You are on page 1of 9

Journal of Hydrology 487 (2013) 39–47

Contents lists available at SciVerse ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Flood mapping in ungauged basins using fully continuous


hydrologic–hydraulic modeling
Salvatore Grimaldi a,b,c,⇑, Andrea Petroselli d, Ettore Arcangeletti d, Fernando Nardi e
a
Dipartimento per la innovazione nei sistemi biologici, agroalimentari e forestali (DIBAF Department), University of Tuscia, Via San Camillo De Lellis snc, 01100 Viterbo, Italy
b
Honors Center of Italian Universities (H2CU), Sapienza University of Rome, Via Eudossiana 18, 00184 Roma, Italy
c
Department of Mechanical and Aerospace Engineering, Polytechnic Institute of New York University, Six MetroTech Center, Brooklyn, NY 11201, United States
d
Dipartimento di scienze e tecnologie per l’agricoltura, le foreste, la natura e l’energia (DAFNE Department), University of Tuscia, Via San Camillo De Lellis snc, 01100 Viterbo, Italy
e
WARREDOC, Università per Stranieri di Perugia, Piazza Fortebraccio 4, 06123 Perugia, Italy

a r t i c l e i n f o s u m m a r y

Article history: In this work, a fully-continuous hydrologic–hydraulic modeling framework for flood mapping is intro-
Received 15 August 2012 duced and tested. It is characterized by a simulation of a long rainfall time series at sub-daily resolution
Received in revised form 8 February 2013 that feeds a continuous rainfall–runoff model producing a discharge time series that is directly given as
Accepted 14 February 2013
an input to a bi-dimensional hydraulic model. The main advantage of the proposed approach is to avoid
Available online 26 February 2013
This manuscript was handled by Andras
the use of the design hyetograph and the design hydrograph that constitute the main source of subjective
Bardossy, Editor-in-Chief, with the analysis and uncertainty for standard methods. The proposed procedure is optimized for small and unga-
assistance of Niko Verhoest, Associate Editor uged watersheds where empirical models are commonly applied. Results of a simple real case study con-
firm that this experimental fully-continuous framework may pave the way for the implementation of a
Keywords: less subjective and potentially automated procedure for flood hazard mapping.
Flood hazard Ó 2013 Elsevier B.V. All rights reserved.
Flood mapping
Continuous hydrologic model
2D hydraulic model
WFIUH

1. Introduction routing algorithm for the accurate spatially distributed evaluation


of the flow and velocity dynamics.
Flood hazard mapping is a fundamental non-structural measure The more complex and detailed 2D models suffered in the past
implemented by governments for sustainable urban and land plan- of limited amounts of data and of the computational issue. How-
ning, protecting human properties, activities, lives, and preserving ever, latest 2D models, overcoming these limitations, are now
the ecohydrology of river corridors (European Parliament, 2007). increasingly used as the new standard approach for accurately sim-
While hydrogeomorphic methods, that make extensive use of Dig- ulating spatial and temporal dynamics of the flooding process (e.g.
ital Elevation Models (DEMs) and terrain analysis have been re- Horritt et al., 2007; Di Baldassarre et al., 2009). As a result, 2D sim-
cently adopted for preliminary floodplain characterization at the ulations represent a mandatory choice for detailed small scale ur-
basin scale (e.g. Dodov and Foufoula-Georgiou, 2006; Nardi et al., ban development projects as well as for large scale ones, especially
2006), detailed inundation maps are generally based on advanced for complex urban settings where the 1D hypothesis is often not
hydrologic and hydraulic modeling. In particular, the hydrologic applicable and the socio-economic impact of floods must also be
forcing is estimated through a rainfall–runoff model that quantifies assessed (European Parliament, 2007).
flood peak discharges or a flow hydrograph of given return period Nevertheless, current hydrologic–hydraulic modeling proce-
(T). The hydraulic analysis is, developed, by means of a mono dure for flood mapping is characterized by three main issues: (1)
dimensional (1D) surface water model in a steady state condition the availability of detailed topographic information; (2) choosing
and/or by means of an unsteady 1D or bidimensional (2D) flood the flood propagation hydraulic model (3) the impact that the
hydrologic forcing has on flood mapping results (i.e. design hydro-
graph estimation). The first two issues benefit of the latest techno-
logical advancements, in particular of the increasing accuracy,
⇑ Corresponding author at: Dipartimento per la innovazione nei sistemi biologici, availability, and usability of high resolution Lidar DEMs and 2D
agroalimentari e forestali (DIBAF Department), University of Tuscia, Via San Camillo
hydraulic models (e.g. Gilles et al., 2012; Sampson et al., 2012).
De Lellis snc, 01100 Viterbo, Italy. Tel.: +39 0761 357326; fax: +39 0761 357453.
E-mail address: salvatore.grimaldi@unitus.it (S. Grimaldi).
This work investigates the third issue for small and ungauged ba-

0022-1694/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jhydrol.2013.02.023
40 S. Grimaldi et al. / Journal of Hydrology 487 (2013) 39–47

