Development of a Marine Sediment Pollution Index

Development of a Marine Sediment Pollution Index

Environmental Pollution 113 (2001) 281±291 www.elsevier.com/locate/envpol Development of a Marine Sediment Pollution Index P.K.S. Shin *, W.K.C. Lam...

315KB Sizes 0 Downloads 230 Views

Environmental Pollution 113 (2001) 281±291

www.elsevier.com/locate/envpol

Development of a Marine Sediment Pollution Index P.K.S. Shin *, W.K.C. Lam Department of Biology and Chemistry, City University of Hong Kong, Tat Chee Avenue, Kowloon, Hong Kong, People's Republic of China Received 11 March 2000; accepted 7 August 2000

``Capsule'': Principal component analysis was used to identify parameters used to develop an index to determine marine sediment quality. Abstract To facilitate translation of the state of marine sediment quality for public information, a pollution index was developed from the results of a routine monitoring program. Principal component analysis (PCA) of 24 variables at 64 monitoring stations from 1987 to 1997 was carried out to identify the most important parameters that were applied in the index formulation. Of the 24 variables, six (Cr, Cu, Ni, chemical oxygen demand, Zn, Mn) were selected on the basis of their high PCA loadings. The derived Marine Sediment Pollution Index, rating from 0 to 100, re¯ected the general trend in the monitoring areas. The index was also found to have a signi®cant negative correlation (P<0.05) with the benthic species diversity and toxicity of the sediment, indicating its usefulness in re¯ecting marine sediment quality. The application of PCA to identify important variables from a monitoring program would reduce sampling resources, as parameters that did not show signi®cant spatial or temporal variations could be analyzed in a lesser frequency than those that were identi®ed to be more important from the results of PCA. # 2001 Elsevier Science Ltd. All rights reserved. Keywords: Marine sediment; Principal component analysis; Chemistry; Species diversity; Toxicity

1. Introduction Environmental monitoring usually generates a large set of data that may be dicult to analyze and interpret due to the complex inter-relationships between the measured variables. Such data are also often dicult to translate into simple terms so that an interested member of the public can understand the meaning of the monitoring results. To this end, the use of environmental indices can play an important role in the translation process of aggregating complex technical data into understandable forms (Train, 1972). Ott (1978) further identi®ed six possible applications of environmental indices: (1) assisting environmental managers in prioritization of pollution prevention resources; (2) ranking of environmentally degraded sites; (3) assessment of compliance with environmental standards; (4) determination of temporal changes of environmental conditions; (5)

* Corresponding author. Tel.: +852-27887720; fax: 27887406. E-mail address: [email protected] (P.K.S. Shin).

+852-

public communication; and (6) evaluation of monitoring programs. In the aquatic environment, many environmental indices have been developed for water quality interpretation in rivers or lakes (Horton, 1965; Prati et al., 1971; Yu and Fogel, 1978; Miller et al., 1986; Tyson and House, 1988; Oberdor€ and Hughes, 1992; Dojlido et al., 1994; Lobinske et al., 1997). Read et al. (1982) also developed a similar water quality index for the marine environment. Some of these indices included physicochemical parameters in their formulations whereas others involved biological information. However, there are few sediment quality indices published in the literature for classi®cation of sediment conditions in freshwater (Bergstedt and Bergersen, 1997) and estuarine environments (Zou et al., 1988). The purpose of this study is to develop an index for identi®cation of sediment quality in marine coastal waters from routine physico-chemical monitoring data, and to correlate the index with biological and toxicity criteria. We also attempt to demonstrate the usefulness of this approach in re®ning marine sediment monitoring programs.

0269-7491/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved. PII: S0269-7491(00)00192-5

282

P.K.S. Shin, W.K.C. Lam / Environmental Pollution 113 (2001) 281±291

2. Materials and methods 2.1. Study area The Environmental Protection Department of Hong Kong has been conducting a marine sediment monitoring program since 1986. The program covers 64 stations in 10 water control zones (Fig. 1). Grab sediment samples of the top 10 cm of the seabed are collected on a half-yearly basis. Sediment at the surface 2 cm is then sub-sampled for the determination of 24 variables: electrochemical potential, total volatile solids, total solids, chemical oxygen demand (COD), total organic carbon, total Kjeldahl nitrogen, ammoniacal nitrogen, total phosphorus, sul®de, cyanide, metals (As, Al, B, Cd, Cr, Cu, Fe, Hg, Mn, Ni, Pb, Zn), polynuclear aromatic hydrocarbons, and polychlorinated biphenyls. Data from 1987 to 1997 were obtained from the Environmental Protection Department for statistical analysis and development of a pollution index for re¯ecting marine sediment quality. 2.2. Data treatment Data from 1987 to 1997 were examined for missing values and for identifying variables that had values

