(MW) Databases - ACS Publications - American Chemical Society

Article pubs.acs.org/est

Tracking Temporal Trend Breaks of Anthropogenic Change in Mussel Watch (MW) Databases Carlos Guitart,*,† Adrian Hernández-del-Valle,‡ J. Miguel Marín,§ and José Benedicto† †

Spanish Institute of Oceanography, Murcia Oceanographic Center, 30740 Murcia, Spain Department of Economics, National Polytechnic Institute, 07738 Mexico DF, Mexico § Department of Statistics, University Carlos III, 28903 Madrid, Spain ‡

S Supporting Information *

ABSTRACT: The potential for structural changes in time trend concentrations of mercury (Hg), lead (Pb), cadmium (Cd), zinc (Zn), and copper (Cu) in the Mediterranean mussel, Mytilus galloprovincialis, was examined in Mussel Watch (MW) databases of metal pollution at eighteen coastal stations over a decadal period, from 1992 to 2007. Simultaneously, by using two statistical methods representing both the classical hypothesis-testing and the Bayesian approaches, we found single and multiple trend breaks for Hg (28% of the stations), Cd (17%), and Pb (11%) within trends in connection with anthropogenic and subtle natural environmental changes. Also called change point problems, if not accounted for, these could bias time trend investigations and interpretations. We calculated trend rate differences of 39% and switches up to 1 order of magnitude from classical linear trend assessments. We discuss sampling, analytical, and environmental (both natural and anthropogenic) sources of data set variabilities, showing that, in practice, the overall 16-year analytical performance could be as elevated as the yearly sampling reproducibility. We demonstrate that environmental time trend interpretations benefit from undertaking prior structural change analysis. After decades of MW marine chemical pollution assessments these have proven extremely useful, although the occurrence of trend breaks directly affects the long-term marine environmental monitoring strategies. Our results suggest a broader concept to design monitoring programs in agreement with rapid global anthropogenic and environmental changes.

■

INTRODUCTION Marine environment assessments based on the “Mussel Watch” concept have been conducted for almost 30 years providing unique marine chemical pollution long-term databases.1 In essence, instead of complex and expensive off-shore coastal sampling designs (including extensive use of research vessels and seawater trace chemical analysis), surveillance along coastal areas using widespread marine filter-feeding organisms as sentinels (e.g., mussels, oysters, etc.) should provide a straightforward time-integrated measure of a chemical’s concentration in the surrounding coastal waters.1,2 While the use of marine organisms as sentinels to monitor the seas and ocean’s anthropogenic pollution is frequently debated,3,4 the concept became global and has indicated upward and downward coastal pollution trends. The International Mussel Watch (IMW), the longest-running program for the monitoring of spatial distributions and temporal trends of chemical contaminant concentrations in different coastal environments, began worldwide in 1991,5 following the pioneering U.S. Mussel Watch Program started in 1976 by E.D. Golberg and the National Status and Trends Program (NS&T) from the U.S. National Oceanographic and Atmospheric Agency (NOAA).6 As a result, national mussel watch programs were implemented globally to cover numerous geographical areas of coastal oceans and seas as a primary component of marine © 2012 American Chemical Society

monitoring programs, such as the MEDPOL Program in the Mediterranean Sea.7 Within such programs, rigorous analytical quality assurance programs were established to ensure both sampling methods and chemical measurements of pollutants in biota tissues are made on a comparable basis worldwide. As a premise, temporal trend monitoring implies that sampling and analytical procedures do not change over the studied period and that data sets comply with a certain analytical quality threshold. To this end, analytical quality assurance and quality control (QA/QC) programs and marine certified reference materials (CRMs) were developed and supported by international organizations.8 The World Mussel Watch Database includes coastal chemical pollution data sets from more than fifty countries worldwide9,10 affording a huge amount of scientific research. In environmetrics it is frequently difficult to determine significative temporal trends in time series or longitudinal data. In particular, marine environmental databases such as MW databases, are often statistically compromised with restricted sample sizes over long periods of time and are hindered by Received: Revised: Accepted: Published: 11515