sin, comparing event-based approaches with respect to new exper- means that flow measurements are not available even for regional-
imental procedures based on the implementation of continuous ization. For these small and ungauged basins, where only empirical
rainfall–runoff methods. analytical procedures are applicable, we assume that the following
For event-based approach (EBA) we refer to those schemes that data are available: rainfall registration (at daily and sub-daily res-
implement a rainfall–runoff transformation for estimating the de- olution), a DEM of 20–30 m resolution, and soils information.
sign hydrograph once given the design hyetograph of assigned re- The three approaches, depicted in Fig. 1, include five steps: (1)
turn period, (Alfieri et al., 2008; Hsieh et al., 2006; Laforce et al., rainfall input, (2) rainfall excess, (3) design hyetograph estima-
2011; Soczyñska et al., 1997). EBA procedures are widely adopted tions, (4) rainfall–runoff transformation, and (5) hydraulic flood
in engineering practice for their simplicity, even if some significant routing. A brief summary of steps 2, 4, and 5, shared by the three
drawbacks must be accounted for, and in particular: the predefined procedures, is hereafter reported, while a more detailed descrip-
hyetograph shape and/or estimation method, the critical rainfall tion of the different methodological steps 1 and 3 for EBA, SCA
duration parameter, the difficulty in assigning the pre-event soil and FCA is included in the next corresponding subsections.
moisture conditions, and the pivotal hypothesis of assigning the The rainfall excess (step 2) is estimated using the Soil Conserva-
same T value for design rainfall to the corresponding hydrograph tion Service – Curve Number (SCS-CN) method (USDA, 1986).
(Rahman et al., 2002; Verhoest et al., 2010). While we are aware that this approach is not appropriate for a
As a result, in recent times, the use of various continuous ap- sub-daily rainfall series (Grimaldi et al., in press-b, in press-c), it
proaches has been investigated (Blazkova and Beven, 2009; Calver is here applied because it does not influence the results of the com-
et al., 2009; Grimaldi et al., 2012b, in press-a; Moretti and Monta- parison tests.
nari, 2008). In those schemes, a runoff time series is obtained The rainfall–runoff transformation (step 3) is implemented
through the implementation of a continuous rainfall–runoff trans- using a parsimonious version of the Width Function Instantaneous
formation. The hydrologic input is a long synthetic or observed Unit Hydrograph (WFIUH), namely WFIUH-1par (Grimaldi et al.,
rainfall time series. The corresponding simulated runoff signal is 2012a) that optimizes the available DEM by means of advanced
then statistically analyzed for extrapolating a Synthetic Design and automated terrain analysis techniques. The procedure for per-
Hydrograph (SDH) related to a return period (Pramanik et al., forming this step is summarized in the following three sub-steps:
2010; Serinaldi and Grimaldi, 2011). In the current work, this ap-
proach is referred to as the Semi-Continuous Approach (SCA). (a) DEM pits and flat areas are removed using the Physically-
SCA is characterized by the fact that statistical analysis is not per- based Erosion Model for PIT removal PEM4PIT (Grimaldi
formed during the rainfall estimation step (as for the EBA), but on et al., 2004, 2007; Santini et al., 2009).
the derived simulated hydrographs. (b) Flow directions are defined using a mixed flow direction
Several works evaluated the EBA and SCA performances in approach, combining an advanced single-flow approach in
terms of hydrograph properties and uncertainty of results. Statisti- the channel and a multi-flow approach on the hillslopes
cal and comparative analyses of the hydrograph peak discharge, (Nardi et al., 2008). The river network is then extracted using
volume, and duration suggest that EBA tend to underestimate the the drop-analysis approach (Tarboton et al., 1991; Tarboton
flood volume (Grimaldi et al., 2005, 2012b; Nishat et al., 2010; and Ames, 2001).
Nnadi et al., 1999; Viglione and Bloschl, 2009; Viglione et al., (c) Given the drainage network, the WFIUH-1par is obtained by
2009). In particular, several researchers demonstrated the uncer- estimating a residency time distribution for rainfall drops
tainty that is induced through a selection of subjective parameters flowing throughout the domain. Surface hillslope flow veloc-
and the simplification of a physical hydrologic and hydraulic pro- ities are estimated using the NRCS formula (Grimaldi et al.,
cess that indirectly influences the flood inundation mapping proce- 2010; NRCS, 1997), while channel flow velocity is calibrated
dure (e.g. Aronica et al., 2002; Bates et al., 2004; Di Baldassarre so that the WFIUH base time is equal to the basin concentra-
et al., 2010; Hall et al., 2005; Kalyanapu et al., 2012; McMillan tion time (Tc). Tc is estimated using the Giandotti empirical
and Brasington, 2008; Merwade et al., 2008; Pappenberger et al., formula widely used in Italy (Giandotti, 1934). The adopted
2006). WFIUH-1par approach does not need observed data to be
In the following, we introduce a Fully Continuous Approach calibrated (Grimaldi et al., 2012a).
(FCA) and we test its potential to diminish the impact of subjective
preliminary steps in hydrologic–hydraulic modeling. This is The last step (5) of the flood routing is developed by imple-
achieved mainly by skipping the SDH estimation, still present in menting the 2D commercial hydraulic model FLO-2D (http://
the SCA approach, and routing the entire simulated flow time ser- www.flo-2d.com/). The FLO-2D scheme is based on the dynamic
ies along the floodway. FCA performances are then compared to wave momentum equation solved on a numerical grid of square
EBA and SCA by qualitatively analyzing inundation results. cells through which the hydrograph is routed propagating the sur-
The paper is organized as follows: after this brief introduction, face flow along the eight cardinal directions (O’Brien et al., 1993;
Section 2 provides a more detailed description of EBA, SCA and FLO-2D Reference Manual, 2012).
FCA approaches. In Sections 3 and 4, the case study and the pro-
posed procedure application are presented, while in Section 5, re- 2.1. Event Based Approach (EBA)
sults are discussed followed by conclusive remarks of Section 6.
Standard EBA is characterized by an empirical rainfall input
estimation procedure that is developed as follows. A critical rainfall
2. Event-based, semi-continuous and fully-continuous depth for the predefined return period is quantified using the
approaches Intensity–Duration–Frequency (IDF) curve assuming the wa-
tershed concentration time as duration. This critical rainfall depth
The three methodologies, EBA, SCA and FCA, are described and is distributed over time through a synthetic hyetograph. As previ-
summarized in the flow diagram inserted in Fig. 1. ously underlined, this design hyetograph characterization method
The described procedures are valid for small and ungauged ba- (Alfieri et al., 2008) is affected by four main issues: the concept of
sins. However, the general concept and related application may be critical rainfall for a given duration that tends to underestimate the
extended to any watershed. The term small refers to basin for runoff volume, the uncertainty in the estimation of the time of con-
which rainfall is assumed spatially uniform, while ungauged centration (Grimaldi et al., 2012c), the design hyetograph itself
S. Grimaldi et al. / Journal of Hydrology 487 (2013) 39–47 41

