Xio-kng Xin*,Wei Yin,Ke-feng Li

aState Key Laboratory of Hydraulics and Mountain River Engineering,Sichuan University,Chengdu 610065,China

bChangjiang Water Resources Protection Institute,Changjiang Water Resources Commission,Wuhan 430051,China

Estimation of non-point source pollution loadsw ith flux method in Danjiangkou Reservoir area,China

Xiao-kang Xina,b,*,Wei Yinb,Ke-feng Lia

aState Key Laboratory of Hydraulics and Mountain River Engineering,Sichuan University,Chengdu 610065,China

bChangjiang Water Resources Protection Institute,Changjiang Water Resources Commission,Wuhan 430051,China

Abstract

The estimation of non-point source pollution loads into the Danjiangkou Reservoir is highly significant to environmental protection in the watershed.In order to overcome the drawbacks of traditional watershed numericalmodels,a base flow separation method was established coupledw ith a digital fi lteringmethod and a fluxmethod.The digital fi lteringmethod hasbeen used to separate the base flowsof the Hanjiang, Tianhe,Duhe,Danjiang,Laoguan,and Qihe rivers.Based on daily discharge,base flow,and pollutant concentration data,the fluxmethod was used to calculate the pointsource pollution load and non-pointsource pollution load.The results show that:(1)In the year2013,the total inflow of the six riversmentioned above accounted for 95.9%of the total inflow to the Danjiangkou Reservoir.The total pollution loads of chemical oxygen demand(CODMn)and total phosphorus(TP)from the six riverswere 58.20×103tand 1.863×103t,respectively,and the non-point source pollution loadswere 39.82×103tand 1.544×103t,respectively,indicating that the non-pointsource pollution isamajor factor(w ith a contribution rate of 68.4%for CODMnand 82.9%for TP).(2)The Hanjiang River is themost significant contributor of pollution loads to the Danjiangkou Reservoir,and itsCODMnand TPcontribution rates reached 79.3%and 83.2%,respectively.The Duhe River took the second place. (3)Non-point source pollution mainly occurred in the wet season in 2013,accounting for 80.8%and 90.9%of the total pollution loads of CODMnand TP,respectively.It is concluded that the emphasis of pollution control should be placed on non-point source pollution.

?2017 Hohai University.Production and hosting 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/).

Keywords:Danjiangkou Reservoir；Non-point source pollution；Base flow separation；Digital fi ltering method(DFM)；Fluxmethod

1.Introduction

The Danjiangkou Reservoir plays an important role in optimal allocation ofwater resources in China because of the M iddle Route Project of the South-to-North Water Diversion Project(Ma et al.,2014),and the protection of its water resources has received w idespread attention from all over the world(Li et al.,2009).Over the past 20 years,local governments around the Danjiangkou Reservoir have graduallyclosed down or relocated the point sources of pollution distributed around the reservoir,in order to ensure that clean water is transferred(Xin et al.,2015).Non-point source pollution has become a dominant factor influencing thewater quality of the Danjiangkou Reservoir(Wang et al.,2011；Li et al.,2014).Plenty of research has been conducted on the problem of non-point source pollution in the Danjiangkou Reservoir(Zhao and Hu,2007；Jiang etal.,2010；Huang etal., 2012).In the Danjiangkou Watershed,the non-point sources mainly include losses of pesticides and fertilizers,livestock wastes,and rural domestic wastewater(Zhao and Hu,2007). The process of non-point source pollution is accompanied by rainfall(water and soil losses),and the contribution rate of non-point source pollution is about 75%(Yin et al.,2011).

The estimation of non-pointsource pollution loads into the Danjiangkou Reservoir is highly significant to the determ ination of the pollution control scheme and environmental management planning.Researchers have adopted the export coeffi cientmethod(Fang etal.,2011),annualized agricultural non-point source(AnnAGNPS)model(Polyakov et al.,2007；Tong,2008),and soil and water assessment tool(SWAT) model(Qiao et al.,2013)to study the non-point source pollution load in the Danjiangkou Reservoir area.

However,most of the models mentioned above require complete digital elevation model(DEM)data,land use data, soilattribute data,meteorological data,and hydrological data. The model establishment process is very complicated and difficult(Hong et al.,2012；Chowdary and Adiga,2001),so the current research scope is extremely limited and does not cover the whole watershed.Based on the features of clear physical conception,easy data acquisition,and dependable accuracy,the base flow separation of non-point source pollution load estimation models in hydrology has received w idespread attention in recent years(Liet al.,2010；Zhao,2010).

