Tuyển tập Hội nghị Khoa học thường niên năm 2018. ISBN: 978-604-82-2548-3

A SIMULATION OF INFLOW TO DAK WI RESERVOIR
BY THE MEAN OF SWAT MODEL
Nguyen Thanh Thuy
Faculty of Hydrology and Water resources, Thuyloi University
Email:

1. INTRODUCTION

May to October. Each season lasts for 6
The flow regime has been changed under months, however, 80-90% of yearly
precipitation is in the rainy season.
the climate and land use change conditions.
There was no hydrological station at the
To support the river basin management, a
outlet of Dak Wi reservoir. Therefore, the
mathematical model which can simulate the
influence of these two factors on the flow catchment with the outlet at Trung Nghia
hydrological station was selected for
should be used to simulate different
scenarios. The goal of this study is applying a calibration and validation. Dak Wi was
defined as a sub-basin of Trung Nghia basin
such model to simulate the flow under
current condition. The calibrated and (Figure 1.).
validated model will be the primary condition
to conduct further studies such as water
quality modelling and assessment of climate
change on water quantity and water quality.
The hydrological models are typically
classified in stochastic, lumped conceptual

and detailed physically-based models.
Among these models, only physically-based
models consider spatial distribution of
driving factors and describe the physical
processes by mathematic equations. In this
study, the semi-distributed model QSWAT
model is selected to simulate the flow to Dak
Figure 1. Trung Luong basin
Wi reservoir in Kon Tum province.
The DEM with resolution of 90 m was used
2. STUDY AREA AND DATA
( />Dak Wi reservoir is located in Dak Wi =OTSRTM.042013.4326.1). The soil type
river which is a branch of Krong Po Co river. and land use maps were subtracted from
The catchment area with the outlet at the dam Global
Land
Use
map
2
of Dak Wi reservoir is 83 km . The observed ( />data at Kon Tum meteorological station gy.php) and Digital World Soil map
indicates that the annual precipitation is ( rv/en/metada
around 2600-3000 mm and divided into dry ta.show?id=14116), respectively. There are 5
and rainy seasons. Dry season is from rain gauges including Dak Glei, Dak To, Sa
November to April and rainy season is from Thay, Kon Plong and Kon Tum are located
603


Tuyển tập Hội nghị Khoa học thường niên năm 2018. ISBN: 978-604-82-2548-3

inside and close to Trung Nghia catchment. 4. RESULTS AND DISCUSSIONS
The observed daily discharge is available

Calibration
during the period 1991-1997 at Trung Nghia
Figure a shows that the water balance was
station. Therefore, the meteorological and
simulated very well with the small
hydrological data since 1988 to 1997 was discrepancies between the simulated and
used for model calibration and validation.
observed accumulated volume for the period
1991-1995. The differences increase due to
3. METHODOLOGY
the overestimation of flow in the flood season
The SWAT model is a semi-distributed of year 1992. The gaps in accumulated
model developed by the United States volume between the simulation and
Department of Agriculture (USDA). In this measurement are eliminated in the end the
study, we used the QSWAT interface for flood season of year 1994 due to the
QGIS. The SWAT model divides the overestimation of flow in this flood season.
watershed into sub-basins connected by a In total, the accumulated volume in the
stream network, and further delineates each simulation is solely 7% higher than the
sub-basin into hydrologic response units counterpart in the measurement. Overall, the
(HRUs), which consist of unique flow in the dry season is simulated better
combinations of land cover, soil type and/or than the flow in the wet season with higher
slope. It is assumed that there is no values of NASH and R coefficients (Table 1).
interaction between HRUs (Neitsch et al.
2011). SWAT simulates each HRU separately
and calculates daily water balances (Eq. 1).
t

SWt  SWo   ( Rday  Wsurf  Ea  wse ep  Qgw )

(1)


i1

Where SWt is the final soil water content
(mm H2 0), SWo is the initial soil water
content (mm H2 0), t is the time (days), R day is
the amount of precipitation on day i (mm
H2 0), Qsu rf is the amount of surface runoff on
day i (mm H2 0), E a is the amount of
evapotranspiration on day i (mm H2 0), wseep
is the amount of evapotranspiration on day i
(mm H2 0), wseep is amount of percolation and
by pass flow exiting the soil profile bottom
on day i (mm H2 0) and Qgw is the amount of
return flow on day i (mm H2 O).
The SWAT model has been widely
calibrated and validated through the
comparison of observed and simulated
streamflow data. Most of the studies indicates
that CN2 , AWC, ESCO and SURLAG are the
most sensitive parameters to surface while
GW_ALPHA, GW_DELAP, GW_REVAP and
REVAPMN are the most sensitive parameters
to baseflow (Arnold et al. 2012).

Figure 2. Accumulated volume (top)
and flow (bottom) in calibration and
measurement at Trung Nghia station f or
the period 1991-1995
Figure b presents the flow time series in

simulation and measurement. The low flow
and median flow are simulated very well.
However, there are large differences between
high flow in simulation and high flow in

604


Tuyển tập Hội nghị Khoa học thường niên năm 2018. ISBN: 978-604-82-2548-3

measurement. The highest discharge is 2300
With the calibrated values of parameters,
m3 /s in the historical data while it is 4050 the inflow to Dak Wi reservoir is defined
m3 /s in the simulation.
with mean values of 3.43 m 3 /s and standard
Validation
deviation of 6.39 m3 /s (Figure ).
During the first 9 months of year 1996, the
Table 1. Values of NASH and correlation
accumulated volume in the simulation is very
(R) coefficients calculated for monthly
close to the counterpart in the measurement.
flow in dry and wet seasons in calibration
However, the historical data witnesses a
and validation
sharp increase in the flow since beginning of
Dry season
Wet season
October 2016 to end of December 2016. The
NASH (-)

R (-) NASH (-) R (-)
accumulative volume calculated from the
0.93
0.97
0.79
0.94
observed time series is 17% higher than the Calibrat ion
one from simulation results on on 1/1/1997. Validation
0.72
0.88
0.25
0.56
The overestimation of flow in 1997 reduces
the gaps in accumulated volume between
simulation and observation. The difference is
solely 1.3% in the end of 1997. The values of
NASH and R coefficients are high for flow in
the dry season and very low for flow in the
wet season.

Figure 4. Simulated flow time series at the
outlet of Dak Wi sub-basin
5. CONCLUSIONS

The model for Trung Nghia catchment in
SWAT has been calibrated and validated to
measured data. The set values of parameters
cannot simulate the high flow poorly for
several years. However, they perform very
well for dry season and total water volume.

They can be used for the water balance
simulation. To improve the accuracy of
Figure 3. Accumulated volume (top) and f low simulation results, more investigation should
(bottom) in validation and measurement at be conducted for parameters and input data.
Trung Nghia station for the period 1996-1997
6. REFERENCES

The flow in dry seasons are simulated very
[1] Neits ch S.L., Arnold J.G., Kiniry J.R.
well. Nevertheless, the large discrepancies
(2011). Soil & Water Assessment Tool
exist between the flow in flood seasons.
Theoretical Documentation Version 2009.
There are some flood events in the beginning
Texas, US.
of rainny seasons while there is no flood [2] Arnold, Jeffrey G et al. (2012). SWAT:
events observed in the historical data.
Model use, calibration, and validation.
Addtionally, the highest flow in the observed
Biological Systems Engineering: Papers and
3
data is 1610 m /s while it archives the value
Publications . 406.
of 3870 m 3 /s in the simulation.
605