Content uploaded by Marc Vuillet
Author content
All content in this area was uploaded by Marc Vuillet on Apr 14, 2018
Content may be subject to copyright.
Development of an expert-led GIS-based approach for assessing
the performance of river levees: the Digsure method and tool
L. Peyras1, R. Tourment1, M. Vuillet2,3, B. Beullac1, C. Delaunay4and G. Bambara1
1 Irstea, Unité OHAX, Aix-en-Provence, France
2 G2C, ingénierie, Venelles, France
3 EIVP, Paris, France
4 Société du Canal de Provence, Le Tholonet, France
Correspondence
Laurent Peyras, Irstea, UR OHAX, 3275
Route de Cézanne CS 40061,
Aix-en-Provence Cedex 5, 13182, France
Tel: +33442669908
Email: laurent.peyras@irstea.fr
DOI: 10.1111/jfr3.12178
Key words
French regulatory levee hazard study;
levee; levee failure; levee management;
levee performance assessment; risk
analysis.
Abstract
France and other regions around the world frequently undergo devastating flood
events. Any failure of a flood protection structure can lead to human casualties
and material damage. Unfortunately, such long linear structures are often poorly
maintained and have shown repeated signs of weakness. Therefore, river levee
management raises a number of substantial issues for the decision-makers
responsible for ensuring maximum safety for the surrounding population at a
reasonable management cost. The aim of the Digsure projectistoprovidelevee
managers with scientific methods and information technology (IT) tools for
levee management. The Digsure method is a levee assessment method based on a
probabilistic pattern. It produces levee performance indicators and integrates
data and the uncertainty on results. Digsure is a geographic information system
tool that implements the Digsure method. It estimates and displays levee perfor-
mance and associated uncertainty intervals throughout levee assessments.
Introduction
Levee and leveed area
River levees are long linear civil-engineering structures. They
are designed to prevent water inflows in areas naturally
prone to flooding. The extent of a linear levee structure and
its leveed area may vary significantly. Although 10-m long
levees sometimes protect areas smaller than 1 ha, other
levees can stretch over several hundred kilometres to protect
lowlands near a large river. In normal circumstances, they are
mostly dry and seldom suffer loading for many years, hence
the lack or absence of adequate monitoring of their condi-
tion. Levees are often old, poorly known and maintained
structures composed of heterogeneous construction
materials. Although they withstand most seasonal floods,
levees may suddenly collapse as a result of rare or exceptional
natural events.
Levees are intended to protect an area against natural
flooding events or sea storm surges up to a certain flood or
storm water level related to the height of the levee. Even this
level of protection is limited, as there is a risk of failure of the
levee before water reaches the crest of the levee system.
Indeed, the presence of a flood protection system, including
levees, transforms a natural hazard into a combination of
natural (flood/storm) and infrastructure failure hazards
(levee breach). In some cases,due to the increased velocity of
flows through a potential levee breach,the actual level of risk
in the ‘leveed area’ can be higher because of the levee than it
would be without it. A few breaches may jeopardise the
entire flood protection system, as seen in a number of recent
disasters: the Agly flood in 1999 (Paquier, 2010), the Katrina
storm in 2004 (ASCE, 2007) and the Xynthia storm in 2010
(Kolen et al., 2013), etc.
Issues and goals of the Digsure method and tool
Levees give rise to many technical issues. Levee performance
assessments require different skills: civil engineering,
hydraulics, hydrology, geotechnics, geophysics, etc. These
technical aspects, as well as the various deteriorations that
can result from erosions and seepages, can vary significantly
along the levee structure. Due to the highly variable nature of
structures, high costs and limited investigation technologies,
the data available on levees are mostly uncertain or incom-
plete. Consequently, the engineer has two missions: the first
is to interpret the data and, where necessary, according to the
nature and spatial distribution of the levees, to extrapolate or
interpolate them; the second is to perform a qualitative
J Flood Risk Management •• (2015) ••–•• © 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd
expert assessment of the level of performance of the struc-
tures concerned.
Since the 2000s, regulations in France and in other Euro-
pean countries require that levee managers should perform
levee assessments and visual inspections at regular intervals.
Levee managers may then prioritise their maintenance
operations. In many cases, a French regulatory levee risk
study is also required and must be based on precise levee
assessments. In order to perform a levee assessment, engi-
neers from levee management technical services or research
units have to consider a wide range of data and uncertainties
observed along the linear levee structure. A tool designed to
assist engineers in their levee assessment tasks would satisfy
a major need expressed by levee managers.
Levee assessment is based on a multidisciplinary method-
ology that requires a great number of investigations: histori-
cal research, visual inspections, morphodynamic studies,
hydraulic modelling, geotechnical and geophysical explora-
tions, empirical, physical and mathematical models for levee
failure modes, etc. (Mériaux and Royet, 2007). The Interna-
tional Levee Handbook (ILH) is considered to be the most
comprehensive work on levees (Ciria et al., 2013). This
guide includes the latest developments in levee structure
investigation and assessment (Tourment et al., 2013). The
FloodProBE project, co-funded by the European Commu-
nity Seventh Framework Programme for European Research
and Technological Development (2009–2013), highlights
existing methods for analysing the safety modes of dikes. In
this field, the methods listed range from the relatively simple
(expert assessment) to the very complex (mathematical
models): (1) Expert assessment based on previous experi-
ences without explicitly using forms or index-based
methods; (2) index-based methods in which a number of
performance features are assessed to determine safety; (3)
empirical models for levee failure modes where rules can be
established to assess performance; and (4) physical and
mathematical models of levee failure modes to determine the
safety of a levee based solely on physics (FloodProbe, 2012).
