See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/237697191

Risk‐based floodplain management: A case study from Greece

Article in International Journal of River Basin Management · March 2003

DOI: 10.1080/15715124.2003.9635191

CITATIONS READS

75 371

1 author:

Jacques Ganoulis
Aristotle University of Thessaloniki UNESCO Chair/INWEB
135 PUBLICATIONS 1,237 CITATIONS

SEE PROFILE

All content following this page was uploaded by Jacques Ganoulis on 03 February 2015.

The user has requested enhancement of the downloaded file.

 Intl. J. River Basin Management Vol. 1, No. 1 (2003), pp. 41–47
© 2003 IAHR & INBO

Risk-based floodplain management: A case study from Greece

J. GANOULIS, Professor of Hydraulics, Department of Civil Engineering, Aristotle University of Thessaloniki, 54 124 Thessaloniki,
Greece

ABSTRACT
The use of engineering risk analysis and multi-objective decision-making under risk are considered as tools for floodplain management and extreme
flood protection. Distinction is made between (a) the catchment or large scale planning and (b) the local or small-scale design of protection measures.
After defining the risk of flooding at different scales and specifying the multiple criteria that must be considered when choosing between alternative
measures of floodplain protection and management, the methodology used is illustrated in a case study from Greece (Giofyros Basin, Crete Island),
where a devastating flash flood occurred on 13th January 1994. Possible remedial structural and non-structural solutions are analysed in order protect
the inhabited area and important public buildings from future extreme floods.

Keywords: Flood control; detention basins; risk analysis; multi-objective decision-making; floodplain management; numerical
simulation.

1 Introduction 3. In karst areas, which make up more than half of the Mediter-
ranean drainage basin, flash floods are more acute and much
Extreme floods are essentially natural hazards that occur infre- more violent [2,9–11]. Excessive flooding occurs in these
quently. In most cases excessive precipitation is the main cause areas after the karstic cavities are filled by a huge amount
of catastrophic floods. However, anthropogenic factors, such of rainfall water.
as human occupation of floodways, extensive urbanisation and 4. Heavy concentrations of population in urban and residen-
structural measures to mitigate floods (e.g. flood levees and walls, tial areas around the centres of historic cities have, in many
cutting of river meanders and river training) have modified the cases, resulted in the occupation of the beds and floodways
natural characteristics of extreme floods [6,13]. Recent catas- of ephemeral streams. This phenomenon has been recorded
trophic floods both in Europe and the USA (Rhine River, 1995; mainly near the coastal areas, where tourist activity has
Elbe River, 2002; Mississippi River, 1993) have shown that dramatically increased in recent years. As the existing infras-
human activity and traditional river engineering works may result tructure in sewer systems is inadequate and its completion is
in an increase in the frequency of extreme floods and the water very expensive, great volumes of storm water cannot be evac-
stage with serious negative economic consequences such as loss uated after heavy rains. As a consequence, the lower areas
of or damage to property as well as danger to or loss of human of cities become flooded and serious damage to public and
life. private property occurs [9].
In the Mediterranean area, flooding conditions are unique,
In view of the limited economic means of local authorities, the
given the influence of a semi-arid climate, geological character-
implementation of traditional engineering measures to prevent
istics and the socio-economic environment. The main character-
floods, such as the building of dams and drainage tunnels, is very
istics of floods in the Mediterranean basin may be described as
expensive. In populated areas, extension of the existing storm
follows:
sewer system is not easy, due to the high cost of replacing the
1. The presence of heavy rainfall in autumn and winter may pro- existing sewers and the impact of engineering works on urban
duce flash floods in catchments and streams, which remain activities such as trade, tourism and traffic. A risk-based design
dry throughout much of the year. These flash floods are of of alternative measures may be appropriate to reduce costs and
short duration (from a few minutes to a few hours) and have to improve the reliability of the design [8].
high flood peaks (many hundreds of m3 /s). Floodplain management and flood control involve alternative
2. During flash floods, soil erosion and sediment transport are measures (structural: levees, dikes, retention basins, channel
important and may lead to the failure of flood-defensive modifications, or non-structural: flood warning, land uses),
engineering structures (reservoirs, spillways, gates). different natural conditions (type of climate, socio-economic

Received and accepted on Jan 10, 2003. Open for discussion until May 31, 2003.

