Environmental Characteristics of Groundwater: an Application of PCA to Water Chemistry Analysis in Y

Mar. 2007Journal of China University of Mining & TechnologyVol.17 No.1Available online at www.sciencedirect.comSCIENCEJ China Univ Mining & Technol 2007, 17(1): 0073- 0077Environmental Characteristics ofGroundwater: an Application of PCA toWater Chemistry Analysis in YulinDONG Dong-lin, WU Qiang, ZHANG Rui, SONG Ying-xia, CHEN Shu-ke, LI Pei,LIU Shou-qiang, BI Cen-cen, LV Zhen-qi, HUANG Song-linState Key Lab of Coal Resources and Safe Mining, China University of Mining & Technology, Bejing 100083, China :Abstract: For our investigation into the water quality in Yulin city, we colleted 76 typical water samples to be testedfor particle quality. By applying a Romani type classification method the groundwater of Yulin city was classified intonine categories by type, i.e, Ca-HCO3, Na-HCO3, Na-HCO3-SO4-Cl, Na-HCO3-SO4, Na-Cl, Na-CIl-HCO3, Na-Ca-HCO3, Ca-Cl-HCO3 and Ca-HCO3-SO4-Cl. A principal component analysis was carried out in order to analyze thegroundwater environment. From this analysis we considered that the information collected could be represented by 21indices from which we extracted seven principal components, which, respectively, accounted for 37.4%, 13.0%, 8.1%,7.2%, 6.3%, 5.9% and 4.6% of the total variation. The results show that the groundwater environment of this region islargely determined by characteristic components of the natural groundwater background. One part of the water was pol-luted by leaching/eluviation of solid waste generated from coal mining. Another part of the ground water was contami-nated by acid mine water from the coal layer and from improper irigation. In addition, geological and hydrogeologicalconditions also cause changes in the water environment.Key words: water environment; principal component analysis; component loading; groundwater type; Yulin cityCLC number: X 5231 Introductionis deemed imperative. It is necessary to confront themain contradiction of abundant mineral resources andYulin city is located in the north of Shaanxi prov-the lack of an adequate water supply. A first step is toince and is close to the borders of four provinces,analyze the quality of the available water supply,Gansu, Ningxia, Menggu and Shaanxi. It is also at thewhich is characterized by a great many chemicalcenter of abundant resources. Groundwater in use incompounds. In the analysis we would like to use thethe city has a shallow water level, low total salinitysmallest number of variables and still reflect the es-and is suitable for drinking and industrial use; there-sential conditions of our entire aquatic environmentfore it has potential for exploitation”But a shortageobjectively, w ithout losing any information about thisof water resources is the major factor inhibiting de-most important resource. Confronted with this situa-velopment of mineral resources and the establishmenttion, we concluded that principal component analysisof a chemical engineering base. Moreover, a paper(PCA) was the most suitable method to tackle themill in the Maosuwu desert and the Yuyang coal mineproblem. Our reason was that this method is able toboth discharge contaminated water which causesretain most of the primary information by simplifyingaquatic environmental deterioration. This situationand combining high dimensional variables. Hence,severely affects the economic development of Yulinafter choosing 76 typical water samples in our ex-whose ecological environment is very vulnerable.periment, we carefully analyzed the water resourcesTherefore, an analysis of groundwater characteristicsin the region by applying the method of Romani andReceived 06 June 2006; accepted 15 July 2006中国煤化工Project 2004-295 supported by the Trans-century Scientific Great Project of Ministry of Education of C.MHCNMH GCorresponding author Tel: +86-10-62331248; E-mail address: dd@cumtb.edu.cn.74Journal of China University of Mining & TechnologyVol.17 No.1using PCA.major ions in groundwater; totally dissolved solids,such as anions, which amounts to more than 25%2 Sampling and Methodmilliequivalent, are combined with cation. Ground-water