Fig. 1. Flow-chart of the three flood modeling approaches: event-based, semi-continuous, and fully-continuous.

that is arbitrary and far from representing real rainfall storms, and to the rainfall–runoff procedure (step 4), provides the correspond-
the hypothesis that the design hyetograph and corresponding hyd- ing design hydrograph.
rograph have the same return period.
For estimating the IDF, we selected the three-parameter for-
mula (Chow et al., 1988): 2.2. Semi Continuous Approach (SCA)

aðTÞ The above mentioned four issues of the EBA correspond to four
iðd; TÞ ¼ ð1Þ
½bðTÞ þ dcðTÞ significant theoretical and conceptual hypotheses representing
sources of uncertainty and/or user-derived subjective choices
where i(d,T) is the gross rainfall intensity, and d the rainfall dura- affecting the design hydrograph estimation procedure. The SCA
tion. a(T), b(T) and c(T) are determined through a nonlinear least- aims to reduce the impact of these conceptual hypotheses by
squares method. The i(d,T) is estimated on the annual maxima for substituting the IDF-derived critical rainfall with a simulated rain-
the corresponding durations of 1, 3, 6, 12 and 24 h according to fall time series. This calibrates a statistical model on the observed
the empirical cumulative distribution of the Weibull formula: rainfall data and provides it to the rainfall–runoff model without
j 1 any prior estimation of a single design event. In this way, the im-
¼1 ð2Þ pact of the uncertainty of the IDF, critical rainfall, and design hye-
nþ1 T
tograph hypotheses is substituted by the need for the proper
where j = 1, . . . , n represents the jth observation of the samples ar- selection of a rainfall simulation model. The return period is also
ranged in ascending order. estimated directly on the simulated runoff series allowing it to
The selected hyetograph estimation method is the symmetric omit another source of uncertainty given by the assumption that
Chicago hyetograph (Keifer and Chu, 1957), that, given as an input the design discharge preserves the rainfall time period.
42 S. Grimaldi et al. / Journal of Hydrology 487 (2013) 39–47

In this paper, steps 1 and 3 are developed applying the COS- from the watershed divide is 25.8 Km. The selected hydraulic mod-
MO4SUB framework (Grimaldi et al., 2012b) where the rainfall eling domain is located in the proximity of the confluence of the Ti-
simulation is performed using a copula-based daily rainfall gener- ber where the Torbido meanders within a large floodplain. This area,
ator (Serinaldi, 2009) and a continuous-in-scale universal multi- shown in Fig. 2, is 1.5 Km2 large and 2 Km long wide.
fractal model (Schertzer and Lovejoy, 1987) to disaggregate the The DEM, with integer precision and 20 m of spatial resolution, is
daily rainfall. The simulated rainfall time series is given as an input extracted from the Italian Geographic Military Institute (IGMI, 2003)
to the runoff model applied in continuous using a predefined event MATRIX raster dataset. The IGMI Matrix DEM, derived from the
separation time parameter (Ts) that allows to separate the rainfall interpolation of 1:25.000 contour lines, is characterized by a spatial
events for the appropriate implementation of the SCS-CN scheme. and vertical accuracy of respectively 5–10 m and 1–5 m. Land cover
Selected the annual maximum hydrographs from the obtained run- is derived from the CORINE project (European Commission, 2000).
off time series, the SDHs are estimated using the peak discharge as Rainfall measurements include 49 years of observed data at a daily
the driving variable for each return period following the procedure time scale and 10 years at 5 min of resolution (Serinaldi, 2010).
described in Serinaldi and Grimaldi (2011) where the hydrograph Fig. 3 shows the Rio Torbido basin DEM with the simulated river
shape is assumed to be similar to the Generalized Standard Two- network implementing the terrain analysis procedure described in
Sided Power distribution. Section 2. The detailed topography, bathymetry and roughness con-
ditions of the 2D hydraulic modeling project also including the sur-
2.3. Fully Continuous Approach (FCA) veyed river cross sections (Fig. 2) were derived by integrating the
available digital data with the results of an extensive field cam-
The significant issue of the SDH definition for the SCA scheme is paign. The final 2D flood model resolution is 20 m that is the size
that, even if the peak and volume information are generally pre- of the grid cell representing the domain topography. Upstream
served, it tends to provide unrealistic hydrographs characterized and downstream boundary conditions for the hydraulic model are
by symmetric or averaged shapes. This issue is here solved by respectively the inflow hydrograph for the first channel grid ele-
introducing a fully continuous rainfall–runoff algorithm that is ment and the uniform flow assigned to the boundary grid elements.
the proposed FCA that provides as input to the 2D hydraulic prop-
agation model a complete runoff time series (i.e. multi-event hyd- 4. Comparison test
rograph) rather than a single event SDH. In this way, the entire
physical process is represented in continuous from the hydrologic EBA, SCA and FCA are compared by means of flood maps that are
forcing (i.e. rainfall) to the river flow routing reducing the subjec- derived using the above described procedure. Although a case
tivity and uncertainty of user-derived hypotheses. Moreover, a study with available observations would be better representative
more complete dataset is available for the inundation mapping and would allow a quantitative FCA validation, the aim here is to
interpretation and in particular, a water depth time series (and provide a qualitative evaluation of the differences among the three
other hydraulic parameters like velocity, flow directions, and the procedures. The comparison is here proposed to visualize the im-
energy impact of the flood wave among others) for each node of pact of the fully continuous approach that, while neglecting the
the floodplain domain. subjectivity affecting standard EBA procedure, show its potential
The simulation of the entire runoff series is limited, today, by practical advantage for inundation mapping.
computer hardware and memory for handling this enormous The conditions and parameters used in the method implemen-
amount of data. As a result, given the entire synthetic runoff time tation are the following:
series, the 2D hydraulic model is applied only on the n maximum
annual hydrographs that are selected here using the peak dis- – The rainfall scenario is obtained simulating 500 years of a syn-
charge as driving variable and the duration as a continuous se- thetic rainfall time series at 5 min of resolution. The IDF pre-
quence of no-zero runoff values. However, this does not affect sented in the EBA is estimated directly on this rainfall
the validity and generalization of the procedure. The nth flood scenario using formula (1).
map is derived by selecting the highest simulated water depths – Properties and conditions of the simulated 500 years of 5 min
for each node and the synthetic inundation surface is characterized rainfall are reported in Grimaldi et al. (in press-a).
by water surface peaks related to different simulation time steps. – For the SCS-CN method, adopted in all three procedures for the
The final inundation map for each return period is the result of net rainfall estimation, the parameters are fixed to: k = 0.2 as
the cell-by-cell statistical analysis of the n flood maps. In order suggested in USDA-NRCS (2010), and CN = 74.6. This value is
to estimate the corresponding return period, the n water depth val- estimated averaging the CN assigned to homogeneous subareas
ues for each cell are analyzed using the mixed Gumbel distribution of the basin referring to the CORINE soil classification.
formula (Kedem et al., 1990; Yoo et al., 2005): – The time of concentration, useful either for EBA (for design hye-
tograph estimation and rainfall–runoff model application) and
GðxÞ ¼ ð1  pÞHðxÞ þ pFðxÞ ð3Þ for SCA and FCA (for only rainfall–runoff model application), is
estimated using the Giandotti formula and fixed to 4.5 h.
where x is the random variable for the flow depth for each cell, p is – The separation time Ts, useful in SCA and FCA for continuous
the probability of wet cell (x > 0), 1  p is the probability of a dry application of rainfall–runoff transformation, is set to 24 h as
event (x = 0), G(x) is the mixed distribution function considering suggested in Grimaldi et al. (2012b).
both wet and dry events, H(x) is the step function: H(x) = 0, if – Roughness conditions that is Manning values, for the hydraulic
x = 0; H(x) = 1, if x > 0, and F(x) is the probability density function
model, are estimated using literature values (Chow et al., 1988)
estimated for non-zero values. of respectively 0.04 for the main channel and in the range of
0.05–0.12 for the floodplain.
3. The Rio Torbido case study
5. Results and discussion
The Rio Torbido, a 61.67 Km2 river basin located in central Italy
draining directly into the main channel of the Tiber River, is selected The results of the application of the EBA, SCA and FCA proce-
for this case study. Elevations range from 85 m to 625 m, the average dures are included here and discussed with specific regard to the
slope is 21.9% and the maximum hydrologic distance of the outlet design hydrograph and hydraulic modeling results of steps 4 and 5.
S. Grimaldi et al. / Journal of Hydrology 487 (2013) 39–47 43

