第 2 7卷第 1期 长 江 科 学 院 院 报 Vol. 27 No. 1
2 0 1 0 年 1 月 Journa l of Yangtze R iver Sc ien tif ic Research Institute Jan . 2 0 1 0
Rece ived da te: 2009207206
Correspond ing author:WANG Zhao2hui(19782) , male, Master D. , graduated from Institute of Subtrop ical Agriculture, Chinese Academy of Sci2
ences, China, in 2004. H ismain interests are involved in research of water resource p rotection and app lication of“3S”tech2
nology. ( Tel. ) 027282826895 ( E2mail) wangzh@mail. crsri. cn
Article ID : 1001 - 5485 (2010) 01 - 0057 - 05
Research on Simulation of Non2point Source Pollution in
Q ingjiang R iver Basin Based on SWATModel and GIS
W ANG Zhao2hu i1, 2 , ZHAO D eng2zhong1 , CAO Bo1 , L IANG D ong2ye1
(1. Yangtze R iver Scientific Research Institute, W uhan 430010, China;
2. School of Resource and Environmental Science, W uhan University, W uhan 430010, China)
Abstract:A ssessment of pollution of water bodies from non2point sources (NPS) is a comp lex, requisite long series data and
time2consum ing task. The accuracy of NPS pollution models depends to a great extent on how well a model selects input pa2
rameters describing the relevant characteristics of the watershed. It is certain that the p romoting p recision of input parameters
affects simulation results of runoff, sediment and nutrients yield for the entire watershed. In this study, a basic database,
which includes DEM, soil sort and landuse map, climate data, and land management data, is established using GIS. The
generation and formation of non2point source pollution involves a great uncertainty which makes pollution monitoring and con2
trolling very difficult. Understanding the main parameters that affect NPS pollution uncertainty are necessary for p lanning and
design of control measures. On the basis of the results of parameter sensitivity analysis, the sensitive parameters of soil and
water assessment tool ( SWAT) model are identified, and then model parameters related to stream flow and nutrient loadings
are calibrated and validated by observed values. The results show that simulated values are reasonably compared with observed
data. Spatial2temporal distribution features of NPS pollution in the Q ingjiang R iver basin are revealed. NPS pollution mainly
takes p lace in flood season. The critical risk areas of soil erosion are identified. Stream flow and nutrient loadings ( including
total nitrogen ( TN ) and total phosphorus ( TP) ) in Q ingjiang R iver Basin are simulated. The surface runoff and nutrient
yield results indicate that average annual runoff and the output of TN and TP p rovide better understanding on stream flow and
nutrient loadings corresponding to various variation conditions of land use mode, agricultural tillage operation and natural rain2
fall etc.
Key words: SWAT model and GIS; sensitivity analysis; stream flow; sediment; nutrient loadings ; Q ingjiang R iver Basin
CLC num ber:X522 D ocum en t code:A
1 I NTROD UCT IO N
W ith rap id change and econom ic development,
particularly in the densely populated and econom ically
developed regions of the m iddle reach of Yangtze R iver
Basin, dramatic pollutant loadings have deteriorated
river water quality. Though point source pollution con2
trolling has achieved outstanding results, there are still
serious p roblem s because the non2point source (NPS)
pollution accounts for a large part of water pollution.
Many researches in different countries have p roved that
NPS pollution was one of the main reasons deteriorating
water quality. Finding out the effect of NPS pollution
has instructive significance on sustainable water re2
source development and water quality management in
Yangtze R iver Basin.
Hydrologic models are useful tools for understand2
ing and analyzing watershed hydrologic parameters’
change p rocesses, interactions, testing research hypot2
heses, and assessing management scenarios[ 1 ] . The
soil and water assessment tool ( SWAT) is a model de2
veloped by D r. Jeff A rnold for the USDA2ARS ( agri2
cultural research service) , which is based on physics
and p resents continuous time. It is mostly used to p re2
dict impact of land management p ractices on water,
sediment, and nutrient yields over long periods of
time[ 2 ] .
