided into

three equal safety levels throughout all studies,

ensuring a consistent overall assessment.

Table 9. Safety Criteria for the Schwabenheim lock test case

Rδ

RDist

RTotal

70 %

80 %

50 %

100 %

70 %

30 %

50 %

20 %

1 %

30 %

< 30 %

< 50 %

< 20 %

< 1 %

< 30 %

Table 10. Safety Criteria for the Reibersdorf test case

Rδ

RDist

RTotal

70 %

80 %

30 %

100 %

70 %

30 %

50 %

10 %

1 %

30 %

< 30 %

< 50 %

< 10 %

< 1 %

< 30 %

Table 11. Safety Criteria for the Niedermörmter test case

Rδ

RDist

RTotal

70 %

50 %

100 %

70 %

30 %

20 %

1 %

30 %

< 30 %

< 20 %

< 1 %

< 30 %

For the distance reserve, safety margins specified in

the Richtlinien für Regelquerschnitte in

Binnenschifffahrtskanälen [7] were used, aiming for a

consistent assessment throughout multiple test cases.

In inland waterway channels, the minimum distance

between the swept area width of a ship and a sloping

bank should not fall below 1.50 m, for vertical banks

below 4.00 m [7]. These margins are used to mark the

threshold to the lowest safety level. The 4.00 m margin

is applied to the Schwabenheim lock test case, as the

lock approach channel is outlined by a vertical bank.

The approach channel to Niedermörmter rest harbour

is confined by a sloping bank in the simulation set up,

thus the 1.50 m margin is applied. The safety margin

for the Reibersdorf test case is referenced to the

navigation channel instead of the shore, so a 1.50 m

margin is applied. The ship’s breadth defines the

threshold to the highest safety level, creating a

consistent reference across scenarios. A safety reserve

exceeding the ship’s breadth is deemed sufficient in all

scenarios.

Finally, we averaged each reserve over a defined set

of simulations for each scenario. This procedure

provides a representative result for the selected group,

reducing the effect of outliers and helping to identify

critical areas recurring across simulations. Simulations

were filtered by the hydraulic scenario, ship type and

pilot navigating.

264

The ship handling simulations from each test case

were assessed using the introduced method. The

analysis was conducted with respect to waterway-

kilometres (ww-km), enabling direct comparison of

scenarios in relation to local waterway characteristics.

4 RESULTS AND DISCUSSION

We structured this chapter as follows:

1. We demonstrated the applicability of our

methodology for each of the outlined test cases,

including the pilots’ feedback.

2. We discussed the method’s advantages and

shortcomings in terms of highlighting local critical

areas and integrating the human factor

4.1 Spatially referenced safety assessment

4.1.1 Schwabenheim lock

Figure 8 illustrates a simulation of CEMT class Va

ship Baden-Württemberg entering Schwabenheim lock

from upstream on the inland electronic navigation

chart (IENC). The ship’s waterline area shape is

displayed for every 10th simulation time step. Part of

the navigation channel that is coloured green is used to

calculate the distance reserve and cut off right before

the southern lock chamber. Whenever a ship shape

overlaps with the channel’s outline, the shape is

coloured red. As depicted in Figure 8, the ship collides

with the lock walls, which in a ship-handling setup is

not entirely avoidable.

Figure 9 displays the safety assessment for this

specific scenario. The safety levels and colour code

correspond to Table 9. The individual reserves are from

top to bottom the rudder reserve Rδ; the main propeller

reserve Rn; the bow thruster reserve RB; the distance

reserve RDist and finally, the total reserve RTotal. Each

reserve is depicted in relation to the river-km. The

simulation starts at Ne-km 18.30, the narrowing of the

channel begins roughly at Ne-km 17.90 and the middle

of the lock is at Ne-km 17.70.

