Multivariate analysis of surface water quality in the Three Gorges area of China and implications fo

Available online at www.sciencedirect.com、JOURNAL OFENVIRONNENTAL[n],“ScienceDirectSCIENCESISSN 10-07402JESJoumal of Environmental Sciences 2011, 23(9) 1460-1471www.jesc.ac.cnELMultivariate analysis of surface water quality in the Three Gorges area ofChina and implications for water managementJian Zhao, Guo Fu*, Kun Lei, Yanwu LiChinese Research Academy of Emvironmental Sciences, Bejing 100012, China E-mail: zj103823@ 163.comReceived 15 November 2010; revised 01 April 2011; accepted 06 April 2011Abstract .Multivariate statistical techniques, cluster analysis, non-parametric tests, and factor analysis were applied to analyze a water qualitydataset including 13 parameters at 37 sites of the Three Gorges area, China, from 2003- -2008 to investigate spatio-temporal variationsand identify potential pollution sources. Using cluster analysis, the twelve months of the year were classified into three periods of low-flow (LF), normal-Aow (NF), and high-flow (HF); and the 37 monitoring sites were divided into low pollution (LP), moderate pollution(MP), and high pollution (HP). Dissolved oxygen (DO), potassium permanganate index (CODMn), and ammonia-nitrogen (NH4*-N)were identified as significant variables affecting temporal and spatial variations by non-parametric tests. Factor analysis identified thatthe major pollutants in the HP region were organic matters and nutrients during NF, heavy metals during LF, and petroleum during HF.In the MP region, the identified pollutants primarily included organic matter and heavy metals year around, while in the LP region,organic pollution was significant during both NF and HF, and utrient and heavy metal levels were high during both LF and HF. Themain sources of pollution came from domestic wastewater and agricultural activities and runof; however, they contributed diferentlyand LP regions, water pollution was more likely from the combined effects of agriculture, domestic wastewater, and chemical industry.These results provide fundamental information for developing better water pollution control strategies for the Three Gorges area.Key words: water quality; spatial variations; seasonal variations; multivariate statistical techniques; the Three GorgesDOI: 10.1016/S1001-0742(10)60599-2Citation: Zhao J, Fu G, Lei K, Li Y W, 2011. Multivariate analysis of surface water quality in the Three Gorges area of China andimplications for water management. Jourmal of Environmental Sciences, 23(9): 1460-1471Introductionsuch as cluster analysis (CA), principal component anal-ysis (PCA). factor analysis (FA), and discriminant analysisSurface water quality is controlled by complex anthro-(DA) offers superior interpretation of complicated data setspogenic activities and natural factors (Jarvie et al, 1998; .to better understand water quality (Zhang et al.. 2009).Ravichandran, 2003). Seasonal changes in natural process-Clustering analysis is an unsupervised multivariate tech-es, such as temperature, precipitation, and hydrologicalnique used to classify objects into categories or clusterscondition, infuence water quality such that it presentsbased on their nearness or similarity (Vega et al, 1998).different characteristics in different seasons (Vega et al., .Hierarchical clustering is the most common approach in1998; Ouyang et al., 2006). Accurate assessment of thewhich clusters are formed sequentially, by starting with thetype and extension of water pollution is a dificult taskmost similar pair of objects and forming higher clusters(Huang et al,. 2010). Long-term surveys and water qualitystep by step (Singh et al, 2005). Factor analysis providesmonitoring prograrns are an adequate approach to improveinformation regarding the most meaningful parameters,knowledge of river hydrochemistry and pollution, but theywhich describe whole data set rendering for data reductionproduce large data sets that are often dificult to analyzewith minimum loss of original information (Vega et al,and interpret (Shin and Fong, 1999). Applying advanced1998; Wunderlin et a1.. 