floodplain management strategies for flood attenuation in the river po

RIVER RESEARCH AND APPLICATIONS

River Res. Applic. 27: 1037–1047 (2011)

Published online 7 June 2010 in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/rra.1405

FLOODPLAIN MANAGEMENT STRATEGIES FOR FLOOD ATTENUATION INTHE RIVER PO

A. CASTELLARIN,a*,y G. DI BALDASSARREbz and A. BRATHax

a Department of Civil and Environmental Engineering (DISTART), School of Civil Engineering, University of Bologna, Italyb Department of Hydroinformatics and Knowledge Management, UNESCO-IHE Institute for Water Education, Delft, The Netherlands

ABSTRACT

This paper analyses the effects of different floodplain management policies on flood hazard using a 350 km reach of the river Po (Italy)as a case study. The river Po is the longest Italian river, and the largest in terms of streamflow. The middle-lower Po flows East some350 km in the Pianura Padana (PoValley), a very important agricultural region and industrial heart of Northern Italy. This portion of theriver consists of a main channel (200–500m wide) and a floodplain (overall width from 200m to 5 km) confined by two continuousartificial embankments. Floodplains are densely cultivated, and a significant portion of these areas is protected against frequentflooding by a system of minor dykes, which impacts significantly the hydraulic behaviour of the middle-lower Po during major floodevents. This study aims at investigating the effects of the adoption of different floodplain management strategies (e.g. raising, loweringor removing the minor dyke system) on the hydrodynamics of the middle-lower Po and, in particular, on flood-risk mitigation. This is acrucial task for institutions and public bodies in charge of formulating robust flood risk management strategies for the river Po.Furthermore, the results of this study are of interest for other European water-related public bodies managing large river basins, in thelight of the recent European Directive 2007/60/EC on the assessment and management of flood risks. The analysis is performed bymeans of a quasi-2D hydraulic model, which has been developed on the basis of a laser-scanning DTM and a large amount ofcalibration data recorded during the significant flood event of October 2000. Copyright # 2010 John Wiley & Sons, Ltd.

key words: floodplain management; flood risk mitigation; dyke-protected floodplains; LiDAR topographic survey

Received 13 September 2009; Revised 1 February 2010; Accepted 12 March 2010

INTRODUCTION

Over the past decade a number of major floods in Europe,

such as the catastrophic central European flooding in 2002

and the UK flash flooding in 2004 and 2007, have triggered

the common perception that the flood risk at European level

is increasing (e.g. European Environment Agency, 2005;

Merz et al., 2007;Wilby et al., 2008). Environmental change

(e.g. land-use change, inappropriate river management,

climate variability) might be responsible for higher runoff

and deteriorated river discharge capacity problems, increas-

ing the chance of flooding at the local scale (Brath et al.,

2003). The correct schematization and accurate reproduc-

tion of the actual hydraulic behaviour of structural measures

for flood hazard mitigation, such as river engineering works

or different floodplain management strategies, is a crucial

task for European water-related institutions and public

bodies in charge of formulating robust flood risk

*Correspondence to: A. Castellarin, Department of Civil and EnvironmentalEngineering (DISTART), School of Civil Engineering, University ofBologna, Italy. E-mail: [email protected] professor.zLecturer.xProfessor.

Copyright # 2010 John Wiley & Sons, Ltd.

management strategies as required by the recent Directive

2007/60/EC of the European Parliament on the assessment

and management of flood risks (European Parliament,

2007).

During the last two centuries, the height of river

embankments has increased significantly to protect flood

prone areas, and rivers have become more and more

controlled (Janssen and Jorissen, 1997). Nevertheless, the

heightening of embankments represents a component of a

vicious cycle, by which people may feel safer and

investments in the prone area may increase. This may

produce the so-called ‘levee effect’, first identified by

Gilbert White in the 1940s, whereby, paradoxically, flood

defences actually increase overall vulnerability as protection

from regular flooding reduces perceptions of risk and

encourages inappropriate development that is then vulner-

able to high-consequence and low-probability events. At the

same time, with steadily increasing embankment heights,

the potential flood depth increases, which in turn increases

the flood damage if a failure occurs (see e.g. Vis et al., 2003;

Di Baldassarre et al., 2009a). As an example, Figure 1a

reports the geometry of a river Po cross section—located in

Pontelagoscuro (District of Ferrara, FE), approximately

80 km from the coast and just before the apex of the river Po

-8.0

-4.0

0.0

4.0

8.0

12.0

16.0(a)

(b)

5004003002001000Y (m)

Z (m

a.s

.l.)

2005

1878

1

1.5

2

2.5

3

3.5

4

4.5

200019601920188018401800years

Floo

d de

pth

(m)

500

1000

1500

2000

2500

Leng

th o

f lev

ee s

yste

m (K

m)

Flood depth

Length of the levee system

Figure 1. River Po at Pontelagoscuro, Italy: (a) river cross-section in 1878and 2005; (b) evolution in time of the length of the overall main embank-ment system for the river Po and its main tributaries and correspondingincrease of the maximum water depth at Pontelagoscuro observed during

floods (see Di Baldassarre et al., 2009a)

1038 A. CASTELLARIN ET AL.

delta—clearly showing the heightening and widening of

main embankments in the period 1878–2005. Figure 1b

shows the increase of the length of the entire levee system

(river Po and tributaries) and the corresponding increase of

the largest observed water depth at Pontelagoscuro during

the same historical period.

An effective and desirable alternative to embankment

heightening would be the increase of the lateral space that

rivers can occupy during floods (e.g. ‘Room-for-River’,

Hooijer et al., 2004). Literature reports several solutions for

implementing this strategy: dyke relocation; flood bypasses

or ‘green rivers’; floodplain lowering; retention areas

external to main embankments (Silva et al., 2001). All

these measures aim at creating additional space for storage

and discharge during floods in dyke-protected parts of

valleys and alluvial plains. As a result these measures can

contribute significantly to increase discharge capacity and

peak attenuation, or ‘peak shaving’ (e.g. Hooijer et al.,

2004). Aside from floodplain lowering, these measures are

unsuitable or too expensive to implement for significant

portions of the lower reaches of many large European rivers.

