Journal of Applied Water Engineering and Research

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tjaw20

The design flood under two approaches: synthetic

storm hyetograph and observed storm hyetograph

Catine António Chimene & José Nilson B. Campos

To cite this article: Catine António Chimene & José Nilson B. Campos (2020) The design flood
under two approaches: synthetic storm hyetograph and observed storm hyetograph, Journal of
Applied Water Engineering and Research, 8:3, 171-182, DOI: 10.1080/23249676.2020.1787242

To link to this article: https://doi.org/10.1080/23249676.2020.1787242

Published online: 01 Jul 2020.

Submit your article to this journal

Article views: 26

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tjaw20
Journal of Applied Water Engineering and Research, 2020
Vol. 8, No. 3, 171–182, https://doi.org/10.1080/23249676.2020.1787242

The design flood under two approaches: synthetic storm hyetograph and observed storm
hyetograph
Catine António Chimenea,b∗ and José Nilson B. Camposb
a Higher School of Rural Development, Eduardo Mondlane University, Maputo, Mozambique; b Departamento de Engenharia
Hidráulica e Ambiental, Universidade Federal do Ceará, Fortaleza, CE, Brazil
(Received 21 October 2018; accepted 11 May 2020 )

This paper evaluates the design peak flow for several return periods using two approaches: 1) a flood frequency analysis;
2) rainfall-runoff model applied to a design storm. The design storm originates from the design rainfall depth, which is
distributed throughout the rainfall duration via synthetic hyetographs. In the flood frequency analysis, the design peak
flow was obtained from a 30-year series of flows estimated from rainfall observations at a five-minute time step. The
design storm was routed with the same rainfall-runoff model and parameters to yield the design peak flow. The hyetographs
tested were categorized as having advanced, delayed, alternating, uniform and empirical blocks that were built from rainfall
observations. All design storms analysed undersized the hydraulic structures. The results, which were valid for the study
region, indicated the need to rethink the concept of design storms, which associate the design storm depth with synthetic
hyetographs, for a conservative standard.
Keywords: Design flood; design storm; syntheti hyetographs; HEC-HMS

Table i. Simbology used in this paper.

Candela et al. 2014; Gericke and Smithers 2014; Dalezios
and Eslamian 2016). Methods vary from finding flood fre-
Symbols Description
quencies based on the statistical analysis of observed or
A2 Anderson-Darling Statistics
estimated river flows to design storm models (Saghafian
Q Discharge or direct flow
et al. 2014). Flood frequency or risk analysis for flood haz-
DFH Design Flood Hydrograph
ards is an important tool of the engineering practice whose
DRD Design Rainfall Depth
aim is to obtain relationships between design variables cor-
Prec Precipitation
responding to a chosen hydrologic risk (Wang et al. 2015;
PFE Peak Flow Error
Sraj et al. 2016). Equating the probability of design storms
PFFA Peak Flow obtained from Frequency
to the probability of design floods has been a challenge for
Analysis
many decades (Beran 1973).
PFDS Peak Flow obtained from Design Storm
In engineering practices, establishing a consistent
Tr Return Period
design flood hydrograph (DFH ) is a way of ensuring that
Tp Time to peak
the infrastructure meets the proper design standards. A
tc Time of concentration
DFH can be obtained from flow records or precipitation
data associated with a rainfall-runoff model. The choice
1. Introduction between one method or another depends on engineering
Reliable design flood estimations for a given return period judgement after analysing all available data (Xiao et al.
(Tr) remain a problem for the sizing of hydraulic struc- 2009).
tures, such as spillways, bridges, and urban drainage sys- During the application of a flood frequency analysis,
tems. The concept of a design flood has been thoroughly hydrologists usually focus on peak flows. However, it is
revised with time due to the result of new hydrological important to also consider the variables of flood volume
knowledge and advanced mathematical modelling capa- and duration. Therefore, the bivariate statistical analy-
bilities. The evolution of associated concepts, practices, sis, peak flow and volume have more reliability in sizing
and methods is present in the scientific literature (Watt reservoir spillways (Mediero et al. 2010). In Europe, the
and Marsalek 2013; Zhou et al. 2013; Ball and Ara 2014; legislation on flood risk assessments for infrastructure with

*Corresponding author. Email: [email protected]

