Nat Hazards (2013) 69:869–887

DOI 10.1007/s11069-013-0741-8

ORIGINAL PAPER

Quantification of riverbank erosion and application

in risk analysis

L. Nardi • L. Campo • M. Rinaldi

Received: 15 May 2011 / Accepted: 16 May 2013 / Published online: 26 May 2013
 Springer Science+Business Media Dordrecht 2013

Abstract The estimation of river bank erosion requires the knowledge of both local
hydrodynamic and erodibility characteristics. Models exist in literature that allow the
estimation of the river bank shear stress, the fundamental parameter in evaluating the
retreat given the discharge flow and the geometry of the river channel. In this study, two
hydrodynamic models (1-D and 2-D) were combined with three shear stress models in
order to obtain an estimation of the retreat on a study case on the river Cecina in Tuscany,
Central Italy. A calibration of the models was performed based on observations from aerial
photos of the region over a period of 10 years (1994–2004), and the results of the different
combinations of the models are discussed and compared. A framework was developed for
the risk analysis of land loss due to bank erosion based on the analyses of discharge flow
time series and an excess shear stress erosion model. An application to the study case is
provided by using the results of fluvial erosion modelling.

Keywords Riverbank retreat  Shear stress  Hydraulic erosion  Fluvial erosion risk

1 Introduction

Bank erosion is a natural geomorphologic process affecting several environmental and

socio-economic aspects (Bravard et al. 1999). It is the result of a complex interaction
between channel hydraulic conditions and physical characteristics of the banks, both of
which are highly variable in nature (Nanson and Hickin 1986; Lawler et al. 1997). Bank
erosion is one of the most controversial management issues in alluvial corridors. In fact,
although many negative impacts of bank erosion have been highlighted through extensive

L. Nardi (&)  L. Campo  M. Rinaldi

Department of Civil and Environmental Engineering, University of Florence, Via S. Marta 3,
50139 Florence, Italy
e-mail: laura.nardi@dicea.unifi.it

123
870 Nat Hazards (2013) 69:869–887

literature, a reconsideration of its effect is taking place, and nowadays there is also an
awareness of the beneficial aspects.
The adverse impacts of bank retreat include loss of land (e.g. Amiri-Tokaldany et al.
2003), damage to properties and infrastructures (Darby and Thorne Darby and Thorne
1996; Piégay et al. 1997), turbidity problems (e.g. Bull 1997; Eaton et al. 2004), and
sediment, nutrient or contaminant dynamics (e.g. Reneau et al. 2004). Downs and Simon
(2001) found that bank erosion can have negative effects on channel morphology and flood
carrying capacity further downstream by means of the supply of sediment and large woody
debris. The cost of protecting US stream banks in 1981 was about US $l billion annually
(US Army Corps of Engineers 1983), and according to more recent technical reports
(Department of the Army US Army Engineer District, Alaska 2006), the city of Kenai
incurs an average annual loss of $151,000 due to the reduced value of lands and buildings
and relocation of buildings and utilities, caused by the erosion of the banks along the river.
In contrast with previous aspects, in certain contexts, bank retreat can be regarded as a
positive phenomenon, being a key factor in the morphodynamic equilibrium of meandering
or braiding rivers (Bravard et al. 1986; Church 1992). Nevertheless, beneficial aspects
mostly involve ecological issues (Piégay et al. 1997, 2005). For instance, Florsheim et al.
(2008) found bank processes to be a desirable attribute of rivers, since they play an
important role in promoting riparian vegetation and in creating dynamic habitats crucial for
aquatic and riparian plants and animals. Furthermore, in incised rivers, bank erosion can
contribute to self-restoration by supplying coarse sediments (e.g. Bravard et al. 1999).
A detailed review of the different geomorphic tools available to define the extent of the
erodible corridor was proposed by Piégay et al. (2005). These tools include simple rules of
thumb such as the evaluation of the equilibrium meander amplitude (Malavoi et al. 1998)
or analyses of historical maps and sequential aerial photographs based on GIS technology
to overlay different time series (Downward et al. 1994; Marston et al. 1995; Wasklewicz
et al. 2004).
Besides empirical tools, another viable approach is represented by numerical modelling.
Since riverbank retreat mostly occurs by fluvial erosion and mass failure processes, a
combination of bank stability methods with hydrodynamic models is required to estimate
its rate. The knowledge of the boundary shear stress values responsible for the fluvial
erosion generally represents one of the most challenging parts of the research dealing with
riverbank retreat. In fact, the near-bank shear stresses can vary largely both in space and in
time, and direct measures during flow event are often impracticable (Rinaldi and Darby
2008). Thus, they can be prudently estimated by applying distributions of near-bank shear
stress models obtained from laboratory channels (e.g. Leutheusser 1963; Kartha and
Leutheusser 1972; Simons and Senturk 1977). An alternative is represented by the
application of computational fluid dynamics (CFD) models. Several advances have been
achieved in the field of hydrodynamic modelling, as demonstrated by the increasing dif-
fusion of a wide range of numerical codes: the CFD model SSIIM (Olsen 2002) to model
sediment transport with movable bed in a complex geometry; Delft3D, developed by Delft
Hydraulics, which investigates hydrodynamics, sediment transport, morphology and water
quality for fluvial, estuarine and coastal environments; CCHE3D by the National Center
for Computational Hydroscience and Engineering at the University of Mississippi or
TRIVAST by Cardiff University and Trim3D (Cassuli and Cattani 1994) are only some
examples. At the same time, several advances have been made during recent years in
modelling bank failures at the scale of single bank profiles to take account of the effects of
negative pore water pressure (e.g. Rinaldi and Casagli 1999; Simon et al. 2000) or to
include the presence of vegetation (Abernethy and Rutherfurd 2000; Pollen and Simon

