Resumes & CVs

A Statistical Resampling Method To Calculate Biomagnification Factors Exemplified with Organochlorine Data from Herring ( Clupea harengus ) Muscle and Guillemot ( Uria aalge ) Egg from the Baltic Sea

A Statistical Resampling Method To Calculate Biomagnification Factors Exemplified with Organochlorine Data from Herring ( Clupea harengus ) Muscle and Guillemot ( Uria aalge ) Egg from the Baltic Sea
of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  A Statistical Resampling Method ToCalculate Biomagnification FactorsExemplified with OrganochlorineData from Herring ( Clupea harengus ) Muscle and Guillemot( Uria aalge ) Egg from the BalticSea K A T R I N L U N D S T E D T - E N K E L , *  , † , ‡ M A T ST Y S K L I N D ,  § J O H A N T R Y G G ,  | P E T E RS C H U ¨  L L E R ,  + L I L L E M O R A S P L U N D ,  @ U L L A E R I K S S O N ,  @ L I S B E T H H A  ¨  G G B E R G ,  @ T J E L V A R O D S J O ¨  ,  ‡ M A T S H J E L M B E R G ,  ‡ M A T S O L S S O N ,  ‡ , @  A N D J A N O ¨  R B E R G  † Environmental Toxicology, Department of Physiology and Developmental Biology, Evolutionary Biology Centre,Uppsala University, Norbyva¨gen 18A, SE-752 36, Sweden, and Contaminant Research Group, Swedish Museum of Natural History, P.O. Box 50007, SE-104 05 Stockholm, Sweden,Environmental Chemistry and Research Group for Chemometrics, Department of Chemistry, Ume  å  University,SE-901 87 Ume  å , Sweden, InfiDyne AB, Murargatan 30A,SE-754 37 Uppsala, Sweden, and Institute of Applied Environmental Research, Stockholm University,SE  - 106 91 Stockholm, Sweden A novel method for calculating biomagnification factors ispresentedanddemonstratedusingcontaminantconcentrationdata from the Swedish national monitoring programregarding organochlorine contaminants (OCs) in herring( Clupea harengus  ) muscle and guillemot ( Uria aalge  ) egg,sampled from 1996 to 1999 from the Baltic Sea. With thisrandomlysampledratios(RSR)method,biomagnificationfactors (BMF RSR ) were generated and denoted withstandard deviation (SD) as a measure of the variation.The BMF RSR  were calculated by randomly selecting oneguillemot egg out of a total of 29 and one herring out of a total of 74, and the ratio was determined between theconcentrationofagivenOCinthateggandtheconcentrationof the same OC in that herring. With the resampling technique, this was performed 50 000 times for any givenOC, and from this new distribution of ratios, BMF RSR  foreach OC were calculated and given as geometric mean (GM)with GM standard deviation (GMSD) range, arithmeticmean (AM) with AMSD range, and minimum (BMF MIN ) aswell as maximum (BMF MAX ) biomagnification factors. The 14analyzed OCs were  p  , p  ′ DDT and its metabolites  p  , p  ′ DDEand  p  , p  ′ DDD, polychlorinated biphenyls (PCB congeners:CB28, CB52, CB101, CB118, CB138, CB153, and CB180),hexachlorocyclohexane isomers ( R -,   -, and  γ HCH), andhexachlorobenzene(HCB).Multivariatedataanalysis(MVDA)methods, including principal components analysis (PCA),partial least squares regression (PLS), and PLS discriminantanalyses (PLS-DA), were first used to extract informationfrom the complex biological and chemical data generatedfrom each individual animal. MVDA were used to modelsimilarities/dissimilarities regarding species (PCA, PLS-DA),sample years (PLS), and sample location (PLS-DA) togive a deeper understanding of the data that the BMFmodeling was based upon. Contaminants that biomagnify, that had BMF RSR  significantly higher than one, were p  , p  ′ DDE, CB118, HCB, CB138, CB180, CB153,   HCH, andCB28.Thecontaminantsthatdidnotbiomagnifywere p  , p  ′ DDT, p  , p  ′ DDD, R HCH,CB101,andCB52.Eventualbiomagnificationfor  γ HCH could not be determined. The BMF RSR  for OCspresent in herring muscle and guillemot egg showed a broadspan with large variations for each contaminant. To beable to make reliable calculations of BMFs for differentcontaminants, we emphasize the importance of using databased upon large numbers of, as well as well-defined,individuals. Introduction The aim of this paper is to introduce a novel method tocalculate biomagnification factors (BMF), the randomly sampled ratios (RSR) method that denotes BMFs with anestimateofvariation.Withthemethodcurrentlyinuse,BMFsarecalculatedastheratiobetweenthearithmeticmean(AM)or geometric mean (GM) concentration of a contaminant inthe predator and the AM or GM concentration of the