Journal of Hydrology: Regional Studies 42 (2022) 101165

Contents lists available at ScienceDirect

Journal of Hydrology: Regional Studies

journal homepage: www.elsevier.com/locate/ejrh

Peak discharge mitigation effects in different rainfall patterns at a

paddy plot with a runoff control plate
Yushi Suzuki *, Kimihito Nakamura, Takehide Hama
Division of Environmental Science & Technology, Graduate school of Agriculture, Kyoto University, Oiwake-cho Kitashirakawa Sakyo-ku, Kyoto
606–8502, Japan

A R T I C L E I N F O A B S T R A C T

Keywords: Study region: Tawaramoto Town, Nara Prefecture, Japan

Rice paddy plot Study focus: In Japan, steep topography and humid climate make the country vulnerable to
Runoff control plate flooding and the importance of the flood mitigation function of rice paddy fields has been
Flood mitigation
emphasized. The ‘paddy field dam project’ is being promoted to enhance the flood mitigation
Rainfall pattern
function by installing a runoff control plate in a drainage box. However, there are a few studies
that have examined the effects of differences in rainfall patterns and rainfall intensity on the
function of paddy field dam by examining possible future rainfall patterns. We generated simu­
lated rainfall patterns from the probability-rainfall intensity formula and from the stochastic
analysis of historical rainfall data. For each rainfall scale, we investigated the relationship be­
tween rainfall characteristics and the function of the runoff control plate.
New hydrological insight: When different rainfall patterns simulated using stochastic methods were
input to the model, the peak ponding depth, and difference in peak ponding depth compared to
paddy plot without runoff control plate might be larger than when the unimodal rainfall with the
almost same 24-h rainfall were input. If the rainfall scale and hourly peak rainfall are determined,
we can estimate the possible range of the peak discharge mitigation rate and the difference in the
peak ponding depth between the plot with and without the runoff control plate.

1. Introduction

In recent years, several studies have shown that the frequency of extremely heavy rainfall events has been increasing under climate
change (Min et al., 2011). Global warming affects rainfall at the spatial and temporal scale given that convection is enhanced under
warmer conditions, increasing the amount of water vapor pulled in from the surrounding environment (Westra et al., 2014). Similar
relationships have also been observed in studies using climate models at mesoscales, including at the regional level (Attema et al.,
2014). According to the latest Intergovernmental Panel on Climate Change (IPCC) report, the increase in the frequency of heavy
rainfall is estimated to accelerate with increasing global warming as a global trend, with 10- and 50-year return period rainfall events
approximately double and triple, respectively, for 4 ◦ C of warming (IPCC, 2021; Li et al., 2021).
Flooding is the most common natural disaster in both developed and developing countries; damage caused by flooding is one of the
most costly and dangerous hazards and therefore, has become a marked social concern (Ahern et al., 2005; Knapp et al., 2008;
Hallegatte et al., 2013; Westra et al., 2014). Some studies have predicted changes in flooding due to climate change on a global scale

* Corresponding author.
E-mail address: [email protected] (Y. Suzuki).

https://doi.org/10.1016/j.ejrh.2022.101165
Received 26 April 2022; Received in revised form 25 June 2022; Accepted 30 June 2022
Available online 6 July 2022
2214-5818/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

(Milly et al., 2002; Hirabayashi et al., 2008). Hirabayashi et al. (2013) predicted global flood risk for the end of this century based on
the outputs of 11 climate models and demonstrated a substantial increase in flood frequency in some regions. For the period from 1980
to 2011, reported flood losses (adjusted for inflation) have increased from an average of $7 billion per year in the 1980 s to $24 billion
per year from 2001 to 2011 (Kundzewicz et al., 2013). Likewise, in monsoon Asia, considered the most vulnerable region worldwide to
flooding, there are indications that the frequency and intensity of flooding may increase in the future, with serious damage to human
lives, infrastructure, and rice yield—a major crop (Mirza, 2011; Auffhammer et al., 2012; Loo et al., 2015). In Japan, the steep
topography and humid climate with frequent typhoons make the country particularly vulnerable to flooding. Japan has been
addressing flood control issues on long timescales (Takahasi and Uitto, 2004). The cost of flood damage caused by extreme rainfall with
various return periods was estimated using numerical simulation and was found to increase nearly linearly with increases in extreme
precipitation because of Japan’s topography (Kazama et al., 2009). In recent years, river flooding caused by heavy rains in western
Japan in 2018 and the Kyushu region in 2020 resulted in severe human and property damage. During the heavy rainfall event in
western Japan in 2018, 1316 mm of precipitation was recorded over 72 h in the area that received the highest rainfall, and 231 people
were killed or went missing (Ushiyama et al., 2019). During the heavy rainfall event in 2020, 497 mm of precipitation was recorded in
24 h in the area that received the heaviest rainfall and 67 people were killed or missing in the Kumamoto Prefecture in the Kyushu
region, Japan (Cabinet Office, Government of Japan, 2021).
Conventional flood protection and mitigation require the construction of hard infrastructure (Wheater, 2006). However, in Japan,
the Ministry of Land, Infrastructure, Transport and Tourism (MLIT) has promoted a policy of “River Basin Disaster Resilience and
Sustainability by All” in which all the stakeholders involved work together to mitigate flood damage in the whole basin, besides
conventional flood control measures previously conducted. In fields related to agriculture, retarding basins, reservoirs, and agricul­
tural land could be used for flood control (MLIT, 2021). In 2021, the government approved a law on these policies. Similar initiatives
are known as Catchment-based Flood Management (CBFM) and Natural Flood Management (NFM) in the UK. NFM has been defined in
many ways in literature (Lane, 2017). The essence of the concept is to apply general flood management and hydrological principles
holistically to develop techniques that replicate or enhance natural processes and demonstrably reduce the flood risk at all scales
within a basin (Cooper et al., 2021). In this framework, offline storage areas are most effective at reducing local flood peaks for a small
flash event in a small rural catchment and peak flow could be reduced by over 30% at downstream receptors (Nicholson et al., 2020).
Catchment-scale tree planning and in-channel woody debris lead to modest mitigation of downstream flows. This attenuation has been
shown to influence not only the duration but also the timing of outfall inundation (Ferguson and Fenner, 2020). However, while the
participation and support of multiple stakeholders in the upper basin and local communities are essential for the success of NFM
projects, some stakeholders represented by farmers and landowners have a more cautious view of these changes (Howgate and Kenyon,
2009; Bark et al., 2021).
Rice paddy fields and irrigation and drainage activities have multiple functions, such as production, eco-environmental, and living-
associated functions (Kim et al., 2006; Huang et al., 2006; Matsuno et al., 2006). In multifunctionality in paddy fields, the flood
mitigation function has been attracting attention as a method of reducing flood damage in the basin by using agricultural land enclosed
by footpaths for flooding. When large precipitation events occur, paddy fields store rainwater temporarily and gradually discharge it
downstream, reducing direct runoff. In Taiwan, substantial increases in runoff and decreases in groundwater recharge occur following
conversion of rice paddy fields for other uses (Wu et al., 2001).
An approach used to improve the flood mitigation function of rice paddy areas is the ‘Paddy field dam project’. The main aim of this
project is to reduce peak discharge from the paddy fields by controlling the water discharged during heavy rainfall events. In the paddy
field dam project, a runoff control plate with an opening smaller than the original outlet in the drainage box is installed. By narrowing
the outlet for water drainage from the paddy plot, storage can be increased and peak runoff can be reduced. However, the paddy
ponding depth is easily increased when paddy field dams are constructed. If the ponding depth increases and water overflows from the
levees, the possibility of levee collapse increases as water overflow is one of the most common causes of levee collapse (Mihara et al.,
1994). The submergence of rice plants during growth can reduce yield (Kotera et al., 2005; Kotera and Nawata, 2007; Minakawa et al.,
2016). Therefore, the relevant indicator is important for assessing the yield of paddy cultivation.
Many studies have evaluated the beneficial effects of paddy field dams. In Japan, the peak discharge mitigation effects of the runoff
control plate have been examined at the plot level (Yoshikawa et al., 2009a; Kobayashi et al., 2021) and district level (Yoshikawa et al.,
2009b, 2010a; Miyazu et al., 2012). The effects of different runoff control plate shapes on mitigation of peak discharge have also been
investigated (Yoshikawa et al., 2010b; Takeda and Asaoka, 2018). An economic evaluation of implementing the paddy field dam
project has also been conducted (Yoshikawa et al., 2011a; Miyazu et al., 2013; Oishi et al., 2019).
Yoshikawa et al. (2009a) used center- and backward-concentrated wave patterns obtained from probability rainfall intensity
formula to study the functions of runoff plates. They reported that the effects of runoff control plates, which can increase peak
discharge mitigation and ponding depth, differ within the same paddy plot based on rainfall pattern and rainfall intensity. However,
the extent of this variation under various rainfall patterns is not clear. Kato and Satoh (2002) investigated effects of peak flow
mitigation of agricultural reservoirs during flood using several actual and stochastic rainfall waveforms, and found that the effects
were more variable when actual rainfall waves were used in the model. However, the effect of peak discharge mitigation and the
degree of increase in peak ponding depth changes with the rainfall pattern in paddy field dam are not clear. A recent study examined
the effects of differences in various rainfall patterns and rainfall intensity on the function of paddy field dam by examining possible
future rainfall scenarios. Minakawa and Miyazu (2022) evaluated the effects of runoff control plate, peak discharge mitigation rate,
and difference in peak ponding depth on a paddy plot using simulated rainfall [generated by the method (Minakawa et al., 2014)
described in Section 3.2]. However, Minakawa and Miyazu (2022) evaluated only the ensemble mean as a representative value for a
given rainfall intensity, and the distribution of rainfall was not considered. Hence, other factors such as the occurrence of some rainfall