2001). It is a powerful techniquestatistical methods to these data sets without missingfor patterm recognition and attempts to explain the varianceuseful information is imperative to extract meaningfulof a large set of inter-correlated variables and transform itinformation such as spatial and temporal patterns, signifi-into a smaller set of independent variables (Shrestha andcant parameters and latent pollution sources (Shrestha andKazama, 2007). Discrimirent aplbis id= .statisticalKazama, 2007; Zhang et al.. 2009).classification of sarYH中国煤化工it priorThe application of multivariable statistical techniquesknowledge of membeCNMHGargroupsor clusters. Further, DA helps in grouping samples sharing串Coresponding author. E-mail: Fuguo551225@sina .com1462Joumal of Environmental Sciences 2011, 23(9) 1460 -1471 /Jian Zhao et al.Vol. 23al., 2009). The discharge of TN and TP from livestocktal data standardized through z-scale transformation (theand poultry breeding was 2.3 x 108 and 0.1 x 108 kg, mean and variance were set to zero and one, respective-respectively, and COD was 26.5 X 108 kg in 2009 (Wang ly) to minimize the influence of variable variance andet al, 2009).eliminate the infuence of different units of measurementThirty-seven water quality monitoring sites were de and render the data dimensionless (Singh et al., 2005).signed to cover a wide range of determinants at key sites,All mathematical and statistical computations were madewhich reasonably represent the water quality of the river using Microsoft Office Excel 2003 and STATISTICA 6.0.system in the Three Gorges area (Fig. 1).1.2 Monitored parameters and analytical methods2 ResultsData sets from 37 water quality monitoring sites com- 2.1 Cluster analysisprising 13 water quality parameters monitored monthlyover six years (2003- -2008), were obtained from thedata set with a view to group the periods (temporal) andEnvironmental Monitoring Center of Chongqing Munic-sites (spatial) spread over the river stretch and in theipality and Sichuan Province. The monitored parametersincluded temperature, pH, electrical conductivity (EC),resulted dendrogram. The linkage distance is reported asdissolved oxygen (DO), 5-day biochemical oxygen de-Dink/Dmax, which represents the quotient between the link-mand (BODs), potassium permanganate index (CODMn),age distance for a particular case divided by the maximalammonia-nitrogen (NH4*-N), petroleum, mercury (Hg),distance, multiplied by 100 as a way to standardize thelead (Pb), total nitrogen (TN), total phosphorus (TP), andlinkage distance represented on Y-axis (Wunderlin et al.,fecal coliforms (E. coli). The water quality parameters and2001; Simeonov et al, 2003).specific analytical method are presented in Table 1. TheTemporal CA generated a dendrogram grouping thesampling, preservation, transportation, and analysis of thetwelve months into three groups at Din:/Dmaxx100 < 80water sarmples were performed following standard methods(Fig. 2), with the differences among the groups being sig-(State Environment Protection Bureau of China, 2002).nificant. Group 1 consisted of March -May, correspondingto the normal fow (NF) period, group 2 consisted of June-13 Statistical analysisSeptember, corresponding to the high fow (HF) period,In this study, hierarchical agglomerative cluster analysisand group 3 included October to the following February,was performed on the normalized data set by Ward'scorresponding to the low fow (LF) period.method, using squared Euclidean distances as a measure20 rof similarity. A non-parametric Mann-Whitney U-test wasused to detect possible differences between the means of0owater quality variables among the group data sets (spatialor temporal) determined by cluster analysis. We appliedfactor analysis to obtain composite variables, which wasexpected to identify pollution factors afecting water qual-^ 60ity and latent pollution sources. Before the factor analysis,he Kaiser-Meyer Olkin (KMO) and Bartlett's test were 这40}performed to examine the suitability of the data set forfactor analysis.20}Multivariate statistical analysis of the river water qualitydata set was performed through cluster analysis, factorMayApr MarAug向Sep Jun Dec NovOct Feb Jananalysis, and Mann-Whitney U-test