Fig. 2. Rio Torbido watershed. Case study 2D hydraulic modeling domain.

Fig. 3. Rio Torbido watershed. DEM and drainage network.

Table 1 represents the main properties of the design hydro-


graphs obtained from the EBA and SCA with information on peak Table 1
discharge (Qp), volume (V) and duration (D), and the main statisti- Summary of design hydrograph characteristics obtained with EBA, SCA, and FCA
approaches.
cal properties (minimum, maximum, average, standard deviation
and quantiles at 25% (q25) and at 75% (q75)) of the 500 annual ‘‘Event Based Approach (EBA)’’
maximum FCA hydrographs. For the FCA, the Qp, V and D values T (years) 2 5 10 20 50 100
Qp (m3/s) 16.6 46.4 68.6 95.9 146.5 186.0
are not summarized using the return time since V and D are not 3 4
V (m  10 ) 16.3 45.6 66.8 93.0 140.9 178.0
necessarily maximum annual values . D (h) 6.13 6.7 6.7 6.8 7.0 7.1
The EBA is characterized by peak discharges that range from
‘‘Semi Continuous Approach (SCA)’’
16.6 m3/s to 186 m3/s, while the SCA ranges from 23.2 m3/s to T (years) 2 5 10 20 50 100
210.5 m3/s with an average difference of 34.1%. The FCA peak dis- Qp (m3/s) 23.2 65.2 99.1 140.8 175.7 210.5
charges go from 0 m3/s to 434.8 m3/s. While it seems that the var- V (m3  104) 37.5 105.3 159.9 227.2 283.5 339.6
iability associated with the FCA peak discharges is higher than the D (h) 15.2 17.3 19.0 21.0 22.8 24.5

EBA and SCA, the differences in the quantiles of the peak distribu- ‘‘Fully Continuous Approach (FCA)’’
tion are limited. min q25 average q75 max St. dev.

The comparison of volume and duration of the simulated hydro- Qp (m3/s) 0 5.7 38.9 56.7 434.8 47.7
graphs shows more significant differences: EBA and SCA range V (m3  104) 0.0 7.2 63.4 81.3 526.9 86.0
D (h) 0 7.2 13.8 18.8 49.1 9.5
from 163.017 m3 to 1.780.300 m3 respectively and from
44 S. Grimaldi et al. / Journal of Hydrology 487 (2013) 39–47

374.929 m3 to 3.395.933 m3 with an average percentage difference


of 122.8%. This is consistent with previous studies (Grimaldi et al.,
2012b) demonstrating that the SDH duration is approximately
twice as much as the design hydrograph obtained with EBA. The
FCA results also show the volume and duration differences for
EBA and SCA.
Figs. 4 and 5 illustrate the design hyetographs and the design
hydrographs related to the EBA for the corresponding return time
T, while in Fig. 6, the SDHs for the SCA are reported. Fig. 7 repre-
sents the maximum annual hydrograph related to the 0.98 peak
quantile obtained with FCA.
The interpretation of results for the design hydrographs clearly
shows a significant difference between the EBA and the SCA/FCA
schemes in terms of peak and, even more predominantly, event
volume, while the proposed FCA is able to preserve the realistic Fig. 5. Design hydrographs estimated with the EBA approach.

shape of the hydrologic forcing.