is classified into 49 kinds. On the other hand,2.1 Samplingthe Piper trilinear diagram method consists of twoIn our investigation, water samples were obtainedtriangles and a rhombus and, as well, every borderby using flint glasss conrepresents the milliequivalent of six major ions; to-containers with a capacity of be-tally dissolved solids are stween one and two liter, to collect water from 76 typi-in the rhombus. Their chemical characterization iscal places, such as in the vicinity of coal mines, rivers,based on a discrimination between the diferent re-residential areas, Hongjianzhuo and elsewhere. Awater analysis had never been carrI out in the Yulingions where water samples have been collected. Theirregion. For the different ions which would be identi-collctive characteristics use these sixnajor ions asfied there are various ways for retaining them. Forthe basis of the an; analysis while neglecting the chemi-example, when collected, nitrite and free carbon di- .cal behavior of other ions and relevant geochemicaloxide samples need to be send to the laboratory im-problems. They cannot represent the “special”mediately and measured th; same day if this cannotchemical behavior of groundwater because of theirbe done in the field. Samples of ammonium should belimitations.send to the laboratory at once and should be meas-To solve this problem, ameliorated Romani dia-ured within three days. Small samples (< 100 mg/L)grams have been proposed, given the limitations Ro-of fluorine ions, chloride ions, bicarbonate anions,mani experienced in his research with using trilinearhydroxide radicals, sulfateradicals, nitrate, sodiumdiagrams'. This method added comments on specificions, calcium ions, magnesium ions and silicic acidions in order to classify， groundwater correctly. In hiswere also immediately sent to the laboratory and thestudy of the chemical behavior and associated geo-analyses were completed in ten days. The measure-chemical problems he was able to ascertain the effi-ment of ferrous iron was carried out as follows: weciency of his method for special uses. The advantageused a vinyon bottor flint glass container to obtainis that all water characteristics cae studied in thea sample of 250 mL, then added a sulphuric acid so-area of a square without losing any merits of the arealution (1+1) of about 2.5 mL and a sulfate ammoniaof the rhombus.solution about 0.5- -1.0 g. In the end, this was pres-For the cassification of water, only the proportionssurized by olefin and stored in glass vessels for aof the principal cations and anions, in terms of pmaximum of thirty days.centages and epm values are plotted in each triangle(Fig. 1). The two triangles are further divided into2.2 Method of measurementseven fields, representing the following types of wa-After collecting the water samples, we immediately ter: 1) cation triangle: CI calcium type, C2 magne-measured their pH values on the spot byusing a go-sium type, C3 sodium type, C4 sodium-calcium type,niophotometer.hen used a High Resolution in-C5 calcium-magnesiumtype, C6 sodium-magnesiumductance. coupled partile mass spectrographtype, C7 calcium-magnesium-sodium type; 2) anion(HR-ICP-MS (Elemnent I )) to measure all main posi-triangle: AI bicarbonate type, A2 sulphate type, A3tive ions in all the samples; we used ion chroma-chloride type, A4 chloride-bicarbonate type, A5 bi-tograpy (IC, DIONEX-500) to measure .; the amountscarbonate-sulplA6 chloride-sulphate type,of fluorine ions, chloride ions, sulfate radicals, nitrateA7bicarbonafesupyaree-sulphate-chloride type. When plottingetc; titration methods were applied to measure theanions and cations, nitrate was combined with chlo-amounts of carbonic acid anions, bicarbonate anionsride. We plotted the data of our 76 on-the -spot waterand hydroxide radicals (we used the Metrohmsamples from Yulin on the Romani diagram. These785DMP automatic titrator).are presented in Table 1. In this table, the water sam-ples are represented by calcium-bicarbonate type (C13 Results and DiscussionAl), sodium- bicarbonate type (C3 A1), sodium-bi-carbonate-sulphate-chloride type (C3 A7), sodium-3.1 Characteristics of