123
Nat Hazards (2013) 69:869–887 871

2005; Pollen-Bankhead and Simon 2008). Notwithstanding, the undoubted importance of

the combined action of different processes, only relatively few attempts have been made to
couple the effects of interacting bank erosion mechanisms (e.g. Simon et al. 2003, 2006;
Dapporto and Rinaldi 2003; Darby et al. 2007). Rinaldi et al. (2008) and Luppi et al.
(2009), following the approach developed by Darby et al. (2007), fully coupled fluvial
erosion, finite element seepage and limit equilibrium stability analyses.
Although many efforts have been made with numerical simulations providing a detailed
description of the single processes, their application is often limited to the scale of the flow
event. The employment of such complex models in a wider temporal scale (multiyear) is
highly time-consuming which, in turn, makes them not suitable for risk analyses.
Moreover, the estimation of near-bank shear stress still represents a challenge, although
important improvements were achieved through the development of the analytical model
from Kean and Smith (2006a, b) and its application in recent studies by Darby et al. (2010).
Few attempts to investigate the risk due to fluvial erosion through accurate models and
quantification of retreat processes have been made. A methodology for the delimitation of
hazard zones along riverbanks was developed by Mahdi (2007) by taking into account both
flood risks and possible induced landslides employing an hydraulic model coupled with a
slope stability model. However, risk analyses require tools that may be employed on the
scale of years, instead of single flow events.
The general goal of this research is to investigate the potential of fluvial erosion
modelling on a multiyear temporal scale for risk analysis. In particular, two specific
objectives can be identified: (1) to explore modelling methods of fluvial erosion at reach
scale, with particular focus on the inclusion of near-bank shear stresses and on the use of
relatively simple models suitable for practical application to risk analysis; (2) to build a
framework for the risk analysis of land loss due to bank erosion and to provide an
application by using the results of fluvial erosion numerical modelling.

2 Case study

The case study area (Figs. 1, 2) is located in the middle–lower part of the Cecina basin
(Central Tuscany, Italy). The catchment area is 905 km2, and the river has a total length of
79 km. The middle and lower parts of the catchment are dominated by hilly slopes con-
sisting of erodible fluvio-lacustrine and marine sediments (Rinaldi et al. 2008; Luppi et al.
2009).
The main flow gauging station for the catchment (Ponte di Monterufoli; drainage area
634 km2, located 2 km downstream the study area) provided river stage and flow discharge
data.
Field data were collected at the study reach in 2007, including a detailed topographic
survey using a differential GPS, several grain size distributions of bank materials and
measures of the small scale bank roughness at 5 cm intervals on representative reaches of
two eroding banks (named M1 and V1, upstream and downstream banks, respectively).
The average density of the measurements of the topographic survey was about 1 point
every 10 m2 within the active channel in the meander area (Fig. 2); while in the high bars
and floodplain areas, the point density was coarser. Additional topographic surveys were
carried out downstream by acquiring 20 cross-sections for a total survey length of about
2 km.
The eroding banks within the meander of the study reach have heights ranging from 2.0
to 3.8 m. Sieve curves were undertaken at representative banks along the case study. Bed

123
872 Nat Hazards (2013) 69:869–887

Fig. 1 Cecina basin and location of the study reach

Fig. 2 2004 Aerial photograph showing the study reach and monitored bank and the location of the 1994
channel (aerial photograph reproduced by permission of Provincia di Pisa)

surface material was characterized by means of systematic pebble counting along grids.
Four different sedimentary layers were identified along the banks of the meander which
can be described as follows: (1) coarse gravel and cobble (D50 = 10.9 mm;
D90 = 58.4 mm; SD = 3.6); (2) a layer of in situ packed and slightly cemented gravel
(D50 = 6.7 mm; D90 = 22.2 mm; SD = 1.94); (3) well sorted sand (D50 = 0.061 mm;
D90 = 0.136 mm; SD = 0.840); and (4) a massive sandy silt (D50 = 0.038 mm;
D90 = 0.044 mm; SD = 0.550). Not all of these layers are always present at the same time

123
Nat Hazards (2013) 69:869–887 873

on the banks. For simplicity, during the following analyses, riverbanks were considered as
being composite with a non-cohesive layer at the bank toe (with thickness of 1.25 and
1.60 m for M1 and V1, respectively) and an upper cohesive layer (with thickness of 1.10
and 1.35 m for M1 and V1, respectively). Moreover, measurements of the critical shear
stress for cohesive materials (sandy silt) were made through the cohesive strength meter
(Vardy et al. 2007), resulting in 1.25 and 1.58 N/m2 for the upstream and downstream
bank, respectively.
GIS-based analyses on a multitemporal series of maps and high-resolution aerial photos
showed rates of riverbank retreat ranging from 2.7 to 3.8 m/year in the period 1994–2004.

3 Methodology

In the present research, the rate of fluvial erosion was estimated using the widely accepted
excess shear stress formula (Partheniades 1965; Arulanandan et al. 1980):
e ¼ kd ðs  sc Þa ð1Þ
where e (m s-1) is the fluvial erosion rate per unit time and unit bank area, s (Pa) is the boundary
shear stress applied by the flow, kd (m3 (Ns)-1) and sc (Pa) are erodibility parameters (erod-
ibility coefficient and critical shear stress, which are characteristic of the bank material,
respectively) and a (dimensionless) is an empirically derived exponent, generally assumed
equal to 1. According to this formula, the fluvial erosion starts when the threshold value given
by the critical shear stress is exceeded, and its rate (e) is proportional to the difference between
the shear stress exerted by the flow at the boundary and the critical shear.
Two hydraulic models and two near-bank shear stress models, described with more
details in the next sections, were selected in order to obtain the parameters required by the
fluvial erosion model (Partheniades 1965; Arulanandan et al. 1980). Given the purpose of
developing practical procedures as a tool for risk analyses, the present research was carried
out with a view to employing simple models. In fact, due to their complexity and to the
input data required for their calibration, models at the state of the art, described in Sect. 1,
appear to be better suited for research purposes, rather than being employed as tools for
assisting in river corridor management practices.
Different combinations of models were tested, as shown in Table 1. The erodibility
parameter and the critical shear stress were obtained by field measurements when possible,
otherwise via calibration. The analyses focused on the M1 and V1 cross-sections.

3.1 Hydrodynamic models

The hydrodynamic simulations were performed with two different numerical models: the
one-dimensional model HEC-RAS 4.0 and the two-dimensional model River 2D 0.95a.

Table 1 Tested combinations of

Combination Hydrodynamic Near-bank shear
hydrodynamic models with near-
model stress model
bank shear stress model
1 Hec-Ras Simons and Senturk
2 Hec-Ras Kean and Smith
3 River2D River2D
4 River2D Kean and Smith

123
874 Nat Hazards (2013) 69:869–887

The hydraulic modelling was performed based on field surveys carried out to define the bed
topography (by means of a differential GPS), the sieve curves to characterize the bed
material grain sizes and hydrological data. Both models were used in steady state condi-
tions, for different discharge values, starting from 5 up to 670 m3/s.

