The Research Based on SWAT Model of Rainfall-

runoff Process Simulation in Zimu River in Nanxi

River Basin

Zhenling1,2, Zhengtao Shi1*,Wenfei Xi1,JieLi1,BinSu1

1
College of Tourism and Geographic Sciences, Yunnan Normal University,
Kunming 650050, Yunnan, China.
2
College of Urban and Rural Construction & Engineering Management ,
Kunming University, Kunming 650214, Yunnan, China.
*Corresponding author, e-mail: Shi [email protected]

Abstract—This study, based on the experimental observation Scholars at home and abroad have done a lot of research work
of small watershed, uses the GIS platform and the remote about runoff simulation. The research group headed by Arnold,
sensing data as the distributed hydrological model SWAT.It one of the SWAT model developers [5], selected the regions
uses the measured data to simulate the rainfall runoff process with different hydrological conditions and different terrain
in the small watershed of Zi Mu river. The results show that features. The team carried out the application of SWAT model
the runoff is more consistent and the model efficiency according to the national scale with the medium and large
coefficient is 0.59 after the parameters are determined. The basin scale and the small watershed scale respectively. It is in
model can better simulate the rainfall runoff process in the order to verify the applicability of the model in the runoff
study area. Through the determined sensitivity parameters， simulation; Bingne R [6] uses the SWAT model to study the
the sensitivity parameters of the rain season runoff in the study Mississippi River Branch Goodwin River Basin. The results
area were identified which will provide reference for the show that the annual runoff simulation values of all subbasins
hydrological modeling of the Nanxi river state. in 10 years have reached 76% of the observed values and the
simulated values of most of the sub basins have reached 90%
Keywords—SWAT Model;Rainfall-runoff Process Simulation; of the observed values. Meanwhile, the study also pointed out
Zimu River,Nanxi River Basin that the SWAT model is for long-term watersheds. The
hydrological simulation results were significantly better than
the shorter ones. In China, Wang Zhonggen [7] introduced the
I. INTRODUCTION SWAT model successfully into the above-watershed above the
In the context of global climate change, ecological Luoxia Gap of the Heihe River. It was conducted a local daily
hydrology and watershed management science have developed runoff simulation using GIS. The simulation accuracy was
rapidly in the world [1]. Recently research needs to consider 0.83 and the effect was satisfactory. Sheng Chunshu et al. [8]
ecological processes and the hydrological processes at used the SWAT model to predict the runoff in the Feng Le
different scales of time and space evolution with mutual River Basin which estimate and predict the local hydrological
influence, interaction and coupling evolution which as a risk through the integrated model and the meteorological
theoretical basis for the integrated management and prediction model. It obtains the distribution process of the
sustainable development of the river basin[2].As a new type of annual runoff in the basin. The results can provide the basis
hydrological research and watershed management tool, for the water resource management of the basin. Influenced by
watershed hydrological simulation models are increasingly the complex parameters of the SWAT model which the
accepted by hydrology and watershed managers. classification standard of soil and land cover and the
SWAT(Soil and Water Assessment Tool) model is a inconsistency of the domestic.The SWAT model has been
scale-based long-term distributed watershed hydrological applied in the experimental application stage of the typical
model developed on the basis of the SWRRB model [3, 4]. It is typical basin in China and the high precision [9] has been
able to fully consider climate and underlying surface factor of obtained in the simulation of the large and middle basin.
uneven distribution of space and the fact that the real Considering SWAT model is less studied in the application of
simulation of rainfall runoff formation. The model can small watershed in complex mountain areas.This paper selects
objectively reflect the influence of the real world that the the Zi Mu River in the red Henan River Basin as the study
climate and the spatial distribution of underlying surface area and discusses the applicability of SWAT model in the
factors on the impact of the change of runoff and water application of runoff simulation in a small watershed.
reserves.

