Content uploaded by Marius Ulm
Author content
All content in this area was uploaded by Marius Ulm on Apr 28, 2020
Content may be subject to copyright.
FIELD MEASUREMENTS IN THE KIEL CANAL, GERMANY: SHIP1
WAVES, DRAWDOWN AND SEDIMENT TRANSPORT2
Marius Ulm1, Sebastian Niehüser2, Bernhard Kondziella3, Arne Arns4, Jürgen Jensen5, and3
Klemens Uliczka6
4
1M.Sc., University of Siegen, Paul-Bonatz-Str. 9-11, 57076 Siegen, Germany, E-Mail:5
marius.ulm@uni-siegen.de6
2M.Sc., University of Siegen, Paul-Bonatz-Str. 9-11, 57076 Siegen, Germany, E-Mail:7
sebastian.niehueser@uni-siegen.de8
3Dipl.-Ing. (FH), Federal Waterways Engineering and Research Institute, Wedeler Landstr. 157,9
22559 Hamburg, Germany, E-Mail: bernhard.kondziella@baw.de10
4Dr.-Ing., University of Siegen, Paul-Bonatz-Str. 9-11, 57076 Siegen, Germany, E-Mail:11
arne.arns@uni-siegen.de12
5Prof. Dr.-Ing., University of Siegen, Paul-Bonatz-Str. 9-11, 57076 Siegen, Germany, E-Mail:13
juergen.jensen@uni-siegen.de14
6Dr.-Ing., Federal Waterways Engineering and Research Institute (retired), Wedeler Landstr. 157,15
22559 Hamburg, Germany, E-Mail: klemens.uliczka@baw.de16
ABSTRACT17
Ship waves and ship-induced flows are the main hydrodynamic loads on waterway beds and18
embankments. But the underlying physical processes are not fully understood yet. Recent field19
measurements, conducted in the Kiel Canal, Germany, allow a better understanding of these loads20
and the resulting (ship-induced) sediment transport. The measurements include high-resolution21
time series of pressure, three-dimensional flow velocities, and turbidity, collected using stationary as22
well as vessel-mounted sensors. The focus of the paper is on two aspects: First, existing drawdown23
1 Final Draft
estimation approaches are reviewed and validated against field measurements. Based on this, a new24
approach is derived to improve the general description of ship waves in confined waters. Second, a25
new approach to estimate the ship-induced sediment transport in the Kiel Canal is developed using26
turbidity and flow measurements and validated against dredging volumes. Our results show that27
about 10% of of the totally transported sediment volume in the Kiel Canal can be attributed to28
ship traffic, while the remaining volume is mainly transported during regular dewatering periods.29
This paper provides an empirically based method to estimate ship-induced sediment transport in30
artificial waterways as basis for future canal management strategies.31
INTRODUCTION32
The maintenance of waterways is an enduring and expensive task for federal authorities. Run-33
ning costs are incurred because of permanent dredging and disposing of accumulated sediments34
as well as repair and reinforcement of bank revetments. Annually, more than 40 ×106m3have to35
be dredged from German coastal waterways to maintain safe ship navigation and port operation36
(Weilbeer 2014), occasioning costs of several million Euros. Up to 20% of the total dredging37
volume arise from the Kiel Canal alone (Brockmann et al. 2008), one of the most frequented38
artificial waterways in the world. Lowering running dredging and maintenance costs can mainly39
be achieved in two ways. First, ship-induced loads on waterways need to be reduced. Here, ship-40
induced loads are defined as the hydrodynamic effects resulting from a passing vessel including41
water level depression, flow velocities and ship waves. For example, Rapaglia et al. (2015) propose42
stricter speed limits for vessels in the Venice Lagoon, Italy, to reduce ship-induced loads on the43
waterway beds in the lagoon. Pesce et al. (2018) use these findings as one component within44
their analysis of alternative navigation routes. Second, running costs could also be lowered by45
enhancing current waterway design approaches but therefore detailed knowledge of the underlying46
processes is required. Ship-induced loads appear to have a determining influence on future ship47
traffic management. In order to meet these challenges, the German Federal Waterways Engineering48
and Research Institute (BAW) started a joint research project including a field campaign to answer49
these open research questions to (i) quantify the ship-induced loads on waterways and to (ii) derive50
2 Final Draft
management recommendations for responsible authorities. Specifically, the drawdown and the51
ship-induced amount of the totally transported sediments are addressed within this study. A chal-52
lenge of future ship traffic management will be the minimization of the impact of individual vessels53
on waterways, for which a quantification of the ship-induced loads is essential. The described54
investigations summarize the main results of the research project based on the field measurements.55
First laboratory and field experiments of the BAW on ship-waterway interaction date back to56
the 1990s showing that ship-induced flows and pressure changes lead to resuspension and transport57
of bed sediments (Flügge and Uliczka 1996). Since physical transport processes and the impact of58
passing ships on these processes were not fully explored yet, the need for further research in the field59
of ship-induced sediment transport was then already pointed out e.g. by Flügge and Uliczka (1996).60
A key challenge is the separation of vessel-resuspended sediment from the total amount of sediment61
within the water column. Measuring vessel loads under laboratory-like conditions are required for62
distinguishing between both components, since natural influences like tidal or discharge currents63
can be neglected. The Kiel Canal provides these laboratory-like conditions since both ends of the64
canal are closed by locks and dewatering currents occur only at specific times.65
In 2012, the BAW conducted a field campaign to assess the aforementioned research questions66
for the Kiel Canal by recording hydrodynamic loads as well as sedimentological parameters of about67
500 vessel passages (Uliczka and Kondziella 2016). Similar field campaigns show the importance68
of a detailed assessment of waterway morphology. Zaggia et al. (2017) highlight that heavy ship69
traffic is the main driver of shoreline changes at the Malamocco Marghera Channel in the Venice70
Lagoon. Vessel-induced drawdown flows led to an estimated total sediment loss of 1.19 ×106m3
71
within 47 years in this area. The relationship between drawdown and sediment resuspension was72
also described by Göransson et al. (2014), who investigated ship-induced loads on bed and banks73
of the Göta Älv waterway in Sweden. Houser (2011) uses field measurements as well and focuses74
on ship waves and associated sediment transport in the estuarine Savannah River, USA. Houser75
(2011) highlights differences between subcritical and supercritical cruise and shows the impact of76
turbulent flow velocities expressed as turbulent kinetic energy. Houser (2011) concluded, that there77
3 Final Draft
is a strong dependency between an increasing turbulent kinetic energy leading to an accelerated78
sediment resuspension.79
The ship-induced hydrodynamic loads on waterways are primarily caused by the movement of a80
vessel relative to the movement of the surrounding water body. In shallow and confined waters like81
harbors, propeller wash is also one of the most important loads, as described e.g. by Hamill et al.82
(1999). The presented research focuses on large waterways, where propeller wash can be neglected83
due to a sufficient under-keel-clearance and a constant vessel speed. Hence, it is essential to improve84
the understanding of ship wave generation as the main hydrodynamic load. In general, ship waves85
are separated into a primary wave system, which is characterized by a remarkable water level86
depression called drawdown with long periods and a secondary wave system with shorter periods.87
The drawdown occurs due to the ship’s water displacement and consequently the ship reduces the88
canal cross-section. Since the volumetric flow rate is constant, flow velocity increases in the area89
where the displacement reduces the canal cross-section. Following Bernoulli’s principle, the water90
surface elevation is reduced in these areas with increased flow velocities. The so called return flow91
between ship and canal bed may exceed sediment-dependent critical flow velocities resulting in a92
resuspension and transport of bed material. The wave systems additionally impact the waterway93
banks. Fig. 1 a) shows the resulting characteristic wave pattern of a vessel in displacement mode.94
Fig. 1 b) shows corresponding dimensions which either describe the wave pattern (e.g. drawdown95
𝑧𝑆, bow wave 𝑠𝐵, stern wave 𝐻𝑠), or directly influence the wave heights (e.g. relative vessel speed 𝑣𝑟 𝑠
