22-1, 307 -310 (2012 )

"ICAL 1 / DT X"
Refereed Paper

Journal of Arid Land Studies

Effect of Difference in Rainfall Patterns and Intensities

on Runoff Simulation Results in Arid Land
Osamu YOKOHAGI1), Hironori TABUCHI1), Hideki SUGANUMA1),
Katsuhiko KUROSAWA1) and Toshinori KOJIMA*1)2)
Abstract: Currently, a sustainable carbon fixation system with large scale afforestation in arid land has been established for the
mitigation of global warming. We have set a research site near Leonora (rainfall 200 mm: evapotranspiration 2,400-2,600 mm) in
Western Australia. In arid land, the most important problem is water shortage for afforestation, because of the small amount of
rainfall. In addition, large amount of runoff water flows out without being used by plants in this area. In this study, we developed an
original runoff model for understanding water behavior to select appropriate afforestation sites. This model consists of three
simulation processes which are penetration, evaporation and runoff as water behavior. This model has two fitting parameters. One
is corrective coefficient of penetration (PR) and the other is equivalent roughness coefficient (N). To estimate these two parameters, a
closed hydrographic basin (6 5 km2) including a water collecting pond and its DEM (Digital Elevation Model) data with spatial
resolution of 10 meters were used. Rainfall gauges with rainfall amount resolution and time resolution as 0.2 mm and 0.5 s,
respectively, were set near the main creek flowing into the pond, and water level indicator was set in the pond. This model simulation
was conducted using one hour cumulative data or one minute cumulative data calculated from these rainfall measurement data
obtained on 24th March 2003. We also evaluated effects of various patterns of rainfall intensities on calculated results of runoff water
behavior using other rain data.
Key Words: Afforestation, Arid land, Carbon fixation, Hydrologic simulation, Runoff model

2. Materials and Methods

1. Introduction
For the mitigation of global warming, we have to establish
a sustainable carbon fixation system, thus large scale
afforestation of arid land which was not used for crop
cultivation, has been proposed and tested in Western Australia.
The biggest problem for establishment of afforestation is the
lack of available water because of the rainfall shortage and the
large amount of runoff water loss by evaporation. To avoid
such disadvantages, our afforestation system has adopted the
water-harvesting system. Also the afforestation area should
be selected to use runoff water efficiently (Yamada et al., 2003).
In the present series of studies, we aimed to develop an
original runoff model for arid land afforestation, and select
large scale afforestation by numerical calculation about water
behavior. This study focused on the effect of difference in
rainfall patterns and intensities of input rainfall data on original
runoff model. We compare different patterns of rainfall data
input (1 minute cumulative input and 1 hour cumulative input).
Also we used not only ideal rainfall data with almost same
rainfall at the three rainfall gauges but also other rainfall data
comparing results of numerical calculation.
* Corresponding Author: [email protected]
3-3-1, Kichijoji-kitamachi, Musashino, Tokyo, 180-8633, Japan
1) Department of Materials and Life Science, Faculty of Science and Technologies, Seikei University, Japan
2) Centre for Ecosystem Management, School of Natural Sciences, Edith Cowan University, Australia

2.1. Research area

We have set a research place (6 5 km2) which includes an
enclosed basin (inside dot-line) near Sturt Meadows station
(STM), Leonora, Western Australia as shown in Figure 1. In
this area, most runoff flows into a catchment pond called Jims
pool. Average annual rainfall of STM is about 200 mm and
evapotranspiration is about 2,400-2,600 mmA), which is
classified into arid land (Kojima et al., 2006). Furthermore,
hardpan layer with extremely low permeability locates between
about several tens centimeters deep and a few tens meters deep.
2.2. Observed data
Rainfall gauges (resolution: 0.2 mm, measurement unit: 0.5
s) were installed at three points along the main creek to
catchment pond (shown by the squares in Fig. 1). Water level
gauge installed at catchment pond (shown by the circle in Fig. 1)
and water level data were obtained for every hour (Hamano et
al., 2010). In this study, we adopted a rainfall data recorded on
24th March 2003. Those data recorded with the three rainfall
gauges were very similar, and it was considered as a uniform
rainfall event throughout the target area. We used the data as

of the fitting parameters in this model was employed for

correcting difference between experiment and actual
penetration rate as shown in Eq. (4) (Kojima et al., 2010).