Due to complex levee failure modes (Simm et al., 2012)
and to the considerable uncertainty on the data available on
levees (Vuillet et al., 2013), a high level of expertise is
required to provide assessments of these structures (Mériaux
and Royet, 2007). Therefore, the engineer must review the
available data, and identify and carry out further investiga-
tions until sufficient information is obtained to provide a
good levee performance assessment according to his expert
judgment. Then, he must identify homogeneous segments
and assess performance as a function of the different levee
failure modes (Vuillet et al., 2013).
Different levee management organisations have developed
different methods and tools to conduct levee assessments. A
levee assessment involves determining the reliability of a
levee for the main or all types of levee failure modes (Van der
Meij et al., 2012). Many reliability analysis tools can be used
to determine the overall reliability of a levee, such as the Risk
Assessment for Strategic Planning method in the UK
(Gouldby et al., 2008), and the flood early warning system –
dike analysis module (FEWS-DAM) in the Netherlands
(Knoeff et al., 2011), which use physics-based models, per-
formance indicators or expert assessment. A general frame-
work for levee assessment and levee performance estimation
is defined in the FloodProbe WP03-01-12-24. ‘Combining
information for urban levee assessment’ (FloodProbe, 2012)
found in Chapter 5 of the ILH handbook, ‘Levee inspection,
assessment and risk attribution’ (Ciria et al., 2013).
Since 2004 in France, Irstea has been developing a
geomatic application [système d’information à référence
spatiale – Digues (SIRS-Digues)] designed to help local levee
managers handle information on such structures (Maurel
et al., 2004). It is currently used under operational condi-
tions by several levee managers in France. A method was
developed to improve the tool by including ability function
to assess levees based on visual data from inspections (Serre
et al., 2007, 2008).
Reviewing the literature and investigating national and
international levee management practices reveals, on the one
hand, the use of streamlined methods for assessing levee
conditions and assisting levee management without prior
diagnostics or preliminary analyses and, on the other hand,
spatially referenced information systems for levees without
levee performance assessment. The main challenge here is to
develop methods and geographic information system (GIS)
tools that can assist levee management by providing geo-
referenced information and knowledge on levees, including
performance assessment and assistance for diagnostic
analysis.
The research and development project Digsure (i.e.
SafeDike), performed in 2009, aims to satisfy such require-
ments (Tourment et al., 2012). It comprises two approaches:
• A research and development approach intended to provide
a levee assessment method referred to as the Digsure
method, which includes the main levee failure modes and
takes data uncertainty into account;
• An IT approach designed to develop an operational GIS
tool, referred to as the Digsure tool, to calculate and spa-
tially reference levee performance according to the various
levee failure modes considered, in order to obtain perfor-
mance indicators and provide levee management research
units and technical services with a levee assessment and
management tool.
Therefore, Digsure is a levee assessment method aimed at
estimating the performance of levees by using a probabilistic
approach. This method does not aim to directly estimate the
probability of levee failure, as is required in a full flood risk
analysis. However, it produces probabilistic distributions for
levee performance indicators. Thus the Digsure method does
2Peyras et al.
© 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd J Flood Risk Management •• (2015) ••–••
not produce fragility curves directly in terms of criteria such
as flood elevation, as was developed in the FloodProbe
Report WP03-01-12-24 ‘Combining information for urban
levee assessment’ (FloodProbe, 2012), but it could be used to
do so.
This article first describes the development of a new
levee assessment method. The latter consists in building
performance indicators that represent levee performance
according to different levee failure modes. Second, it sets
out the development of the Digsure tool by incorporating
levee performance indicators in a GIS to provide a spatially
referenced assessment of levee performance along the
linear structure. The final goal of this tool is to produce
levee performance indicators and related uncertainties, in
order to assist levee managers to estimate flood risk and
prioritise actions on levee segments. These actions
can include further investigations required to reduce
uncertainties.
Developing the Digsure method for levee
performance assessment
Levee functional modelling
Different deterioration mechanisms may have a significant
impact on the components of levees and lead to the risk of
failure and possible breaching. Such deterioration mecha-
nisms have very different and complex forms and possible
combinations. Therefore, a large variety of levee failure sce-
narios may be considered for a given levee (Ciria et al.,
2013).
A levee failure scenario, named after its initiating or pre-
dominant deterioration mechanism, is referred to as a levee
failure mode. It consists of a sequence of deteriorated levee
components, or successive degradations and failures of the
levee’s component functions.
According to the ILH, the main levee failure mechanisms
are: external erosion, internal erosion and instability (Ciria
et al., 2013). The Digsure model is based on these three levee
failure mechanisms. It studies internal erosion, external
erosion through overflowing and scour mechanisms, and
instability through landside and waterside sliding mecha-
nisms. Thus, the five levee failure modes treated in the
Digsure model are:
• Overflowing,
• Internal erosion,
• Scouring,
• Landside sliding,
• Waterside sliding.