96
and the aforementioned ship cross-section 𝐴𝑆and canal cross-section 𝐴). The blocking coefficient97
𝑛combines ship and waterway parameters to account for the interaction between the vessel and the98
characteristics of the observed cross-section. The coefficient is defined as the canal cross-section99
divided by the ship cross-section at the considered location. Furthermore, the distance between100
sailing line and bank determines whether the primary or the secondary wave system causes higher101
impacts on the bank.102
Theoretical approaches for the estimation of ship wave heights base on fundamental principles103
of hydraulics: conservation of mass, energy, and momentum. The application of the consequential104
4 Final Draft
laws (i.e. continuity law, Bernoulli’s law) requires model assumptions like a stable return current105
and negligibleness of secondary waves, friction losses, and vessel dynamics like yaw, pitch, and106
roll movement. Several published studies developed approaches for the estimation of ship-induced107
wave heights, generally described using the drawdown 𝑧𝑆, but building on different assumptions and108
boundary conditions. Commonly used approaches were developed by Krey (1913), Constantine109
(1960), Bouwmeester et al. (1977), Gelencser (1977), Dand and White (1977), Führböter et al.110
(1984), and PIANC (1987). All equations describe, in simplified terms, the drawdown 𝑧𝑆with111
a function of waterway geometry, vessel geometry, and vessel movement (Jensen 1998). The112
mathematical descriptions of the drawdown 𝑧𝑆can be based on purely analytically determined113
approaches or existing physical relationships. Often a combination of both is used. Since these114
approaches are all adjusted to specific conditions by including empirical parameters, the results115
show large differences when calculating the drawdown with each approach using the same database,116
as shown within the discussion. The aforementioned simplifications include but are not limited117
to neglecting friction effects or assuming homogeneous cross-sections, which both influence the118
results and the associated uncertainties. The existing empirical approaches have been developed119
for specific engineering or research issues and mostly rely on less input data. To overcome these120
limitations and to reduce the uncertainties, Rapaglia et al. (2011) and Gelinas et al. (2013) evaluated121
field measurements in the Venice Lagoon based on a regression approach of Schoellhamer (1996)122
and are able to explain 81% of the variability of the drawdown. Based on their findings, Rapaglia123
et al. (2011) conclude that ship speed is more important regarding the drawdown than the ship size.124
Furthermore, the regression approach of Schoellhamer (1996) has also been derived and adjusted to125
the dataset in an earlier state of the presented research project and is actually replaced by the newly126
developed drawdown estimation approach, described in this paper. The present dataset of field127
measurements allows a derivation of an approach completely based on physical laws and represents128
an added value (see Methods: Drawdown estimation).129
Previous studies have also focused on numerical simulations to describe ship-induced waves130
and flows in waterways (e.g. Dam et al. 2008; Gharbi et al. 2010; Ji et al. 2014). However, such131
5 Final Draft
numerical models are generally restricted by the underlying boundary conditions like the ship’s132
hull geometry or inlet/outlet conditions as well as simplified canal geometries. The results help133
to describe flows in detail but still lack the unknown processes contained in the data from field134
measurements.135
STUDY AREA136
The Kiel Canal has a total length of 98.64 km connecting Brunsbüttel at the tidal Elbe River137
with Kiel at the Baltic Sea (see Fig. 2 a) (e.g. Brockmann et al. 2008, Thormählen 2010). In 2012138
about 35,000 vessels travelled through the Kiel Canal. Using the Kiel Canal reduces the route139
between the North Sea and the Baltic Sea by several hundreds of kilometers because the ships can140
avoid travelling around northern Denmark. This highlights the essential meaning of the Kiel Canal141
for regional economics as well as for the trans-European infrastructure. The distance and time142
benefits generate a great economic advantage for international shipping. The shortcut also leads to143
a reduction of CO2emissions due to fuel savings (Brockmann et al. 2008). The Kiel Canal has a144
trapezoidal cross-section with a waterline width of approximately 160 m after the last widening in145
1966. Only the last 20 km off Kiel are still limited to about 100 m. Two-way traffic and overtaking146
of large ships is limited to wider passing places in this canal section. The bottom width of 90 m147
and the mean water depth of 12 m lead to a slope angle of about 1:3. The attached embankment148
has a flat slope which is designed to withstand wave impacts, water level depression, and return149
currents due to the passing vessels (IKC 2013). The locks at both ends of the Kiel Canal and the150
homogeneous cross-section lead to constant and stationary conditions, for example with regard to151
the flow velocities, which only reach a magnitude of approximately 0.15 m/sduring dewatering152
periods. The sediments at the measurement site are mainly comprised of very fine sand (19%), fine153
sand (38.4%), and medium sand (17.6%). This composition covers the principal grain fractions154
in German coastal waterways, so that results are mostly transferable. The following investigations155
focus on these three main grain diameters which yield a cumulated mass fraction of 75%. The grain156
fractions were determined by using four soil samples, taken only during the measurements (Aqua157
Vision BV, Suspended sediment measurements in the Nord-Ostsee-Kanal, unpublished report).158
6 Final Draft
Therefore, no reliable measurements regarding the spatial and temporal development are available.159
However, since there is no significant sediment input from outside the Kiel Canal (e.g. dumping of160
dredged material or larger inflows), it can be assumed that the sediment composition at the bottom161
only changes marginally. Furthermore, the Kiel Canal is regularly kept in the same condition by162
dredging and hence should be quasi-stationary in time. Due to the length of the canal and the163
different landscape types, it is reasonable that there are also different sediment compositions in164
other canal sections which are, however, not relevant for the intended investigations based on a165
specific cross-section.166
In general, the traffic in restricted waterways shows an increase of larger and faster ships167
which generate surface waves and currents that have the potential to cause serious damage to168
embankments and seawalls (e.g. Bezuijen and Köhler 1996; Jensen 1998). The development of the169
traffic, especially in the Kiel Canal, can be described based on two field campaigns performed by170
BAW in 1998 and 2012. The latter will be introduced and analyzed subsequently. Vessels passing171
the Kiel Canal are classified by traffic classes depending on the ship’s geometry. The higher the172
class number, the larger are the vessels. The comparison clearly shows a shift in the distribution173
within the classes. In 1998 65% of all vessels were in class 3, reducing to 40% in 2012. In the same174
time, the percentage of vessels assigned to class 4 and 5 increased from 21% and 7% up to 22%175
and 32%. Since 2014, a new lock is under construction in Brunsbüttel to meet the requirements of176
the increasing traffic after its completion in 2021. This, however, also means that the ship-induced177
loads will increase with larger vessels and therefore higher effort is needed to protect the waterway.178
This affects both the embankments as well as the bottom of the canal requiring more dredging.179
Annual dredging volumes for the western half of the Kiel Canal fairway (west of km 49.460) were180
provided by the local Waterways and Shipping Office (WSA Brunsbüttel, personal communication,181
2016) and are available for the years 2006 to 2008 as well as for 2013 and 2014, as shown in Tab. 1.182
The numbers vary significantly from year to year in the range of a few 1×104m3/aand more183
than 1×105m3/asince the recorded volumes do not differentiate between annual maintenance184
dredging and individual measures, e.g. due to construction. Furthermore, these numbers do not185
7 Final Draft
include dredging works in the Brunsbüttel locks and the lock harbors, which were in the order186
of 6×106m3/ato 7.6×106m3/abetween 1998 and 2006 (Brockmann et al. 2008). The federal187
maintenance budget for the Kiel Canal ranges from 5 to 9 million Euro per year in the past eight188