In the alluvial plains of major rivers, close to the main

embankments, industrial, commercial, agricultural and


residential areas are often massively present. These areas

have steadily grown during centuries and should necessarily

be delocalized for implementing a number of the afore-

mentioned engineering interventions. When these circum-

stances hold, floodplain lowering and anthropogenic

rejuvenation (e.g. Baptist et al., 2004) appear to be a

feasible strategy. Nevertheless, regional and local regula-

tions on sediment exploitation and preservation of natural

habitats, or the presence of shallow unconfined aquifers,

may limit considerably the excavation depth, and therefore

the effectiveness of the flood risk mitigation measure.

This study focuses on dyke-protected floodplains (DPFs),

investigating their potential and effectiveness in terms of flood

mitigation. The scientific and technical community is still

debating on possible advantages and disadvantages associated

with this floodplain management strategy. For example, the

literature highlights possible environmental drawbacks of

DPFs, due to the possibility that in these areas concentration

of chemically persistent organic and inorganic (i.e. heavy

metals) pollutants exceeds natural background level (see e.g.

Japenga and Salomons, 1993). The objective of the study is to

quantify the advantages in terms of flood hazard mitigation of

a system of DPFs located within the two main embankments

of the middle-lower reach of the river Po (Italy).

The river Po is the largest river in Italy and the analysis of

the hydraulic behaviour of its middle-lower reach is a

complex hydraulic problem, which is of interest from both

scientific and administrative viewpoints. Figure 2 illustrates

the middle-lower reach of the river Po from Isola S. Antonio

to Pontelagoscuro (about 350 km). This river reach is located

in the centre of a large flat alluvial plain, the Pianura Padana

(Po Valley), which is a very important agricultural region

and industrial heart of Northern Italy. This portion of the

river is characterized by a stable main channel, whose width

ranges from 200 to 500m, and a floodplain confined by two

continuous main embankments, whose overall width varies

from 200m to 5 km. The embanked floodplain is densely

cultivated, and some areas are protected against frequent

flooding by a system of minor dykes (the dyke-protected

areas are illustrated in Figure 2). The aim of this study is the

analysis of the effects of possible modification of this minor-

dyke system on the hydraulic behaviour of the middle-lower

Po during major flood events.

This study utilizes a quasi-2D hydraulic model, which has

been developed on the basis of laser-scanning DTM

(resolution: 2m, topographic survey: 2005) and a set of

calibration data recorded during the high magnitude flood

event of October 2000. The model is used as a tool for a

medium-to-large scale study that aims at quantifying the

effects on flood hazard of different floodplain management

strategies using as hydrological input two synthetic design

hydrographs, characterized by a return period of about

200 years.

River Res. Applic. 27: 1037–1047 (2011)

DOI: 10.1002/rra

Figure 2. Po River Basin (left panel—from Wikipedia Commons); �350 km study reach of river Po from Isola S. Antonio to Pontelagoscuro, dashed boxhighlights the sub-reach with most of the dyke-protected floodplain areas (top right panel—system of dyke-protected floodplain areas, DPF; lower right

panel—main embankments: thin solid line; DPF: shaded areas)

FLOODPLAIN MANAGEMENT STRATEGIES FOR FLOOD ATTENUATION 1039

HYDRAULIC MODEL

Topographic data

The Po River Basin Authority (AdB-Po, www.adbpo.it)

coordinates the management for the entire drainage basin

area (see Figure 2). AdB-Po recently commissioned the

construction of a 2m DTM for the riverbed of a 350 km

reach of the middle-lower portion the river Po (Figure 2).

The DTM was built on the basis of the data collected in year

2005 during numerous flights, using two different laser

scanners (3033 Optech ALTM and Toposys Falcon II), from

altitudes of approximately 1500m. Below the water surface,

channel bathymetry of the navigable portion was acquired

during the same year by boat surveying using a multi-beam

sonar. Elsewhere, these data were supplemented with ground

survey of about 200 cross sections conducted by the

Interregional Agency for the Po River (AIPO, www.agen-

ziainterregionalepo.it). The resulting DTM (Figure 3) was

validated against the data achieved through a DGPS. Mean

quadratic residuals between DGPS survey and DTM were

less than 0.13m for approximately 25 000 control points

located over a 150 km reach. Also, the validation procedure

Figure 3. Two metres DTM of the river Po, main embankments and minordykes are visible—3D visualization of a 12� 6 km2 area—(courtesy of

Camorani et al., 2006)


confirmed the absence of local systematic differences

(Camorani et al., 2006).

Model implementation

A quasi-two-dimensional (quasi-2D e.g. Willems et al.,

2002) hydraulic model was built using the UNET code

(Barkau, 1997). UNET, now available as part of the software

package HEC-RAS (Hydrologic Engineering Center, 2001),

numerically solves the Saint-Venant equations, through an

algorithm that uses a classical implicit four-point finite

difference scheme (Preissmann, 1961).

In this study, UNET was used in a quasi-2D approach.

Specifically, areas protected by the minor dyke system were

modelled as storage areas, connected to the main channel by

means of lateral weirs, which represent the minor dyke

elevations. Compared to the fully 2D approach, the quasi-2D

approach has the advantage that the computational cost is

very low (e.g. Willems et al., 2002). This is extremely

important in this study where a large region, the entire

middle-lower portion the river Po (Figure 2), is modelled.

The model has been developed by following the general

criteria of objectivity and parsimony. The criterion of

objectivity was followed by sticking to the information

contained in the 2005 DTM during the construction of

the model, and avoiding the inclusion of any element of

subjectivity in the schematization of the river geometry.