© 2020 IAHR and WCCE

172 C. A. Chimene and J. N. B. Campos

a Tr ranging from 50 to 500 has established that fre- If the PFE is close to zero, then the design storm
quency analyses are the method to be used (Botero and method (for the adopted synthetic hyetograph) correctly
Francés 2010). The statistical method has been compared sizes the hydraulic structure. Otherwise, if this error is less
with the rainfall-runoff method in a data rich catchment than zero, the design storm method undersizes the water
in southern China. For the simulated discharges, they structure. On the other hand, for errors greater than zero,
used two models: the Xinanjiang (XAJ) and Hydrolo- the design storm method oversizes the water structure.
giska Vattenbalansadelning (HBV). The results showed This paper studies five synthetic hyetographs: alternating
that both models yielded acceptable results. The maxi- blocks hyetograph, advanced blocks hyetograph, delayed
mum error was 11% for a Tr equal to 100 years. The blocks hyetograph, empirical hyetograph built from the
XAJ model oversized, whereas the HBV model under- most critically observed rainfall, and a classic uniform
sized the design flood (Zeng et al. 2016). Okoli et al. hyetograph. The focus of this paper is to analyse the differ-
(2019) used two different methods for the estimation of ences between two approaches (a flood frequency analysis
design floods such as statistical method and hydrological and design storm) in a modelled river basin. In doing so,
concepts (event-based simulation “EBS” and continuous we also analyse the hydraulic structure safety that is sized
simulation “CS”). by the design storm method. All of the basic data used for
Nevertheless, statistical inferences from the observed the two approaches are the same, including the parameters
river flows have some difficulties. First, in many places of the modelled watershed. This is not a case study for the
of the world, there are no reliable long-term records of basin itself but a comparative study between the two meth-
discharges to statistically extrapolate and estimate floods ods. Therefore, the entire watershed utilizes the parameters
with large return periods. Second, the hydrological char- for the hydrologic model and gives an idea on the validity
acteristics of river basins and land and soil covers are of the results.
normally modified during the period of the flow record and
keep changing after structure construction. These changes
have significant impacts on the peak flow (Barasa and 2. Data and methods
Perera 2017; Salami et al. 2017). As water engineering
2.1. Characterization of study area & modelled
projects are performed to the future, rainfall-runoff mod-
watershed
elling can permits the improved representation of the future
watersheds. The Sítios Novos River basin is located at the geograph-
The physical characteristics of the study region and ical coordinates − 3°46’22.36” E and − 38°57’31.29”
the observed rainfall series have played important roles N, which lies along the coast of Ceará (Caucaia) in an
in the development of design storms. The local hydro- area of Fortaleza in Brazil, provides the rainfall-runoff
logical signature can be very important for translating model parameters. The watershed controls an area equal
rainfall return periods into responsive return periods for to 446 km2 , average elevation (H) 552 m, travel length
river flows (Knighton and Walter 2016). The application (L) 51.43 km and slope (S) 0.01063 m/m. The reservoir
of the design storm method with an alternating block capacity is 126 million cubic metres, and the lake has a
hyetograph is very applicable in technical and scientific lit- surface equal to 2,010 hectares. It has a territorial area of
erature (Sordo-Ward et al. 2014; Chin 2017; Yang et al. 1,293 km², which represents 0.82% of the area of the state
2018) (PDDU 1998). The region presents the characteristics of a
The method applied here for evaluating the safety of coastal region, with a typical vegetation cover composed of
the design storm method, in comparison with the flood fre- mangroves and dense, shrubby caatinga, which have soils
quency analysis, consists of using the same rainfall-runoff that are classified as planosols and solodic planosols. The
model and parameters as those of river basins to obtain mean annual precipitation is approximately 950 mm and
the peak flow. The peak flow from the frequency anal- is concentrated during the period from January to May.
ysis (PFFA) derives from a series of maximum annual The mean temperature varies from 26 °C to 30°C (Paulino
flow yielded by a rainfall-runoff model, which is applied 2008). Figure 1.
to a series of precipitation records over short time-interval. Recently, Companhia de Gestão dos Recursos Hídricos
The peak flow from design storms (PFDS) derives from do Ceará (COGERH) developed a study for evaluating the
running the design storm in the rainfall-runoff model for hydrological safety of a spillway. The procedure applied
each Tr. design storms with a duration equal to 24 h (see Sule and
The safety of the design storm method in terms of siz- Alabi 2013; Yannopoulos et al. 2015; Duka et al. 2017; Ilić
ing hydraulic structures for each synthetic hyetograph and et al. 2018) and a temporal distribution according to the
Tr is analysed by the peak flow error (PFE), as shown in alternating block hyetograph. The models used were the
Equation 01: NRCS-CN (for the loss function) and the NRCS dimen-
sionless unit hydrograph (for the transformation function)
because of its accessibility and easy handled with the local
(PFDS − PFFA) data provided. Some authors such as Ponce and Hawkins
PFE = 100 (1)
PFFA
 Journal of Applied Water Engineering and Research 173

Figure 1. Sítios Novos river basin localization.

