Analysis of Historical Sources of Heavy Metals in Lake Taihu Based on the Positive Matrix Factorization Model

International Journal of Environmental Research and Public Health

Article Analysis of Historical Sources of Heavy Metals in Lake Taihu Based on the Positive Matrix Factorization Model

Yan Li 1,2,†, Liping Mei 3,†, Shenglu Zhou 1,2,*, Zhenyi Jia 1, Junxiao Wang 1, Baojie Li 1, Chunhui Wang 1 and Shaohua Wu 1

1 School of Geographic and Oceanographic Sciences, Nanjing University, 163 Xianlin Road, Nanjing 210023, Jiangsu, China; [email protected] (Y.L.); [email protected] (Z.J.); [email protected] (J.W.); [email protected] (B.L.); [email protected] (C.W.); [email protected] (S.W.) 2 Key Laboratory of Coastal Zone Exploitation and Protection, Ministry of Land and Resources, Nanjing 210008, Jiangsu, China 3 School of Chemistry and Chemical Engineering, Nanjing University, 163 Xianlin Road, Nanjing 210023, Jiangsu, China; [email protected] * Correspondence: [email protected] † These authors contributed equally to this work.

Received: 15 May 2018; Accepted: 18 July 2018; Published: 20 July 2018

Abstract: Analysis of sediment grain sizes and heavy metal correlations in the western part of Lake Taihu shows that the grain size of the sediment is stable as a whole. With increasing depth, the grain size tends to decrease. Heavy metals such as Cr, Cd, Pd and Sr are strongly correlated and influence each other. Based on the positive matrix factorization (PMF) model, this study classified the origin of heavy metals in the sediments of western Lake Taihu into three major categories: Agricultural, industrial and geogenic. The contributions of the three heavy metal sources in each sample were analyzed and calculated. Overall, prior to the Chinese economic reform, the study area mainly practiced agriculture. The sources of heavy metals in the sediments were mostly of agricultural and geogenic origin, and remained relatively stable with contribution rates of 44.07 ± 11.84% (n = 30) and 35.67 ± 11.70% (n = 30), respectively. After the reform and opening up of China, as the economy experienced rapid development, industry and agriculture became the main sources of heavy metals in sediments, accounting for 56.99 ± 15.73% (n = 15) and 31.22 ± 14.31% (n = 15), respectively. The PMF model is convenient and efficient, and a good method to determine the origin of heavy metals in sediments.

Keywords: sediment; positive matrix factorization; heavy metal; source resolution

1. Introduction Lake Taihu, located in Southern Jiangsu Province, is the third largest freshwater lake in China, with an area of 2338 km2, and plays a decisive role in regional and socioeconomic development. Heavy metals enter aquatic systems via surface runoff or atmospheric deposition and accumulate in sediments through the adsorption and sedimentation of suspended matter. Heavy metals are important pollutants of aquatic ecosystems because of their environmental persistence, toxicity and potential for accumulation in food chains [1–4]. Lake Taihu water ﬂows in mainly from the west, and the environmental quality of the western part of the lake plays a crucial role in the environment of the entire lake. However, the sources and history of heavy metals in the sediments of West Lake Taihu are

Int. J. Environ. Res. Public Health 2018, 15, 1540; doi:10.3390/ijerph15071540 www.mdpi.com/journal/ijerph Int. J. Environ. Res. Public Health 2018, 15, 1540 2 of 12 Int. J. Environ. Res. Public Health 2018, 15, x FOR PEER REVIEW 2 of 12 unclear.Taihu Therefore,are unclear. it Therefore, is necessary it is to necessary investigate to investigate these issues these and issues provide and a basisprovide for a the basis monitoring for the andmonitoring management and ofmanagement heavy metal of pollution.heavy metal pollution. AtAt this this stage, stage, the the modelsmodels of of source source determination determination forfor sediment sediment heavy heavy metals metals can be can divided be divided into intotwo two main main types: types: diffusion diffusion and and receptor-oriented receptor-oriented models. models. The Thediffusion diffusion model model (which (which involves involves a a mathematicalmathematical modelmodel combined with certain certain assumptions assumptions to to select select a aseries series of of parameters parameters to tocalculate calculate thethe diffusion diffusion and and migration migration of of pollutants pollutants inin aa simulationsimulation of the actual actual situation), situation), has has the the pollution pollution sourcesource as as the the main main research research object object [ 5[5,6].,6]. TheThe diffusiondiffusion equation can can be be used used to to calculate calculate the the source source contributionscontributions according according to theto pollutantthe pollutant discharge, discha distancerge, distance between between the study the areastudy and area the and discharge the source,discharge physical source, and physical chemical and properties chemical of properti the pollutants,es of the discharge pollutants, rate, discharge weather conditions,rate, weather and otherconditions, parameters. and other On the parameters. other hand, On thethe receptor-orientedother hand, the receptor-oriented model (a digital model model (a anddigital method model for identifyingand method and for analyzing identifying the differentand analyzing sources the and differ contributionent sources rates and ofcontribution receptor contaminants) rates of receptor takes contaminants) takes a contaminated area as the research object [7–10]. The positive matrix a contaminated area as the research object [7–10]. The positive matrix factorization (PMF) model is factorization (PMF) model is one of the source resolution methods recommended by the US one of the source resolution methods recommended by the US Environmental Protection Agency [11] Environmental Protection Agency [11] (USEPA). The method is mainly applied to air pollution, (USEPA). The method is mainly applied to air pollution, water pollution, sediment, etc. [7,12,13] and water pollution, sediment, etc. [7,12,13]. and has been shown to be a powerful model for analyzing has been shown to be a powerful model for analyzing the source of material in the environment [13,14]. the source of material in the environment [13,14]. InIn this this study, study, we we investigated investigated thethe heavyheavy metalmetal concentrations concentrations in in the the sediment sediment of ofLake Lake Taihu. Taihu. TheThe speciﬁc specific objectives objectives were were as as follows: follows: (1)(1) to analyze characteristics characteristics of of the the grain grain size size of ofLake Lake Taihu Taihu sediments;sediments; (2) (2) to to measure measure the the concentrations concentrations ofof heavyheavy metals in in the the sediments; sediments; and and (3) (3) to toanalyze analyze the the sourcessources of of heavy heavy metals metals in in the the sediments sediments of of western western Lake Lake Taihu Taihu and and quantify quantify thethe historicalhistorical changeschanges in thein sources the sources of heavy of heavy metals metals in the in the sediments sediments using using the the PMF PMF model. model.