3.1.1 HEC-RAS

Developed by the US Army Corps of Engineers (USACE), HEC-RAS is one of the most
popular 1D models used to perform steady and unsteady flow calculations in operational
practice (US Army Corps of Engineers 2002a, b, c). Based on the topography of cross-
sections, channel roughness, flow rate and boundary conditions, HEC-RAS computes the
water surface profile from one cross-section to the next by iteratively solving the energy
equation.
In the present study, the extent of the reach modelled with HEC-RAS was about 3 km,
from the flow gauging station of Ponte di Monterufoli upstream to the meander of interest
in this study. Along this reach, 39 cross-sections were surveyed and subsequently inter-
polated every 10 m, in order to avoid convergence problems.
Regarding the boundary conditions, a normal depth, approximated with the upstream
slope of the channel (0.0018), was set at the upstream end. At the downstream end, a
critical depth condition was selected, due to the presence of a transverse structure utilized
as a riverbed protection under the Monterufoli bridge.
The Manning’s n coefficient, required by HEC-RAS to characterize the roughness, was
obtained for each cross-section basing on the D90 of the sieve curve retrieved through the
field surveys.

3.1.2 River2D

River2D is a two-dimensional, depth averaged, finite elements model developed at the

University of Alberta (Ghanem et al. 1995; Steffler and Blackburn 2002), specifically for
gravel bed rivers. River2D is based on the Saint–Venant equations for the conservation of
mass and momentum which are used along with the Petrov–Galerkin implicit method to
solve depth and discharge intensities in lateral and longitudinal directions (Steffler and
Blackburn 2002). The model is able to accommodate supercritical/subcritical flow tran-
sitions, variable wetted area and ice covers. The inputs of the model include bed topog-
raphy and equivalent roughness height together with boundary conditions and initial flow
conditions.
The River2D domain geometry was obtained by a detailed and distributed GPS survey
limited to the meander area (about 700 m along the centreline). The digital terrain model
was represented by a Triangular Irregular Network (TIN) methodology. A computational
mesh was created with different resolution depending on the distance from the eroded
banks (between 0.5 and 10 m). This allowed us to find a compromise between obtaining a
detailed mesh in the areas of interest and reducing the computational time (Fig. 3).
The boundary conditions were defined by imposing a specific total discharge at the
inflow section and fixed water surface elevations at outflow section. To minimize the effect
of boundary condition uncertainties, the ends were located some distance from the area of
interest. Values of water stage were set for each run at the inflow section as the initial guess
in the iterative solution procedure. The values of water stages, both at the inflow and
outflow sections, were assumed being equal to the ones obtained through the 1D modelling.

123
Nat Hazards (2013) 69:869–887 875

Fig. 3 Computational grid and topographic elevation for the meander area used in River2D model and
geometry of the sections M1 and V1

By means of high-resolution orthophotos, the study area was divided into 4 different
zones characterized by the presence of channel/bars, high bars, herbaceous or bushy
vegetation and trees. For each one of these areas, the effective roughness heights were
defined, by assigning the corresponding value to the nodes of the mesh. In the case of
channel and bars, ks was defined through the average D90 of sediments. This was obtained
by means of grain size distributions of channel bed and bars, collected at different locations
along the study reach. In the vegetated areas, values of ks were based on tables in the
literature (Barnes 1967, Herbaceous/Bushy area, ks = 0.35 m; Area with trees,
ks = 1.00 m).

3.2 Near-bank shear stress models

In the present research, three different models were tested for the computation of the near-
bank shear stress: the empirical distribution from Simons and Senturk (1977), the com-
puting scheme of the River2D hydrodynamic model and the analytical model from Kean
and Smith (2006a). The models are described as follows:

3.2.1 The Simons and Senturk distribution

According to the Simons and Senturk model (1977), for sufficiently wide channels and
specifically if the ratio between the width and the depth of the channel is larger than 3, it is

123
876 Nat Hazards (2013) 69:869–887

possible to estimate the highest value of the bank shear stress as 75 % of the average value
of the shear stress of the section. It is important to note that this model was developed using
empirical data sets obtained from laboratory flumes. Thus, the main limitation of this
distribution, as well as others in literature obtained by means of experimental tests
(Leutheusser 1963; Kartha and Leutheusser 1972; Knight et al. 1984), is that they should
be applied only for simple geometry, as they are not able to reproduce the effects of the
natural irregularities of the banks.
The Simons and Senturk distribution was coupled with the 1D hydraulic model HEC-
RAS as shown in Table 1. Moreover, the effect of the presence of the bend on the near-
bank shear stresses was taken into account using the diagram from the Soil Conservation
Service (1977), which expresses the ratio between the boundary shear stress on the outer
bank and the averaged shear stress on the cross-section as a function of the ratio between
the radius of curvature and the channel width.

3.2.2 The River2D model

The computation of the shear stresses along the bank nodes was also obtained from
River2D. For each node of the mesh, the model calculates the non-dimensional Chezy
coefficient. This coefficient is related to the effective roughness height ks of the boundary
and the depth of flow H through (Steffler and Blackburn 2002):
 
H
CS ¼ 5:75 log10 12 ð2Þ
ks
The shear stresses are assumed to be related to the magnitude and direction of the depth-
averaged velocity U, through the Chezy parameter, such as:
jU j2
s¼q ð3Þ
Cs2

3.2.3 The Kean and Smith model

An estimation of the boundary shear stress was also made by employing the Kean and
Smith (2006a, b) analytical model for partitioning the drag on bank roughness elements
into form (sd) and skin (ssf) components. According to the model, the roughness elements
were modelled as Gaussian-shaped features. The form drag on an individual roughness
element is determined using the drag coefficient (CD) of the individual element and a
reference velocity (uref) that includes the effects of roughness elements further upstream.
The square of the reference velocity is defined to be the average of the square of the
velocity that would be present if the element was removed from the flow. The velocity in
this area is primarily affected by three interdependent regions each having turbulent
processes which scale differently: an internal boundary layer region, a wake region and an
outer boundary layer region. The model of Kean and Smith (2006a) describes the velocity
profiles in each region separately and joins them together using appropriate matching
conditions. Specifically, the total stress on the boundary has the following expression:
1 H
sT ¼ ssf þ sd ¼ qhuIBL i2 þ qCD u2ref ð4Þ
2 k
where q is the density of water, u*IBL the shear velocity inside the internal boundary layer,
H is the protrusion height of the element and k is the spacing of the Gaussian elements.