An example of flood inundation results is inserted in Fig. 8 that
include the maximum floodplain flow depth that is the maximum
height of the flow above the top of the channel banks for the design
event (EBA and SCA) or for the entire simulation time (FCA). Fig. 8
includes, in fact, the three flood maps corresponding to the routing
of the EBA, SCA and FCA design hydrograph along with the selected
hydraulic modeling domain. Differences of the three flooding sce-
narios are quite significant. Note that some areas are not associated
to any flow depth meaning that the channel is sufficient to convey
the routed flow in that area. In particular, the EBA underestimates
the inundated area with respect to the SCA and FCA that provide
similar results. Table 2 summarizes results for the three applica-
tions in quantitative terms and for all the return periods with the
corresponding flood area and volume. In Table 3, the percentage
differences among the three approaches are shown confirming
Fig. 6. Design hydrographs estimated with the SCA approach.
the major differences for the EBA with respect to the SCA and
FCA. These flood areas and volumes are sensibly greater than the
standard EBA case. In particular, in Table 3, it is noted that the dif-
ferences between the SCA and FCA are limited to a maximum of
16.9% for flood volumes and 5.6% for flood areas. This was esti-
mated by analyzing the floodplain and not the channel. While dif-
ferences between the SCA and FCA with respect to the EBA go up to
a maximum of 201.9% for flooded volumes and 127.8% for flooded
areas.
Results show that inundation mapping based on a design hyd-
rograph using EBA should be used with caution since the flood vol-
ume may be significantly underestimated. This is particularly
important when the extra-fluvial process forced by extreme hydro-
logic conditions requires 2D unsteady hydraulic model applica-
tions whose results are significantly impacted by the hydrograph
shape and volume rather than just the flood peak. While the pres-
ent work shows differences between the design hydrograph de- Fig. 7. Example of hydrograph estimated with the FCA approach.
rived from the SCA and the proposed FCA, the corresponding

flood maps seem to be characterized by similar results. However,


this could be due to either the simple flood modeling domain that
is characterized by a large flat area limited by a well defined break
in slope and by the fact that the hydrograph shape does not mean-
ingfully influence the inundation dynamics.
In any case, results show the accomplishment of the two goals
of the present work: (1) to underline differences in the methodol-
ogies for design hydrograph estimation using the standard EBA
with respect to the SCA and FCA, keeping in mind that EBA is still
used as the standard method by professionals. In fact, EBA is opti-
mized to estimate the flood peak discharge as the main input for
steady 1D hydraulic models without paying much attention to
the other flood hydrograph components. In particular, it indicates
that the volume, shape and duration govern the modeling results
of 2D hydraulic modeling applications. (2) to present a simple
Fig. 4. Design hyetographs estimated with the EBA approach. and parsimonious procedure for hydrologic forcing characteriza-
S. Grimaldi et al. / Journal of Hydrology 487 (2013) 39–47 45

Fig. 8. Water depth distribution in selected inundation domain related to 20-years return period: (a) EBA approach; (b) SCA approach; (c) FCA approach.

tion reducing at best the conceptual hypotheses and the corre- investigated. By taking into consideration small and ungauged riv-
sponding uncertainties that significantly impact the EBA, and also er basins that usually requires subjective and empirical approaches
in part, the SCA methods. for rainfall–runoff modeling, the following three methodologies
have been selected and compared with specific regard to the even-
6. Conclusions tual advantages of the FCA.
The so-called event-based approach, or EBA, is based on a syn-
In this paper three hydrograph estimation procedures for defin- thetic representation of both the rainfall input (i.e. IDF method)
ing the design hydrologic input of flood modeling and mapping are and runoff output for hydrograph characterization. The EBA has
46 S. Grimaldi et al. / Journal of Hydrology 487 (2013) 39–47

Table 2
Summary of flood volumes and areas for six return period values estimated for the EBA, SCA and FCA.

T (years) 2 5 10 20 50 100
‘‘Event Based Approach (EBA)’’
V (floodplain and channel) (m3  103) 69.8 115.6 142.9 182.2 269.1 341.5
V (floodplain)(m3  103) 0.0 2.0 7.6 23.6 80.0 134.7
A (floodplain and channel) (m2  103) 34.8 42.0 58.8 80.8 130.4 158.0
A (floodplain) (m2  103) 0.0 7.2 24.0 46.0 95.6 123.2
‘‘Semi Continuous Approach (SCA)’’
V (floodplain and channel) (m3  103) 76.0 134.5 175.5 236.2 297.7 365.7
V (floodplain) (m3  103) 0.0 5.2 20.4 56.6 101.2 153.5
A (floodplain and channel) (m2  103) 34.8 51.2 77.6 122.8 143.6 160.8
A (floodplain) (m2  103) 0.0 16.4 42.8 88.0 108.8 126.0
‘‘Fully Continuous Approach (FCA)’’
V (floodplain and channel) (m3  103) 77.4 136.1 180.7 249.8 310.3 382.9
V (floodplain) (m3  103) 0.0 5.6 23.0 66.1 110.7 166.9
A (floodplain and channel) (m2  103) 34.8 51.2 80.0 126.4 147.2 165.2
A (floodplain) (m2  103) 0 16.4 45.2 91.6 112.4 130.4

