Journal of Japan Society of Civil Engineers, Ser.

B1 (Hydraulic Engineering), Vol. 73, No. 4, I_151-I_156, 2017.

EFFECTS OF DIFFUSIVE WAVE AND FLOOD

INUNDATION ON TIME OF CONCENTRATION

Khai Lin CHONG1, Takahiro SAYAMA2, Kaoru TAKARA3, Ismail ABUSTAN4

1Member of JSCE, Civil and Earth Resources Engineering, Kyoto University
(Graduate School of Engineering, Kyoto University, Nishikyo, Katsura, 615-8530 Kyoto, Japan.)
E-mail: catherine_cho28@yahoo.com
2Member of JSCE, Associate Professor, Disaster Prevention Research Institute, Kyoto University
(Gokasho, Uji, Kyoto 611-0011, Japan)
E-mail: sayama.takahiro.3u@kyoto-u.ac.jp
3Fellow of JSCE, Professor, Disaster Prevention Research Institute, Kyoto University
(Gokasho, Uji, Kyoto 611-0011, Japan)
E-mail: takara.kaoru.7v@kyoto-u.ac.jp
4Professor, School of Civil Engneering, Universiti Sains Malaysia
(Nibong Tebal, Seberang Perai Selatan, Pulau Pinang 14300, Malaysia)
E-mail: ceismail@usm.my

Time of concentration Tc is defined as the wave travel time from the most hydraulically remote point to
the point of study. Tc is an important element in hydrological studies, especially in drainage system designs
and the estimations of flood arrival time. The common approach in the estimation of Tc is based on
Kinematic Wave (KW) approximation for both overland flow and river routing. This approximation,
however, may not be appropriate on a flood plain with inundations. The main objective of this study is to
propose the estimation of Tc with Diffusive Wave (DW) approximation considering the effect of flood
inundation. The proposed method is demonstrated in the Kelantan River basin, Malaysia, focusing on a
severe flood event in December 2014. This study compares the estimated Tc with other estimations based a
Rainfall-Runoff-Inundation simulation and with a simple correlation method between rainfall amount in
different durations and peak river discharges. In consequence, the proposed method with DW
approximation with flood inundation showed closer estimations of Tc by the other two methods.

Key Words: RRI model, time of concentration, kinematic wave, diffusive wave, inundation extent

1. INTRODUCTION the surface runoff at the catchment outlet3). An alter-

Flood is an environmental hazard that happens
every year in Malaysia, especially on the east part of
the country due to the northeast monsoons. Historical
records have shown extensive damages on human en-
vironment caused by flooding1).
In Malaysia, common practice to design a peak
discharge for urban drainage system uses a rational
method. Based on Urban Stormwater Management
Manual2), the design rainfall intensity used in the ra-
tional method is a function of time of concentration
Tc. In addition, Tc is used to estimate the timing of
flood arrival at different river sections. Hence Tc is
an important parameter in drainage system designs
and provide early warnings during flooding.
Tc is defined as the wave travel time needed for the
most remote point of the catchment to contribute to
native operating definition is the amount of time re-
quired for all portions of a catchment that bring all
surface runoff to the contribute runoff to the outlet3).
The concept of the time of concentration and its
relation to the maximum runoff during the
continuance of a uniform rate of rainfall was ex-
plained in previous studies4), 5). The affecting factors
of Tc include the size, shape and inclination of catch-
ments. There are experimental studies conducted by
the Corps of Engineers6) and Izzard7), which showed
rainfall intensity and rainfall duration also have
significant impacts on Tc.
In the estimations of Tc, there are essentially two
theoretical approaches, (i) lumped approach which
treats Tc of overland flow and channel flow as a single
process and (ii) separated approach in which two of
the processes are treated separately. Yen et al.8), Kib-
ler and Aron9) suggested that the separated approach