41
42 J. Ganoulis

environment) and various_preferences (economic, environmen- of Heraklion to the Aegean Sea. In terms of its catchment area,
tal, aesthetics, etc.). For the management of risks related to the Giofyros is one of the biggest streams of this Mediterranean
floods, various hydrological, economic and environmental uncer- island, although it has a constant flow only during the humid part
tainties should be assessed and quantified. The US Army Corps of the year (i.e. autumn and winter). The main geomorphological
Engineers’ flood reduction structures use traditional principles characteristics of the catchment are:
of risk-based design, in that they only consider hydrological
Total area: 189 km2
risk, which maximizes the net economic benefits from the project
Max hydraulic route: 31 km
under various uncertainties [3].
In this paper, the general principles of engineering risk anal- Max altitude: 1000 m
ysis are used to develop a multi-objective risk-based approach Mean altitude: 353 m
to floodplain management. The various steps taken in a compre- Mean slope: 0.22
hensive application of engineering risk analysis to flood control The soil is mainly alluvial and contains a relatively high per-
are: (1) identification of hazards, (2) risk quantification, (3) con- centage of clay, and some areas of rock. The area is constantly
sequences of risk, and (4) risk management. This multi-objective cultivated and covered mainly by vineyards and olive trees, with
risk-based floodplain management approach is illustrated in the some forests. The climate is typically Mediterranean, with hot,
case of the Giofyros Basin near the city of Heraklion, on the dry summers and mild winters. Rainfall is quite considerable
island of Crete. Hydro-meteorological data for this area, as well during the winter period (from October to March) and the mean,
as data from the 1994 flood, are available. maximum and minimum annual precipitations are:
Mean annual precipitation: 827 mm
2 Impact of the 1994 flood Max. annual precipitation: 1217 mm
Min. annual precipitation: 469 mm
2.1 Geological and hydrological characteristics
Data from the Aghia Varvara hydrometeorological sta-
As shown in Figure 1, the hydrologic basin of the Giofyros tion from 1954–1994 show that the monthly precipitation in
(189 km2 ) lies in the northern part of the island of Crete. The December and January has exceeded 550 mm/month at least once
Giofyros stream outfalls through the western suburbs of the city during the last 40 years. Many rainfall gauge stations, and some

N
100 Km

Bulgaria
Yugoslavia
Alb
an
ia

Turk
Ae

ey
ge
an
se
a
Io
ni
on
se
a

AEGEAN SEA

HERAKLION

GIOFYROS
STREAM

Profitis Ilias
Station

–2 0 2 4 6 Kilometers

Ag. Varvara
Station

Figure 1 Location of the Giofyros catchment and hydrometeorological stations.