Base flow separation is a hydrological method that can separate the riverbase flow,which is relatively stable,from the surface flow,which is the fluctuant part of the total flow.The point source pollution load and the natural background load are stable and show little change throughouta year.They can be represented by the base flow flux.The non-point source pollution load along w ith the rainfall and surface flow rushes into rivers,w ith a characteristic of volatility,and can be represented by the surface runoff flux.

According to thedaily discharge dataof sixmain tributaries of the Danjiangkou Reservoir,the Hanjiang,Tianhe,Duhe, Danjiang,Laoguan,and Qihe rivers,in the hydrological year 2013,this study sought,fi rst,to separate the base flow from the total flow using the digital fi lteringmethod(DFM),then to calculate the non-point source pollution loads using the flux method based onmonthly water qualitymonitoring data from 2013,and,finally,to investigate the spatial and temporal characteristics of pollution loads.

2.Study area and basic data

2.1.Study area

The Danjiangkou Reservoir is located in the upper reaches of the Hanjiang River and at the junction of northwest Hubei Province,southwest Henan Province,and southeast Shaanxi Province.Thewatershed is in the transition zone of the Daba Mountains,Qinling Mountains,and Jianghan Plain.The Danjiangkou Reservoir is a large reservoir w ith a variety of functions,such as water supply,flood control,power generation,and navigation.

The fi rst construction stage of the Danjiangkou Reservoir began in September1958andwascompletedattheend of1973. Thenormalwater levelwas157m and thedeadwater levelwas 139m.The totalwater capacity was 20.97×109m3,the dead storagewas7.23×109m3,and the reservoir surface areawas 745 km2at the normalwater level.In order to implement the M iddle Route Project of the South-to-North Water Diversion Project,adam-heightening projectwascarried outin 2005,and the normalwater level increased to 170m,w ith a totalwater capacity of29.05×109m3.A fter thewetseason in 2014,there were about 9.5×109m3of water(the annual average value) transferred to Beijing City,Tianjin City,and Henan Province.

The area of the watershed upstream of the Danjiangkou Dam is over 95200 km2,w ith an average annual inflow of 1230.4m3/s and an annual runoff of 38.8×109m3.There are seven counties in the reservoir watershed:Xixia,Xichuan, Yunxi,Yunxian,Danjiangkou,Maojian,and Zhangwan (Fig.1).There are approximately 200 tributaries around the reservoir.The sum of the drainage areas of the sixteen largest tributaries account for 95%of the total drainage area.Information about these sixteen tributaries isgiven in Table 1.Only six tributaries have been equipped w ith hydrological monitoring stations:the Hanjiang River(Baihe Station),the Tianhe River(Jiajiafang Station),the Duhe River(Zhushan Station), the Danjiang River(Jingziguan Station),the Qihe River (Xiping Station),and the Laoguan River(Xixia Station) (Fig.1).The sum of the drainage areas of these six tributaries is86549 km2,accounting for90.9%of the totalwatershed area of the upper Danjiangkou Dam,and the sum of the annual discharge is 1179.5 m3/s,accounting for 95.9%of the total discharge.Therefore,the analysis of the non-point source pollution loads of the six main tributaries is representative.

2.2.Basic data

2.2.1.Hydrological data

According to the daily hydrological datameasured in 2013 in the six main tributaries,the monthly average discharges were calculated and are listed in Table 2.The discharge was mainly concentrated in the wet season(from June to September),accounting for 73.4%of the annual discharge. The discharge in the dry season(from December to March) accounted for only 10.5%,and the discharge in the normal season(October,November,and April)accounted for 16.1%.

From thepointofview of the tributaries(Table2),theannual discharge of the Hanjiang River is the largest(533.11 m3/s, accounting for 78.3%of the total annual discharge of the six tributaries),and followed by the Duhe River(109.97m3/s,accounting for 16.2%),while the other four small tributaries contribute a small part of the total discharge(37.61m3/s,accounting for5.5%).