I_151
should be adopted as the two flow components are x1 and travels to x = x2 in either overland or channel
the distinct systems, and therefore the separated ap- flow phase, the wave travel time tw is written as equa-
proach provide better estimations of Tc. tion (1),
In Malaysia, Urban Stormwater Management x2 dx
Manual2) suggests the use of separated approach tw (1)
based on Kinematic Wave (KW) approximations.
x1 c
where x is the distance measured along the flow path
However, during a severe flood event in the Kelantan
and c is the wave celerity. The celerity of the wave
River basin, such as the one in December 2014, the
depends upon an expression for the resistance as
estimated flood arrival time based on the KW ap-
well as the wave type. In general, the wave celerity
proach was much shorter than the actual flood arrival
is expressed by equation (2),
time. Therefore, there is a need to estimate more prac-
Q dh
tically reasonable Tc. The working hypothesis of this c (2)
study is that the short Tc is due to the KW assumption, h dAx
which may be invalid in fairly flat areas and also no where Q is the discharge, h is the flow depth, and Ax
inundation consideration. Both of their effects may is the flow cross-section area.
reduce the estimations of Tc. The relationship between discharge Q, flow
Zakeri et al.10) introduced the general form in the depth h and flow cross-section area Ax based on
derivation of Tc based on Saint-Venant equations and Manning resistance law can be written as equation
DW approximation. They found DW approximation (3).
improved the results of Tc compared to KW approxi- 2
Q R 3 S f 2 b h
1 1
mation especially for flat topography10). Regarding (3)
the flow depth profile, DW estimation shows the ef- n
fect of pressure gradient allowing more water to re- Here n is the Manning's roughness coefficient, R is
tain in the channel. Wong11) studied the effect of river hydraulic radius, Sf is friction slope, b is width and
channel shape on wave travel time and detention stor- h is flow depth.
age in seven types of river channel. The findings Based on flow cross-section area Ax the hydrau-
show river channel with high detention storage ca- lic radius R can be determined as the functions of
pacity provide longer Tc. flow depth h. Friction slope, Sf can be assumed as
The main objective of this research is to propose bed slope, So when the acceleration and pressure gra-
an equation for calculating Tc with DW approxima- dient terms are assumed negligible in the
tion for river channel followed by Zakeri et al.10) with conservation of the dynamic wave momentum for
considering the effect of inundation extent. Further- overland flow.
more, the relationship between the proposed methods
Q 5 1 3 1 2
and other simulation and data based approaches are h So 2 b (4)
compared with the estimation of Tc. h 3 n
dh 1
(5)
dAx b
2. METHODOLOGY By substituting equation (4), (5) in the wave ce-
lerity equation (2) and replacing the computed c into
This section presents the methods to calculate Tc the wave travel time equation (1) for a rectangular
based on KW and DW approximations for overland channel, Tcr-KW is expressed as
flow and river channel flow. In addition, this section 3
3 x2 dx
explains the application of a hydrologic model simu- Tcr KW n 5 2
(6)
lating for rainfall-runoff and flood inundation for the 5 x1
Q x S o x b
2 3
5 10 5
Tc estimate with continuous rainfall input.
If the wave travel time is computed for x1 = 0 and
(1) Theoretical derivations x2 = L with the assumption of So(x) are constant along
a) Tc based on Kinematic Wave (KW) x, Tcr-KW can be further approximated as;
By referring to Saghafian and Julien12), we sepa- 3 2

rated that Tc into two phases: Tcs-KW refers to the 3 n5 b5 L dx

time of concentration based on KW approximation
Tcr KW (7)
5 S 103 Qx
0 2
5
on a hillslope to the beginning of stream flow. Tcr- o

KW refers to the time of concentration with KW ap- Now let assume that rainfall of constant intensity
proximation along river channels from an upstream falling over an impervious watershed. If the uniform
point to a downstream point. rainfall intensity, i falls under a uniform rectangular
As a wave originates from a point at a distance x = plane, along the length of the slope SL, the kinematic

I_152
wave based Tc for slope overland flow5) can be esti- 3

mated by inserting Qx i x b and the unit width 5 2

h 3 n
(i.e. b = 1) in equation (7). Q 2S L i
5
(15)
x 5 b S 12
0.6
S L n 0.6 f
Tcs KW 0.4 0.3 (8) Wave travel time in diffusive wave for x1 = 0 and x2
i So = L the integral may similarly be defined based on
where SL is length along the slope, n is the Man- h
ning's roughness coefficients, i is rainfall intensity, equation (6) by replacing S f S o and the
So is the topographic slope.
x
wave travel time in river channel flow Tcr-DW is
Hence the Tc-KW can be summed up from the two
expressed as
phases: the wave travel time for overland flow Tcs-KW 3
3 L dx
n5
and river channel flow Tcr-KW.
Tcr DW
Tc KW Tcs KW Tcr KW (9) 5 0 3
h 10 5
2S x i Qup