May 28, October October October

2012 2, 2012 5, 2012 5, 2012

dx.doi.org/10.1021/es3021183 | Environ. Sci. Technol. 2012, 46, 11515−11523

Environmental Science & Technology

Article

Figure 1. Geographical distribution of the Spanish “Mussel Watch” Network in the Western Mediterranean Sea coast (see Table S2 in Supporting Information for detailed location information). Base map obtained from Google Earth.

missing data, thus limiting statistical methods that are applicable. Moreover, trend monitoring programs can be explicitly designed to fulfill the requirements of the statistical methods used,11,12 therefore, there is a reciprocal relationship between statistics and sampling theories.13,14 Typically, nonparametric statistical methods are employed to assess temporal trends, such as Sen’s slope and Kendall-tau-b or Spearman’s rho correlations. These are insensitive to gross data errors, outliers, or missing data.15,16 Alternatively, data sets are frequently forced to fulfill normal probability distribution criteria (e.g., by excluding prejudiced data or outliers, logarithmic data transformations, etc.), if for example, significative statistical linear trends with an arbitrary statistical power are to be expected.17,18 Usually, the use of log-transformed data sets from experimental sampling and monitoring is recurrent in environmental sciences, thus logarithmic data sets frequently exhibit a parametric behavior amenable to statistical analysis. However, the logarithmic function itself minimizes the magnitude of the data sets variability at the expense of the simplification of environmental processes. Therefore, between the typical and directed temporal trend analysis methods, several options such as smoothing techniques19 and polynomial functions for seasonality trends20 are also in use to accommodate time series data sets. Nevertheless, on occasion, none of these approaches are capable of convincingly explaining coastal pollution trends, despite their statistical assessments. Structural breaks or change point problems analysis has been made popular in econometrics to forecast macroeconomic data,21,22 such as gross domestic product (GDP),23 or applied to climate sciences, such as rainfall historical patterns.24 Often in MW time series apparent high data dispersion can be observed even in high-quality validated data sets, thus, having potential trend break (TB) or regime switch features. Those abrupt changes in time series can be statistically assessed to understand the phenomenon which deviates from their starting

or expected trends and connected with existing chronological events. Therefore, these represent a growing statistical alternative for temporal trend investigations without, for example, obvious monotonic linear upward or downward trends.25 Despite the growing number of statistical methods, one of the most used is the Bai and Perron26 methodology which is based on a least-squares regression to estimate the locations of the changes (i.e., trend breaks). The optimal model, in terms of the possible number of change points, is obtained by an information criterion which takes into account the complexity and accuracy of the model at the same time, and then, specific locations of change point are estimated. Other alternative methods, which allow the introduction of prior relevant information to discover possible change points, are based on the Bayesian methodology following the approach of Barry and Hartigan.27 The method computes the probability distribution of a change point at each location in a sequence of observations, instead of specific locations. In this study, we evaluate and discuss the occurrence and significance of trend breaks to investigate temporal trends in validated MW databases. We use alternative statistical methodologies for the time series analysis representing the hypothesistesting and Bayesian statistical points of view. In practice, both were useful to identify trend breaks, while the Bayesian methodology provided the possibility of their occurrence in terms of probability. Thus, both methods can be complementary. Heavy metals are considered as one of the most important group of marine chemical pollutants due to their toxicity and accumulative behavior in organisms. Therefore, we have re-examined our MW database of metal pollution in mussel (Mytilus galloprovincialis) over a 16-year period in the Mediterranean Sea. 11516