The representative hydrological data from 2013 were analyzed.Theannualdischargesof thesix tributariesin2013 are significantly less than the average annual discharge.According to the Pearson type IIIhydrological frequency analysismethod, the hydrological frequency of the Hanjiang River in 2013 was 75.3%,and the hydrological frequencies of the Tianhe River, Duhe river,DanjiangRiver,QiheRiver,and Laoguan Riverwere 80.3%,83.3%,90.5%,90.3%,and 96.5%,respectively.Therefore,itwasadry year in2013 for the fi rstthree tributariesand an extremedry year for the last three tributaries.Theanomaly ratio is the difference between themeasured value and the annual average,so themonthly anomaly ratio can reflect the degree ofwetness and dryness.The flow anomaly ratios of the six tributaries fordifferentmonthsin 2013areshown in Table3.Overall, the inflow of theDanjiangkouReservoirwas lessin2013 than in anormalyear.

Fig.1.Locations of hydrological and water qualitymonitoring stations in Danjiangkou Reservoir area.

2.2.2.Water quality data

Nineteen water quality indices were monitored at all the monitoring stations of the sixteen tributaries,including dissolved oxygen(DO),chemical oxygen demand(CODMn), five-day biochemical oxygen demand(BOD5),total phosphorus(TP),ammonia nitrogen(NH3-N),sulfide(S2-),cyanide(CN-),fluorine ion(F-),hexavalent chrom ium(Cr6+), arsenic(As),mercury(Hg),selenium(Se),copper(Cu),lead (Pb),cadm ium(Cd),zinc(Zn),volatile phenols,and oils.The monitoring frequency was once per month.The month-tomonth variation trend graphs for typical indices,the CODMnand TP concentrations,in 2013 are shown in Fig.2.

Non-pointsource pollutantsare often discharged into rivers with rainfall runoff.Therefore,in general,pollutant concentration ishigherduring thewetseason than during thedry seasonor normal season,indicating that the riverwater quality is significantly affected by the non-point source pollution.The Tianhe River flows through YunxiCounty and the Laoguan River flows through Xixia and Xichuan counties.Because these three counties are more industrialized and the proportions of point source pollution and non-point source pollution are equal,the correlations between concentrations and discharge are less apparent.The other four tributaries are rural rivers,and thecontamination ismainly caused by agriculturalnon-pointsource pollution.As we can see from the relationships between the CODMnand TPconcentrationsand dischargeQ(Fig.3),CODMnand TPconcentrationshavestrong positive correlationsw ith the discharge,w ith thecoefficientofdeterminationR2ranging from 0.39 to0.78and0.51 to0.80,respectively.Therefore,wedrew an inference that the CODMnand TP concentrations in the Danjiangkou Reservoir are significantly influenced by non-point source pollution.

Table 1Hydrological parameters for sixteen large tributariesw ithin study area.

Table 2Monthly average discharges of six main tributaries in 2013.

Table 3Flow anomaly ratio of six main tributaries for differentmonths in 2013.

3.Research methods

3.1.Digital fi ltering method

Base fl ow separation methods separate the base fl ow from the total flow.Due to different interpretations of base flow,there are different theories and methods of separation. The most commonly used methods are the linear segmentation method,slash segmentation method,hydrological modeling method,water balance method of Kalinin,environmental isotope method,and digital fi ltering method (Chen et al.,2006；Ni et al.,2005).The digital fi ltering method is a popular method used to separate highfrequency signals from low-frequency signals in the field of digital signal analysis and processing.In the hydrological course of the rainfall-runoff process,surface flow responds to rainfall and changes rapidly,representing the highfrequency signal,while the base flow changes slow ly w ith rainfall and represents the low-frequency signal.Therefore, separating the base flow from the total flow is just like separating the low-frequency signal in signal processing (Eckhardt,2005).

Fig.2.Monthly change of CODMnand TP concentrations of six main tributaries in 2013.

Fig.3.Correlation between typical pollution indices and flow rate.

In the field of hydrology,the digital fi ltering method was fi rst used by Lyne and Hollick(1979),and the core of the method is represented by the fi lter equation:

whereqjis the fi ltered flow on thejth day(qjandqj-1are surface flow)；Qjis the total flow on thejth day；andαandβ are the fi lter parameters.A fter the surface flow isobtained,the base flow is equal to the total flow minus the surface flow. There have been no good ways to validate fi lter parameters until now.Nathan and M cmahon(1990)compared the results of the digital fi ltering method and the manual segmentation method,and proposed that the reference values ofαandβ were 2.0 and 0.925,respectively.