below analytical detection limits at the 64 monitoring stations. For variables with missing values at a particular station, these were substituted by the average value obtained at that station (Tabachnick and Fidell, 1996). For variables with values below analytical detection limit, these were replaced with one-half of that detection limit value (Gilliom and Helsel, 1986). Each variable in the entire data set was tested for its normality p and, as necessary, transformed using log (x+1) or x, where x=data value (Zar, 1996). Since the 24 variables had di€erent units, the data were standardized to zero mean and unit variance before subjecting the data set to principal component analysis (PCA) (Jolli€e, 1986). Prior to the analysis, multivariate normality of the data was tested by checking the linear correlation between the squared Mahalanobis distance and the Chi-squared value of each sample as described in Sharma (1996) and Tabachnick and Fidell (1996). PCA of the sediment data was conducted using the computer software package, STATISTICA1 (StatSoft, 1995). In the analysis, rotation of the PCA loadings was not performed. The principal components and associated loadings resulted from PCA were examined to select the most important variables for the derivation of the Marine Sediment Pollution Index (MSPI). Variables were selected for the index

Fig. 1. Water control zones in Hong Kong and sediment sampling stations for this study (the name of the water control zones and number of routine sediment monitoring stations in each zone are listed in Table 4).

P.K.S. Shin, W.K.C. Lam / Environmental Pollution 113 (2001) 281±291

calculation if the absolute value of its component loading was greater than 0.7, as suggested by Comrey and Lee (1992). 2.3. Calculation of MSPI The important variables selected from PCA were used to calculate the MSPI based on the modi®ed arithmetic weighted formula derived by the Scottish Development Department (SDD, 1976), as follows: MSPI ˆ …Sqi wi †2 =100; where qi is the sediment quality rating of the ith variable and wi the weight attributed to the ith variable. The sediment quality rating, qi, ranging from 0 (best possible quality) to 100 (worst possible quality), was developed through a ranking system. For each variable, the rating was based on the percentile in the data set. Except for electrochemical potential, a 10 mark was given to the rating if the value of the variables lied on the 0±10 percentile, a 20 mark to the value with 10±20 percentile, and so on. A zero mark was assigned to variables with value below analytical detection limit. For electrochemical potential, a reverse of the above ranking procedure was adopted as the lower the electrochemical potential the more serious pollution would be. The weight, wi, attributed to each variable in the MSPI was computed on the basis of the proportion of eigen values obtained from the results of PCA. For public communication purposes, the MSPI was classi®ed into ®ve bands depicting the sediment conditions: excellent (MSPI 0±20), good (MSPI 21±40), average (MSPI 41±60), poor (MSPI 61±80) and bad (MSPI 81±100). 2.4. Correlation of MSPI with biological and toxicity data The MSPI was calculated primarily from physicochemical variables selected from PCA of the sediment monitoring data set. To assess the validity of the MSPI in re¯ecting the state of marine sediment quality in Hong Kong waters, sediment samples at 18 monitoring stations (Fig. 1) were collected in 1997 for analysis of macrobenthic infaunal composition and for toxicity tests. At each station, ®ve replicate sediment samples were obtained using a 0.05 m2 van Veen grab. Each of the replicate samples was washed gently through 1- and 0.5-mm mesh sieves and the resultant residues were preserved in 5% formalin for subsequent laboratory sorting, identi®cation and enumeration of benthic animals. Species diversity H0 at these stations was calculated from the pooled replicates using the following formula (Shannon and Weaver, 1963):