techniques. Non-Fg 2 Dendrogram showing temporal similaries of monitoring periodsparametric tests were applied on raw data, whereas, cluster produced by cluster analysis.analysis and factor analysis were applied on experimen-Thble1 Water quality parameters, units, analytical methods and detection limit as measured during 2003- 2008 for the Three Gorges areaParametersAnalytical methodsDetecection limitTemp(C)ThermometerGlass electrodeBC (mS/sec)ElectrometricDO (mg/L)lodimetry)2CODMn (mg/L)esler's reagent spectrophotomnetry0.5BODs (mg/L)Diution and inoculation test.5H4*-N (mg/L)N-reagent colorimetry0.05TN (mg/L)Alkaline potassium pe sulfate digestion-UV spectropbotometryTP (mg/L)Ammonium molybdate spectrophotometryE coli (num/L)Multi-tube zymolytic method/membrane fhter method中国煤化工”Petroleum (mg/L)Infrared spectrophotometyCold vapor atomic absorpcion spectro photometayCNMH GOHg (mg/L)MHGo0o05Pb (mgL)Atomic absorption spectrophotometry0.01No. 91463Spatial cluster analysis, like temporal cluster analysis,Min River (M1-M2), Dadu River (M3), Fu River (F1) andgenerated a dendrogram grouping all 37 monitoring sitesJinsha River (JS1-JS2).into three groups at Dink/DmaxX100 < 25 (Fig. 3). Group2.2 Non-parametric testA comprised sites T1-T6, and M4 -M6, group B comprisedsites W1, C2- C8, JL1-JLA, F2-F3, and Q1-Q2, and groupA temporal Mann-Whitney U-test was performed onC comprised sites F1, F4, M7, T7, C1, JIL5- -JL6, M1-M3,original data after dividing the whole data set into threeand JS1-JS2. The clustering procedure generated threegroups (group 1, group 2, and group 3), as delineated bygroups of sites in a very convincing way, as the sitestemporal cluster analysis. The results showed that tem-in these groups have similar characteristic features andperature, DO, CODMn and NH4*-N were different amongnatural background source types. In group A, six sites werethe three groups (Table 2). No significant differences werelocated in the main stream of the Tuo River (T1-T6) andfound between group 1 and group 3 for pH and TP, andthree sites were located in the middle reaches of the Minno significant differences were observed between group 2River (M4 M6). In group B, six sites were located in theand group 3 for EC, BODs TN and E. coli. No significantThree Gorges Reservoir (C5- C8, W1 and J1), and ten sitesdiferences were found between the group 1, group, 2 andwere located in the infuence areas (C2- C4, JL2- JLA, F2-group 3 for petroleum, Hg and Pb. As identified by Mann-F3 and Q1-Q2) of the Three Gorges Reservoir. In groupWhitney U-test, box and whisker plots of the selectedC, five sites were located most upstream of Min Riverparameters showing temporal variations are given in Fig. 4.(M7), Tuo River (T7), Fu River (F4) and Jialing River (JIL5Results from the spatial Mann-Whitney U-test showedand JL6), and six sites were located most downstream ofsignificant differences among the data sets of groups A,Table 2 Comparison of mean values between pollution regions and periods*ParametctsPeriod mcan valueRegion mean valueGroup 1Group2Group 3Group AGroupBGroup CTemperature (°C)17.06ab13.71a18.63b16.47.93a792b7.95a774h7.91ab8.13a，EC (mS/)169.9b145.7a155.8a135.09b249.02ab33.4aDO (mg/L)7.53ab7.0668.29a5.76b8.11ab8.59aCODMn (mg/L)3.25ab3.34b2.68a4.32b2.69ab2.56aBODs (mg/L2.01b1.53a1.57a2.84b1.46ab1.05aNH4* -N (mg/L)0.76ab0.34b0.46a1.26b0.26ab0.23aTN (mg/L)2.23b1.68a1.69a3. 9b1.51ab .1.01aTP (mg/L)0.15a0.15b0.14a0.11a0.12aE coli (num/L)44228b63359a74853674720b14389aPetroleum (mg/L)Hg (mg/L)0.0000240.0000220.000022a0.000031b0.000021ab0.000019Pb (mg/L)0.0065a0.0067a0.0063a0.0059b0.0053ab0.0088a●The comparison of means betwcen pollution regions and periods of all variables were according to Mann-Whitney U-test. The valucs followed by thesame ltter are not significantly diferent at the 0.05 probability level.20r100804C2(3533.中国煤化工THCNMH GFl3 Dendrogram showing spatial simiaritics of monitoring sitcs produced by chuster analysis.140Jourmal of Bnvironmental Sciences 2011, 23(9) 1460 -1471 / Jian Zhao et al.Vol. 230.0r20.0TemperatureDO15.00.0+。