groundwater ion classir.bicarbonate-sulphate(C3 A5), sodium- chloridecationtype (C3 A3), sodium-chloride-bicarbonate type (C3In studies of groundwater quality, domestic and in-A4), sodium-calcium- bicarbonate type (C4 Ai), cal-termational scholars apply various methods and dia-cium-chloride-bicarbonate type (Ci A4) and cal-grams to analyze and explain the chemical compo-cium-bicarbonate-sulphate-chloride type (Cl A7).nents of groundwater. Of these methods, the PiperIn order to study water quality, plotting in the twotrilinear diagram and lyaye lassification are thetriangles is ext中国煤化工Fig.1). Thisbest known and used abroad for teaching purposes2.field is divided_oup II (G2).The lIyraneclassification isbased on six kinds ofIn group I theYHCNMHGexceedthebicarbonate anions whereas in group II the reverse isDONG Dong-lin et alEnvironmental Characteristics of Groundwater: an Application of PCA to7:true. The water samples in group I have permanentFe3+ (13)，Fe2+ (14), NH4*(15), CI(16), SO2(17),hardness without residual sodium carbonate (RSC).HCO3 (18), CO3 (19), NO2 (20) and NO3(21). AThe water samples in group II have temporary hard-preliminary analysis of the groundwater chemistryless with RSC. Apart from showing this primarydata in the mining area of Yulin showed that thecharacteristic of water, the field is used to study thedominant components (ion concentrations) are corre-combination of water and other geochemical prob-lated. Bartlett's sphericity test, a test for the inde-lems similar to the diamond-shaped fields of a Piperpendence of the 21 variables, was used to determinediagram. Out of 76 samples, RSC is present in 23whether the correlation matrix R was an identity ma-samples (30.3%) and absent in 53 samples (69.7%). Ittrix. The result indicated a probability of 0.000, farshows the salinity of the water and also the suitabilitybelow the 0.05 level, which is_ in line with the pre-of the water for rrigation.conditions for the use of PCAl+ 7]Table 1 Groundwater types based on Romani methodPCA constructs independent linear combinations ofthe original variables, i.e., the principal components.TypeSample numbersampleStarting with the first component, each explains a15.7.8,9,10.11.12.13.14,15,16,17,26.29.333decreasing amount of variation. In our case, the vec-C1 A14.36,37.38.39.42.43.44.45.46.47.49.50.55.564tor of eigen values( λ ) of the R matrix represents the,57.58.6.64.65,66.68.69.70.73.74,75.76variance of each component. It is seen (Table 2) thatC3Al 3.4.6.19.20.21.23,25,28.31.32.40.53.60.71for the first six components, λ > 1, which, cumula-C3A74.30.52tively, explains 78% of the total variation of theC3.A5 27original variables, less than the required 80%, so oneC3A341additional component was needed to raise the totalC3A4 51explained variation to 82.5% >80%. Our explanationof the chemical characteristics of the water samples isC4A1 2.18.22.35.48.54,61.62.67based on our interpretation of these first seven prin-cipal components. See also Fig. 2.C1A759Table 2 Initial eigenvaluesExtraction sums ofInitial eigenvalues0100squared loadingsCom-ponentTotalVarianceCumu-Vari- Cumu-(%)Lative Totalance 2aeyeA37.845 37 .35637.356 7 .845 37 356 3 7.3562.73413.01850.374 2.734 13.018 50.3741.7068.1258.4981.706 8.124 58.4981.511 7.19565.693 1.511 7.195 65.693121.328 6.3272.019 1.328 6.326 72.019100d。1.2435.91977.938 I .243 5.919 77 9380.960 4.5782.510 0.960 4.573 82.5100.780 3.71686.227巴|\C6~ C41 0.001 0.007 10.0000C7C5In the first principal component, salinity, Na' , total2dissolved solids, oxygen consumption, Mg，Cl，SO:+CI+NO3MgSO4, HCO3, CO32 " and fluoride have the highestloads, i.e, the greatest absolute coefficients of theFig. 