2. Materials2. Materials and and Methods Methods

2.1.2.1. Study Study Area Area and and Sampling Sampling LakeLake Taihu Taihu Basin Basin is ais key a key area area in easternin eastern China China because because of itsof highits high population. population. Rainfall Rainfall in thein the area, controlledarea, controlled by the Paciﬁc by the monsoon,Pacific monsoon, is moderately is moderat highely (905–1956 high (905–1956 mm/a), mm/a), and precipitation and precipitation in summer in is approximatelysummer is approximately 37% of the total37% annualof the total rainfall. annual The rainfall. average The annual average evaporation annual isevaporation approximately is ◦ 1001approximately mm, and the 1001 mean mm, annual and the temperature mean annual is 16temperatC. Inure this is study, 16 °C. we In selectedthis study, the we western selected part the of Lakewestern Taihu, part shown of Lake in FigureTaihu,1 shown, and collected in Figure a1, 45-cm and collected long sediment a 45-cm columnlong sediment using acolumn non-interfering using a gravitynon-interfering sampler. Thegravity column sampler. was The sliced column into 1was cm slic samplesed into (a 1 totalcm samples of 45 samples) (a total of and 45 samples) freeze-dried and at −40freeze-dried◦C for 48–72 at − h.40 °C for 48–72 h.

FigureFigure 1. 1.Location Location ofof thethe study area and and sampling sampling sites. sites.

Int. J. Environ. Res. Public Health 2018, 15, 1540 3 of 12

2.2. Analysis of Sediment A laser grain size analyzer (Malvern Mastersizer 2000) was used to measure the grain size (a separate sample aliquot of 2 g of sediment was pretreated using 15 mL of 0.6% (NaPO3)6 for 24 h to be dispersed and homogenized). The total heavy metal concentrations and 210Pb in the prepared samples were determined with a HCl-HNO3-HF-HClO4 extraction [15]. Approximately 100 mg of sample was digested with 3 mL of 37% HCL, 1 mL of 65% HNO3, 6 mL of 65% HF, and 0.5 mL of 65% HClO4. There were two steps as follows: in the first step, the temperature was increased to 200 ◦C in 15 min, and in the second step it was maintained at 200 ◦C for approximately 20 min to continue digestion. The sample digestion solutions were evaporated to near dryness after the microwave pretreatment and then dissolved with 65% HNO3, after which 20 mL of deionized water was added to each sample. Prior to the experimental analysis, the solution was stored in 25-mL high-density polyethylene vials. The concentrations of K, As, Mn, Cu, Fe, Zn, Al, Ni, Sr, Ti, Ni and Cr were determined by inductively coupled plasma optical emission spectrometry (ICP-OES), and the concentrations of Cd and Pb were determined with an inductively coupled plasma mass spectrometer (ICP-MS). The accuracy and precision were verified using standard reference materials (soil GBW07405, Chinese geological reference materials). In addition, blank and duplicate samples were also prepared and analyzed (Tables S1 and S2), and the detection limit, quantification limit and recovery rate are shown in Table S4. The activity of 210Pb was determined by γ analysis, using a high-purity germanium detector (EC & GORTEC, Meriden, CT, USA), digital spectrometer and multi-channel analysis system. The sample measurement time was 40,000 s, with a measurement error controlled at the 95% confidence level and measured data errors of less than 5%.

2.3. Data Analysis The Al is the main component of clay minerals, and its content is an important indicator of the coarse grain size. It is an ideal element to normalize the effect of grain size on the content of heavy metals. In this study, Al was used to normalize heavy metal elements [16–18], using the following formula: i i Al CN = C0/CO (1) i i where CN is the normalized value of the i element in the sediment; C0 is the measured value of the i Al element in the sediment; CO is the measured value of the Al in the sediment [19]. PMF multivariate statistical analysis was proposed by Paatero and Tapper [20]. PMF is applied mainly through the least squares method to determine the source of the receptor media and its contribution rate. Its mathematical principle is expressed as follows: deﬁne matrix X, which can be expressed as n × m, where n represents the number of samples and m represents the number of contaminants; decompose matrix X into two factor matrices and one residual matrix, G (n × p), F(p × m) and E (n × m), respectively, where p is the number of pollution sources; the mathematical formula is expressed as X = GF + E; the matrix F represents the factor loading, the matrix G represents the factor contribution, and the matrix E represents the residual matrix, deﬁned as the difference between the actual data and the result of the analysis. The matrix conversion is performed as follows:

= p + Xij ∑k=1 gik fkj eij (2) where Xij represents the content (concentration) of the jth pollutant in the ith sample; gik represents the contribution rate of the kth source to the ith sample; fkj represents the content of the jth pollutant in the source k concentration; eij represents the residual matrix. At the same time, PMF deﬁnes an objective function: e 2 ( ) = n m ( ij ) Q E ∑i=1 ∑j=1 (3) uij Int. J. Environ. Res. Public Health 2018, 15, 1540 4 of 12

where uij denotes the jth pollutant uncertainty of the ith sample. The uncertainty is calculated based mainly on the uncertainty of the sample measurement (MU) and the method detection limit (MDL). WhenInt. the J. Environ. sample Res. contentPublic Health (concentration) 2018, 15, x FOR PEER is REVIEW less than or equal to the MDL, the uncertainty 4 of [21 12 ] u is calculated as: When the sample content (concentration) is less5 than or equal to the MDL, the uncertainty [21] u is u = × MDL (4) calculated as: 6 u = × MDL When the sample content (concentration) is greater than the MDL, the uncertainty(4) u is calculated as: When the sample content (concentration)q is greater than the MDL, the uncertainty u is calculated as: u = (MU × concentration)2 + (MDL)2 (5) u = ( × ) +() (5) PMFPMF analysis analysis was was performed performed using using thethe USEPAUSEPA PMF PMF 5.0 5.0 model model (USEPA, (USEPA, 2014). 2014). The TheIBM IBMSPSS SPSS StatisticsStatistics 21 package 21 package was was used used for for data data description description and correlation correlation analysis. analysis.