 Risk-based floodplain management: A case study from Greece 43

full meteorological stations are located in the watershed and in which was still under construction at the time. Many of the plant’s
the neighbouring basin. The data from these stations are not eas- reservoirs, made of concrete, were rendered unserviceable or
ily and fully exploitable due to errors or missing periods. The completely destroyed by the force of the incoming water.
best and most reliable rainfall data and rainfall intensity data are
available for a period of less than a decade from the official mete-
orological station installed at Heraklion airport, which is out of 3 Risk-based floodplain management
the catchment area on the coast. The analysis of these data showed
a good representation of conditions for the catchment area. By definition, floodplain management is an integrated consider-
ation of all structural (engineering) and non-structural (admin-
istrative) measures to minimize losses due to flooding on the
2.2 The 1994 flash flood catchment scale. Selection among alternative measures to prevent
On 13th January 1994, a devastating flash flood occurred in the floods may be made on different scales. It is useful to distinguish
Giofyros basin. The extreme flood resulted in a series of events, between
which may be summarised as follows: (1) catchment scale planning: a large scale or regional scale
(a) Heavy rainfall. The total rainfall recorded on the day “optimal” selection between various alternative measures,
of the flood was about 185 mm, which is equal to about half (2) local scale design: a small-scale area (sub-basin) design
of the mean annual precipitation in the region of Heraklion. of hydraulic structures.
A maximum rainfall intensity of 37 mm/h was recorded at the On both scales, the risk of flooding is traditionally related to
hydro-meteorological station of Aghia Varvara (Figure 2). In 6 h, hydrological uncertainties (hydrological risk). If the engineer-
which is about the retention time for the Giofyros basin, a total ing risk is defined as the probability of failure [7,8] then on the
rainfall of 143 mm was recorded. catchment scale (regional scale) we have:
(b) Soil conditions. Rainfall of a light intensity had persisted risk of flooding = P(Q > QT ) (1)
several days before the critical storm of 13th January 1994. The
soil was almost completely saturated and runoff was high during where P(.) is the probability, Q is the actual flood in the catchment
the critical storm. area, and QT is the T-year flood.
(c) Other. Deforestation and the removal of several hectares of On the local scale of a hydraulic structure, the risk may be
vineyards during the months preceding the storm probably also defined as the probability of overtopping. As shown in the case
influenced the increased intensity of the flood. of a simple flood levee (Figure 3), failure occurs when h0 +z > H.
Many houses located downstream, near the coast, were For this case the flooding risk can be expressed as:
flooded and material damage was evaluated at several hundreds
risk of failure = P(h0 + z > H) (2)
of thousands of Euros. The most important effect of the flood
was the damage caused to the city’s wastewater treatment plant, where h0 is the mean water level, z is the surrevelation for a given
flood, and H is the height of the levee.
40 Traditional risk-based design (US Corps of Engineers) incor-
porates uncertainty analysis under risk into an optimisation
35 framework [3]. The objective is to select the optimal risk-based
design that maximizes the net economic benefits.
If CD (x) is the expected annual damage cost due to flooding
30
failure, CI (x) the annual installation cost and x a vector of decision
variables relating to structural sizes, then the optimal risk-based
25
design may be expressed as
i(mm/h)

20 min{CI (x) + CD (x)} (3)

x

under some design specifications g(x) = 0.

15

10
h
z
5

H
0 h0 : normal
5 15 25 water level
0 10 20
Time t(h)
Figure 2 Relationship between rainfall intensity i (mm/h) and duration
t (h) between 13–14 January 1994 (Ag. Varvara Station). Figure 3 The flood levee problem.
44 J. Ganoulis

COSTS (US $) In selecting the most appropriate method, important criteria are
2 × 105 the kind of objectives (quantitative or qualitative), the num-
ber of decision-makers (one or a group) and whether objectives
1.5 × 105
are involved a priori, a posteriori or interactively. ELECTRE I
Optimum
CT to III techniques are more suitable for qualitatively expressed
1 × 105
criteria [1]. Game and team theories [4] are mainly interac-
5 × 104 tive techniques. Uncertainties and risk may be quantified by
Cp
CD using probabilities or fuzzy sets, and can be handled better by
0 compromise programming techniques.
0 1.39 2 4 6 8 10
Multi-criteria decision-making analysis is actually under
–ln pF = safety
investigation for the catchment scale planning of flood defensive
Figure 4 Total cost-reliability relationship for a flood levee. measures in the Giofyros stream.