2
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

pattern, which can result in a greater increase in ponding depth, were difficult to assess.
The present study aimed to clarify the range of effects and impacts of all possible rainfall patterns on the peak discharge mitigation
rate as well as on the increase in paddy ponding depth. It would be useful to clarify the effects in a single rice paddy plot to provide the
potential flood mitigation effect and the potential rice inundation risk at the regional level. In this study, we assessed the flood
mitigation function of a paddy plot with various simulated rainfall patterns. We developed the discharge model to evaluate the
discharge mitigation function of the runoff control plate. An initial set of rainfall patterns was generated from the probability-rainfall
intensity formula for the study area. We created 10-, 50-, and 100-year return period simulated rainfall events with a single peak. For
each event, we created three rainfall events with forward-concentrated (peak position: 0.0), central-concentrated (peak position: 0.5),
and backward-concentrated (peak position: 1.0) wave patterns and input them in the discharge model. We conducted sensitivity
analysis of the discharge model parameters for each rainfall pattern and examined other factors on these parameters. A second set of
rainfall patterns were generated stochastically from observed rainfall data using the method proposed by Minakawa et al. (2014). The
method used for reproducing heavy rainfall is based on the Monte Carlo method, which can generate simulated heavy rainfall that
reproduces the wave patterns of previously recorded heavy rainfall events. We created 1000 rainfall patterns for 100, 150, and 200 mm
d− 1. For each rainfall scale, we investigated the relationship between each rainfall characteristic and the functions of the runoff control
plate, and finally estimated the possible range of the functions.

2. Materials and methods

2.1. Hydrological observations in the research plot

A paddy plot in Tawaramoto Town, Nara Prefecture, Japan in the Yamato River basin (34.5440◦ N–135.8074◦ E), was selected as
the study site (Fig. 1). In the Yamato River basin, 158 tributary rivers gather in a radial pattern and join within the basin to form the
Yamato River. The Yamato River basin is surrounded by mountains and is a typical low-lying area with depressions in the plains where
rainwater pools. Because of these topographical features, a typhoon that hit the Nara Prefecture in August 1982 caused flooding below
floor level in over 10,000 houses (Nara Prefecture, 2018a). Therefore, some parts of the region are already working on the paddy field
dam project with a focus on regional flood control.
We conducted temporal hydrological observations on the paddy plot (short side: 10.9 m, long side: 104.5 m, 1 drainage box) where
a runoff control plate had been installed. The cross-sectional structure of the drainage box with the runoff control plate is shown in
Fig. 2.
The depth of the paddy pond and the water in the drainage box (the depth upstream and downstream of the runoff control plate) in
the plot were measured. Precipitation values were converted to 5-min intervals by bisecting the 10-min interval precipitation values
from the AMeDAS (Automated Meteorological Data Acquisition System) Tawaramoto station, which is 2 km away from the study site.
The paddy ponding depth and the water depth in the drainage box were measured at 5-min intervals with an accuracy of 1 mm using
capacitance-type water gauges (WT-HR, Intech Instruments Ltd.). The discharge from the drainage box was equal to the discharge from
the opening in the runoff control plate and was calculated from the difference in the water level upstream and downstream of the
runoff control plate and the structure of the drainage box and the submerged opening in the runoff control plate. The discharge from
the submerged opening is expressed hydraulically in Eq. 1:
√̅̅̅̅̅̅̅̅̅̅
Q = CAorif 2gH ′ (1)

Fig. 1. Location of the research plot.

3
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

Fig. 2. Cross-sectional structure of the drainage box with a runoff control plate.

where Q is discharge (m− 3 s− 1), C is the discharge coefficient, Aorif is the cross-sectional area of the opening on the runoff control plate
(m2), g is the acceleration of gravity (= 9.81 m s− 2) H′ is the difference in the water level between the upstream and downstream of the
runoff control plate (m). When the water level downstream from the runoff control plate is lower than the top of the opening, Aorif in
Eq. 1 is the cross-sectional area where the opening is submerged in the water downstream of the runoff control plate. The value of the
discharge coefficient C was set to 0.6 based on the results of a calibration experiment in which several known constant flow rates
entered the drainage box using a pump.

2.2. Discharge model of a paddy plot with a runoff control plate

2.2.1. Discharge from the paddy plot into the drainage box
As a model for the runoff from the paddy plot into the drainage box, we constructed a hydrological model that accounts for the
temporal runoff and infiltration based on the water balance. The structure of the model is shown in Fig. 3. Evapotranspiration was not
included in the model because it would not have played a substantial role during a short rainfall event. There was no inflow, such as
irrigation water or flood water, to the plot other than rainfall. This model could also estimate the discharge from the paddy plot where a
runoff control plate had not been installed (hereafter called ’plot without runoff control plate’). The calculation time interval was set to
1 s and the water balance was calculated for each time step to update the depth of the paddy pond. The calculations are given in Eqs.
2–4:

(2)
3
Q1 = a(H1 − z)2

I = bH1 (3)

dH1
= P − I − Q1 (4)
dt

where Q1 is the discharge (mm s− 1), a is the parameter for runoff over the weir plate (mm− 1/2 − 1
s ), H1 is the paddy ponding depth on

Fig. 3. Discharge model from the paddy plot without a runoff control plate.

4
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

the soil surface (mm), z is the height of the weir plate (height from the soil surface to the top of the weir plate) (mm), I is the infiltration
rate (mm s− 1), b is the parameter for infiltration (s− 1), and P is the amount of precipitation (mm s− 1). Given that the overflow from the
weir plate in the paddy plot could be assumed to be the overflow from a simplistic rectangular weir, the exponent of Eq. 2 was set to 3/
2. Q1, I, and P in Eqs. 2–4 can be expressed in the unit of (m− 3 s− 1) by multiplying by the plot area Aplot (m2) and applying the
appropriate unit conversion. The a and b parameters were identified so the observed discharge from the plot and the calculated
discharge using the model were fitted during the period when the runoff from the paddy plot to the drainage box was equal to the water
discharged from the drainage box into the drainage channel.