1 Water chemistry in a Romani diagramcomponent, which explains 37.4% of the total varia-3.2 Application of PCA to groundwater chemis-tion. From past, anecdotal, information of Yulin, thearea belongs to a high fluoride zone'. On the othertry characteristics in Yulinhand, the other groups have a more conventional wa-In order to tabulate and conveniently chart theter origin, so the first principal component can bechemicals in the 76 water samples tested, identifica-defined as a natural water component. From Table 3tion numbers 1 to 21 have been assigned, respectively,we see that salinity, Na , total dissolved solids, oxy-to the chemicals as follows: total acidity (1), totalgen consumpti中国煤化工03 ，CO3alkalinity (2), soluble SiO2 (3), total dissolved solidsind_ fluoride aYHcorrelation(4), oxygen consumption (5), salinity (6), copper (7),coefficient rangc N M H Gerefore, thefluoride (8), pH (9), Na+(10), Ca2+ (11), Mg-+ (12), .first principal component not only controls salinity,.76Journal of China University of Mining & TechnologyVol.17 No.Nat, total dissolved solids,and SO4, but also .monia. As well, ammonia is brought about by wastedisposable oxygen consumption, fluoride, HCO3~ andwater discarded from chemical and petrochemicalCO32. In the second principal component, total acid-plants, coal gas, coking plants, chemical fertilizersity and concentration of the Cat+ payload are rela-and other industries. We can conclude that the thirdtively high. This component accounts for 13.0% ofprincipal component is affected by sewage and indus-trial effluents. It is seen that cumulatively, the firstexplained variation.three components explained 58.5% of the total varia-Table 3 Correlation matrix of water chemistrytion in the original variables. Fig. 2b shows the load-componentsings of the third component on the first with the threemain variables plotted in the ellipse. The fourth prin-1.00 -0.06 0.49 -0.06 0.10 -0.02 0.250.16cipal component is dominated by nitrate, which has0.06 1.00 -.0.22 0.52 0.37 0.54 -0.03the highest payload. It explains 7.2% of the total0.48 -0.22 1.00 .-0.35 .0.23 .-0.36 0.14- 0.02variation. Farmland drainage and returned irrigation0.06 0.52 0.35 1.00 0.62 0.97 -0.02water in Yulin contains a lot of nitrate, so the fourth0.100.37 -0.23 0.62 1.00 0.61 0.1principal component can be regarded as the contribu-6 -0.02 0.54 -0.36 0.97 0.61 1.00 -0.01-0.02tion of irrigated farmland. Fig 2c demonstration ofthe loadings of this component. In the fifth principal0.25 -0.03 0.14 -0.02 0.15 -0.01 1.000.30component Fe't has the highest payload and its con-8.0.26 0.30 -0.50 0.67 0.41 0.64 - 0.08 .- 0.06tribution to explained variation is 6.3%. The mainsource of iron is related to geological and hydro-0.16 .0.03 -.0.02 .0.02 0.28 .0.02 0.3000geological conditions, chiefly due to ion migrationand accumulation, after weathering and water solubleGiven the source of the total acidity, mining andions in the iron ore strata of the plains. Iron accumu-solid waste leaching into the groundwater, this com-lates in the plains, it oxidizes continuously and thenponent can be regarded as external forces from coalseeps into the groundwater. In addition, a mixture ofmining and from the chemical composition ofsilt and organic material which is rich in the aquifergroundwater. The variables identified in the first twohas an effct on absorption and accumulation of iron. .principle components are plotted, respectively, in theTherefore, the fifth principal component can belarge and small ellipses in Fig. 2a. The third principalviewed as an infection of the groundwater environ-component accounts for 8.1% of the total variationment by geological and hydrogeological conditionsdue to the high loadings of total alkalinity and am-This is shown in Fig. 2d. The sixth principal compo-monia. Ammonia comes mainly from human andnent accounts for 5. 9%. Here, the copper ion concen-animal wastes, which amounts to an average 2.5- 4.5tration has the highest load. On the basis of a sitekg per person per year. The use of agricultural fertil-analysis, waste water turns out to be the main sourceizer and rain runoff is another major source of am-0，0.60.8.60.40.2.49817 '.2 t0.0 t3 0.04-0.2-0.42-0.6-0.6 -00.6 T.cComponent 1Component I(a) Componcnt 1 and2(b) Component 1 and 3(C) Cormponent 1 and 40.6 [.4.).2 t.0t3..0 t0.0-0.2 t14-0.4 t .J14-0.6 -0..2 0.6 1.0(d) Componcnt I and 5(c) Component 1 and 6中国煤化工。