Along with the DTM, the study considered several

additional sources of topographic information, such as a

traditional topographic survey and topographic maps with

scales 1:25.000 and 1:10.000. Given that the general aim

was to develop a quasi-2D model for large-scale investi-

gation of the hydraulic behaviour of the middle-lower Po,

engineering works and civil infrastructures interfering with

River Res. Applic. 27: 1037–1047 (2011)

DOI: 10.1002/rra


the riverbed (e.g. bridges, diversions, etc.) were not

represented. This choice is supported by the results of

previous studies on the same river reach (see e.g. Magis. Po –

SIMPO, 1982, 1984, Consorzio Italcopo, 2002), which

showed for the study reach that the effects of such structures

on simulated water surface elevations and discharges are

always negligible (i.e. a few centimetres at most) and

definitely smaller than standard simulation uncertainties

(e.g. Samuels, 1995). When possible, the river morphology

was described by resorting to the information retrieved from

the DTM alone. All DPFs were contoured (see Figure 2) by

identifying their levees in the DTM (see Figure 4). Then the

volumes associated with different water levels for all DPFs

were computed to describe the capacity of each one of them

in terms of the volume-water level curve (see Figure 4).

It is worth noting that the reach under study includes

approximately 50 minor DPFs, corresponding to a global

storage of roughly 450Mm3 computed at the maximum

retention level of each reservoir. The specific volume

capacity for the reach of interest is 1.6Mm3 km�1, with the

highest concentration between the towns of Cremona and

Revere-Ostiglia (see Figure 2).

The hydraulic model describes the main channel and the

unprotected floodplain through a series of cross-sections

extracted from the DTM, following the indications reported

in the scientific literature on the optimal average spacing

between cross-sections (Samuels, 1990; Castellarin et al.,

2009). The DPF areas were modelled as storage areas (e.g.

Figure 4) connected to the main channel (and/or the

unprotected floodplain) by means of lateral weirs that,

according to the criterion of objectivity (see above),

reproduce the elevation of the minor dyke system, as

described in the 2005 DTM. The same type of information

Figure 4. River Po near the dyke-protected floodplain area of S. Benedetto Po; lefthick lines), dyke-protected floodplain areas (grey areas), minor dykes (thin black liadopted in themodel (thick dashed lines), stream centrelines for the main channel a

volume—water level curve for the dyke-prote


was used to model the connections between contiguous

storage areas. The hydraulic behaviour of each dyke-

protected area is modelled as a sequence of hydrostatic

steps, during which the water level is controlled by

the volume-level curve of the storage area and by the

volume of water exchanged (i.e., inflow or outflow,

depending on the simulated water levels) with the main

channel (and/or the contiguous floodplain). Figure 4 shows

an example of this for the large dyke-protected area located

near the town of S. Benedetto Po (District of Mantua, MN,

see also Figure 2).

The implementation of the model followed also the

general criterion of parsimony. The model differentiates

friction coefficients between the main channel and

the unprotected floodplain. In particular, the model uses

only one friction coefficient for the entire unprotected

floodplain and, according to the indication reported in

previous studies (Magis. Po – SIMPO, 1982, 1984) and the

information contained in the 2005 orthophotoplan of the

study reach, adopts a sub-division of the main channel into

nine hydromorphologically homogeneous sub-reaches,

characterized by different friction coefficients. The 10

different friction coefficients were calibrated using the

information available for the significant flood event of

October 2000.

Calibration of the model

The numerical model was calibrated for the recent flood

event of October 2000 in the light of the event magnitude

(the estimated recurrence interval is approximately 50 years)

and the completeness of the available flood data. The

information used for the calibration of the model is briefly

t panel: main embankments (black thick lines), main channel (between greynes), surveyed river cross-sections (very thin black line), river cross-sectionsnd lateral unprotected floodplains (black and grey arrowed lines); right panel:cted floodplain area of S. Benedetto Po

River Res. Applic. 27: 1037–1047 (2011)

DOI: 10.1002/rra


discussed herein; further details can be found in Consorzio

Italcopo (2002) and Coratza (2005).

The flow hydrograph observed by ARPA Piemonte

(Agenzia Regionale per la Protezione dell’Ambiente,

ARPA; Piedmont Regional Environmental Protection

Agency) at Isola S. Antonio (District of Alessandria, AL)

was used as upstream boundary condition (see Figure 5).

The stage hydrograph observed at Pontelagoscuro by ARPA-

Emilia Romagna was used as downstream boundary

condition (see Figure 5). All major tributaries were

represented as concentrated lateral inflows and the flow

hydrographs were retrieved from previous studies (see e.g.

Consorzio Italcopo, 2002).

The 10 values of the friction coefficient used for the main

channel were identified by optimizing the performance of

the model relative to (i) stage hydrographs observed at seven

intermediate hydrometers and high watermarks surveyed

after the flood; (ii) inundation dynamics of the DPF areas

(i.e. starting time of the inundation and maximum water

0

2500

5000

7500

10000

12500

22-Oct20-Oct18-Oct16-Oct14-Oct

Dis

char

ge (m

3 /s)

Isola Sant'Antonio (upstream)

Figure 5. Calibration of the quasi-2D model: upst

10

20

30

40

50

60

70

16-O

ct

17-O

ct

18-O

ct

19-O

ct

20-O

ct

21-O

ct

Elev

atio

n (m

a.s

.l.)

ModelObservations

Ponte Becca

Piacenza

CremonaCasalmaggiore

Sermide

BorettoBorgoforte

5000

7500

10000

12500

15000

220

Pea

k flo

w (m

3 /s)

Figure 6. Calibration of the quasi-2D model: observed versus simulated hydrograuncertainty


level in each reservoir); (iii) observed flood peaks at all

intermediate hydrometric stations. The calibrated values

obtained for the Manning’s friction coefficients of the nine

homogeneous sub-reaches are similar to each other and very

close to the mean value, 0.04 sm�1/3 (computed as the

weighted average by weighting each coefficient proportion-

ally to the length of the corresponding sub-reach). After

calibration we obtained a friction coefficient for the active

floodplains equal to 0.10 sm�1/3. The values of the

Manning’s friction coefficient obtained in this study for

the main channel and active floodplains are in agreement

with what reported in the scientific literature and previous

studies on the same reach (see e.g. Chow, 1959; Magis. Po –

SIMPO, 1982, 1984; Consorzio Italcopo, 2002; Di

Baldassarre et al., 2009a)

Figure 6 shows some of the results obtained in model

calibration. In particular, the figure compares observed

versus simulated stage hydrographs at seven intermediate

hydrometric gauges and observed versus simulated peak

5.0

7.5

10.0

12.5

22-Oct20-Oct18-Oct16-Oct14-Oct

Elev

atio

n (m

a.s

.l.)