The design level x is the structural size H for the flood levee,
which may be related to the hydrological risk PF or the hydro- 4 Application to the Giofyros basin
logical reliability (ln pF ) [10]. The result of minimizing the
expression (3) for a special case [10] is shown in Figure 4. 4.1 Large scale planning
This approach has only one objective: the total cost or the total
A distinction should be made between: (a) the downstream plain
net benefit of the project, in other words maximized (benefit) or
area of the Giofyros stream and (b) the upstream catchment area.
minimized (cost) as a function of the flooding risk. The proce-
The downstream plain area represents about 20% of the total area
dure is suitable mainly for small-scale design (e.g. sizing a flood
of the basin and has a mean hydraulic length of 11 km. This is
levee or a hydraulic structure), where a trade-off between costs
about 1/3 of the mean hydraulic length of the basin ( 30 km). Apart
(or benefits) and risk (reliability) may be obtained through the
from some minor hydraulic works in the plain area, no other struc-
optimisation procedure.
tural measures (such as reservoirs, regulation structures, etc.)
On the catchment or regional scale planning process, a multi-
have yet been implemented in the entire basin.
objective approach to flood control alternatives is recommended
The hydraulic risk of flooding was first evaluated for the
[1]. The main objectives or criteria to be taken into considera-
entire catchment area. Because no gauged data are available for
tion are
flow rates, the discharge-frequency relationship was estimated by
1. Economic objectives: Costs and benefits such as project
analysing the maximum rainfall-frequency data. Then, a rainfall-
cost, operation and maintenance costs, external costs, reduction
runoff model, such as the HEC-1, was used to estimate runoff.
of flood damage benefits, land enhancement, indirect benefits;
The relationship between the maximum rainfall height (mm)
2. Environmental objectives: These may be positive or neg-
and the return period T(yr) is shown in Figure 5 for a 2-h rainfall
ative environmental impacts, such as increase or decrease in
duration. In order to evaluate different uncertainties that influence
the number of species, flora and fauna modifications, losses of
the extrapolation results over 50- and 100-year return periods,
wetlands, landscape modification; and
three different methods were applied [9] including: (a) fitting
3. Social objectives: Risk of extreme flooding, duration of
a Gumbel distribution, (b) fitting the data, and (c) fitting the
construction, employment increase, impacts on transportation.
A and B coefficients. These coefficients appear in the following
After the definition of the objectives, the steps to be undertaken
relationship:
for the multi-objective planning of flood control alternatives are
the following [8]: h(t, T) = A(T)t 1−B(T) (4)
1. Define a set of alternative actions, which include structural
h (mm)
and non-structural measures of flood protection;
200.00
2. Evaluate the outcome or impact matrix, i.e. assign rates to each
specific objective, corresponding to each particular action;
and 150.00 BEST FIT OF
DATA
3. Rank the alternative actions, using an appropriate multi- A &B

objective analysis technique.

100.00
Different techniques are available for multi-criteria decision-
making [5,12,14] and recently, distance-based techniques have 50.00
GUMBEL

been most developed, such as the following:

– ELECTRE I to III 0.00
– Compromise programming 1 10 100
– Goal programming RETURN PERIOD T (yr)
– Sequential multi-objective optimisation Figure 5 Maximum rainfall height (mm) versus the return period T(yr)
– Game theory. for 2-hour rainfall duration.
 Risk-based floodplain management: A case study from Greece 45

Table 1 Estimated maximum rainfall height 4.2 Local scale flood protection
hmax (mm) and peak flood discharge Qmax (m3 /s) for
return period T = 30, 50 and 100 years. Local authorities expressed their desire for an urgent undertaking
of the necessary flood protection measures for the city’s wastewa-
T h (max) (mm) Q max (m3 /s) ter treatment plant. The issue was to determine the size of the flood
30 125 450 levees around the sewage treatment facility in order to protect
50 152 580 important civil and mechanical equipment from future extreme
100 193 900 floods. Emphasis was placed on safety rather than cost, because
of the importance of the plant and the relatively small volume of
the levees.
For the design of the flood levees on the local scale, a two-
where t is the rainfall duration (min), h the rainfall height (mm) dimensional mathematical model was used to propagate the flood
and T the return period (yr). hydrograph. Different hydrographs representing the historical
The maximum rainfall height and the peak flood discharge flood (13 January 1994) and the T = 30-, 50-, 100-year return
corresponding to T = 30, 50, and 100 years are summarized in periods were simulated. The mathematical model consists of the
Table 1. following mass continuity and Saint–Venant equations:
Alternative measures for floodplain protection are combina-
tions of three different approaches: ∂h ∂qx ∂qy
+ + =0 (5)
∂t ∂x ∂y
1. Regulation of the downstream cross-section of the Giofyros