3. Results3. Results and and Discussion Discussion

3.1. Sediment3.1. Sediment Properties Properties FigureFigure2 shows 2 shows the variation the variation in particle in particle size withsize with depth depth in the in sedimentary the sedimentary column. column. Most Most samples couldsamples be classiﬁed could be as classified sandy silt, asclayey sandy silt, clayey or silty si sandlt, or silty [22]. sand In the [22]. sediment In the sediment column, column, the silt the fraction was predominant,silt fraction was and predominant, its proportion and variedits proportion between varied 73% between and 82% 73% with and an 82% average with an content average of 76%. The proportioncontent of 76%. of theTheclay proportion fraction of variedthe clay between fraction varied 9% and between 27% with 9% and an average27% with contentan average of 17%. content of 17%. The sand fraction of the sediment samples gradually increased from the bottom to The sand fraction of the sediment samples gradually increased from the bottom to the surface, and the the surface, and the variation range was 0–16%. In general, the particle size in the sediment column variationtended range to increase was 0–16%. from Inthe general,bottom to the the particle surface size and in remained the sediment relatively column stable tended as a whole, to increase fromindicating the bottom that to the surfacesediment and column remained particle relatively diameter stable had not as abeen whole, affected indicating by historical that the floods sediment column[23,24]. particle diameter had not been affected by historical ﬂoods [23,24].

Figure 2. Changes of sediment grain size with depth. Figure 2. Changes of sediment grain size with depth.

3.2. 210Pb Geochronology 3.2. 210Pb Geochronology The profiles of excessive 210Pb activity in the sediment core are shown in Figure 3. The vertical Theprofiles profiles of the of excessive excessive 210Pb210 activityPb activity in the in sediment the sediment core exhibited core are an shown approximately in Figure 3monotonic. The vertical 210210 profilesdecline of the with excessive depth. The PbPb activityex values in in thethe sedimentcore were significantly core exhibited correlated an approximately with sample depths monotonic 2 210 decline(R with= 0.765), depth. which The indicatesPbex thatvalues the slices in the from core both were of significantly the cores cancorrelated reflect historic with change. sample The depths constant initial concentration (CIC) mode of 210Pb requires a better index distribution of the 210Pbex

Int. J. Environ. Res. Public Health 2018, 15, 1540 5 of 12

(R2 = 0.765), which indicates that the slices from both of the cores can reﬂect historic change. Int. J. Environ. Res. Public Health 2018, 15, x FOR PEER REVIEW 5 of 12 The constant initial concentration (CIC) mode of 210Pb requires a better index distribution of the 210 specificPbex speciﬁc activity activity with depth, with depth, so the sostudy the studyarea wi areath a with stable a stabledepositional depositional environment environment tends to tends use to usethe the CIC CIC mode mode to tocalculate calculate the the year year [25,26]. [25,26 In]. this In this study, study, the thedates dates were were calculated calculated based based on a CIC on a of CIC 210 of 210PbPb model model [25,26]. [25,26 ].

210 Figure 3. Vertical profiles of 210Pbex in Lake Taihu. Figure 3. Vertical profiles of Pbex in Lake Taihu. 3.3. Heavy Metal Characteristics 3.3. Heavy Metal Characteristics On the basis of the dating, the description and analysis of the heavy metals in the upper part of theOn sediment the basis column of the (after dating, 1978 the ± description0.7838) and the and lower analysis part of(before the heavy 1978 ± metals 0.7838) in showed the upper that part the of thecontent sediment of heavy column metals (after such 1978 as± Cr,0.7838) Cd, Sr and and the Pb lowerwere obviously part (before higher 1978 in± the0.7838) upper showed part than that in the contentthe lower of heavy part. metalsIn Table such 1, the as Cr,average Cd, Srconcentrations and Pb were of obviously Cr and Cd higher in the in upper the upper part were part than70.6 mg in the − lowerkg−1 part.and 465.1 In Table μg1 ,kg the−1, averagerespectively; concentrations in the lower of Crpart, and they Cd inwere the 51.7 upper mg part kg were−1 and 70.6 136.0 mg μ kgg kg1−1and, 465.1respectively.µg kg−1, respectively;The coefficients in theof variation lower part, for theythe ot wereher elements 51.7 mg were kg −1 notand significantly 136.0 µg kg large,−1, respectively. and the Thecoefficients coefficients of ofvariation variation forfor various the other metals elements after th weree 1980s not were significantly slightly larg large,er than and those the coefficients before the of variation1980s. Correlation for various analysis metals after (CA) the was 1980s performed. were slightly larger than those before the 1980s. Correlation analysis (CA) was performed. Table 1. Heavy metals in sedimentary columns around 1978 (± 0.7838).

UpperTable Part 1. Heavy (After metals 1978 ± in0.7838) sedimentary columns around Lower Part 1978 (Before (± 0.7838). 1978 ± 0.7838) Minimum Maximum Mean CV Minimum Maximum Mean CV Cr (mg kg−1)Upper Part51.3 (After 197884.0± 0.7838)70.6 14.9 43.1 Lower Part67.8 (Before 197851.9± 0.7838) 10.8 Cu (mg kg−1) Minimum13.5 Maximum27.8 Mean19.7 24.2 CV Minimum17.6 Maximum36.2 Mean23.9 18.1 CV −1 CrFe (mg (mg kg kg−1)) 22,284.051.3 27,289.9 84.0 70.624,972.3 14.95.3 22,747.9 43.1 28,965.8 67.8 25,869.8 51.9 6.3 10.8 CuMg (mg (mg kg kg−1−)1) 5286.613.5 27.87783.3 19.76407.2 24.211.0 5996.2 17.6 9648.3 36.2 23.98216.9 11.2 18.1 FeMn (mg (mg kg kg−1−)1) 22,284.0504.2 27,289.91231.0 24,972.3810.7 27.1 5.3 22,747.9487.6 28,965.81044.2 25,869.8811.9 20.4 6.3 MgNi (mg (mg kg kg−−11)) 5286.630.6 7783.392.9 6407.242.5 11.039.4 5996.234.4 9648.347.1 8216.941.1 8.8 11.2 −1 MnK (mg(mg kgkg−1)) 15,576.8504.2 1231.018,928.6 810.717,341.9 27.15.9 15,240.7 487.6 21,698.1 1044.2 18,935.6 811.9 9.7 20.4 −1 NiSr (mg (mg kg kg−1)) 131.430.6 92.9268.5 42.5198.6 39.419.5 67.8 34.4 47.1255.9 41.1123.8 38.6 8.8 K (mg kg−1) 15,576.8 18,928.6 17,341.9 5.9 15,240.7 21,698.1 18,935.6 9.7 Ti (mg kg−1) 3984.0 5963.2 4492.8 9.8 3894.2 5128.1 4374.5 7.0 Sr (mg kg−1) 131.4 268.5 198.6 19.5 67.8 255.9 123.8 38.6 Zn (mg kg−1) 89.2 153.8 114.7 17.7 94.4 129.2 110.9 9.0 Ti (mg kg−1) 3984.0 5963.2 4492.8 9.8 3894.2 5128.1 4374.5 7.0 −1 ZnCd (mg (mg kg kg−1)) 0.12589.2 153.80.836 114.70.465 17.761.8 0.082 94.4 129.20.244 110.90.136 28.5 9.0 CdPb (mg (mg kg kg−1−1)) 0.12517.4 0.83629.4 0.46523.6 61.817.2 18.1 0.082 0.24423.7 0.13620.8 7.6 28.5 PbAl (mg (mg kg kg−1−1)) 17.422.0 29.434.9 23.627.0 17.211.8 21.5 18.1 23.743.6 20.835.0 15.4 7.6 AlAs (mg (mg kg kg−1−)1) 22.016.3 34.933.6 27.021.5 11.818.1 14.9 21.5 43.643.5 35.025.5 27.8 15.4 −1 As (mg kg ) 16.3 33.6CV: coefficient 21.5 of variation 18.1 in 14.9%. 43.5 25.5 27.8 CV: coefﬁcient of variation in %. Table 2 lists the correlation coefficients between the deposition parameters (clay, silt, sand) and metals (Cr, Cu, Fe, Mg, Mn, Ni, K, Sr, Ti, Zn, Cd, Pb, Al and As). The CA of the data showed that