长江科学院院报 2010年
This model has been widely used. The SWAT is
p robably attributed to the comp rehensive considerations
of hydrologic, biological and environmental p rocesses,
the incorporation of management scenarios, the availa2
bility of parameter databases, and its robustness, flexi2
bility, as well as user friendliness. Heuvelmans p res2
ented a discussion in model parameter transferability
for simulating the impact of land use on catchments hy2
drology[ 3 ] . W hile the SWAT is widely app lied to a
broad range of conditions, however, a few studies have
reported on the variability and transferability of the
model parameters, and on evaluation of its crop
growth, soil water, and groundwater modules using ex2
tensive field experimental data at the p rocess scale[ 4 ] .
In this paper, the Q ingjiang R iver Basin, a main
branch of m iddle Yangtze R iver Basin, is selected as a
research area. W ater discharge and pollutant loadings
are simulated based on SWAT model to reconstruct riv2
er nutrients evolution. The critical risk areas of soil e2
rosion are identified. According to the simulation re2
sults, soil erosion moduli are calculated in each sub2
basin and soil erosion intensities are classified. Stream
flow and nutrient loadings ( including total nitrogen
( TN) and total phosphorus ( TP) ) in Q ingjiang R iver
Basin are simulated.
F ig. 1 Geograph ic loca tion and wa ter system of
Q ingjiang R iver Ba sin
2 STUDY AREA D ESCR IPT IO NS
Q ingjiang R iver Basin lies in the m iddle reach of
Yangtze R iver Basin, near the Three Gorges Project,
app roximately between 29°33′- 30°50′N and 108°35′
- 111°35′E. It is a mother river with long history of
peop le living and agricultural development. The famous
Tujia Nationality lives in this area. The terrain of
Q ingjiang R iver Basin is intensely fluctuated, from the
west high mountains and hills to the east low altitude
( Fig. 1). Climate iswarm and wet, it belongs typ ical2
ly subtrop ical monsoon climate. The average annual
temperature is 15216℃, and average annual p recip ita2
tion and evaporation are 1460 mm and 600 - 800 mm
respectively. The basin is covered naturally by ever2
green and deciduously broad2leaved, m ixed forest.
Three major soil types are found in the basin including
red2yellow soil, yellow cinnamon soil and paddy soil.
3 DATA AND M ETHOD S
3. 1 Da ta descr iption
The data for SWAT include a land use database, a
soil map, digital elevation model data (DEM) , a digital
river network and a climate database. The scale of DEM
data is 1 /50 000, and mesh size is 25 m ×25 m. In or2
der to elim inate the errors of river network extracted
from DEM data , the digital river network is revised by
water system map s in scale of 1 /10 000[ 5 ] . Besides the
DEM data, all the other data are p rojected to a 25 m ×
25 m grid data under the same reference frame.
3. 2 S im ula tion m ethods
The first level of sub2division in SWAT is sub2ba2
sin which contains at least one hydrologic response unit
(HRU ) , a tributary channel and a main channel or
reach. HRU s possess unique land2use, management
and soil attributes.
During the simulation based on SWAT, the whole
Q ingjiang R iver Basin was divided into fifteen sub2ba2
sins, and every sub2basin was composed of many
HRU s. Before simulation, parameter sensitivity analy2
sis were conducted, the sensitive parameters of SWAT
model were identified, and then model parameters re2
lated to stream flow and nutrient loadings were calibra2
ted and validated by observed values.
In order to research the impacts of nutrient load2
ings caused by human activities, nutrient loading were
simulated from year 1980 to 2005. NPS pollution dis2
tribution and critical risk areas of soil erosion and nu2
trient loss were identified. The spatial2temporal distri2
bution features of NPS pollution in the Q ingjiang R iver
basin were revealed.
