DSD-INT 2017 Long-term morphodynamics of muddy backbarrier basins - Canestrelli

Long-term morphodynamics of
muddy backbarrier basins:
the relative importance of lateral
bank erosion and vertical platform drowning.
Alberto Canestrelli
University of Florida
Giulio Mariotti
Louisiana State University

Back-barrier systems are sheltered basins located
behind barriers which are parallel to the coast.
Salt marshes form mud is allowed to deposit.
Virginia Coast Reserve
3 – Long term evolution of back-barrier ecosystems: fill in or empty out?
Approximately half of US salt marshes are
located along the Gulf Coast.
Salt marshes are the most productive ecosystem known.
Filters out toxins from incoming water.
Accumulate organic material with time, forming a dense
layer called peat.
Important sinks for atmospheric CO2
Act as effective storm buffer and effective sediment trap
- withstanding large wave forces with no significant vertical erosion
Moller et al. (2014)

Allen and Haslett. (2014)
Marsh progradationMarsh retreat
Can marshes withstand high rates of SLR and be in morphodynamic equilibrium?1
2 Are marshes eroding by drowning or horizontal retreat?
Marsh erosion versus progradation
Organic and inorg. deposition versus SLR
Mariotti and Carr, 2014

ISSUES OF CLASSICAL/SIMPLE BANK EROSION APPROACH
(SUCH AS THE ONE EMPLOYED IN DELFT3D)
BANK EROSION VERTICAL BANK DISPLACEMENT=
Very low
velocity/wave
energy
No further
erosion
Canestrelli et. al.,2016
High velocity/wave
energy close to the
bank.
=> Bank retreat
Analytical solution
Simplified approach
𝑝 Δ𝑡𝑖 =
𝛹 𝑘
𝐴
𝑙 𝑘
𝑘
𝐴: area
Marsh
𝛹 𝑘
𝑙 𝑘
𝐴: area

Suspended Sediment Concentration of mud co [mg/l]
- Rate of Relative sea level rise (R)
-Wind distribution:
Wind speed [m/s]
-Landward transgression not possible due to the
presence of human settlements and infrastructure
along the coastline (Doody, Coastal squeeze, 2004).
- The barrier island is fixed in time seawalls:
SWAN
MODEL
Tide: 2 m mean tidal range, spring-neap

Marsh edge erosion
E = α P
Wave powerRetreat rate
Validated with field data (Leonardi et al., 2016)
Dorg = Korg
B
Bmax
Marsh vegetation processes:
• Biomass is a parabolic function of inundation depth
• Organic sediment production prop. to biomass:
• Increase in drag
• Increase in critical shear stress
Morris et al, 2013
MSL MHW

co = 0 mg/l
R = 1 mm/yr
SAND vs MUD
Bassin d’Arcagnon (France)
NO WIND
WIND

Formation of marshes (R = 1 mm/yr)
co = 0 mg/l co = 50 mg/l
2 km
At equilibrium:
Marsh extent is
~70% of the basin
Virginia Georgia
TIDE
+
WAVE
Mud: Mud:

R=10 mm/y
Which concentration do I need to
withstand a R=10 mm/y
What happens if I cut
out sediment input?
C0=0R=1mm/y

Bad news:
Marsh retreat increases with R
Because sediment needs to fill accommodation space,
less available for marsh progradation
Because marsh edge perimeter decreases
Basin-wide marsh edge erosion slows down
as the marsh erodes away
Good news:

Marsh does not drown
for R up to 10 mm/yr
even without external
sediment supply!
Good news:
Because of internal
sediment recycle
With SLR <5 mm/yr, such as those
currently experienced in the US
Atlantic Coast, marsh retreat is a
more likely modality of marsh loss
than marsh drowning!!

Bank erosion PREVENTS DRAWNING!

Open issues:
How is this picture changing with river inflow?
How does the picture change when the boundaries are not fixed:
-barrier can be overtopped and transgresses over the
salt marshes
Kirwan, 2016
-marshes can migrate inland
Raabe and stumf, 2016
Overwash resolving wave model

Take-home message
• Edge erosion “empties” backbarrier basins
• Retreat slowed down and reversed by external sediment
supply, which promotes marsh progradation
• Internal sediment recycle allow marsh to accrete vertically
even in the absence of external sediment supply
= • Sea level rise increases the rate of marsh retreat
Low to moderate SLR(<5 mm/yr), such as those currently experienced in the US Atlantic Coast
(Engelhart et al., 2009), marsh retreat is a more likely modality of marsh loss than marsh drowning!!
• If no/low external sediment, basin can empty also with no SLR
• “Filled basin” dynamic equilibrium with large channels