Pontelagoscuro (downstream)

ream and downstream boundary conditionsP

ON

TELA

GO

SC

UR

O

BO

RG

OFO

RTE

BO

RET

TO

CR

EM

ON

A

PIA

CE

NZA

BEC

CA

520000420000320000000Chainage (m)

Model Observations

phs and peak flows (observed peak flows are reported together with a 10%band)

River Res. Applic. 27: 1037–1047 (2011)

DOI: 10.1002/rra


discharges along the study reach of the river Po. It is worth

noting that the latter comparison takes into account the

uncertainty of river discharge observations (e.g. Di

Baldassarre and Montanari, 2009). The difference in time

between simulated and observed inundation of the dyke-

protected area resulted to be always smaller than 6 h, and the

simulated maximum water level inside each storage area

resulted to be practically coincident to the observed

maximum water level of the corresponding DPF. Calibration

errors are summarized in Table I.

NUMERICAL STUDY

The quasi-2D model was utilized as a tool to assess the

hydraulic behaviour of the middle-lower reach. In particular,

the model was used to quantify (1) the contribution of the

system of minor dykes to flood peak attenuation and (2) the

effects on the flood peak attenuation of alternative floodplain

management strategies.

The analysis was performed for a sub-portion of the

modelled reach that extends from Cremona to Pontela-

goscuro (length: �190 km, see Figure 2). A �135 km

upstream part of this reach (from Cremona to the town of

Revere-Ostiglia) is characterized by wide lateral flood-

plains. This part incorporates the majority of the DPF areas

located along the river Po. The downstream part (from

Revere-Ostiglia to Pontelagoscuro, �55 km) presents a

radically different morphology, the average width of the

cross-section is much smaller and only a limited number

of small DPF areas are present downstream the town of

Revere-Ostiglia.

Water levels in the �55 km reach between Revere-

Ostiglia and Pontelagoscuro, due to the reduced width of this

reach, are particularly sensitive to the peak attenuation that

takes place in the upstream floodplain (dyke-protected and

unprotected). Hence, the effects of different floodplain

management strategies were analysed along this portion of

the river.

Table I. Calibration: mean and standard deviation of relative errors and redischarges and hydrographs (seven intermediate stages) and high watestandard deviation of the differences between observed and simulated i

Relative error

Mean

Peak discharge 0.57%Hydrograph (water level) �0.23%High water mark 3.92%

Absolute errorMean

Inundation dynamics �0.8 h


The numerical simulations considered three different

geometric configurations of the floodplain. These can be

listed as follows:

� 0

latrmnu

S

S

5_DPF: Current configuration of DPF, with altimetric

configuration of minor dykes retrieved from the 2005

DTM;

� M
odDPF: The elevation of all minor dykes is increased to
1m below the corresponding main embankments, which

is the maximum possible minor-dyke height, according to

the current AIPO directive.

� N
o_DPF: The minor dyke system is completely removed.
It is worth noting that this type of strategy is also

considered in the framework of the ‘Room for the Rhine

branches’ (removal of minor river dykes; Asselman and

van Wijngaarden, 2002).

These configurations include the current geometry of the

river reach as described by the 2005 DTM (05_DPF) and two

further geometric scenarios, ModDPF and No_DPF, which

represent two limit cases. These cases are particularly

suitable for discussing the hydraulic efficiency of different

floodplain management strategies for the middle-lower

reach of the Po.

Two synthetic flood hydrographs, built for the hydro-

metric gauge of Cremona, are used as upstream boundary

conditions for the numerical simulations: ONDA2 and

TR200. The flood waves represent major flood events

that can be associated with a recurrence interval of about

200 years, which is the recurrence interval adopted by

AdB-Po and AIPO to test the adequacy of the system of

main embankments for middle-lower Po and its major

tributaries. These flood hydrographs were defined during

two previous studies. In particular, ONDA2 was defined

within the SIMPO hydraulic study (Magis. Po – SIMPO,

1982, 1984) as a 5% increase of the flood wave recorded

at Cremona gauge during the catastrophic flood event

of November 1951. The second synthetic hydrograph,

TR200, derives from the application of the method

proposed by, Maione et al. (2003) to the available

ive errors in absolute value between observed and simulated peakarks (simulated and surveyed at 205 cross-sections); mean andndation starting time for five major DPF systems

Relative error (abs. value)

td. Dev. Mean Std. Dev.

4.96% 3.65% 3.07%1.69% 1.53% 1.21%4.98% 4.64% 4.31%

Absolute error (abs. value)td. Dev. Mean Std. Dev.5.3 h 4.3 h 2.2 h

River Res. Applic. 27: 1037–1047 (2011)

DOI: 10.1002/rra


flood hydrographs recorded at Cremona up to year 2000.

TR200 can be globally associated with a recurrence

interval of 200 years, since its peak discharge is the 200-

year flood quantile, and the maximum flood volume

returned by hydrograph for any given duration (e.g. 3, 6,

12, 24 h etc.) is the 200-year flood volume for that

particular duration (Maione et al., 2003). Figure 7 shows

ONDA2 and TR200 in terms of river discharge and flood

volume exceeding a given discharge threshold, and

compares them with the hydrograph recorded during the

October 2000 flood event.

For all runs the lateral inflows were neglected and a rating

curve at Pontelagoscuro was adopted as the downstream

boundary condition.

DISCUSSION

Figure 8 illustrates some of the results obtained for ONDA2

hydrograph and all three geometric configurations con-

sidered in the study. The diagrams compare the three

hydrographs simulated at Pontelagoscuro both in terms of

discharge and water surface elevation (in the first case the

upstream boundary condition, i.e. the hydrograph at

Cremona, is also illustrated).