■

Article

EXPERIMENTAL SECTION Sampling Strategy and Analytical Methods. Mussel samples were collected at eighteen stations, which constitute part of the Spanish “Mussel Watch” Network, located along 2000 km of the Western Mediterranean coast, from the Gulf of Lion (Spain−France boundary) to the Gibraltar Strait facing the Atlantic Ocean (Figure 1). At each station, the mussel population was sampled (before the spawning period) by taking 150 individuals to generate 3 pooled samples of 50 individuals each, every year between April and June. Within the pool, mussels were measured and classified by shell length in 10 groups, containing a group of 5 individuals from 3.0 ± 0.05 to 3.9 ± 0.05 cm (i.e., 3.0, 3.1, 3.2, 3.3, etc.) to minimize the variability due to the mussel size effect. At the same time, 12 individuals of shell length size 4.0 ± 0.05 cm were sampled to calculate the condition index (CI), as CI = body dry weight/ shell weight, as a measure of the mussel physiological status in relation to the environmental conditions which might affect their contaminant burden.28 In addition, measurements of salinity (database mean: 37.0 ± 1.1 psu; RSD: 3%) and temperature (17.6 ± 2.2 °C; RSD: 13%) were also recorded in the field. At the laboratory, samples were pretreated and processed for trace metal determinations. The mussels were homogenized, lyophilized, bottled, and stored until chemical analysis. The Hg, Cd, Pb, Zn, and Cu analytical determinations were carried out by atomic absorption spectrometry (AAS).29−31 Briefly, following a microwave sample digestion with HNO3 (nitric acid), Zn and Cu were determined by flame AAS (F-AAS, AAnalyst 100), Cd and Pb were determined by graphite furnace with Zeeman background correction (ZGF-AAS, 4110 ZL, Perkin-Elmer), and Hg was determined by cold vapor generation using a flow injection mercury system (CV-AAS, FIMS 400, Perkin-Elmer). Analytical Quality Assurance and Database Validation. Between 1992 and 2008, three successive certified reference materials, CRM 278, CRM 278R, and ERM 278, were employed within batches of analyzed samples as internal laboratory quality assurance procedures and were purchased from the European BCR Office and the Institute of Reference Materials and Measurements (IRMM, Belgium, Europe). The overall (16-year) analytical reproducibilities for trace metal determinations in mussel matrix, as a percentage of the residual standard deviation (%RSD), were 8.2% (N = 175) for Cd, 15.8% (N = 190) for Hg, and 7.9% (N = 187) for Pb. These were considered as indicative of the possible maximum analytical variability (from CRMs) over the studied period regardless of the sample median concentration ranges, although this constitutes an approximation (Figure 3a, b, and c). In addition, the laboratory external quality assurance included the participation, twice a year, in the European round tests for trace metal determinations in marine matrices through QUASIMEME (Quality Assurance of Information for Marine Environmental Monitoring in Europe).32 Since 1992 our laboratory accounted 94.7% satisfactory results (Z-Score < ± 2). Moreover, the laboratory also participated in the several Worldwide Interlaboratory Exercises and the MEDPOL regional round tests for the determination of trace metals in several marine matrices organized by the IAEA for the same purpose.33 The stations and sampling protocols did not change from 1992 to 2007, providing a coherent and reproducible field