SUPPLEMENTARY MATERIAL
MARSHES

• Low gradient of shelf
• Abundant sediment supply
• Low tidal energy with respect to wave energy
How do barrier islands form?
How does the picture change when salt marshes form in open
ocean, like at Florida Big Bend?
Improving ecological coupling
Modeling multi-species and zonation
(spartinas ,needlerush, mangroves)
Storing vertical stratigraphy Peat formation
(Petroleum
Research Fund)

This effect could explain why marsh edge erosion in the
Mississippi Delta (LA), a region with very high rates
of sea level rise (and subsidence), can be as fast as 10
m/yr [Penland et al., 2002], a rate that is several times
faster than in other eroding marshes within the U.S.
[Mariotti and Fagherazzi, 2013b; McLoughlin et al.,
2014]

DSD-INT 2017 Long-term morphodynamics of muddy backbarrier basins - Canestrelli

Bank erosion prevents drowning!!!
Marsh edge erosion, mudflat deepening, and channel deepening release sediments (Figure 10)
[Day et al., 1998]. Furthermore, marsh edge erosion increases the mudflat area, which could
be then eroded by surface erosion from bed shear stress. The model predicts that wave
erosion would increase the suspended sediment concentration in the mudflat to 100–200
mg/L, which agrees with field measurements in the backbarrier basins of the Virginia Coast
Reserve (VA, USA) [Lawson et al., 2007].

Even when the marsh enters the runaway erosion and eventually disappears, the marsh platform keeps
pace with sea level rise up to 10 mm/yr, i.e., marsh loss by drowning does not take place. Of particular
notice is that the marsh platform keeps up even in the absence of an external sediment supply (co50). In
this case, because the maximum accretion rate by organic sediment was set equal to 5 mm/yr, inorganic
sediment contributed to 5 mm/yr of accretion or more. If only organic sediment were contributing to vertical
accretion, a marsh starting at MHW (here set 1 m above msl) would have drown in 157 years considering
a parabolic biomass-elevation relationship (see section 2.2) (Figure 9g). Where did the additional sediment
come from if the external sediment concentration was zero? Clearly these sediment resulted from erosion
of the basin, i.e., by internal recycling

Equilibrium:
ebb dominated currents
that establish in channels with extensive intertidal areas
No tidal flats => no marsh progradation

remarkably similar, especially considering that the wave
forcing is stochastic
MF=400

However, this formulation fails to reproduce the expected marsh edge
retreat for a given wave power (equation (5)). For example,
such a formulation predicts a fast (slow) edge retreat when the mudflat
in front of the marsh is shallow
deep), given that bed shear stress of locally generated
waves decreases with water depth [Fagherazzi et al., 2006].
In reality, a faster marsh retreat is expected where the marsh
edge faces a deeper mudflat, which would allow a larger
wave power to reach the marsh edge.

The ability of marshes to reduce sediment erodability and increase drag is simulated in a simplified way: if a
cell is unvegetated the critical shear stress for erosion is set equal to scr,unveg and the Chezy coefficient for
bed friction is set equal to Cunveg, if a cell is vegetated the critical shear stress is set equal to scr,veg and the
Chezy coefficient is set equal to Cveg. Here we set the vegetated critical shear stress equal to 1 Pa, noticing
that values above 0.8 Pa are already sufficient to make the soil virtually nonerodible [Marani et al., 2010].
This choice is consistent with the observation that the marsh platform is able to withstand large wave forces
[M€oller, 2006].
s range from 2.5 mm/yr [D’Alpaos et al., 2007] to 10
mm/yr [Blum and Christian,
2004
The wave power
causing marsh retreat is calculated on the unvegetated
cell facing a marsh boundary cell as
c is the specific weight of the water, Hs is the significant wave height, and cg is the wave group velocity,
which are obtained from the SWAN model.

marsh edge erosion is complex and poorly understood and includes a variety of masswasting
geotechnical processes such as undercutting and cantilever failure, toppling failure, and stress
failure, as well as hydrodynamic processes that occur at small spatial scales (0.1–1 m), such as wave
breaking and dissipation over the edge [Schwimmer, 2001; McLoughlin et al., 2014; Priestas et al., 2015;
Bendoni et al., 2016]. For simplicity, the effect of these processes are lumped into an empirical coefficient
aw, which is highly variable, site specific, and not predictable a priori. Rather than selecting a value that
fits a specific field site, we choose a value that gives marsh retreat rates on the order of a meter per year
(see section 3.4), in range with field measurements
Because of the large gird size (100 m) and the relatively small time step (2 min), the probability p is at most
1024
For the U.S. Atlantic Coast, organic matter contributes to about 50% to
vertical accretion [Morris et al., 2016], therefore, we used a rounded value and set fox to 50% in our model