H0 ˆ

283

s

S…Ni N†log2 …Ni N†; 1

where s is the total number of species, N the total number of individuals and Ni the number of individuals of the ith species. A 96-h toxicity test was undertaken for the 18 sediment samples using a local gammaridean amphipod Melita longidactyla. In the test, a 2-cm layer of sediment from a homogenized 10 cm sediment collected from the grab samples was placed on the bottom of a 1-l bottle with 800 ml of continuously aerated, ®ltered seawater. Amphipods collected from the ®eld were acclimatized in running seawater for 24 h and 10 animals of size between 0.5 and 0.7 cm were introduced to each test bottle. The number of living, dead and missing (presumably dead) amphipods in each bottle was determined after 96-h exposure by sieving the sediment through a 0.5-mm sieve. Control tests were performed using clean seawater and a reference, clean sediment collected at a location 20 km from Hong Kong's territorial waters in South China Sea. A total of ®ve replicates were used in each sediment sample run. The toxicity of the sediment was measured by the mean percent survival of the amphipods as compared to the controls (Swartz et al., 1982). The MSPI for these 18 sediment samples was calculated using the 1997 data obtained from the Environmental Protection Department. Correlation of the MSPI from these 18 monitoring stations with species diversity values and percent survival of the amphipods in the sediment toxicity test was analyzed using the Pearson product±moment procedures (Zar, 1996). Prior to the correlation analysis, the MSPI and percent survival data were arcsine transformed to conform to data normality. 3. Results The squared Mahalanobis distance and Chi-squared values of the data set demonstrated a linear relationship (correlation r=0.96, P=0.000). Results of PCA showed that the ®rst six principal components were found to explain 70% of the total variation in the sediment data set. The eigen values and loadings of these six components are shown in Table 1. Of the 24 variables analyzed, six had loading values greater than 0.7, including Cr, Cu, Ni, COD, Zn and Mn. The weight, wi, attributed to each of these six variables in the MSPI calculation is listed in Table 2, according to the proportion of eigen values of the principal components. The corresponding sediment quality rating, qi, for each of these variables is shown in Table 3. Table 4 summarizes the MSPI for the 64 stations in the 10 water control zones (WCZs) over the monitoring

284

P.K.S. Shin, W.K.C. Lam / Environmental Pollution 113 (2001) 281±291

Table 1 Eigen values and loadings of the ®rst six principal components (PCs) for the marine sediment data set in Hong Konga

Eigen value Cumulative eigen values Cumulative % variation Loadings of variables AN TKN TP COD TVS TS CN S TC As Al B Cd Cr Cu Fe Hg Mn Ni Pb Zn PCBs PAHs Eh

PC 1

PC 2

PC 3

PC 4

PC 5

PC 6

6.538 6.538 27

3.869 10.407 43

2.093 12.500 52

1.834 14.334 60

1.363 15.697 65

1.017 16.714 70

0.306 0.439 0.555 0.090 0.094 0.076 0.152 0.337 0.464 0.285 0.099 0.107 0.496 0.163 0.075 0.326 0.152 0.064 0.301 0.254 0.126 0.060 0.181 0.078

0.137 0.229 0.238 0.243 0.195 0.144 0.125 0.339 0.090 0.087 0.151 0.400 0.043 0.006 0.255 0.208 0.079 0.200 0.315 0.693 0.464 0.430 0.192 0.291

0.307 0.142 0.197 0.185 0.418 0.395 0.111 0.378 0.324 0.320 0.114 0.307 0.209 0.271 0.195 0.326 0.167 0.010 0.110 0.017 0.179 0.159 0.126 0.011

0.432 0.080 0.136 0.134 0.125 0.016 0.013 0.175 0.396 0.013 0.388 0.443 0.333 0.040 0.051 0.179 0.033 0.111 0.130 0.035 0.007 0.055 0.209 0.110

0.508 0.489 0.328 0.759b 0.532 0.529 0.063 0.498 0.157 0.194 0.406 0.229 0.563 0.876b 0.832b 0.231 0.668 0.037 0.728b 0.301 0.754b 0.479 0.535 0.530

0.106 0.294 0.143 0.015 0.402 0.478 0.023 0.174 0.337 0.525 0.569 0.480 0.056 0.029 0.285 0.671 0.371 0.791b 0.096 0.490 0.241 0.581 0.455 0.523

a AN, ammoniacal nitrogen; TKN; total Kjeldahl nitrogen; TP, total phosphorus; COD, chemical oxygen demand; TVS, total volatile solids; TS, total solids; CN, cyanide; S, sul®de; TC, total carbon; PCBs, polychlorinated biphenyls; PAHs, polynuclear aromatic hydrocarbons; Eh, electrochemical potential. b Loading value of being equal to or greater than 0.7.

Table 2 Weights for each of six variables selected from principal component analysis (PCA) for calculation of the Marine Sediment Pollution Index in Hong Kong marine sediments PC

Eigen value

Relative eigen value

Variable

Loading value

Relative loading value on same PC

Weight (relative eigen valuerelative loading value)

1

6.538

0.628

Cr Cu Ni CODa Zn

0.876 0.832 0.728 0.759 0.754 Total

0.222 0.211 0.184 0.192 0.191 1.000

0.139 0.133 0.116 0.121 0.119

3.869 10.407

0.372 1.000

Mn

0.372

1.000

0.372 1.000

2

Total a

COD, chemical oxygen demand.