A functional model was developed in order to analyse
these levee failure modes, using reliability methods: the
functional analysis and the Failure Mode and Effect Analysis
(Modarres, 1993; Peyras et al., 2006). These methods are
used to identify all the functions of levee components, their
functional failure modes and the causes and effects of these
functional failure modes. In cooperation with a group of
experts specialising in hydraulic structures, the application
of these methods resulted in the definition of three groups of
variables in the model (Figure 1):
• Basic indicators: these are fundamentals that detail the
information to be considered for determining each cri-
terion. These can comprise raw data or information inter-
preted or inferred from measures, observations,
computations or material: visual data, historical data,
geotechnical tests, modelling, etc.
• Functional criteria: decision-making items used to assess
levee component performance. They help to determine
how well the levee component functions are performed. A
number of basic indicators must be reviewed before a cri-
terion can be determined.
• Performance indicators: these determine levee perfor-
mance against levee failure modes by combining several
functional criteria.
Based on these variables, the functional model provides a
representation of levee failure modes as sequences of succes-
sive failures of technical functions according to functional
criteria and their related basic indicators (Vuillet et al.,
2012). Figure 2 shows an example of how the overflowing
Figure 1 Generic structure and variables of the Digsure model.
Assessing the performance of river levees 3
J Flood Risk Management •• (2015) ••–•• © 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd
levee failure mode can be modelled on the basis of its specific
functional criteria.
The data used for functional modelling are the same as
those used by levee engineers. They include visual data stem-
ming from field studies, geotechnical data resulting from
tests and data produced from mechanical, hydraulic and
numerical models. These data are critical information for
levee assessment. Finally, 26 functional criteria and nearly 90
related basic indicators were defined for the levee assess-
ment. The levee failure modes were modelled as causal dia-
grams (Vuillet et al., 2012).
For example, the ‘levee body prevents overflowing’ func-
tion is estimated in relation to four functional criteria: ‘pres-
ence of low elevation points in crest’, ‘operation of spillway’,
‘flow obstruction factors’ and ‘levee crest elevation’. The last
functional criterion is determined by taking into account the
basic indicators ‘levee crest longitudinal profile’ and ‘longi-
tudinal profile of flood water elevation’.
Developing levee performance indicators
Levee assessment and the associated functional modelling of
levees are based on the identification of homogeneous levee
segments (Ciria et al., 2013). Each homogeneous levee
segment is defined by its type, the association of different
components that form a specific cross-sectional geometry
and shape, and its real physical properties (Ciria et al., 2013).
No single method exists to guide the division of levee struc-
tures into homogeneous segments in terms of type and avail-
able data. According to FloodProbe WP03-01-12-24
(FloodProbe, 2012), the two main types of segmentation are
the division of levees into segments of equal length and
dynamic segmentation.
To optimise levee assessment resolution, the Digsure
method models levees as continuous linear objects divided
by dynamic segmentation. Thus, the performance of levee
segments is estimated in relation to cross-sections that are
the smallest unit for which information and therefore per-
formance can be considered as homogeneous. A homo-
geneous levee segment is a continuous set of local cross-
sections that match a homogeneous performance estimation
(Vuillet et al., 2012). Information on levees must be pro-
vided linearly with geo-referenced guide marks in order to
identify homogeneous levee segments. In some cases, seg-
ments (of zero length) may contain isolated irregularities.
According to the principle explained in FloodProbe WP03-
01-12-24 (FloodProbe, 2012), the method has been designed
to be integrated into GIS software to calculate levee perfor-
mance. This tool includes a spatially referenced database
gathering all the information available on basic indicators,
functional criteria and performance indicators. The software
splits the linear levee structure into homogenous spatial seg-
ments, according to changes in the basic indicator values of
each functional criterion. Therefore, each change of a cri-
terion value corresponds to a change from one levee segment
to another.
In agreement with the group of expert engineers respon-
sible for following up this project, a double performance
scale is proposed: (1) a discrete preference scale with five
conditions to formalise and model expert knowledge, and
(2) a continuous preference scale between 0 and 10 related to
the previous discrete preference scale (Figure 3).
Quantitative analyses are then possible for the different
functional criteria values.
A multicriteria method based on a single criterion is
implemented to calculate performance indicators for the
Figure 2 Modelling a levee overflowing failure mode.
4Peyras et al.
© 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd J Flood Risk Management •• (2015) ••–••
homogeneous levee segments in comparison with the differ-
ent levee failure modes considered (Vuillet et al., 2012).
These are calculated by combining aggregation operators
according to rules specific to each levee failure mode. Cj,M is
used to determine the j criterion for levee failure mode M.
The single criterion is estimated through intermediate
aggregations Ak,M, where A is the kth aggregation (k =1, 2, . . .,
n) of the functional criteria connected with a functional
sequence of the levee failure mode M. These intermediate
aggregations are based on mathematical operators. There-
fore, performance indicators for the different levee failure
modes considered can be formulated according to analytical
functions (Figure 4). PIMis the performance indicator for
levee failure mode M.
For each levee failure mode, the expert group will test and
adjust the formulated performance indicators for different
types of levees and related environments.
Functional criteria values are estimated according to the
type of basic indicators. For quantitative basic indicators,
rule-based procedures are defined to move from the basic
indicators to the different functional criteria (Serre et al.,
2008). The procedures can include predefined rules such as
‘IF . . . THEN’ (Serre et al., 2008).