123
Nat Hazards (2013) 69:869–887 877

Kean and Smith (2006a) estimate the drag coefficient using an empirical function derived
from the experimental data of Hopson (1999):
 r
CD ¼ 1:79 exp 0:77 ð5Þ
H
where r is the streamwise length scale.
In the present research, the geometries of the topographic bank features were measured
at 5 cm intervals along two straight edges placed at 2 different levels in order to char-
acterize both the gravel and cohesive layers which compose the banks of the case study.
Moreover, the velocity in the outer boundary layer region was computed by means of the
two hydrodynamic models. In the application of this model, only the skin friction part of
the shear stress was used (Darby et al. 2010). In fact, although the total stress comprises
two components, the skin and the form drag, only the former is exerted directly onto the
sediment grains; it is, thus, the component which is relevant in the context of bank erosion.
The form drag component can be considered to be the component that is dissipated by the
macro topography, thereby reducing the total stress available for geomorphic work.

4 Calibration of the erodibility parameters and calculation of bank retreat

The application of the excess shear stress formulation (Eq. 1) to estimate the rate of fluvial
erosion requires the knowledge of the erodibility parameters sc and kd.
Recent studies (Hanson and Cook 2004; Wynn 2004; Wynn et al. 2008) addressed the
significant spatial and temporal variability in soil erodibility. By means of monthly in situ
measurements using a multiangle submerged jet test device, Wynn et al. (2008) found that
kd was 2.9 and 2.1 times higher during the winter than in the spring/fall.
At the reach examined in the present study, the banks are considered as composite rather
than being composed of gravel toe. From the computational point of view, the bank profiles of
M1 and V1 cross-sections were represented as vertically subdivided into 23 nodes. The retreat
was computed separately for each node taking into account for different bank materials
properties and hydraulic conditions. The retreat of the whole bank was assumed to be equal to
the maximum retreat over the vertical profile. This is equivalent to consider that, on a multiyear
temporal scale, the retreat at the toe of the bank extends to the cohesive part throughout mass
failures not explicitly modelled in the present study. This assumption is justified by field
evidence showing subvertical bank profiles induced by the failure of cantilevered blocks
having confined thickness. This supports the fact that fluvial erosion at the toe of the bank may
be recognized as the dominant process which controls the overall retreat in the long term.
Critical shear stress for loose gravels can be estimated by methods for incipient motion,
whereas in the case of partly packed and cemented sediments, the analytical model from
Millar and Quick (1993) can be employed. The authors provide a modification of the Lane
(1998) criterion based on an empirical coefficient. It is more difficult to determine the
value of the erodibility parameter kd, because it varies seasonally, and no reliable devices
are available to measure it (Rinaldi and Darby 2008). In view of this lack in data, the model
Bank Stability and Toe Erosion Model developed at the USDA-ARS, National Sedimen-
tation Laboratory (Simon and Curini 1998) applied the analytical model proposed by
Hanson and Simon (2001), although these results were obtained through jet testing on
cohesive materials. Rinaldi et al. (2008) and Luppi et al. (2009) in their studies based on
field monitoring along the Cecina River determined the erodibility coefficient via cali-
bration at the single flow event time scale.

123
878 Nat Hazards (2013) 69:869–887

In the present study, the mean values of the critical shear stress for gravel (sc_g), the
erodibility coefficient for gravel (kd_g) and the erodibility coefficient for cohesive material
(kd_c) were determined via calibration based on the bank retreat that occurred in about
10 years, measured by means of GIS analysis of high-resolution aerial photos. Given the
fact that only 3 aerial photos were available, and data were not enough to allow for a
validation.
The critical shear stress for cohesive material (sc_c) was assumed being equal to the
in situ measured value. The calibration procedure was performed by looking for those
values providing the observed riverbank retreat. In detail, the total amount of lateral retreat
has been subdivided into 2 periods: erosion occurred in the years 1994–2000 and in
2000–2004. Table 2 provides the values of riverbank retreat that occurred in these periods
for two sections of reference with an estimation of the error affecting the measurements.
The error depends on the aerial photo resolution and on the total residual error arising from
the georeferencing process. Hourly time series of discharge measured within the periods of
the photos were collected. Thus, the flow events were discretized considering constant
values during each time-step (=1 h). For each time-step, hydraulic models combined with
near-bank shear stress models provided the values of the boundary shear stress. Therefore,
lateral erosion was computed multiplying the erosion rate e, computed with Eq. (1), by the
time-step.
For each value of near-bank shear stress (obtained for the selected banks by applying a
different combination of models) and for each period (1994–2000, 2000–2004 and the total
period 1994–2004), the calibration of the parameters was obtained by applying the Nelder-
Mead algorithm (Nelder and Mead 1965) in order to find the values of the unknown
parameters which minimized the error between the observed retreat and the values of
retreat estimated through Partheniades formulation (Eq. 1). The Nelder-Mead algorithm is
a direct-search method in which the space of parameters is explored by mean of a simplex
that is iteratively stretched and reflected. The minimum error for each case was found by
following a multistart approach, in which the minimization was initialized with a range of
first guess values (based on values from literature) of critical shear stress for gravel
material (sc_g) and erodibility coefficient for both cohesive (kd_c) and gravel (kd_g)
sediments.
Most of the minimizations led to a small final error, so the following criteria were
adopted in order to select a unique optimal triplet between the several ones provided by the
research algorithm for each case:
• Error (difference between simulated and observed retreat) \2 m
• Value of critical shear stress for gravel similar to the values provided from Millar and
Quick (1993)
• 10-9 \ kd_c \ 10-6, 10-6 \ kd_g \ 10-4 and kd_c \ kd_g
The first criterion was chosen as a consequence of the resolution of the aerial photos.
Concerning the second criterion, the Millar and Quick (1993) formula was selected since it

Table 2 Retreat estimates in the

Frame-time Upstream bank (M1) Downstream bank (V1)
two banks considered in the study
(year) retreat (m) retreat (m)
obtained by comparison of aerial
photos. The error is estimated
based on the spatial resolution of 1994–2000 11 ± 2.1 12.5 ± 2.1
the different images 2000–2004 19.5 ± 1.5 1 ± 1.5
1994–2004 30.5 ± 1.3 13.5 ± 1.3

123
Nat Hazards (2013) 69:869–887 879

was specifically developed for partly packed and cemented sediments, resulting in 5.68 and
4.26 N/m2 for the upstream and downstream bank, respectively. The last criterion was
defined considering the values reported in literature (Hanson and Simon 2001) for similar
types of sediment.