Int. J. Environ. Res. Public Health 2018, 15, 1540 6 of 12 Int. J. Environ. Res. Public Health 2018, 15, x FOR PEER REVIEW 6 of 12

heavyTable metals2 lists such the correlationas Cu, Fe, coefficientsMg and Al between were po thesitively deposition correlated parameters with the (clay, clay silt, fraction sand) andand metalsnegatively (Cr, Cu,correlated Fe, Mg, with Mn, Ni,the K, sand Sr, Ti, fraction Zn, Cd, (significant Pb, Al and As).at p The< 0.01), CA ofreflecting the data the showed adsorption that heavy and metalsenrichment such asof heavy Cu, Fe, metals Mg and on Alfine were clay positively [27,28]. However, correlated Cr, with Cd, the Pd, clay Sr fractionand other and heavy negatively metals correlatedwere negatively with the correlated sand fraction with fine (significant clay (significant at p < 0.01), at p reflecting < 0.01), and the there adsorption was no and correlation enrichment for ofTi heavyand Fe. metals With onthe finedecrease clay [27of ,fine28]. grain However, content, Cr, Cd,these Pd, heavy Sr and metal other contents heavy increased, metals were reflecting negatively the correlatedimpact of human with fine activities clay (significant on the import at p < of 0.01), heavy and metals there into was lake no correlation sedimentsfor [29,30]. Ti and Fe. With the decrease of fine grain content, these heavy metal contents increased, reflecting the impact of human activities3.4. Source on Identification the import ofby heavyPMF metals into lake sediments [29,30]. We used PMF to analyze the sources of heavy metals in the sediments of the western Lake 3.4. Source Identification by PMF Taihu. The PMF model was simulated with three to six factors, and the starting point of each run was Wedifferent. used PMFWe selected to analyze a random the sources seed pattern of heavy with metals 20 random in the sedimentsorigins and of checked the western for three Lake to Taihu.six factors. The PMF Finally, model the was PMF simulated analysis with identified three to three six factors, appropriate and the factors starting in pointthe sediments of each run of was the different.western Lake We selectedTaihu (the a random uncertainty seed patternof the factor with 20 changed random by origins 10%, andand checkedits analysis for threeresults to sixare factors.shown Finally,in Table the S3; PMF the analysisanalysis identifiedof the model three with appropriate four to factorssix factors in the is sedimentsshown in ofTable the westernS5). The Laketentative Taihu identification (the uncertainty of the of factor the factor profiles changed was bybased 10%, on and the its characteristic analysis results geochemical are shown signature in Table S3;from the different analysis sources. of the model with four to six factors is shown in Table S5). The tentative identification of theThe factor source profiles composition was based of on the the three-factor characteristic solution geochemical is summarized signature in from Figure different 4, with sources. factor 1 accountingThe source for 29% composition of the total of heavy the three-factor metals in th solutione sediments is summarized of western Lake in Figure Taihu,4, withdominated factor by 1 accountingAs, K and forCu. 29% According of the total to previous heavy metals studies, in the fertilization sediments can of westerneffectively Lake increase Taihu, dominatedthe concentration by As, Kof and K in Cu. the According soil [31,32]. to Elements previous studies,such as fertilizationAs and Cu are can commonly effectively used increase to make the concentration pesticides [33]. of KThe in themain soil agricultural [31,32]. Elements practice such in the as study As and area Cu is are the commonly growing of used rice to and make tea. pesticides To increase [33 output,]. The main local agriculturalfarmers often practice use pesticides in the study and areachemical is the fertilizers growing ofsuch rice as and sodium tea. To methylarsonate increase output, and local potassium farmers often[34]. Therefore, use pesticides factor and 1 indicated chemical agricultural fertilizers suchsources. as sodiumFactor 2 methylarsonatehad a degree of interpretation and potassium of [46%34]. Therefore,and was dominated factor 1 indicated by Cd, Cr, agricultural Pb, Sr and sources. Mn. The Factor research 2 had area a degree mainly of has interpretation metallurgical of industry, 46% and wasbuilding dominated materials by Cd, industry, Cr, Pb, Srceramic and Mn. industry The research and soarea on [34,35]. mainly hasThese metallurgical industries can industry, produce building large materialsquantities industry, of pollutants ceramic containing industry andCd, soCr, on Pb [34 an,35d]. Mn These in industriesthe process can of produceproduction large [36–38]; quantities thus, of pollutantsfactor 2 represented containing industrial Cd, Cr, Pb sources. and Mn inFactor the process 3 accounted of production for 25% of [36 the–38 total]; thus, heavy factor metals 2 represented and was industrialmainly composed sources. of Factor Ti, Al 3 accountedand Ni, while for 25%Mg also of the showed total heavy some metalsloadings and on was this mainly factor. composedTi, Al and Ni of Ti,mainly Al and come Ni, whilefrom the Mg erosion also showed of the some riverbed loadings and dissolution on this factor. of Ti,minerals; Al and thus, Ni mainly factor come 3 represented from the erosiongeogenic of sources the riverbed [39]. and dissolution of minerals; thus, factor 3 represented geogenic sources [39].