Example: C1,o ‘Levee crest elevation’functional criterion is
scored in terms of the difference between levee crest eleva-
tion and reference flood elevation, according to the rule
presented in Table 1.
For other, more complex, basic indicators (geotechnical,
hydraulic, etc.), it is suggested that each functional criterion
value should be directly determined by expert assessment. To
assist the engineer to do this, the Digsure method provides
them with the information required to estimate functional
criteria values. The engineer is then able to estimate the
functional criteria values with all the available basic indica-
tors according to standardised and uniform scoring.
Example: C6,o ‘Levee Body resistance to Overflowing’ func-
tional criterion is scored according to the basic indicators
presented in Table 2.
The rules and the data set defined to estimate the func-
tional criteria make this approach appropriate for assessing a
set of parameters through expert judgment.
Taking uncertainties on data into account for the
Digsure method
The data used to determine the basic indicators and possibly
complete a levee performance assessment are prone to many
imperfections: uncertainties on the representativeness of a
local investigation, incomplete data due to the absence of
levee monitoring during floods, etc. Such imperfections,
when combined with variable deteriorations occurring in
the levees, can prevent the engineer from providing unique
values for the different functional criteria Cj,M.
Figure 3 Levee performance assessment scale (Vuillet et al.,
2012).
Figure 4 Performance indicator for overflowing levee failure mode.
Table 1 Estimation rule for functional criterion C1,o ‘Levee crest
elevation’
Estimation of C1, o ‘Levee Crest Elevation’
Note
C1,o
ΔZ: difference between levee crest
elevation and reference flood elevation Phenomenon
0ΔZ≤−50 m Overflowing
1–2 −50 cm <ΔZ≤−25 cm
3–4 −25 cm <ΔZ≤0cm
50cm>ΔZ≥25 cm Uncertain
6–7 25 cm >ΔZ≥50 cm No overflowing
8–9 50 cm >ΔZ≥100 cm
10 100 cm >ΔZ
Assessing the performance of river levees 5
J Flood Risk Management •• (2015) ••–•• © 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd
Appraisal by expert assessment based on subjective prob-
ability methods should be adopted (Vuillet et al., 2012).
These methods consist in reviewing data fields relevant to
the basic indicators for each levee segment in order to esti-
mate functional criteria values Cj,M in a probabilistic pattern.
The method proposed consists in considering each func-
tional criterion Cj,M as a random variable by providing two
different parameters for its density function: a modal value
and a dispersion interval. The modal value, which corre-
sponds to the ‘most plausible’ value for the criterion, is esti-
mated in connection with its dispersion interval. The
dispersion interval, which is the interval for the plausible
values, is determined using 5% and 95% quantiles.
The modal value parameter seems to be even more suit-
able for the research herein as it allows the engineer to esti-
mate the most plausible condition for the criterion rather
than estimate probabilities directly.
Determining the dispersion interval requires that the
lower and upper limits of the interval and the quantiles best
suited for this study should be selected carefully. Interval
limits are determined by the criteria evaluation scale for a
continuous interval (0; 10). 5% and 95% quantiles were
chosen because they are the quantiles commonly used by
civil engineers. Furthermore, they are the same as those
approved for characteristic values in semi-probabilistic
methods such as Eurocodes (Peyras et al., 2010). The disper-
sion can be asymmetrical and must reflect the engineer’s
confidence in the modal value estimated (Figure 5).
A number of probability distributions are determined to
estimate how uncertain the performance values are: normal
distribution, log-normal distribution, mirror log-normal
distribution and uniform distribution. These distributions
are known by the engineers and help them to model all the
uncertainty factors for the functional criteria Cj,M used in
levee assessments. Regarding the notation scale applied to the
indicators and functional criteria, the distributions must be
truncated at interval (0; 10). Once they have been modelled
using the Digsure method, the density functions are submit-
ted to the engineer for confirmation or adjustment of the
values assessed for density function parameters (modal value
and 5% and 95% quantiles of functional criteria Cj, M).
An automated procedure is used to match these values
with a probability distribution function based on a normal
distribution, log-normal distribution or mirror log-normal
distribution, according to the potential asymmetry associ-
ated with the dispersion of the distribution.
Table 2 Basic indicators for functional criterion C6,0 ‘Levee body resistance to overflowing’
Criteria Basic Indicators Required Information Sources
C6,o
Levee body resistance
to Overflowing
Flood duration Overflowing duration Hydrological study
Geometry Height and landside slope Visual inspection
Topographical study
Levee body erodibility Nature of levee body material
Particle size
Heterogeneity
Resistance to erosion
Visual inspection
Exploration
Geotechnical study
Geophysical study
Geometric irregularity Nature of landside crest, slope and toe Visual inspection
Topographical study
Figure 5 Log-normal distribution implemented for a modal value 2 and an interval (1, 5) for 5% and 95% quantiles.
6Peyras et al.
© 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd J Flood Risk Management •• (2015) ••–••
A Monte Carlo simulation process is incorporated in our
model to propagate the uncertainties related to the func-
tional criteria Cj,M. Monte Carlo simulations are necessary
because of the size of the sample of input variables and
because the Digsure method includes a large number of
probability distributions, empirical distributions and trun-
cated distributions. This process has been already used by
Sayers et al. (2010) in risk-based asset management tools.