sampling conditions. Moreover, environmental parameters and mussel CI were recorded to account for the variability due to natural environmental factors. Therefore, the potential sources of data sets variability were both minimized and controlled over the studied period (detailed analytical quality assessments for Hg, Cd, and Pb (Figures S1, S2, and S3) and plots of the analytical Z-scores and CI (Figure S4) are provided as SI). Database Treatment, Structural Breaks, and Trend Analysis. The median contaminant concentration (mg/kg dry weight mussel tissue) and the interquartile range (IQR) using the 3 pooled mussel samples were calculated at each coastal station. The resulting time series (n = 90) was tested for normality (i.e., Jarque-Bera test), independence (i.e., Ljung-Box test), and homoscedasticity (i.e., ARCH test). Overall, 82% of the untransformed time series data sets were normally distributed (n = 74) and 96% (n = 87) exhibit independence and homoscedasticity (i.e., homogeneous variance). We further assumed the split data sets within a time series to fulfill the same criteria, but the time series not satisfying the statistical criteria were not further discussed. The Bai-Perron and Bayesian methodologies applied, assuming normally distributed data sets, independence, and homoscedasticity over the time periods, although the independence condition might be relaxed for the latter.27 Initially, the time series for Hg, Cd, Pb, Zn, and Cu containing data compiled over a 10-year period (11−16 years) were completed to cover the whole 1992−2007 period using both linear and neural network interpolation methods to evaluate their effect on statistical structural break analysis. The linear interpolation estimates the missing values through a lineal function between the closest data, while the neural network considers the whole time series data set to estimate missing data. Then, the time series followed the structural break analysis. The time series exhibiting TBs were carefully examined to fulfill the statistical criteria (see a statistical resume (Figure S5) for all the time series (Table S2 and S3) in SI). Within the Bayesian methodology, the posterior probability of change (PPC) describes the degree of statistical confidence to assert a TB in terms of probability and the posterior means estimates a concentration value for each year within a time series. Additionally, to elucidate the significance of the trend breaks at each location in relation to the whole geographical area, a background posterior probability of change (BPPC) was calculated as the mean PPC results from all the stations [Hg (n = 150), Cd (n = 195), and Pb (n = 180)], except the stations with TBs, to provide background TB probability threshold value for each metal. Therefore, time series' with PPC higher than BPPC were considered as exhibiting significative TB. However, the interpolation methods distorted the results of the TB analysis by modifying the PPC and Bai-Perron test significance, e.g. especially when data are missing at the beginning and at the end of the data sets (see an example in SI). Therefore, further statistical analysis and discussion focused on examples from continuous and original data sets at three selected stations, namely, Algeciras, Barcelona, and Cadaqués (for the latter, years 2000 and 2001 were interpolated, see Section 4 in SI). These are representative stations exhibiting strong trend breaks for Hg (1994−2007), Cd (1993−2005), and Pb (1992−2007), respectively. Besides, to investigate the significance of the TBs in trend interpretations, the linear trends by the nonparametric Sen’s slope method were calculated after splitting the time series using the highest probable TB year (given by the Bai-Perron 11517



11518

−0.006 (0.858) −0.029 (0.367) −0.064 (0.013) 5−10 years) for different aims and needs, while going into detail considering both the geographical scales and the case particularities to plan sampling frequencies and station distributions. Thus, such stations without present trends could be the best to monitor chronic pollution in the marine environment under a long-term objective, although program components should be periodically reviewed. In addition, local, regional, and global marine surveillance and long-term monitoring programs should put in context preindustrial studies,40 as well as consider present and future economic developments potentially linked to anthropogenic environmental impacts49 to help design, as well as revise marine monitoring activities. So far, IMW programs have proven extremely useful for the monitoring of the past anthropogenic impacts in the marine environment, which were recorded in MW network databases, such as the Chernobyl nuclear incident,50 as well as helped to support emergency monitoring activities, such as after the Prestige tanker oil spill.51,52 In connection to our study, these environmental incidents can be seen either as TBs or “outliers” under different time scales, in a manner similar to what can be expected to be recorded in MW coastal networks from the recent Fukushima nuclear plant incident and the Gulf of Mexico oil spill. Therefore, statistical assessment methodologies implemented in trend monitoring programs would benefit moving away from the simplification of the environmental data sets to underpin the complexity of the environmental processes in a changing environment. The objective of the long-term monitoring should be about finding, controlling, and understanding sources of variability, while observing ecosystem responses and implementing restoration programs, beyond strict statistics. To this end, Bayesian statistical approaches (i.e., probability results) could be more resolving to take action within marine environmental assessments and restoration programs and should be further investigated. In summary, we demonstrated that marine chemical pollution MW databases unveil statistical trend break features which could modify trend interpretations and predictions of potential chemical pollution hazards in the marine environment. The sustained quality assurance within marine monitoring programs is shown to be essential. Our study also suggests that marine pollution monitoring programs could be adapted and revised to better understand present and long-term global environmental changes.