Figure 8. Illustration of a simulation with CEMT class Va ship

Baden-Württemberg entering Schwabenheim lock on the

IENC. The ship’s waterline area shape is displayed for every

10th time step. Part of the navigation channel that is used to

calculate the distance reserve is coloured green. The red

coloured ship shapes indicate collisions with the lock walls.

Figure 9. Safety Assessment of CEMT class Va ship Baden-

Württemberg entering Schwabenheim lock. The simulation

starts at Ne-km 18.30, where reserves are mostly sufficient.

Reserves start to drop around Ne-km 17.90, rudder and bow

thruster are used to a great extent. The simulation ends in the

lock at Ne-km 17.70, where the ship collides with the lock

walls.

For the first part of the simulation until Ne-km

17.90, the parameters retain sufficient reserve with

short sections of higher rudder and propeller usage.

The reserves start to drop around the channel

narrowing up until the lock entry, where rudder and

bow thruster are used to a great extent. The collision

with the lock walls between Ne-km 17.70 and 17.80 is

clearly visible. The total reserve decreases from

acceptable to tolerable over the course of the

simulation

A simulation of CEMT class Va ship Odeon entering

Schwabenheim lock from upstream on an IENC is

pictured in Figure 10. The corresponding safety

assessment is presented in Figure 11. Both figures

reveal a collision with the lock walls. In contrast to

Figure 9, the rudder is used more frequently to a great

extent, whereas the propeller retains a sufficient

reserve. The total reserve is almost identical to the

previous simulation.

Figure 10. Illustration of a simulation with CEMT class Va

ship Odeon entering Schwabenheim lock on a navigation

chart. Figure description remains the same as for Figure 8.

The red coloured ship shapes show that the ship collides with

the lock walls.

265

Figure 11. Safety Assessment of CEMT class Va ship Odeon

entering Schwabenheim lock. Reserves are mostly sufficient

towards the start and drop at Ne-km 18.00. The rudder shows

the lowest reserves during the simulation.

Figure 12 and Figure 13 display the averaged

reserves for Baden-Württemberg and Odeon over a

total of five simulations each, which were performed

by the same pilot. The figures reflect the observed

trends: the reserves are sufficient at the start of the

manoeuvre but decrease around the channel

narrowing up until lock entry. The distance reserve

indicates that collisions with the lock walls cannot be

excluded. Overall, the reserves suggest only minor

risks to safety and ease at the approach channel

narrowing which is consistent with the pilot feedback.

However, neither reserve reveals a noticeable

difference in navigation between the two ship types —

the key objective of the investigation.

Figure 12. Averaged safety assessment of CEMT class Va ship

Baden-Württemberg entering Schwabenheim lock over five

simulations. Reserves only decrease significantly around the

channel narrowing and the lock entry.

Figure 13. Averaged safety assessment of CEMT class Va ship

Odeon entering Schwabenheim lock over five simulations.

Reserves are sufficient, except for the strong use of rudder

from Ne-km 17.90 to 18.00 and the collision with the lock

walls.

4.1.2 Reibersdorf river bend

Figure 14 illustrates a simulation of the coupled

convoy travelling downstream Reibersdorf river bend

on the IENC. The floodplain channel flow rate in this

scenario is Q = 0 m

⁄s. As before, the ship’s waterline

area shape is displayed for every 10

simulation time

step. The navigation channel used to calculate the

distance reserve is coloured green. As to be seen by the

red coloured ship shapes, the ship surpasses the

channel outline during the change of direction from the

left-hand to the right-hand bend. Overall, the coupled

convoy occupies much of the available navigation

space, particularly in the narrower right-hand bend.

Figure 14. Illustration of a simulation with the coupled

convoy travelling downstream Reibersdorf river bend. The

floodplain channel flow rate is Q = 0 m

⁄s. The ship’s

waterline area shape is displayed for every 10th time step.

The navigation channel is coloured green. The ship surpasses

the channel outline in the transition from the left-hand to the

right-hand river bend.