FigureFigure 4.4. SourceSource proﬁlesprofiles obtainedobtained fromfrom thethe positivepositive matrixmatrix factorizationfactorization (PMF)(PMF) model.model. ((aa)) FactorFactor 11 agriculturalagricultural sources;sources; ((bb)) FactorFactor 22 industrialindustrial sources; sources; ( c(c)) Factor Factor 3 3 geogenic geogenic sources. sources.

Int. J. Environ. Res. Public Health 2018, 15, 1540 7 of 12

Table 2. Pearson’s correlation of metal concentrations for Lake Taihu sedimentary column.

Cr Cu Fe Mg Mn Ni K Sr Ti Zn Cd Pb Al As Sand Silt Clay Cr 1 Cu −0.057 1 Fe −0.166 0.286 1 Mg −0.487 ** 0.691 ** 0.594 ** 1 Mn 0.264 0.669 ** 0.282 0.490 ** 1 Ni 0.343 * 0.477 ** 0.180 0.224 0.435 ** 1 K −0.208 0.629 ** 0.599 ** 0.863 ** 0.570 ** 0.277 1 Sr 0.557 ** −0.213 0.087 −0.227 0.387 ** 0.194 0.073 1 Ti 0.188 −0.254 −0.013 −0.257 −0.452 ** 0.006 −0.153 −0.163 1 Zn 0.437 ** 0.709 ** 0.355 * 0.364 * 0.814 ** 0.600 ** 0.475 ** 0.357 * −0.163 1 Cd 0.801 ** 0.140 −0.116 −0.373 * 0.457 ** 0.472 ** −0.191 0.562 ** −0.083 0.659 ** 1 Pb 0.652 ** 0.343 * 0.083 −0.092 0.570 ** 0.487 ** 0.077 0.335 * −0.092 0.772 ** 0.861 ** 1 Al −0.526 ** 0.349 * −0.127 0.451 ** 0.060 −0.030 0.254 −0.597 ** −0.104 −0.176 −0.504 ** −0.278 1 As −0.247 0.191 0.477 ** 0.293 −0.011 0.070 0.216 −0.127 −0.039 0.179 −0.065 0.023 −0.043 1 Sand 0.599 ** −0.507 ** −0.340 * −0.766 ** −0.030 −0.058 −0.556 ** 0.652 ** 0.049 0.043 0.646 ** 0.361 * −0.639 ** −0.246 1 Silt 0.260 −0.468 ** −0.186 −0.410 ** −0.181 −0.081 −0.267 0.433 ** 0.344 * −0.122 0.181 −0.011 −0.311 * −0.142 0.370 * 1 Clay −0.573 ** 0.588 ** 0.328 * 0.780 ** 0.106 0.062 0.562 ** −0.677 ** −0.168 0.013 −0.588 ** −0.282 0.626 ** 0.234 −0.943 ** −0.648 ** 1 * Signiﬁcantly correlated at the 0.05 level (2-tailed); ** Signiﬁcantly correlated at the 0.01 level (2-tailed). Int. J. Environ. Res. Public Health 2018, 15, x FOR PEER REVIEW 8 of 12

Int. J. Environ.The PMF Res. model Public Health was 2018successfully, 15, 1540 applied to the identification of sediment pollution sources 8 of in 12 western Lake Taihu. Figure 5 shows that before the 1980s, the contributions of industry, agriculture Int. J. Environ. Res. Public Health 2018, 15, x FOR PEER REVIEW 8 of 12 and geological processes to the origin of heavy metals were relatively stable, with a significant The PMF model was successfully applied to the identiﬁcation of sediment pollution sources in contributionThe PMFof geological model was processessuccessfully and app liedagricu to thelture, identification and a certain of sediment degree pollution of contribution sources in of western Lake Taihu. Figure5 shows that before the 1980s, the contributions of industry, agriculture and industry.western After Lake the Taihu. 1980s, Figure the 5 showscontribution that before of th gee 1980s,ological the contributionsprocesses to of sediment industry, agricultureheavy metals geological processes to the origin of heavy metals were relatively stable, with a signiﬁcant contribution decreasedand geological rapidly, whileprocesses that to of the industry origin ofrose heavy rapidly metals and were reached relatively stability stable, in withapproximately a significant 2003. of geological processes and agriculture, and a certain degree of contribution of industry. After the Beforecontribution the reform ofand geological opening processesup of China, and theagricu contlture,ributions and aof certain industry, degree agriculture of contribution and geological of 1980s, the contribution of geological processes to sediment heavy metals decreased rapidly, while processesindustry. to sediment After the heavy 1980s, metals the contribution were 20.26 of± 15.61%geological (n =processes 30), 35.67 to ± 11.70%sediment (n heavy = 30) andmetals 44.07 ± that ofdecreased industry rapidly, rose rapidlywhile that and of reachedindustry rose stability rapidly in and approximately reached stability 2003. in approximately Before the reform 2003. and 11.84% (n = 30), respectively. After the reform and opening up, their contributions to heavy metals openingBefore up the of China,reform and the opening contributions up of China, of industry, the cont agricultureributions of industry, and geological agriculture processes and geological to sediment were 56.99 ± 15.73% (n = 15), 31.22 ± 14.31% (n = 15) and 11.79 ± 9.83% (n = 15), respectively (Figure heavyprocesses metals wereto sediment 20.26 heavy± 15.61% metals (n were= 30), 20.26 35.67 ± 15.61%± 11.70% (n = 30), (n 35.67= 30) ± and11.70% 44.07 (n =± 30)11.84% and 44.07 (n ±= 30), 6a,b). In recent years, the proportion of heavy metals in sediments originating from industrial respectively.11.84% ( Aftern = 30), the respectively. reform and After opening the reform up, their and contributions opening up, their to heavy contributions metals wereto heavy 56.99 metals± 15.73% sources has declined. (n = 15),were 31.22 56.99± ± 14.31%15.73% ( n = 15), 15) 31.22 and 11.79± 14.31%± 9.83% (n = 15) (n and= 15), 11.79 respectively ± 9.83% (n (Figure= 15), respectively6a,b). In recent (Figure years, 6a,b). In recent years, the proportion of heavy metals in sediments originating from industrial the proportion of heavy metals in sediments originating from industrial sources has declined. sources has declined.