Authorized licensed use limited to: NWFP UNIV OF ENGINEERING AND TECHNOLOGY. Downloaded on July 27,2021 at 08:39:07 UTC from IEEE Xplore. Restrictions apply.
 Tab. 1 Research area data

Data type Data description Data source Data Property

Yunnan surveying
Vegetation type is and mapping
Vegetation
based on Geo-1 data of geographic Grid layer
database
August 25, 2017 information
bureau.
Daily maximum
Yunnan
temperature, minimum
meteorological Attribute
temperature,
Meteorologic bureau/Yunnan
precipitation, solar data
al data hydrology and
radiation, average daily
water resources
wind speed and relative
bureau.
humidity;
Soil bulk density, Soil
mechanical Soil institute of
composition and Soil Chinese academy Vector/raster
Fig 1 Research Area Soil database
organic matter content, of sciences And layer.
Root depth, Soil Sampling
II. THE PROPOSED EXTRACTION METHOD thickness.

The operation of SWAT hydrological model requires a Hydrological The flow observation
Yunnan Attribute
lot of data including meteorological data, DEM, land use, soil hydrology and data
observation data of nanxi river
water resources
type, hydrological data and so on. It is necessary to use GIS, data water station.
bureau.
RS, mathematical statistics and other means to process these
data into a specific format that can be identified by the model
to establish the spatial database and attribute database of the 2.3 The theory of SWAT model
research area for use in the model running. Due to the large The process of surface runoff simulation of SWAT
number of data required for model operation and the need to includes hillside confluence and river network confluence.The
unify geographic coordinates and projections. The study former refers to the process of net rain flowing into the river
selected Albers Conical Equal Area projections.The network from the surface of the slope.The latter refers to the
geographic coordinates were chosen as WGS-1984. process of the flow of slope land into the river network in the
2.1 Research area river network. [12]
Nanxi river is a tributary of Honghe River in Yunnan According to the hydrological cycle theory, the SWAT
province which originated from MABA mountain in the hydrological model is based on the following water balance
southeast of Ming Jiu town in Mengzi. The East is adjacent to equation.
t
the Panlong River and the west is adjacent to the Red river.
SW t  SW 0   ( R i  Q i  ET i  Pe i  QR i )
Mengzi, Pingbian and estuarine counties of Honghe i 1
state,Wenshan county and Maguan County of Wenshan.It SWt is the ultimate soil water content (mm); SW0 is the
remit to Honghe River at the Honghe exit of the Hekou county original soil water content that can be absorbed by vegetation
where elevation of 76.4 m is the lowest place in Yunnan on the day of i, and defined as the original soil moisture
province. The total length is 169.2 km, the drop is 2045 m, content minus the withering point water content (mm); ti is
and the catchment area is 3354.7 km 2 [10]. time (days); Ri is the i day precipitation (mm); Pei is the
Zi Mu river basin is located in the Laofanzhai permeability and flow rate (mm) existed at the bottom of the
Mountains in the lower reaches of the Nanxi River and soil profile , and QRi was the groundwater return volume (mm)
belongs to the sub-basin of the Nanxi River. It is a tributary of on the first day of i.
the Red River. The drainage area is 24.3km2, the length of the
river is 17.0km. The southwest monsoon and the southeast III. MODEL CONSTRUCTION
monsoon alternated which belonged to the humid climate of
tropical montane rain forest. On the banks of the Zimu River, 3.1 Sub-basin division
there is a tropical primitive gully rain forest landscape. Sub basin division is the basis of the SWAT model and
Because of its low altitude, hot climate and abundant rainfall, the result of the partition directly affects the accuracy of
forest resources are unique. [11] subsequent simulation [13].In order to make the sub basin
2.2 Data accord with the real distribution law of the river network as far
The data used in the model include two parts: spatial as possible. The DEM was corrected using the measured river
data and attribute data. Among them, spatial data includes network data before subwatershed division. In this study, the
meteorological data, water system, land use, soil type. The subbasin division should first determine the outlet position of
attribute data includes daily runoff, precipitation, wind speed, the sub basin. The nearest grid flow in the upper reaches of the
solar radiation,air relative humidity, maximum air temperature, two river channels is taken as the drainage area if the location
minimum temperature, soil properties, and so on. The data of the sub basin is unknown. Manillapalli and others found
sources and specific conditions of the research area are shown that the runoff prediction of SWAT model has a certain
in Table 1. sensitivity to the number of watershed and hydrological