The safety assessment for this scenario is depicted

in Figure 15. The safety levels and colour code

correspond to Table 10. The simulation starts around

Do-km 2320.00 and ends at Do-km 2318.00. The

floodplain channel branches into the river roughly at

Do-km 2318.30. The assessment reveals that reserves

remain mostly sufficient during the left-handed river

band. In the transition phase, rudder usage increases as

the ship reaches the channel outline. Propeller thrust is

raised until the floodplain channel inflow is passed,

where the rudder usage increases once more. The bow

thruster reserve remains sufficient throughout the

entire manoeuver.

266

Figure 15. Safety Assessment of the coupled convoy

travelling downstream Reibersdorf river bend with a

floodplain channel flow rate of Q = 0 m

⁄s. Sufficient reserve

remains from the simulation start at Do-km 2320.00 to the

change of direction, where rudder usage increases as the ship

reaches the channel outline. Propeller thrust is raised until

the channel inflow is passed at roughly Do-km 2318.30,

where rudder usage increases once more. Bow thruster

reserve remains completely sufficient.

A simulation of a simulation with the coupled

convoy for a floodplain channel flow rate of Q = 120

⁄s is depicted in Figure 16. In contrast to Figure 14,

the channel outlined is surpassed twice during the

right-handed river bend. The corresponding safety

assessment is presented in Figure 17. Rudder is

engaged frequently throughout the manoeuver,

leaving only low reserve. Propeller and bow thruster

show the same pattern as in Figure 15.

Figure 16. Illustration of a simulation with the coupled

convoy and a floodplain channel flow rate of Q = 120 m

⁄s.

The channel outline is surpassed twice during the right-

handed river bend.

Figure 17. Safety Assessment of the coupled convoy

travelling downstream Reibersdorf river bend with a

floodplain channel flow rate of Q = 120 m

⁄s. Rudder is

engaged frequently throughout the manoeuver, while

propeller and bow thruster reserve show the same pattern as

in Figure 15.

The averaged reserves over five simulations for the

Q = 0 m

⁄s scenario are displayed in Figure 18. The

figure reveals that the bow thruster was used in at least

one simulation towards the floodplain channel inflow.

Figure 19 illustrates the averaged reserves for the Q =

120 m

⁄s scenario, containing four simulations. The

rudder is frequently used over the entire manoeuver,

while the propeller is engaged the most while

travelling the right-hand river bend. In contrast to

Figure 17, safety distances leave higher reserve and

show fewer points of contact with the channel outline

over a series of simulations.

Overall, both scenarios indicate sufficient reserves

through the left-hand river bend. The right-hand bend

is more challenging, as it demands greater use of

control units, particularly the propeller, and occasional

contacts with the channel outline might occur. The total

reserves do not differ significantly between scenarios,

suggesting that safety and ease remains steady across

both floodplain channel flow rates. This is consistent

with the pilot’s feedback, stating that travelling the

right-handed river bend poses a navigational

challenge, while no difference in navigation between

the two scenarios was perceived.

Figure 18. Averaged safety assessment of the coupled convoy

for a floodplain channel flow rate of Q = 0 m

⁄s over five

simulations. In contrast to Figure 15, the bow thruster was

engaged towards the floodplain channel inflow.

Figure 19. Averaged safety assessment of the coupled convoy

for a floodplain channel flow rate of Q = 120 m

⁄s over four

simulations. While the rudder is frequently used, the

propeller is engaged the most during the right-hand river

bend. In contrast to Figure 17, safety distances leave higher

reserve and show fewer points of contact with the channel

outline.

267

4.1.3 Niedermörmter rest harbour

Figure 20 illustrates a simulation of CEMT class Vb

ship Vigilia entering Niedermörmter rest harbour at

low water level. As before, the ship shape is displayed

for every 10th simulation time step. As the available

navigation area changes with different water levels,

individual waterway outlines were set up for each of

the four considered water levels to calculate the

distance reserve. The navigation area available at low

water is coloured green in the figure. As depicted, the

ship stayed within the available navigation area.

