Turbidity plumes due to trenching : practical points

of attention for numerical modelers

1st Vincenzo Marco Gatto 2nd Ludovica Sartini 3rd Eleonora Duchini
RINA Consulting S.p.a. RINA Consulting S.p.a. RINA Consulting S.p.a.
Roma, Italy Genova, Italy Genova, Italy
[Link]@[Link] [Link]@[Link] [Link]@[Link]

Abstract—The expansion of offshore energy infrastructure trade-off between complexity/level of detail and operational
requires engineering interventions that can affect marine ecosys- feasibility (available field and project data, computational
tems. Dredging and trenching of the seabed can release suspended power, time, etc.), especially during the early stage of a
sediment into the water column, impacting the biota through
increased turbidity and re-deposition. project.
We explore key modeling choices for early-stage feasibility The results will provide valuable insight into the strengths
assessments of such interventions, balancing operational viability and limitations of existing techniques, contributing to the
and conceptual accuracy. Using MIKE software, we simulate development of more accurate and robust environmental as-
sediment transport associated with submarine cable trenching in sessments.
the Tyrrhenian Sea. We propose different model configurations
to account for the deposition and resuspension cycles, the II. M ETHODOLOGY
associated bed-shear stress thresholds, and the background levels
of turbidity. A. Case study and approach
The analysis indicates that schematized configurations, al- Submarine power cables must be installed in the seabed
though necessary for viability, tend to overestimate sediment
deposition while underestimating turbidity. However, increasing of the central Tyrrhenian Sea, off the Italian coast (Fig. 1).
the complexity of the model makes it difficult to discern the Nine parallel trenches (cross section 0.75 m2 each, length ≈13
investigated effects from the background dynamics. km) will be excavated by a ROV (plough- or jet trencher).
Hence, a relatively large degree of schematization is advisable The composition of the seabed features a stepwise transition
despite its limitations. While we encourage the enhancement of between almost uniform sand (nearshore) and mud (offshore).
predictive reliability for environmental impact assessments, we
also stress the importance of expert judgement by showing that
The generation of turbidity plumes can be assumed to affect
simple scaling considerations can correctly identify the magnitude only the muddy region. However, mud can be transported to
of the interesting variables. the sandy nearshore region, where several conservation areas
Index Terms—turbidity, plume, dredging, trenching, modeling, are present.
MIKE The possible impacts are quantified by the persistence of
Suspended Sediment Concentrations (SSC henceforward), and
I. I NTRODUCTION
by the thickness of the mud layer that deposits on the seabed
The installation of submarine cables in marine protected ar- (MLT). The national laws do not specify threshold magnitudes
eas presents significant environmental challenges, particularly for such variables, below which impacts are deemed negligible
concerning the dispersion of sediments. During excavation or acceptable. This quantification is far from trivial, as it
and laying operations, seabed sediments are disturbed, leading depends on multiple, case-by-case environmental factors that
to increased Suspended Sediment Concentrations (SSC) that are not within the scopes at hand.
can impact local ecosystems (1; 2; 3). The extent and per- A conservative approach is a general guideline for such sce-
sistence of these effects depend on several factors, including narios. The rationale is to show that relatively small thresholds
hydrodynamic conditions, sediment properties, and installation are not - or barely - exceeded despite a general overestimation
techniques. of the corresponding variables. Following common practice,
Numerical simulations play a crucial role for the assess- we consider the water as ’clear’ if SSC<2 mg/l (6). We
ment of these impacts, as various scenarios can be forecast. assume an exceedance of this threshold lasting for at least 24
However, assumptions and parameterizations made by the user consecutive hours, and a MLT=2 cm as reference threshold
strongly influence the reliability of models. Hence, tailoring an conditions. For each grid cell, the SSC are aggregated in time
appropriate modeling strategy is a key aspect of environmental and space by referring to the 98th percentiles of the time series,
impact assessments (4; 5). and to the near-bed value of the vertical profile, respectively.