years (expenses set by the annual federal budget laws 2012 to 2019). These expenses explicitly189
exclude construction and training measures which additionally demand several million Euros each190
year.191
DATA AND METHODS192
Field data193
Field measurements took place in the cross-section at canal station km 17.925 (see Fig. 2 b)194
starting September 17, 2012, 10:00 am (CEST) until the recovery of the measurement equipment195
September 25, 2012, 12:30 pm (CEST). In this period a total number of 509 vessels were recorded.196
The direction of ship traffic was found to be balanced between the locks in Brunsbüttel and Kiel.197
The canal was chosen for the field campaign because of its laboratory-like ambient conditions and198
the presence of a representative sediment grain size distribution for German coastal waterways.199
Due to ship locks at both canal ends, the measurements were neither affected by tidal currents nor by200
naturally varying discharges. As shown in Fig. 3 d), three-dimensional flow velocities range around201
zero before and after the ship-induced disturbances. Considerable flows only occurred during202
regular dewatering periods when water was drained through the Brunsbüttel lock. Furthermore, the203
chosen cross-section is within a straight canal section, so that maneuvering effects are also reduced204
to a minimum. A nearby bridge was used for a photo documentation and to set up a data logger for205
the Automatic Identification System (AIS) signals transmitted by passing vessels. The AIS signal206
provides all relevant ship and maneuvering parameters together with a timestamp, including name,207
identifier, geometry (such as the ship length 𝐿, the beam 𝐵or the draft 𝑇), and current position208
of a vessel. These parameters are referred to as AIS-parameters in the following. Furthermore,209
traveling speed as speed over ground, draft, and direction are also broadcasted by the AIS-system210
and referred to as sailing parameters in the following (IMO 2015).211
8 Final Draft
The instruments for the field campaign were located at the canal bed as sketched in Fig. 2 b).212
The probes are almost equally distributed in one half of the approximately symmetrical cross-213
section: at the center of the bed (probe 3, 82.85 m), near the shore (probe 1, 43.31 m), and between214
the forenamed locations (probe 2, 66.61 m). All mentioned cross-channel distances refer to the215
southern bank of the measurement cross-section.216
At these three locations, conductivity (C), temperature (T), and depth (D) were recorded using217
a CTD profiler with a temporal resolution of 8 Hz. Depth, recorded as pressure and important218
for the further analysis, has an accuracy of ±0.01%, according to the manufacturer information.219
Turbidity of the canal water, recorded in Nephelometric Turbidity Units (NTU), is used as surrogate220
variable of the suspended sediment concentration (SSC). Several water samples were taken during221
the measurements and evaluated in the laboratory to calibrate the conversion from turbidity to SSC.222
The given sediments yield a factor of 1.5 g/m3per NTU, which is used for conversion between223
SSC and NTU throughout the study whenever necessary. Turbidity was measured at 8 Hz using224
optical backscatter probes (Seapoint Turbidity Meter) at three locations in the canal cross-section at225
km 17.925, as shown in Fig. 2 a). The accuracy of the measurement is specified by the manufacturer226
with ±2%. Together with the SSC probes, three 3D flow probes (Nortek Vector) were deployed227
at the canal bed. Each probe measured canal-parallel (𝑥direction), cross-canal (𝑦direction), and228
vertical (𝑧direction) flow velocities at 32 Hz during the entire field campaign. The recorded flow229
velocities can be assumed to be representative for the entire Kiel Canal due to the homogeneous230
cross-section of the canal. The probes are declared with a typical error of 1% for the used measuring231
range of ±2 m/s. Due to favorable conditions at the canal bed with sufficient suspended particles232
and no considerable air bubbles in the measuring volume, the errors are expected to be within this233
range specified by the manufacturer. Furthermore, a vessel-mounted Acoustic Doppler Current234
Profiler (ADCP) was used to estimate the SSC distribution in the cross-section during three days235
of the field campaign. The backscatter measurements were performed and evaluated by Aqua236
Vision BV, resulting in cross-section profiles of SSC distribution of 373 individual timestamps.237
Additionally, another optical backscatter probe was regularly lowered to the canal bed during the238
9 Final Draft
vessel-based measurements to record vertical turbidity profiles. The corresponding sediment grain239
sizes were determined using soil samples taken and analyzed also by Aqua Vision BV (Aqua Vision240
BV, Suspended sediment measurements in the Nord-Ostsee-Kanal, unpublished report). Due to241
the homogeneous and stationary hydrodynamic conditions in the Kiel Canal, it is assumed that the242
measurements only deviate within the typical error ranges specified by the probe manufacturers.243
Furthermore, errors are assumed to be normally distributed, resulting in a constant signal noise244
rather than a drift over time.245
The observed parameters are exemplary shown in Fig. 3 considering the passage of one in-246
dividual vessel. The grey dashed line is located at time step 120 s, a relative time with respect247
to the starting point of the passage. This point was chosen as a common reference point for all248
observed passages and is defined as the point of time at which the maximum drawdown occurs.249
Fig. 3 also highlights the fluctuation in the water level which correlates with the recorded currents250
in 𝑥-direction. After the drawdown, the turbidity starts to increase rapidly, reaching a maximum251
with the stern wave peak and decreasing slowly towards its initial level.252
Drawdown estimation253
Different existing empirical approaches to estimate the drawdown of a vessel in confined254
waters (Krey 1913, Constantine 1960, Bouwmeester et al. 1977, Gelencser 1977, Dand and White255
1977, Führböter et al. 1984, PIANC 1987, and the adjusted regression approach of Schoellhamer256
1996) have been reviewed and used to calculate the drawdown in confined waters using the field257
measurement data in the Kiel Canal. The results are summarized in Fig. 7. All approaches show258
shortcomings in the proper description of the actual drawdown at the ship hull since they were259
developed for either very specific boundary conditions (e.g. Constantine 1960) or for the design of260
revetments focusing on the drawdown at the canal bank (e.g. Bouwmeester et al. 1977).261
All methods to estimate the drawdown 𝑧𝑆have in common, that they are subject to certain262
simplifications such as neglecting friction or return flow assumed being uniform. In addition,263
most of the existing methods are restricted to homogeneous cross-section or small cross-section264
ratios. Based on Bouwmeester et al. (1977), the underlying simplifications and assumptions are265
10 Final Draft
highlighted exemplary. Specifically, the ship speed in an approximately uniform rectangular or266
trapezoidal channel is assumed constant. The channel itself is straight and of infinite length. The267
ship hull is considered uniform, meaning that the effective shape is not taken into account. The268
return flow around the ship is constantly distributed over the entire cross-section of the channel269
and the water level depression in the channel cross-section due to the passing vessel is constant270
along the length of the ship. Finally, the squat in the longitudinal axis of the ship is equal to271
the water level depression and frictional losses are not taken into account. Both, the applied272
mass and pulse conservation laws in Bouwmeester et al. (1977) are based on these assumptions.273
Similarly, the approaches of Krey (1913), Constantine (1960) and Führböter et al. (1984) are based274
on simplifications. However, the approaches may vary due to the differences in the applied physical275
equations or the assumed boundary conditions. Constantine (1960), for example, is based on the276
continuity equation and the law of energy conservation. Führböter et al. (1984) by contrast used the277
law of conservation of energy and momentum. In more empirical approaches, such as Gelencser278
(1977), model observations are combined with dimensionless factors such as the Froude number279
𝐹and correction terms derived from the dataset, e.g. for the wave run-up within the determination280
of the drawdown 𝑧𝑆. An empirically based approach relying on model experiments can be found281