2 2

b) Tc based on Diffusive Wave (DW) 5 So b

x
L
In the absence of lateral flow, conservation of the
dynamic wave momentum for overland flow over a (16)
wide plane can be expressed as follows: Note that the solution becomes imaginary
h numbers when Sf is smaller or equal to zero10). In such
S f So (10) a case Zakeri et al.10) proposed to divide the integral
x into two parts with the boundary of xb calculated by
With the function of discharge Q and friction slope the equation (17).
Sf, flow depth h / x can be expressed as
5 3
5 2 n 2
3 3
h h Q h S f xb S o

13
(2S L ) 2 i 2 Qup (2S L ) 1 i 1
(11) 4
3 b
x Q x S f x
(17)
S f
S 2h 10)
Refer to Zakeri et al. , wave travel time Tcr-DW is
o 2
x x x calculated as equation (18).
S o 2h h 3 2
0 and 2 0 , 3
Tcr DW n 5 b 5 3
1 1.01xb dx
By assuming
x x x
can be


2S x i Qup
2
5 0

written as equation (12). S o 10 L

5

h h Q (18)
(12)
x Q x L dx
1.01xb 2

3

(2S L x i Qup ) 5 ( S o )10
Based on the Manning resistance law, flow depth h h
can be written as x
3
5 Since the equations (16) and (18) have no analytical
nQ solution, it has to be numerically integrated to obtain
h 1 the Tcr-DW. Note that if xb is within the range of 0 and
bS 2
f L, the equation (18) has to be used instead of (16). In
3 other words, if the following condition (19) is satis-
5 2 fied with inflow Qup from the upstream, xb will be al-
h 3 n 5
ways smaller than 0 (i.e. outside the range of 0 to L).
Q 5 b S 12 Q (13)
In such a case the equation (16) can be used.
f
5

3
5 n
13 5 5
2 2
Discharge at the position x along river is expressed as Qup So 2S L i 2 (19)
Qx 2S L x i Qup
4 2
3 b
Q The above derivation focused on the diffusive wave
2S L i (14) for river channel Tcr-DW. The equation can be derived
x in the similar way as the Tcs-DW.
where Q(x) is discharge at distance x, Qup is discharge c) Tc based on diffusive wave (DW) with flood
from upstream catchment that flow into the river chan- inundation
nel, b is the width. Fig. 1 shows the assumption that made for
By substituting equations (13) and (14) in equa- calculate Tcr which considers DW and flood
tion (12), inundation.