and 7Fig. 2 Loading plots of component 2 to 7 on component 1 of the groulMYHCNM HG.DONG Dong-lin et alEnvironmental Characteristics of Groundwater: an Application of PCA toof Cu, discharged by electroplating, metallurgy, metalmaining water is polluted, caused by leaching / eluvi-processing, mining, petrochemical and chemical ination of solid wastes generated from coal mining, bydustries and other sectors, so it is concluded that theacid mine water of the coal layer and by impropersixth principal component can be viewed as the in-irrigation. In addition, geological and hydrogeologi-dustrial waste water contribution to the pollution ofcal conditions also affct the aquatic environment.groundwater (see Fig. 2e). Nitrite has the highestpayload in the seventh component which contributes4 Conclusions4.6% of the variation. Nitrite is very unstable. Gener-ally speaking, nitrite does not exceed 0.1 mg/L in1) With the aid of the Romani method, the ground-natural water. A large amount of nitrite is dissolved inwater was classified into nine categories in Y ulin city,the soil. The soil pH increases when ammoniai.e., Ca-HCO3 type, Na-HCO3 type, Na-HCO3-SO4-evaporates and is gasified. During this time, nitriteCl type, Na-HCO3-SO4 type, Na-Cl type, Na-Cl-bacteria are still very active and able to produce aHCO3 type, Na-Ca-HCO3 type, Ca-Cl-HCO3 type andlarge amount of nitrite. However, nitrite is unsuitableCa-HCO3-SO4-Cl type.for bacteria under acid conditions so that instead, ni-2) A principal component analysis was carried outtrite accumulates in the soil. When saturation within order to evaluate the sources of groundwater qual-nitrite occurs, the pH of the soil decreases and theity. The results show that the first seven componentsaccumulated nitrite evaporates and escapes into theaccounted for 37.4%， 13.0%， 8. 1%，7.2%， 6.4%,air. The mechanism of producing nitrogen dioxide gas5.9% and 4.6% of variation respectively, amountingconstitutes a hazard in agricultural cultivation. Theto an explanation of 82.5% of the total variation ex-high ammonia levels are the result of making use ofhibited in the groundwater environment of this region.excessive amounts of ammonium bicarbonate, urea,This result was largely determined by characteristicnitration ammonium, ammonia, ammonium chloridecomponents in the natural groundwater background.and ammonium sulfate. Therefore, the seventh prin-At the same time, however, one part of the water wascipal can be seen as excessive nitrogen in the ground-polluted by leaching/eluviation of solid waste gener-water environment and is shown in the small oval of ated from coal mining and another part was contami-Fig. 2f.nated by acid mine water in the coal layer and by im-The seven plots in Fig. 2 are an accurate reflectionproper irrigation. In addition, geological and hydro-of much of the state of conditions in the Yulingeological conditions also caused changes in the wa-area.They feature normal water conditions, coincid-ter environment.ing with a background of fluoride and the presence ofdominant, naturally occurring common ions. The re-References1] Dong D L, YuX Y, Wu Q. Assessment system for water environmentof coal mines in energy resource base of Mongolia,Shanxi and Shaanxi - a case study in north Shaanxi. Journal of China University of Mining & Technology, 2003, 32(3):247- 250. (In Chinese)2] Wang D C, Zhang R Q, Shi Y H. Hydrogeology. Beijing: Geology Press, 1998.3] Srinivasa Y R, Reddy T V K, Nayudu P T. Groundwater quality in the Niva River basin, Chittoor district, Andhra Pradesh,India. Environmental Geology, 1997, 32(1): 56- -63.4] Fung T, Le D E. Application of principal components analysis to change detection. Photogrammetric Engineering and Re-mote Sensing, 1987, 53(12): 1649- 1658.Feng L H. Principal component analysis of environmental quality. Mathematics in Practice and Theory, 2003, 33(8): 32- -35.6] He B, Meng Q. Some notes on the principal component analysis method. Journal of Yunnan Normal University, 2002, 22(2):-8.7] Mrklas 0, Bentley1 L R, Lunn s R D. Principal component analyses of groundwater chemistry data during enhanced biore-8]YuJ Y, He X H. Data statistics and application of SPSS. Beijing: Posts and Telecom Press, 2006.9] Su Y. A preliminary study on the formation of groundwater with high fluorine in Xianyang city. Journal ofXianyang TeachersCollege, 1999, 14(3): 40- 43.中国煤化工MHCNMH G.