The Digsure method finally provides a probability distri-
bution for the different performance indicators worthy of
consideration. This format is suggested for each levee failure
mode and is subjected to a number of confirmation tests
performed by the expert group on different types of levees.
Applying the Digsure method to an
existing levee
The approach for calculating a performance indicator for a
homogeneous levee segment of an old levee (late 18th
century), located in the French Alps and built as a 400-m
linear structure (Figure 6), is presented in this chapter. For
this example, the approach for assessing the overflowing
levee failure mode for a 1% flood (the design flood for this
levee) through the Digsure method is presented below.
The reference hydraulic features for analysing the levee
segment are as follows:
• Flow for the one per cent per annum event: 387 m3/s;
• Freeboard before crest overflowing, at least 20 cm for the
1% per annum event;
• Maximal stream cross-section: 184 m2;
• Average width of bed: 47 m;
• Water speed for the 1% per annum event: 2.6 m3/s;
• River bed slope: 0.75%.
The general levee features on this homogeneous segment are
the following (Figure 7):
• 2.0 m as maximum levee height;
• Wide levee crest (about 11.7 m) with a road running on
top of it;
• A waterside mortared stone facing in satisfactory
condition;
• Waterside levee toe riprap in satisfactory condition;
• Grassed landside levee slope without specific treatment;
• The levee body mainly comprises loosely compacted
sandy-silty gravel stemming from river deposits, with a
small proportion of fillers and a permeability of 10−3m/s
to 10−5m/s;
• The waterside and the landside levee slopes have gradients
of 1H/1V and 3H/2V, respectively;
• The foundation contains materials that are similar to those
in the levee body.
Estimating the functional criteria
The estimation process for the six functional criteria Cj,o
(j =1–6 and M =o corresponding to the overflowing failure
mode) required the calculation of the performance indicator
PIo, specific to the overflowing levee failure mode for the
homogeneous segment studied and the reference one per
cent per annum event, is described below.
Such functional criteria Cj,o include:
•C
1,o: ‘levee crest elevation’;
•C
2,o: ‘flow obstruction factors’;
Figure 6 Location of levee segment studied.
Figure 7 Theoretical levee cross-section studied.
Assessing the performance of river levees 7
J Flood Risk Management •• (2015) ••–•• © 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd
•C
3,o: ‘presence of low elevation points in crest’;
•C
4,o: ‘operation of spillway(s)’;
•C
5,o: ‘resistance to overflowing of crest protection, landside
levee slope protection and waterside levee toe’;
•C
6,o: ‘levee body resistance to overflowing’.
Assessing criterion C1,o ‘Levee crest elevation’:
The modal value for the ‘levee crest elevation’ functional
criterion is estimated based on an expert rule (Table 1) and
the freeboard value before crest overflow on this levee
segment for the 1% per annum event. Given this 20-cm
freeboard, the criterion is deemed poor (see Figure 3), with a
value of 5.
Regarding the segment crest elevation variations and due
to the uncertainties on the longitudinal profile of the water
elevation for the 1% per annum event, a (4; 6) interval
ranging from poor to acceptable (Figure 3) is then selected
by the experts in order to define limits for the modal value 5.
Therefore, this criterion evaluation is modelled following a
normal probability distribution (Figure 8).
Assessing criterion C2,o ‘Flow obstruction factors’:
Although a bridge standing on the landside of the
segment is a potential obstacle that should be taken into
account, the risk of interfering with the flow due to bed
obstruction remains very low and almost insignificant on
this segment when taking into account its straight profile,
the average bed width (47 m) and the 30 m bed width near
the landside bridge. In these circumstances, the experts
calculated a modal value of 9 for the functional criterion
‘flow obstruction factors’ and an associated interval of
(7; 10), ranging from acceptable to good (see Figure 3).
Therefore, this criterion estimation can be modelled
following a mirror log-normal probability distribution
(Figure 9).
Assessing criterion C3,o ‘Presence of low elevation points in
crest’:
The low points to be considered can be very localised.
They were identified on the ground and not from the avail-
able low resolution topographic survey.
Figure 8 Normal distribution adjusted to the functional criterion ‘Levee crest elevation’.
Figure 9 Mirror log-normal distribution adjusted to the functional criterion ‘Flow obstruction factor’.
8Peyras et al.
© 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd J Flood Risk Management •• (2015) ••–••
The levee crest displays a regular geometry and included
no isolated low point. The bridge lying downstream was the
only isolated irregularity of the segment studied but did not
present low point either. Because no low point was identi-
fied, the experts assessed the ‘presence of low elevation
points in the crest’ criterion as ranging from acceptable to
good (see Figure 3), with a value included in the (7; 10)
interval, with 8 as a modal value rated as good. Therefore,
this criterion evaluation is modelled following a mirror log-
normal probability distribution (Figure 10).
Assessing criterion C4,o ‘Operation of spillway(s)’:
There is no spillway near the considered homogeneous
levee segment. Therefore, the ‘operation of spillway(s)’ cri-
terion was not assessed.
Assessing criterion C5,o ‘Resistance to overflowing of crest
protection, landside levee slope protection and waterside levee
toe’:
The different basic indicators studied in order to assess the
criterion ‘resistance to overflowing of crest protection,
landside levee slope protection and waterside levee toe’
include: ‘flood duration’, ‘crest erodibility’, ‘erodibility of
landside levee slope materials’, ‘erodibility of landside levee
toe’ and ‘geometrical irregularities’.