samecontaminant in the prey. Therefore, these BMFs are alwaysdenoted as a single value with no estimate of the variation.There are many papers that deal with the issue of error/uncertainty propagation regarding bioaccumulation of con-taminants, for instance, to evaluate how model outputchanges with variations in input variables ( 1 ,  2  ) and/or how uncertainties in individual parameters can affect uncertain-ties in the results ( 3 ,  4 ). These papers use Bayesian meth-odologies (e.g., Monte Carlo or Marcov Chain Monte Carlosimulations) to generate new distributions of values for, forexample,contaminantconcentrationsusedasinput/outputvalues in the modeling. Often, such papers generate new distributions after a low number of analyzed samples fromshort-termexperiments(i.e.,thegenerateddistributionsarenotbasedonthedistributionofcontaminantsinthenaturalenvironment). The U.S. EPA recommends the use of tech-niques to clarify the impact/sources of uncertainty on theresult of an ecological risk assessment but at the same timecautionsthatthepoorexecutionofanymethodcanobscurerather than facilitate understanding ( 5  ). In fact, Linkov andBurmistrov ( 6  ) found that modeler uncertainty and choicesmade by modelers contributed to as high as 7 orders of magnitude differences in model predictions.In a food-web of a specific ecosystem, nutrients aretransportedbetweenabioticandbioticcompartmentsaswellas between individual organisms within the biotic compart-ment.Contaminantsaretransportedinasimilarwayforming a contaminant-web, where fluctuations in concentrationsare caused by variations in biotic factors (e.g., species com-position, age, and health status of individuals) as well as inabiotic factors (e.g., temperature, wind, and precipitation). * Corresponding author phone:  + 46 18 4716498; fax:  + 46 18518843; e-mail: † Uppsala University. ‡ Swedish Museum of Natural History. § Environmental Chemistry, Department of Chemistry, Ume å University. | Research Group for Chemometrics, Department of Chemistry,Ume å  University. + InfiDyne AB. @ Stockholm University. Environ. Sci. Technol.  2005 ,  39,  8395 - 8402 10.1021/es048415y CCC: $30.25  󰂩  2005 American Chemical Society VOL. 39, NO. 21, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY  9 8395 Published on Web 09/21/2005   All together, this leads to variations in contaminant con-centrations in any given individual at any given time ( 7  - 9  ). When modeling contaminant levels, transport, and bio-magnification in an ecosystem, a thorough knowledge of variations in concentrations is important. Without this,modeling is difficult to perform at best, or glaringly faulty at worst ( 6  ,  8  ,  10  - 17  ). Biomagnification.  As our present knowledge regarding quantitativedataonuptakeroutes(e.g.,gastrointestinaltract,respiratoryorgans,andintegument)ofcontaminantsinmostanimalsinthewildislimited,thispaperusesawidedefinitionofbiomagnification( 18  ).Ifthelipidnormalizedconcentrationof a substance is significantly higher in the predator than inthe prey, regardless of how that higher concentration hasbeen achieved, biomagnification of that substance hasoccurred.Properties of the contaminant essential for biomagnifi-cation,mainlybioavailabilityandpersistence,aredeterminedby a number of physical/chemical characteristics such asmolecularsize,structure( 19  -  21 ),presenceofreactivegroups,lipophilicity,andwatersolubility.Othervariablesimportantforbiomagnificationarerelatedtotheanimals: thepredatorand its prey (e.g., feeding habits and habitats, lipid charac-teristics and fat content, sex, age, and the animals’ capacity to biotransform/excrete contaminants). The fact that con-taminantsbiomagnifytovariousdegrees(  22  )ultimatelyleadsto specific contaminant patterns in different animal species withinthesameecosystem.Theultimateaimofourresearchis to study and model contaminants’ physical - chemicalproperties,includingtheirstructures,inrelationtothevarying contaminant patterns in different animal classes to gain aninsight to biomagnification processes (i.e., factors favoring or counteracting biomagnification). To reach this goal, wefirsthavetodeterminereliableBMFswithanestimateofthevariation.In our present article, we have used large numbers of individuallyanalyzedherring( Clupeaharengus  )muscleandguillemot ( Uria aalge  ) egg from the Swedish monitoring program (  23 ) carried out in the Baltic Sea to calculatebiomagnification