曾10.020.0-宁吕s.o宁宁宁0.0.0 LGroup 1Group 2Group 3Group2Group310.0 rCODmaNH:-N.0 t.06.0 tT置2.0|4.0Hz0当白会1GroupFg4 Box and whisker plots of tempornl variations of the selected parameters.B, and C for pH, EC, DO, CODMn. BODs, NH4*-N, TN, considered in FA as its concentration was very low inpetroleum, Hg and Pb (Table 2). There were no signifcanteach group and it was insensitive to analysis (Huang et al,diferences between group A and group B for temperature2010).and E. coli, and no significant difference between groupThe KMO results for the three group data sets (LP, MP,B and group C for TP. Box and whisker plots of theand HP regions) were 0.590, 0.596 and 0.624, respectively,selected parameters by Mann-Whitney U-test showingand the significance levels by Bartlett's test are all closespatial variations are given in Fig.5.to 0 (< 0.05), showing that factor analysis was effectivein reducing dimensionality. The factor analysis of the2.3 Factor analysisthree data sets (LP, MP and HP regions) yielded fourTo identify the main pollution factors, factor analysisvariance factors (VFs) for all groups with eigenvalueswas performed on standardized log-transformed data sets》1, explaining 63.9%， 61.7% and 61.2% of the totalseparately for the three groups (group A, group B, andvariance, respectively. The factor analysis results includ-group C), as delineated by cluster analysis techniques.ing the loadings, variance contribution rate of each VFThe water quality of some sites (F4, M7, T7 and JL5- and cumulative variance contribution rate are presentedJL6) in group C remained pristine over the six study yearsn Table 3. The VF loading plot (Fig. 6) for the three(2003- -2008), therefore source identification analysis wasgroups showed the relationships among the parameters; thenot performed on these sites. In addition, Hg was notsmaller distance, and the stronger correlation between thelable 3 Loadingsof 12 experimental variables on significant variance factors (VFs) for group A, group B and groupCParametersGroup A (four significant VFs)Group B (four significant VFs)Group C (four signicant VFS)VF1 VF2VF3 VF4 VF1VF2 VF3_ VF4_ VFI_ VF2 VF3_ VF4-0.07 0.01-0.83 -0.14 0.030.89 0.03-0.06 0.86-0.03 0.30 -0.09-0.18 0.59-0.090.18 0.58-0.26 0.39-0.277 -0.46-0.460.10 0.830.14 -0.13 0.46-0.15 -0.200.29. -0.17 0.09. -0.52 - 0.04)o-0.76 0.380.30050.00-0.89 0.002-0.87-0.16 0.05 -0.20:ODwa0.17-0.01 0.00 0.010.43 0.550.520.310.660.19 0.440.49-0.11-0.28 0.020.02 0.100.84-0.21.81-0.17 0.01NHA*.N0.25-0.08-0.13 -0.360.640.17 0.31-0.77 0.12IN0.150.13 0.580.760.14 -0.070.500.18 0.050.85 0.09E coli0.090.08 0.68Petrolcum0.04-0.03 0.000.38023 0.420.350.570.34 -0.13 -0.01 -0.02 0.85中国煤化工0.73220520922172 159 14.7CNMHG 639MH63.9Bold and italic values indicate strong and moderatc loadings, respectively.No.9Multivariate analysis of surface water quality in the Three Gorges area of China and implications for water management1465l0厂80pFEC600-400 I舌8宁皇宁旨200宁Group 1Group 2Group 310.0 r.0 rCODaNH,*-N.0T.0 t.o-GrouplGroup2Group3.04-8.0e-s-HgPt03↑6.0e-54.0e-5}出2.0-s50.01 It2.00.0020.0 rBOD,15.0曾10.0早s.0-z2号。0.010.0πPetroleum.3-高0.2f0.2-.0-! 0.1-自专Group IGroup中国煤化工YHCNMH GFig.5 Box and whisker plots of spatial variations of the selected parameters.No.91467Table 4 Results of KMO and Bartet's sphericity teatsbetween temperature and DO is a natural process becausewarmer water becomes saturated more easily with oxygen PeriodsKMOBartet's sphericityand it holds less dissolved oxygen (Shrestha and Kazama,Approx.Significance2007). Results also showed that VF3 explained 14.7% ofChi-Squarethe total variance and had strong positive loading on Pb andmoderate positive loading CODMn. This factor representedLF period0.572157.7340.000heavy metal pollution from industrial sewage, while VF4,NF period206.908HF period0.513157.294which explained 13.2% of the total variance and had strongGroup Bpositive loading on BODs and moderate positive loadings642.216on NH4*-N and CODMn, represented organic pollutionNE pernod451 2010.607455.219from domestic wastewater, industrial sewage, and a littleHF pernodfrom living garbage and bilge water (Zheng et al., 2008).0.600235.852L astly, for group C, VFI explained 17.2% of the total0.552181.416variance and had strong positive loading on temperature0.491152.842and strong negative loading on DO, as