(1996) emphasize that the SCS-CN method has had appli- elements in a basin to model the river discharge. The
cations susceptible to basins up to 250 km2 and in case of HEC-HMS software has the necessary tools and models
superiority there should be a subdivision into subbasins. to satisfy this objective. In summary, we have just one
However, there are works developed in areas larger than conceptual basin and the two following procedures to esti-
250 km2 and having reached approximately 2000 km2 ; mate the design peak discharge: 1) applying the statistical
such as Askar 2014; Tailor and Shrimali 2016; Stuti 2017; method to a series of maximum yearly peak discharges
Matomela et al. 2019. obtained via rainfall-runoff modelling using the observed
precipitation; and 2) applying the statistical method to the
maximum daily precipitation to obtain the design storm,
2.2. Rainfall Data which is associated with several synthetic hyetographs and
In the area near the basin, there are 30 years (from 1970 to rainfall-runoff models and results in the design storm peak
1999 see Elhassnaoui et al. 2019) of available precipitation flow.
data at a temporal scale of five minutes, which are obtained To obtain the PFFA, the following steps were uti-
via pluviography at the climate station at the Universidade lized: 1) the 30-year series (1970–1999) of five-minute
Federal do Ceará. More recent records are not available. precipitation was organized in an Excel sheet, 2) the
These data are the only rainfall data available at the sub- most critical precipitation each year was simulated via
daily scale, with a five minutes time step, in this area. the rainfall-runoff model to obtain the maximum peak
This set of data was used to build the intensity-duration discharge, 3) a 30-year series of the maximum daily dis-
frequency (idf ) for Fortaleza (Equation 2), which was used charge was obtained, 4) this series was statistically studied
to build the synthetic hyetographs (Silva et al. 2013): using Minitab to select the best probability density func-
tion (pdf ), and 5) the return period for the peak discharge
2345.29TR0.173
i= (2) was determined using the best pdf.
(t + 23.81)0.904 To obtain the PFDS, the following steps were utilized:
where Tr – return period; i – average rainfall intensity in 1) organizing the 30-year series of precipitation in an Excel
mm/hour for a given storm duration; and t – storm duration sheet; 2) establishing a 30-year series of maximum annual
in minutes. daily precipitation; 3) analysing the statistical series of
maximum daily precipitation using the Anderson–Darling
criteria to find the best pdf ; 4) using the best pdf , the
3. Methods design rainfall depth (DRD) was estimated for the selected
The method consists of adopting, via the two approaches, return period; 5) by associating the DRD to the synthetic
the same representation of hydrological processes and hyetographs, a group of design precipitation hyetographs
174 C. A. Chimene and J. N. B. Campos

(DPHs) was obtained; 6) the DPH s were routed in the rainfall becomes flow through the use of the unit hydro-
modelled watershed using the rainfall-runoff model via the graph. The CN parameter defines the characteristics of land
HMS software. use and cover, soil physical and hydrological conditions
and antecedent moisture conditions (Oliveira et al. 2016;
Tailor and Shrimali 2016; Stuti 2017; Valle Junior et al.
3.1. The rainfall-runoff model 2019) and can be found in tables developed by NRCS of
The software HEC-HMS was the tool used to transform the US Departament of Agriculture (Ilić et al. 2018).
rainfall into runoff. For the hydrological sub-basin, two The loss model, NRCS-CN, assumes that for a given
hydrological processes were applied to obtain the resul- watershed throughout the duration of a rain event, the ratio
tant hydrograph: the loss and transform processes. For the of the effective precipitation (Pe), that represents the vol-
loss process, we used the traditional NRCS curve number ume of direct flow, compared to the total rainfall volume
model. For the transform process, the NRCS dimension- (P) is equal to the ratio of the accumulated infiltration vol-
less unit hydrograph was used. The NRCS-CN method ume until that period of time compared to the potential
establishes the empirical relationship between the basin’s maximum retention volume (S). The initial abstraction (Ia)
storage capacity and the Curve Number index where actual is equal to 0.2 S, and the Pe is shown by Equation 3:
⎡
(P − Ia )2
⎢ if (P − Ia ) > 0
Table 1. Maximum annual peak flow obtained from
Pe = ⎣ P + 0.8S (3)
observed precipitation via the rainfall-runoff model. 0 elsewhere
Peak flow Peak flow Peak flow where Ia = initial abstraction (mm); P = total rainfall
Year (m3 .s−1 ) Year (m3 .s−1 ) Year (m3 .s−1 )
depth (mm) at given time step; Pe = effective rainfall
1970 224.1 1980 589.4 1990 288.6 (mm) at given time step; and S = potential maximum soil
1971 1463.8 1981 622.4 1991 862.4 retention (mm).
1972 224.2 1982 668.1 1992 380.5 The S value can be obtained from Equation. 4:
1973 487.6 1983 578.9 1993 672.3
1974 1085.6 1984 642.6 1994 636.2 25400
1975 240.9 1985 720.6 1995 1896.1 S= − 254 (4)
1976 498.3 1986 755.8 1996 567.2
CN
1977 625.5 1987 666.7 1997 1832.0 CN ranges from zero to 100, and S is related to the
1978 999.1 1988 812.1 1998 614.9 soil and cover conditions of the watershed linked to CN .
1979 246.3 1989 427.7 1999 365.5
Equations 3 and 4 are solved iteratively via HEC-HMS. In

Figure 2. Different types of synthetic hyetograph.