Average marsh retreat rate during the first 100 years for
the three rates of sea level rise. Given that the areal
retreat rate is about
0.1 km2/yr for R51 mm/yr, 0.2 km2/yr for R55 mm/yr, and
0.4 km2/yr for R510 mm/yr (Figure 9a), and
given that the total marsh edge perimeter during the same
time is about 300 km (supporting information
Figure S4),

Thriving along protected shorelines, they are a common
habitat in the U.S., where salt marshes can be found on
every coast. Approximately half of the nation's salt
marshes are located along the Gulf Coast.
Salt marshes are coastal wetlands that are flooded and
drained by salt water brought in by the tides. They are
marshy because the soil may be composed of deep mud
and peat.
Back-barrier systems are coastal basins who are
protected from storm by mean of barriers which are
parallel to the coast.
Salt marshes form in these relatively quiet environments,
in which mud is allowed to deposit.
Virginia Coast Reserve

IMPORTANCE OF SALT MARSHES
“The salt marsh is the most productive ecosystem
known. Grasses, algae, and phyto-plankton can
produce up to 10 tons of organic matter per acre per
year. …Because more organic matter is produced
than is used by salt marsh inhabitants, this community
is continually exporting nutrients … enriching the
surrounding estuary. The importance of the salt marsh
to marine ecosystems cannot be overemphasized.”
Janet McMahon in: “Forests, Fields, and Estuaries:
A Guide to the Natural Communities of Josephine
Newman Sanctuary”

IMPORTANCE OF SALT MARSHES
Act as effective storm buffer
- withstand large wave forces with no significant vertical erosion
- a reduction in wave height of 20% after crossing 40 metres.
- Effective sediment trap
In places where marshes have been destroyed, winter storms
are more damaging.
Erosive nature of tides is also dampened by wetland plants
Tides carry in nutrients that stimulate plant growth in the marsh and carry out
organic material that feeds fish and other coastal organisms
Ability to filter out and break down toxins and sediments from
incoming water.
Salt marshes are important sinks for atmospheric CO2
Over time, salt marshes accumulate organic material,
forming into a dense layer called peat.
Moller et al. (2014)

Distribution of Salt Marshes
3 – Salt marshes SUPPLEMENTARY MATERIAL

Values of Korg used in previous models range from 2.5 mm/yr [D’Alpaos et al.,
2007] to 10 mm/yr [Blum and Christian, 2004], here we choose an
intermediate value of 5 mm/yr.
3 – Marshes SUPPLEMENTARY MATERIAL
if a cell is unvegetated the critical shear stress for erosion is set equal to τcr,unveg and the Chezy coefficient for
bed friction is set equal to Cunveg, if a cell is vegetated the critical shear stress is set equal to τcr,veg and the
Chezy coefficient is set equal to Cveg. Here we set the vegetated critical shear stress equal to 1 Pa, noticing that
values above 0.8 Pa are already sufficient to make the soil virtually nonerodible [Marani et al., 2010]. Such
choice is consistent with the observation that the marsh platform is able to withstand large wave forces
[Möller, 2006
𝑊 = 1 16 𝛾𝑐 𝑔 𝐻𝑠
2
" for conditions typical of the USA East Coast, organic matter probably contributes to about
60% to vertical accretion [Morris et al 2016], therefore we used a rounded value of 50%."
VALUES USED IN THE MODEL

COMPARISON WITH LEONARDI AND FAGHERAZZI CELLULAR MODEL
Note that a probabilistic approach has also been proposed in the context of a simple
cellular automata model [Leonardi and Fagherazzi, 2014], which differs from our
approach in many aspects.
In Leonardi and Fagherazzi's approach, cells had no dimension and no bed elevation.
Eroded material was not redistributed, therefore neglecting the dissipating effect that
slumped blocks have on wave energy.
Moreover, the erosion term was divided by the erosion rates in all the exposed edge of
all the domain. This made the method global, which we think is not suitable for a
physically based model. In fact, the erosion rate of a marsh cannot depend on the
erosion rate on a different distal marsh, potentially located in a hydraulically
disconnected region of the domain.
Our method has the advantage to be local and it can be directly linked to the
deterministic approach: the expected value of the erosion with the probabilistic method
coincides with the erosion predicted with the deterministic method.
Leonardi and Fagherazzi 2014