in Dand and White (1977). In order to adjust the Suez Canal for larger ships, three model ships282
were analyzed to investigate the influences of different manoeuvers. Based on their measurements283
of water level depressions, a directly solvable approach was developed. Schoellhamer (1996) focus284
on the relationship between the depth-based Froude number 𝐹ℎand the blocking coefficient 𝑛using285
dimensionless wave heights from the relationship between draft and water depth. The optimization286
yielded an empirical, exponential relationship based on field data. In Rapaglia et al. (2011), a field287
data based optimization also resulted in an exponential relationship of Froud number, blocking288
coefficient and draft.289
For the evaluation of the existing approaches, mainly large vessels were used (𝑛≤60;𝐴𝑆≥290
31 m2). This reduces the analysis to vessels which potentially induce high loads and generate large291
water level fluctuations. Analogous to Niehüser et al. (2016), the calculation of the drawdown is292
11 Final Draft
based on a sample assumed being unaffected by complete measured values from all recorded ship293
passages. To assure that there are no influences from other passages, a time interval of two minutes294
before and 10 minutes after a ship’s passage has been chosen in which no other traffic occurred.295
Another criteria ensures that the distance of a recorded ship passage from the nearest probe is less296
than 2 m. Finally, 251 ship passages were used. 189 of the 251 ships have been maneuvered above297
probe 3 in the center of the canal, 61 vessels over probe 2 and one vessel over probe 1, located close298
to the southern bank.299
In order to estimate the drawdown at the hull of a vessel without using empirical factors or300
exponents, a new approach was derived purely from fundamental physical equations: the law of301
mass conservation in the form of the continuity equation (Eq. 1) and the law of energy conservation302
in the form of the Bernoulli equation (Eq. 2). Both laws describe the conservation for the transition303
between two control sections, indexed with 1and 2:304
𝐴1·𝑣1=𝐴2·𝑣2,(1)305
306
𝑣2
1
2𝑔+𝑝1
𝜌𝑔 +𝑧1=𝑣2
2
2𝑔+𝑝2
𝜌𝑔 +𝑧2,(2)307
where 𝐴is the canal cross-sectional area, 𝑣is the flow velocity, 𝑝is the pressure, and 𝑧is the308
geodetic height component. Constants are the gravitational acceleration 𝑔and the water desity 𝜌.309
For the calculation of the drawdown 𝑧𝑆the two cross-sections 1-1 and 2-2, as shown in Fig. 4 b),310
are used. Eq. 1 can be solved to 𝑣2and rewritten as following with reference to the given cross-311
sections:312
𝑣2=𝐴1·𝑣1
𝐴2
,(3)313
314
(𝑣𝑟𝑠 +𝑣𝑅)=𝐴1·𝑣𝑟𝑠
(𝐴1−𝑧𝑆·𝑏1−𝐴𝑆+𝑧𝑆·𝐵−𝑠·𝐵),(4)315
where 𝐴𝑆is the cross-sectional area blocked by the vessel, 𝑏1is the canal width at the water surface,316
𝐵is the vessel width, 𝑣𝑟𝑠 is the vessels speed relative to the water column, and 𝑣𝑅is the return317
flow velocity around the vessel. The dynamic squat effect, causing a non-uniform, stern-focused318
12 Final Draft
increase in the vessel’s draft by 𝑠and therefore causing an increase in 𝐴𝑆at the far end of the vessel319
is neglected, since the squat could not be recorded as part of the field measurements. In Fig. 4 b),320
the squat is indicated as the difference in draft 𝑇between bow and stern of the ship. However, a321
vessel geometry dependent correction will be applied in the following steps, accounting for errors322
due to model assumptions and with that also accounting implicitly for the squat effect.323
The Bernoulli equation (Eq. 2) can be reduced and rearranged since pressures 𝑝1and 𝑝2are324
equal in an open channel and the velocities are given as depicted in Fig. 4:325
𝑧𝑆=𝑧1−𝑧2=1
2𝑔·𝑣2
2−𝑣2
1,(5)326
327
𝑧𝑆=1
2𝑔·(𝑣𝑟𝑠 +𝑣𝑅)2−𝑣2
𝑟𝑠 .(6)328
Substituting (𝑣𝑟𝑠 +𝑣𝑅)in Eq. 6 with Eq. 4 yields an approach to estimate the drawdown 𝑧𝑆
329
solely based on easily quantifiable parameters. All required parameters are available from survey330
(cross-sectional area of the canal 𝐴1, canal width at the water surface 𝑏1) or from data transmitted331
by the AIS-parameter of the vessels (relative vessel speed 𝑣𝑟𝑠 , vessel width 𝐵, cross-sectional area332
of the vessel 𝐴𝑆derived from draft and width):333
𝑧𝑆=1
2𝑔·"𝐴1·𝑣𝑟𝑠
(𝐴1−𝐴𝑆−𝑧𝑆·𝑏1+𝑧𝑆·𝐵)2
−𝑣2
𝑟𝑠 #.(7)334
The strong simplifications of the chosen approach, i.e. the drawdown is assumed being constant335
over the entire water surface width and the squat is assumed being negligible, result in a general336
underestimation of the drawdown. Therefore the approach has been modified in a second step by337
replacing the canal width 𝑏1with a, yet unknown, reduced canal width 𝑏∗
1for which the assumption338
of a constant drawdown is correct. The errors in the estimation due to neglecting the squat are also339
covered by this approach, since the correction process does not differentiate between error origins.340
To allow later an application of the drawdown approach, 𝑏∗
1has to be derived from given AIS and341
canal parameters for each vessel. The ratio between canal and vessel cross-section 𝑛, which is a342
13 Final Draft
dimensionless parameter describing the canal size in relation to the vessel size, has been chosen as343
a proxy. 𝑛∗is defined as following using water depth ℎto describe the reduced cross-sectional area344
of the canal 𝐴∗
1which can be assumed to be rectangular:345
𝑛∗=𝐴∗
1
𝐴𝑆
=𝑏∗
1·ℎ
𝐴𝑆
.(8)346
Different AIS-parameter and parameter combinations were finally tested using correlation347
analyses to describe 𝑛∗. Largest correlations were found using the ratio between water depth and348
vessel draft ℎ/𝑇for the description of 𝑛∗. A linear relationship with a 95% confidence interval was349
derived:350
𝑛∗=7·ℎ
𝑇−5.4[±1.7].(9)351
Using Eq. 9, 𝑛∗can be estimated for any vessel. 𝑏∗
1can be calculated subsequently (cf. Eq. 8).352
Substituting 𝑏1with 𝑏∗
1in Eq. 7 yields the final drawdown estimation approach which is further353
applied to the observational data:354
𝑧𝑆=1
2𝑔·©«
𝑏∗
1·ℎ·𝑣𝑟𝑠
𝑏∗
1·ℎ−𝐴𝑆−𝑧𝑆·𝑏∗
1+𝑧𝑆·𝐵ª®®¬
2
−𝑣2
𝑟𝑠
.(10)355
Transport estimation356
The main scope of the developed sediment transport estimation approach is the comparison357
of (i) transported sediment volume in a canal with frequent vessel traffic with (ii) the transported358
volume in case that there is no traffic in the same canal. The difference of the scenarios (i) and359
(ii) describes the ship-induced change in sediment transport while the residual volume is due to360
natural transport processes. Applied to the recorded time series, the approach does not yield results361
for individual vessels but rather for the general traffic, which is more interesting and helpful from362
a management perspective. The estimation approach is divided into three steps: 2D turbidity363
approximation, transport estimation based on field data, and transport estimation based on adjusted364
data.365
14 Final Draft
A 2D turbidity approximation is needed to extrapolate from the three point-wise turbidity366
measurements at the canal bed to a turbidity distribution in the canal cross-section. A computational367
grid with quadratic cells with edge lengths of 1 m was chosen to discretize the cross-section, as368
shown in Fig. 6 a). At the three cells where the probes are located the recorded turbidity time369
series are used as boundary conditions. Necessary assumptions to extrapolate the turbidity from370
these cells were made by using the vessel-based recorded SSC distributions as reference. The SSC371
distributions, based on ADCP-measurements by Aqua Vision BV (Aqua Vision BV, Suspended372
sediment measurements in the Nord-Ostsee-Kanal, unpublished report), show the most important373
distribution defining patterns: During periods with no ship-induced disturbance the SSC at the374
water surface is approximately at 2
/3of the bed SSC. Furthermore, the surface SSC does not375
significantly change during periods with disturbance. Therefore, in a first step, a base turbidity376
time series was derived from the point-wise records at the canal bed. Signal noise was smoothed377
using a moving-average low-pass filter and the maximum base turbidity was defined as median of378
the local minimal turning points in the smoothed signal. This method allows a robust estimation379
of the upper limit of the base turbidity using the entire time series, since outliers and temporary380
turbidity increases during ship passages are ignored by the median. The maximum base turbidity381
calculates to 64 NTU and is used to create a base turbidity time series by truncating all values in382
the smoothed signal above the maximum base turbidity (see Fig. 5).383
The resulting time series is multiplied by 2
/3as boundary condition for the surface cells in384
the computational grid. As a fourth sampling point at the canal bed, besides the three probes,385