5 Results and proposals for the risk evaluation

5.1 Results of the modelling and calibration

Figure 4 presents the maximum values of the excess near-bank shear stress (s–sc in Eq. 1)
along the height of the upstream bank (M1) obtained through steady flow analyses for
different discharges. Different models or combinations of models provided different results
both in quantitative and qualitative terms.
According to the 1-D hydraulic model, the excess shear stress on the bank increases
with the discharge. In the case of application of the Simons and Senturk distribution
(Fig. 4a), it is possible to observe a local decrease in correspondence of 147 m3/s due to
the beginning of the bar overtopping, followed by a continuous increase of the excess shear

Fig. 4 Mean value of the excess near-bank shear stresses (i.e. difference between shear stress exerted by the
flow and critical shear stress of the sediment) estimated for different discharges with different models at the
upstream riverbank. SS: Simons and Senturk (1977); HR: HEC-RAS; R2D: River2D; KS: Kean and Smith
(2006a, b)

123
880 Nat Hazards (2013) 69:869–887

stress after the total submersion of the bar. In the Kean and Smith case (Fig. 4b), the excess
shear stress increases rapidly until a discharge flow of 147 m3/s, and then, it reaches the
maximum value more slowly due to the increase of the wet section area. Since Simons and
Senturk is based on the average velocity on the whole section, while Kean and Smith
method is sensitive to the local velocities near the bank, in the latter case, the slight
decrease of the velocity computed by HEC-RAS for higher discharge flows causes a
decrease of the excess shear stress. Concerning the other two cases shown in Fig. 4 (panels
c and d), the values of excess boundary shear stress obtained by River2D reach a maximum
in correspondence to a discharge of 147 m3/s and decrease for increasing discharges. In
fact, in these cases, the presence of the bends most likely induces a shifting of the main
flow along a chute channel in the central part of the cross-section during high flows,
thereby causing a reduction in near-bank shear stresses. Similar findings were reported in
Rinaldi et al. (2008), although the research was carried out at a different reach along the
Cecina river and a different hydrodynamic model (Delft3D) was applied. The different
values of the excess near-bank shear stress are mostly due to the fact that in the Kean and
Smith case, only the skin component of the shear stress was used (see Eq. 4), whereas
River2D model only considers the total value.
These results confirm how one-dimensional models are not appropriate to describe
processes of those riverbanks located along meanders, although their employment is
definitely easier, especially in terms of required input data.
Since different combinations of models provided important differences in the estimation
of near-bank shear stresses, it was not possible to find a unique triplet of erodibility parameters
producing a rate of retreat with acceptable errors at the same time for all of the cases.
The erodibility parameters determined via calibration are reported in Table 3 for all the
combinations of the models. Due to the intermittent behaviour of the retreat, especially at
the downstream bank (see Table 2), it was not always possible to find a triplet in which
calibrated parameters were in the range of values taken from literature and, at the same
time, that produced acceptable errors. This happened in two combinations of models for
the downstream bank: by combining the Kean and Smith and the one-dimensional model,
and by computing the value of boundary shear stress through the 2D-modelling. For all the
other cases, small errors between the simulated and observed values of retreat were found.
The results of the calibration highlight the differences between the models employed:
the erodibility parameter for gravel (kd_g) is higher in the case of the two-dimensional
model to balance the lower shear stresses computed with the River2D model (or with the
River2D combined with the Kean and Smith model). It has to be highlighted that the
observed retreat during the 2000–2004 period at section V1 was particularly small
(1 ± 1.5, see Table 2). Thus, during the calibration process, the need to match such small
observed retreat led to the enhancement of the differences between the shear stress dis-
tribution provided by the two near-bank shear stress models (SS and KS).
As an example, Fig. 5 shows the cumulative bank retreat through the period 1994–2004
for site M1 computed with two different combinations of models: River2D-Kean and
Smith and HEC-RAS-Simons and Senturk. The Figure shows the comparison with the
observations retrieved from the aerial photos. As it is possible to see, the shape of the
cumulative retreat curve for the River2D-Kean and Smith case confirms that the range of
effective discharges is narrower with respect to the HEC-RAS-Simons and Senturk
combination of models. In fact, in the R2D-KS case, the erosive events are less numerous
and more intense (few large jumps in the corresponding curve in Fig. 5, likely closer to the
actual behaviour); while in the HR-SS case, the erosive events are more distributed along
the whole period (several small jumps in Fig. 5).

123
Nat Hazards (2013) 69:869–887 881

Table 3 Erodibility parameters obtained by mean of the calibration of the case study
Section Hydrodynamic Shear stress sc_g (Pa) kd_c (m3/Ns) kd_g (m3/Ns) Error (m)
models. model.

1994–2004
M1 HR SS 5.61 5.07E-07 1.07E-06 \0.5
M1 HR KS 5.59 5.24E-07 1.38E-06 \0.5
M1 R2D R2D 5.67 2.42E-07 6.85E-06 \0.5
M1 R2D KS 5.65 4.04E-07 2.75E-05 \0.5
V1 HR SS – – – –
V1 HR KS 4.39 2.89E-08 6.47E-07 \0.5
V1 R2D R2D – – – –
V1 R2D KS 4.09 2.58E-07 7.06E-05 \0.5
1994–2000
M1 HR SS 5.56 7.25E-07 1.15E-06 \0.5
M1 HR KS 5.41 6.70E-07 1.54E-06 \0.5
M1 R2D R2D 5.64 1.41E-07 9.41E-06 \0.5
M1 R2D KS 5.61 1.01E-08 4.59E-05 \0.5
V1 HR SS 4.96 6.78E-08 3.54E-07 \0.5
V1 HR KS 4.56 2.92E-07 2.25E-06 \0.5
V1 R2D R2D 2.72 7.22E-07 2.57E-04 \0.5
V1 R2D KS 4.23 5.03E-07 2.93E-04 \0.5
2000–2004
M1 HR SS 5.74 4.05E-08 1.08E-06 \0.5
M1 HR KS 5.71 9.71E-07 1.33E-06 \0.5
M1 R2D R2D 5.67 7.54E-07 5.92E-06 \0.5
M1 R2D KS 5.64 3.02E-07 2.17E-05 \0.5
V1 HR SS 29.13 2.89E-07 3.53E-07 \0.5
V1 HR KS 8.01 7.36E-08 1.26E-07 \0.5
V1 R2D R2D 4.5 1.12E-07 3.72E-06 1.06
V1 R2D KS 4.28 4.28E-07 9.03E-06 \0.5