period of 1987 to 1997. Sediment quality at most of these sampling stations was classi®ed `average'. Only in one area (WCZ3, Port Shelter) the sediment was classi®ed `good', whereas in WCZs 2 (Tolo Harbor) and 7 (Victoria Harbor) the sediment was identi®ed as `poor'

(see Fig. 1 for location). Fig. 2 presents the temporal changes of the MSPI in the WCZs. A signi®cant increase (analysis of variance [ANOVA], P<0.05) in the mean MSPI values from 1987 to 1997 was noted in three WCZs (WCZ1, WCZ5, WCZ7). Overall, the

P.K.S. Shin, W.K.C. Lam / Environmental Pollution 113 (2001) 281±291

285

Table 3 Sediment quality rating for each of the six variables selected from principal component analysis for calculation of the Marine Sediment Pollution Index in Hong Kong marine sedimentsa Sediment quality rating

Cr (mg kg 1)

Cu (mg kg 1)

Ni (mg kg 1)

COD (mg kg 1)

Zn (mg kg 1)

Mn (mg kg 1)

10 20 30 40 50 60 70 80 90 100

BDL110

BDL420

BDL42

BDL27,000

BDL340

BDL800

a

COD, chemical oxygen demand; x, data value in dry weight basis; BDL, below detection limit.

Table 4 Summary of Marine Sediment Pollution Index (MSPI) (meanS.D.) computed for 10 water control zones in Hong Kong over 1987 to 1997 (MSPI values are rounded up to whole number) Water control zone (locality name)

No. of monitoring stations

MSPI

No. of observations

Classi®cation of sediment quality based on mean MSPI

1 (Mirs Bay) 2 (Tolo Harbor) 3 (Port Shelter) 4 (Junk Bay) 5 (Eastern Bu€er Waters) 6 (Southern Waters) 7 (Victoria Harbor) 8 (Western Bu€er Waters) 9 (Western Waters) 10 (Deep Bay)

16 5 8 1 4 7 11 3 5 4

5510 6213 3814 5716 4618 5315 6215 5813 5214 5513

195 126 136 17 79 139 251 39 104 71

Average Poor Good Average Average Average Poor Average Average Average

mean MSPI for the territorial waters of Hong Kong has also shown an increasing trend (ANOVA, P=0.025), indicating the continual degradation of the marine sediment quality over the years (Fig. 3). A total of 76 species and 2200 individuals of macrobenthic infauna were recorded from the 18 sediment locations in Hong Kong. In terms of species composition, the dominant faunal groups were polychaetes (59.2%), crustaceans (19.7%) and molluscs (7.9%). In terms of individuals, polychaetes were numerically abundant (56.5%), followed by crustaceans (10.7%) and molluscs (2.2%). Among the sampling stations, species diversity H0 ranged from 1.59 to 4.62. Results of the toxicity test showed that the mean amphipod survival varied from 46 to 72% for the 18 sediment samples. Figs. 4 and 5 depict scatter plots of correlation between MSPI and species diversity H0 values and mean percent survival of the amphipod M. longidactyla in the sediment toxicity test. Signi®cant negative correlation was found between MSPI and H0 (P=0.019), and between MSPI and mean percent amphipod survival data (P=0.000).

4. Discussion The approach that was adopted in developing the MSPI index in this study was based on PCA of the sediment chemistry data. As a multivariate statistical technique, PCA has largely been used in identi®cation of patterns of environmental contamination (Pavoni et al., 1988; Evers et al., 1993; Poulton et al., 1996; Emmerson et al., 1997; Ausili et al., 1998; Naes and Oug, 1998) and relationships between abiotic and biotic factors (Ibanez and Dauvin, 1988; Rakocinski et al., 1991; Santos, 1993; Bilos et al., 1998). The use of PCA for development of a MSPI has, however, not been demonstrated. Vogt (1990) applied PCA to interpret sediment chemical composition and calculated ``pollution scores'' on the basis of a regression model derived from the PCA results. One main function of PCA is to reduce the dimensionality of the complex data set, so as to make sense of the variations observed (Tabachnick and Fidell, 1996). In the present study, results showed that only six of the 24 variables were found to be important in contributing

286

P.K.S. Shin, W.K.C. Lam / Environmental Pollution 113 (2001) 281±291

Fig. 2. Temporal changes of Marine Sediment Pollution Index (MSPI) (meanS.D.) in the 10 water control zones (WCZs) of Hong Kong (trend of temporal changes is represented by regression line and equation, and tested by ANOVA).

P.K.S. Shin, W.K.C. Lam / Environmental Pollution 113 (2001) 281±291

287

Fig. 3. Temporal changes of Marine Sediment Pollution Index (MSPI) (meanS.D.) in the entire territorial waters of Hong Kong (trend of temporal changes is represented by regression line and equation, and tested by ANOVA).