 Journal of Applied Water Engineering and Research 175

addition, government agencies in Brazil have been recom- 3.2. The synthetic hyetographs
mended the use of the CN method for the rainfall-runoff Five types of synthetic hyetographs were analysed: the
model and meeting the specificities of each region accord- classic alternating blocks, the advanced blocks, the delayed
ing to the tabulated values (Oliveira et al. 2016; Valle blocks, the uniform blocks, and the empirical blocks. All
Junior et al. 2019). For this case study, the Cearense Foun- of these blocks, except those that are empirical, are built
dation of Meteorology and Water Resources (FUNCEME) in such a way that they accommodate the critical rainfall,
has adopted the average value of CN calculated from with a duration equal to tc, for a given Tr in time inter-
several information/characteristics of soil use and land, val tc, which is within a 24-hour duration. The empirical
surface condition, soil class and antecedent moisture con- hyetograph uses the shape of the most critical observed
dition (Paulino 2008). rainfall.

Table 2. Peak flow associated with the return period 3.3. Alternating blocks, advanced blocks and delayed
obtained via the flood frequency analysis.
blocks
Tr(years) 1000 500 200 100 50 The alternating blocks were built using the intensity-
duration frequency (idf ) equation in order to generate
PFFA(m3 .s−1 ) 3336.6 2963.3 2502.1 2176.1 1668.3
synthetic storm patterns (Duka et al. 2017). The method
establishes a time interval for the analysis (T) and forms
n blocks, which are successive and have durations equal to
Table 3. Series of maximum 24-hour observed precipi-
tation at the Campus do Pici climate station. 1) T and 2) T, . . . , n. For each of these blocks, the total
precipitation for a given Tr is obtained by applying the idf
Year P 24h (mm) Year P 24h (mm) Year P 24h (mm) equation. From these n blocks, another set of n blocks was
formed, all with a duration equal to T. The blocks are
1970 52.9 1980 91.9 1990 57.5
1971 145.0 1981 95.3 1991 104.4 block 1 = idf (Tr, t = 1*5)), block 2 = idf (Tr, t = 2*5)
1972 57.4 1982 63.4 1992 76.9 – block 1, block 3 = idf (Tr, t = 3*5) – block 2 through
1973 95.6 1983 81.9 1993 99.6
1974 123.4 1984 100.6 1994 83.0
1975 63.6 1985 102.8 1995 171.2 Table 4. Design rainfall depth associated with the return period.
1976 79.5 1986 97.0 1996 84.3
1977 85.4 1987 87.9 1997 160.9 Tr(years) 1000 500 200 100 50
1978 110.0 1988 98.5 1998 81.9
1979 53.4 1989 75.9 1999 80.5 DRD (mm) 277.71 212.82 193.13 178.20 163.22

Figure 3. Adjusting the annual peak flow to a probability density function.

176 C. A. Chimene and J. N. B. Campos

block n = idf (Tr, t = n*5) – block (n-1). The notation idf to the left of the antecedent block. For a 24-hour rain-
(Tr, t = x) represents the precipitation for return period Tr fall duration at a time step of five minutes, there are 288
and a rainfall duration (t) equal to x. blocks.
For the alternating blocks, these blocks are organized
into a time sequence with the maximum intensity occur-
ring at the centre of the required duration and the other
blocks are set in descending order alternately to the left and 3.4. The empirical hyetograph
right of central block to form the design hyetograph (Duka For the empirical hyetograph, the maximum daily rainfall
et al. 2017; Ilić et al. 2018). For the advanced blocks, was selected for the historical series. For this precipita-
block 1 is set at the beginning of the hyetograph, and all tion type, a dimensionless hyetograph with 288 blocks and
of the others are sequentially set to the right of the previ- a duration equal to five minutes was shaped. The 24-h
ous block. For the delayed blocks, block 1 is set at the end design storm was distributed according to the dimension-
of the hyetograph, and all of the others are set sequentially less hyetograph that was formed.

Table 5. Peak flow from the flood frequency analysis and design storm for the return period and five design storm hyetographs.

Tr (Years) 1000 500 200 100 50

Peak Flow from Frequency Analysis

PFFA(m3 .s−1 ) 3336.6 2963.3 2502.1 2176.1 1868.3
Peak Flow from Design Storm/ Deviate from PFFA
Empirical PFDS(m3 .s−1 ) 2795.9 2545.8 2212.5 2006.1 1771.6
Error (%) − 16.2 − 14.1 − 11.6 − 7.8 − 5.2
Retarded PFDS(m3 .s−1 ) 2746.9 2521.2 2223.4 1998.6 1774.1
Error (%) − 17.7 − 14.9 − 11.1 − 8.2 − 5.0
Alternating PFDS(m3 .s−1 ) 2606.6 2380.0 2079.9 1855.7 1633.3
Error (%) − 21.9 − 19.7 − 16.9 − 14.7 − 12.6
Advanced PFDS(m3 .s−1 ) 2012.1 1827.2 1586.2 1406.5 1229.6
Error (%) − 39.7 − 38.33 − 36,6 - − 35,4 − 34,2
Uniform PFDS(m3 .s−1 ) 2075.6 2031.2 1785.1 1608.7 1421.6
Error (%) − 37.8 − 31.5 − 28.7 − 26.1 − 23.9