3.2.Fluxmethod

After completing the base flow separation,the total pollution load,point source pollution load,and non-point source pollution load can be estimated w ith the flux method.On the basis of division of river base flow and surface flow,we consider that the pollution load transported by the base flow to be thesum of thenaturalbackground load and thepointsource pollution load(hereafter referred to as the point source pollution load).The pollution load transported by the surface flow is regarded as the non-point source pollution load.The above-mentioned concept can be rew ritten as Eq.(2):

whereWtis the total pollution loads transported in the river channel；Wpis the pointsource pollution load；Wnpis the nonpointsource pollution load；tis time；Cp（t）is the pointsource pollutant concentration at timet,represented by the monitoring data of water quality in the dry season(in this study,Cp（t）was approximately replaced by themean concentration from December to March)；Qp（t）is the base flow at timet；Cnp（t）is thenon-pointsource pollutantconcentration at timet, which can be represented by the mean concentration from May to October；andQnp（t）is the surface flow at timet.Due to the lack of continuous water quality monitoring data,the integral equation Eq.(2)should be transformed into the discrete equation Eq.(3):

whereCpiis the pointsource pollutantconcentration in theith month,Qpiis the average base flow in theithmonth,Cnpiis the non-pointsource pollutant concentration in theithmonth,Qnpiis the surface flow in theithmonth,nis the number ofmonths in a year,andΔtis the period of theithmonth.

At fi rst,we should calculate the totalpollution load w ith the monthly flow dataand pollutantconcentration dataw ith Eq.(4):

whereCiis the pollutant concentration monitored in theith month,since thewater quality datamonitoring frequency was once permonth in this study；andQiis the average discharge in theith month.Thus,the non-point source pollution loads can beexpressed asWnp=Wt-Wp,which can be rew ritten as

Eq.(5)is the non-point source pollution load computation formula based on the base flow separation method,in which the non-point source pollution load equals the total pollution load m inus the point source pollution load,while the point source pollution load is estimated by the product of the base flow and pollutant concentration during the dry season.

4.Results and discussion

4.1.River base flow separation

W ith the help of daily flow data from the six main tributaries around the Danjiangkou Reservoir,the base flow andsurface flow,which were calculated w ith the digital fi ltering method,are listed in Table 4.In 2013,the average base flows of the Hanjiang River,Tianhe River,Duhe River,Danjiang River,Qihe River,and Laoguan River were 229.56,4.90, 37.79,7.71,1.54,and 2.04 m3/s,respectively.The ratios of total runoff taken up by the base flow varied from 30.1%to 52.1%.The profi lesof the total flow and base flow can be seen in Fig.4.

4.2.Pollution load calculation results and pollution structure interpretation

The established method was used to calculate the point source pollution loadsand non-pointsource pollution loads of the six main tributaries,and the results are shown in Table 5. The total pollution loadsof CODMnand TP transported by the sixmain tributarieswere 58.20×103tand 1.863×103t,ofwhich about39.82×103tof CODMnand 1.544×103tof TP came from thenon-pointsource pollution,reaching 68.4%and 82.9%of the total,respectively.The calculated results agreed w ith the resultsof the exportcoefficientmethod,which shows that thenon-pointsource pollution load of TPwas1.86×103t (Fang etal.,2011).Thisnearly agreesw ith the results from the SWATmodel,which shows that the non-pointsource pollution load of TPwas 1.58×103t(Qiao et al.,2013).

Table 4Base fl ows ofmain tributaries around Danjiangkou Reservoir.

Fig.4.Profi les of total flow and base flow for six main tributaries in 2013.

Table 5Pollution loads ofmain tributaries in Danjiangkou Reservoir area.

The total CODMnloads of the six tributaries were 46.13×103,1.02×103,8.18×103,1.85×103,0.35×103, and 0.67×103t,respectively,and the proportions of the nonpoint source pollution load varied from 65.7%to 80.0%.The total TP loads of the six tributaries were 1.550×103, 0.042×103,0.180×103,0.073×103,0.006×103,and 0.016×103t,respectively,and the proportions of the nonpoint source pollution load varied from 50.0%to 83.8%.