4 PARAM ETERS SENS IT IV ITY A2
NALY S IS
In this study, parameter sensitivity was analyzed
85
第 1期 WANG Zhao2hui, et al Research on Simulation of Non2point Source Pollution on W ater Quality of Q ingjiang R iver Basin Based on SWATModel and GIS
by LH2OAT method p roposed byMorris in SWAT mod2
el. The advantages of method one2factor2at2a2time
(OAT) and the method Latin2Hypercube are extracted
and adop ted by the method LH2OAT[ 6 ] . The merits of
LH2OAT method are that it lessens the range of param2
eters, decreases the number of parameters being adjus2
ted and imp roves the efficiency of simulation.
The purpose of conducting sensitivity analysis for
all parameters is to understand the influences of the ten
top parameters on stream flow, sediment, total phos2
phorous and total nitrogen in order ( seen in Tbale 1).
Because model calibration and validation depend on the
actual physical p rocess, these listed parameters were
acted as reference from observed values.
Table 1 The results of param eter sen sitiv ity ana lyses
Stream
flow
Sediment
loading
Total
phosphorus
Total
nitrogen
CN2 CN2 SOL2DRCP SOL2ORGN
ESCO SPCON SOL2Z SOL2Z
SOL2Z SOL2Z CN2 CN2
SOL2AWC B IOM IX AWC SOL2AWC
CANMX SLOPE ESCO ESCO
B IOM IX SOL2AWC SOL2LABP SOL2LABP
SOL2K USLE2P SLOPE SLOPE
GWQMN ESCO ALPHA2BF ALPHA2BF
RCHRG2DP SURLAG USLE2P USLE2P
ALPHA2BF CANMX CANMX CANMX
5 MOD EL CAL IBRAT IO N AND
VAL IDAT IO N
The aim of model calibration and validation is to
find out the parameter value which coincides with sim2
ulated datum and observed value. It is indispensable
for simulation, which is used to assess model p redicted
results[ 7 ] . Observed data from three stations including
Shuibuya Station, Geheyan Station and Changyang Sta2
tion were used in this research. Only data of Changy2
ang Station were used to calibrate and validate sedi2
ment and nutrition because of observed data in the ab2
sence of former two stations.
In this study Nash2Sutcliffe coefficient ( ENS) and
correlation coefficient were adop ted in model calibra2
tion and validation. In comparson simulated values
with observed values, if a reasonable stream flow was
achieved, same parameters were used for calibration of
the sediment and nutrient yield. The Nash2Sutcliffe co2
efficient and correlation coefficient reached 0. 81 and
0. 991 respectively in flow simulation. The Nash2Sut2
cliffe coefficient and correlation coefficient reached
0. 752 and 0. 964 respectively in sediment simulation.
The Nash2Sutcliffe coefficient and correlation coeffi2
cient of nutrient calibration both reached satisfied value
using the same method .
6 ANALY S IS AND D ISCUSS IO N
6. 1 Iden tif ica tion of the cr itica l r isk area s of so il
erosion
Soil erosion quantity was calculated in each sub2
basin. According to the simulated results of the soil ero2
sion based on SWAT model during 198022005, soil ero2
sion intensities were classified, and spatial distribution
characters were revealed. The simulation results showed
that the area of low2grade soil erosion accounted for
68. 87% , that of m iddle2grade soil erosion 26. 74% and
that of drastic soil erosion 4. 39%. (Fig. 2).
F ig. 2 The d istr ibution of so il erosion cr itica l r isk area
The result shows that there are intimate relation2
ship s between soil erosion, p recip itation, slope length
and slope gradient[ 8 ] , that the relative coefficient be2
tween soil erosion and p recip itation is 0. 59, and that
the relative coefficient between soil erosion and slope
length is 0. 23, while the coefficient between soil ero2
sion and slope gradient is 0. 39.
6. 2 S im ula tion on stream flow
The simulated average annual stream flow was
1. 57 ×1010 m3 ·yr- 1 from 1980 to 2005 ( Fig. 3 ).
The maximum runoff occurred in 1983 , following with
1998, viz. 2. 46 ×1010 m3 and 2. 39 ×1010 m3 respec2
tively. The m inimum stream flow was in 1997, viz.
1. 03 ×1010 m3.
95
长江科学院院报 2010年
F ig. 3 S im ula ted stream flow dur ing 1980 - 2005
The annual stream flow mainly occured from May
to Sep tember, which was 61. 3% of the whole year.