Figure 4. Adjusting the series of maximum 24-hour annual rainfall to a probability density function.
 Journal of Applied Water Engineering and Research 177

3.5. The uniform hyetograph where F*(xi) is the distribution function of the speci-
For the uniform hyetograph, rainfall is assumed to have fied distribution, xi represents the ordered data, and N
a duration equal to three times the time interval (tc). The represents the sample size.
hyetograph has three blocks, all with a duration equal to The A2 value is compared with a critical value (CV).
tc. The block in the centre is equal to idf (Tr, t = tc). If A2 > CV, then the distribution is rejected. Among the
The others two blocks are (set to the left and right) equal accepted distributions, the one that yields the smallest A2
to block = ½[(idf (Tr, t = tc) - idf (Tr, t = 2tc)]. In fact, is used as the best pdf to represent the empirical data. The
the uniformity of the hyetograph is restricted to the central pdfs tested were Gumbel, log-normal, gamma and Weibull.
block, with a duration equal to the basin tc. The software Minitab was used for this task.

4. Results and discussion

3.6. Statistical methods for selecting the probability
density function 4.1. Design peak flow from the flood frequency
analysis (PFFA)
The Anderson–Darling statistic (A2 ) is a general test used
for fitting the observed frequencies of a random variable The HMS software is structured into hydrologic ele-
(X) to a theoretical probability density function (Anderson ments, such as sub-basins, junctions, reaches, sources,
and Darling 1954). A2 is computed by Equation 5: sinks, reservoirs and diversions. Inside of the sub-basin,
the loss, transform, and baseflow hydrologic processes
occur. In this paper, the simulation is accomplished for the
1  sub-basin hydrologic element via the loss and transform
A2 = −N − (2i − 1) hydrologic processes. For the loss function, which esti-
N
mates the effective rainfall from the total rainfall, the curve
× {logF ∗ (Xi) + log1 − F ∗ (Xn+1−i )} (5) number method is used, with CN = 78 (weighted average

Figure 5. Design storm with delayed blocks in Sitios Novos River Basin, in Ceará, Brazil.
178 C. A. Chimene and J. N. B. Campos

value, recommended by FUNCEME for this watershed). rainfall-runoff model yielded, for each return period,
For the transform function, which models the hydro- the PFDS.
graph shape from the hyetograph and effective rainfall
amount, the NRCS dimensionless unit hydrograph is used.
A lag time (tL ) of 275 min and tc of 475 min was 4.3. Design Rainfall Depth (DRD)
calculated (using Equation 6, 7). The basin area was
To find the DRD, a series of maximum annual observed
446 km2 . The series of annual peak flows are shown in
24-hour rainfall was formed (Table 3).
Table 1.
The best pdf function, for describing the series of
maximum 24-hour rainfall, was obtained by applying the
L1.155 Anderson–Darling criteria using Minitab (Figure 3). The
tc = 57 (6)
ΔH 0.385 best fitted pdf was the Gumbel, which yielded A2 = 0.454
tL = 0.6tc (7) and a p-value > 0.250. The DRDs for the defined Tr
were found using the frequency analysis, using Gumbel pdf
The best pdf for the peak flow series was a lognormal (Table 4).
distribution that yielded A2 = 0.580 and a p-value > 0.12
(Figure 2). The PFFAs for each Tr are in Table 2.
4.4. Peak Flow from the Design Storm (PFDS)
There were five DRDs distributed over time, according to
4.2. Design Peak Flow from the Design Storm (PFDS) the synthetic hyetographs, which formed five design storm
The design storms for a given Tr were formed by (DS), one for each Tr. Then, by running the DSs in HMS,
the design rainfall depth (DRD), which was dis- 25 PFDS values were obtained. These values were com-
tributed over a 24-hour duration according to the syn- pared with the PFFAs and taken as references to yield
thetic hyetographs. The design storm applied to the the deviations or errors, assuming that the PFFA was the

Figure 6. Design storm with advanced blocks in Sitios Novos River Basin, in Ceará, Brazil.
 Journal of Applied Water Engineering and Research 179