These results are consistentw ith the calculation results of non-point source pollution loads in other basins in China(Li and Huang,2005；Hao et al.,2004).Therefore,we can draw a conclusion that non-point source pollution is the dominant factor in the water quality in the Danjiangkou Reservoir.In terms of different pollution indicators,CODMnhas a strong relationship w ith the decentralized rural domestic sewage and TP is closely associated with the application of pesticidesand chem ical fertilizers(Qiao et al.,2013).

4.3.Distribution characteristics of pollution loads

4.3.1.Spatial distribution characteristics

Table 6 shows the contribution ratesof the six tributaries to totalpollution loadsand non-pointsource pollution loads.Thecontribution rates of tributaries to the total CODMnpollution loads are ranked in the follow ing descending order:the Hanjiang River,Duhe River,Danjiang River,Laoguan River, Tianhe River,and Qihe River；and those of CODMnnon-point source pollution loads are ranked in the follow ing descending order:the Hanjiang River,Duhe River,Danjiang River,Tianhe River,Laoguan River,and Qihe River.As we can see,the contribution rate of the Hanjiang River to the total CODMnpollution loadswas79.3%and the contribution rate to thenonpoint source CODMnpollution loadswas 78%,demonstrating that the contribution of the Hanjiang River to the pollution loads of the Danjiangkou Reservoir is the most significant, followed by that of the Duhe River.The contribution rate of the Duhe River to the totalCODMnpollution loadswas14.1%, and the contribution rate to the non-point source CODMnpollution loadswas 15.0%.

Table 6Contribution rates of each tributary to total and non-point source pollution loads.

Sim ilarly,the contribution rates of tributaries to total TP pollution loadsare ranked in the follow ing descending order: the Hanjiang River,Duhe River,Danjiang River,Tianhe River,Laoguan River,and Qihe River,and the order of the contribution rates of tributaries to non-point source pollution loads is the same.The contribution rates of the Hanjiang River,followed by the Duhe River,occupy a dom inant position aswell,w ith 83.2%and 84.1%for the total pollution loads and non-point source pollution loads,respectively. This phenomenon is strongly related to the infl ows of the tributaries:the proportion of discharge from the Hanjiang River was 78.4%in 2013,and that of the Duhe River was 16.1%,while there were no significant differences among the CODMnand TP concentrations of the six main tributaries.

The Danjiangkou Watershed can be divided into two parts: the Han Part,which is located in Shaanxi Province and includes the Hanjiang,Tianhe,and Duhe rivers；and the Dan Part,which ismainly located in Hubei Province and Henan Province and contains the other three tributaries.Further analysis shows that non-point source pollution loads from the Dan Partare 2.1×103t for CODMnand 0.074×103t for TP, while those from the Han Part are 37.71×103t for CODMnand 1.469×103t for TP.This spatial distribution characteristic shows that non-point source pollution loads of the Danjiangkou Reservoir come mainly from Shaanxi Province, followed by Hubei Province,and the contribution rate of Henan Province is relatively small.

Table 7Temporal distribution characteristics of pollution loads.

The em ission intensity of CODMnin the Dan Part is 0.52 t/(km2$year)and thatof TP is0.02 t/(km2$year),while the em ission intensity ofCODMnin theHan Partis0.16 t/(km2$year) and that of TP is 0.006 t/(km2$year).Therefore,the emission intensity of non-point sources from the Han Part is larger than that from theDan Part.Thischaracteristic isconsistentw ith the areaof farm landand the ruralpopulation distribution in thewater source area of the M iddle Route Projectof the South-to-North Water Diversion Project:according to the statistics yearbooks of ShaanxiProvince,HubeiProvince,and Henan Province,the ruralpopulationsare9807.5×103,2014×103,and59.15×103in Shaanxi Province,Hubei Province,and Henan Province, respectively；and the farm landarea is536.8×103ha forShaanxi Province,169.9×103ha forHubeiProvince,and 79.4×103ha forHenan Province.4.3.2.Temporal distribution characteristics

Table 7 shows the distribution of pollution loads in different water periods.Aswe can see,both the totalpollution loadsand non-point source pollution loads of CODMnand TPare ranked in the following descending order:thewet season,the normal season,and the dry season.During the wet season,the total pollution load of CODMnfrom the six main tributaries is 45.568×103t,accounting for78.3%；and thenon-pointsource pollution load is 32.167×103t,accounting for 80.8%.The total pollution load of TP from the six main tributaries is 1.633×103t,accounting for 87.7%；and the non-pointsource pollution load is 1.404×103t,accounting for 90.9%.These results further prove that the pollution in the Danjiangkou Reservoirmainly comes from non-point sources,and the nonpoint source pollution is accompanied by rainfall.