This study aims to evaluate how the modeling strategy
affects the predicted extent and duration of turbidity plumes B. Hydrodynamic model
caused by submarine cable installations. By comparing dif- We adopt a software package that is recognized as one
ferent model schematizations, we seek to determine the best of the market standards, MIKE (8). A flexible mesh with
 production of 80 kg/s, of which 25% (20 kg/s) is spilled into
the near-bed layer of the water column. After a pause of 8 h
at the end of the track, the operation restarts from the offshore
limit. The operation is repeated 34 times until the end of the
simulation.
The time required to complete one trench is short with
respect to the timescale of forcing variability. Hence, this
schematization ensures that a satisfactory variability of dy-
namic responses is captured with a single simulation. The
timescale of the entire simulation is short with respect to
the timescale of morphological changes, which are therefore
neglected.
The actual grain-size distribution of the study site allows
to conveniently simplify the parameterization of the seabed
by considering a single mud fraction, with a settling velocity
ws = 0.5 mm/s when the SSC is ≤ 10 mg/l (reference value
for the onset of flocculation). All other model settings are
discussed by analyzing different simulations in the remainder
of the paper.
Fig. 1. Zoom onto the study site (UTM32N coordinates). The area of the
trenches is filled in white. The dotted black lines mark the conservation areas.
III. A NALYSIS AND D ISCUSSION
(inlay) Average current speed ([V c] = m/s) rose (direction is ”going to”) at
the transition between muddy and sandy seabed (-25 m+MSL)
A. Reference simulation
The standard schematization isolates the processes due
to trenching by neglecting the ’natural’ erosion due to the
variable resolution (≈10 m near trench and coastline, ≈ 2 hydrodynamics. A characteristic dry density for moderately
km at the offshore boundary) is used. The simulation period consolidated deposits (ρd = 300 kg/m3 ), and an arbitrary
is April 2021 to June 2021. This period well represents the critical bed-shear stress for erosion that can never be exceeded
(inter)annual variability of the forcing. Input and calibra- (e.g., τe = 100 Pa) are assigned to the whole computational
tion data are obtained from local measurements, EMODNet grid. In this way, only the sediment spilled by the trencher can
(bathymetry), TPXO (tide), Copernicus (waves, currents, and be mobilized from the seabed.
wind) (13). Fig. 2A shows that the area where the SSC can exceed the
A decoupled workflow is adopted. Firstly, waves are sim- reference threshold of 2 mg/l extends for ≈ 10 km along the
ulated with the MIKE21 SW module (depth-averaged). The coastline. The shape of the colored area resembles the shape of
output is fed into the MIKE3 FM module (three-dimensional, the current rose in Fig. 1. Given the relatively low production
4 vertical layers), which simulates currents and water levels. of the trencher, near-bed SSC above 10-30 mg/l are hardly
The resulting hydrodynamics are dominated by longshore reached, except for a small area close to the track. Depth-
currents of average speed ≤ 0.5 m/s from SE to NW, and averaged SSC are about a factor 2 smaller. SSC percentiles
wave heights ≤ 4 m (average ≈1 m) mostly coming from are often used as the main indicator, possibly because they
the sector 180-270°N. Consequently, we expect that the mud are intuitive to visualize. However, they miss key information,
spilled from the trench is transported to and along the coast. i.e. the persistence of the SSC above the threshold.
This case study embodies a typical example of the dynamics Fig. 3 (solid black line) shows that persistencies are gener-
for a sandy open coast with muddy offshore area, fairly regular ally quite shorter than 24 h. Non-negligible SSC are essentially
bathymetry, and mild average forcing. correlated to the passage of the trencher, especially if forcing
conditions (for which the time series of bed-shear stress is
C. Mud transport model shown as a proxy) are stronger than average. The combination
The MIKE MT module uses the output of the hydrodynamic of these two events is sporadic and quasi-instantaneous with
model as input to compute erosion, transport, and settling respect to the time scale of the simulation. In other words, the
(including concentration-dependent flocculation) of mud. The model exhibits an intermittent behavior.