As a result, we observed that the overall maximum analytical reproducibility (%RSD) calculated from CRMs over a 16-year period and applied to the median concentrations across the whole time series could be similar to the sample yearly field reproducibilities (Figure 3), even under rigorous quality controlled data sets. Overall, Figure 3 shows similar analytical variability (error bars) within the time series', while in Figures 3a and 3b those are somehow enlarged in higher median values (due to a larger concentration range time series), although the yearly field measurement reproducibilities follows a random pattern over time. In this study, as these two main sources of data sets variability can be minimized and controlled (i.e., analytical and field reproducibilities), the gross metal concentration changes also observed within the time series and statistically assessed as TBs point to an anthropogenic origin, as mentioned earlier. On the other hand, the fitted smoothing curves appear to coincide well solely in Figure 3a with TB interpretations, while in Figure 3b and 3c its combination with the Bayesian posterior mean estimations improves the interpretation of anthropogenic pollution trends. Regardless of the statistical methods applied, the magnitude of the variability sources within environmental time series' need to be scaled and interpreted according to the length of the investigated periods, as much as reduced, controlled, or eliminated. Yet, the different sources of variability can be differentiated from intermittent or interrupted anthropogenic pollution episodes in MW databases over a decadal period by studying TBs occurrence prior to trend interpretations as carried out and resumed as follows: (a) undertaking systematic sampling operations, (b) high-quality database validation, (c) performing structural change analysis, and (d) applying classical temporal trend statistical methods. Unpredictability of the Long-Term Marine Monitoring Programs. Monitoring programs play a fundamental role in environmental pollution and experimental sciences, being the basis of environmental impact assessments (EIAs) and natural resource damage assessments (NDRAs).48 Undoubtedly, longterm monitoring programs have to choose initial sampling time frames and spatial scales and also require coherent statistical considerations a priori for sample collection. Some authors suggested theoretical periods (e.g., decades) to detect significative changes of linear temporal trends (in trend %) with an arbitrary and maximized statistical power, that is, controlling for false positives (α, type I error), as well as a true positive decisions (i.e., 1 - type II error, β), respectively.11,12 However, type II error can not be arbitrarily fixed nor predicted, and therefore, this may lead to stringent monitoring programs with increasing amounts of costly sampling activities which could reveal to be statistically ineffective in the long term. Moreover, both the assessment of linear trends and to pursue statistically powerful monitoring programs has proved more complex than expected,17 and in general, it is argued that when additional data become available statistical tests will be more valid (i.e., statistically significative). However, according to our study, significative statistical results (under the hypothesis-testing statistical approach) could not be guaranteed even with an infinite number of years of monitoring activities, due to the occurrence and unpredictability of environmental changes (i.e., trend breaks), including those derived from anthropogenic activities. Even so, the majority of the stations for Hg, Cd, and Pb, and almost all for Zn and Cu, did not show TBs occurrence neither upward nor downward trends. Thus, a question arises from our

■

ASSOCIATED CONTENT

S Supporting Information *

(1) Database validation assessment: analytical quality assurance for Cd, Hg, and Pb determinations from 1992 to 2008 and the Z-scores and CI plots; (2) statistical resume for all the evaluated time series (n = 90); (3) an example of the comparison between interpolation methods for missing values with regard to the Bayesian posterior probabilities of change variability; (4) full detailed statistical results for the Bai-Perron and Bayesian methodologies at the selected stations; and (5) complete data set correlation tables at the selected stations. 11521