Given the coarse and largely non-erodible materials con-
tained in the levee crest, the gradient of 2/3 of the grassed
landside levee slope, the wide structure (11.7 m) and the
short duration of reference floods, the criterion is rated as
ranging from acceptable to good (see Figure 3), within an
interval of (6; 9) with 8 as the most plausible modal value.
Therefore, this criterion evaluation is modelled following a
mirror log-normal probability distribution (Figure 11).
Assessing criterion C6,o ‘Levee body resistance to
overflowing’:
The course materials contained in the earth fill levee body
(sandy-silty gravel) which were loosely compacted when the
structure was built, were used by the experts as reference
parameters to assess the ‘levee body resistance to overflow-
ing’ criterion. In view of these features in the levee segment,
Figure 10 Mirror log-normal distribution adjusted to the functional criterion ‘Presence of low elevation points in crest’.
Figure 11 Mirror log-normal distribution adjusted to the functional criterion ‘resistance to overflowing of crest protection, landside
levee slope protection and waterside levee toe’.
Assessing the performance of river levees 9
J Flood Risk Management •• (2015) ••–•• © 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd
this criterion was rated as ranging from poor to acceptable
(see Figure 3), within the interval (4; 7) with the most plau-
sible modal value rated as poor (5). Therefore, this criterion
evaluation is modelled following a log-normal probability
distribution (Figure 12).
Propagation of uncertainties and use of results
Functional criteria uncertainties are propagated using
Monte Carlo simulations based on the deterministic model
of levee behaviour and the aggregation laws developed. The
simulations provide an empirical distribution of the perfor-
mance indicator IPofor the overflowing levee failure mode
considered (Figure 13).
The empirical distribution helps determine the modal
value that represents the central trend of the performance
indicator, and the 5% and 95% fractiles, which represent the
uncertainty on the indicator IPo(Figure 14).
This is carried out all along the levee by analysing homo-
geneous segments of levee according to the data used by the
Digsure method. Thus, the most plausible value for the per-
formance indicator related to a specific levee failure mode
can be seen along the entire linear levee structure, as can the
uncertainty on the identification of the performance indica-
tor. In Figure 15, segment 1 is the homogeneous levee
segment presented above. Segments 2, 3 and 4 refer to other
parts of the linear levee structure studied, and which are not
described here.
Developing the Digsure GIS tool for
levee assessment and management
The Digsure method is designed to assess the performance of
levees. A GIS tool was selected to build the Digsure tool
structure in order to use only one tool for collecting, analys-
ing and providing a geographical representation of data and
the levee assessment results.
The Digsure tool is designed to satisfy the different man-
agement requirements:
• Data review (data on levees and other structures in the
protected area);
• Data input by agents responsible for routine inspections
and inspections during floods;
• Estimation of levee functional criteria by expert engineers;
• Estimation of performance indicators of the different
homogeneous levee segments using the Digsure method.
Levee performance assessment
The Digsure tool contains a spatially referenced database
gathering available information on basic indicators and the
probabilistic model for calculating the performance indica-
tors. The software process relies on the following steps:
• Collecting and referencing data about basic indicators;
• Splitting the linear structure into homogeneous levee seg-
ments according to the basic indicators for each functional
criterion;
• Applying the probabilistic procedure for estimating the
functional criteria, as developed by the engineers;
• Splitting the linear structure into homogeneous segments
according to the functional criteria values, for each levee
failure mode;
• Calculating the performance of each homogeneous levee
segment.
The first step consists in collecting and referencing the
available data on levees in the GIS database. These data
include all the known features of the levees as well as the
results from regular inspection visits and specific studies
carried out on these structures. They are referenced and
geo-located in the tool as basic levee indicators estimated by
expert assessment according to the information contained in
the data. Figure 16 shows how the estimated values for the
different basic indicators specific to the‘resistance to external
erosion of landside levee slope’ criterion are entered in the
Figure 12 Log-normal distribution adjusted to the functional criterion ‘Levee body resistance to overflowing’.
10 Peyras et al.
© 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd J Flood Risk Management •• (2015) ••–••
Figure 13 Propagation of functional criteria uncertainties using Monte Carlo simulations for the overflowing levee failure mode performance indicator IPo.
Assessing the performance of river levees 11
J Flood Risk Management •• (2015) ••–•• © 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd
Digsure tool: ‘erosive flow obstructions’; ‘local water speed’;
‘morphodynamic context’; ‘type of protection’.
The next step consists in completing a spatial decompo-
sition of the linear levee structure into homogeneous seg-
ments. This is done for each functional criterion according
to the variable values of their specific basic indicators. There-
fore, each functional criterion considered will result in a
different specific decomposition of the linear levee structure.
The probabilistic estimation procedure for the functional
criteria values is based on the values related to the specific
basic indicators for each homogeneous levee segment.
Figure 16 presents an example of how the linear structure
can be decomposed and shows the results from the proba-
bilistic calculation of values for the criterion ‘resistance to
external erosion of landside levee slope’ according to the
values of all four specific basic indicators relevant for that
levee segment. Indeed, the ‘Start’ and ‘End’ columns contain
the limits identified (in meters on the linear levee structure)
for the basic indicator values, and the resulting homo-
geneous segments for the relevant criterion.