2.2.2. Discharge model of a paddy plot with a runoff control plate

The effect of peak discharge mitigation in the paddy plot where the runoff control plate has been installed (hereafter called ’plot
with runoff control plate’) was evaluated by inputting the simulated rainfall into the discharge model as shown in Section 3. We
constructed the hydrological model to account for the effect of the runoff control plate in the drainage box as shown in Fig. 4. As
parameters related to the drainage box, Q2 is the discharge through the opening (m− 3 s− 1), H2 is the height from the center of the
opening to the water level upstream of the runoff control plate (m), and h is the height from the center of the opening to the top of the
weir plate (m).
For calculating the discharge from the drainage box, cases (a) and (b) are classified according to the water depth in the drainage
box.
(a) When H2 < h the upstream water level of the runoff control plate is lower than the top of the weir plate.
Until the upstream water level in the drainage box overflows from the top of the weir plate, the relationship between the runoff Q1
from the plot to the drainage box and the discharge Q2 from the drainage box is expressed by Eq. 5:
dH2
Q1 = S + Q2 (5)
dt

where S is the bottom area upstream of the runoff control plate. The first term (right) is less than 1% of the second term (right) because
S is relatively low (1.6 ×10− 2 m2). For simplicity, the first term on the right was ignored here. Therefore, the discharge Q2 per
calculation time step when H2 < h is given in Eq. 6.
Q2 = Q1 (6)
The water depth H2 in the drainage box at each calculation time step was calculated as follows, considering the effect of submerged
discharge.
In the drainage box with the runoff control plate (Fig. 4), the water running over the weir plate falls to the upstream side of the
runoff control plate, moves to the downstream side through the opening and discharges into the drainage channel through the pipe in
the drainage box. The cross-sectional area of the opening Aorif is less than the cross-sectional area of the pipe Apipe, so the discharge
from the drainage box is defined by the size of the opening (Eq. 7).
Q3 = Q2 (7)
Q3 was calculated with the assumption that the outlet pipe is a weir in line with the method proposed by Yoshikawa et al. (2011b).
We conducted a calibration experiment using a pump for the water to flow into a drainage box at several known constant flow rates

Fig. 4. Discharge model from the paddy plot with a runoff control plate.

5
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

with a runoff coefficient for the drainage pipe of 0.64. This value was close to the runoff coefficient of 0.62 obtained by Yoshikawa et al.
(2011b) for a similar pipe. Eq. 8 was obtained from Eqs. 1 and 7.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Q3 = Q2 = CAhole 2g(H2 − H3 ) (8)
Therefore, if H3 and Q3 (or Q2) are known, H2 can be found and the relationship between discharge Q2 (= Q3) and H2 is obtained.
Therefore, H2 was successively updated according to Q1 at each calculation time step.
(b) When H2 > h, the water level upstream of the runoff control plate is higher than the top of the weir plate.
When the water level upstream of the runoff control plate rises above the top of the weir plate, then the water surface in the paddy
and the drainage box are linked. Here, the change in the water depth in the drainage box at each calculation step is expressed by Eq. 9:
( ) dH2 dH2
Aplot + S ≈ Aplot = P − I − Q2 (9)
dt dt

and H2 was updated successively according to this equation. Similar to the method described above, S was assumed to be negligible in
Eq. 9 because it was relatively low compared to the area of the plot Aplot. The discharge Q2 from the drainage box was obtained from the
relationship between the discharge Q2 and H2.
By switching between the above two cases according to the water depth H2 in the drainage box, we calculated the discharge from
the plot with the runoff control plate. The actual height the two cases switch might be several centimeters higher than h, but for
simplicity, this analysis assumes that the switch occurs at the height of h. We have confirmed that this has little effect on the results.

2.3. Simulation using the discharge model

2.3.1. Parameter identification for the discharge model

The Nash–Sutcliffe efficiency (NSE) (Nash and Sutcliffe, 1970) was used as an index to show concordance between the observed
discharge and the simulated discharge from the discharge model. NSE is expressed by Eq. 10:
∑n
(obsi − simi )2
NSE = 1 − ∑i=1
n 2
(10)
i=1 (obsi − obs)

and is widely used to describe the accuracy of hydrological models. Therefore, it was used to identify the model parameters. In this
equation, obsi and s im i are the observed and simulated values at time step i, respectively, n is the total number of time steps and obs is
the average of the observed values during the calculation period. The a and b parameters of the model were identified so the observed
discharge and ponding depth and the simulated discharge and ponding depth would be fitted when the water level upstream of the
runoff control plate was lower than the top of the weir plate during the selected precipitation event. The selected precipitation events
for parameter identification were: [event 1] 00:00, July 6, 2020–15:00, July 7, 2020 (total precipitation: 52.5 mm, 5-min maximum
precipitation: 2.00 mm) and [event 2] 7:00, July 7, 2021–19:00, July 8, 2021 (total precipitation: 81.5 mm, 5-min maximum pre­
cipitation: 2.25 mm). Given that z is expressed as the weir height, we used the actual weir height measured with a ruler several days
before the two precipitation events as the value for z. The a and b parameters were adjusted, so the observed discharge and the ponding
depth and the simulated discharge and ponding depth were matched during the two precipitation events using the SCE-UA method
(Duan et al., 1992).

2.3.2. Simulation conditions

The same simulated rainfall was input in the parameterized discharge model in the plot with and without the runoff control plate,
and the output results for the plot with and without the runoff control plate were compared and the effect of the runoff control plate
was evaluated. The simulated input rainfall that was input in the model is described in Section 3. The conditions set for the simulation
using simulated rainfall have been described subsequently.
(A) The plot area for 1 drainage box (weir width: 31.8 cm) was set at 1130 m2.
This simulation was based on the plot area of the research field and the number of drainage boxes (one drainage box per paddy plot
with an area of approximately 1130 m2).
(B) The ponding depth has not exceeded the top of the runoff control plate and the height of the levee.
The discharge analysis in the model was conducted for one plot. In this simulation, it was assumed that the ponding depth did not
exceed the height of the levee and there was no overflow beyond the levees. The top of the runoff control plate was also assumed to be
high enough, and there was no overflow from the top of the runoff control plate.
(C) The weir height is the value at [event 1] described in Section 2.3.1, and the initial ponding depth is fixed at the weir height.
The height of the weir plate and the ponding depth at the beginning of rainfall vary according to the precipitation events. In this
simulation, the height of the weir plate was set to the value for [event 1] (=130 mm). The initial ponding depth was also fixed at the
height of the weir plate (=130 mm); thus, there was no space for precipitation to pool in the paddy plot at the beginning of the
simulation.

6
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

3. Rainfall patterns input in the discharge model

3.1. Simulated rainfall patterns generated from the probability-rainfall intensity formula

The rainfall patterns input in the discharge model were decided based on the probability-rainfall intensity formula of Nara Pre­
fecture, Japan, and the method of generating probability rainfall used in river flood control projects and the design standards to
regulate reservoirs for disaster prevention (Nara Prefecture, 2018b; Sato, 1987). We created 10-, 50-, and 100-year return period
simulated rainfall events with a single peak, and for each event, we created three rainfall events with forward-concentrated (peak
position: 0.0), central-concentrated (peak position: 0.5), and backward-concentrated (peak position: 1.0) wave patterns. Given that for
the technical standards for the paddy field dam project in Nara Prefecture, Japan (Nara Prefecture, 2018c) a standard duration of 24 h
was used for the planned rainfall, the duration of the simulated rainfall generated in this simulation was also set to 24 h. The
probability-rainfall intensity formula for Nara Prefecture, Japan, and the characteristics of the simulated rainfall are shown in Table 1.

3.2. Simulated rainfall patterns generated from the probabilistic statistical method

Minakawa et al. (2014) developed a method for reproducing heavy rainfall events, which reproduces the wave patterns of pre­
viously recorded heavy rainfall events, based on the Monte Carlo method. Using this method as a reference, simulated rainfall patterns
were created. The methods described below are not all the same as those of Minakawa et al. (2014). The flowchart to generate the
simulated rainfall patterns is shown in Fig. 5.