Article

(16) O’Connor, T. P.; Lauenstein, G. G. Trends in chemical concentrations in mussels and oysters collected along the US coast: Update to 2003. Mar. Environ. Res. 2006, 62, 261−285. (17) Assessment of Data Collected under the Co-Ordinated Environmental Monitoring Programme (CEMP); ISBN1-904426-77-8; OSPAR Commission, 2005; Appendix 4, p 57. http://www.ospar.org/ documents/DBASE/Publications/p00235_CEMP%20report.pdf. (18) CEMP Assessment Manual. Co-Ordinated Environmental Monitoring Programme Assessment Manual for Contaminants in Sediment and Biota, No. 379/2008; ISBN978-1-906840-20-4; OSPAR Commission, 2008. http://www.ospar.org/documents/dbase/publications/p00379_ CEMP_assessment_manual.pdf. (19) Fryer, R. J.; Nicholson, M. D. Using smoothers for comprehensive assessments of contaminant time series in marine biota. ICES J. Mar. Sci. 1999, 56, 779−790. (20) Surveillance du Milieu Marin. Travaux du RNO; ISSN1620-1124; Ifremer et Ministère de l’Aménagement du Territoire et de l’Environnement; Ifermer (RNO): Nantes, 2000. (21) Busetti, F.; Taylor, A. M. R. Variance shifts, structural breaks, and stationary tests. J. Bus. Econ. Stat. 2003, 21, 510−531. (22) Harvey, D. I.; Leybourne, S. J.; Taylor, A. M. R. Robust methods for detecting multiple level breaks in autocorrelated time series. J. Econometrics 2010, 157, 342−358. (23) Busetti, F. Testing for (Common) stochastic trends in the presence of structural breaks. J. Forecasting 2002, 21, 81−105. (24) de la Casa, A.; Nasello, O. Breakpoints in annual rainfall trends in Córdoba, Argentina. Atmos. Res. 2010, 95, 419−427. (25) Monson, B. Trend reversal of mercury concentrations in piscivorous fish from Minnesota lakes: 1982−2006. Environ. Sci. Technol. 2009, 43, 1750−1755. (26) Bai, J.; Perron, P. Computation and Analysis of Multiple Structural Change Models. J. Appl. Econometrics 2003, 18, 1−22. (27) Barry, D.; Hartigan, J. A. A Bayesian Analysis for Change Point Problems. J. Am. Stat. Assoc. 1993, 88, 309−319. (28) Benedicto, J.; Andral, B.; Martinez-Gomez, C.; Guitart, C.; Deudero, S.; Cento, A.; Scarpato, A.; Caixach, J.; Benbrahim, S.; Chouba, L.; Bouladhid, S.; Galgani, F. A large scale survey of trace metal levels in coastal waters of Western Mediterranean Basin using caged mussels (Mytilus galloprovincialis). J. Environ. Monit. 2011, 13, 1495−1505. (29) Besada, V.; Fumega, J.; Vaamonde, A. Temporal trends of Cd, Cu, Hg, Pb and Zn in mussel (Mytilus galloprovincialis) from the Spanish North-Atlantic coast 1991−1999. Sci. Total Environ. 2002, 288, 239−253. (30) Benedicto, J.; Martínez-Gómez, C.; Guerrero, J.; Jornet, A.; Rodriguez, C. Metal contamination in Portman Bay (Murcia, SE Spain) 15 years after the cessation of mining activities. Cienc. Mar. 2008, 34, 1−10. (31) Besada, V.; Andrade, J. M.; Schultze, F.; Fumega, J.; Cambeiro, B.; González, J. J. Statistical comparison of trace metal concentrations in wild mussels (Mytilus galloprovincialis) in selected sites of Galicia and Gulf of Biscay (Spain). J. Mar. Syst. 2008, 72, 320−331. (32) Wells, D. E.; Chiffoleau, J. F.; Klungsoyr, J. QUASIMEME: a preliminary study on the effects of sample handling on the measurement of trace metals and organochlorine residues in mussels. Mar. Pollut. Bull. 1997, 35, 109−124. (33) Azemard, S., de Mora, S., Guitart, C., Wyse, E. Report on the World Wide Intercomparison Exercise for the determination of Trace Metals and Methyl Mercury in Fish Homogenate IAEA-436; IAEA/AL/ 157-IAEA/MEL/77; IAEA Publications, 2006. (34) R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria; ISBN3900051-07-0; R Development Core Team, 2008; http://www.Rproject.org. (35) Zeileis, A.; Kleiber, C.; Krämer, W.; Hornik, K. Testing and Dating of Structural Changes in Practice. Comput. Stat. Data Anal. 2003, 44, 109−123.

This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected], [email protected]; phone: +34 968179432; fax: +34 968 184441. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We sincerely thank the life-work of laboratory manager Mr. Juan Guerrero (deceased in 2009) who performed the majority of the metal analyses in mussel samples included in this work. Support grants were received from the Spanish Ministry of the Environment, and Rural and Marine Environments to conduct the Mediterranean Pollution (MEDPOL) Programme, through the Spanish Institute of Oceanography − Marine Pollution Monitoring Program. We are grateful to Scott W. Fowler and J.W. Readman for valuable comments.