Simulation results indicate that the maximum flow oc2
curred in August ( Fig. 4).
F ig. 4 S im ula ted m on thly output changes of
stream flow dur ing 1980 - 2005
6. 3 S im ula tion on changes of TP and TN
Average annual nutrient outputs of TN and TP
were 1. 559 ×106 kg and 1. 56 ×105 kg respectively.
Seasonal variations of the nutrient output were differ2
ent. The result of 1980 is not included in Table 2 be2
cause its values are abnormal.
Table 2 S im ula ted changes of TN and TP
from 1981 - 2005 ( 105 kg)
Year N P Year N P Year N P
1981 19. 98 2. 44 1990 12. 65 1. 52 1999 12. 11 1. 41
1982 21. 49 2. 63 1991 16. 74 2. 01 2000 15. 6 1. 77
1983 22. 38 2. 72 1992 10. 83 1. 26 2001 9. 92 1. 16
1984 18. 02 2. 20 1993 17. 55 1. 98 2002 14. 23 1. 65
1985 15. 76 1. 90 1994 13. 32 1. 55 2003 13. 07 1. 48
1986 15. 51 1. 86 1995 14. 00 1. 59 2004 14. 06 1. 59
1987 17. 51 2. 08 1996 16. 32 1. 89 2005 13. 70 1. 56
1988 15. 43 1. 85 1997 12. 37 1. 43 Average 15. 59 1. 83
1989 18. 69 2. 24 1998 18. 55 2. 10
The concentration and output of TN and TP
changed according to season transformation. The maxi2
mum output of TN and TP occured in late sp ring and
summer ( from Ap ril to Sep tember) ( Fig. 5 ) , the
m inimum output occured in winter. There are two rea2
sons for this phenomenon. The most important one is
agricultural tillage operation, such as cultivation and
fertilizing in sp ring[ 9 ] . The other one is temperature.
Grass and leaves are rotted rap idly in higher tempera2
ture, which makes water carry p lenty of pollutants and
enhances nutrient loadings[ 10 ] .
F ig. 5 S im ula ted m on thly output changes of TN
and TP dur ing 1980 - 2005
7 SUMM ARY AND CO NCL US IO N
A lthough the SWAT model is commonly used in
simulation on stream flow, sediment and nutrient load2
ings, it is influenced not only by natural conditions
such as p recip itation, soil type and topography but also
human activities. Land use, tillage operation and p re2
cip itation are the main reasons influencing surface run2
off and nutrient concentration and output. The outputs
of nutrient loadings are in accordance with tillage oper2
ation and fertilizer app lication seasons.
In Q ingjiang R iver basin the area of low2grade soil
erosion accounted for 68. 87% , that of m iddle2grade
soil erosion 26. 74% , and that of drastic soil erosion
4. 39% respectively. The average annual stream flow
was 1. 57 ×1010 m3 ·yr- 1 from 1980 to 2005. Stream
flow mainly distributed from May to Sep tember, which
owns 61. 3% of the whole year. The average annual
output of TN and TP were 1. 559 ×105 kg and 1. 56 ×
106 kg respectively. The surface runoff and nutrient
yield result indicates that the average annual runoff and
06
第 1期 WANG Zhao2hui, et al Research on Simulation of Non2point Source Pollution on W ater Quality of Q ingjiang R iver Basin Based on SWATModel and GIS
output of TN and TP p rovides better understanding in
stream flow and nutrient loadings corresponding to vari2
ations of land use conditions, agricultural tillage opera2
tion and natural rainfall etc.
REFERENCES:
[ 1 ] LUO Yi, HE Chan2sheng, SOPHOCLEOUS Marios, et
a l. A ssessment of crop growth and soil water modules in
SWAT2000 using extensive field experiment data in an
irrigation district of the Yellow R iver basin [ J ]. Journal
of Hydrology, 2008, 352: 139 - 156.