study is carried out in a project phase during which neither The same applies to the deposition process. Default settings
calibration data, nor the technical specifications of the trencher consider that mud can settle only when the bed-shear stress
are available. This is a further motivation to proceed with a drops below a critical value for deposition (τd = 0.07 Pa),
conservative approach. which is almost always exceeded (Fig. 4i-ii-iii). Hence, the
We assume that the trencher always operates along the MLT inside the conservation areas hardly reaches 2 mm at
centerline of the track, proceeding uninterruptedly for 61 h the end of the simulation (Fig. 4A). Note that the locations
at 200 m/h from offshore to nearshore, and performing a experiencing maximum SSC and maximum deposition do not
Fig. 2. 98th percentiles (colorbar in logarithmic scale) of near-bed SSC [mg/l]
for different simulation settings (check subsections corresponding to panel
letters). The star pinpoints the location where the largest values (> 400 mg/l)
are simulated. The solid black line traces the trajectory of the trencher.

coincide. The observation point is kept at the former location

for consistency.
Fig. 3. (left vertical axis, in logarithmic scale) Time series of near-bed SSC
A limitation of this schematization is that mud deposited [mg/l] for different simulation settings, at the location pinpointed by the star
inside the conservation areas cannot be resuspended ever in Fig. 2. (right axis) Time series of bed-shear stress at the same location. The
again. In the case of fresh deposits, ρd and τe are up to one time discretization on the horizontal axes refers to the passage of the trencher
nearby, which is every ≈3 days.
order of magnitude smaller than in the case of consolidated
deposits (10). In reality, fresh coastal deposits are likely to
undergo cycles of resuspension and (re-)deposition driven by
larger turbidity in the area. In time, suspended sediment is
the local hydrodynamics.
bound to spread over a much larger area and settle again (Fig.
Hence, this model configuration is expected to overestimate
4B).
the MLT. In contrast, the (persistencies of) SSC in the conser-
vation areas can be underestimated. In conclusion, we deem With this schematization, the 98th SSC percentile is not an
this schematization to be somewhat crude, and therefore we appropriate indicator (Fig. 2B). Very large values are observed
proceed by adding further degrees of complexity. in a wide region that stretches the whole coastline. Nearby
the offshore edge of the conservation areas, SSC decrease
B. Cycles of deposition and resuspension with a steep and almost regular cross-shore gradient over a
This schematization aims to mimic the possible removal of narrow shore-parallel belt. However, the time series (B) of
fresh mud deposits from conservation areas, while still neglect- near-bed SSC at the observation point is basically the same as
ing the background dynamics. We prescribe the variability of in simulation A during the first 7 passages of the trencher
critical bed-shear stress for erosion and sediment dry density (Fig. 3i). Then, a period of relatively strong forcing starts
(τe = 0.1 Pa and ρd = 100 kg/m3 inside the conservation resuspending and (re-)depositing fresh mud in the conservation
areas, τe = 100 Pa and ρd = 300 kg/m3 elsewhere). An initial areas (Fig. 3ii).
MLT of 0 m is assigned to the bed of all grid cells. In this In time, this process tends to homogenize the turbidity in the
way, the bed sediment can be mobilized by the hydrodynamics area. Simulation B diverges more often from simulation A (Fig.
only in the conservation areas, and only after a deposit of mud 3iii). After ≈2 months since the beginning of the simulation,
originating from a trench has formed. the two time series start to fully diverge (Fig. 3iv, the last ≈20
Such a schematization represents erosion and deposition days are omitted for the sake of conciseness). Quite larger SSC
processes more realistically. While the MLT is bound to mono- can now persist in spite of prolonged mild forcing. During the
tonically increase in the reference simulation, local longshore last shown week, the critical bed-shear stress for erosion (τe =
currents can resuspend the fresh mud deposits in this simula- 0.1 Pa) is never exceeded, indicating that the local suspension
tion (Fig. 4iii). However, resuspension leads to a progressively is due to transport from neighboring locations.
 of SSC at the observation point are quite similar (Fig. 3i).