Fig. 4. A scatter plot of the correlation between Marine Sediment Pollution Index (MSPI) and species diversity H0 (MSPI values were arcsine transformed; straight line shows slope of negative correlation).

to the variations of the monitoring data. Of these six variables identi®ed from the PCA results, COD, Cr, Cu, Ni and Zn are all indicators of sediment pollution from anthropogenic sources (Fergusson, 1990). The high Mn values, on the other hand, are possibly due to

the natural enrichment process, in which the anoxic part of the sediment is reduced in the presence of high COD. Subsequent migration of Mn in pore water and reprecipitation on the sediment surface may result in elevated levels of Mn in the sediment. In Hong Kong, the

288

P.K.S. Shin, W.K.C. Lam / Environmental Pollution 113 (2001) 281±291

Fig. 5. A scatter plot of the correlation between Marine Sediment Pollution Index (MSPI) and mean percent amphipod survival (both MSPI and mean percent amphipod survival values were arcsine transformed; straight line shows slope of negative correlation).

poor sediment quality in Victoria Harbor is attributable to the indiscriminant discharges of untreated domestic and industrial wastewater over the years (Sin et al., 1996). In Tolo Harbor the high sediment pollution index value can be due to limited dispersion of pollutants resulting from poor ¯ushing within the semienclosed embayment (Li and Lee, 1996). Most of the pollutants are thus trapped in the sediment. In terms of absolute concentration of metals, the sediment in Victoria Harbor falls under the Category H of the marine sediment classi®cation scheme used in Hong Kong (EFB, 2000). The scheme divides sediment into three classes: (1) Category L (low contamination) Ð sediment with all contaminant levels not exceeding the lower chemical exceedance level (LCEL); (2) Category M (moderate contamination) Ð sediment with any one or more contaminant levels exceeding the LCEL and none exceeding the upper chemical exceedance level (UCEL); and (3) Category H (high contamination) Ð sediment with any one or more contaminant levels exceeding the UCEL. The LCEL and UCEL for each contaminant (e.g. metals) are derived from data on natural background level in the marine sediments of Hong Kong and from a literature review of the potential adverse ecological/toxicological e€ects of the contaminant on the marine biota (Lau-Wong et al., 1993). The scheme also stipulates the dredging and disposal methods for the di€erent sediment categories in order to protect the marine ecosystem from the release of the contaminants. For the sediment in Victoria Harbor, the

average concentration of Cr and Cu exceeds the UCEL of 160 and 110 mg kg 1, respectively, whereas that of Ni and Zn exceeds the LCEL of 40 and 200 mg kg 1. In Tolo Harbor, the sediment is also classi®ed under Category H, in which the mean concentration of Zn is higher than the UCEL of 270 mg kg 1. The only area that shows good sediment quality is Port Shelter, where industrial activities are minimal. Similar ®ndings were reported by the Hong Kong Environmental Protection Department in their annual monitoring report (EPD, 1998). Thus, the categorization of MSPI into ®ve bands depicting the general conditions of the marine sediment is relevant to the local conditions. In addition, the MSPI further reveals the signi®cant degradation of sediment quality in the territorial waters of Hong Kong, especially in WCZ1 (Mirs Bay), WCZ5 (Eastern Bu€er Waters) and WCZ7 (Victoria Harbor). The derived MSPI formulation from this study is useful for providing simple, understandable information on the state of marine sediment pollution for the public's reference. The methodology developed also relies on less sediment chemistry data, thereby providing a basis for re®ning existing monitoring programs. Parameters that do not show signi®cant spatial or temporal variations can be analyzed in a lesser frequency than those that are identi®ed to be more important from the results of PCA. This will save resources in long-term routine monitoring. The goal of PCA is to extract maximum variance from a data set with a few orthogonal components (Tabachnick and Fidell, 1996). This is successfully