strating the expected different behavior on the three approaches


Table 3
Summary of flood volume and area percentage differences for the EBA, SCA and FCA. while using a topographic resolution of common use for practical
flood mapping in ungauged basins. More specifically, results show
T (years) 2 5 10 20 50 100
that the three methodologies provide significantly different out-
‘‘Semi continuous’’ – ‘‘event based’’ puts with major differences characterizing the EBA with respect
DV (%) (floodplain and 8.9 16.4 22.8 29.6 10.6 7.1
to the SCA and FCA. Those differences confirm that the EBA should
channel)
DV (%) (floodplain) – 162.9 167.8 139.3 26.4 14.0 not be used for flood modeling applications that are mainly gov-
DA (%) (floodplain and 0.0 21.9 32.0 52.0 10.1 1.8 erned by the extra-channel physical processes and dynamics. For
channel) these cases the hydrograph shape, duration and volume play the
DA (%) (floodplain) – 127.8 78.3 91.3 13.8 2.3 major role rather than just the flood peak. This is always the case
‘‘My continuous’’ – ‘‘event based’’ of high return time flood mapping studies that require an accurate
DV (%) (floodplain and 10.9 17.7 26.5 37.1 15.3 12.1 and detailed characterization of the input hydrograph and the
channel)
DV (%) (floodplain) – 183.3 201.9 179.7 38.3 23.9
application of a 2D hydraulic model. It is also shown that the dif-
DA (%) (floodplain and 0.0 21.9 36.1 56.4 12.9 4.6 ferences in design hydrograph estimation of the SCA with respect
channel) to the FCA are not transferred into the corresponding flood map-
DA (%) (floodplain) – 127.8 88.3 99.1 17.6 5.8 ping, but this is reasonable considering that the two approaches
‘‘Fully continuous’’ – ‘‘semi continuous’’ differ only for the hydrograph shape.
DV (%) (floodplain and 1.8 1.1 3.0 5.8 4.2 4.7 The described proof of concept suggests to prefer the FCA pro-
channel)
cedure since it focuses on the implementation of a more realistic
DV (%) (floodplain) – 7.7 12.7 16.9 9.4 8.7
DA (%) (floodplain and 0.0 0.0 3.1 2.9 2.5 2.7 representation of the rainfall–runoff process by means of a simple,
channel) parsimonious model that could be easily implemented by profes-
DA (%) (floodplain) – 0.0 5.6 4.1 3.3 3.5 sionals for practical flood modeling and mapping applications in
small and ungauged river basins.
been the most simple and commonly applied procedure for dec- Future research shall validate the FCA on a small watershed
ades, but it is severely impacted by conceptual hypotheses, where flood maps as well as runoff observations shall be available
assumptions, related errors and uncertainties in hydrologic terms. with the aim of investigating the potential use of such method for
The more recent Semi Continuous Approach, or SCA, overcomes gauged basins. As a result several DEMs at different resolution
most of the conceptual hypotheses of the EBA and introduces the should be selected to evaluate the topographic data impact on
use of a simulated rainfall time series by which the rainfall–runoff model performances. Moreover, the validation, using a DEM with
model is applied using a continuous implementation. This allows proper resolution and precision, (<5 m of resolution and <1 m of
the extrapolation of a design hydrograph from the runoff synthetic precision), shall be also developed. Finally, since the proposed
time series. This SCA also removes the conceptual steps of the EBA FCA method provides flow height and velocity time series for each
method in the rainfall input estimation, but it is still predomi- cell of the inundated domain, a specific statistically-based post
nantly synthetic in the extrapolation of the corresponding processing analysis shall be implemented for the estimating the
hydrograph. uncertainty of flood inundation maps with the final aim of provid-
The proposed Fully Continuous Approach, or FCA, neglects any ing to floodplain managers and urban planners a detailed decision
preliminary conceptualization of the hydrologic modeling proce- making framework.
dure for design hydrograph estimation by implementing a method
that provides, as input to the hydraulic model, the entire simulated
runoff time series. In this way, there is no conceptualization until Acknowledgments
the last step of flood modeling and mapping analysis. Furthermore,
it performs the last necessary analytical and statistical evaluation, This research was partially supported by the Honors Center of
without prior simplification, of the hydraulic modeling results to Italian Universities.
identify the flood inundation map and the corresponding design Authors are particularly grateful to Dr. Francesco Serinaldi for
flow depths and velocities for each cell of any given return period. his contribution in the statistical analysis of the inundation maps,
Even if the selected coarse resolution DEM is not appropriate for to the eponymous reviewer, Dr. Giuliano Di Baldassarre, and to the
a detailed quantitative evaluation, it is appositely used for demon- anonymous reviewer for their useful comments.
S. Grimaldi et al. / Journal of Hydrology 487 (2013) 39–47 47