Figure 5. Variation in the composition of pollution sources of sediment heavy metals with time based Figure 5. Variation in the composition of pollution sources of sediment heavy metals with time based Figureon 5. the Variation positive matrixin the compositionfactorization (PMF) of pollution model. sources of sediment heavy metals with time based on the positive matrix factorization (PMF) model.

Figure 6. Sources of sediment heavy metals before and after the Chinese economic reform. (a) before Chinese economic reform; (b) after Chinese economic reform.

Figure 6. Sources of sediment heavy metals before and after the Chinese economic reform. ( a) before Chinese economic reform; (b)) afterafter ChineseChinese economiceconomic reform.reform.

Int. J. Environ. Res. Public Health 2018, 15, 1540 9 of 12 Int. J. Environ. Res. Public Health 2018, 15, x FOR PEER REVIEW 9 of 12

This trend is in good agreement with historic historicalal policies and economic development development in in China. Figure7 7 shows shows thethe relationshiprelationship betweenbetween thethe GDPGDP ofof JiangsuJiangsu ProvinceProvince (Jiangsu(Jiangsu StatisticalStatistical Yearbook,Yearbook, 1975–2015) [[40]40] andand thethe ratioratio ofof industrialindustrial sourcessources ofof heavyheavy metalsmetals basedbased onon thethe PMFPMF model.model. Before 2003, there was a signiﬁcantsignificant positive correlation between the GDP of Jiangsu Province and ratio of industrial sources of heavy metals. With With the rapid increase in GDP, thethe proportionproportion of heavy metals produced byby industrial industrial activities activities rose rose quickly. quickly. However, However, after 2003,after the2003, two the variables two variables showed oppositeshowed trends,opposite and trends, economic and developmenteconomic development did not increase did not the increase proportion the ofproportion heavy metals of heavy produced metals by industrialproduced by activities. industrial This activities. phenomenon This phenomenon has also been has reported also been in thereported records in ofthe lakes records and of marine lakes sedimentsand marine in sediments other regions in other [41– 43regions]. For [41–43]. example, For Li [ex11ample,] and Wan Li [11] [44 ]and reported Wan [44] that reported the pollutants that the in Chaohupollutants and in GonghaiChaohu reachedand Gonghai their maximum reached their level maximum in the late level 20th centuryin the late and 20th began century to decline and began in the earlyto decline 21st century.in the early These 21st research century. results These are research in agreement results with are oursin agreement and indicate with that ours the and PMF indicate model inthat the the present PMF model study isin highlythe present reliable. study is highly reliable.

Figure 7. Correlation analysis of the proportion of industrialindustrial sources of heavy metals based on the PMF model and the GDP of the study area.

Before the 1980s, China’s scientific, technological and economic conditions were limited, with Before the 1980s, China’s scientiﬁc, technological and economic conditions were limited, with agriculture as the mainstay. After the reform and opening up in 1978, China’s economy developed agriculture as the mainstay. After the reform and opening up in 1978, China’s economy developed rapidly, and the output value of industry and agriculture soared [40]. At the same time, this rapidly, and the output value of industry and agriculture soared [40]. At the same time, this development brought about serious environmental pollution problems. The proportion of heavy development brought about serious environmental pollution problems. The proportion of heavy metal metal pollution caused by industrial production was highest in the early 21st century. With the pollution caused by industrial production was highest in the early 21st century. With the promulgation promulgation and implementation of a series of environmental laws, China’s environment has and implementation of a series of environmental laws, China’s environment has improved greatly improved greatly after the 2000s. The proportion of heavy metal pollution brought about by after the 2000s. The proportion of heavy metal pollution brought about by industrial production has industrial production has been curbed and is showing a downward trend, as shown in Figure 5. been curbed and is showing a downward trend, as shown in Figure5.

4. Conclusions With thethe development development of of the the Chinese Chinese economy, economy, heavy heavy metal metal pollution pollution became became increasingly increasingly serious. Inserious. general, In thisgeneral, period this can period be divided can be into divided two stages. into Beforetwo stages. 1978 ±Before0.7838, 1978 the ± heavy 0.7838, metals the heavy in the studymetals area in the were study mainly area were of geogenic mainly andof geogenic agricultural and agricultural origins. After origins. 1978 ± After0.7838, 1978 rapid ± 0.7838, industrial rapid developmentindustrial development occurred, andoccurred, the heavy and the metals heavy in metals the sediments in the sediments originated originated mainly from mainly industry from andindustry agriculture and agriculture and were and less were affected less by affected geological by geological processes. processes. This result This is in result good agreementis in good withagreement the GDP with of the the GDP study of the area. study The area. PMF The model PMF and model quantitative and quantitative analysis analysis of sediment of sediment for heavy for metalheavy sourcesmetal sources from the from bottom the depths bottom to thedepths top provideto the antop efﬁcient provide and an convenient efficient and combination convenient of analyticalcombination methods. of analytical methods.

Int. J. Environ. Res. Public Health 2018, 15, 1540 10 of 12

Supplementary Materials: The following are available online at http://www.mdpi.com/1660-4601/15/7/1540/ s1, Table S1: Operating parameters of ICP-MS (ELAN 9000, Perkin-Elmer SCIEX) for the determination of element concentrations, Table S2: Operating parameters of ICP-OES (Optima 5300DV, Perkin-Elmer SCIEX) for the determination of elemental concentrations, Table S3: Uncertainty analysis of the model, Table S4: Detection limit, Quantification limit and Recovery rate for the heavy metals, Table S5: Factor Profiles (% of factor total). Author Contributions: Conceptualization, Y.L.; Data curation, Y.L. and L.M.; Formal analysis, L.M.; Funding acquisition, L.M. and S.Z.; Investigation, Y.L., L.M. and Z.J.; Methodology, Y.L. and S.W.; Project administration, S.Z.; Resources, Z.J.; Software, Y.L. and J.W.; Supervision, S.Z. and C.W.; Validation, J.W.; Visualization, B.L.; and Writing—original draft, Y.L. and L.M. Funding: The National Natural Science Foundation of China: 41771243; the National Key Research and Development Program of China: 2017YFD0800305; the Scientific Research Foundation of the Graduate School of Nanjing University: 2016CL04; and the Special Fund for Research in the Public Interest of Ministry of Land and Resources: 201511001-03; Project funded by China Postdoctoral Science Foundation: 2018T110478, 2017M621699. Conflicts of Interest: The authors declare no conflicts of interest.