7500

10000

12500

15000

967248240-24-48-72-96Time (hours)

Dis

char

ge (m

3 /s)

ONDA2TR2002000 flood (obse

Figure 7. Synthetic hydrographs at Cremona compared with the hydrograph obseflood volume exceeding a giv

Figure 8. Numerical simulations: simulated hydrographs at Pontelagoscuro (


Figure 8 shows how the rising limb of hydrographs

simulated at Pontelagoscuro is significantly attenuated

relative to the upstream condition (Cremona hydrograph)

in all simulations except for the case No_DPF (i.e. absence

of minor dyke system). The difference between No_DPF and

the pair 05_DPF (i.e. current configuration) and ModDPF

(i.e. minor dyke heightening) can be ascribed to the storage

of part of the initial flood volume into the system of DPF

areas. This consideration, along with the analysis of the

hydrographs simulated at Pontelagoscuro for the geometric

configurations 05_DPF and ModDPF, points out that the

system of protected floodplains begins to accumulate flood

volumes for relatively low discharges (the main levees of

river Po downstream Cremona can withstand discharges that

are significantly larger than 10 000m3 s�1). Evidently,

stored volumes are initially smaller for ModDPF than for

05_DPF, as minor-dykes forModDPF are higher than minor-

dykes for 05_DPF. This aspect is illustrated in Figure 9,

which depicts the temporal dynamics of the overall

utilization of the storage volume for the compound of dyke

protected floodplains (i.e. lateral storage areas) referring to

ONDA2 hydrograph and the three geometric configurations

considered in the study. Volumes are expressed in per cent of

the maximum simulated volume for the 05_DPF scenario.

Figure 9 shows that significant volumes are accumulated in

120

rved)

7500

10000

12500

15000

180013509004500Volumes (Mm3)

Dis

cgar

ge (m

3 /s)

rved during the October 2000 flood (left panel: river discharge; right panel:en discharge threshold)

downstream end) for the synthetic flood wave ONDA2 (see Figure 7)

River Res. Applic. 27: 1037–1047 (2011)

DOI: 10.1002/rra

Figure 9. Progression in time of the overall storage volume for the com-pound of dyke protected floodplains for the three geometric scenarios andsynthetic flood wave ONDA2 (volumes are expressed in per cent of the

maximum simulated volume for the 05_DPF scenario)

- 0,5

0,0

0,5

1,0

1,5

510 520 530 540 550 560Chainage (km)

05_DPF ModDPF No_DPF Optimal

ONDA2

Diff

eren

ces

(m)

-0,5

0,0

0,5

1,0

1,5

510 520 530 540 550 560Chainage (km)

TR200

Diff

eren

ces

(m)

Figure 10. �55 km reach between Revere-Ostiglia and Pontelagoscuro(see also Figure 1): differences between maximum water surface elevationssimulated for the reference geometric configuration (i.e. 05_DPF) and fouralternative configurations (i.e. 05_DPF, No_DPF, ModDPF and Optimal)

using two different synthetic hydrographs (i.e. ONDA2 and TR200)


the initial part of the event for the No_DPF scenario, which

is also unable to fully exploit the accumulation potential

(maximum accumulated volume is 80% of the 05_DPF

scenario). Also, Figure 9 shows that the largest maximum

overall volume is associated with 05_DPF, but ModDPF

utilizes the available storage volume in a more efficient way.

Relative to 05_DPF, ModDPF delays the volume accumu-

lation saving a significant portion of the retention volume for

when the flood peak occurs. As a result, simulated

streamflows at Pontelagoscuro (see Figure 8) are higher

for ModDPF than for 05_DPF up to discharges around

10 000m3 s�1; for higher discharge values, the hydrograph

simulated for 05_DPF is above theModDPF one. Obviously,

given the downstream boundary condition adopted in these

numerical experiments, these considerations hold also in

terms of water level (see Figure 8). Hydrographs simulated

using the synthetic flood wave TR200 confirm these findings

and are not reported herein.

Figure 10 reports the results of the numerical simulations

for the entire �55 km reach from Revere-Ostiglia to

Pontelagoscuro, using the two hydrographs (TR200 and

ONDA2). The results are reported for cross-sections used in

the model as a function of the river chainage. The results are

illustrated in terms of differences between maximum water

surface elevations simulated using the current geometric

configuration (05_DPF)and the twoalternative configurations

(No_DPF, ModDPF). Positive differences for a given river

cross-section indicate that the maximum water surface

elevation simulated for the particular geometric configuration

(e.g. NO_DPF, ModDPF, etc.) is lower than for configuration

05_DPF (i.e. advantage in terms of flood risk mitigation). To

better characterize the magnitude of the differences reported

in Figure 10, the diagrams for TR200 and ONDA2 report an


additional reference curve. This curve is still a difference

between maximum water surface elevations simulated for

05_DPF and for a hypothetical geometric configuration

(Optimal in Figure 10). The hypothetical configuration

replaces all DPFs with a single theoretical flood control

reservoir with a storage capacity of 446Mm3 (i.e. overall

storage capacity of the DPF system) located in Revere-

Ostiglia. This geometric configuration is used to quantify the

theoretical maximum reduction in water surface elevations

that, hypothetically, can be achieved by an optimal floodplain

management strategy. The results illustrated for the �55 km

reach inFigure10arealso summarized in termsofaverageand

maximum differences in Table II.

The comparison between the results obtained for

geometric configurations 05_DPF and No_DPF enables

one to quantify the modifications in the peak attenuation

capacity of the river associated with a complete removal of

the minor dyke system, a totally hypothetical scenario.

Equivalently, this comparison enables one to assess the

increase of the flood peak attenuation associated with the

presence of the current (i.e. year 2005) minor dyke system.

From the differences reported in Figure 10 and Table II it can

be observed that the hypothesis of a complete removal of the

minor dyke system would result in a significant increase of

the flood hazard along the �55 km reach downstream

Revere-Ostiglia for both considered synthetic hydrographs.