P.K.S. Shin, W.K.C. Lam / Environmental Pollution 113 (2001) 281±291

applied to the present study in identifying the important measured variables that contribute towards variation from the data set, as the sampling program is extensive covering a wide range of sediment conditions in the territorial waters of Hong Kong. However, in a more limited data set, where samples are taken from similar sediment conditions, it may be dicult to identify the important measured variables, since they would have low loading values on the principal components owing to small variance. To some extent, such an e€ect can be minimized by ensuring rigorous sampling that comprises a wider monitoring area, preferably covering a spectrum of `clean' to `contaminated' sediments in the sampling regime. This will allow data from di€erent ecological habitats to be included in the PCA and represented in the formulation of the MSPI. The MSPI should be viewed as a simple summary of the state of the marine sediment quality based on routine chemistry monitoring data for public information. The MSPI is di€erent from the more complex Sediment Quality Triad which encompasses sediment chemistry, biological community and toxicity data (Chapman et al., 1997; Long and Wilson, 1997) or the benthic index of sediment contamination and toxicity derived from discriminant model (Engle et al., 1994). We appreciate the complex characteristics of marine sediments which should be best described by a variety of responses and measures and a single unit-less index may not provide adequate scienti®c information on the sediment quality (Chapman et al., 1997). However, for e€ective communication with laymen, a single index may have its value to allow the public to comprehend the general status of the sediment conditions through translating complex scienti®c data into a relatively simple term (Train, 1972). Our results also indicate signi®cant negative correlation between the MSPI and benthic species diversity as well as sediment toxicity in the study area. This suggests that the MSPI is useful in re¯ecting the state of the benthic communities (as represented by within-habitat species diversity) and ex-situ toxicity level (as represented by mean percent amphipod survival) of the sediment. The regression model developed by Vogt (1990) also showed that low diversity values were associated with increased pollution scores in a southern Norwegian fjord. However, the use of species diversity alone as indicator of sediment contamination should be cautioned. There are many studies which show that species diversity is not very sensitive as pollution indicator, apart from the results of eutrophication and organic enrichment (Gray et al., 1990; Warwick and Clarke, 1992; Olsgard and Hasle, 1993; Oug et al., 1998). As shown by Olsgard and Gray (1995), diversity indices exhibit varying responses depending on the type of pollutant in the sediment. For example, values of diversity indices will not usually decrease even though the sediment is toxic, since both the total number of species and

289

the number of individuals decrease. The use of a variety of benthic endpoints, such as the incorporation of benthic mortality (ecotoxcity) data in the present study and/or changes in trophic composition (Oug et al., 1998) may circumvent the problems of relying on diversity indices alone. Applications of multivariate methods are also deemed more sensitive in analysis of the response of benthic community changes to sediment contamination (Olsgard and Gray, 1995; McRae et al., 1998). Ordinary PCA does not rely on any hypothesis on the underlying data structure (Tabachnick and Fidell, 1996). The analysis, however, is not invariant under a change in scale. When variables are measured in di€erent units, the data should be mean-corrected or standardized (Sharma, 1996). Moreover, if the parameters derived in PCA should be used further as in the present study, it might be necessary to ensure multivariate normal distribution of the data set (Tabachnick and Fidell, 1996). Multivariate normality implies that the sampling distributions of means of the various variables (parameters) and all linear combinations of them are normally distributed. In the present study, the linear relationship between the squared Mahalanobis distance and Chisquared values suggesting that the data set ful®lls the assumption of multivariate normality. This also shows that it is appropriate to test the normality of each variable and transform the data, as necessary, prior to subjecting the data set to PCA. With modern computerbased statistical software packages, it is easy for users to employ inappropriate techniques to analyze data without fully understanding the assumptions of various statistical methods. Violation of the normality assumption may a€ect the power of a test statistic and result in committing Type II errors, i.e. increasing the probability of failing to reject the null hypothesis when in fact it is false (Sharma, 1996). To improve reliability of the correlation among variables, a large sample size is recommended for PCA. Comrey and Lee (1992) suggested that at least 200 cases should be included for a meaningful result. In our study, a total of 1157 sets of data (Table 4) were used in the derivation of the MSPI. This thus enhances the reliability of identifying the most important parameters from the monitoring program for the formulation of the index. Acknowledgements We would like to thank the Environmental Protection Department of the Hong Kong Special Administrative Region Government in allowing access to the sediment database, and Prof. Peter Tanner for his critical reading of the manuscript. Thanks are also extended to the three anonymous reviewers who provided invaluable comments on the manuscript.