References

1. Khan, M.U.; Malik, R.N.; Muhammad, S. Human health risk from heavy metal via food crops consumption with wastewater irrigation practices in Pakistan. Chemosphere 2013, 93, 2230–2238. [CrossRef][PubMed] 2. Akinyele, I.O.; Shokunbi, O.S. Comparative analysis of dry ashing and wet digestion methods for the determination of trace and heavy metals in food samples. Food Chem. 2015, 173, 682–684. [CrossRef] [PubMed] 3. Mendoza-Carranza, M.; Sepúlveda-Lozada, A.; Dias-Ferreira, C.; Geissen, V.Distribution and bioconcentration of heavy metals in a tropical aquatic food web: A case study of a tropical estuarine lagoon in SE Mexico. Environ. Pollut. 2015, 210, 155–165. [CrossRef] [PubMed] 4. Shahbaz, M.; Hashmi, M.Z.; Malik, R.N.; Yasmin, A. Relationship between heavy metals concentrations in egret species, their environment and food chain differences from two headworks of Pakistan. Chemosphere 2013, 93, 274–282. [CrossRef][PubMed] 5. Sun, P.; Xie, B.; Song, Y.; Yang, H.; Wang, Y. Historical Trends of Polycyclic Aromatic Hydrocarbons in the Reservoir Sediment of the Dianchi Watershed, Southwest China. Bull. Environ. Contam. Toxicol. 2017, 99, 117–124. [CrossRef][PubMed] 6. Komárek, M.; Ettler, V.; Chrastný, V.; Mihaljevic, M. Lead isotopes in environmental sciences: A review. Environ. Int. 2008, 34, 562–577. [CrossRef][PubMed] 7. Guan, Q.; Wang, F.; Xu, C.; Pan, N.; Lin, J.; Zhao, R.; Yang, Y.; Luo, H. Source apportionment of heavy metals in agricultural soil based on PMF: A case study in Hexi Corridor, northwest China. Chemosphere 2017, 193, 189–197. [CrossRef][PubMed] 8. Men, C.; Liu, R.; Xu, F.; Wang, Q.; Guo, L.; Shen, Z. Pollution characteristics, risk assessment, and source apportionment of heavy metals in road dust in Beijing, China. Sci. Total Environ. 2018, 612, 138–147. [CrossRef][PubMed] 9. Fernández, S.; Cotos-Yáñez, T.; Roca-Pardiñas, J.; Ordóñez, C. Geographically weighted principal components analysis to assess diffuse pollution sources of soil heavy metal: Application to rough mountain areas in Northwest Spain. Geoderma 2016, 311, 120–129. [CrossRef] 10. Ha, H.; Olson, J.R.; Bian, L.; Rogerson, P.A. Analysis of Heavy Metal Sources in Soil Using Kriging Interpolation on Principal Components. Environ. Sci. Technol. 2014, 48, 4999–5007. [CrossRef][PubMed] 11. Li, H.; Cheng, Y.; Liu, Y.; Gao, J. Environmental signiﬁcange of heavy metals in the core sediments of the Yalu River estuary. Acta Sci. Circumstantiae 2017, 37, 2296–2306.

12. Kim, E.; Larson, T.V.; Hopke, P.K.; Slaughter, C.; Sheppard, L.E.; Claiborn, C. Source identiﬁcation of PM2.5 in an arid Northwest U.S. City by positive matrix factorization. Atmos. Res. 2003, 66, 291–305. [CrossRef] 13. Wang, C.; Wu, S.; Zhou, S.; Wang, H.; Li, B.; Chen, H.; Yu, Y.; Shi, Y. Polycyclic aromatic hydrocarbons in soils from urban to rural areas in Nanjing: Concentration, source, spatial distribution, and potential human health risk. Sci. Total Environ. 2015, 527, 375–383. [CrossRef][PubMed] 14. Hu, X.; Zhang, Y.; Luo, J.; Wang, T.; Lian, H.; Ding, Z. Bioaccessibility and health risk of arsenic, mercury and other metals in urban street dusts from a mega-city, Nanjing, China. Environ. Pollut. 2011, 159, 1215–1221. [CrossRef][PubMed] Int. J. Environ. Res. Public Health 2018, 15, 1540 11 of 12