5.Conclusions

(1)Non-pointsource pollution is themost important factor in the water quality of the Danjiangkou Reservoir.A simple and feasible base flow separation method coupled w ith the digital fi ltering method and flux method was established to separate the pointand non-pointsource pollution loads.Itwas verified to be accurate and effective.

(2)There isa strong correlation between the CODMnand TP concentrationsand thedischargesofmain tributaries.Therefore, the established method can be used to calculate the pollution loads of these two indicators,and the calculation results show thatabout58.20×103tofCODMnand 1.863×103tof TPwere transported into theDanjiangkou Reservoir through thesixmain tributaries in 2013.However,therewere 39.82×103tCODMnand 1.544×103t TP com ing from the non-pointsource pollution,and the contribution proportionswereashigh as68.4%and 82.9%,respectively.These resultsagreew ith the results from the exportcoefficientmethod and the SWATmodel.

(3)The spatial distribution characteristicsof pollution loads show thatthenon-pointsourcepollution loads from theDan Part are2.1×103t forCODMnand 0.074×103t for TP,while those from the Han Partare 37.71×103tand 1.469×103t,respectively.The non-pointsource pollutantemission intensity of the Han Part is larger than thatof the Dan Part.The Hanjiang and Duhe rivershave the largestnon-pointsourcepollution loads,so more pollution controlemphasisshould be paid to the Hanjiang Basin and Duhe Basin.

Chen,L.Q.,Liu,C.M.,Li,F.D.,2006.Review of the research on basic fl ow. Prog.Geogr.Sci.25(1),1-15.http://dx.doi.org/10.11820/ dlkxjz.2006.01.001(in Chinese).

Chowdary,V.M.,Adiga,S.,2001.Assessmentof non-pointsource pollution on watershed basis using remote sensing,GIS,and AGNPSmodel.In:Proceeding of the 22nd Asian Conference on Remote Sensing.National University of Singapore,Singapore.

Eckhardt,K.,2005.How to construct recursive digital fi lters for baseflow separation.Hydrol.Process.19(2),507-515.http://dx.doi.org/10.1002/hyp.5675.

Fang,N.F.,Shi,Z.H.,Li,L.,2011.Application of export coefficientmodel in simulating pollution load of non-point source in Danjiangkou Reservoir Area.J.Hydroecol.32(4),7-11(in Chinese).

Hao,F.H.,Zhang,X.S.,Yang,Z.F.,2004.A distributed non-point source pollution model:Calibration and validation in the Yellow River Basin.J. Environ.Sci.16(4),646-650.http://dx.doi.org/10.3321/j.issn:1001-0742.2004.04.025.

Hong,Q.,Sun,Z.,Chen,L.,Liu,R.,Shen,Z.,2012.Small-scalewatershed extended method for non-point source pollution estimation in part of the Three Gorges Reservoir Region.Int.J.Environ.Sci.Technol.9(4), 595-604.http://dx.doi.org/10.1007/s13762-012-0094-y.

Huang,W.M.,Bi,Y.H.,Hu,Z.Y.,Wang,H.,Feng,R.J.,Hu,J.D.,2012.Study on characteristics of agricultural non-point source pollution in Danjiangkou Reservoir.Environ.Sci.Manag.37(1),33-38(in Chinese).

Jiang,S.Y.,Han,P.,Jia,Z.B.,Mao,X.L.,Si,J.,Ye,F.F.,2010.Evaluationw ith PSRmodel and GIS analysis of agricultural non-point source pollution in Danjiangkou reservoir of the M id-route of the South-to-North water transfer project.J.Agro Environ.Sci.29(11),2153-2162(in Chinese).

Li,C.M.,Huang,Z.L.,2005.Study on the pollution load of the Three Gorges Reservoir:I.Research on the currentsituation of the pollution load before the storage.Resour.Environ.Yangtze Basin 14(5),610-622(in Chinese).

Li,H.F.,Bai,J.F.,Wu,B.B.,2014.Estimate and analysis of rural non-point source pollution in the Danjiangkou Reservoir.J.Nanyang Norm.Univ. 13(9),43-48(in Chinese).