■

REFERENCES

(1) Goldberg, E. D. The Musssel Watch - a first step in global marine monitoring. Mar. Pollut. Bull. 1975, 6, 111. (2) Goldberg, E. D. The Mussel Watch concept. Environ. Monit. Assess. 1986, 7, 91−103. (3) Bayne, B. Watch on mussels. Mar. Pollut. Bull. 1976, 7, 217−218. (4) Goldberg, E. D.; Bertine, K. K. Beyond the Mussel Watch - new directions for monitoring marine pollution. Sci. Total Environ. 2000, 247, 165−174. (5) Tripp, B.; Farrington, J.; Goldberg, E. D.; Sericano, J. International Mussel Watch: the implementation phase. Mar. Pollut. Bull. 1992, 24, 371−373. (6) O’Connor, T. P. The National Oceanic and Atmospheric Administration (NOAA) National Status and Trends Mussel Watch Program: National Monitoring of Chemical Contamination in the Coastal United States. Chapter 19 in Environmental Statistics, Assessment and Forecasting; Cothern, C. R., Ross, N. P., Eds.; Lewis Publishers: Boca Raton, FL, 1993. (7) Review of MEDPOL Marine Monitoring Activities-Assessment of the State of the Environment and Pollution Trends; UNEP(DEPI)/MED WG.343/3; Mediterranean Action Plan, UNEP: Athens, November 2009; http://195.97.36.231/acrobatfiles/09WG343_3_eng.pdf. (8) de Mora, S.; Villeneuve, J. P.; Wyse, E. An overview of laboratory performance studies conducted by the IAEA for marine pollution studies of metals and organic contaminants. Accred. Qual. Assur. 2007, 12, 587−592. (9) Cantillo, A. Y., Ed. World Mussel Watch Database. NOAA Technical Memorandum NOS ORCA 109; NOAA: Silver Spring, MD, April 1997; http://docs.lib.noaa.gov/noaa_documents/NOS/ORCA/ TM_NOS_ORCA/nos_orca_109.pdf. (10) Tanabe, S., Ed. Mussel Watch: Marine Pollution Monitoring in Asian Waters; Center for Marine Environmental Studies (CMES) Ehime University, 2000. (11) Nicholson, M. D.; Fryer, R. J.; Ross, C. A. Designing monitoring programmes for detecting temporal trends in contaminants in fish and shellfish. Mar. Pollut. Bull. 1997, 34, 821−826. (12) Nicholson, M. D.; Fryer, R. J. The statistical power of monitoring programmes. Mar. Pollut. Bull. 1992, 24, 146−149. (13) Gilbert, R. O. Statistical Methods for Environmental Pollution Monitoring; Van Nostrand Reinhold: New York, 1987. (14) Lohr, S. L. In Sampling: Design and Analysis; Julet, M., Ed.; Brooks/Cole: USA, 2010. (15) O’Connor, T. P. Trends in Chemical concentrations in mussels and oysters collected along the US coast from 1986 to 1993. Mar. Environ. Res. 1996, 41, 183−200. 11522