3.2.1. Preparation for a simulated heavy rainfall event

The precipitation data observed at the AMeDAS Tawaramoto station every 10 min for 27 years from January 1994 to December
2020 were used. The duration of heavy rainfall was set at 24 h, which is the same as the duration of the planned rainfall described in
Section 3.1. A rainfall threshold was set and events exceeding the threshold were extracted as heavy rainfall. We extracted 80 heavy
rainfall events (2.96 events per year on average) over 27 years by setting the threshold for heavy rainfall at 50 mm d− 1. The rainfall
scale generated was fixed at 100, 150, and 200 mm d− 1. For each rainfall scale, 1000 heavy rainfall events were generated. The
Gumbel distribution was applied to the annual maximum daily rainfall data observed at the AMeDAS Tawaramoto station over the last
40 years, and the return periods for each rainfall scale were estimated to be 4, 18, and 86 years, respectively. Because the probability-
rainfall intensity formula used in Section 3.1 is for Nara Prefecture (not for Tawaramoto Town), the estimated return period for each
rainfall scale differs slightly from the return period for the unimodal rainfall produced in Section 3.1.

3.2.2. Determination of the distribution ratio to the hourly rainfall

The total rainfall was distributed to the 10-min rainfall for the 1000 heavy rains generated for each rainfall event. The 10-min
rainfall values were calculated by multiplying the total rainfall by the distribution ratio of each time step. The length of one array
was 144 (24 h duration, 10-min intervals). The distribution ratio dt at time step t was the value of [0,1] obtained according to the beta
distribution. The parameters of the beta distribution were determined by trial and the shape of the distribution was exponential so that
the rainfall intensity distribution obtained was close to the observed distribution. This was because the intensity distribution of short-
duration rainfall during heavy rainfall can be approximated using an exponential distribution. In this study, the rainfall scale was fixed
at 100, 150, and 200 mm d− 1. Therefore, the intensity distribution was close to that of the observed rainfall which was extended to
each rainfall scale. For extending the observed rainfall, the actual rainfall is sometimes extended to equal the planned rainfall for river
flood control planning, and the extension ratio should be limited to approximately 2.0. We extracted 17, 12, and 9 observed rainfall
events by limiting the extension ratio of 0.80–1.25, 0.56–1.80, and 0.48–2.10 at the rainfall scale of 100, 150, and 200 mm d− 1,
respectively. Fig. 6 shows the distribution of the 10-min rainfall occurrence for each rainfall scale for the extended observed rainfall
and the simulated rainfall events.
The distributed rainfall arrays were reordered to match the autocorrelation coefficient of the simulated rainfall patterns to that of
the observed rainfall patterns. For this purpose, a quasi-rainfall array x was generated. Assuming the elements of the array were xi
(i = 1, 2,., 144), the values were determined using exponential distribution (Eq. 12):

Table 1
Characteristics of the probability-rainfall intensity formula of the Nara pref. and the model rainfall patterns.
Return Max Max Total rainfall Formula
period (mm 10 min− 1) (mm h− 1) (mm 24 h− 1)
1 ≤ t ≤ 90 91 ≤ t

10-year 19.4 51.8 166.9 4669 241

I = I =
t + 30.18 t0.5 − 3.29
50-year 25.8 69.5 215.0 6307 308
I = I = 0.5
t + 30.75 t − 3.56
100-year 28.5 77.0 235.8 6990 337
I = I = 0.5
t + 30.83 t − 3.64

I: rainfall intensity (mm h− 1)

t: time (min)

7
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

Fig. 5. Flowchart of the simulated rainfall patterns generated.

F(xi ) = ui = 1 − exp( − λxi ) (12)

where λ is the scale parameter of the exponential distribution. This value represents the inverse of the expected value of the distribution
by definition. First, x1 was a randomly selected value from the 144 rainfall values to be sorted. u1 was in the range [0,1]. We needed to
consider the correlation of x2 with x1, where we could determine the value of u2 considering the correlation with u1, and then x2 could
be determined by Eq. 12. The value of u2 was assumed to follow a triangular distribution with u1 as the most frequent value and the
minimum value Ltri and maximum value Rtri as the values obtained from Eq. 13.
′
Ltri = max(0, u1 − w )

(13)
′
Rtri = min(0, u1 + w )
The parameter w’ was determined by trial and error so that the autocorrelation coefficient of the final quasi-rainfall array was a
good fit for the autocorrelation coefficient of the observed rainfall events. After generating the quasi-rainfall values, the distributed
rainfall values were reordered so the rank order of the distributed rainfall array would be the same as the rank order of the quasi-
rainfall array.
The autocorrelation coefficients for the observed rainfall and the simulated rainfall were calculated and compared. Given that the
autocorrelation coefficients were different for each event, we calculated them individually for each event, and the average of the whole
was the representative value. The correlogram for the observed and simulated heavy rainfall series for each rainfall scale is shown in
Fig. 7. The parameter w′ was determined with the importance that the autocorrelation coefficient of the simulated rainfall should be
fitted to the autocorrelation coefficient of the observed rainfall when the lag time was low. These results demonstrated that generated
rainfall represents the autocorrelation coefficient of the observed rainfall at each rainfall scale.

4. Results and Discussion

4.1. Parameter identification results for the discharge model

Parameters a and b in the discharge model in Fig. 3 were identified so the observed and calculated discharge and ponding depth
would match each other during the [event 1] and [event 2] precipitation events shown in Section 2.3.1. For the analysis, the units of

8
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

Fig. 6. Distribution of 10-min rainfall occurrence for each rainfall scale for extended observed rainfall and simulated rainfall events.

Fig. 7. Correlogram of the observed and simulated heavy rainfall series for each rainfall scale.

9
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

parameters a and b are expressed in (mm− 1/2 5 min− 1) and (5 min− 1), respectively. Therefore, for identifying the parameters, a
= 4.80 × 10− 3 (mm− 1/2 5 min− 1) and b = 4.87 × 10− 5 (5 min− 1). Applying an appropriate unit conversion to parameter a and
converting it to the unit system of the runoff coefficient Crect in the rectangular weir formula (Eq. 14).

(14)
3
Q = Crect BH 2
We obtained Crect= 1.80. This value does not differ substantially from the value of Crect= 1.90 from the study by Yoshikawa et al.
(2009a). In this equation, B is the width of the weir plate (m) and H is the overflow depth (m). The parameter z was set at z = 130 (mm)
for [event 1] and z = 147 (mm) for [event 2] to match the actual height of the weir plate during the precipitation events. Fig. 8 shows
the time changes for the observed discharge and the calculated discharge for both precipitation events using the discharge model. The
NSE calculated from the temporal observed and calculated discharge for both events was NSE = 0.83 for [event 1] and NSE = 0.93 for
[event 2]. NSE is considered a good fit for hydrological models with a value > 0.70 (Moriashi et al., 2015); thus, the model could
effectively reproduce the actual discharge process from the paddy plot. [event 1] shows a brief increase in discharge at 13:20 on July 6,
but no substantial rainfall was observed at the AMeDAS Tawaramoto station. However, the observed ponding depth substantially
increased, indicating that highly localized rainfall occurred around the study plot. The model could not reproduce this discharge. Fig. 8
also shows the temporal changes in the observed and depth during both precipitation events. The error was evaluated by the RMSE,
which is expressed with Eq. 15.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
∑n 2
i=1 (obsi − simi )
RMSE = (15)
n
The RMSE for both precipitation events was 1.7 mm for [event 1] and 3.6 mm for [event 2], with a slightly higher value for [event
2] with consistent trends. Therefore, the discharge model was used after the parameter identification to simulate the different input
rainfall patterns.

4.2. Inputting the rainfall patterns generated by the probability-rainfall intensity formula

For each simulation, the calculation time was set to 48 h, and a unimodal rainfall pattern was set to the first 24 h. Fig. 9 shows the
temporal change in the discharge from the plot with and without the control plate under the unimodal rainfall patterns generated
according to the method described in Section 3.1. This figure represents results for forward, central, and backward-concentrated

Fig. 8. Temporal changes in the observed and calculated discharge Q and The observed and calculated ponding depth H1 using the discharge model
for both precipitation events.