290

P.K.S. Shin, W.K.C. Lam / Environmental Pollution 113 (2001) 281±291

References Ausili, A., Mecozzi, M., Gabellini, M., Ciu€a, G., Mellara, F., 1998. Physico chemical characteristics and multivariate analysis of contaminated harbour sediments. Water Science and Technology 37, 131±139. Bergstedt, L.C., Bergersen, E.P., 1997. Health and movements of ®sh in response to sediment sluicing in the Wind River, Wyoming. Canadian Journal of Fisheries and Aquatic Science 54, 312±319. Bilos, C., Colombo, J.C., Persa, M.J.R., 1998. Trace metals in suspended particles, sediments and Asiatic clam (Corbicula ¯uminea) of the Rio de la Plata Estuary, Argentina. Environmental Pollution 99, 1±11. Chapman, P.M., Hyland, J., Ingersoll, C., Carr, S., Engle, V., Green, R., Hameedi, J., Harmon, M., 1997. General guidelines for using the sediment quality triad. Marine Pollution Bulletin 34, 368±372. Comrey, A.L., Lee, H.B., 1992. A First Course in Factor Analysis, 2nd Edition. Lawrence Erlbaum Associates, Hillsdale, NJ, USA. Dojlido, J., Raniszewski, J., Woyciechowska, J., 1994. Water quality index applied to rivers in the Vistula River basin in Poland. Environmental Monitoring and Assessment 33, 33±42. EFB, 2000. Management of Dredged/Excavated Sediment (Works Bureau Technical Circular No. 3/2000). Environment and Food Bureau, Hong Kong. Hong Kong Special Administrative Region Government, Hong Kong. Emmerson, R.H.C., O'Reilly-Wiese, S.B., Macleod, C.L., Lester, J.N., 1997. A multivariate assessment of metal distribution in inter-tidal sediments of the Blackwater Estuary, UK. Marine Pollution Bulletin 34, 960±968. Engle, V.D., Summers, J.K., Gaston, G.R., 1994. A benthic index of environmental condition of Gulf of Mexico estuaries. Estuaries 17, 372±384. EPD, 1998. Marine Water Quality in Hong Kong in 1997. Environmental Protection Department. Hong Kong Special Administrative Region Government, Hong Kong. Evers, E.H.G., Klamer, H.J.C., Lane, R.W.P.M., Govers, H.A.J., 1993. Polychlorinated dibenzo-p-dioxin and dibenzofuran residues in estuarine and coastal North Sea sediments: sources and distribution. Environmental Toxicology and Chemistry 12, 1585±1598. Fergusson, J.E., 1990. The Heavy Elements: Chemistry, Environmental Impacts and Health E€ects. Pergamon Press, Oxford, UK. Gilliom, R.J., Helsel, D.R., 1986. Estimation of distributional parameters for censored traced level water quality data. I. Estimation techniques. Water Resources Research 22, 135±146. Gray, J.S., Clarke, K.R., Warwick, R.M., Hobbs, G., 1990. Detection of initial e€ects of pollution on marine benthos: an example from the Eko®sk and Eld®sk oil®eld, North Sea. Marine Ecology Progress Series 66, 285±299. Horton, R.K., 1965. An index-number system for rating water quality. Journal of Water Pollution Control Federation 37, 300±306. Ibanez, F., Dauvin, J.C., 1988. Long-term changes (1977 to 1987) in a muddy ®ne sand Abra alba - Melinna palmata community from the western English Channel: multivariate time-series analysis. Marine Ecology Progress Series 49, 65±81. Jolli€e, I.T., 1986. Principal Component Analysis. Springer-Verlag, New York, USA. Lau-Wong, M.M.M., Rootham, R.C., Bradley, G.C., 1993. A strategy for the management of contaminated dredged sediment in Hong Kong. Journal of Environmental Management 38, 99±114. Li, Y.S., Lee, J.H.W., 1996. Hydraulics of Deep Bay and Tolo Harbour. In: Coastal Infrastructure Development in Hong Kong. A Review. Proceedings of the Symposium on Hydraulics of Hong Kong Waters, Hong Kong, 1995. Hong Kong Government, Civil Engineering Oce, Civil Engineering Department, pp. 67±72. Lobinske, R.J., Ali, A., Stout, I.J., 1997. Benthic macroinvertebrates and selected physico-chemical parameters in two tributaries of the Wekiva River, central Florida, USA. Medical Entomology and Zoology 48, 219±231.