  
stream in order to increase the hydraulic capacity. Due to ∂qx ∂ qx2 ∂  qx qy  ∂h
some constraints (existing bridges) the maximum hydraulic + + = −gh − (Ifx − I0x ) (6)
∂t ∂x h ∂y h ∂x
capacity can reach the 20-year flood (Q ∼ = 300 m3 /s). Envi-
   
ronmentally sound regulation may avoid any concrete scaling ∂qy ∂  qx qy  ∂ qy2 ∂h
and stream training: regulation should be based on the enlarge- + + = −gh − (Ify − I0y ) (7)
∂t ∂x h ∂y h ∂y
ment of the cross-section, use of natural materials for fixing
the bed and earthen flood levees, and should be well-integrated where h is the flood stage in m, qx , qy the flow rates per unit width
into the landscape. in m3 /s/m, I0x , I0y the bed slopes, and Ifx , Ify the friction slopes.
2. Design and construction of a multi-purpose reservoir to retain The Manning formula was used to compute Ifx and Ify as func-
a substantial volume of the critical flood. Two reservoirs of tions of qx , qy and h. Numerical integration of the above equations
different capacities were proposed: was performed over a two-dimensional grid using finite differ-
2.1 A 28 × 106 m3 reservoir to be realized by an earthen dam ences. A 100 m grid size was selected. The model was validated
of about 70 m in height by comparing the numerical results with data available from the
2.2 A smaller reservoir with a total capacity of 15 × 106 m3 historical flood of 13th January 1994. On that day, the maximum
It should be noted that the net annual water balance for the water levels at different locations inside the wastewater treatment
catchment is estimated at 20×106 m3 , although the maximum plant were recorded.
volume of a 50-year flood is about 5 × 106 m3 . Results of the numerical simulation indicating the contour
3. Use of a storm water detention basin network distributed over lines of the water stage during the 1994 flood are shown in
the catchment. The principal function of the system should Figure 6. For the same flood, water stage hydrographs computed
be to reduce the peaks of the flood hydrographs. At the same at characteristic locations are shown in Figure 7. After defining
time, significant volumes of water may be retained locally for the size of the flood levees around the wastewater plant, results
agricultural purposes. of the simulation of the T = 100-year flood are shown in Figure
Design of the detention basin system (i.e. size and site 8. It can be seen that the space where the wastewater treatment
of flood detention reservoirs) should be adequate to sustain plant is located is well protected from this extreme flood. Further
floods of T = 30-, 50-, or 100-year return period. protection of the local area will be provided after implementation
By combining the above three structural solutions, the follow- of the flood detention basin network in the upstream catchment
ing alternatives are currently under investigation: area as described in Section 4.1.

1. Regulation of the downstream level of the stream (R) and

construction of a large capacity reservoir (LR)
5 Conclusions
2. (R) + Construction of a small capacity reservoir (SR)
3. (R) + Detention Basin network of T = 30-yr floods (DB30)
Special attention should be paid to the floodplain management
4. (R) + Detention Basin network of T = 50-yr floods (DB50)
measures in areas with semi-arid climates. In these areas, flash
5. (R) + Detention Basin network of T = 100-yr floods (DB100)
floods in ephemeral streams can be violent and unpredictable.
The main objectives for ranking the above 5 alternatives are: Risk-based design methodologies for protection measures may
(a) costs and benefits, (b) risk of failure, (c) environmental impact, result in trade-offs between risk and costs, as well as having
and (d) social effects. environmental and social impact.
46 J. Ganoulis

1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 13

65 65

55 55

45 45

X 100m
35 35
X 100m

25 25

15 15

5 5

1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 13
X 100m X 100m

Figure 6 Contour lines of water stage for the 1994 flood (no flood levees Figure 8 Contour lines of water stage for the T = 100-year flood, after
around the wastewater treatment plant). construction of the flood levees around the wastewater treatment plant.