the turbidity under the ship’s keel is linearly interpolated using the adjacent probes. A quadratic386
function is finally fitted to the sampling points and turbidity values for each bed cell are computed387
using this function. An inverse distance-weighting of the bed cells yields a similar distribution388
as observed with maximum turbidity under the keel and a quick lateral decrease in turbidity. The389
turbidity of the cells between surface and bed was finally computed by linear interpolation between390
the assumed surface turbidity and the calculated bed turbidity. The assumption of an almost linear391
turbidity change between bed and surface was confirmed by the vertical turbidity profiles. This392
15 Final Draft
2D turbidity approximation approach was applied to each time step of the 8 Hz time series. This393
approximation is exemplarily highlighted in Fig. 6 b) while Fig. 6 c) shows the measured SSC394
distribution for the same point of time. The approach was subsequently validated using SSC395
distributions, which were not used for the calibration exercise.396
In the next step, the approximated distributions for each time step are applied to the transport397
estimation. Measured flow velocities are used to estimate the transport of suspended matter along398
the canal axis. Prior to this, preinvestigations were conducted to determine the relevant components399
of the 3D flow data. An Eulerian specification of the flow field during vessel passages reveals, that400
canal-parallel (𝑥direction) flows highly dominate the resulting flow direction while cross-canal (𝑦401
direction) and vertical velocities (𝑧direction) can be neglected. Therefore, only flows in 𝑥direction402
are used for the transport estimation. Furthermore, the measured velocities at the canal bed are403
assumed to be representative for the entire canal cross-section since there were no more detailed404
measurements for the water column during the current field campaign. However, Plate and Keil405
(1971) show that this assumption is reasonable in terms of the vertically uniform distribution using406
vertical flow profiles from the Kiel Canal during vessel traffic. The 32 Hz 𝑥direction flow data407
are down-sampled using a moving average filter to meet the 8 Hz turbidity sampling rate. For each408
timestep 𝑡the sediment transport rate Φ𝑆𝑆𝐶 in g/(m2·s)is finally computed by combining the409
sediment concentration 𝛽𝑆 𝑆𝐶 in g/m3and canal-parallel flow velocity 𝑣𝑥in m/s:410
Φ𝑆𝑆𝐶 (𝑡)=𝛽𝑆 𝑆𝐶 (𝑡) · 𝑣𝑥(𝑡),(11)411
where values of Φ𝑆𝑆𝐶 may be positive or negative, depending on the sign of 𝑣𝑥. Due to the412
alignment of the flow probes with the canal axis, positive 𝑥flows are defined as flows towards the413
lock at the Brunsbüttel canal end. Subsequently transport rates are directed and sediment fluxes414
can be interpreted directly.415
In a last step the previously conducted transport estimation is repeated with adjusted flow and416
SSC data to assess the ship-induced sediment transport. While the already conducted estimation417
16 Final Draft
yields transport rates as occurred, the estimation with adjusted data should describe the sediment418
transport as if there were no vessel traffic during the field campaign. Therefore ambient conditions419
are derived from the measurements and used in Eq. (11). Ambient turbidity conditions are already420
derived from the turbidity measurements for the first step and are reused at this point. The ambient421
flow conditions are derived from the canal-parallel flow velocities using a moving average filter422
with a 60 min moving window. This filter configuration was tested against other window sizes423
and finally chosen because it successfully removes temporary changes in flow velocity, i.e. due to424
passing vessels, but still retains velocity changes on a hourly scale which occur due the periodic425
canal dewatering. The averaged flow time series is then used in the transport estimation together426
with the ambient SSC conditions.427
RESULTS AND DISCUSSION428
Drawdown estimation and approach adjustment429
The measurements at the cross-section in the Kiel Canal show maximum water level depres-430
sions up to 𝑧𝑆=1.01 m and minimum values of 𝑧𝑆=0.07 m. For the comparison between the431
measurements of the drawdown 𝑧𝑆and the values calculated with the empirical formulas each set432
of computed drawdown data is shown as kernel density function in Fig. 7 a). Furthermore, the433
residuals between the measurements and the calculated values are shown as error bars in Fig. 7 b)434
with the mean plotted as dot and the standard deviation as solid line representing the symmet-435
ric distance above and below the mean. The differences in the spread are shown by the kernel436
densities of the different empirical approaches. The observed values of 𝑧𝑆are scattered widely,437
which is shown by the flat density function while the kernel densities of the empirical formulas438
are compressed with significantly higher peaks in the frequencies compared to the observations.439
The peaks of the calculated drawdown also show an offset to the left illustrating that the empirical440
formulas underestimate the observations. This is confirmed by the mean residuals which show441
differences in the magnitude of 14 cm to 25 cm except of the regression approach of Schoellhamer442
(1996). In the latter case, the mean residual of zero is reasonable because the observed values443
of the drawdown 𝑧𝑆have been used to fit the regression coefficients of the depth-based Froude444
17 Final Draft
number 𝐹ℎand the blocking coefficient n. The standard deviation ranges between 8 cm regarding445
Bouwmeester et al. (1977) and 22 cm regarding PIANC (1987). Bouwmeester et al. (1977) also446
exhibits the lowest mean residual. In general, the empirical approaches based on applied physical447
laws (e.g. Bouwmeester et al. 1977) show better results regarding the mean and the spread than448
analytical (e.g. Gelencser 1977) or diagram based (e.g. PIANC 1987) approaches. In particular,449
diagram based approaches tend to weight the individual input parameter higher than others. Ac-450
cordingly, when determining these approaches, it depends especially on the intended use and the451
available data sets (e.g. individual vessel types or model specification). Entirely for all analyzed452
approaches, it is difficult to quantify and discuss exactly to which parameters the residuals can be453
attributed. Considering the age of the approaches beginning with Krey (1913) in the early 20th454
century and ending in late 1980s, it should be taken into account whether the findings obtained from455
the comparison of the empirical approaches are due, for example, to changes in the ship geometry456
over the past decades (see Study Area). With regard to the ship size, an increase of the drawdown457
𝑧𝑆is directly associated and would explain the underestimation of the observed values because458
actual vessel sizes were not even considered in the derivation of the approaches at that time. It459
is likely that the main differences in the residuals presented in this study between the empirical460
approaches and the observed drawdown 𝑧𝑆are related to this fact. Furthermore, scale effects in461
physical models can distort the achieved results. The specific characteristics of the Kiel Canal can462
also cause discrepancies in the comparison. For example, in the Kiel Canal the speed over ground463
is limited by law. In case that the speed over ground is highly weighted in an empirical approach,464
it is reasonable that differences occur. Overall, the relative influence of the required parameters in465
the approaches plays an important role if a dependency to the drawdown 𝑧𝑆is assumed which is466
not provided by the field measurements presented here.467
Since all of the approaches base on limited data and limiting assumptions, a development of a468
more robust analytical approach has been conducted also to adapt for larger vessels. The aim of the469
development and the following optimization was to avoid unsteady correction functions by using470
fundamental physical equations and by improving the basic assumptions of the approach. Taking471
18 Final Draft
a reduced canal width 𝑏∗
1into account, the flow around a vessel is more accurately reflected in the472
approach. The reduced canal width is determined by the relationship between the reduced cross-473
sectional ratio 𝑛∗, water depth, and draft. A further correction of the results can be omitted due to474
this optimization. The application of this approach to the passages in the Kiel Canal shows that the475
drawdown can be estimated based on openly available data, i.e. AIS-based parameters and water476