Figure 20. Illustration of a simulation with CEMT class Vb

ship Vigilia entering Niedermörmter rest harbour at low

water level. The ship’s water area shape is displayed for

every 10th time step. The available navigation area at low

water level is coloured green; it is not surpassed in the

simulation.

The safety assessment for this scenario is presented

in Figure 21. The safety levels and colour code

correspond to Table 11. A reference axis was used for

the rest harbour analysis and is referred to NieMoe-km

in the following. The simulation starts at NieMoe-km

1.20, the outflow area of the flood channel is at

NieMoe-km 2.00 and the end of the harbour channel is

reached at NieMoe-km 2.30. While sufficient reserve is

maintained for the first half of the simulation, a

significant reduction in rudder and propeller reserves

occurs towards NieMoe-km 1.90 due to the turning

manoeuvre. Reserves recover again after passing the

flood channel outflow. The total reserve reflects these

observations clearly. Bow thruster and safety distances

remain sufficient throughout the entire simulation.

Figure 21. Safety Assessment of CEMT class Vb ship Vigilia

entering Niedermörmter harbour at low water level. The

simulation starts at NieMoe-km 1.20, the outflow area of the

flood channel is at NieMoe-km 2.00 and the end of the

harbour channel is reached at NieMoe-km 2.30. Reserves

remain sufficient except for a drop in rudder and propeller

reserves from NieMoe-km 1.80 to 2.10 due to the turning

manoeuvre.

Figure 22 illustrates a simulation with Vigilia at

high water mark I (German: ‘Hochwassermarke I’), a

regulatory threshold for a specific high water level

used in inland navigation on German waterways. At

this water level, parts of the surroundings are flooded,

leading to a larger navigational space. Again, the ship

stays withing the available navigation space. Figure 23

depicts the corresponding safety assessment. For most

part of the simulation, control units are used to a lower

extend except for a brief and strong use of rudder

between NieMoe-km 1.80 and 1.90.

Figure 22. Illustration of a simulation with CEMT class Vb

ship Vigilia entering Niedermörmter harbour at high water

mark I (German: ‘Hochwassermarke I’). Due to the flooding,

the available navigation space is larger than at low water

level. The ship stays withing the available navigation space.

Figure 23. Safety Assessment of CEMT class Vb ship Vigilia

entering Niedermörmter harbour at high water mark I.

Reserves remain mostly sufficient except for a brief and

strong use of rudder between NieMoe-km 1.80 and 1.90.

The averaged reserves for low water level over four

simulations are displayed in Figure 24 and for high

water mark I over two simulations in Figure 25.

Simulations were performed by the same pilot. Both

figures confirm the observations from the individual

simulations: reserves are sufficient throughout the first

part of the manoeuvre until rudder and propeller

reserves decrease in the turning between NieMoe-km

1.90 and 2.10. Comparing Figure 24 and Figure 25, the

data indicate that higher reserve is maintained at high

water mark I than at low water level. This suggests a

lower level of safety and ease at low water level.

268

Interestingly, this is confirmed through pilot feedback:

at Niedermörmter, higher flow velocities associated

with high water levels from the flood channel improve

the ship’s turning ability, reducing the demand of

steering devices.

Figure 24. Averaged safety assessment over four simulations

with CEMT class Vb ship Vigilia entering Niedermörmter

harbour at low water level. Merely rudder and propeller

reserve decrease in all averaged simulations around NieMoe-

km 1.90.

Figure 25. Averaged safety assessment over two simulations

with CEMT class Vb ship Vigilia entering Niedermörmter

harbour at high water mark I. Merely rudder reserve

decreases throughout the averaged manoeuvres at NieMoe-

km 1.90.