best estimate. These values are in Table 5. The alternat- blocks undersize the peak flow by approximately 40%.
ing blocks are commonly used as conservative methods Therefore, these hyetographs should not be considered
because they conduct to the highest estimate of peak flows when sizing safe hydraulic structures.
creating critical flood situations. They are considered as The results also show that the errors for all
a powerful engineering tool that results in reliable design synthetic hyetographs increase as Tr increases. This
values that does not seek to represent real storm patterns means that the errors yielded by the design storm
(Sordo-Ward et al. 2014; Duka et al. 2017; Ilić et al. 2018; method increase as the return period increases. In other
Elhassnaoui et al. 2019). words, the errors are greater for structures that need
The alternating blocks, may be the most used syn- higher safety regulations than those for higher return
thetic hyetograph for design storms, underestimate 21.9% periods.
of the design peak flow for Tr = 1,000 years. This value The reason for higher values when using the delayed
significantly reduces the safety of hydraulic structures. blocks compared with the advanced and alternating blocks
Lee and Ho (2008) points out that the precision for peak is that for the delayed blocks, the most intense blocks
discharge estimation vary along return period (lower the of rainfall concentrate after reduced intensity blocks have
return period, the estimation error decreases and higher supplied the moisture deficit in the soil (Figure 4). The
return period, the estimation increases significantly) for the same did not happen with the advanced (Figure 5) and
alternating blocks method. alternating blocks (Figure 6). Regarding the empirical
The empirical and delayed blocks yield similar results. hyetograph (Figure 7), it had a peculiar behaviour com-
For Tr = 1,000, the empirical method underestimates pared with the blocks. While the composed blocks were
16.2%, while the delayed blocks underestimate 17.7%. built with a balanced Tr (i.e. if the design rainfall has
These errors are still high (compared with reference value), a Tr = 1,000 years), none of the time intervals in the
but the hyetographs yield lower (safer) results than the design hyetograph had a Tr > 1,000. Because the blocks in
alternating blocks. Finally, the uniform and advanced the empirical hyetograph are unbalanced, certain rainfall

Figure 7. Design storm with alternating blocks in Sitios Novos River Basin, in Ceará, Brazil.
180 C. A. Chimene and J. N. B. Campos

Figure 8. Design storm with empirical hyetograph in Sitios Novos River Basin, in Ceará, Brazil.

episodes, with Tr > 1,000, in the total number of rainfall Table 6. Return period obtained from the design storm
events will occur. Figure 8. method compared with that obtained from the flood fre-
quency analysis.

Rainfall
4.5. Return periods for peak discharges from synthetic distribution
hyetographs shape Return period (years)
In this section, the scientific question to be answered is as Tr (flood frequency 1000 500 200 100 50
follows: if one uses a design storm with a return period of analysis)
Tr years, which is applied to a rainfall-runoff model, what Empirical 362 219 109 69 40
is the return period for the flood yielded? To answer that, Retarded blocks 329 209 111 68 41
the return period of the PFDS for a given design storm Alternating blocks 248 155 62 33 13
Uniform with tc 81 73 42 28 18
Tr was computed from the pdf obtained from the flood Advanced blocks 70 46 26 17 11
frequency analysis. The results are in Table 6.
Table 6 shows that the peak flow return period obtained
from the design storm is significantly less than the return
period of the design storm (see Prodanovic and Simonovic
2004). That is, it is a mistake assuming that a 1,000- 5. Conclusions
year return period for rainfall (associated with alternating The results showed that all synthetic hyetographs stud-
blocks) yields a peak flow with a return period close to ied undersized the hydraulic structures. The validity of the
that. In fact, in that specific case, the Tr from that method conclusion is restricted to the hydrological signature of the
is only 248 years. Therefore, the hydraulic structure is less study area; nevertheless, it can be possible that the same
safe than expected by the engineer. can happen in other rainfall regimens. Similar studies in
 Journal of Applied Water Engineering and Research 181