10
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

Fig. 9. Temporal change in the discharge from the plot with and without A runoff control plate when the unimodal rainfall patterns are input in the
discharge model. (50-year return period; forward-, center-, backward- rainfall patterns).

rainfall with a 50-year return period input in the model. The results of the other rainfall patterns input in the discharge model are
summarized in Table 2. The peak discharge mitigation rate is defined by Eq. 16:
Qp,dam
ρ = 1− (16)
Qp,norm

where r is the peak discharge mitigation rate, Qp,dam is the peak discharge from the plot with the runoff control plate, and Qp,norm is the
peak discharge from the plot without the runoff control plate. The difference in the peak ponding depth between the plot with and
without the runoff control plate is defined by Eq. 17:
δ = Hp,dam − Hp,norm (17)

where d is the difference in the peak ponding depth between the plot with and without the runoff control plate, Hp,dam is the peak
ponding depth of the plot with the discharge control pate, Hp,norm is the peak ponding depth of the plot without the runoff control plate.
This is an indicator of the extent to which the peak ponding depth increases when the runoff control plate is installed compared to
when no plate is installed.
These results showed that for each return period, the peak discharge mitigation rates for center- and backward-concentrated
precipitation events were similar and higher than the peak discharge mitigation rate for forward-concentrated precipitation events.
This may be due to center- and backward-concentrated rainfall patterns contributing more to the peak discharge from the paddy plot
without the runoff control plate than for the forward-concentrated rainfall pattern. The difference in the peak ponding depth was
slightly larger for the input of centralized rainfall than for the input of forward and backward-concentrated rainfall, with a difference of
17.2 mm for the 100-year return period centralized rainfall. This could be due to the centralized rainfall pattern having a larger
amount of precipitation near the peak precipitation, which increases the ponding depth more than the forward and backward-
concentrated rainfall patterns.

Table 2
Changes in the discharge from the paddy plot with and without the runoff control plate when unimodal rainfall wave patterns are input in the
discharge model.
Return Wave Peak ponding depth (mm) Difference in peak ponding Peak discharge Peak discharge mitigation rate
period pattern depth (mm 5 min− 1) (%)
(mm)
Without With Without With
control control control control
plate plate plate plate

10-year Forward 171.6 174.6 3.0 1.259 0.872 30.8

Central 189.9 197.1 7.2 2.180 0.896 58.9
Backward 195.7 200.1 4.4 2.504 0.899 64.1
50-year Forward 184.0 193.7 9.7 1.866 0.893 52.2
Central 206.0 220.0 14.0 3.111 0.921 70.4
Backward 213.1 222.5 9.4 3.564 0.923 74.1
100- Forward 189.1 202.1 13.0 2.137 0.902 57.8
year Central 212.5 229.7 17.2 3.523 0.931 73.6
Backward 220.4 232.4 12.0 4.042 0.934 76.9

11
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

4.3. Sensitivity analysis for model parameters and initial ponding depth

The function of the runoff control plate is influenced by the characteristics of the plots and the ponding depth at the start of the
rainfall. We conducted sensitivity analysis to determine how the peak discharge mitigation rate changes with the input rainfall return
period and the peak location when the model parameters a, b, the height of the weir plate z, and the initial ponding depth Hinit were
changed. The parameters were based on those obtained in Section 4.1 (the value of parameter z was used for [event 1]), and the
targeted parameters were changed. The initial ponding depth was fixed at the value of parameter z, and the discharge delay due to the
room for precipitation to pool was not considered an initial condition. Fig. 10 shows the change in the peak discharge mitigation rate
for input of nine types of rainfall with different return periods and peak locations when the parameters a, b, z, and Hinit were changed.
The meaning of parameter a is a factor that defines the amount of discharge over the weir plate of the drainage box in the actual
plot, because the larger a is, the larger the discharge when the overflow depth is equal. The width of the weir plate installed in the
drainage box, the runoff coefficient of the weir plate, and the slope toward the paddy plot outlet are considered the main factors.
Parameter a was changed in the range of 2.5 × 10− 3–6.5 × 10− 3 (mm− 1/2 5 min− 1). If we convert to the unit system of the runoff
coefficient Crect in Eq. 14, the range of Crect is 0.94–2.44. For each rainfall input to the model, the larger the value of parameter a, the
larger the peak discharge mitigation rate. The difference in the peak discharge mitigation rate became smaller as the value of
parameter a increased when both forward and backward-concentrated rainfall patterns were entered in the discharge model for the
same return period rainfall.
The meaning of parameter b is the infiltration rate of the plot. Parameter b varied in the range of 0.0–1.0 × 10− 3 (5 min− 1). From
Eq. 3, this corresponds to an infiltration rate of 0.0–43.2 mm d− 1 in a plot where the ponding depth is always maintained at 150 mm.
For each rainfall input in the model, the peak discharge mitigation rate decreased as the value of parameter b increased. Change in the
peak discharge mitigation rate with the change in parameter b was small and almost constant for each rainfall.
The meaning of parameter z is the length from the surface of the plot to the top of the weir plate. Parameter z was changed from 0 to
150 mm. This is zero for the non-irrigation period when the weir plate was not installed and is the height of the weir plate installed by
farmers for the irrigation period. The initial ponding depth Hinit was fixed to the value of parameter z. For each rainfall input to the
model, the peak discharge mitigation rate decreased as the value of parameter z increased. Change in the peak discharge mitigation
rate with the change in parameter z was small, as resulted from the sensitivity analysis for parameter b.
The initial ponding depth Hinit at the beginning of the rainfall changed from 0 to 130 mm. The value of the 130 mm is the value of

Fig. 10. Change in the peak discharge mitigation rate for input for nine types of rainfall with different return periods and peak locations when the
parameters are changed.

12
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

parameter z in [event 1]. When the 100-year return period center- and backward-concentrated rainfall patterns were input in the
model, the peak discharge mitigation rate was almost constant regardless of Hinit. However, when other rainfall patterns were entered,
the discharge mitigation rate in the plot with the runoff control plate increased as Hinit increased.
These results indicate that the peak discharge mitigation rate changes depending on the characteristics of the paddy plot with the
runoff control plate, the return period, and the peak location of the rainfall events.

4.4. Inputting simulated rainfall patterns generated from the probabilistic statistical method

A total of 3000 simulated rainfall events were generated for three rainfall scales of 100, 150, and 200 mm d− 1, 1000 events for each
scale. The discharge mitigation rate, the peak ponding depth in the plot with and without the runoff control plate, and the difference
between them were investigated. The calculation time was 72 h, and simulated rainfall was entered in the first 24 h.
Fig. 11 shows the distribution of the peak discharge mitigation rate for the plot with the runoff control plate and the difference in
the peak ponding depth between the plot with and without the runoff control plate for each rainfall scale. The distribution shape was
shown using a violin plot, which is a combination of line segments showing the maximum, median, and minimum values, dots showing
mean values, and a histogram. The distribution of the peak discharge mitigation ratio showed that the maximum, minimum, median,
and mean values of the peak discharge mitigation ratio increased with increasing rainfall, and the range of the peak discharge
mitigation ratio decreased. The relationship between the rainfall scale and the difference in the peak ponding depth in the plot with
and without the runoff control plate showed that the maximum median, and mean values for the difference in the peak ponding depth
increased as the rainfall scale increased, and the range of the difference in the peak ponding depth increased. Minakawa and Miyazu
(2022) only assessed the mean values; however, the results of the present study suggest that the mean value for each rainfall pattern
did not necessarily coincide with the mode and median of the distribution.
The frequency distribution of the peak ponding depth for each rainfall scale in the plot with and without the runoff control plate is
shown in Fig. 12. These histograms showed that the distribution shifted toward the higher peak ponding depth with the increase of the
rainfall scale in both the plots with and without the runoff control plate. As the rainfall scale increased, the distribution shifted more
toward the larger peak ponding depth in the plot with and without the runoff control plate, and the difference in the peak ponding
depth between them became greater. When carrying out the paddy field dam project, it is important to maintain and manage the
storage function of rice paddy fields to make the runoff control plate effective (Yoshikawa and Tsubaki, 2016). Results obtained from
the simulations indicate that the peak ponding depth will be less than 260 mm even during large rainfall events. The possibility of
overflow from the levees is small if the levee height is greater than 260 mm in this study area.
Fig. 13 shows a scatter plot of the hourly peak rainfall on the horizontal axis and the peak discharge mitigation rate on the vertical
axis for 1000 simulated rainfall events generated at each of the three rainfall scales (3000 events). The results of using unimodal
rainfall patterns with three 50-year return periods (total rainfall: 215 mm d− 1, hourly peak precipitation: 69.5 mm h− 1) in the model
are also represented in Fig. 13. The range of the peak discharge mitigation rate shifts higher as the rainfall scale increases. The range of
hourly peak precipitation distribution shifts higher as the rainfall scale increases. Despite slight differences in daily precipitation
compared to 200 mm d− 1, the results for unimodal rainfall, especially backward concentrated rainfall input to the model, were located
at the top of the distribution of dots in Fig. 13. Fig. 14 shows a scatter plot of the hourly peak rainfall on the horizontal axis and the
difference in the peak ponding depth in the plot with and without the runoff control plate on the vertical axis. The results of using three
(forward-, center-, and backward-concentrated) 50-year return period unimodal rainfall patterns (total rainfall: 215 mm d− 1, hourly
peak precipitation: 69.5 mm h− 1) in the model are also shown in Fig. 14. The range of the difference in the peak ponding depth be­
tween the plot with and without the runoff control plate becomes wider as the rainfall scale increases. When the three unimodal
rainfall patterns were used in the model, the differences observed in peak ponding depth between the plots with and without control
plate were all situated at the bottom of the distribution of dots in Fig. 14.