4.2 Discussion of advantages and shortcomings

The test cases demonstrate that navigability reserves

are highly sensitive to local waterway conditions. In

Schwabenheim, for instance, low reserves emerge

immediately before the lock entrance, predominantly

affecting the rudder and distance reserve due to the

channel’s narrower cross-section at this point. At

Niedermörmter, the outflow area of the flood channel

produces the most critical reductions in reserves

whereas at Reibersdorf the resulting cross current does

not have an impact on safety and ease. Interestingly,

higher flow rates had contrasting effects in different

locations: while they had no severe impact on ship-

dynamic parameters at Reibersdorf, they enhanced

manoeuvrability at Niedermörmter when entering the

rest harbour. These site-specific results cannot be

detected without spatial resolution of reserves.

A mean value criterion is there for insufficient

because it does not include a spatial resolution.

Consider the Niedermörmter rudder reserves depicted

in Figure 21 and Figure 23: at low water, the mean

value is 67% and at high water it is 80%. These figures

suggest that low water means safer navigation, but

they conceal the sharp reduction in reserves near the

flood channel outflow, which makes low water level

the more challenging condition in practice. Similarly,

the average rudder reserves at Reibersdorf with 75%

for Q = 0 m

⁄s in and 64% for Q = 120 m

⁄s in Figure 17

indicate safer navigation without any floodplain

discharge. However, the mean values do not indicate if

this matter can actually be linked to the floodplain

channel. Consequently, reserves must be evaluated in

a spatially resolved manner to enable detailed

comparisons between scenarios and provide

policymakers with valuable information on hydraulic

measures.

Averaging across multiple simulations that were

performed by the one pilot deepens the analysis by

reducing the impact of outliers and distinguishing

recurrent impairments from incidental events in a

single simulation. In our study, the feedback from the

pilots proved necessary for correctly interpreting the

results of the data analysis. However, it should be

noted that the pilot sample must be sufficiently large

and representative; a small or homogeneous group

presents the risk of confusing a shared navigational

habit with a hydraulic constraint. Averaging across

pilots could confirm whether local critical situations

arise consistently and independent of human factors,

and should be explored in future research.

A major advantage of our method is its flexibility:

weighting factors can be adjusted to fit the

requirements of various scenarios and additional

reserves, such as water depth or parameters relevant to

autonomous shipping, can be incorporated. In cases

like turning manoeuvres, which are confined to a

relatively small investigation area, a time-based

analysis seems more appropriate. The framework can

also adapt to different ship types. Furthermore,

qualitative feedback from pilots collected prior to each

exercise, can be translated into quantitative metrics,

providing a basis for reserve thresholds in operational

experience. Nevertheless, the selection of weights and

thresholds remains largely subjective at this stage. A

systematic sensitivity analysis and cross-site validation

would be required in the future before the method

could be applied in a prescriptive manner in regulation

or infrastructure design.

Finally, including a total reserve in the evaluation is

more informative than assessing rudder, bow thruster

and engine reserves in isolation, since manoeuvrability

is a composition of all three. Whether a dedicated

manoeuvring reserve should be defined separately is

an open question and a valuable area for future

research.

5 CONCLUSIONS

In the present study, we introduced a novel data-

driven method to quantify safety and ease of inland

waterway navigation using ship handling simulator

data. Our method is based on the evaluation of ship

269

kinematic parameters as well as waterway-related

parameters, translated into non-dimensional, spatially

resolved reserves. We considered rudder angle, engine

and bow thruster RPM, and the relative distance

between the ship and the closest waterway boundary.

For each parameter, we computed its reserve at each

simulation time step. We combined these non-

dimensional reserves into a single metric, with each

reserve weighted to reflect the importance of a

parameter in a given manoeuvre. Following the works

of PIANC, Gronarz et al. [3] and Mansuy et al. [4], [5],

[6], we defined three safety levels: acceptable, tolerable

and inacceptable. We analysed each test case with

regard to the ship’s position along the waterway. The

results offered a spatially resolved assessment,

enabling accurate identification of local navigational

bottlenecks.