It has to be noted that the geometric schematization of the

riverbed (i.e. main channel and floodplain) for the No_DPF

configuration still adopts a lateral storage area representa-

tion for all the floodplain areas that are currently protected

against frequent flooding by minor dykes. This schematiza-

tion implies that the velocity of stored water volume is equal

to zero. In fact, if the minor dyke system was totally

River Res. Applic. 27: 1037–1047 (2011)

DOI: 10.1002/rra

Table II. Average (and maximum) differences in terms of simulated maximum water surface elevations with respect to the referencegeometric configuration (i.e. 05_DPF) for the �55 km reach between Revere-Ostiglia and Pontelagoscuro

Synthetic hydrograph No_DPF ModDPF Optimal

TR200 �36 cm (max: 0 cm) 10 cm (max: 14 cm) 121 cm (max: 145 cm)ONDA2 �27 cm (max: 0 cm) 15 cm (max: 18 cm) 114 cm (max: 134 cm)


removed, the streamflow velocity over the floodplain,

although very low, could not be neglected. Aureli et al.

(2006) performed a comparison of fully-2D and quasi-2D

(storage areas with zero-velocity) schematizations for two

test sites in the floodable regions located in the middle reach

of the river Po. The authors show that differences in terms of

water level are negligible (a few cm in Figure 6 on p. 1748;

Aureli et al., 2006). Therefore, we deemed the adopted

quasi-2D schematization suitable for the hydraulic problem

at hand and the geographical scale of investigation.

The comparison between the results obtained for 05_DPF

and ModDPF configurations enables one to quantify the

increase of the flood peak attenuation corresponding to

alternative floodplain management strategies that slightly

modifies the current altimetric configuration of the minor-

dyke system. The results of the study point out that the

increase of flood peak attenuation associated with the

considered floodplain management strategy is not negli-

gible, even though the advantages in terms of reduction of

maximum water surface elevation are rather limited (see

Table II and Figure 10).

The floodplain management strategy mimicked by

ModDPF voluntarily refers to a simple policy, which adopts

the same criterion along the entire river reach (i.e. raising

and widening of all minor dykes, whose crest is currently,

i.e. in year 2005, 1m lower than the corresponding main

embankment crest). Obviously, differentiating this criterion

along the reach under investigation by allowing local

variations in the elevation of minor dykes or shaping the

longitudinal profile of the dykes themselves to optimize their

hydraulic behaviour for a particular flood event would result

in an even stronger peak attenuation that what obtained for

ModDPF. Nevertheless, it has to be remembered that the

elevation of minor dykes has to meet the previously

mentioned AIPO directive. Also, a uniform floodplain

management strategy is more suitable for the present

investigation, which is performed for a medium-to-large

spatial scale (i.e. whole middle-lower reach of the river Po).

Concerning the two synthetic flood waves considered in

this study, which can be associated with recurrence intervals

larger than 200 years, the results obtained in the study

highlight that remodelling the minor dyke system according

to the AIPO Directive leads to rather limited flood

hazard mitigation. In particular, the limitedness of these

improvements can be better appreciated by a comparison


with the theoretical geometric configuration denoted as

‘Optimal’ in Figure 10 and Table II. It has to be noted that

this ad hoc optimization is to be evaluated carefully in view

of the errors that may affect the definition of hydrological

inputs (ONDA2 and TR200) as well as the uncertainty that

always affects hydraulic analysis. Moreover, these limited

improvements would be associated with widespread actions

on the territory, which are also definitely expensive to

put into practice. Concerning this point, though, a few

considerations should be made.

First, the study considers only the modification of the

minor-dyke system according to the AIPO directive as a

possible intervention, without hypothesizing a concurrent

floodplain lowering (floodplain rejuvenation, e.g. Baptist

et al., 2004). Integrating the minor-dyke system remodelling

with floodplain rejuvenation would definitely improve the

hydraulic effectiveness of the intervention.

Moreover, in the light of the results of this study and given

the wide margin of improvement represented by the

hypothetic optimal geometric configuration (see Figure 10

and Table II) it could be recommended to revise the AIPO

directive allowing a smaller minimum difference between

the crest elevation of the main embankments and the one of

the minor dykes (currently set to 1m). The AIPO directive

should be revised on the basis of an extensive analysis which

considers a number of possible alternatives. This point is

currently under investigation.

Second, the result of the study strengthen the conclusions

drawn by the analysis of the evolution in time of the

maximum observed water level at Pontelagoscuro versus the

increase of the overall length of the Po main levee system

(see Figure 1). Main embankments cannot be the only flood

risk mitigation measure along the middle-lower reach of the

river Po. Different strategies should be identified for

reducing flood risk for residential, agricultural and industrial

areas of the Po Valley for all flood events that exceed the

design flood (e.g. for the Po River Basin Authority the 200-

year flood). For extraordinary floods, or flood events larger

than the design event, a reliable flood risk mitigation plan

cannot rely only on the embanked area within river Po main

embankments; the effective utilization of less vulnerable

and less valuable areas outside main embankments should

be foreseen instead (controlled flooding; see e.g., Di

Baldassarre et al., 2009b). This strategy belongs to the

‘Room-for-River’ approach and represents an attractive

River Res. Applic. 27: 1037–1047 (2011)

DOI: 10.1002/rra


flood mitigation strategy, which definitely requires

additional investigations. The implementation of controlled

flooding techniques for the river Po represents a very

complex issue from technical and socioeconomic view-

points due to the high population density and number of

industrial and agricultural areas currently protected by the

main embankments. Sticking to technical aspects, which are

probably the easier to deal with, the implementation of a

‘Room-for-River’ strategy that adopts a retention area

external to main embankments requires a compound of

engineering and socioeconomic analyses and political

directives and regulations aiming at (i) identifying the most

suitable areas to inundate during extraordinary flood events;

(ii) reducing the vulnerability of goods that are present in

these areas; (iii) empowering of the minor artificial channel

network in charge of draining the inundated areas after the

flood event; (iv) locally reinforcing the main embankments

that has to resist against overtopping and has to serve as the

inlet structure for the retention area.