MODES OF MARSH EROSIONS
UndercuttingSlumping

Marsh emptying
Channel widening is caused by waves, not currents
co = 0
R = 1 mm/yr
co = 30 mg/l
R = 10 mm/yr

WAVE ENERGY IN THE BASIN
Although sediment resuspension within channels is generally dominated by currents, our results show that
current-induced bed shear stress is high only in the relatively narrow channels (Fig. 12). In the wide km-wide
channels, wave-induced bed shear stress dominates, especially on channel flanks (Fig. 12). We therefore suggest
that km-wide channels are needed to generate waves large enough to trigger wave-induced sediment
resuspension.

Synthetic time series of wind speeds
Weibull probability distribution function for the hourly wind speed U is:
where Uo is a scale parameter and b is a form parameter.
Analysis of wind time series in the Atlantic and
Gulf coastal areas of the US gave values of 1.9–2 for the
form parameter b and 5.6–7.2 m/s for the scale
parameter [Mariotti and Fagherazzi, 2012]. For an average
wind speed of 5.76 m/s, corresponding values of 2 and 6.5
m/s for the form (distribuzione di Rayleigh) and the scale
parameter were chosen.
.

Key West: The longer term trend computed by using all the data is +2.36
mm/year.
Key west SLR

Pagliaga salt marsh – Venice Lagoon
Tidal meanders
1932 20121970
Gaggian creek

4 – Storm surge modeling and forecast
Courtesy R. Scharroo – Altimetrics LLC
Katrina: Satellite Altimeter Measurements

The Delft3D model
 =0 (water level=mean sea level) at Gibraltar
Atmospherical forcings:
ECMWF pressure fields
ECMWF Wind velocity fields
Small forecast error
Large forecast error

L
The presence of a low atmospheric
pressure event on the Northern part of
the Tyrrhenian sea results in a South-
Eastern wind (Scirocco) on the Adriatic
sea
A difference in atmospheric
pressure of 1 hPA between the two
extremities of the Adriatic sea
produces a difference in the level of
marine surface of approx. 1 cm!
Main factors for storm surge in Venice: pressure differences and winds

Forecast errors for storm larger than 50 cm
(cross validation 2010-2011)
Forecast Analysis
Lead time (h)Lead time (h)
MAE(cm)MAE(cm)

DUD (Does not use derivatives)
Delft3D is non linear!!
(Ralston and Jennrich, Dud, A Derivative-Free
Algorithm for Nonlinear Least Squares,1978)
whereas the drag coefficient depends on both wind speed and wave state
8 wind directions plus 8 velocity intervals
64
parameters
Model is strongly sensitive on the shear stress at the air-water interface
I embedded in Cd the uncertainty on the
velocity field and on Cd itself
Cd is my calibration
paramenter
The goal is to find a vector of parameters xopt that minimizes:
C
Given:
a set y of n observed data points
a model prediction fi for the ith data point.
a model with a vector x of p uncertain parameters

• Application of storm surge satellite data to the coastal protection of
Northern Cuba
Luis Cordova, F. Reale, F. Dentale, R. Lamazares, M.Buccino, E. Pugliese Carratelli …
• CUGRI (Universities of Naples &Salerno), and CUJAE (Instituto Superior Politécnico José Antonio Echeverría) of
Havana
1) Evaluating different solutions for the
protection of El Malecon (Sea front of Avana)
2) Simulating hurricane wave fields and storm
surge
3) Acquiring and assessing satellite data to
provide useful information for points 1 and 3
HurricaneIvan,2005

4– Surges - SUPPLEMENTARY MATERIAL
DUD algorithm

Take home message:
Data assimilation largely improves storm surge forecast.
An appropriate calibration of the model can remove its bias.
Satellites provide a large amount of data which have to be assimilated efficiently
in order to have real time predictions.
But we need more satellite. Now the frequence at which satellites pass over the
Adriatic Sea is once per day.
4– Surges - SUPPLEMENTARY MATERIAL

DSD-INT 2017 Long-term morphodynamics of muddy backbarrier basins - Canestrelli

More Related Content

What's hot (20)

Similar to DSD-INT 2017 Long-term morphodynamics of muddy backbarrier basins - Canestrelli (20)

More from Deltares (20)

Recently uploaded (20)

DSD-INT 2017 Long-term morphodynamics of muddy backbarrier basins - Canestrelli