Li,K.,Lin,K.R.,Jiang,T.,Chen,X.H.,2010.Application of a digital fi lter in the differentiation of pollutant loads from point and non-point sources.Res.Environ.Sci.23(3),298-303.http://dx.doi.org/10.13198/ j.res.2010.03.52.lik.008(in Chinese).

Li,S.Y.,Cheng,X.L.,Xu,Z.F.,Han,H.Y.,Zhang,Q.F.,2009.Spatial and temporalpatternsof thewaterquality in theDanjiangkou Reservoir,China. Hydrol.Sci.J.54(1),124-134.http://dx.doi.org/10.1623/hysj.54.1.124.

Lyne,V.,Hollick,M.,1979.Stochastic time-variable rainfall-runoff modelling.In:Proceedings of the Hydrology and Water Resources Symposium. Institution of Engineers National Conference Publication,Canberra, pp.89-93.

Ma,F.B.,Li,C.H.,Wang,X.,Yang,Z.F.,Sun,C.C.,Liang,P.Y.,2014.A Bayesian method for comprehensive water quality evaluation of the Danjiangkou Reservoir water source area,for the m iddle route of the South-to-North Water Diversion Project in China.Front.Earth Sci.8(2), 242-250.http://dx.doi.org/10.1007/s11707-013-0395-6.

Nathan,R.J.,M cmahon,T.A.,1990.Evaluation of automated techniques for base fl ow and recession analyses.Water Resour.Res.26(7),1465-1473. http://dx.doi.org/10.1029/WR026i007p01465.

Ni,Y.Q.,Zhang,W.H.,Guo,S.L.,2005.Analysis and discussion on flow process line segmentationmethod.Hydrology 25(3),10-19(in Chinese).

Polyakov,V.,Fares,A.,Kubo,D.,Jacobi,J.,Sm ith,C.,2007.Evaluation of a non-point source pollution model,AnnAGNPS,in a tropical watershed. Environ.Model.Softw.22(11),1617-1627.http://dx.doi.org/10.1016/ j.envsoft.2006.12.001.

Qiao,W.F.,Niu,H.P.,Zhao,T.Q.,2013.Temporal-spatial distribution of agricultural non-point source pollution in the Danjiangkou Reservoir Watershed based on SWATmodel.Resour.Environ.Yangtze Basin 22(2), 219-225(in Chinese).

Tong,W.H.,2008.Research on Non-point Source Pollution of Heimiaogou Watershed in Danjiangkou Reservoir Area Based on ANNAGNPSModel. M.E.Dissertation.Huazhong Agricultural University,Wuhan.

Wang,X.,Li,Z.B.,Li,P.,Zhang,L.H.,2011.Surface pollution and countermeasures in Danjiang-Hanjiang watershed in Shaanxi Province.Bull. SoilWater Conserv.31(6),186-189(in Chinese).

Xin,X.K.,Li,K.F.,Finlayson,B.,Yin,W.,2015.Evaluation,prediction,and protection of water quality in Danjiangkou Reservoir,China.Water Sci. Eng.8(1),30-39.http://dx.doi.org/10.1016/j.wse.2014.11.001.

Yin,W.,Shi,Z.H.,Lei,A.L.,2011.Research on water environment of Danjiangkou reservoir.Yangtze River 42(13),90-94(in Chinese).

Zhao,G.J.,2010.Development and Application of a Non-point Sources Pollution Model for Hydrological Process and Nutrient Loadings in the Xitiaoxi Catchment in South China.M.E.Dissertation.Kiel University, Kiel.

Zhao,W.Y.,Hu,J.Q.,2007.Present status analysis to non-point source pollution in Danjiangkou Reservoir.South North Water TransfersWater Sci.Technol.5(2),50-52(in Chinese).

Received 1 June 2016；accepted 18 October 2016 Available online 31 May 2017

Thiswork was supported by the National Key Research and Development Program of China(Grants No.2016YFC0402204 and 2016YFC0402207).

*Corresponding author.

E-mail address:xin.xiaokang@163.com(Xiao-kang Xin).

Peer review under responsibility of Hohai University.

http://dx.doi.org/10.1016/j.wse.2017.05.001

1674-2370/?2017 Hohai University.Production and hosting 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/).