Authorized licensed use limited to: NWFP UNIV OF ENGINEERING AND TECHNOLOGY. Downloaded on July 27,2021 at 08:39:07 UTC from IEEE Xplore. Restrictions apply.
 response units. The larger scale threshold affects the actual Tab. 2 Sensitivity analysis and rate determination results of SWAT model
parameters
distribution of CN values thus reducing the accuracy of the
prediction. In this study the catchment area is 25ha, and 29 sub Parameter Adjust the final value arrange mode
watershed maps are drawn out.
3.2 HRU division Alpha_Bf 0.075 Repalce by Value
The most basic unit of calculation for the SWAT model CN2 -20 Add to value
is the hydrological response unit (HRU). The model provides
ESCO 0.85 Repalce by Value
two methods of dividing HRUs: one is to consider the
dominant soil type and type of land use. Then it is to CH_K2 25 Repalce by Value
determine a threshold for soil type and land use type and to be
SOL_K +20 Add to value
divided into the dominant type below this threshold type. Soil
types and land use within the basin where the area ratio GWQMN 3 Repalce by Value
exceeded the threshold were determined as HRUs (Multi-type
SLOPE 0.35 Repalce by Value
HRUs) within the study area. The other is to consider the main
soil types and land use combinations in sub basins. Only one Sol_Awc -0.06 Add to value
HRU (Domainant HRUs) is formed in each sub basin. In this LAT_TTIME 3 Repalce by Value
study the hydrological response unit was divided by the first
method and the land use and soil thresholds were 10%,165 SLSOIL 160 Repalce by Value
hydrological response units were divided in 29 sub basins. *Alpha_Bf is the coefficient of the groundwater flow response to the supply
3.3 Parameter sensitivity analysis and rate determination change;CN2 is the parameters of the hydrological response unit scale; ESCO
is the control parameters of groundwater through capillary action and soil
The sensitivity analysis in the SWAT model uses the
fracture compensation for soil evaporation ;CH_K2 is the flow loss caused by
LH-OAT sensitivity analysis method [14] which refers to the the resupply of groundwater;SOL_K is the movement rate of soil moisture
OAT sensitivity analysis at each sampling point (LH sampling along the hydraulic gradient direction;GWQMN is to control the recharge of
method). The final sensitivity is the average of the local surface runoff from shallow groundwater;SLOPE is the average slope of a
subbasin;Sol_Awc is the amount of water available to plants in the
sensitivity. Using this method, we can effectively get the main
soil;LAT_TTIME is the change of flow rate of runoff in soil in
parameter factors that affect the results of the model and time.;SLSOILis the slope of the soil is long.
greatly improve the availability of the model. If the sensitivity (2) The SWAT model can better simulate the rainfall
factor is less than 0.05, the sensitivity is low. If the factor runoff process in the study area then the determining
value is more than 1, the sensitivity is high. Based on the coefficient and the Nash coefficient of the simulated value and
results of LH OAT sensitivity analysis of watershed the measured value are 0.67 and 0.59 respectively in the study
parameters, we have to adjust the sensitive parameters that are area. In the case of no rainfall or less rainfall, the simulation
more influential. In this paper, the sensitivity analysis based results of the model are relatively good. In the case of long
on the measured data and not based on the measured data is rainfall with longer duration, the model simulation results after
carried out in the sensitivity analysis of the parameters. The a period of rainfall are poor.
use of observation data has no effect on the first four model
parameters but the sensitivity of the parameters is obviously
improved after the use of the observed data. Therefore, The
actual measurement data should be used as much as possible
in the parameter sensitivity analysis of the SWAT model.
3.4 Simulation Results Analysis
The 20 day runoff process at upstream of Nanxi
hydrological station is simulated by using the SWAT model.
The Nash efficiency coefficient, mean relative error and
correlation coefficient are used to reflect the goodness of fit
and correlation degree between simulated values and observed
values, respectively.
(1) The measured data has an important influence for Fig 2 The runnoff line of the Observed and the Simulated
parameter sensitivity analysis in the SWAT model. So the IV. CONCLUSIONS
measured data should be used in the model parameter
adjustment. Except the 26 parameters in sensitivity The evaluation index value of the runoff simulation
analysis,lateral flow travel time (LAT_TIME) and slop length accuracy of the Nanxi River Valley is within the permitted
for lateral subsurface flow(SLSOIL)are also sensitive for the range. The developers of the SWAT model think that the
simulated result. After calibrating, the best parameters for model efficiency coefficient of the runoff simulation is above
runoff simulation in SWAT model are: 0.5 which means that the model can reach the reasonable
simulation requirement. The model efficiency system number
is 0.59 after the parameter rate is fixed. The model after the
rate is determined can be set. It is better to simulate the rainfall