f t = f c + ( f 0 f c ) exp( K f t ) =

2.00 10 5 + 3.06 10 5 exp 3.42 10 3 t

(1)

f = 1.530 10 3 3 + 1.487 10 3
2 4.782 10 4 + 7.115 10 5

=W/0.15

(2)

(3)

F = f PR
( 4)

Fig. 1. Research area and locations of instruments.

an ideal rainfall event for the present model development and

validation.
2.3. Concept of original runoff model
The present runoff model has been developed to predict
water behavior in arid area by inputting only Digital Elevation
Model (DEM) data and rainfall data. However, this model
needs some measured parameters such as soil depth and initial
volumetric water content of soil [-] as local environmental
conditions. In this study, we adopted 0.15 m for soil depth as
the measured averaged data and 0.05 for the initial volumetric
water content of soil, the lowest value of the range of possible
value between 0.05 and 0.4 (Kojima et al., 2010), because of
the actual dry condition.
Time difference method with an interval (time step) for
approximating real-time progress was employed to the model.
As the water movement expression, penetration, evaporation
and runoff after rainfall were calculated in all meshes. DEM
with 10m grid spacing (10 m mesh) was purchased from
Kevron Aerial Survey Pty. Ltd. and employed as the elevation
data in this model. Previous studies revealed that calculation
results were affected by time step. The optimal time step
depends on the mesh size (Kojima et al., 2010). In this study,
optimal time step was set as 0.5 s for 10 m mesh.
2.4. Penetration
Penetration rate was estimated by analyzing the results of
cylinder intake rate method with Hortons equation (Eq. (1))
(Horton, 1939). In our model, we modify Eq. (1) giving time
variation of penetration rate to give Eq. (2) by giving
penetration rate as a function of volumetric water content of
soil , which is calculated by Eq. (3)
Also penetration rate adjustment factor (PR) which was one

ft : Permeation rate at time, t [m/s], f0 : Initial permeation rate

[m/s], fc : Saturated hydraulic conductivity [m/s], Kf : Constant
value [s-1], t : Time [s], f : Permeation rate at volumetric water
content of soil, [m/s], : Volumetric water content of soil [-],
W : penetrated water depth [m], F : Corrected permeation rate
[m/s], PR : Correction factor for f [-]
2.5. Evaporation
Evaporation rate was evaluated as the sum of that from
surface water and that from soil. Water surface evaporation
rate was given as annually average pan evaporation rate as
shown in Eq. (5). Evaporation rate from soil was divided into
two drying steps from the result of evaporation experiment
from soil. Constant evaporation rate ES1 was considered to be
almost same as evaporation rate from surface water. Thus,
ES1 was supposed to be equal to EL as shown in Eq. (6). And
also, decreasing evaporation rate ES2, which is slower
evaporation rate with decreasing water content of soil, was
estimated from evaporation experiment from soil as shown in
Eq. (7) (Kojima et al., 2010).