other regions can show the local safety of the design storm José Nilson Campos, Graduated in Civil Engineer and MSc
method. degree holder in water resources management from the Federal
Based on the concept developed in this paper, which University of Ceará. Obtained his Ph.D. in Water Resources Man-
agement at the Colorado State University in 1987. He worked
concerns the flood frequency analysis method compared as an Engineer at the National Department of Drought Works
with the design storm method, we conclude that all of the (DNOCS) until 1989 where he elaborated many hydrological
synthetic hyetographs used for defining design storms and studies. He is actually retired professor but he is still collabo-
the advanced, delayed, alternating, uniform and empirical rating with the post-graduate process by teaching and research.
blocks undersize the water structure. The method used to He was Head of the Brazilian Association of Water Resources for
the 2003-2005 biennium. Deputy Dean of the Engineering I at
define design storms with alternating blocks undersizes the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior
peak flow in 21.9% of the total, which indicates a significa- (CAPES) in the 2007-2010 triennium. He works mainly in the
tive reduction in structure safety; a hydraulic structure that following subjects: planning and management of water resources,
utilizes a design storm with a Tr = 1,000 years results in a climate studies and hydrological regime of the Semiarid, cli-
peak flow with Tr = 248 years. matic variability. He outlined hydrological studies of more than
50 reservoirs in the Northeast of Brazil. He has been a member of
The empirical hyetograph built from the observed the Dams Safety Panel in Northeast Brazil since 1996. He is the
rainfall in the studied area, which preserves the local author of a reservoir design methodology applied in the Northeast
hydrological signature, yields the safest results among all since 1992. He has 50 papers published in refereed journals and
of the studied hyetographs, but even this synthetic hyeto- more than 100 complete articles published in annals of national
graph undersizes the structure in approximately 16% of and international events. He is a peer-reviewed of Elsevier’s Jour-
nal of Hydrology and Springer’s Water Resources Management
the total. This undersize error increases with the return Journal. He is editor of the Sanitary and Environmental Engi-
period in all synthetic hyetographs; that is, the more safety neering Journal. He is a member of the Brazilian Commission
the water structures need, the more the design method together with the International Hydrological Program because of
undersizes the structure. his knowledge in the semiarid theme.
Finally, it is important to note that the results that are
related to a given hydrological signature can be taken as a
warning for the evaluation of the design storm method in
other places of the world. References
The point that remains a challenge for hydrologists is Anderson TW, Darling DA. 1954. A test of goodness of fit. J
equating the design storm return period to the design flood American Stat Assoc. 49(268):765–769.
return period: the following two methods can be the objec- Askar MK. 2014. Rainfall-runoff model using the SCS-
tive of future studies: 1) establishing, in a regional base, CN method and geographic information systems: a case
a factor to correct the design rainfall depth and volume study of Gomal river watershed. WIT Trans Ecol Envt.
178:159–170.
by equating the two methods; and 2) developing (from Ball JE, Ara J. 2014. Variability in design flood flows from alter-
the observed rainfall) a regional design hyetograph that native rainfall temporal patterns. In Hydrology and Water
equates the two methods. Resources Symposium 2014, HWRS 2014 - Conference
Proceedings.
Barasa BN, Perera EDP. 2017. Analysis of land use change
6. Geolocation information impacts on flash flood occurrences in the Sosiani river basin
Kenya. Int J River Basin Manag. 16(2):1–38.
The study area is located in Northeast coast of Brazil. Beran MA. 1973. Estimation of design flood and the prob-
lem of equating the probability of rainfall and runoff.
In: Symposium on the design of water resources projects
Acknowledgements with inadequate data. Madrid: International Association of
We would like to extend our gratitude to CAPES (Coordenação Hydrological Sciences; p. 459–471. https://goo.gl/ctS3HW.
de Aperfeiçoamento de Pessoal de Nível Superior) for the schol- Botero BA, Francés F. 2010. Estimation of high return period
arship provided and Campus de Pici at the UFC (Universidade flood quantiles using additional non-systematic information
Federal do Ceará) for the meteorological data provided for this with upper bounded statistical models. Hydrol Earth Syst
case study. Sci. 14(12):2617–2628.
Candela A, Brigandì G, Aronica GT. 2014. Estimation of syn-
thetic flood design hydrographs using a distributed rainfall-
runoff model coupled with a copula-based single storm
Notes on contributors rainfall generator. Nat Haz Earth Syst Sci. 14(7):1819–1833.
Catine António Chimene, Former researcher and lecturer at Chin DA. 2017. Designing bioretention areas for stormwater
the higher School of Rural Development (Escola Superior de management. Environmental Processes. 4(1):1–13.
Desenvolvimento Rural), Eduardo Mondlane University (UEM). Dalezios N, Eslamian S. 2016. Regional design storm of Greece
He is currently a Ph.D student enrolled in 2015 (on going) at within the flood risk. Int J Hydrol Sci Technol. 6(1):
the Universidade Federal do Ceará, Departamento de Engen- 82–102.
haria Hidráulica e Ambiental, Fortaleza – Ceará, Brazil. He Duka MA, Lasco JDD, Veyra Jr CD, Aralar AB. 2017. Compara-
accomplished his MSc degree in Municipal Water Infrastructure, tive Assessment of different methods in generating design
Water Supply Engineering at the UNESCO-IHE, Delft in the storm hyetograph for the Philippines. J Env Sci Manag.
Netherlands. 21(1):82–89.
182 C. A. Chimene and J. N. B. Campos