levels. This allows an application of the approach to remotely collected data, e.g. with a temporary477
AIS antenna or to data provided via online services. Nevertheless, a verification of estimation478
results using direct measurements is still necessary, especially after applying the approach on a new479
canal geometry.480
The developed approach to estimate the drawdown at the hull of a vessel (Eq. 10) has been481
applied to the measured data of the Kiel Canal field campaign. Using AIS vessel data and the known482
waterway parameters, the drawdown has been estimated for each vessel, that passed over one of483
the three probes. The estimation results are compared with the measured water level depression484
and plotted against each others, as shown in Fig. 8 a). In Fig. 8 b) the performance of the approach485
compared to the empirical approaches is shown analogous to Fig. 7. The residuals between the486
measurements and the calculated values are plotted as error bars with the mean plotted as dot and487
the standard deviation as solid line. The mean of 5.6 cm is significantly lower than the mean of488
the seven empirical approaches. The improvement can be assigned to the purely physics-based489
approach with as few assumptions as possible. Nevertheless, the simplification of the complex490
processes results in the remaining error which could be reduced using further measurements. Only491
the regression approach of Schoellhamer (1996) shows a smaller mean error, since this approach492
is explicitly adapted to the Kiel Canal measurements and is therefore not transferable to other493
waterways. The remaining, visible spread in Fig. 8 a), expressed as standard deviation in Fig. 8 b),494
origins from measuring uncertainties as well as local processes (e.g. wind waves) not covered by the495
analytical approach. However, the main goal, improving the overall description of the drawdown,496
compared to the reviewed approaches, was achieved.497
In terms of ship traffic management, empirical approaches for the estimation of the drawdown498
19 Final Draft
yield a practical use in engineering. Analyses of field measurements or model experiments benefit499
from the simplicity of the physical fundamentals, in the form of conservation laws, and can be500
flexibly adapted to the problems at hand. Through the analytical adaptation of the theoretical bases501
of observation to the measurement results, optimizations are possible to an extent that allows further502
practice-oriented application and can be incorporated into rules and technical regulations for the503
design of embankments for waterways.504
Transport estimation505
Both transport estimation runs, with conditions as recorded and with ambient conditions, reveal506
a traffic induced net sediment transport towards the lock at Brunsbüttel. Fig. 9 a) shows the estimated507
sediment transport in metric tons for both runs. Starting at 0, signed values of Φ𝑆𝑆𝐶 for each time508
step were accumulated and plotted against time for the duration of the field campaign. The upper509
solid line indicates the transport estimation for conditions as recorded (with ships) while the lower510
solid line indicates the transport estimation for ambient conditions (without ships). Overall, both511
lines show a positive trend but the development of the cumulative sediment transport over time512
follows a step-wise increase. Four strong increases (“steps”) can be found during the field campaign513
and are highlighted with dotted lines in Fig. 9 a). These steps match with periods of increased514
ambient flow conditions and can be linked to four dewatering events that took place during the515
measurements. In Fig. 9 b) these events are visible as periods where ambient flow velocities reach516
up to 0.15 m/sfor a few hours. During times of no dewatering, flow velocities are around zero and517
noisy disturbances only arise from local and short lasting effects such as wind gusts, since the locks518
of the Kiel Canal prevent tidal currents or naturally varying discharges. Between the dewatering519
events both estimations slightly drift apart resulting in an approximately 10% lower accumulated520
sediment mass for the run with the data reduced by the ship traffic. This proportion of about 10%521
of the entirely transported sediment can be attributed to ship-induced resuspension.522
The annual dredging volumes listed in Tab. 1 are used to validate our estimates. Therefore the523
resulting transport of 580 t during 8 days with 509 vessel passages has to be extrapolated to meet524
the number of ca. 35 000 vessels that used the canal in 2012. The resulting mass of 39 695 t is then525
20 Final Draft
converted to a transported sediment volume using an average sediment density of 1.857 t/m3. The526
average density is calculated from the grain size distribution determined by Aqua Vision BV based527
on sediment samples, and densities for each grain fraction which were estimated e.g. by Eiffert528
et al. (2013), who evaluated soil parameters in order to estimate critical shear stresses for a nearby529
cross-section of the Kiel Canal (km 26.020). The estimation of the totally transported sediment530
volume in 2012 amounts to 21 376 m3. Comparing this estimate with the actual dredging volumes531
in Tab. 1 has to be done carefully, since the actual volumes are from the entire western Kiel Canal532
while the estimate only considers the transport through a specific cross-section. Nevertheless, the533
estimate is in the same order of magnitude as the actual volumes.534
This result gives an insight in the complex sediment transport processes since it quantifies the535
ship-induced amount of transported sediments for the first time. Regarding future management536
strategies, this estimation demonstrates, that only a rather small part of the dredging can be affected537
e.g. by traffic restrictions like lower speed limits, as Rapaglia et al. (2015) exemplary propose for538
the Venice Lagoon.539
CONCLUSIONS540
In the presented paper, we describe newly derived drawdown and sediment transport estimation541
approaches based on field measurements in the Kiel Canal, Germany.542
The analytical approach to estimate the drawdown is based on fundamental physical laws.543
Therefore, it should be transferable to any confined waterway. As another advantage, the appli-544
cation only requires openly available data. Future research should, on the one hand, focus on the545
application of the developed drawdown estimation approaches to different datasets from other field546
measurements in order to cover a wide range of boundary conditions and vessel geometries. On547
the other hand, a better understanding of the wave transformation from the ship to the shore is548
necessary to use our drawdown estimations as input for revetment dimensioning.549
The main finding of the developed engineering approach for estimating ship-induced sediment550
transport is, that ship-induced sediment resuspension combined with even small flow rates yields551
a significant sediment transport in the laboratory-like Kiel Canal. The ship-induced proportion552
21 Final Draft
of the entirely transported mass is in the order of 10%, meaning, in return, that the by far largest553
part is not transported due to vessel-related currents but rather due to ambient conditions. In the554
artificial Kiel Canal these ambient condition are mainly driven by dewatering currents. Since there555
are no more detailed dredging volumes or transport masses available, there was no possibility to556
calibrate the estimation. However, the presented approach is solely based on field measurements557
in the Kiel Canal in order to minimize uncertainties due to assumptions. The presented estimation558
of the impact of ship traffic on the local sediment regime is based on an extensive calibration of the559
NTU measurements against water samples of the considered cross-section in the Kiel Canal. As560
also shown for example by Göransson et al. (2014), the NTU measurements can be used as a proxy561
for the complex process of sediment resuspension since this process directly affects the turbidity562
in the water column. In the first place, the conducted sediment transport estimation is limited to563
the examined cross-section. But the uniformity of the Kiel Canal allows the transfer of the results564
to other sections of the waterway. The approach allows a first estimation of the impact of ship565
traffic on the local sediment regime and a rough estimate of the ship-induced proportion of the566
entirely transported sediment mass. This meets the main objective of the research cooperation by567
providing authorities and decision-makers a better understanding of the interaction between vessel568