Then, curve C diverges to consistently larger magnitudes.
In time, curve B gradually shifts from overlapping curve A
to overlapping curve B. In other words, after ≈ 2 months
of simulation, schematization B leads to enough sediment
circulating in the conservation area to match the nearly-
constant background SSC of schematization C (Fig. 3iv).
D. Critical bed-shear stress for deposition
The presented schematizations result in a on-and-off be-
havior of the model. That is realistic with regard to the
generation of turbidity plumes, but disputable with regard to
the deposition process. The cause is sought in the conceptual
model behind MIKE MT.
The model assumes that suspended sediment can start set-
tling only if the bed-shear stress is smaller than a user-defined
threshold τd . Since τd ≤ τe , erosion and deposition cannot
happen at the same time. The validity of this assumption
has been questioned from both experimental and theoretical
viewpoints (7; 12). However, it is still used as τd can be
interpreted as a calibration (or probabilistic) factor.
We attempt to simulate a more progressive MLT formation
in the conservation area by comparing the reference simulation
(A, τd = 0.07 Pa) with two variants that violate the assumption
τd ≤ τe = 0.1 Pa. Enhanced (D1) and continuous (D2)
deposition are prescribed, respectively (Fig. 5).

Fig. 4. (A-B-C) MLT at the end of the simulation [mm]. (i-ii-iii) Time series
of MLT [mm] (left vertical axis, in logarithmic scale) and bed-shear stress
(right axis) for different simulation settings, at the location pinpointed by the
star.

C. Background concentration
We can argue that neglecting the possible presence of
a background SSC field due to natural processes (as per Fig. 5. Time series of MLT [mm] (left vertical axis, in logarithmic scale) and
schematizations A and B) is inconsistent with the rationale bed-shear stress (right axis) for different simulation settings (A: τd = 0.07
of the impact evaluation. A relatively minimal contribution of Pa, D1: τd = 0.5 Pa, D2: τd = 100 Pa), at the location pinpointed by the
star in Fig. 4.
trenching may add just enough to the background turbidity to
exceed the impact threshold. Schematization D1 lead to a final MLT similar to A,
Hence, we test an idealized schematization that prescribes although the bulk of the deposition flux is anticipated before
an unlimited availability of consolidated mud in the seabed the end of the stormy period (24-01). Henceforward and
below -25 m+MSL (initial MLT= 1000 m, τe = 0.5 Pa and until the last days of simulation, mud plumes do not reach
ρd = 300 kg/m3 ), and an ’empty’ bed of fresh mud elsewhere the observation point, as bed-shear stresses smaller than the
(initial MLT= 0 m, τe = 0.1 Pa and ρd = 100 kg3 ). deposition threshold cause all the mud to immediately deposit
The background dynamics become dominant, and the effects in the vicinity of the spilling point. Schematization D2 leads
of trenching cannot be discriminated (Fig. 2C and 4C). The to the latter scenario at all times, and no mud is deposited
average magnitude of SSC caused by trenching is possibly in outside of the trenching area.
the same order of the model precision. A sufficiently realistic In conclusion, these schematizations tend to change the
simulation of the background sediment dynamics (including dynamics without solving the issue. Hence, τd should be tuned
e.g., sand transport, graded grain-size distribution, etc.) raises by testing combinations with other settings, which are briefly
the level of complexity beyond the operational conditions set discussed in the following sub-section.
by the objective of the paper. However, this schematization
provides a useful term of comparison for schematizations A E. Other model settings
and B. A number of other settings are left to default, simplified, or
During the first ≈ 10 days of simulation, the 3 time series excluded from the analysis as they raise the level of complexity
beyond the objective of the paper. A notable example is with H = 15 m water depth, z = distance from the bottom
the parameterization of horizontal and vertical dispersion, (z > a), a = 0.01 m reference height, k = 0.4 Von Karman
which is based on the turbulence model. Results suggest that constant, τe = 0.1 Pa, ws = 0.001 m/s (mud flocs), and β =
default settings overhomogenize the turbidity, which helps calibration coefficient.
with numerical stability. However, calibration requires data Fig. 6i shows that the Rouse profile with the typical β =
that are typically unavailable. 1 matches the simulated 98th percentile of SSC near the bed.