I_153

Fig. 1 Schematic of assumption of Q(x)
Tcr for DW and flood inundation can be calculated
by equation (16) or (18) with simply replacing the b
with bI to represent the area of inundation extent.
3
3 L dx
Tcr DWI n
5
5
0 3
h
2S x i Qup S o bI
2 10
5
L
x channel was deepen by 30 m from
(20)
where bI is width of the inundation boundary, Qup is
discharge from upstream catchment that flow into the
river channel, L is distance along of the river channel.
Time of concentration of DW and flood inundation
Tc-DWI is calculate in two parts: the wave travel time
on slope Tcs-DW equation (8) and river channel with
inundation Tcr-DWI equation (20).
Tc DWI Tcs DW Tcr DWI (21)
(2) Rainfall-runoff-inundation model simulation
In this study, Tc is estimated also by a hydrologic
model considering runoff on slope, stream flow and
flood inundation. Here Tc is approximated by the time
for discharge to reach a steady state where a dis-
charge reaches 97% of given constant rainfall fol-
lowed by previous studies13). Saghafian and Julien12)
called the time estimated in this way for more com-
plex catchment with varied catchment characteristics
as time to equilibrium Te. The definitions of Tc and Te
are different because the former one expresses the ar-
rival of wave time from hydraulically longest point to
the outlet regardless the conditions of steady rainfall,
while the latter one presumes the steady state condi-
tions. However, in the theoretical estimations of Tc
presented above, we assume the steady state condi-
tions, and therefore the following sections compare
the variables without distinguishing Tc and Te.
As a hydrologic model, this study used the RRI
model, simulating for both rainfall-runoff and inun-
dation simultaneously15). The topographic data is ob-
tained from HydroSHEDS dataset with 15 arc-second
resolution. For the model parameters, we assume no
infiltration to make it consistent with the theoretical
approach described above, leading to two model pa-
rameters; the Mannings roughness coefficients for
rivers and slopes. Based on simulations for the 2014
flood event, they were determined as 0.03 m-1/3s and
0.4 m-1/3s, respectively, used also for the theoretical
Tc calculations.
The steady state simulation was conducted with 1
mm/h for 360 hours to simulate the base flow condi-
tions followed by various rainfall intensities from
5mm/h to 25 mm/h to reach the equilibrium condi-
tions. Moreover, the RRI model simulations were
executed with different parameters including kine-
matic wave, diffusive wave and with and without in-
undation processes (Table 1).
Table 1 Abbreviation for simulated parameters
Test Code Descriptions
2
5 KW Kinematic wave. The depth of river
original depth to avoid flood inunda-
tion.
DW Diffusive wave. The depth of river
channel was deepen by 30 m from
original depth to avoid flood inunda-
tion.
DW + Inu Diffusive wave. The observed river
cross sections were used to estimate
river width and depth. Flood inunda-
tion can occur in the model.
(3) Empirical study on correlations of rainfall and
discharge
Tc can be also estimated empirically by examining
the relationship between peak discharge and rainfall
with different durations. We used 29-year (1986-
2014) records of observed hourly rainfall and dis-
charge. The rainfall from 17 stations were interpo-
lated by Thiessen polygon method to compute catch-
ment average rainfall (Fig. 2). By setting the annual
peak discharges as the ending time for each year, we
calculated catchment average rainfall prior to the dis-
charge peaks by changing the durations from 6 hours,
12 hours, 18 hours, 24 hours, 2 days, 3 days, 4 days,
5 days, 6 days, 7 days, 2 weeks, 3 weeks and 4 weeks.
Then the duration showing the highest coefficient of
determination with the tolerance level of 0.01 were
selected as the empirically estimated Tc in this study.
3. STUDY AREA
The size of the Kelantan River basin is 13,100 km2.
It originates from Ulu Sepat Mountain and discharges
into the South China Sea16). The length of the main
river is about 284 km and its tributaries include the
I_154
Lebir River (2500 km2) and the Galas River (8000 representative slope length SL of each sub-catchment
km2) 17). Annual rainfall ranges between 2,200 mm is estimated by equation (22)18).
and 4,000 mm and almost 85% of the total rainfall A
occurs in north eastern monsoon period from Novem- SL (22)
2 TL
ber to next February.
In December 2014, two waves of extreme mon- where A is the area of sub-catchment, TL is total
soon rainfall hit the catchment. The upstream rain length of the river channel.
gauges, Gunung Gagau received 1,765 mm in a week.
The water level at Kelantan station exceeded the dan-
ger level of 25 m and reached 34 m. The overtopped 4. RESULTS AND DISCUSSIONS
river flow and prolonged rainfall caused inundation
in almost 90 % of the Kelantan state. The damage was Table 2 shows the summary of Tc in hour with dif-
devastating with 25 deaths and Ringgit Malaysia 2.81 ferent methods of calculations and different rainfall
billion loss. intensity.