and waterway. Future design or management decisions can be based on this understanding. This569
may include traffic restrictions, a widening and/or deepening of the canal, or improved revetment570
design. Furthermore, the gained knowledge about the sediment transport processes can be used571
to improve the dredging schedule. In a next step, the developed sediment transport estimation572
approach should be applied to different waterways to improve the understanding of the impact of573
ambient conditions on the totally transported sediment mass. Tidal rivers, like the Elbe estuary,574
have to cope with large amounts of sediment, which are transported upstream due to tidal currents.575
The quantification of the ship-induced proportion can help to improve the waterway management,576
similar to the management decision support for the Kiel Canal, provided with the presented study.577
DATA AVAILABILITY STATEMENT578
Some or all data, models, or code generated or used during the study are available from the579
22 Final Draft
corresponding author by request. Available are: field measurement data, unpublished measuring580
reports, and scripts used for the presented analyses.581
REFERENCES582
Bezuijen, A. and Köhler, H.-J. (1996). “Filter and revetment design of water imposed embankments583
induced by wave and draw-down loadings.” Geosynthetics: Applications, Design and Construc-584
tion, M. B. De Groot, G. Den Hoedt, and R. J. Termaat, eds., Rotterdam, The Netherlands,585
Balkema, 1007–1023.586
Bouwmeester, J., van de Kaa, E. J., Nuhoff, H. A., and van Orden, R. G. J. (1977). “Recent587
studies on push-towing as a base for dimensioning waterways.” Report No. 194, Delft Hydraulics588
Laboratory.589
Brockmann, J., Heeling, A., Pohl, M., and Uliczka, K.(2008). “The Kiel Canal.” Die Küste, Vol. 74,590
Heide, Holstein, Germany, Boyens, 317–332, <https://hdl.handle.net/20.500.11970/101614>.591
Constantine, T. (1960). “On the movement of ships in restricted waterways.” Journal of Fluid592
Mechanics, 9(2), 247–256.593
Dam, K. T., Tanimoto, K., and Fatimah, E. (2008). “Investigation of ship waves in a narrow594
channel.” Journal of Marine Science and Technology, 13(3), 223–230.595
Dand, I. W. and White, W. R. (1977). “Design of navigation canals.” Symposium on Aspects of596
Navigability of Constraint Waterways including Harbour Entrances, Delft, The Netherlands.597
Eiffert, F., Niekamp, O., and Meesenburg, S. (2013). “Schiffsbedingte Erosion am Nord-Ostsee-598
Kanal [Ship-induced erosion in the Kiel Canal].” HANSA International Maritime Journal, 150(9),599
116–119 (in German).600
Flügge, G. and Uliczka, K. (1996). “Schiffsbedingte Wellen unter den spezifischen Randbedin-601
gungen von Seewasserstraßen [Ship waves under the specific boundary conditions of coastal602
waterways].” Dresdner Wasserbauliche Mitteilungen, Vol. 9, TU Dresden, 75–90. (in German).603
Führböter, A., Dette, H.-H., and Jensen, J. (1984). “Untersuchungen über schiffserzeugte Wellen604
an der Unterelbe in den Jahren 1980/81 [Investigations on ship waves on the Lower Elbe in605
1980/81].” Report No. 546, Leichtweiß-Institut für Wasserbau, TU Braunschweig. (in German).606
23 Final Draft
Gelencser, G. J. (1977). “Drawdown surge and slope protection, experimental results.” Proceedings607
of the 24th International Navigation Congress, Leningrad, Russia.608
Gelinas, M., Bokuniewicz, H., Rapaglia, J. P., and Lwiza, K. M. M. (2013). “Sediment Resuspension609
by Ship Wakes in the Venice Lagoon.” Journal of Coastal Research, 29(1), 8–17.610
Gharbi, S., Valkov, G., Hamdi, S., and Nistor, I. (2010). “Numerical and field study of ship-induced611
waves along the St. Lawrence Waterway, Canada.” Natural Hazards, 54(3), 605–621.612
Göransson, G., Larson, M., and Althage, J. (2014). “Ship-Generated Waves and Induced Turbidity613
in the Göta Älv River in Sweden.” Journal of Waterway, Port, Coastal, and Ocean Engineering,614
140(3).615
Hamill, G. A., Johnston, H. T., and Stewart, D. P. (1999). “Propeller Wash Scour near Quay Walls.”616
Journal of Waterway, Port, Coastal, and Ocean Engineering, 125(4), 170–175.617
Houser, C. (2011). “Sediment Resuspension by Vessel-Generated Waves along the Savannah River,618
Georgia.” Journal of Waterway, Port, Coastal, and Ocean Engineering, 137(5), 246–257.619
Initiative Kiel-Canal e. V. (IKC) (2013). Lebensader Nord-Ostsee-Kanal [Lifeline Kiel Canal],620
<https://initiative-kiel-canal.de/wp-content/uploads/2016/10/Journal_IKC.pdf>. (in German).621
International Maritime Organization (IMO) (2015). Resolution A.1106(29) - Revised Guidelines622
for the onboard operational use of shipborne automatic identification systems (AIS).623
Jensen, J. (1998). “Ermittlung von schiffserzeugten Belastungen an Wasserstraßen [Investigation of624
ship-generated loads on waterways].” Numerische Verfahren in der Wasserbaupraxis, J. Jensen625
and A. Braxein, eds., Siegen, Germany, fww. (in German).626
Ji, S., Ouahsine, A., Smaoui, H., and Sergent, P. (2014). “3D Modeling of sediment movement627
by ships-generated wakes in confined shipping channel.” International Journal of Sediment628
Research, 29(1), 49–58.629
Krey, H. (1913). “Fahrt der Schiffe auf beschränktem Wasser [Ship cruise in limited waters].”630
Schiffbau, 14(12–17) (in German).631
Niehüser, S., Ulm, M., Arns, A., Jensen, J., Kelln, V., Uliczka, K., and Kondziella, B. (2016).632
“Interaction between ship-induced stress and associated characteristics of turbidity records.”633
24 Final Draft
Proceedings of the 4th MASHCON 2016. International Conference on Ship Manoeuvring in634
Shallow and Confined Water with Special Focus on Ship Bottom Interaction, Vol. 4, 16–26.635
Permanent International Association of Navigation Congresses (PIANC) – Permanent Technical636
Committee I – Working Group 4 (1987). Guidelines for the Design and Construction of Flexible637
Revetments Incorporating Geotextiles for Inland Waterways.638
Pesce, M., Terzi, S., Al-Jawasreha, R. I. M., Bommarito, C., Calgaro, L., Fogarin, S., Russo, E.,639
Marcomini, A., and Linkova, I. (2018). “Selecting sustainable alternatives for cruise ships in640
Venice using multi-criteria decision analysis.” Science of the Total Environment, 642, 668–678.641
Plate, U. and Keil, G.-W. (1971). “Sediment-Transport in einem Seeschiffahrt-642
skanal [Sediment transport in a coastal waterway].” Die Küste, Vol. 21, 59–65,643
<https://hdl.handle.net/20.500.11970/101001>. (in German).644
Rapaglia, J. P., Zaggia, L., Parnell, K., and Lorenzetti, G. (2015). “Ship-wake induced sediment645
remobilization: Effects and proposed management strategies for the Venice Lagoon.” Ocean &646
Coastal Management, 110, 1–11.647
Rapaglia, J. P., Zaggia, L., Ricklefs, K., Gelinas, M., and Bokuniewicz, H. (2011). “Characteristics648
of ships’ depression waves and associated sediment resuspension in Venice Lagoon, Italy.”649
Journal of Marine Systems, 85(1–2), 45–56.650
Schoellhamer, D. H. (1996). “Anthropogenic Sediment Resuspension Mechanisms in a Shallow651
Microtidal Estuary.” Estuarine, Coastal and Shelf Science, 43(5), 533–548.652
Thormählen, C. (2010). “Modernisation of the Brunsbüttel locks.” PIANC Yearbook, 131–134.653
Uliczka, K. and Kondziella, B. (2016). “Ship-Induced Sediment Transport in Coastal Waterways654
(SeST).” 4th MASHCON – International Conference on Ship Manoeuvring in Shallow and655
Confined Water with Special Focus on Ship Bottom Interaction, K. Uliczka, C.-U. Böttner, M.656
Kastens, K. Eloot, G. Delefortrie, M. Vantorre, M. Candries, and E. Lataire, eds., Karlsruhe,657
Germany, Federal Waterways Engineering and Research Institute.658
Ulm, M., Niehüser, S., Arns, A., Jensen, J., Kondziella, B., and Uliczka, K.659
(2017). “Estimating ship-induced sediment transport in confined waters.” EGU Gen-660
25 Final Draft
eral Assembly Conference Abstracts, Vol. 19, (online), European Geosciences Union,661
<http://meetingorganizer.copernicus.org/EGU2017/EGU2017-5070.pdf>.662
Weilbeer, H. (2014). “Sediment Transport and Sediment Management in the Elbe Estuary.” Die663
Küste, Vol. 81, Karlsruhe, Germany, Federal Waterways Engineering and Research Institute,664
409–426, <https://hdl.handle.net/20.500.11970/101703>.665
Zaggia, L., Lorenzetti, G., Manfé, G., Scarpa, G. M., Molinaroli, E., Parnell, K. E., Rapaglia, J. P.,666
Gionta, M., and Soomere, T. (2017). “Fast shoreline erosion induced by ship wakes in a coastal667
lagoon: Field evidence and remote sensing analysis.” PLoS ONE, 12(10), 1–23.668
26 Final Draft
TABLE 1. Annual dredging volumes for the Kiel Canal fairway west of km 49.460, excluding
locks and lock harbors
Year 2006 2007 2008 2009–2012 2013 2014
Volume [m3]25 954 59 440 65 329 N/A* 109 462 53 375
* Unknown due to a lump-sum contract with a dredging company.