5.2 A characterization of the risk due to fluvial erosion

Notwithstanding the fact that positive aspects are to a certain degree related to the mobility
of river channels (Piégay et al. 1997, 2005; Florsheim et al. 2008), and riverbank retreat
represents a hazard when it occurs in urbanized areas.
There are several existing definitions for similar principles within the risk management
processes, as recognized in Pliefke et al. (2007). In the present research, the risk is mea-
sured in terms of monetary units per time unit.
Compared with most natural hazards, fluvial erosion is a relatively frequent process that
can occur when a critical value of the shear stress is exceeded by the flow, and the sediment
transport is not at its maximum capacity (Wiberg and Smith 1989; Duan 2005).
Furthermore, the analyses carried out along the Cecina River to estimate near-bank
shear stress showed that the amount of the retreat was not proportional to the discharge,
due to the nonlinear relationships between flows and shear stresses in the different models

123
882 Nat Hazards (2013) 69:869–887

Fig. 5 Cumulative bank retreat through the period 1994–2004 for site M1 computed with two different
combination of models: River2D-Kean and Smith (blue line) and HEC-RAS-Simons and Senturk (red line).
The employed erodibility parameters were obtained by the calibration on the 1994–2004 period

combinations. These considerations led to the development of an algorithm for a charac-

terization of the risk due to fluvial erosion that is summarized in the following steps:
(1) Computation of the flow duration curve of the river;
(2) Estimation of the near-bank shear stress for each value of the discharge;
(3) Application of the fluvial erosion model for each value of the near-bank shear stress;
(4) Computation of the local loss duration curve.
The result of the algorithm is shown in the co-axial graphs represented in Fig. 6, where
for each step of the procedure, there is a corresponding quadrant, reading clockwise.
For the study case, the first quadrant presents the flow duration curve obtained from
hourly time series of discharge flows in the period 1974–2007.
The analyses of the near-bank shear stress, combined with the calibrated erodibility
parameters, provided, for increasing discharges, values of the shear stress responsible for
the fluvial erosion (excess shear stress sexc) defined as:
sexc ¼ maxðs  sc ; 0Þ ð6Þ
where s is the shear stress along the bank profile, and sc is the critical shear stress.
In the third step, the local loss (here expressed in €/mdays) was defined through the rate
of retreat and the economical value of the eroded soil (in this case the land use was
agricultural). As an example, the price of land for the case study was set at 5€/m2. The line
in the third quadrant was obtained by applying the excess shear stress linear formulation
from the (Eq. 1) (Partheniades 1965; Arulanandan et al. 1980) and multiplying it by the
price of the land in order to obtain directly the monetary loss per unit of longitudinal bank
length per day.

123
Nat Hazards (2013) 69:869–887 883

Fig. 6 Co-axial graph representing the risk analysis for the M1-R2D-KS case (a) and M1-HR-SS case (b)

123
884 Nat Hazards (2013) 69:869–887

The graphic in the fourth quadrant represents the local loss duration curve, expressed, in
analogy to the flow duration curve, through a complementary cumulative probability
distribution function; given a certain value of loss, this function yields the number of days
in the year in which such value is exceeded. Therefore, a characterization of the mean
hydrological year in terms of local loss is defined.
The value of risk can now be evaluated as follows:
1Zyear

R¼ Ldt ð7Þ
0

where R is the local (i.e. for the considered river section) risk in the mean hydrological
year, and L is the daily local loss.
Since the erodibility parameters were determined via calibration for each combination
of hydrodynamic models and near-bank shear stress models, the resulting final values of
the risk are very similar. The different distributions in the mean hydrological year are more
interesting.
Figure 6 shows two scenarios obtained by applying two different combinations of
models. The examples reported here were selected to highlight differences resulting from
the models. In fact, M1-HR-SS represents the combination of the most straightforward
models, HEC-RAS and Simons and Senturk shear stress distribution, whereas M1-R2D-KS
is the combination of the most complex models applied in the present study, namely
River2D and Kean and Smith models. From Fig. 6, it can be observed, for the M1-HR-SS
and M1-R2D-KS cases, that the non-monotonic relationships between the discharge flow
and the excess shear stress reflect in different ways on the shape of the loss duration curve.
This co-axial graph may be used as a nomogram to yield the expected loss corresponding
to a discharge flow value with a known occurrence frequency in the mean hydrological
year. In Fig. 6 some examples of such paths are shown.
It is important to note that the characterization of the risk in the present research is only
statistical since we are not able to predict when the loss will occur.
It should be also remarked that these analyses were carried out taking account only of
the fluvial erosion, although mass failures represent an important process responsible for
channel migration. Moreover, the presence of structures such as bridges, near-river roads
was not taken into account in order to provide a ‘‘continuous’’ relation between the retreat
and the damage. Taking into account, also these aspects would require more complex
models that are able to perform stability analyses. Finally, only the retreat of the basal layer
composed by gravel was considered since it is reasonable to assume that fluvial erosion
mainly occurs in this part of the bank (Thorne 1982).

6 Conclusions

Different estimations of the near-bank shear stress exerted by the flow were carried out by
employing two hydrodynamic models combined with different shear stress models, with
the specific objective to define an operative tool for assisting in riverbank retreat risk
assessment. A calibration of the erodibility parameters required for the analyses of fluvial
erosion was carried out for different combinations of models and for different periods at
two representative eroding banks.