E L = 7.67 10 8
(5)
(6)
ES 1 = E L

ES 2 = E L 54.0 ( 0.05)

2.677

(7 )

EL : Water surface evaporation rate [m/s], ES1 : Constant

evaporation rate [m/s], ES2 : Decreasing evaporation rate [m/s]
2.6. Runoff
Runoff from the target mesh to the adjacent meshes (four
directions) was calculated by using Mannings equation (Ven Te
Chow, 1959). Mannings equation includes roughness coefficient,
n representing stream resistance, while in the present model,
equivalent roughness coefficient N was introduced as one of the
fitting parameters (Eq. (8)). Mannings equation is the equation
for a river with uniform flow, and it couldnt express water
movement at a catchment pond. Thus, in this model, error trap

was introduced in the runoff calculation. In order to avoid

phenomenon of water level oscillation, the flow of water is stopped
when levels become horizontally flat. Thus, water level
difference before one time step of runoff calculation cant be
reversed after one time step of runoff calculation (Kojima et al.,
2010).
1 2 1
v = R 3 I 2
(8)
N
v : Velocity [m/s], N : Equivalent roughness coefficient [sm-1/3]
I : Water surface inclination [m/m], R : Hydraulic radius [m]
2.7. Fitting parameters
This model includes two fitting parameters. The fitting
parameters are correction factor for ft (PR) to correct the
difference between experimental penetration rate and actual
penetration rate and equivalent roughness coefficient (N) that
represents stream resistance of soil surface. Both of the
parameters also include the averaging effect inside a wide mesh.
In the previous study, sensitivity study of N and PR was
conducted (Kojima et al., 2010).

which gives the best fit for one hour cumulative input. The
numerical results for one minute cumulative input give
seriously higher value of water level than the observed one and
numerical results from one hour cumulative input. The
present results indicate strong effects of instantaneous rainfall
intensity on the runoff phenomena. Consequently we decided
Table 1. One hour cumulative rainfall data.
Time[h]
0-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10

Rainfall[mm]
2003/3/24
2004/6/1
0
0
1.6
1.8
9.2
8.4
2.8
0.8
0
0
0
0.4
0
0
0
0
0
0
0
0

Table 2. Maximum rainfall intensities for two rainfall input patterns.

Max Rainfallintensity [mm/s]

2.9. Other rainfall data

In the previous study the runoff model was applied only to
the rainfall data on 24th March 2003. In the present study, we
also conducted the calculation for the rainfall data on 1st June
2004 and decided fitting parameters. One hour cumulative
rainfall data of the latter were shown in Table 1.

3. Results and Discussion

3.1. Effect of difference in rainfall patterns
Difference in overall cumulative rainfall data between one
hour cumulative and one minute cumulative rainfall input data
is shown in Figure 2. Difference in rainfall intensity is also
shown in Table 2. From Table 2, one minute cumulative
rainfall input gives more instantaneously strong rainfall
intensity than one hour cumulative rainfall input. The
numerical results of time variation of the water level of the
catchment pond for the different rainfall patterns are shown in
Figure 3, with the parameters of N = 0.011 s m-1/3, PR = 24.5

1 hour cumulative
1 min cumulative

0.00256
0.00667

Cumulative rainfall [mm]

20
18

1 h cumulative rainfall data

16

1 min cumulative rainfall data

14
12
10
8
6
4
2
0
0

60

120

Time [min]

180

240

300

Fig. 2. Overall cumulative rainfall data using one hour cumulative

and one minute cumulative rainfall data.
Water level at catchment pond[mm]

2.8. Rainfall input

The previous study used 1 hour cumulated rainfall data
converted from observed rainfall data (Kojima et al., 2010).
However, instantaneous intensity of rainfall is often much
larger than the time averaged one. Thus in the present study,
one minute cumulated rainfall input was used and the
calculation results of water behavior was compared with that of
1 hour cumulated rainfall input to evaluate the effect the time
resolution for rainfall.

1000
Observation

950

Calculated with
1 h cumulative
data
Calculated with
1 min cumulative
data

900
850
800
750
700
650
600
0

60

120 180 240 300 360 420 480 540 600

Time [min]

Fig. 3. Observed time variation of water level of catchment pond on

24th March 2003 and numerical results with two different
rainfall input patterns (N = 0.011 s m-1/3, PR = 24.5).