Authorized licensed use limited to: NWFP UNIV OF ENGINEERING AND TECHNOLOGY. Downloaded on July 27,2021 at 08:39:07 UTC from IEEE Xplore. Restrictions apply.
 runoff process in the study area. In this paper, the SWAT [4] Krysanova V, Srinivasan R. Assessment of climate and land use change
impacts with SWAT[J]. Regional Environmental Change, 2015,
model is used to simulate the rainfall runoff process in small 15(3):431-434.
watershed. It is short in the study period and does not consider [5] Srinivasan R, Arnold J G, Jones C A. Hydrologic modeling of the
the influence of vegetation and terrain factors on runoff. In United States with the soil and water assessment tool[J]. Water
future research, we will introduce a model study on the Resources Development，1988, 4(3):315 -325．
interaction between vegetation, terrain change and [6] Krysanova V, White M. Advances in water resources assessment with
hydrological processes from the perspective of ecology and SWATâ an overview[J]. International Association of Scientific
Hydrology Bulletin, 2015, 60(5):771-783.
geography. The flow curves of the model simulation and
[7] Lv leting, Peng qiuzhi, Guo yuanyuan, et al. Runoff simulation of
measured flow are shown in the fig.2. It can be seen from the dongjiang river basin based on SWAT model [J]. Journal of natural
fig.2 that the simulation results of the model are relatively resources, 2014(10):1746-1757.
good in the absence of rainfall or small rainfall, and in the case [8] Sheng Chunshu, Luo Dinggui. Based on AVSWAT, Feng Le River
of long-duration rainfall, the model simulation results after a Basin Hydrological Prediction [J]. Chinese agronomy bulletin, 2006, 22
(9): 493- 497.
period of rainfall are poor.
[9] Yuan Y Z, Zhang Z D, Meng J H. Impact of changes in land use and
ACKNOWLEDGMENT climate on the runoff in Liuxihe Watershed based on SWAT model[J].
Ying Yong Sheng Tai Xue Bao, 2015, 26(4):989-998.
We thanks to our tutors Professor Shi zhengtao in [10] Yunnan hydrology and Water Resources Bureau. Yunnan river lake [M].
Yunnan Normal University.He gives our guidance from paper Kunming: Yunnan science and Technology Press, 2010.
topics, structural design, materials collect and write.We [11] Zhu Juying. Analysis of the characteristics of Nanxi River in Honghe
appreciate the assistance from the Ministry of Forests,Yunnan River Basin [J]. people's Pearl River, 2012 sixth, P19-21.
Provincial Government and Hydrological Bureau of Honghe, [12] Luo Qiao, Wang Kelin, Wang Qingxue. Study on the runoff simulation
of land use change scenarios in Xiangjiang basin based on SWAT model
Yunnan Province. [J]. Chinese Journal of ecologic agriculture, 2011, 19 (6): 1431 - 1436.
[13] Volk M, Sauvage S, Sánchez-Pérez J M, et al. Development and
REFERENCES applications of the SWAT model to support sustainable river basin
[1] Bussi G, Dadson S J, Prudhomme C, et al. Modelling the future impacts management on different scales[J]. Sustainability of Water Quality &
of climate and land-use change on suspended sediment transport in the Ecology, 2016, 8:1-3.
River Thames (UK)[J]. Journal of Hydrology, 2016, 542:357-372. [14] Li Lei, Dong Xiaohua, Yu Dan, and so on. Runoff simulation in
[2] Wang Chao. Changes of vegetation runoff in the Tian-La-Chi watershed Qingjiang basin based on SWAT model [J]. Yangtze River, 2013, 44
on the upper reaches of the Heihe River on rainfall-runoff process (22): 25-29.
[D] .2013.
[3] Chen Y R, Yu B. Impact assessment of climatic and land-use changes on
flood runoff in southeast Queensland[J]. International Association of
Scientific Hydrology Bulletin, 2015, 60(10):1759-1769.

Authorized licensed use limited to: NWFP UNIV OF ENGINEERING AND TECHNOLOGY. Downloaded on July 27,2021 at 08:39:07 UTC from IEEE Xplore. Restrictions apply.