15. Qinghua, L.I.; Wan, S.; Jun, H.E.; Anchun, L.I. Human impact on the intertidal environment in Beibu Gulf over the last 200 years. Mar. Geol. Quat. Geol. 2014, 34, 57–64. 16. Soto-Jiménez, M.F.; Páez-Osuna, F. Distribution and Normalization of Heavy Metal Concentrations in Mangrove and Lagoonal Sediments from Mazatlán Harbor (SE Gulf of California). Estuar. Coast. Shelf Sci. 2001, 53, 259–274. [CrossRef] 17. Balachandran, K.K.; Raj, C.M.L.; Nair, M.; Joseph, T.; Sheeba, P.; Venugopal, P. Heavy metal accumulation in a flow restricted, tropical estuary. Estuar. Coast. Shelf Sci. 2005, 65, 361–370. [CrossRef] 18. Paatero, P.; Tapper, U. Positive matrix factorization: A non-negative factor model with optimal utilization of error estimates of data values. Environmetrics 1994, 5, 111–126. [CrossRef] 19. Yixing Municipal People’s Government. Yixing Yearbook; China Local Records Publishing: Yixing, China, 1989–2017. (In Chinese) 20. Polissar, A.V.; Hopke, P.K.; Paatero, P.; Malm, W.C.; Sisler, J.F. Atmospheric aerosol over Alaska: 2. Elemental composition and sources. J. Geophys. Res. Atmos. 1998, 103, 19045–19057. [CrossRef] 21. Shepard, F.P. Nomenclature Based on Sand-silt-clay Ratios. J. Sediment. Res. 1954, 24, 151–158. 22. He, Q.; Walling, D.E. An investigation of the spatial variability of the grain size composition of floodplain sediments. Hydrol. Process. 2015, 12, 1079–1094. [CrossRef] 23. Pan, B.; Pang, H.; Zhang, D.; Guan, Q.; Wang, L.; Li, F.; Guan, W.; Cai, A.; Sun, X. Sediment grain-size characteristics and its source implication in the Ningxia–Inner Mongolia sections on the upper reaches of the Yellow River. Geomorphology 2015, 246, 255–262. [CrossRef] 24. Oldfield, F.; Appleby, P.G.; Battarbee, R.W. Alternative 210Pb dating: Results from the New Guinea Highlands and Lough Erne. Nature 1978, 271, 339–342. [CrossRef] 25. Li, Y.; Zhou, S.; Zhu, Q.; Li, B.; Wang, J.; Wang, C.; Chen, L.; Wu, S. One-century sedimentary record of heavy metal pollution in western Taihu Lake, China. Environ. Pollut. 2018, 240, 709–716. [CrossRef][PubMed] 26. Huang, S.; Wan, Z.; Zhang, C.; Wan, L. Effect of grain size of sediment on the adsorption of heavy metals. J. Hydraul. Eng. 1994, 10. 27. Han, X.; Lu, X.; Zhang, Q.; Wuyuntana; Hai, Q.; Pan, H. Grain-size distribution and contamination characteristics of heavy metal in street dust of Baotou, China. Environ. Earth Sci. 2016, 75, 468. [CrossRef] 28. Li, Z.; Ma, Z.; van der Kuijp, T.J.; Yuan, Z.; Huang, L. A review of soil heavy metal pollution from mines in China: Pollution and health risk assessment. Sci. Total Environ. 2014, 468, 843–853. [CrossRef][PubMed] 29. Zeng, H.; Wu, J.; Liu, W. Two-century sedimentary record of heavy metal pollution from Lake Sayram: A deep mountain lake in central Tianshan, China. Quat. Int. 2014, 321, 125–131. [CrossRef] 30. Dong, H.Z.; Kong, X.Q.; Li, W.J.; Tang, W.; Zhang, D.M. Effects of plant density and nitrogen and potassium fertilization on cotton yield and uptake of major nutrients in two fields with varying fertility. Field Crops Res. 2010, 119, 106–113. [CrossRef] 31. Niu, J.; Zhang, W.; Ru, S.; Chen, X.; Kai, X.; Zhang, X.; Assaraf, M.; Imas, P.; Magen, H.; Zhang, F. Effects of potassium fertilization on winter wheat under different production practices in the North China Plain. Field Crops Res. 2013, 140, 69–76. [CrossRef] 32. Quazi, S.; Sarkar, D.; Datta, R. Effect of soil aging on arsenic fractionation and bioaccessibility in inorganic arsenical pesticide contaminated soils. Appl. Geochem. 2010, 25, 1422–1430. [CrossRef] 33. Han, Y.; Jin, Z.; Cao, J.; Posmentier, E.S.; An, Z. Atmospheric Cu and Pb Deposition and Transport in Lake Sediments in a Remote Mountain Area, Northern China. Water Air Soil Pollut. 2007, 179, 167–181. [CrossRef] 34. Lin, C.; Ma, R.; Xiong, J. Can the watershed non-point phosphorus pollution be interpreted by critical soil properties—A new insight of different soil P states. Sci. Total Environ. 2018, 628, 870–881. [CrossRef] [PubMed] 35. Lin, C.; Wu, Z.; Ma, R.; Su, Z. Detection of sensitive soil properties related to non-point phosphorus pollution by integrated models of SEDD and PLOAD. Ecol. Indic. 2016, 60, 483–494. [CrossRef] 36. Ma, X.; Zuo, H.; Tian, M.; Zhang, L.; Meng, J.; Zhou, X.; Min, N.; Chang, X.; Liu, Y. Assessment of heavy metals contamination in sediments from three adjacent regions of the Yellow River using metal chemical fractions and multivariate analysis techniques. Chemosphere 2016, 144, 264–272. [CrossRef][PubMed] 37. Zhang, Y.; Su, Y.; Liu, Z.; Sun, K.; Kong, L.; Yu, J.; Jin, M. Sedimentary lipid biomarker record of human-induced environmental change during the past century in Lake Changdang, Lake Taihu basin, Eastern China. Sci. Total Environ. 2018, 613, 907–918. [CrossRef][PubMed] Int. J. Environ. Res. Public Health 2018, 15, 1540 12 of 12

38. Wang, S.; Wang, Y.; Zhang, R.; Wang, W.; Xu, D.; Jing, G.; Li, P.; Yu, K. Historical levels of heavy metals reconstructed from sedimentary record in the Hejiang River, located in a typical mining region of Southern China. Sci. Total Environ. 2015, 532, 645–654. [CrossRef][PubMed] 39. Li, C.; Huo, S.; Yu, Z.; Wei, G.; Xi, B.; He, Z.; Zeng, X. Historical records of polycyclic aromatic hydrocarbon deposition in a shallow eutrophic lake: Impacts of sources and sedimentological conditions. Acta Sci. Circumstantiae 2016, 41, 261–269. [CrossRef][PubMed] 40. National Bureau of Statistics of the People’s Republic of China. The 2016 Jiangsu Statistical Yearbook; China Statistics Press: Beijing, China, 1975–2016. (In Chinese) 41. Guo, J.Y.; Wu, F.C.; Liao, H.Q.; Zhao, X.L.; Li, W.; Wang, J.; Wang, L.F.; Giesy, J.P. Sedimentary record of polycyclic aromatic hydrocarbons and DDTs in Dianchi Lake, an urban lake in Southwest China. Environ. Sci. Pollut. Res. Int. 2013, 20, 5471–5480. [CrossRef][PubMed] 42. Liu, L.Y.; Wang, J.Z.; Wei, G.L.; Guan, Y.F.; Wong, C.S.; Zeng, E.Y. Sediment records of polycyclic aromatic hydrocarbons (PAHs) in the continental shelf of China: Implications for evolving anthropogenic impacts. Environ. Sci. Technol. 2012, 46, 6497–6504. [CrossRef][PubMed] 43. Fang, C. The urbanization and urban development in China after the reform and opening-up. Econ. Geogr. 2009, 29, 19–25. 44. Wan, D.; Song, L.; Yang, J.; Jin, Z.; Zhan, C.; Mao, X.; Liu, D.; Shao, Y. Increasing heavy metals in the background atmosphere of central North China since the 1980s: Evidence from a 200-year lake sediment record. Atmos. Environ. 2016, 138, 183–190. [CrossRef]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).