We validated our approach on three representative

test cases on German inland waterways: a lock entry on

the River Neckar at Schwabenheim, a sharp river bend

on the River Danube near Reibersdorf, and a rest

harbour entry on the River Rhine at Niedermörmter.

Each case involved multiple pilots with varying

experience and multiple hydraulic scenarios,

demonstrating the method's robustness across

different navigational conditions.

Our study highlighted that a mean reserve per

simulation may be insufficient for safety assessment.

Critical navigational situations are more accurately

identified through spatial resolution. This was

illustrated clearly at Niedermörmter, where mean

reserve values suggested safer navigation at low water,

while spatially resolved reserves revealed the opposite.

Moreover, the analysis showed that low reserves

exclusively concentrate around the floodplain channel

outflow. Averaging reserves across multiple

simulations further refines the assessment by reducing

the influence of outliers and helping to distinguish

systematic hydraulic constraints from incidental

events in a single simulation. Pilot feedback proved

essential for correctly interpreting the quantitative

results and remains an integral part of the assessment

process.

Our methodology is flexible by design: weighting

factors and safety thresholds can be adapted to

different scenarios and ship types, and additional

reserves may be incorporated to further extend the

method. A time-dependent assessment is equally

possible, as the recorded data carries a temporal

dimension alongside the spatial one.

Future work should thus address the application of

the methodology to localized navigational challenges

such as manoeuvres in turning basins. A quantification

of the human factor, for instance by averaging

simulations across pilots or quantifying qualitative

feedback, should be further studied in order to

distinguish individual piloting behaviour with critical

sections along the waterway.

ACKNOWLEDGEMENTS

The results presented in this paper originated from federally

funded research projects by the Federal Waterways and

Shipping Administration (WSV). Funding and support are

acknowledged. The authors would like to thank Christian

Metz, Sabine Schlenker and Sandra Stober from BAW for

providing data relevant to the analyses.

REFERENCES

[1] PIANC. Design guidelines for inland waterway

dimensions. InCom Working Group Report Nr. 141.

Brussels, Belgium: PIANC, 2019, pp. 45-46

[2] B. Söhngen. Driving Dynamics of Inland Vessels: Vessel

Behaviour on European Inland Waterways and

Waterway Infrastructure with Special Respect to German

Waterways. Karlsruhe, Germany: BAW, 2016, pp. 15-16.

[3] A. Gronarz, M. Tenzer and T. Elsner. “Investigation of

Safety and Ease of Traffic on the River Danube by Real-

Time Simulations”. In: International Conference on

Marine Simulation and Ship Manoeuvrability (MARSIM),

2018, pp. 238–246.

[4] M. Mansuy; M. Candries; K. Eloot; B. Wéry.

“Optimization of a Ship Turning Basin Using Real Time

Simulations – A Case Study for the Quai Des Trois

Fontaines (Chooz, France)”. In: TransNav 13.2 (2019), pp.

357–363. DOI: 10.12716/1001.13.02.12

[5] M. Mansuy; M. Candries; K. Eloot; S. Page. “Simulation

Study to Assess the Maximum Dimensions of Inland

Ships on the River Seine in Paris”. In: PIANC. 2022, pp.

186–200. DOI: 10.1007/978-981-19-6138-0_17.

[6] M. Mansuy; M. Candries; K. Eloot; S. Page. “Simulation

Study to Assess the Effect of Ship Beam on the Navigable

Flow Conditions in Paris”. In: TransNav 17.1 (2023), pp.

25–31. DOI: 10.12716/1001.17.01.01.

[7] Bundesministerium für Verkehr, Bau und

Stadtentwicklung, Richtlinien für Regelquerschnitte von

Binnenschifffahrtskanälen. Bonn, Germany, 2011, p. 2