123
Nat Hazards (2013) 69:869–887 885

Based on the results obtained in the present study, a statistical characterization of the
damage was proposed. Specifically, a method for the estimation of the risk due to the
fluvial erosion in monetary terms was developed. This algorithm requires the flow duration
curve, the values of the near-bank shear stress by means of the appropriate models and the
values of the erodibility parameters as input data for the fluvial erosion model.
The analyses proposed here allow a statistical evaluation of the risk for a specific cross-
section of the channel, in terms of economical damage due to the loss of land. The presence
of structures and infrastructures was not taken into account in these analyses.
Notwithstanding uncertainties about values of the near-bank shear stress exist since it
was not possible to make direct measurements of the velocities, this conceptual model
represents a potential basis for assisting in river management.
Future development could take account of the presence of different layers of material in
composite banks.

References

Abernethy B, Rutherfurd ID (2000) The effect of riparian tree roots on the mass-stability of riverbanks.
Earth Surf Process Landf 25:921–937
Amiri-Tokaldany E, Darby SE, Tosswell P (2003) Bank stability analysis for predicting land loss and
sediment yield. J Am Water Resour Assoc 39:897–909
Arulanandan K, Gillogley E, Tully R (1980) Development of a quantitative method to predict critical shear
stress and rate of erosion of natural undisturbed cohesive soils. Report GL-80-5, US Army Engineers,
Waterways Experiment Station, Vicksburg
Barnes HH (1967) Roughness characteristics of natural channels. USGS Water Supply Paper 1849
Bravard JP, Amoros C, Pautou G (1986) Impact of civil engineering works on the successions of com-
munities in a fluvial system. A methodological and predictive approach applied to a section of the
Upper Rhone river, France. Oikos 47(1):92–111
Bravard JP, Kondolf GM, Piégay H (1999) Environmental and societal effects of channel incision and remedial
strategies. In: Darby SE, Simon A (eds) Incised river channels. Wiley, New York, pp 303–341
Bull LJ (1997) Magnitude and variation in the contribution of bank erosion to the suspended sediment load
of the River Severn, UK. Earth Surf Process Landf 22:1109–1123
Cassuli V, Cattani E (1994) Stability, accuracy and efficiency of a semi-implicit method for three-dimen-
sional shallow water flow. Comput Math Appl 27:99–112
Church M (1992) Channel morphology and typology. In: Calow P, Petts GE (eds) The rivers handbook.
Hydrological and ecological principles. Blackwell Scientific Publications, Oxford, pp 126–143
Dapporto S, Rinaldi M (2003) Modelling of river bank retreat by combining fluvial erosion, seepage and
mass failure. Geophys Res Abstracts EGS-AGU-EUG Joint Assembly, Nice, France, 6–11 April 2003
Darby SE, Thorne CR (1996) Development and testing of river-bank stability analysis. J Hydraul Eng
122(8):443–454
Darby SE, Rinaldi M, Dapporto S (2007) Coupled simulations of fluvial erosion and mass wasting for
cohesive riverbanks. J Geophys Res 112:1–15
Darby SE, Trieu HQ, Carling PA, Sarkkula J, Koponen J, Kummu M, Conlan I, Leyland J (2010) A
physically based model to predict hydraulic erosion of fine-grained riverbanks: the role of form
roughness in limiting erosion. J Geophys Res 115:F04003. doi:10.1029/2010JF001708
Department of the Army US Army Engineer District, Alaska (2006) Kenai river bank erosion. Technical
Report Kenai, Alaska
Downs PW, Simon A (2001) Fluvial geomorphological analysis of the recruitment of large woody debris in
the Yalobusha river network, Central Mississippi, USA. Geomorphology 37:65–91
Downward SR, Gurnell AM, Brookes A (1994) A methodology for quantifying river channel planform
change using GIS: variability in stream erosion and sediment transport. In Olive LJ, Loughran RJ,
Kesby JA (eds) International Association of Hydrological Sciences, Publication 224. Proceedings of
the Canberra Symposium, Canberra, 1994, pp 449–456
Duan JG (2005) Analytical approach to calculate rate of bank erosion. J Hydraul Eng 131(11):980–990
Eaton BC, Church M, Millar RG (2004) Rational regime model of alluvial channel morphology and
response. Earth Surf Process Landf 29:511–529