References Kalyanapu, A., Judi, D., McPherson, T., Burian, S., 2012. Monte Carlo-based flood
modelling framework for estimating probability weighted flood risk. J. Flood
Risk Manage. 5, 37–48. http://dx.doi.org/10.1111/j.1753-318X.2011.01123.x.
Alfieri, L., Laio, F., Claps, P., 2008. A simulation experiment for optimal design
Kedem, B., Chiu, L.S., Karni, Z., 1990. An analysis of the threshold method for
hyetograph selection. Hydrol. Process. 22, 813–820.
measuring area-average rainfall. J. Appl. Meteorol. 29 (1), 3–20.
Aronica, G., Bates, P.D., Horritt, M.S., 2002. Assessing the uncertainty in distributed
Keifer, C.J., Chu, H.H., 1957. Synthetic storm pattern for drainage design. ASCE, J.
model predictions using observed binary pattern information within GLUE.
Hydraulics Div. 83(HY4), 1–25.
Hydrol. Process. 16, 2001–2016.
Laforce, S., Simard, M.-C., Leconte, R., Brissette, F., 2011. Climate change and
Bates, P.D., Horritt, M.S., Aronica, G., Beven, K., 2004. Bayesian updating of flood
floodplain delineation in two southern quebec river basins. J. Am. Water Resour.
inundation likelihoods conditioned on flood extent data. Hydrol. Process. 18,
Assoc. 47 (4), 785–799.
3347–3370.
McMillan, H.K., Brasington, J., 2008. End-to-end flood risk assessment: a coupled
Blazkova, S., Beven, K., 2009. A limits of acceptability approach to model evaluation
model cascade with uncertainty estimation. Water Resour. Res. 44, W03419.
and uncertainty estimation in flood frequency estimation by continuous
http://dx.doi.org/10.1029/2007WR005995.
simulation: Skalka catchment, Czech Republic. Water Resour. Res. 45, W00B16.
Merwade, V., Olivera, F., Arabi, M., Edleman, S., 2008. Uncertainty in flood
Calver, A., Stewart, E., Goodsell, G., 2009. Comparative analysis of statistical and
inundation mapping: current issues and future directions. J. Hydrol. Eng. 13
catchment modelling approaches to river flood frequency estimation. J. Flood
(7), 608–620.
Risk Manage. 2, 24–31.
Moretti, G., Montanari, M., 2008. Inferring the flood frequency distribution for an
Chow, V.T., Maidment, D.R., Mays, L.W., 1988. Applied Hydrology. Mc-Graw-Hill,
ungauged basin using a spatially distributed rainfall–runoff model. Hydrol.
New York.
Earth Syst. Sci. 12, 1141–1152.
European Commission. 2000. CORINE (Coordination of Information on
Nardi, F., Vivoni, E.R., Grimaldi, S., 2006. Investigating a floodplain scaling relation
Environment) Database, a Key Database for European Integrated
using a hydrogeomorphic delineation method. Water Resour. Res. 42, W09409.
Environmental Assessment. Programme of the European Commission,
http://dx.doi.org/10.1029/2005WR004155.
European Environmental Agency (EEA).
Nardi, F., Grimaldi, S., Santini, M., Petroselli, A., Ubertini, L., 2008. Hydrogeomorphic
Di Baldassarre, G., Castellarin, A., Montanari, A., Brath, A., 2009. Probability
properties of simulated drainage patterns using DEMs: the flat area issue.
weighted hazard maps for comparing different flood risk management
Hydrol. Sci. J. 53 (6), 1176–1193.
strategies: a case study. Nat. Hazards 50 (3), 479–496.
Nishat, S., Guo, Y., Baetz, B.W., 2010. Antecedent soil moisture conditions of
Di Baldassarre, G., Schumann, G., Bates, P.D., Freer, J.E., Beven, K.J., 2010. Flood-plain
different soil types in South-western Ontario, Canada. Hydrol. Process. 24,
mapping: a critical discussion of deterministic and probabilistic approaches.
2417–2424.
Hydrol. Sci. J. 55 (3), 364–376.
Nnadi, F.N., Kline, F.X., Wary, H.L., Wanielista, M.P., 1999. Comparison of design
Dodov, B., Foufoula-Georgiou, E., 2006. Floodplain morphometry extraction from a
storm concepts using continuous simulation with short duration storms. J. Am.
high resolution digital elevation model: a simple algorithm for regional analysis
Water Resour. Assoc. 31 (1), 61–85.
studies. IEEE Geosci. Remote Sens. Lett. 3 (3), 410–413.
NRCS. 1997. Ponds-Planning, Design, Construction. Agriculture Handbook N. 590.
European Parliament. 2007. Directive 2007/60/Ec of the European Parliament
US Natural Resources Conservation Service, Washington DC.
and of the Council of 23 October 2007 on the Assessment and Management
O’Brien, J.S., Julien, P.Y., Fullerton, W.T., 1993. Two-dimensional water flood and
of Flood Risks. <http://ec.europa.eu/environment/water/flood_risk/key_docs.
mudflow simulation. J. Hydraulic Eng. 119 (2), 244–261.
htm#Directive> (accessed 15.03.2011).
Pappenberger, F., Matgen, P., Beven, K.J., Henver, J.B., de Pfister, L., Fraipont, P., 2006.
FLO-2D. 2012. FLO-2D Reference Manual, <https://www.flo-2d.com/downloa/>.
Influence of uncertainty boundary conditions and model structure on flood
Giandotti, M., 1934. Previsione delle piene e delle magre dei corsi d’acqua. vol. 8,
inundation predictions. Adv. Water Resour. 29, 1430–1449.
Istituto Poligrafico dello Stato. pp. 107–117.
Pramanik, N., Panda, R.K., Sen, D., 2010. Development of design flood hydrographs
Gilles, D., Young, N., Schroeder, H., Piotrowski, J., Chang, Y.-J., 2012. Inundation
using probability density functions. Hydrol. Process. 24 (4), 415–428.
mapping initiatives of the iowa flood center: statewide coverage and detailed
Rahman, A., Weinmann, P.E., Hoang, T.M.T., Laurenson, E.M., 2002. Monte Carlo
urban flooding analysis. Water 4, 85–106.
simulation of flood frequency curves from rainfall. J. Hydrol. 256, 196–210.
Grimaldi, S., Teles, V., Bras, R.L., 2004. Sensitivity of a physically-based method of
Sampson, C.C., Fewtrell, T.J., Duncan, A., Shaad, K., Horritt, M.S., Bates, P.D., 2012.