Other examples include, e.g., spatial variability and stratifi- However, SSC are progressively more overestimated upwards
cation of different sediment fractions in the bed, bed rough- (exceeding the maximum simulated values), as the Rouse
ness, sand transport, rheological interactions between sand profile neglects the horizontal sediment fluxes. The effect of
and mud (11). Furthermore, there are aspects that deal with such fluxes can be mimicked by setting a much larger vertical
operational feasibility (time, computational power, etc.), such decay rate of SSC with height (β = 8).
as mesh resolution and number of vertical layers. Choices may We estimate the persistence of SSC by defining a time scale
often be not optimal, yet necessary. for settling (Ts ) as
The parameterization of the engineering intervention plays H
Ts = (3)
a role as well. In principle, we have adapted the conceptual ws
model of MIKE MT, which is based on a floating dredger, to In agreement with model results, Fig. 6ii shows that the
the case of a trencher rolling on the sea floor. This affects persistence is shorter than 12 h for the characteristic settling
assumptions such as the depth at which sediment is spilled velocity of mud particles we consider (ws = 0.0005 m/s), and
into the water column. shorter than 6 h for mud flocs (ws ≈ 0.001 m/s).
In order to encompass a sufficient variability of external
forcing within a single simulation, we assume that the excava-
tion is carried out as a single and continuous operation, which
is different from reality. Moreover, 34 trenches are excavated,
while only 9 are necessary. In time, the combination with
overestimated dispersion results in increasing turbidity, and
plumes reach arguably too far from their origin.
This may be prevented by subdividing the simulation into
a number of overlapping time windows. Each window covers
extra time after the trenches are completed, so that plumes can
fully vanish. In this case, the effect is only partially achieved
by allowing a pause of 8 h at the end of each trench. However,
we deem the schematization is reasonable given the relatively
small production of the trencher, which spends most of the
time in the deep offshore region, from where plumes cannot
reach the conservation areas.

F. Scaling considerations
The presented analysis highlights that sensitive schemati- Fig. 6. (i) Comparison between simulated and theoretical SSC profiles at
zations are required to cope with operational limitations. We at the location pinpointed by the star in Fig. 4. (ii) Time scale of settling.
propose relatively simple scaling considerations to support the Horizontal axes in logarithmic scale.
interpretation of model results, especially in case validation
data are unavailable.
IV. S UMMARY AND CONCLUSIONS
We estimate the near-bed SSC by referring to the water-
sediment mixture in a dredging pipeline. We assume a char- This study shows how modeling choices significantly in-
acteristic volumetric concentration Cv = 0.25 (25%). The fluence the environmental impact assessment of submarine
corresponding near-bed SSC in weight (Ca ) reads cable installation in protected areas. The analysis of different
simulation schemes highlights that the dynamics of suspended
ρs Cv ρs Cv sediment transport and deposition strongly depend on the
Ca = = ≈ 460 mg/l (1)
ρm ρw (1 − Cv ) + ρs Cv assumptions made regarding physical and operational param-
with ρs = 2650 kg/m3 sediment quartz density, ρm = mixture eters.
density, and ρw = 1025 kg/m3 water density. The reference scheme isolates the effects of plumes gen-
At first order, we assume that the vertical SSC distribution erated by trenching operations by completely neglecting the
(Cz ) follows the Rouse profile (9), i.e. natural sediment dynamics in the background. Results indicate
that the area affected by turbidity extends approximately 10 km
 βws √ ρτw
along the coast, with generally low sediment concentrations,

H/z − 1 k e
Cz = Ca (2) persistence times of less than 24 hours, and up to about 2 cm
H/a − 1
thickness of redeposited mud. However, the absence of a rep- [7] Gatto V.M., van Prooijen, B.C. and Wang, Z.B. (2017)
resentation for resuspension processes leads to underestimated Net sediment transport in tidal basins: quantifying the
turbidity and overestimated deposition. tidal barotropic mechanisms in a unified framework.