Long, E.R., Wilson, C.J., 1997. On the identi®cation of toxic hot spots using measures of the sediment quality triad. Marine Pollution Bulletin 34, 373±374. McRae, G., Camp, D.K., Lyons, W.G., Dix, T.L., 1998. Relating benthic infaunal community structure to environmental variables in estuaries using nonmetric multidimensional scaling and similarity analysis. Environmental Monitoring and Assessment 51, 233±246. Miller, W.W., Joung, H.M., Mahannah, C.N., Garrett, J.R., 1986. Identi®cation of water quality di€erences in Nevada through index application. Journal of Environmental Quality 15, 265±271. Naes, K., Oug, E., 1998. The distribution and environmental relationships of polycyclic aromatic hydrocarbons (PAHs) in sediments from Norwegian smelter-a€ected fjords. Chemosphere 36, 561±576. Oberdor€, T., Hughes, R.M., 1992. Modi®cation of an index of biotic integrity based on ®sh assemblage to characterize rivers of Seine Basin, France. Hydrobiologia 228, 117±130. Olsgard, F., Hasle, J.R., 1993. Impact of waste from titanium mining on benthic fauna. Journal of Experimental Biology and Ecology 172, 185±213. Olsgard, F., Gray, J.S., 1995. A comprehensive analysis of the e€ects of o€shore oil and gas exploration and production on the benthic communities of the Norwegian continental shelf. Marine Ecology Progress Series 122, 277±306. Ott, W.R., 1978. Environmental Indices. Theory and Practice. Ann Arbor Science, Michigan, USA. Oug, E., Naes, K., Rygg, B., 1998. Relationship between soft bottom macrofauna and polycyclic aromatic hydrocarbons (PAH) from smelter discharge in Norwegian fjords and coastal waters. Marine Ecology Progress Series 173, 39±52. Pavoni, B., Marcomini, A., Sfriso, A., Orio, A.A., 1988. Multivariate analysis of heavy metal concentrations in sediments of the lagoon of Venice. The Science of the Total Environment 77, 189±202. Poulton, D.J., Morris, W.A., Coakley, J.P., 1996. Zonation of contaminated bottom sediments in Hamilton Harbour as de®ned by statistical classi®cation systems. Water Quality Research Journal of Canada 31, 505±528. Prati, L., Pavanello, R., Pesarin, F., 1971. Assessment of surface water quality by a single index of pollution. Water Research 5, 741±751. Rakocinski, C., Heard, R.W., Simons, T., Gledhill, D., 1991. Macroinvertebrate associations from beaches of selected barrier islands in the northern Gulf of Mexico: important environmental relationships. Bulletin of Marine Sciences 48, 689±701. Read, P.A., Anderson, K.J., Matthews, J.E., Watson, P.G., Halliday, M.C., Shields, G.M., 1982. Water quality in the Firth of Forth. Marine Pollution Bulletin 13, 421±425. Santos, R., 1993. A multivariate study of biotic and abiotic relationships in a subtidal algal stand. Marine Ecology Progress Series 94, 181±190. SDD, 1976. Development of a Water Quality Index (Report AR3). Scottish Development Department, Edinburgh, UK. Shannon, C.E., Weaver, W., 1963. The Mathematical Theory of Communication. University of Illinois Press, Urbana, IL, USA. Sharma, S., 1996. Applied Multivariate Techniques. John Wiley, New York, USA. Sin, W.S., Chan, P.K., Chau, K.M., 1996. Sewage and stormwater disposal. In: Coastal Infrastructure Development in Hong Kong. A Review. Proceedings of the Symposium on Hydraulics of Hong Kong Waters, Hong Kong, 1995. Hong Kong Government, Civil Engineering Oce, Civil Engineering Department, pp. 343±352. StatSoft, 1995. STATISTICA1 for the Windows Operating System. Release 5. StatSoft, Tulsa, OK, USA. Swartz, R.C., DeBen, W.A., Sereu, K.A., Lamberson, J.O., 1982. Sediment toxicity and the distribution of amphipods in Commencement Bay, Washington, USA. Marine Pollution Bulletin 13, 359± 364.

P.K.S. Shin, W.K.C. Lam / Environmental Pollution 113 (2001) 281±291 Tabachnick, B.G., Fidell, L.S., 1966. Using Multivariate Statistics, 3rd Edition. HaperCollins College Publishers, New York, USA. Train, R.E., 1972. The quest for environmental indices. Science 178, 121. Tyson, J.M., House, M.A., 1988. The application of a water quality index to river management. Water and Science Technology 21, 1149±1159. Vogt, N.B., 1990. Multivariate ecotoxicological mapping of the relationships between sediment chemical composition and fauna diversity. The Science of the Total Environment 90, 149±161.

291

Warwick, R.M., Clarke, K.R., 1992. Comparing the severity of disturbance: a meta-analysis of marine macrobenthic community data. Marine Ecology Progress Series 92, 221±231. Yu, J.K., Fogel, M.M., 1978. The development of a combined water quality index. Water Resource Bulletin 14, 1239±1250. Zar, J.H., 1996. Biostatistical Analysis, 3rd Edition. Prentice-Hall International, NJ, USA. Zou, J., Zhang, J., Wu, J., Zhang, F., Gu, T., Wu, Y., 1988. On organic pollution and its control in the Haihe estuarine area of the Bohai Bay. Studies of Marine Sinica 29, 1±20.