Table 2 Summary of Tc in hours based on different methods

Lebir Kota Bahru

Rainfall,
Method 5 10 15 20 25 5 10 15 20 25
mm
KW 20 16 14 12 10 22 18 14 12 12
RRI DW 24 18 15 15 15 30 24 20 30 60
Model DW +
76 74 71 72 74 210 236 250 250 260
Inu
Correla-
Empiri-
tion 72 - 168 120 - 336
cal
(0.01)
KW 6 4 4 3 3 12 9 8 7 6
DW 8 6 5 5 4 14 11 10 9 8
Theory
DW +
Inu 16 12 11 10 9 205 158 136 123 114
bI=5,000m

According to the theoretical calculations, the re-

sults indicate that Tc decreases with the increase of
rainfall intensity both at Lebir and Kota Bahru. This
result agrees to the previous study3), 4), 7), which stated
Fig. 2 Sub-catchement and discharge point for upstream and
rainfall intensity is an important influence to the Tc.
downstream
For the Lebir, Tc calculated by using KW and DW
show almost the same values. Table 2 shows Tc of
In terms of the topographic parameters assumed in
KW decreases from 6 hours to 3 hours with rainfall
the theoretical Tc estimations, we used the Hy-
intensities 5 mm to 25 mm, while Tc of DW decreas-
droSHEDS DEM with 15-arc second which was also
ing from 8 hours to 4 hours. On the other hand, for
used for the RRI simulation. As shown in Fig. 2, we
the Kota Bahru, Tc that is calculated by DW has
defined river grid cells with flow accumulation
longer Tc compared to the one by KW by 2 hour. Ta-
greater than or equal to 5. We divided the main rivers
ble 2 shows Tc at Kota Bahru with 5 mm rainfall is 12
into seven segments at the gauging stations. Then the
hours for KW and 14 hours for DW. Followed by 10
average gradients of each river segment were calcu-
mm rainfall, 9 hours Tc for KW and 11 hours Tc for
lated as the average of gradients of all river grid cells
DW. Tc of KW and DW slightly decreases 1 hour
in the sub-catchment. The river widths were esti-
each from 10 mm rainfall to 25 mm rainfall. This can
mated by the measurement data with Acoustic Dop-
be explained as DW effect occurs at mild slope area3),
pler Current Profiler (ADCP) at total 31 cross sec- 10), 12)
. Tc that includes the effect of DW and inunda-
tions and calculated the average widths for each seg-
tion shows even higher increase in the time compared
ment. As for the sub-catchment topographic parame-
to the Tc without inundation. At 5 mm rainfall, Tc of
ters, the slope is calculated from the average slopes
DW is 8 hours on the Lebir and 14 hours at the Kota
of all grid cells within the sub-catchment. Finally, the