The introduction of deposition and resuspension cycles Ocean Dynamics 67, 1385-1406
improves the realism, simulating that sediment can be lifted [8] I.R. Warren, H.K. Bach, MIKE 21: a modelling system
up and redistributed multiple times, That results in smaller for estuaries, coastal waters and seas, Environmental
deposition, but increased turbidity over a wider area and for Software, Volume 7, Issue 4, 1992, Pages 229-240.
a longer period. The inclusion of a background concentration [9] Rouse, H. (1938) Experiments on the mechanics of
reveals that the effects of excavation are difficult to distinguish sediment suspen-sion. Proc., 5th Int. Congress of Applied
from natural dynamics, making the impact assessment even Mech., 550–554.
more complex. [10] Van Rijn, L.C. (1993) Principles of Sediment Transport
The parameterization of the critical shear stress for deposi- in Rivers, Estuaries and Coastal Areas. Aqua Publica-
tion significantly influences the formation of the mud layer on tions, Amsterdam.
the seabed, but its calibration is challenging without specific [11] Van Rijn, L.C. (2020). Erodibility of Mud–Sand
observational data. Other factors, such as turbulent dispersion Bed Mixtures. Journal of Hydraulic Engineering. 146.
and sediment stratification, could further refine the model, but 04019050. 10.1061.
would require a higher level of complexity and resources. [12] Winterwerp, J.C., van Kesteren, W.G.M., van Prooijen,
In conclusion, the reference scheme is advisable despite its B.C. and Jacobs W. (2012) A conceptual framework for
limitations. This study confirms the importance of a conser- shear flow-induced erosion of soft cohesive sediment
vative approach in the preliminary phases of environmental beds. J Geophys Res 117:C10020.
impact assessment, favoring configurations that tend to over- [13] [Link] [Link]
estimate rather than underestimate the effects. [Link]
The analysis suggests that balancing simplification and
realism is a challenge for reliable evaluations. Especially in
contexts with operational constraints and limited data avail-
ability, interpretation and expert judgment is essential. Even
simple scaling considerations can be effective as a first-order
validation of model results.
R EFERENCES
[1] Paul L.A. Erftemeijer, Bernhard Riegl, Bert W. Hoek-
sema, Peter A. Todd, Environmental impacts of dredging
and other sediment disturbances on corals: A review, Ma-
rine Pollution Bulletin, Volume 64, Issue 9, 2012,Pages
1737-1765.
[2] P.L.A. Erftemeijer, R.R.I.I.I. Lewis. Environmental im-
pacts of dredging on seagrasses: a review. Marine Pollu-
tion Bulletin, 52 (2006), pp. 1553-1572.
[3] M.D. Fortes. The effects of siltation on tropical coastal
ecosystems. Oceanographic Processes of Coral Reefs:
Physical and Biological Links in the Great Barrier Reef,
CRC Press, Boca Raton (2001), pp. 93-112.
[4] Amoudry, L. O., and Souza, A. J. (2011) Deterministic
coastal morphological and sediment transport modeling:
a review and discussion. Review of Geophysics, 49,
RG2002, doi:10.1029/ 2010RG000341.
[5] Chavarrias, V., Schielen, R.M.J., Ottevanger, W. and
Blom, A. (2019). Ill posedness in modelling two-
dimensional morphodynamic problems: Effects of bed
slope and secondary flow. Journal of Fluid Mechanics.
868. 461-500. 10.1017/jfm.2019.166.
[6] De Padova D, Attolico A, Ungaro N, Mossa M. Environ-
mental Management of Dredging Activities: Monitoring
System and an Empirical Approach to Estimating Val-
ues of Background and Limit Turbidity. Environmental
Sciences Proceedings. 2022; 21(1):12.