Then the GIS software decomposes the linear levee struc-
ture into homogeneous segments for each levee failure mode
according to the values of their relevant functional criteria.
The single criterion-based assessment model automatically
provides the performance indicator for each levee failure
mode. The GIS tool automatically identifies the homo-
geneous segments according to the type of levee perfor-
mance considered. The tool performs propagations of
uncertainties for the functional criteria used for probability
distributions. A resulting probability distribution is then
suggested for each performance indicator. The tool will
display the modal value as well as the 5% and 95% distribu-
tion quantiles for the performance indicators along the
linear levee structure (Figure 17).
Figure 14 Empirical distribution of the performance indicator for overflowing levee failure mode IPo.
Figure 15 Performance indicator IPomodal value and 5–95% fractiles along the linear levee structure studied.
12 Peyras et al.
© 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd J Flood Risk Management •• (2015) ••–••
The engineer then enters the procedure to appraise the
functional criteria, and control the assessment process and
the results obtained by the GIS software. The levee perfor-
mance assessment model proves very useful for research unit
engineers and levee managers. In addition, the semi-
automatic levee assessment process ensures significantly
streamlined tasks for the expert engineer. Using subjective
probabilities makes it possible to include uncertainty ranges
for performance indicators. They can be represented by
matching the central trend as the modal value with the 5%
and 95% distribution quantiles as the related uncertainty
ranges (Figure 17).
Along the entire linear levee structure the Digsure tool
displays the most plausible value for the performance indi-
cator related to a specific levee failure mode and the uncer-
tainty on this performance indicator value (Figure 17).
Once the Digsure tool has been set up and adjusted to the
local context by engineers for each routine inspection,
during floods or post-flood, the engineer can edit or re-enter
information on the specific functional criteria and update
the performance indicator calculation. Indeed, the user can
access continuously updated information on each segment,
on any changes occurring in the performance of their struc-
tures and their related uncertainties.
The automatic sorting of segments into different perfor-
mance categories in addition to the consideration of flood
consequences for the leveed areas studied leads to a risk
estimation that will help levee managers to continuously
optimise their actions and improve responsiveness by:
• Prioritising the investigations required to improve infor-
mation quality and reduce uncertainties related to perfor-
mance indicators;
• Prioritising and preparing rehabilitation actions to
improve the performance of structures and reduce risks.
Conclusion
The purpose of the Digsure project was to develop levee
performance indicators and incorporate them in an opera-
tional GIS levee assessment tool designed for research units
and levee management technical services.
The Digsure method includes a functional levee behaviour
model for all levee failure modes: overflowing, internal
erosion, scouring, landside sliding and waterside sliding.
Also, for each levee failure mode, the method includes func-
tional criteria used to assess the levees and their related basic
indicators. On that basis, levee performance indicators were
developed for each levee failure mode. Based on GIS
Figure 16 Digsure GIS tool: spatial decomposition of linear levee structure and calculated values for the functional criterion ‘Resistance
to external erosion of landside levee slope’.
Assessing the performance of river levees 13
J Flood Risk Management •• (2015) ••–•• © 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd
software, the Digsure tool provides values for performance
indicators along the linear levee structure.
The Digsure method also includes a probabilistic
approach, taking into account all the uncertainties on the
functional criteria used in the levee performance assessment.
Therefore, performance indicators are indicated as probabil-
ity distributions with uncertainty ranges.
The Digsure tool nested in the GIS module will help engi-
neers and managers to identify levee segments for which
further investigations are needed to reduce uncertainties and
to prioritise and carry out rehabilitation actions for those
levee segments identified as having poor quality and/or high
probabilities of failure.
Digsure GIS is a proprietary tool owned by the different
project partners: Irstea, Société du Canal de Provence and
G2C. These partners use the Digsure tool in commercial
engineering activities performed for different levee
managers.
To ensure completeness, it should be noted that the
Digsure project also includes research on vulnerable leveed
areas, a topic not addressed herein. This approach has led to
the development of vulnerability indicators for leveed areas.
Indeed, three groups of vulnerability indicators have been
defined: economic, ecological and social vulnerability. Those
readers who are interested can refer to the following publi-
cation (Allouche et al., 2013).
Acknowledgements
The authors wish to address their warm thanks to the
Conseil Régional Provence Alpes Cote d’Azur, which
co-funded the work presented herein, and the project
partner SYMADREM, which provided data and validated
the method.
References
Allouche A., Nicolas L. & Tourment R. Approche socio-
anthropologique pour l’évaluation de la vulnérabilité sociale
des zones protégées par les digues fluviales du Rhône aval.
Houille Blanche-Revue Internationale de l’eau 2013, 2013,
(n°4), 57–63.
ASCE (2007) The New Orleans hurricane protection system: what
went wrong and why. Reston, VA: ASCE Hurricane Katrina
External Review Panel. ISBN: 978-0-78440-893-3
Ciria, MEDE (Ministère de l’Ecologie du Développement
durable et de l’Energie) & USACE (US Army Corps of Engi-
neers). The International Levee Handbook. London: CIRIA,
2013.
FloodProbe (2012) Combining Information for Urban Levee
Assessment, Report Number: WP03-01-12-24. Available from:
http://www.floodprobe.eu
Figure 17 Digsure GIS tool: overflowing performance indicator assessment result.