CONCLUSIONS

The study presents an extensive hydraulic analysis performed

bymeansof a quasi-2D(Willemset al., 2002)hydraulicmodel

to a 350 km reach of river Po, the largest Italian river. The

model was developed according to the general principles of

objectivity and parsimony. Principle of objectivity, as the river

geometry schematization (i.e. main channel, floodplain and

DPF areas) was entirely based upon a high quality DTM. No

subjective elements were included in the model, which

consists of amain channel, left and right floodplain and lateral

storage areas schematizing the DPF areas. Furthermore, the

model was developed according to the general principle of

parsimony because only 10 parameters were used: nine

friction coefficients corresponding to nine morphologically

homogenous sub-reaches of the main channel; one friction

coefficient to characterize the unprotected floodplain areas.

The calibration of the model was based upon the compre-

hensive information available for a recent significant flood

event (October 2000),with an estimated recurrence interval of

50years (Castellarin et al., 2009;DiBaldassarreet al., 2009a).

In particular, the calibration was aimed at optimizing the

reproduction of observed stage hydrographs at seven

intermediate cross-sections, the seven observed peak dis-

charge values, the maximum water surface elevation for the

entire �350 km reach and the recorded inundation dynamics

for the complex system of minor dykes and DPFs (inundation

times and maximum water levels).

This large-scale study aimed at quantifying the hydraulic

effects of an extensive system of minor-dykes, protecting

floodplain areas, during severe flood events. This minor-

dyke system has been built during a number of decades to


protect sensitive and valuable agricultural areas and fields

from frequent flooding and has an overall storage capacity of

about 450Mm3, with a specific volume capacity of

1.6Mm3 km�1. As such, the system of DPF areas can also

be seen as a strategic resource for flood risk mitigation in a

large portion of Pianura Padana, area of great socioeconomic

importance for northern Italy.

The results of the study can improve the knowledge on the

hydraulic behaviour of the middle-lower Po and give useful

indications that might be used by stakeholders (above all AdB-

Po and AIPO) in the definition of floodplain conservation and

management plans. In particular, the study showed how

significant the hydraulic behaviour of the minor-dyke system

is for flood risk mitigation in a large portion of the river Po.

The presence of a dense network of minor dykes significantly

improves the capacity of the river to attenuate the peak

discharge during the routing of flood waves associated with

recurrence intervals of about 200 years (reference recurrence

interval selected by AdB-Po and AIPO for designing and

testing the main embankment system). This is due to the fact

that unprotected floodplains are activated during the initial

phases of a flood event and hence their storage volume has a

negligible influence upon the peak of the hydrograph, while

minor dykes of DPFs retard the activation of storage volumes

and may increase peak flow attenuation effects.

The study also highlighted that heightening and empow-

ering all minor dykes to the maximum elevation allowed by

the current AIPO regulation (i.e. 1m below the crest

elevation of the corresponding main levee) would improve

the attenuation of the flood peak, and therefore mitigate

flood risk. Nevertheless, our study pointed out that such

additional asset would be rather limited.

Finally, the outcomes of the study indicate that there is still

remarkable potential for flood risk mitigation associated with

a correct management of the system of DPF areas of the

middle-lower reach of the river Po. In order to exploit a

significant part of this residual potential it seems advisable to

modify the current AIPO directive allowing higher minor

dykes. The definition of the optimal height of the minor-dyke

system is still an open problem, which is currently under

investigation. In fact, the height of the minor dyke system can

be seen as an instrument to control for which event

magnitudes (recurrence interval) the flood peak retention is

maximized. Evidently, the optimal dyke height is associated

with the considered event; optimal height identified with

respect to a recurrence interval 200 years may result in higher

water levels in the downstream reach in case of small and

medium flood events (e.g. 10-year flood event). Along with

minor-dykes heightening additional measures could be

recommended, such as floodplain rejuvenation (i.e. exca-

vation). In addition, both measures, minor dykes heightening

and floodplain excavation may have a positive impact on the

accumulation of organic and inorganic persistent pollutants in

River Res. Applic. 27: 1037–1047 (2011)

DOI: 10.1002/rra


the floodplain areas (see e.g. Japenga and Salomons, 1993)

due to the reduction of the frequency of floodplain inundation

and the periodic removal of the first layers of soil. These

results are obviously associated with the considered test site;

nevertheless, that the hydraulic conditions of the middle-

lower Po can be considered representative ofmany other large

rivers in Europe and around the world.

As a last remark, our study highlighted that, also for the

middle-lower reach of the river Po like for many other

European rivers (e.g. Silva et al., 2001), a correct strategy for

flood risk mitigation when considering very large or extreme

events necessarily needs to exploit areas outside the main

embankments with the ‘Room-for-River’ approach. This is

particularly true for the Po main embankments, which are

already very tall for large portions of the lower reach andwhose

further heightening is not environmentally sustainable.

ACKNOWLEDGEMENTS

Theauthors areextremelygrateful to the InterregionalAgency

for the Po River (Agenzia Interregionale per il Fiume Po,

AIPO, Italy) andPoRiverBasinAuthority (Autorita diBacino

del Fiume Po, Italy) allowing access to their high resolution

DTMof river Po.Thepreliminary analysis performedbyLuca

Galletti and Valentina Balacchi and the comments of two

unknown reviewers are also acknowledged.

REFERENCES

Asselman NEM, van Wijngaarden M. 2002. Development and application

of a 1D floodplain sedimentation model for the River Rhine in The

Netherlands. Journal of Hydrology 268(1–4): 127–142.

Aureli F, Mignosa P, Ziveri C, Maranzoni A. 2006. Fully-2D and Quasi-2D

Modelling of Flooding Scenarios Due to Embankment Failure, River

Flow 2006. Taylor & Francis Group: London (ISBN 0-415- 40815-6).

Baptist MJ, Penning WE, Duel H, Smits AJM, Geerling GW, Van Der Lee

GEM, Van Alphen JSL. 2004. Assessment of the effects of cyclic

floodplain rejuvenation on flood levels and biodiversity along the Rhine

River. River Research and Applications 20: 285–297.