with VF2 of .group B. For group C, VF2 explained 17.0% of the totalfinal source identifcation results are shown in Table 5 andvariance and had strong positive loadings on BODs andwere similar to the above factor analysis analysis. TheTN, and moderate positive loading on CODMn. This factorresults also provided useful information about pollutionrepresented organic pollution from domestic wastewaterseasonality. For group A, sites were not affected by heavyand industrial sewage. Results showed that VF3 explainedmetal pollution in the HF period, indicating point source15.8% of the total variance and had strong positive loadingpollution. Sites were also afected by oil pollution in groupon TP and negative loading on NH4+-N. This factorC during periods NF and HF. In addition, sites were alsorepresented nutrient pollution. The inverse relationshipaffected by fecal and nutrient pollutions, which were notbetween NH4*-N and TP suggested vast difference in pol-well identified by performing FA on group dataset.lution sources such as point source (domestic wastewater)and non-point source (agriculture activities). Finally, VF43 Discussionexplained 13.9% of the total variance and had moderatepositive loadings on Pb and E. coli. This factor represented3.1 Temporal similarity and variation of water qualitymixed pollution by heavy metal and fecal, and was subjectto the effects of industries and local livestock farm.As shown in Fig. 2, the temporal variation of waterTo identify the spatio-temporal patterns in latent pol-quality in the Three Gorges area was greatly determinedlution factors, the VFs scores were plotted for eachby hydrological conditions (low, normal, and high flowmonitoring site to groups A, B and C (Fig. 7). The plotperiods) or local climate (spring: March to May, summer:showed high associations of the pollution groups to theJun to September, autumn: October to December, andmonitoring sites, and discriminated among the efects ofwinter: January to February).potential sources in different periods (Qu and Kelderman,Characterization of seasonal changes in surface water2001; Kowalkowski et al, 2006); the greater VF score, the quality is an important aspect for evaluating temporalgreater the effect. In group A (Fig. 7a and b), some sitesvariations of river pollution due to natural or anthropogenic(for example e, M4, T1, T2, T3 and T6) were stronglyinputs of point and non-point sources (Quyang et al., ,infuenced by organic and nutrient pollutions in NF period.2006). Among the measured water quality variables, tem-The other sites (for example, M5, M6 and T5) were mainlyperature, DO, CODMn and NH4*-N difered significantlyinfluenced by biochemical pollution during the NF period.among the periods (Table 2). Because of local climate,In addition, T1 received more heavy metal pollution allthe average temperature during the HF period (June toyear, and T6 received more biochemical and oil pollutionSeptember) was clearly higher than that in other periodsduring LF and NF. In group B (Fig. 7c and d), all sites were(Fig. 4). The higher values of CODMn and lower valuespredominantly infuenced by nature factors during the HF of DO during the HF period were infuenced by variousperiod and mixed pollutions (fecal, nutrients, organic andfactors. Intense rainfall and river-flow caused the wash ofheavy metal) all the year. In group C (Fig. 7e and f), mostorganic matter into the surface water, which reduced thesites were infuenced by nature and nutrient pollutionsconcentration of dissolved oxygen through biodegradation,during the HF period. A small number of sites (for examplewhile the increase in temperature caused a decrease inF1, MI and M2) were associated with organic, heavy metaloxygen solubility, and a further reduction in the DOand fecal in NF period.concentrations (Kannel et al., 2008). In addition, lowerFactor analysis was also performed on each period (LF,NH4*-N concentrations were found during the HF period,NF and HF) data set for the three groups (groups A, B andwhich were attributed to point sources that provide aC), considering the effects of temporal differences on VFs.relatively constant input into the river throughout theThe KMO and Bartlett's test results (Table 4) indicated thatyear (Edwards and Withere 20081 Diue 切n dilution effect,factor analysis was useful for all the data sets to provideNH4*-N concentr中国煤化工dingly withsignificant reductions in dimensionality. The proceduresincreasing flow ral|YHcNMHGnsofBODswere the same as the above factor analysis analysis. Theand TN were all nigner' aunng tne Nr period than in1468Joumal of Eavironmental Sciences 2011, 23(9) 1460-1471 /Jian Zhao et al.Vol. 232.52.5-2.0--.5--T6L+T2N-T6、0JI6Nt 0.5-M6GM6N5 0.4N M4261M4RHMSN0.5----- S4 -T3HustBN_ M4N0.5-- MSHTIN-1.0MlLsM6HBLMSHHMSN.1.0_TSL-1.