Fig. 11. Distribution of the peak discharge mitigation rate and the difference in the peak ponding depth between the plot with and without a runoff
control plate at each rainfall scale (the line segments indicate the maximum, median, and minimum values from the top, whereas the dots indicate
the mean values).

13
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

Fig. 12. Frequency distributions of the peak ponding depth for each rainfall scale in the plot with and without a runoff control plate.

Fig. 13. Scatter plot of the relationship between the hourly peak rainfall and the peak discharge mitigation rate for 1000 simulated rainfall events
generated on each rainfall scale P and the results of inputting three (forward-, center-, and backward-concentrated) 50-year return period rainfall
into the model.

As shown in Table 2 and Figs. 11, 12, and 14, when the simulated rainfall patterns created by the method of Minakawa et al. (2014)
were input to the model, the peak ponding depth and the difference in peak ponding depth might be larger than when the unimodal
rainfall waveform with the same 24-h rainfall were input in paddy plot without control plate. Therefore, it would be inadequate to

14
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

Fig. 14. Scatter plot of the relationship between the hourly peak precipitation and the difference in the peak ponding depth for 1000 simulated
rainfall events generated on each rainfall scale P and the results of inputting three (forward-, center-, and backward-concentrated) 50-year return
period rainfall into the model.

consider only unimodal rainfall pattern when considering the risk of levee collapse and yield reduction associated with increased
ponding depths due to use of control plates.

4.5. Relationship between the peak rainfall and the effects of the runoff control plate

Figs. 13 and 14 indicate that the dots are concentrated and distributed within a limited range, respectively. For example, Fig. 13
shows that the dots are concentrated inside the area bounded by the curve that rises from approximately 40 mm hourly peak rainfall
and approximately follows the lower side of the distribution, the curve that rises from approximately 20 mm hourly peak rainfall and
approximately follows the upper side of the distribution, and the horizontal axis. From these results, it was considered possible to
estimate the degree of the peak discharge mitigation rate and the difference in the peak ponding depth in the plot with and without the
runoff control plate based on the rainfall scale and the hourly peak rainfall. The peak discharge mitigation rate curve is defined as the
lower limit curve and the difference in the peak ponding depth is defined as the upper limit curve.
The peak discharge mitigation rate is expressed by Eq. 16. Because Qp,dam in Eq. 16 is almost constant, the value of Qp,norm con­
tributes the most to the peak discharge mitigation rate, and the smaller the value of Qp,norm, the smaller the peak discharge mitigation
rate. The larger the value of Qp,norm, the larger the peak discharge mitigation rate. For a set of rainfall patterns with a determined
hourly peak rainfall at each rainfall scale, the peak discharge mitigation rate when the pattern with the lowest rainfall contributes to
the peak discharge is input in the discharge model is drawn as the lower limit curve while changing the hourly peak rainfall. When the
rainfall patterns with the highest rainfall contributing to the peak discharge are input in the discharge model, the peak discharge
mitigation rate is drawn as the upper limit curve while changing the hourly peak precipitation. Based on this hypothesis, the rainfall
pattern that forms the curve close to the lower limit concentrates the peak rainfall in the first hour, while the rainfall pattern that forms
the curve close to the upper limit continues the peak rainfall for (rainfall scale) / (hourly peak precipitation) hours. If the hourly peak
rainfall is 50 mm for a rainfall scale of 200 mm, the rainfall pattern that forms a curve close to the lower limit is considered the result
from concentrating the peak rainfall of 50 mm in the first hour and distributing 150 mm over the remaining 23 h. However, we
assumed that 150 mm of precipitation could be freely distributed so as not to contribute to the peak discharge in the remaining 23 h,
and it was expected that the rainfall pattern that forms the curve close to the lower limit could be approximated by the rainfall pattern
with the peak rainfall of 50 mm concentrated only in the first hour. The rainfall pattern that forms the curve close to the upper limit
was expected to be approximated to the rainfall pattern with the peak rainfall of 50 mm h− 1 which continued for 4 h.
For each rainfall scale, the rainfall pattern that forms a curve close to the lower limit and a curve close to the upper limit is
generated by changing the hourly peak rainfall from 20 mm to the value of each rainfall scale by 1 mm. The curves drawn from the
lower and upper limits are shown in Fig. 15. According to this figure, the distribution plotted for each rainfall scale is concentrated in
the area bounded by the lower and upper curves, indicating that the hypothesis is valid. Fig. 16 also shows a figure drawn by
superimposing the lower and upper limit curves in Fig. 14. In this figure, the distribution plotted on each rainfall scale is concentrated
in the area surrounded by the lower limit curve and the upper limit curve. However, it can be seen that no dots are present near the

15
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

Fig. 15. Estimated lower and upper limit curves in Fig. 13 at each rainfall scale P.

upper limit curve, which implies that the probability of a rainfall event with such results is extremely low. Further probabilistic an­
alyses can be undertaken to narrow this range in future.
In summary, if the rainfall scale and the hourly peak rainfall are determined, the possible range of the peak discharge mitigation
rate and the difference in the peak ponding depth between the plot with and without the runoff control plate can be estimated.

5. Conclusions

We developed a discharge model to evaluate the discharge mitigation function of the runoff control plate and compared the results
when inputting various rainfall patterns. The main conclusions of the present study are summarized below:

I. The peak discharge mitigation rate and difference in peak ponding depth changed depending on the characteristics of the paddy
plot with the runoff control plate, the return period, and the peak location of the rainfall. The difference in the peak discharge
mitigation rate between the forward and backward-concentrated rainfall input in the discharge model for the same return period
rainfall became smaller as the value of parameter a (the weir width installed in the drainage box, the runoff coefficient of the weir
plate, and the slope of the rice paddy plot to the outlet) increased.
II. When the simulated rainfall patterns created by stochastic method were input to the model, the peak ponding depth, and the
difference in peak ponding depth compared to paddy plot without runoff control plate might be larger than when the unimodal
rainfall waveform with the almost same 24-h rainfall were input. The possible range of the peak discharge mitigation rate in the
plot with the runoff control plate and the difference in the peak ponding depth between the plot with and without the runoff control
plate can be estimated if the rainfall scale and the hourly peak rainfall are determined.

In the future, the model developed in this study, which can calculate the water balance of the plot with the runoff control plate, will
be incorporated into a model that can perform district-level inundation analysis to quantitatively evaluate the effects of runoff control
plates when various rainfall patterns occur in districts with different land uses. Various rainfall patterns were created based on the
rainfall data observed in the past in this study. To verify the effects of the runoff control plate when entering expected future rainfall
under climate change, we will use the output results from the global climate model to examine the role of agricultural land in basin-
level flood control.