I_155
Bahru without inundation. However, Tc of DW with
inundation is 16 hours for upstream and 205 hours for
downstream.
The simulation results showed that KW presents a
smaller Tc which is gradually decreasing from 20
hours to 10 hours with rainfall intensity from 5 mm
to 25 mm at the Lebir; compared with DW which is
24 hours for 5 mm rainfall. Moreover, from the sim-
ulation results, Tc with inundation shows longer Tc
compared to the one without inundation. Tc with in-
undation is about three times longer than Tc without
inundation with 5 mm rainfall, 76 hours for Tc with
inundation and 24 hours for Tc without inundation.
According to the RRI simulated results, although
the Tc values are generally longer than the theoretical
ones possibly due to the different assumptions in the
calculation. One of the reason may be associated to
the assumption of 5,000 m inundation extent for the
theoretical calculation based on Kelantan 2014 flood
event. With the constant inundation extent, the theo-
retical Tc becomes shorter with the higher rainfall in-
tensities. On the other hand, the inundation extent by
the RRI model increases without the limitation of
5,000 m for to the high rainfall intensity cases. This
leads to even longer Tc by the model estimations. Be-
side the inundation extent, different Mannings
roughness coefficient was used for the inundated area
as 0.4 m-1/3s. Hence the reason for the underestima-
tion by the theoretical approach especially for the
high intensity rainfall cases may be related to the as-
sumptions of the inundation extents. Regardless the
discrepancies, our analysis supported that Tc becomes
longer with the considerations of flood inundation ef-
fects.
5. CONCLUSIONS
This study proposed a method to estimate Tc that
considers diffusive wave and inundation effects
along stream flow. From the comparison between the
proposed theoretical approach and RRI model results,
we can conclude that the diffusive wave and inunda-
tion extent slow down the wave travel time. For flat
topography areas, it is therefore important to consider
these factors for estimating more reasonable Tc to be
used to estimate flood arrival times or in designing
drainage systems.
ACKNOWLEDGMENT: The first author acknowl-
edges the doctoral scholarship provided by the Min-
istry of Higher Education Malaysia and this research
was granted by GCOE-ARS Program, Kyoto Univer-
sity. The authors are also thankful to Department of
I_156
Irrigation and Drainage Malaysia for providing rain-
fall and discharge data.
REFERENCES
1) Jonkman, S.N. and Kelman, I.: An analysis of the causes
and circumstances of flood disaster deaths, Disasters,
29(1), pp. 75-97, 2005.
2) Department of Irrigation and Drainage Malaysia: Urbans
storm water management manual for Malaysia, Percetakan
Nasional Malaysia Berhad, 2000.
3) Su, D. and Fang, X.: Estimating Traveling Time of Flat
Terrain by 2- Dimensional Overland Flow Model, In Shal-
low Flows, Editors Gerhard Jirka and Wim Uijttewaal,
Balkema Publishers, 2004.
4) Singh, V. P.: Derivation of time of concentration, Journal
of Hydrology, Vol. 30, pp. 147-165, 1976.
5) Ishihara, T. and Takasao, T.: Fundamental researches on
the unit hydrograph method and its application, In Trans-
actions of the Japan Society of Civil Engineers, Vol. No.
60, Extra Paper (3-3), 1959.
6) Corps of Engineers: Airfield drainage investigation,
Data report, U.S. Army, Los Angeles District for the
Office of the Chief of Engineers, Airfield Branch En-
gineering Division, Military Construction, Washing-
ton, D.C., 1954.
7) Izzard, C.F.: Overland flow tests -Tabulated hydro-
graphs, Bureau of Public Roads, Engineering Socie-
ties Library, New York, N.Y. 1943.
8) Yen, B. C.: Some measures for evaluation and comparison
of simulated models, Proc., 2nd Int. Conf. on Urban Storm
Drainage, B. C. Yen, ed., Water Resources, Littleton,
Colo., pp. 341349, 1982.
9) Kibler, D. F., and Aron, G.: Evaluation of Tc methods for
urban watersheds, Frontiers in Hydraulic Engineering:
Proc., Cambridge Conf., H. T. Shen, ed., ASCE, New
York, pp. 553558, 1983.
10) Zakeri, M. N., Saghafian, B., Golian, S., Moramarco, T.
and Shamsai, A.: Derivation of travel time based on diffu-
sive wave approximation for the time-area hydrograph sim-
ulation, J. Hydrol. Eng., 10.1061/ (ASCE) HE.1943-
5584.0000399, pp. 8591, 2012.
11) Wong, T. S. W.: Effect of channel shape on time of travel
and equilibrium detention storage in channel, J. Hydrol.
Eng., 13(3), pp. 189196, 2008.
12) Saghafian, B. and Julien, P. Y.: Time to equilibrium for
spatially variable watersheds, J. Hydrology, 172(14), pp.
231245, 1995.
13) Machmeier, R.E. and Larson, C.L.: Runoff hydrographs
for mathematical watershed model, J. Hydraulic. Div.
ASCE, 94(6), pp. 1453-1474, 1968
14) Sayama, T. et al.: Rainfall-runoff-inundation analysis of
the 2010 Pakistan flood in the Kabul River basin, Hydro-
logical Sciences Journal, 57 (2), pp. 298-312, 2012.
15) Lehner, B., Verdin, K. and Jarvis, A.: New global hydrog-
raphy derived from space borne elevation data, EOS T. Am.
Geophys. Un., 89, pp. 9394, 2008.
16) Syed Hussain, T. and Ismail, H.: Flood Frequency Analysis
of Kelantan River Basin, Malaysia, World Applied Sci-
ences Journal, 28 (12), pp. 1989-1995, 2013.
17) Ibbitt, R., Takara, K., Mohd. Desa, M. and Pawitan, H. Cat-
alogue of River for Southeast Asia and the Pacific-Volume
IV, Jakarta, Indonesia: UNESCO-IHP, 2002.
18) Takasao, T. and Shiba, M.: Lumping of the basin model
and constants in the kinematic wave method, Disaster Pre-
vention Research Institute Annuals, No. 21B-2, pp. 207-
217, 1978.
(Received September 30, 2016)