-1.5 -1.0 -0.5 0.00.5 11.5 2.04-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.sVF1F3.-LF periodi Lpriod2.0。NEperiodHF poried.5.CSHCNQ214 m.0-12. C75C7HC6NLITEINO2H BLGO2NCSHC6HWHC3FCTN2N4C8NT CINCN CHraNCT.QHJIN十0.5--” WIN-1.0 -出WLt1.5--Q2uC2L.0_J4N★-1-2.5 -2.0 -1.5 -1.0-0.5 0.0 0.5 1.0 1.5 2.0-2.5-2.0-1.5-1.0-0.50.00.51.01.52025VF322.0F1yMIN2.1.Mm1FIL明1H0ZL的MIN_ 0M11g 0.5其0.0●JSIH0.0CNM2PM3IJS2H-0.5fINJSIN-2.-1.5 -1.0 -0.5 0..5 1..5 2.025 2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.01.5 2.0Fig 7 Scatter plot of scores for the four VFs for group A (a and b), group B (c and d) and group C(e and f). Each symbol was short for each monitoringsite in diferent periods (LF, NF and HF), for example,“C2L", means the C2 in the LF period. LF: low flow; NF: normal flow; HF: high flow.Table5 Source ientification results of cach period for three pllution regionsPeriodsVF2VF4VF5Group AHeavy metal pollution Organic pollutionFecal pollutionOil pollutionNF periodNutient pollutionFecal + HeavyNatural pollutionOil plltionmetal pollutionsHFperiod Nutient (TN) +Organic pollutionOil pollutioNutient (TP)fecal polutionspollutionGroup BLFperiod Heavy metal plluion Fecal + 0il pllutions Organic pllutionNutient pllutionNF period Fecal pollutionHeavy metal pollutionHFperiod Nutrient+ fecal+Heavy metal pollution Natural pollutionoil pollutionsGroup C中国煤化工LFperiod Organic pllutionNutrient pollutionHeavy metCNMH GNF period Organic pllutioPhysical-chemistry Nutient +.MYHmctal pollutionshf period Noutrient pllutionFecal + Oil pollutions Organic pollution Heavy metal pollutionPhysical-chemistry pollution1470Jounal of Environmental Sciences 2011, 23(9) 1460-1471 /Jian Zhao et al.Vol. 23related to natural factor, organic pollution, nutrient pol-temperature, DO, CODMn and NH4+-N, showed significantlution, and physicochemical pollution. Zhou et al. (2007)diferences among the periods. Ten variables, i.e, pH,discovered that soil weathering, nutrient pollution, mineral EC, DO, CODmn, BODs, NH4* _N, TN, petroleum, Hgpollution, and physicochemical pollution were the mainand Pb showed signifcant differences among the regions.pollution sources in coastal water in easterm Hong Kong.Furthermore, the pollution of the HP region was relativelyZhang et al. (2009) found that oxygen consuming organicserious and needs to be controlled. (2) The factor analysispollution, toxic organic pollution; heavy metal pollution,helped in identifying six latent pollution sources, i.e, or-fecal pollution and oil pollution were five main types of ganic pollution, nutrient pollution, heavy metals pollution,pollution in Daliao River basin. The numbers of identifed fecal pollution, oil pollution, and nature pollution. Thpollution sources in the above results were less than those pollution patterns varied spatially and temporally. For thefrom our study. The quantity of pollution may be deter-HP region, the major pollutants were organic matter andmined by a number of variables and actual pollution levelsnutrients during the NF period, heavy metals during theand types (Huang et al., 2010). In this study, we identifiedLF period and petroleum during the HF period. For the MPtwelve variables. However, the relatively high numbers of region, the identified pollutants primarily included organicpollution indicated relatively high pollution levels and pol- matter and heavy metal year -around. For the LP region,lution types in the river of the Three Gorges area than thoseorganic pollution was significant during both NF and HFin other areas. In addition, the different pollution types inperiods; nutrient and heavy metal levels were high duringthe above results along with this work indicated that eachboth LF and HF periods. Furthermore, pollution factorsriver had unique physical and chemical characteristics due identification for each period of the three pollution regionsto its different natural and anthropogenic features (Huang provided useful information about seasonality of pollution.et al., 2010).