[ 2 ] GEZA Mengistu and McCray J E. Effects of soil data res2
olution on SWAT model stream flow and water quality
p redictions[ J ]. Journal of Environmental Management,
2008, 88: 393 - 406.
[ 3 ] XU Z X, ZHAO F F, L I J Y. Response of stream flow to
climate change in the headwater catchment of the Yellow
R iver Basin [ J ]. Quaternary International , 2008: 1 -
14.
[ 4 ] HOLVOET K, VAN Griensven A, SEUNTJENS P, et a l.
Sensitivity analysis for hydrology and pesticide supp ly to2
wards the river in SWAT[ J ]. Physics and Chem istry of
the Earth , 2005, 30: 518 - 526.
[ 5 ] CHAPLOT V. Impact of DEM mesh size and soil map
scale on SWAT runoff, sediment, and NO32N loads p re2
dictions[ J ]. Journal of Hydrology, 2005, 312: 207 -
222.
[ 6 ] SHEN Zhen2yao, HONG Q ian, YU Hong, et a l. Parame2
ter uncertainty analysis of the Non2point source pollution
in the daning river watershed of the Three Gorges Reser2
voir Region [ J ]. Science of the Total Environment,
2008, 405: 195 - 205.
[ 7 ] N INGA Shu2Kuang, CHANG N i2B in, JENGC Kai2Yu, et
a l. Soil erosion and Non2point source pollution impacts
assessment with the aid of multi2temporal remote sensing
images [ J ]. Journal of Environmental Management ,
2006, 79: 88 - 101.
[ 8 ] WANG Zhao2hui, CHEN Bei2qing, CHENG Xue2jun.
Impact of land use change on Non2point Source Pollution
Load in Changyang County [ J ]. JOURNAL of Yangtze
R iver Scientific Research Institute, 2010, 27 ( 1) : 17 -
21.
[ 9 ] Trancoso Ana Rosa, B raunschweig Frank, Leitao Pedro
Chambel, et al. An advanced modelling tool for simula2
ting comp lex river system s[ J ]. Science of the Total En2
vironment, 2009, 407: 3004 - 3016.
[ 10 ] TANGA Z, ENGELA B A, P IJANOW SKIB B C, et a l.
Forecasting land use change and its environmental impact
at a watershed scale [ J ]. Journal of Environmental Man2
agement, 2005, 76: 35 - 45.
( Edited by L IU Yun2fei, YI Xin2hua)
基于 GIS和 SWAT模型的清江流域
面源污染模拟研究
汪朝辉 1, 2 ,赵登忠 1 ,曹 波 1 ,梁东业 1
(1. 长江科学院 空间信息技术应用研究所 ,武汉 430010; 2.武汉大学 资源与环境科学学院 ,武汉 430079)
摘要 :水体面源污染评价研究是一项复杂的工作 ,涉及面广和要求有较长时间序列的数据。面源污染模型很大程
度上依赖于对于流域的特征描述的相关参数 ,因此提高输入模型参数的精度 ,有利于提升流域面源污染的径流、泥
沙和营养物质的产出模拟效果。GIS和面源污染模型的有机结合是当前面源污染模拟研究最有效的方法。本研究
建立了基于 GIS的基础数据库 ,其中包括 DEM、土壤类型、土地利用、气象数据以及农业耕作管理数据等。面源污
染模拟的产生和形成具有很大的不确定性 ,这更加增加了监测和控制面源污染的难度。探索影响面源污染的主要
因素 ,研究其不确定性对于提出和制定污染控制措施至关重要。本研究进行了清江流域面源污染的参数敏感性分
析 ,根据观测数据对 SWAT模型进行验证和率定 ,并利用 SWAT模型进行模拟 ,揭示了清江流域面源污染的时空分
布特征 ,确定了清江流域水土流失风险区。结果表明清江流域的面源污染主要发生在丰水期 ,不同的土地利用方
式、农业耕作
以及降水等是影响径流量和营养物质产生的主要因素。
关 键 词 : SWAT模型和 GIS;敏感性分析 ;径流模拟 ;泥沙模拟 ;营养物载荷 ;清江流域
16