16
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

Fig. 16. Estimated lower and upper limit curves on Fig. 14 at each rainfall scale P.

Author Agreement

All authors have seen and approved the final version of the manuscript being resubmitted. They warrant that the article is the
authors’ original work, hasn’t received prior publication and isn’t under consideration for publication elsewhere.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that may have influenced the
work reported in this paper.

Acknowledgements

The authors would like to thank Nara Prefecture, the Representative and Manager of the district, and the landowner. This work was
supported by funding from the Land Improvement Construction General Association of Japan. This work was also supported by JET
SPRING, Grant Number JPMJSP2110.

Appendix A. Supporting information

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.ejrh.2022.101165.

References

Ahern, M., Kovats, R.S., Wilkinson, P., Few, R., Matthies, F., 2005. Global health impacts of floods: epidemiologic evidence. Epidemiol. Rev. 27, 36–46. https://doi.
org/10.1093/epirev/mxi004.
Attema, J.J., Loriaux, J.M., Lenderink, G., 2014. Extreme precipitation response to climate perturbations in an atmospheric mesoscale model. Environ. Res. Lett. 9,
014003 https://doi.org/10.1088/1748-9326/9/1/014003.

17
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

Auffhammer, M., Ramanathan, V., Vincent, J.R., 2012. Climate change, the monsoon, and rice yield in India. Clim. Change 111, 411–424. https://doi.org/10.1007/
s10584-011-0208-4.
Bark, R.H., Martin-Ortega, J., Waylen, K., 2021. Stakeholders’ views on natural flood management: implications for the nature-based solutions paradigm shift?
Environ. Sci. Policy 115, 91–98. https://doi.org/10.1016/j.envsci.2020.10.018.
Cabinet Office, Government of Japan, 2021. Disaster management in Japan. https://www.bousai.go.jp/updates/r2_07ooame/index.html. (Accessed 31 March 2022)
(in Japanese).
Cooper, M.M.D., Patil, S.D., Nisbet, T.R., Thomas, H., Smith, A.R., McDonald, M.A., 2021. Role of forested land for natural flood management in the UK: A review.
WIREs Water 8, 1–16. https://doi.org/10.1002/wat2.1541.
Duan, Q., Sorooshian, S., Gupta, V., 1992. Effective and efficient global optimization for conceptual rainfall-runoff models. Water Resour. Res. 28, 1015–1031.
https://doi.org/10.1029/91WR02985.
Ferguson, C., Fenner, R., 2020. The impact of natural flood management on the performance of surface drainage systems: a case study in the Calder Valley. J. Hydrol.
590, 125354 https://doi.org/10.1016/j.jhydrol.2020.125354.
Hallegatte, S., Green, C., Nicholls, R.J., Corfee-Morlot, J., 2013. Future flood losses in major coastal cities. Nat. Clim. Change 3, 802–806. https://doi.org/10.1038/
nclimate1979.
Hirabayashi, Y., Kanae, S., Emori, S., Oki, T., Kimoto, M., 2008. Global projections of changing risks of floods and droughts in a changing climate. Hydrol. Sci. J. 53,
754–772. https://doi.org/10.1623/hysj.53.4.754.
Hirabayashi, Y., Mahendran, R., Koirala, S., Konoshima, L., Yamazaki, D., Watanabe, S., Kim, H., Kanae, S., 2013. Global flood risk under climate change. Nat. Clim.
Change 3, 816–821. https://doi.org/10.1038/nclimate1911.
Howgate, O.R., Kenyon, W., 2009. Community cooperation with natural flood management: a case study in the Scottish borders. Area 41, 329–340. https://doi.org/
10.1111/j.1475-4762.2008.00869.x.
Huang, C.C., Tsai, M.H., Lin, W.T., Ho, Y.F., Tan, C.H., 2006. Multifunctionality of paddy fields in Taiwan. Paddy Water Environ. 4, 199–204. https://doi.org/
10.1007/s10333-006-0049-3.
IPCC, 2021. Climate change 2021: The physical science basis. https://www.ipcc.ch/report/sixth-assessment-report-working-group-i/. (Accessed 3 December 2021).
Kato, T., Satoh, M., 2002. Flood mitigation function and its stochastic evaluation of the Matsuzawa-ike Pond, Osaka Prefercture, Japan. JSIDRE 70 (6), 637–644.
https://doi.org/10.11408/jsidre1965.2002.637.
Kazama, S., Sato, A., Kawagoe, S., 2009. Evaluating the cost of flood damage based on changes in extreme rainfall in Japan. Sustain. Sci. 4, 61–69. https://doi.org/
10.1007/978-4-431-99798-6_1.
Kim, T.C., Gim, U.S., Kim, J.S., Kim, D.S., 2006. The multi-functionality of paddy farming in Korea. Paddy Water Environ. 4, 169–179. https://doi.org/10.1007/
s10333-006-0046-6.
Knapp, A.K., Beier, C., Briske, D.D., Classen, A.T., Luo, Y., Reichstein, M., Smith, M.D., Smith, S.D., Bell, J.E., Fay, P.A., Heisler, J.L., Leavitt, S.W., Sherry, R.,
Smith, B., Weng, E., 2008. Consequences of more extreme precipitation regimes for terrestrial ecosystems. BioScience 58, 811–821. https://doi.org/10.1641/
B580908.
Kobayashi, K., Kono, Y., Kimura, T., Tanakamaru, H., 2021. Estimation of paddy field dam effect on flood mitigation focusing on Suse region in Hyogo. Jpn. Hydrol.
Res Lett. 15 (3), 64–70. https://doi.org/10.3178/hrl.15.64.
Kotera, A., Nawata, E., Thao, L.V., Vuong, N.V., Sakuratani, T., 2005. Effect of submergence on rice yield in the red river delta, Vietnam. Jpn. J. Trop. Agric. 49,
197–206. https://doi.org/10.11248/jsta1957.49.197.
Kotera, A., Nawata, T., 2007. Role of plant height in the submergence tolerance of rice: a simulation analysis using an empirical model. Agric. Water Manag. 89,
49–58. https://doi.org/10.1016/j.agwat.2006.12.002.
Kundzewicz, Z.W., Kanae, S., Seneviratne, S.I., Handmer, J., Nicholls, N., Peduzzi, P., Mechler, R., Bouwer, L.M., Arnell, N., Mach, K., Muir-Wood, R., Brakenridge, G.
R., Kron, W., Benito, G., Honda, Y., Takahashi, K., Boris, S., 2013. Flood risk and climate change: global and regional perspectives. Hydrol. Sci. J. 59, 1–28.
https://doi.org/10.1080/02626667.2013.857411.
Lane, S.N., 2017. Natural flood management. WIREs Water 4, e1211. https://doi.org/10.1002/wat2.1211.
Li, C., Zwiers, F., Zhang, X., Li, G., Sun, Y., Wehner, M., 2021. Changes in annual extremes of daily temperature and precipitation in CMIP6 models. J. Clim. 34,
3441–3460. https://doi.org/10.1175/JCLI-D-19-1013.1.
Loo, Y.Y., Billa, L., Singh, A., 2015. Effect of climate change on seasonal monsoon in Asia and its impact on the variability of monsoon rainfall in Southeast Asia.
Geosci. Front. 6 (6), 817–823. https://doi.org/10.1016/j.gsf.2014.02.009.
Matsuno, Y., Nakamura, K., Masumoto, T., Matsui, H., Kato, T., Sato, Y., 2006. Prospects for multifunctionality of paddy rice cultivation in Japan and other countries
in monsoon Asia. Paddy Water Environ. 4, 189–197. https://doi.org/10.1007/s10333-006-0048-4.
Mihara, M., Yasutomi, R., Aiba, N., 1994. Levee slope protection measures taking account of soil structural changes. Journal of Japanese Society of Irrigation. Drain.
Reclam. Eng. 62 (4), 315–319. https://doi.org/10.11408/jjsidre1965.62.4_315.
Milly, P.C.D., Wetherald, R.T., Dunne, K.A., Delworth, T.L., 2002. Increasing risk of great floods in a changing climate. Nature 415, 514–517. https://doi.org/
10.1038/415514a.
Min, S.K., Zhang, X., Zwiers, F., Hegerl, G.C., 2011. Human contribution to more-intense precipitation extremes. Nature 470, 378–381. https://doi.org/10.1038/
nature09763.
Minakawa, H., Masumoto, T., Kudo, R., 2014. Development of a diurnal rainfall pattern generator incorporating characteristics of long- and short-term rainfall
intensity. Trans. Japan. Soc. Irrig. Drain. Rural Eng. 82 (3), 147–156. https://doi.org/10.11408/jsidre.82.147.
Minakawa, H., Kitagawa, I., Masumoto, T., 2016. Estimation method of flood damage on rice paddies for watershed management as flood risk mitigation. Transactions
of The Japanese Society of Irrigation. Drain. Rural Eng. 84 (3), I_271–I_279. https://doi.org/10.11408/jsidre.84.I_271〉.
Minakawa, H., Miyazu, S., 2022. Functional evaluation of paddy field dam devices with different characteristics using generated heavy rainfall data. Trans. Jpn. Soc.
Irrig., Drain. Rural Eng. 90 (1), I_157–I_165. https://doi.org/10.11408/jsidre.90.I_157.
Ministry of Land, Infrastructure, Transport and Tourism (MLIT), 2021. Bills related to river basin disaster resilience and sustainability by All. 〈https://www.mlit.go.
jp/river/kasen/ryuiki_hoan/20210202.html〉. (Accessed 3 December 2021) (in Japanese).
Mirza, M., 2011. Climate change, flooding in South Asia and implications. Reg. Environ. Change 11, 95–107. https://doi.org/10.1007/s10113-010-0184-7.
Miyazu, S., Yoshikawa, N., Abe, S., Misawa, S., Yasuda, H., 2012. Development and application of evaluation model for interior flood damage mitigation effect of a
paddy field dam. Trans. Japan. Soc. Irrig. Drain. Rural Eng. 80 (6), 479–488. https://doi.org/10.11408/jsidre.80.479.
Miyazu, S., Yoshikawa, N., Abe, S., Misawa, S., 2013. Economic evaluation of paddy field dam in the case of polder area. J. Jpn. Soc. Civ. Eng. 69, I_1531–I_1536.
https://doi.org/10.2208/jscejhe.69.i_1531.
Moriashi, D.N., Gitau, M.W., Pai, N., Daggupati, P., 2015. Hydrologic and water quality models: performance measures and evaluation criteria. Trans. ASABE 58,
1763–1785. https://doi.org/10.13031/trans.58.10715.
Nara Prefecture, 2018a. Yamato-gawa ryuiki ni okeru sougou-tisui no suisin ni kansuru jyourei nituite. http://hywr.kuciv.kyoto-u.ac.jp/renkei/activities/20181211/
03_hayashi.pdf. Accessed 18 July 2020) (in Japanese).
Nara Prefecture, 2018b. Yamato-gawa ryuiki tyousei-ti gijyutu-kijyun. https://www.pref.nara.jp/secure/187197/tyouseitikijyun.pdf. (Accessed 6 December 2021)
(in Japanese).
Nara Prefecture, 2018c. Yamato-gawa ryuiki suiden-tyoryu ni kansuru gijyutu-kijyun. https://www.pref.nara.jp/secure/187197/suidentyoryuukijyun.pdf. (Accessesd
6 December 2021) (in Japanese).
Nicholson, A.R., O’Donnell, G., Wilkinson, M., Quinn, P.Q., 2020. The potential of runoff attenuation features as a natural flood management approach. J. Flood Risk
Manag. 13 (S1), e12565 https://doi.org/10.1111/jfr3.12565.
Oishi, T., Nakano, R., Matsuno, Y., 2019. Perception and valuation of paddy field dam functions by rural communities: a CVM approach. Paddy Water Environ. 17,
383–390. https://doi.org/10.1007/s10333-019-00733-2.