4.00 Distinction is made between catchment scale planning and

points on line 31
2 local scale protection from floods. On the former scale, a multi-
3
4 criteria decision making approach to areas under risk may help
3.00 5
6 in selecting between different alternatives. In areas without too
7
many constraints (e.g. high population or intensive agriculture)
a storm water detention basin system distributed over the entire
h(m)

2.00
catchment area seems to be the most appropriate. On a local
scale, reliability of the protection measures may be based on
1.00 more traditional techniques involving hydrological and hydraulic
modelling of two-dimensional steady flows.
0.00
The above methodology was applied to the Giofyros basin, on
0.00 10.00 20.00 the island of Crete, Greece.
t (hours)

6.00 Notations
Points at line 38
2
3 A, B = fitting coefficients
4
5 CD = annual damage cost
4.00 6
7 CI = annual investment cost
h = flood stage
h(m)

H = height of the levee

2.00 h0 = mean water level
hmax = maximum rainfall height
i = rainfall intensity
0.00 I0x , I0y = bed slopes
0.00 10.00 20.00 30.00 Ifx , Ify = friction slopes
t (hours)
P(.) = probability
Figure 7 Water stage hydrographs h(t) at characteristic locations. PF = hydrological risk
 Risk-based floodplain management: A case study from Greece 47

Q = flow rate 7. Ganoulis, J. (ed.) (1991). Water Resources Engineering

QT = T-year flood Risk Assessment. NATO ASI Series, Vol. G29, Springer
qx , qy = flow rates per unit width Verlag, Heidelberg, p. 552.
T = return period 8. Ganoulis, J. (1994a). Risk Analysis of Water Pollution:
t = time Probabilities and Fuzzy Sets. VCH, Weinheim, Oxford, NY,
z = surrevelation p. 306.
9. Ganoulis, J. (1994b). “Flood Retention Basins in the
References Mediterranean Urban Areas”, in Rossi et al. (eds), Cop-
ing With Floods. NATO ASI Series, Kluwer Academic,
1. Bogardi, I. and Nachtnebel, H.P. (1994). Multicriteria Dordrecht, pp. 759–765.
Decision Analysis in Water Resources Management, IHP, 10. Ganoulis, J. (1995). “Floodplain Protection and Man-
UNESCO, Paris, p. 469. agement in Karst Areas”, in Gardiner, J. et al. (ed.),
2. Davy, L. (1989). Une catastrophe naturelle: l’averse Defence from Floods and Floodplain Management. NATO
Nimoise du 3 Oct. 1988 et ses consequences. Hydrologie ASI Series, Vol. 299, Kluwer Academic, Dordrecht,
Continentale 4, No. 2, 75–92. pp. 419–428.
3. Dotson, H. and Davis, D. (1995). “Case Study: Risk-Based 11. Ganoulis, J. and Vafiadis, M. (1995). “Urban Flood Con-
Analysis of Flood Reduction Measures”, in Gardiner, J. et al. trol in Karst Areas: the Case of Rethymnon (Greece)”, in
(ed.), Defence from Floods and Floodplain Management. Gardiner, J. et al. (ed.), Defence from Floods and Flood-
NATO ASI Series, Vol. 299, Kluwer Academic, Dordrecht, plain Management. NATO ASI Series, Vol. 299, Kluwer
pp. 479–485. Academic, Dordrecht, pp. 181–192.
4. Fang, L., Hipel, K.W. and Kilgour, D.M. (1993). Inter- 12. Goicoechea, A., Hansen, D.R. and Duckstein, L. (1982).
active Decision-Making – The Graph Model for Conflict Multiobjective Decision Analysis with Engineering and
Resolution, J. Wiley and Sons, N.Y. Business Applications, J. Wiley and Sons, N.Y.
5. Fraser, N.M. and Hipel, K.W. (1984). Conflict Analysis: 13. Rossi, G., Harmoncioglu, N. and Yevjevich, V. (1994).
Models and Resolutions, North Holland, N.Y., p. 377. Coping With Floods. NATO ASI Series E, Vol. 257, Kluwer
6. Gardiner, J., Strarosolszky, O. andYevjevich, V. (eds.) Academic, Dordrecht.
(1995). Defence from Floods and Floodplain Management. 14. Vincke, P. (1989). L’aide multicritere a la decision, Editions
NATOASI Series E,Vol. 299, KluwerAcademic, Dordrecht. de l’Universite de Bruxelles.
View publication stats