14 Peyras et al.
© 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd J Flood Risk Management •• (2015) ••–••
Gouldby B., Sayers P., Mulet-Marti J., Hassan M. & Benwell D.
A methodology for regional-scale flood risk assessment.
Water Management 2008, 161, (3), 169–182.
Knoeff J., Vastenburg E.W., Van den Ham G.A. &
Lopez de la Cruz J. (2011) Automated flood risk manage-
ment. Proceedings of the 5th International Conference on
Flood Management (ICFM5). Japan.
Kolen B., Slomp R. & Jonkman S.N. The impacts of storm
Xynthia February 27–28, 2010 in France: lessons for flood
risk management. J Flood Risk Manag 2013, 6, 261–278. doi:
10.1111/jfr3.12011.
Maurel P., Serre D. & Tourment R. Towards a generic GIS for
Dike management in flood plain areas: from conceptual
design to real applications. In: AGILE 2004 Conf on Geo-
graphic Information Science. Heraklion: Crete University Press
(ed.), 2004.
Mériaux P. & Royet P. (2007) Surveillance, Maintenance and
Diagnosis of Flood Protection Dikes – A Practical Handbook for
Owners and Operators. Quae.
Modarres M. What every engineer should know about reliability
and risk analysis. New York: Marcel Dekker Publisher, Inc.,
1993.
Paquier A. Failure of a dike in a flood environment (Agly 1999).
In: J.-M. Tanguy, ed. Practical applications in engineering, Vol.
4. Hoboken, NJ: John Wiley & Sons, Inc., 2010. doi: 10.1002/
9781118557792.ch10.
Peyras L., Royet P. & Boissier D. Dam ageing diagnosis and risk
analysis: development of methods to support expert judg-
ment. Can Geotechnical J 2006, 43, (2006), 169–186.
Peyras L., Merckle S., Royet P., Bacconnet C. & Ducroux A.
Study on a semi-probabilistic method for embankment
hydraulic works – Application to sliding mechanism. Euro-
pean J Environ Health and Civ Eng 2010, 14, (n°5), 669–691.
Sayers P.B., Wallis M., Simm J.D., Baxter G. & Andryszewski T.
(2010) Towards the next generation of risk-based asset man-
agement tools. In: G. Pender, et al.ed.Flood Risk Science and
Management. Oxford, UK: Blackwell Publishing Ltd, 2010.
Serre D., Peyras L., Curt C., Boissier D. & Diab Y. Evaluation of
hydraulic works in civil genius. Can Geotechnical J 2007, 44,
(2007), 1298–1313.
Serre D., Peyras L., Tourment R. & Diab Y. Levee performance
assessment methods integrated in a GIS to support planning
maintenance actions. J Infrastructure Syst – ASCE 2008, 14,
(n°3), 201–213.
Simm J.D., Wallis M., Smith P., Deniaud Y., Tourment R.,
Veylon G., Durand E., McVicker J., Hersh-Burdick R. &
Glerum J. The significance of failure modes in the design and
management of levees–aperspectivefromtheInternational
Levee Handbook team. In: Frans Klijn & Timo
Schweckendiek, eds. FloodRisk 2012 – 2nd European Confer-
ence on FloodriskManagement. London: Taylor & Francis
Group, 2012. Rotterdam, Netherlands, 20th–22nd November
2012. ISBN 978-0-415-62144-1.
Tourment R., Peyras L., Vuillet M., De Massiac J.-C., Allouche
A., Nicolas L., Casteigts C. & Delaunay C. Digsure method:
decision support indicators and GIS tool for levees manage-
ment. In: Frans Klijn & Timo Schweckendiek, eds. FloodRisk
2012 – 2nd European Conference on FloodriskManagement.
London: Taylor & Francis Group, 2012. Rotterdam, Nether-
lands, 20th–22nd November 2012. ISBN 978-0-415-62144-1.
Tourment R., Van Hemert H., Neutz C., Wallis M., Beullac B.,
Bradner G., Raschke S., Maurel P., Mériaux P., Peyras P. &
Royet P. Chapter 5 levee inspection, assessment and risk
attribution. In: Frans Klijn, ed. The International Levee Hand-
book. London: CIRIA, 2013, pp. 299–394.
Van der Meij R., Van den Ham G.A., Morris M., Lhomme J.,
Tourment R. & Maurel P. Combining different data sources
for assessment of urban flood defences. In: Frans Klijn &
Timo Schweckendiek, eds. 2nd European Conference on
Floodrisk Management. London: Taylor & Francis Group,
2012. Rotterdam Netherlands, 20th–22nd November 2012.
ISBN 978-0-415-62144-1.
Vuillet M., Peyras L., Serre D. & Diab Y. Decision making
method for assessing performance of large levee alignment.
J Decision Syst 2012, 21, (2), 137–160. doi: 10.1080/
12460125.2012.680354.
Vuillet M., Peyras L., Carvajal C., Serre D. & Diab Y. Levee per-
formance evaluation based on subjective probabilities. Euro-
pean J Environ Health and Civ Eng 2013, 17, (5), 329–349.
doi: 10.1080/19648189.2013.785723.
Assessing the performance of river levees 15
J Flood Risk Management •• (2015) ••–•• © 2015 The Chartered Institution of Water and Environmental Management (CIWEM) and John Wiley & Sons Ltd