18
Y. Suzuki et al. Journal of Hydrology: Regional Studies 42 (2022) 101165

Sato, H., 1987. Off-site detention hydraulic calculation program. Kawada-gihou. 6, 156–157. https://www.kawada.co.jp/technology/gihou/pdf/vol06/06_
program03.pdf [Accessesd December 3 (in Japanese)].
Takahasi, Y., Uitto, J.I., 2004. Evolution of river management in Japan: from focus on economic benefits to a comprehensive view. Glob. Environ. Change 14, 63–70.
https://doi.org/10.1016/j.gloenvcha.2003.11.005.
Takeda, T., Asaoka, Y., 2018. A comparison of water storage functions among drainage devices for the rice field dam. J. Jpn. Soc. Civ. Eng., Ser. G (Environ. Res. ) 74
(5), I_125–I_132. https://doi.org/10.2208/jscejer.74.I_125.
Ushiyama, M., Honma, M., Yokomaku, S., Sugimura, K., 2019. Characteristics of victims caused by heavy rainfall disaster in July, 2018. J. Jpn. Soc. Nat. Disaster Sci.
38, 29–54. https://doi.org/10.24762/jndsj.38.1_29.
Wheater, H.S., 2006. Flood hazard and management: a UK perspective. Philos. Trans. R. Soc. 364, 2135–2145. https://doi.org/10.1098/rsta.2006.1817.
Westra, S., Fowler, H.J., Evans, J.P., Alexander, L.V., Berg, P., Johnson, F., Kendon, E.J., Lenderink, G., Roberts, N.M., 2014. Future changes to the intensity and
frequency of short-duration extreme rainfall. Rev. Geophys. 52, 522–555. https://doi.org/10.1002/2014RG000464.
Wu, R.S., Sue, W.R., Chien, C.B., Chen, C.H., Channg, J.S., Lin, K.M., 2001. A simulation model for investigating the effects of rice paddy fields on the runoff system.
Math. Comput. Model. 33, 649–658. https://doi.org/10.1016/S0895-7177(00)00269-7.
Yoshikawa, N., Nagao, N., Misawa, S., 2009a. Evaluation of runoff detention function of the paddy field installed with runoff control devices. Transactions of The
Japanese Society of Irrigation. Drain. Rural Eng. 77 (3), 263–271. https://doi.org/10.11408/jsidre.77.263.
Yoshikawa, N., Nagao, N., Misawa, S., 2009b. Watershed scale evaluation of flood mitigation function of paddy fields installed with runoff control devices.
Transactions of The Japanese Society of Irrigation. Drain. Rural Eng. 77 (3), 273–280. https://doi.org/10.11408/jsidre.77.273.
Yoshikawa, N., Nagao, N., Misawa, S., 2010a. Evaluation of the flood mitigation effect of a Paddy field dam project. Agric. Water Manag. 97, 259–270. https://doi.
org/10.1016/j.agwat.2009.09.017.
Yoshikawa, N., Koide, H., Misawa, S., 2010b. Requirements for vertically installed runoff control boards for the “paddy field dam” and appropriate orifice shapes.
Trans. Jpn. Soc. Irrig., Drain. Rural Eng. 78 (4), 273–279. https://doi.org/10.11408/jsidre.78.273.
Yoshikawa, N., Arita, H., Misawa, S., Miyazu, S., 2011a. Evaluation of social function of paddy field dam and its technical prospects. J. Jpn. Soc. Hydrol. Water
Resour. 24 (5), 271–279. https://doi.org/10.3178/jjshwr.24.271.
Yoshikawa, N., Miyazu, S., Yasuda, H., Misawa, S., 2011b. Development of inundation analysis model for low-lying agricultural reservoir. J. JSCE 67, I_991–I_996.
https://doi.org/10.2208/JSCEJHE.67.I_991.
Yoshikawa, N., Tsubaki, K., 2016. Measure scheme for supporting sustainability of paddy field dam. J. Jpn. Soc. Irrig., Drain. Reclam. Eng. 84 (4), 271–274. https://
doi.org/10.11408/jjsidre.84.4_271.

19