Data provided by WSA Brunsbüttel, personal communication, 2016.
28 Final Draft
List of Figures672
1 Explanation of the Primary and Secondary Ship Wave System (a) and ship induced673
water motion in channels (b). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30674
2 a) Map of the study area with the field campaign location at km 17.925. b) Cross675
section at the field measurement location with probe locations and an exemplary676
vessel geometry (grey box). Adapted from Ulm et al. (2017). . . . . . . . . . . . . 31677
3 Overview of the recorded parameters and their development over time during a678
random vessel passage. From top to bottom: a) simplified ship geometry and679
location of the probes, b) water level fluctuation, c) turbidity, d) flow velocity.680
Adapted from Niehüser et al. (2016). . . . . . . . . . . . . . . . . . . . . . . . . . 32681
4 Considered cross-section (a) and longitudinal section (b) indicating the parameters682
used for the development of the drawdown estimation. . . . . . . . . . . . . . . . . 33683
5 Base turbidity within an exemplary 24 h measurement cut-out. . . . . . . . . . . . 34684
6 Grid for turbidity extrapolation (a), examplary turbidity extrapolation performed685
on the grid for a specific time step (b), and corresponding ADCP field measurement686
atthesametimestep(c). ............................... 35687
7 Calculation of the drawdown based on existing empirical approaches using the688
data of the field measurements in the Kiel Canal: a) normal kernel functions for689
the individual approaches and b) the mean residuals (dots) and the corresponding690
standarddeviation. .................................. 36691
8 a) Observed drawdown vs. estimated drawdown using the newly developed ap-692
proach. b) shows the performance of the approach compared to the existing empir-693
ical approaches, analogous to Fig. 7. . . . . . . . . . . . . . . . . . . . . . . . . . 37694
9 Transport estimation results for the entire traffic between September 17 and Septem-695
ber 25, 2012: a) Transported sediment mass, cumulated over time; b) flow velocity696
in 𝑥-direction. Dewatering periods are highlighted as dotted lines. . . . . . . . . . 38697
29 Final Draft
Transversal Stern Wave
Secondary Waves
Front Wave
19°28'
Water Level Depression
Bank
a)
Primary
Wave Sailing
Course
h
A
As
b
zS
b)
vrs
Hs
sB
Fig. 1. Explanation of the Primary and Secondary Ship Wave System (a) and ship induced water
motion in channels (b).
30 Final Draft
NORTH
SEA
BALTIC
SEA
Hamburg
Kiel
Bruns-
büttel
km 17.925
GERMANY
DENMARK
20 km
-12 m BMWL
Probes
1
2
3
MWL ( const.)
bMWL 160 m →
a)
b)
Fig. 2. a) Map of the study area with the field campaign location at km 17.925. b) Cross section at
the field measurement location with probe locations and an exemplary vessel geometry (grey box).
Adapted from Ulm et al. (2017).
31 Final Draft
-5
0
5
10
15
Depth [m]
Probes 1
2
3
n = 8.87
As = 169.05 m² B = 23 m
T = 7.5 m
L = 148 m
a)
-0.5
0
0.5
Water level [m]
b)
0
200
400
600
Turbidity [NTU]
c)
0 30 60 90 120 150 180 210 240 270 300
Time [s]
-1
-0.5
0
0.5
Flow velocity [m/s]
x-direction
y-direction
z-direction
d)
Fig. 3. Overview of the recorded parameters and their development over time during a random
vessel passage. From top to bottom: a) simplified ship geometry and location of the probes, b)
water level fluctuation, c) turbidity, d) flow velocity. Adapted from Niehüser et al. (2016).
32 Final Draft
b
B
zS
A
2
2
1
1
vrs + vR
vrs
a)
b)
zS
AS
Tbow Tstern > Tbow (squat)
Fig. 4. Considered cross-section (a) and longitudinal section (b) indicating the parameters used for
the development of the drawdown estimation.
33 Final Draft
Turbidity [NTU]
0
100
200
300
400
500
Sept. 22
00:00
Sept. 22
12:00
Sept. 23
00:00
Base turbidity
time series
Measurement
Base turbidity
treshold
Fig. 5. Base turbidity within an exemplary 24 h measurement cut-out.
34 Final Draft
4
8
12
20 40 60 80 100 120 140 160
180 20040 60 80 100 120 140 160
Turbidity [NTU]
Canal width [m]
Canal depth [m]
20 40 60 80 100
SSC [g/m³]
c)
b)
a)
4
8
12
4
8
12
Canal depth [m]
Probe 3 Probe 2
Probe 1
Canal depth [m]
Fig. 6. Grid for turbidity extrapolation (a), examplary turbidity extrapolation performed on the grid
for a specific time step (b), and corresponding ADCP field measurement at the same time step (c).
35 Final Draft
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Drawdown zS [m]
0
0.01
0.02
0.03
0.04
0.05
0.06
Relative frequency [-]
a)
Observed drawdown
Krey (1913)
Constantine (1960)
Bouwmeester et al. (1977)
Gelencser (1977)
Dand and White (1978)
Führböter (1984)
PIANC (1987)
Schoellhamer (1996, adjusted)
-0.2
0
0.2
0.4
Residuals [m]
b)
Fig. 7. Calculation of the drawdown based on existing empirical approaches using the data of the
field measurements in the Kiel Canal: a) normal kernel functions for the individual approaches and
b) the mean residuals (dots) and the corresponding standard deviation.
36 Final Draft
0.2 0.4 0.6 0.8 1.0
1.0
0.8
0.6
0.4
0.2
0
-0.1
0
0.1
0.2
0.3
0.4
0.5
Residuals [m] Estimated drawdown [m]
Observed drawdown [m]
a)
b)
Fig. 8. a) Observed drawdown vs. estimated drawdown using the newly developed approach. b)
shows the performance of the approach compared to the existing empirical approaches, analogous
to Fig. 7.
37 Final Draft
Measurement
60 min. moving average
Flow velocity vx [m/s]
Cumulated sediment transport
[metric tons]
600
500
400
300
200
100
0
0.4
0.2
0
-0.4
-0.2
Sept. 19 Sept. 21 Sept. 23 Sept. 25
a)
b)
Fig. 9. Transport estimation results for the entire traffic between September 17 and September
25, 2012: a) Transported sediment mass, cumulated over time; b) flow velocity in 𝑥-direction.
Dewatering periods are highlighted as dotted lines.
38 Final Draft