Water level at catchment pond[mm]

1000

4. Conclusion

950
Observation

900

N=0.014 s m-1/3 PR=21.0

850
800
750
700
650
600
0

60 120 180 240 300 360 420 480 540 600

Time [min]

Fig. 4. Observed time variation of water level of catchment pond on

Our original model was modified to use one minute

cumulative rainfall input, and as a result it can express more
detail response to instantaneously strong rainfall intensity than
one hour cumulative rainfall input. We decided fitting
parameters as N=0.014 s m-1/3, PR=21.0 for the data on 24th
March 2003, whose values did not differ so much. Though
the calculated results for other rainfall data on 1st June 2004 did
not fit observation data, it is expected that small modification of
the fitting parameters will give better fit of numerical results to
the observation.

24th March 2003 and numerical results with one minute

Water level at catchment pond[mm]

rainfall input data (N = 0.014 s m-1/3 and PR = 21.0).

Acknowledgement
This work was conducted under the support of the Mitsui
& Co., Ltd. Environment Fund, Grant-in-Aid for Scientific
Research (22254002) of Japan Society for the Promotion of
Science and the Global Environment Research Fund of The
Ministry of Environment of Japan (GHG-SSCP project).

800
Observation

750

N=0.014 s m-1/3 PR=21.0

700
650
600

Note
A) Bureau of Meteorology, Australia (http://www.bom.gov.au/
jsp/ncc/climate_averages/evapotranspiration/index.jsp?map
type=6&period=an#maps)

550
500
450
400
0

60

120

180

240

300

360

420

480

540

600

Time [min]

References

Fig. 5. Water level at Catchment Pond on 1st June 2004.

to use one minute cumulative rainfall input, i.e., closer to the

actual rainfall pattern, for further numerical analysis.
3.2. Fitting parameters for one minute cumulative input
We conducted the numerical calculation using original
runoff model with one minute cumulative rainfall data.
Calculation results are shown in Figure 4. The observed
water level was well reproduced with one minute cumulative
rainfall and with fitting parameters of N = 0.014 s m-1/3 and PR
= 21.0. It was demonstrated that the observed data were well
reproduced with a little bit different fitting parameters.
3.3. Calculation of other rainfall data
We conducted the numerical calculation using the original
runoff model with one minute cumulative rainfall data on 1st
June 2004 using the same fitting parameters of Figure 4.
Calculation results are shown in Figure 5. Calculation results
did not well fit the observation data. As shown in the
previous section, small difference among the rainfall gages data
may affect the calculation result of water level of catchment
pond, however small modification may improve the matching
between numerical and observed results.

Hamano H., Takahashi N., Yamada K., Kojima T., Law J.

(2010): Measurement of Runoff Ratio in a Medium-scale
Watershed for Afforestation of Arid Land. Kagaku Kogaku
Ronbunshuu, 36(4): 351-354.
Horton R.E. (1939): Analysis of runoff-plat experiments with
varying infiltration-capacity. Trans. American Geophysical
Union, 20(4): 693711.
Kojima T., Hamano H., Abe Y., Saito M., Tanouchi H.,
Egashira Y.,Takahashi N., Law J., Yamada K. (2006): Basic
Data of Research Project on Large Scale Afforestation of
Arid Land for Carbon Fixation near Leonora in Western
Australia. J. Arid Land Studies, 16(3): 167-174.
Kojima T., Koyanagi S., Tabuchi H., Suganuma H., Kurosawa
K., Takahasi N., Hamano H. (2010): Development of a
Runoff Model for Arid Land Afforestation in Western
Australia. Kagaku Kogaku Ronbunshu, 36(5): 532-538.
Ven Te Chow (1959): Open-Channel Hydraulics. McGraw-Hill
Publishing Company, Inc., New York, pp. 98-123.
Yamada K., Kojima T., Abe Y., Saito M., Egashira Y.,
Takahashi N., Tahara K., John L. (2003): Restructuring
Afforestation of hardpan Area to Sequester Carbon. Journal
of Chemical engineering of Japan, 36(3): 328-332.