Article

(36) Zeileis, A.; Leisch, F.; Hornik, K.; Kleiber, C. strucchange: An R Package for Testing for Structural Change in Linear Regression Models. J. Stat. Softw. 2002, 7, 1−38. (37) Andrews, D. W. K. Tests for parameter instability and structural change with unknown change point. Econometrica 1993, 61, 821−856. (38) Erdman, C.; Emerson, J. W. bcp: An R Package for performing a Bayesian Analysis of Change Point Problems. J. Stat. Softw. 2007, 23, 1−13. (39) Directive 98/70/EC of the European Parliament and of the Council of 13 October 1998 Relating to the Quality of Petrol and Diesel Fuels; EU, 1998; http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri= CELEX:31998L0070:EN:NOT. (40) Alexander, R. B.; Smith, R. A. Trend in lead concentrations in major U.S rivers and their relation to historical changes in gasolinelead consumption. Water Resour. Bull. 1988, 24, 557−569. (41) Elbaz-Poulichet, F.; Dezileau, L.; Freydier, R.; Cossa, D.; Sabatier, P. A 3500-Year Record of Hg and Pb Contamination in a Mediterranean Sedimentary Archive (The Pierre Blanche Lagoon, France). Environ. Sci. Technol. 2011, 20, 8642−8647. (42) Gibraltar Area Quality Environmental Plan (1997−2002). Consejeria de Agricultura, Pesca y Medio Ambiente; http://www. juntadeandalucia.es/medioambiente/site/portalweb/menuitem. 7e1cf46ddf59bb227a9ebe205510e1ca/?vgnextoid= 337fe126cf415310VgnVCM2000000624e50aRCRD&vgnextchannel= 4e44e6f6301f4310VgnVCM2000000624e50aRCRD. (43) Borchardt, T.; Burchert, S.; Hablizel, H.; Karbe, L.; Zeitner, R. Trace metal concentrations in mussels: comparison between estuarine, coastal and offshore regions in the south eastern North Sea from 1983 to 1986. Ecol. Prog. Ser. 1988, 42, 17−31. (44) Swaileh, K. M.; Adelung, D. Levels of trace metals and effect of body size on metal content and concentration in Artica islandica L. (Mollusca, Bivalvia) from Kiel Bay, Western Baltic. Mar. Pollut. Bull. 1994, 28, 500−505. (45) Kim, Y.; Powell, E. N.; Wade, T. L.; Presley, B. J. Relationship of parasites and pathologies to contaminant body burden in sentinel bivalves: NOAA Status and Trends ‘Mussel Watch’ Program. Mar. Environ. Res. 2008, 65, 101−127. (46) Wu, R. S. S.; Lau, T. C.; Fung, W. K. M.; Ko, P. H.; Leung, K. M. Y. An ‘artificial mussel’ for monitoring heavy metals in marine environments. Environ. Pollut. 2007, 145, 104−110. (47) Gonzalez-Rey, M.; Lau, T. C.; Gomes, T.; Maria, V. L.; Bebianno, M. J.; Wu, R. S. S. Comparison of metal accumulation between ‘Artificial Mussel’ and natural mussels (Mytilus galloprovincialis) in marine environments. Mar. Pollut. Bull. 2011, 63, 149−153. (48) Morrisey, D. J. Environmental Impact Assessment − A review of its aims and recent developments. Mar. Pollut. Bull. 1993, 540−545. (49) de Bruyn, S. M.; van den Bergh, J. C. J. M.; Opschoor, J. B. Economic growth and emissions: reconsidering the empirical basis of environmental Kuznets curves. Ecol. Econ. 1998, 161−175. (50) Thebault, H.; Baena, A.M.R.Y.; Andral, B.; Barisic, D.; Albaladejo, J. B.; Bologa, A. S.; Boudjenoun, R.; Delfanti, R.; Egorov, V. N.; Khoukhi, T. E.; Florou, H.; Kniewald, G.; Noureddine, A.; Patrascu, V.; Pham, M. K.; Scarpato, A.; Stokozov, N. A.; Topcuoglu, S.; Warnau, M. 137Cs baseline levels in the Mediterranean and Black Sea: A cross-basin survey of the CIESM Mediterranean Mussel Watch programme. Mar. Pollut. Bull. 2008, 57, 801−806. (51) Soriano, J. A.; Vinas, L.; Franco, M. A.; Gonzalez, J. J.; Nguyen, M. H.; Bayona, J. M.; Albaiges, J. Spatial and temporal trends of polycyclic aromatic hydrocarbons in wild mussels from the Cantabrian coast (N Spain) after the Prestige oil spill. J. Environ. Monit. 2007, 9, 1018−1023. (52) Fernandez, B.; Albentosa, M.; Vinas, L.; Franco, A.; Gonzalez, J. J.; Campillo, J. A. Integrated assessment of water quality of the Costa da Morte (Galicia, NW Spain) by means of mussel chemical, biochemical and physiological parameters. Ecotoxicology 2010, 19, 735−750.

11523


(MW) Databases - ACS Publications - American Chemical Society

Recommend Documents