123
886 Nat Hazards (2013) 69:869–887

Florsheim JL, Mount JF, Chin A (2008) Bank erosion as a desirable attribute of rivers. Bioscience
58:519–529
Ghanem A, Steffler P, Hicks F, Katopodis C (1995) Two dimensional modeling of flow in aquatic habitats.
Water Res Eng Rep No 95–S1. University of Alberta, Canada
Hanson GJ, Cook KR (2004) Apparatus, test procedures, and analytical methods to measure soil erodibility
in situ. Appl Eng Agric 20(4):455–462
Hanson GJ, Simon A (2001) Erodibility of cohesive streambeds in the loess area of the midwestern USA.
Hydrol Process 15:23–38
Hopson TM (1999) The form drag of large natural vegetation along the banks of open channels. M. S. thesis,
University of Colorado, Boulder, p 114
Kartha VC, Leutheusser HJ (1972) Distribution of tractive force in open channels. J Hydraul Div ASCE
96:1469–1483
Kean JW, Smith JD (2006a) Form drag in rivers due to small-scale natural topographic features: 1. Regular
sequences. J Geophys Res 111:F04009. doi:10.1029/2006JF000467
Kean JW, Smith JD (2006b) Form drag in rivers due to small-scale natural topographic features: 2. Irregular
sequences. J Geophys Res 111:F04010. doi:10.1029/2006JF000490
Knight D, Demetriou JD, Hamed ME (1984) Boundary shear in smooth rectangular channels. J Hydraul Eng
110:405–422
Lane SN (1998) Hydraulic modelling in hydrology and geomorphology: a review of high resolution
approaches. Hydrol Process 12:1279–1298
Lawler DM, Thorne CR, Hooke JM (1997) Bank erosion and instability. In: Thorne CR, Hey RD, Newson
MD (eds) Applied fluvial geomorphology for river engineering and management. Wiley, Chichester,
pp 137–172
Leutheusser HJ (1963) Turbulent flow in rectangular ducts. Proc Am Soc Civ Eng 89(HY3):1–19
Luppi L, Rinaldi M, Teruggi LB, Darby SE, Nardi L (2009) Monitoring and numerical modelling of
riverbank erosion processes: a case study along the Cecina River (Central Italy). Earth Surf Process
Landf 34(4):530–546. doi:10.1002/esp.1754
Mahdi T (2007) Pairing geotechnics and fluvial hydraulics for the prediction of the hazard zones of an
exceptional flooding. Nat Hazards 42:225–236. doi:10.1007/s11069-006-9096-8
Malavoi JR, Bravard JP, Piégay H, Herouin E, Ramez P (1998) Determination de l’espace de liberte0 des
cours d’eau. Guide technique no. 2, SDAGE RMC, p 39
Marston RA, Girel J, Pautou G, Piégay H, Bravard JP, Arneson C (1995) Channel metamorphosis, flood-
plain disturbance, and vegetation development: Ain River, France. Geomorphology 13:121–131
Millar RG, Quick MC (1993) Effect of bank stability on geometry of gravel rivers. J Hydraul Eng
119(12):1343–1363
Nanson GC, Hickin EJ (1986) A statistical analysis of bank erosion and channel migration in western
Canada. Geol Soc Am Bull 97:497–504
Nelder JA, Mead R (1965) A simplex method for function minimization. Comput J 7:308–313. doi:
10.1093/comjnl/7.4.308
Olsen NRB (2002) A three-dimensional numerical model for simulation of sediment movements in water
intakes with moving option. User’s Manual, Department of Hydraulic and Environmental Engineering
The Norwegian University of Science and Technology, Trondheim, Norway 2002
Partheniades E (1965) Erosion and deposition of cohesive soils. J Hydraul Div ASCE 91:105–139
Piégay H, Darby SE, Mosselman E, Surian N (2005) A review of techniques available for delimiting the
erodible river corridor: a sustainable approach to managing bank erosion. River Res Appl 21:773–789
Piégay H, Cuaz M, Javelle E, Mandier P (1997) Bank erosion management based on geomorphological,
ecological and economic criteria on the Galaure River, France. Regul Rivers Res Manag 13:433–448
Pliefke T, Sperbeck ST, Urban M, Peil U, Budelmann H (2007) A standardized methodology for managing
disaster risk—an attempt to remove ambiguity. 5th International Probabilistic Workshop, Ghent
Pollen N, Simon A (2005) Estimating the mechanical effects of riparian vegetation on stream bank stability
using a fiber bundle model. Water Resour Res 41:W07025
Pollen-Bankhead N, Simon A (2008) Enhanced application of root-reinforcement algorithms for bank-
stability modeling. Earth Surf Process Landf 34(4):471–480
Reneau SL, Drakos PG, Katzman D (2004) Geomorphic controls on contaminant distribution along an
ephemeral stream. Earth Surf Process Landf 29:1209–1223
Rinaldi M, Casagli N (1999) Stability of streambanks formed in partially saturated soils and effects of
negative pore water pressures: the Sieve River (Italy). Geomorphology 26:253–277
Rinaldi M, Darby SE (2008) Modelling river-bank-erosion processes and mass failure mechanisms: progress
towards fully coupled simulations. In: Habersack H, Piégay H, Rinaldi M (eds) Gravel-Bed rivers 6—

123
Nat Hazards (2013) 69:869–887 887

from process understanding to river restoration. Series developments in earth surface processes, 11.
Elsevier, Netherlands, pp 213–239
Rinaldi M, Luppi L, Mengoni B, Darby SE, Mosselman E (2008) Numerical simulation of hydrodynamics
and bank erosion in a river bend. Water Resour Res 44. doi:10.1029/2008WR007008
Simon A, Curini A (1998) Pore pressure and bank stability: the influence of matric suction. In: Abt SR,
Young-Pezeshk J, Watson CC (eds) Water resources engineering ’98. ASCE, Reston, pp 358–363
Simon A, Curini A, Darby SE, Langendoen EJ (2000) Bank and near-bank processes in an incised channel.
Geomorphology 35:193–217
Simon A, Langendoen EJ, Collison A, Layzell A (2003) Incorporating bank-toe erosion by hydraulic shear
into a bank-stability model: Missouri River, Eastern Montana. Proceedings, EWRL-ASCE, World
Water and Environmental Resources Congress, Cd-Rom, 11
Simon A, Pollen NL, Langendoen EJ (2006) Influence of two woody riparian species on critical conditions
for streambank stability: upper Truckee River, California. J Am Water Resour Assoc 42:99–113
Simons DB, Senturk F (1977) Sediment transport technology. Water Resour Pub, Forth Collins
Soil Conservation Service (1977) Design of open channels. Engineering Div., Technical Release, No 25,
Washington
Steffler P, Blackburn J (2002) River2D: two-dimensional depth-averaged model of river hydrodynamics and
fish habitats. University of Alberta, Edmonton, Canada
Thorne CR (1982) Processes and mechanisms of river bank erosion. In: Hey RD, Bathurst JC, Thorne CR
(eds) Gravel-Bed rivers: fluvial processes, engineering, and management. Wiley, Chichester,
pp 227–271
US Army Corps of Engineers (1983) Sacramento River and tributaries bank protection and erosion control
investigation. California Sediment Transport Studies, Sacramento Dist., US Corps of Eng., Sacra-
mento, CA
US Army Corps of Engineers (2002a) HEC-RAS River Analysis System. Hydraulic Reference Manual.
Version 3.1. US Army Corps of Engineers Hydrologic Engineering Center, Davis, CA, November
2002
US Army Corps of Engineers (2002b) HEC-RAS River Analysis System. User’s Manual. Version 3.1. US
Army Corps of Engineers Hydrologic Engineering Center, Davis, CA, November 2002
US Army Corps of Engineers (2002c) HEC-RAS River Analysis System. Application Guide. Version 3.1.
US Army Corps of Engineers Hydrologic Engineering Center, Davis, CA, November 2002
Vardy S, Saunders JE, Tolhurst TJ, Davies PA, Paterson DM (2007) Calibration of the high-pressure
cohesive strength meter (CSM). Cont Shelf Res 27:1190–1199
Wasklewicz TA, Anderson S, Liu PS (2004) Geomorphic context of channel locational probabilities along
the lower Mississippi River, USA. Geomorphology 63:145–158
Wiberg PL, Smith JD (1989) Model for calculating bed load transport of sediment. J Hydraul Eng
115(1):101–123
Wynn TM (2004) The effects of vegetation on streambank erosion. PhD Dissertation, Virginia Polytechnic
Institute and State University, Blacksburg, Virginia
Wynn TM, Henderson MB, Vaughan DH (2008) Changes in streambank erodibility and critical shear stress
due to subaerial processes along a headwater stream, southwestern Virginia, USA. Geomorphology
97(3–4):260–273

123