Elhassnaoui I, Moumen Z, Serrari I, Bouziane A, Ouazar D, Silva FOE da, Palacio FFR, Campos JNB. 2013. Equacao de
Hasnaoui MI. 2019. Generation of synthetic design storm chuva para Fortaleza-CE com dados do pluviografo da UFC
hyetograph modeling under HEC-HMS for Ziz watershed. [Storm equation for Fortaleza-CE with data from the UFC
Int J Innov Technol Explor Eng. 8(10):3308–3319. pluviograph]. Revista DAE. 61(192):48–59.
Gericke OJ, Smithers JC. 2014. Review of methods used to Sraj M, Viglione A, Parajka J, Bloschl G. 2016. The influence
estimate catchment response time for the purpose of peak of non-stationarity in extreme hydrological events on flood
discharge estimation. Hydrol Sci J. 59(11):1935–1971. frequency estimation. J Hydrol Hydro. 64(4):426–437.
Ilić A, Plavšić J, Radivojević D. 2018. Rainfall-runoff simulation Sordo-Ward Á, Bianucci P, Garrote L, Granados A. 2014. How
for design flood estimation in small river catchments. Archit safe is hydrologic infrastructure design? analysis of fac-
Civil Eng. 16(1):29–43. tors affecting EXTREME flood estimation. J Hydrol Eng.
Knighton JO, Walter MT. 2016. Critical rainfall statistics for 19(12):04014028.
predicting watershed flood responses: rethinking the design Stuti ACP. 2017. Geospatial technique for runoff estimation based
storm concept. Hydrol Processes. 30(21):3788–3803. on SCS-CN method in upper south Koel river basin of
Lee KT, Ho JY. 2008. Design hyetograph for Typhoon rain storms Jharkhand (India). Int J Hydrol. 1(7):213–220.
in Taiwan. J Hydrol Eng. 13(7):647–651. Sule BF, Alabi SA. 2013. Application of synthetic unit hydro-
Matomela N, Tiaxin L, Morahanye L, Bishoge O, Ikhumhen graph methods to construct storm hydrographs. Int J Water
H. 2019. Rainfall-runoff estimation of Bojing lake water- Res Env Eng. 5(11):639–647.
shed using SCS-CN model coupled with GIS for watershed Tailor D, Shrimali NJ. 2016. Surface runoff estimation by SCS
management. J Appl Adv Res. 4(1):16–24. curve number method using GIS for Rupen-Khan watershed,
Mediero L, Álvarez-Jiménez A, Garrote L. 2010. Design flood Mehsana district Gujarat. J Indian Water Res Soc. 36:4.
hydrographs from the relationship between flood peak and Valle Junior LCG, Rodrigues DBB, Oliveira PTS. 2019. Initial
volume. Hydrol Earth Syst Sci. 14:2495–2505. abstration ratio and curve number estimation using rainfall
Okoli K, Mazzoleni M, Breinl K, Di Baldassare G. 2019. A sys- and runoff data from a tropical watershed. RBRH. 24(5):
tematic comparison of statistical and hydrological methods 1–9.
for design flood estimation. Hydrol Res. 50(6):1665–1678. Wang X, Zhou J, Zhang L. 2015. The probability design evolution
Oliveira PTS, Nearing MA, Hawkins RH, Stone JJ, Rodrigues method for flood frequency analysis: a case study of the Nen
DBB, Panachuki E, Wendland E. 2016. Curve number esti- river in China. Water J. 7:5134–5151.
mation from Brazilian Cerrado rainfall and runoff data. J Soil Watt E, Marsalek J. 2013. Critical review of the evolution of
Water Conserv. 71(5):420–429. the design storm event concept. Canadian J Civil Eng.
Paulino WD. 2008. Açude Sítios Novos: Inventário Ambiental. 40(2):105–113.
Fortaleza, Ce-Brazil. http://https://goo.gl/1L3fyh. Xiao Y, Guo S, Liu P, Yan B, Chen L. 2009. Design flood hydro-
[PDDU] Plano Diretor de Desenvolvimento Urbano (Br). 1998. graph based on multicharacteristic synthesis index method.
Plano Estratégico do Município [Strategic Plan of the J Hydrol Eng. 14(12):1359–1364.
Municipality]. Retrieved from https://url.gratis/TYKwE. Yang SY, Chan MH, Chang CH, Hsu CT. 2018. A case
Ponce VM, Hawkins RH. 1996. Runoff curve number: has it study of flood risk transfer effect caused by land develop-
reached maturity? J Hydrol Eng. 1(1):11–19. ment in flood-prone lowlands. Natural Hazards. 91(3):863–
Prodanovic P, Simonovic SP. 2004. Generation of Synthetic 878.
design storms for the upper Thames River Basin CFCAS Yannopoulos S, Eletheriadou E, Mpouri S, Giannopoulou I. 2015.
project: Assessment of water resource risk and vulnerability Implementing the Requirements of the European flood direc-
to changing climate condition. London, Ontario: University tive: the case of ungauged and poorly gauged watersheds.
of Western Ontario. Report No: 049. Env Process. 2(1):191–207.
Saghafian B, Golian S, Ghasemi A. 2014. Flood frequency anal- Zeng Q, Chen H, Xu CY, Jie MX, Hou YK. 2016. Feasibility
ysis based on simulated peak discharges. Natural Hazards. and uncertainty of using conceptual rainfall-runoff models
71(1):403–417. in design flood estimation. Hydrol Res. 47(4):701–717.
Salami AW, Bilewu SO, Ibitoye AB, Ayanshola AM. 2017. Zhou F, Xu Y, Chen Y, Xu CY, Gao Y, Du J. 2013. Hydrological
Runoff Hydrographs using Snyder and SCS synthetic unit response to urbanization at different spatio-temporal scales
hydrograph methods: A case study of selected rivers in simulated by coupling of CLUE-S and the SWAT model in
South West Nigeria. J Ecol Eng. 18(1):25–34. the Yangtze river Delta region. J Hydrol. 485:113–125.