terrain interpolation to variance of input data and to the constraint of an
Use of terrestrial laser scanning data to drive decimetric resolution urban
enforced surface drainage direction. Earth Surf. Process. Land. 29 (5), 587–597.
inundation models. Advances in Water Resources, 41, 1–17, ISSN 0309-1708.
Grimaldi, S., Serinaldi, F., Napolitano, F., Ubertini, L., 2005. A 3-copula function
10.1016/j.advwatres.2012.02.010.
application for design hyetograph analysis. IAHS-AISH Publ. 293, 203–211.
Santini, M., Grimaldi, S., Nardi, F., Petroselli, A., Rulli, M.C., 2009. Pre-processing
Grimaldi, S., Nardi, F., Di Benedetto, F., Istanbulluoglu, E., Bras, R.L., 2007. A
algorithms and landslide modelling on remotely sensed DEMs. Geomorphology
physically based method for removing pits in digital elevation models. Adv.
113 (1–2), 110–125.
Water Resour. 30, 2151–2158.
Schertzer, D., Lovejoy, S., 1987. Physical modeling and analysis of rain and clouds by
Grimaldi, S., Petroselli, A., Nardi, F., Alonso, G., 2010. Flow Time estimation with
anisotropic scaling of multiplicative processes. J. Geophys. Res. 92, 9693–9714.
variable hillslope velocity in ungauged basins. Adv. Water Resour. 33 (10),
Serinaldi, F., 2009. A multisite daily rainfall generator driven by bivariate copula-
1216–1223.
based mixed distributions. J. Geophys. Res. 114, D10103. http://dx.doi.org/
Grimaldi, S., Petroselli, A., Nardi, F., 2012a. A parsimonious geomorphological unit
10.1029/2008JD011258.
hydrograph for rainfall–runoff modelling in small ungauged basins. Hydrol. Sci.
Serinaldi, F., 2010. Multifractality, imperfect scaling and hydrological properties of
J. 57 (1), 73–83. http://dx.doi.org/10.1080/02626667.2011.636045.
rainfall time series simulated by continuous universal multifractal and discrete
Grimaldi, S., Petroselli, A., Serinaldi, F., 2012b. A continuous simulation model for
random cascade models. Nonlin. Processes Geophys. 17, 697–714. http://
design-hydrograph estimation in small and ungauged watersheds. Hydrol. Sci. J.
dx.doi.org/10.5194/npg-17-697-2010.
57 (6), 1–17.
Serinaldi, F., Grimaldi, S., 2011. Synthetic design hydrographs based on distribution
Grimaldi, S., Petroselli, A., Tauro, F., Porfiri, M., 2012c. Time of concentration: a
functions with finite support. Journal of Hydrologic Engineering, 16, (5) 434–
paradox in modern hydrology. Hydrol. Sci. J. 57 (2), 217–228. http://dx.doi.org/
446; doi:10.1061/(ASCE)HE.1943-5584.0000339.
10(1080/02626667), 2011, 644244.
Soczyñska, U., Nowicka, B., Somorowska, U., Kupczyk, E., Suligowski, R., 1997.
Grimaldi, S., Petroselli, A., Serinaldi, F., in press-a. Design hydrograph estimation in
Prediction of design storms and floods. IAHS-AISH Publ. 246, 297–303.
small and ungauged watersheds: continuous simulation method versus event-
Tarboton, D.G., Ames, D.P., 2001. Advances in the Mapping of Flow Networks from
based approach. Hydrological Processes. http://dx.doi.org/10.1002/hyp.8384.
Digital Elevation Data. World Water and Environmental Resources Congress,
Grimaldi, S., Petroselli, A., Romano, N., in press-b. Green-ampt curve number mixed
Orlando, Florida, May 20–24, ASCE.
procedure as an empirical tool for rainfall–runoff modelling in small and
Tarboton, D.G., Bras, R.L., Rodriguez-Iturbe, I., 1991. On the extraction of channel
ungauged basins. Hydrological Processes. http://dx.doi.org/10.1002/hyp.9303.
networks from digital elevation data. Hydrol. Process. 5 (1), 81–100.
Grimaldi, S., Petroselli, A., Romano, N., in press-c. Curve-Number/Green-Ampt
USDA-SCS. 1986. Urban Hydrology for Small Watersheds. Technical Release 55.
mixed procedure for streamflow predictions in ungauged basins: parameter
Washington, DC.
sensitivity analysis. Hydrological Processes. http://dx.doi.org/10.1002/
USDA-NRCS. 2010. National Engineering Handbook, Part 630 – Hydrology. <http://
hyp.9749.
www.info.usda.gov/viewerFS.aspx?hid=21422>.
Hall, J.W., Tarantolo, S., Bates, P.D., Horritt, M., 2005. Distributed sensitivity analysis
Verhoest, N.E.C., Vandenberghe, S., Cabus, P., Onof, C., Meca-Figueras, T.,
of flood inundation model calibration. J. Hydraulic Eng. 131 (2), 117–126.
Jameleddine, S., 2010. Are stochastic point rainfall models able to preserve
Horritt, M.S., Di Baldassarre, G., Bates, P.D., Brath, A., 2007. Comparing the
extreme flood statistics? Hydrol. Process. 24, 3439–3445.
performance of a 2-D finite element and a 2-D finite volume model of
Viglione, A., Bloschl, G., 2009. On the role of storm duration in the mapping of
floodplain inundation using airborne SAR imagery. Hydrol. Process. 21, 2745–
rainfall to flood return periods. Hydrol. Earth Syst. Sci. 13, 205–216.
2759. http://dx.doi.org/10.1002/hyp. 6486.
Viglione, A., Merz, R., Bloschl, G., 2009. On the role of the runoff coefficient in the
Hsieh, L.-S., Hsu, M.-H., Li, M.-H., 2006. An assessment of structural measures for
mapping of rainfall to flood return periods. Hydrol. Earth Syst. Sci. 13, 577–593.
flood-prone lowlands with high population density along the Keelung River in
Yoo, C., Jung, K.S., Kim, T.W., 2005. Rainfall frequency analysis using a mixed
Taiwan. Nat. Hazards 37 (1–2), 133–152.
Gamma distribution: Evaluation of the global warming effect on daily rainfall.
IGMI (Italian Geographic Military Institute). 2003. Raster (Matrix) Numerical DEM
Hydrol. Process. 19 (19), 3851–3861.
of Italy (Internal Factsheet, in Italian). <http://www.igmi.org/pdf/
info_matrix2003.pdf>.

You might also like