(3) Generally, domestic wastewater, agricultural activities,The types of pollution in the monitoring sites for LP,and runoff were responsible for the low water quality inMP and HP regions differed significantly during the threethe whole region. In the HP region, most sites along theperiods (Fig. 7). Obviously, most sites received moremain stream of the Tuo River and the middle stream oforganic pollution during the NF period than that in the the Min River were influenced by point sources, primarilyother two periods. For HP, some sites (T5, T6, M5 and M6) discharged from wastewater treatment plants. In the MPalong the middle stream of the Min and Tuo Rivers wereand LP regions, sites were mainly from multiple pollutionstrongly affected by biochemical and oil pollution during sources originating from agricultural runof and domesticthe LF and NF periods. For SP and MP, sites received morewastewater, as well as pollution from some chemicalnature sources during the HF period. For MP, sites wereindustrial activities.infuenced by mixed pollution (i.e., fecal, oil and nutrientAcknowledgmentspollutions) all year. For SP, sites were mainly affectedby nutrient pollution during the HF period. The level ofThis work was supported by the National Water Spe-pollution in the LF period was in the middle of the othercial Project (No. 2009ZX07526-005) and the Strategictwo periods. The T1 site in HP, M1, M2 and M3 sites in SP Environmental Assessment Project (No. HP1080901). Theand most sites in MP were strongly infuenced by heavy authors sincerely thank Qian Jun, Tong Hongjin and Dengmetal pollution all year. Several studies have been devoted Cunguang for their help in providing data.to seasonal and polluting effects on water quality (Vega etal, 1998; Ouyang, 2006; Zhou et al., 2007; Zhang, 2009;ReferencesRazmkhah et al., 2010). Some discrepancies in the aboveresults along with his work can be atributed to different CAE (Cninese Academy of Engincring). 2008. Strategy Consutingriver environments and different water quality parameters,Report of Water Pollution Control on the Three Gorges Reservoiras well as to the different time periods (i.e., seasonal vs.and is Upstream Areas. China Environment Science Press,overal) used in each study (Ouyang et al.. 2006). However,Beijing, Chinaresults suggested that pollution factors (or water qualityCRAES (Chinese Research Academy of Environmental Sciences),2010. Strategy Environmental Impact Assessment Report forvariables) that play important roles in infuencing riverChengdu-Chongqing Economic Zone Key Industrial Deveiop-water quality in one environment may not be important inment. Chinese Research Academy of Environmental Sciencesanother.Press, Beijing. Chin:Cao M, Cai Q H, Liu RQ, Qu x D, Ye L, 2006. Comparative research4 Conclusionsctors in the frontofThree Gregsreservoier .on physicochemical factors in ,before and after the initiate impounding. Acta HydrobiologicaMultivariable statitical methods were successfully ap-ChaiC, Yu ZM. Shen zL. SongXX,CaoX H, Yao y, 2009. Nutricatplied to evaluate spatio-termporal variations in river watercharacteristics in the Yangtze River Estuary and the adjacent Eastquality and source identification at the monitoring sitesChina Sea before and after impoundment of the Three Gorgesin the Three Gorges area. The main conclusions were asDam. Science of the Total Ervironment. 407(16): 4687. 4695.following. (1) Hierarchical cluster analysis grouped the Chang HI, 2050 Spatia中国煤化工qualiy in thetwelve months of the year into three periods and the 37Han River andis! YHCNMH02. Wuer Airmonitoring sites into three groups based on the similaritySoil Pollution, 161of their water quality characteristics. Four variables, i.e, ChenXQ, Yan YX FuR s, Dou XP, ZDhang E F, 2008. Sediment