Barkau RL. 1997. UNET One dimensional Unsteady Flow Through a Full

Network of Open Channels. US Army Corps of Engineerings, Hydrologic

Engineering Center: Davis.

Brath A, Montanari A, Moretti G. 2003. Assessing the Effects on Flood Risk

of Land-Use Changes in the Last Five Decades: An Italian Case Study,

IAHS Publication n8 278. IAHS Press: Wallingford, UK.

Camorani G, Filippi F, Cavazzini A, Lombardo G, Pappani G, Forlani G.

2006. Il rilievo altimetrico e batimetrico del Fiume Po nel tratto tra

confluenza Ticino e l’incile. Proceedings of X Asita Nat. Conf. (in

Italian).

Castellarin A, Di Baldassarre G, Bates PD, Brath A. 2009. Optimal cross-

section spacing in Preissmann Scheme 1D hydrodynamic models. Jour-

nal of Hydraulic Engineering, ASCE 135(2): 96–105.

Chow VT. 1959. Open-Channel Hydraulics. McGraw-Hill: New York.

Consorzio Italcopo, 2002. Aggiornamento dell’assetto idraulico di progetto

del fiume Po dalla confluenza del Tanaro all’incile del Po di Goro

mediante analisi modellistica numerica in moto vario (in Italian).

Coratza L. 2005. Aggiornamento del catasto delle arginature maestre di Po,

Autorita di Bacino del fiume Po, Parma (in Italian).


Di Baldassarre G, Castellarin A, Brath A. 2009a. Analysis of the effects of

levee heightening on flood propagation: example of the River Po, Italy.

Hydrological Sciences Journal 54(6): 1007–1017.

Di Baldassarre G, Castellarin A, Montanari A, Brath A. 2009b. Probability

Weighted HazardMaps for Comparing Different Flood Risk Management

Strategies. A Case Study, Natural Hazards. Springer: Netherlands,

Dordrecht. DOI: 10.1007/s11069-009-9355-6

Di Baldassarre G, Montanari A. 2009. Uncertainty in river discharge

observations: a quantitative analysis. Hydrology and Earth System

Sciences 13: 913–921.

European Environment Agency. 2005. Vulnerability and Adaptation to

Climate Change: Scoping Report. EEA Technical Report, Copenhagen,

Denmark.

European Parliament, 2007. Council, Directive 2007/60/Ec of the European

Parliament and of the council of 23 October 2007 on the assessment and

management of flood risks. http://eur-lex.europa.eu/en/index.htm

Hooijer A, Klijn F, Pedroli GBM, Van Os AG. 2004. Towards sustainable

flood risk management in the Rhine and Meuse river basins: synopsis of

the findings of IRMA-SPONGE. River Research and Applications 20:

343–357. DOI: 10.1002/rra.781

Hydrologic Engineering Center. 2001. Hydraulic Reference Manual. U.S.

Army Corps of Engineers: Davis, California.

Janssen JPFM, Jorissen RE. 1997. Flood Management in the Netherlands:

RecentDevelopment and ResearchNeeds. InRibamod, RiverBasinModel-

ling, Management and Flood Mitigation, Concerted action, Casale R,

Havno K, Samuels P (eds). RIBAMOD: Kopenhagen, Denmark; 89–104.

Japenga J, Salomons W. 1993. Dyke-protected floodplains: a possible

chemical time bomb? Land Degradation and Development 4(4): 373–

380. DOI: 10.1002/ldr.3400040421

Magis. Po – SIMPO. 1982. Studio e progettazione di massima delle

sistemazioni idrauliche dell’asta principale del Po dalle sorgenti alla

foce finalizzate alla difesa e alla conservazione del suolo e alla utilizza-

zione delle risorse idriche, Convenzione del 14 luglio 1980 (in Italian).

Magis, Po – SIMPO. 1984. Interventi per le golene chiuse del Fiume Po (in

Italian).

Maione U, Mignosa P, Tomirotti M. 2003. Regional Estimation Model of

Synthetic Design Hydrographs. International Journal of River Basin

Management 12: 151–163.

Merz B, Thieken AH, Gocht M. 2007. Flood risk mapping at the local scale:

concepts and challenges, In Flood risk Management in Europe: Inno-

vation in policy and practice. Series: Advances in Natural and Techno-

logical hazards Research, Vol. 25, Chapter 13, Begum S, Stive MJF, Hall

JW (eds). Springer: Dordrecht; 231–251.

Preissmann A. 1961. Propagation of translatory waves in channels and

rivers. Proceedings of First Congress of French Association for Com-

putation; 433–442.

Samuels PG. 1990. Cross section Location in One-dimensional Models. In

International Conference on River Flood Hydraulics, White WR, (ed.).

John Wiley: Chichester; 339–350.

Samuels PG. 1995. Uncertainty in Flood Level Prediction, XXVI Biannual

Congress of the IAHR, HYDRA2000, London, September 1995, Vol. 1 of

Proceedings published by Thomas Telford.

Silva W, Klijn F, Dijkman J. 2001. Room for the Rhine Branches in The

Netherlands;What the research has taught us.WL/Delft Hydraulics: Delft;

RIZA: Arnhem. Delft Hydraulics Report R3294; RIZA report 2001. 031.

Vis M, Klijn F, De Bruijn KM, van Buuren M. 2003. Resilience strategies

for flood risk management in the Netherlands. International Journal of

River Basin Management 1(1): 33–44.

Wilby RL, Beven KJ, Reynard NS. 2008. Climate change and fluvial risk in

the UK: more of the same? Hydrological Processes 22: 2511–2523.

Willems P, Vaes G, Popa D, Timbe L, Berlamont J. 2002. Quasi 2D river

flood modelling. In River Flow, Vol. 2, Bousmar D, Zech Y (eds). Swets

& Zeitlinger: Lisse; 1253–1259.

River Res. Applic. 27: 1037–1047 (2011)

DOI: 10.1002/rra

floodplain management strategies for flood attenuation in the river po

Documents