factors and to quantify the uncertainty inthese calculations. The eggs are sampled from a bird colony  whose feeding area covers the two sampling sites where theherringarecaught;thus,theanalyzedanimalssrcinatefromthe same contaminant-web; see Materials and Methods.Difficultiesmightarisewithmigratorybirdswherethebirds’exposure history is unknown for parts of the year (  24 ). ForguillemotfromtheStoraKarlso¨ colony,whichisastationary bird residing in the Baltic Sea all year-round, this problemdoes not arise. Only a few articles present calculations of BMFs using data from the biota of the Baltic Sea. On thebasisofthelevelsofpolychlorinateddibenzo-dioxins/furans(PCDD/F) in herring muscle and guillemot eggs, de Wit etal. reported a BMF of   ∼ 16 for PCDD/F (  25  ). Materials and Methods Herring. Regardingtransportofenvironmentalcontaminantsin the Baltic Sea food-web, herring ( C. harengus  ) is animportant species; thus, a number of herring specimens aresampled each year from the Baltic Sea (  23 ,  26  -  28  ). It is aplanktivorousfishthatfeedsonselectedpelagiczooplankton(  29  - 31 ), and in turn, herring play an important role as foodfor piscivore animals such as other fish (e.g., cod ( Gadus morhua ) and salmon ( Salmo salar  )), birds (e.g., guillemot( U. aalge  )), and mammals (e.g., grey seal ( Halichoerus  grypus  ))andisanimportanthumanfoodsource( 32  ).Herring specimensusedforthisarticleweresampledfromfishcatchesat two sites, Landsort (L) and Utla¨ngan (U) (  23 ). L and U aresituatedtothenorth(L)andtothesouth(U)oftheguillemotcolony at Stora Karlso¨ (S), and both herring locations are withinthosebirds’largeforagingarea( 33 ).Theherringweresampled autumn (October to November) of 1996, 1997, and1998, and a total of 72 female fish (12 individuals from eachsite, each year) 3 - 4 years of age were later individually analyzed regarding their dorsal muscle concentration of anumber of OCs. All collecting and sample preparation wascarried out and recorded in a standardized manner (  23 ,  34 ).(The sampling procedures as well as the resulting biologicaldataregardinganimalcharacteristicsarepresentedindetailin the Supporting Information). Guillemot.  Guillemot ( U. aalge  ) is a bird species thatresides year-round in the Baltic Sea; it is a piscivore thatfeeds mainly on the two clupeid fish species, herring andsprat( Sprattussprattus  )(  23 , 35  - 37  ).Allguillemoteggswerecollected from the guillemot colony at the island of StoraKarlso¨ (S)(  23 ).Ataround5yearsofage,theguillemotfemalebreeds for the first time, laying one large ( ∼ 110 g) egg per year. The high age at first egg laying, together with the factthatguillemotinvestlargequantitiesoflipidsrelativetotheirbody weight of   ∼ 1000 g to the one egg they lay each year,means that the contaminant composition of the egg mainly reflects the composition of maternal tissue rather than thefemale’s diet at time of yolk formation ( 38  ). The guillemoteggs were from spring (April to May) 1997, 1998, and 1999. A total of 30 guillemot eggs (10 eggs/year) were laterindividually analyzed regarding their (yolk and albumen)concentration of a number of OCs. One guillemot egg from1999waslostduringsamplepreparation,sothetotalnumberof guillemot eggs analyzed for the 3 years were 29 (seeSupporting Information).  Analyzed OCs.  Chemical analyses were carried out at anaccredited laboratory (Institute for Applied EnvironmentalResearch (ITM) at Stockholm University), according topreviously described procedures ( 39  ,  40  ). For both herring andguillemotsamples,thefollowing14OCswereanalyzed:2,2-bis(4-chlorophenyl)-1,1,1-trichloroethane compounds( p , p ′ DDT and its metabolites  p , p ′ DDE and  p , p ′ DDD); poly-chlorinatedbiphenyls(PCBs)withthecongeners’respectiveIUPAC number within parentheses: 2,4,4 ′ -trichlorobipenyl(CB28), 2,2 ′ ,5,5 ′ -tetrachlorobiphenyl (CB52), 2,2 ′ ,4,5,5 ′ -pen-tachlorobiphenyl (CB101), 2,3 ′ ,4,4 ′ ,5-pentachlorobiphenyl(CB118),2,2 ′ ,3,4,4 ′ ,5 ′ -hexachlorobiphenyl(CB138),2,2 ′ ,4,4 ′ ,5,5 ′ -hexachlorobiphenyl (CB153), and 2,2 ′ ,3,4,4 ′ ,5,5 ′ -heptachlo-robiphenyl(CB180);hexachlorocyclohexaneisomers( R -,   -,and  γ HCH); and hexachlorobenzene (HCB). All compounds could be separated from one another, as well as from other PCBs of importance, except for CB138, where the interfering peak caused by CB163 means thatCB138 can be overestimated by up to 20 - 30%. To facilitatereading, CB138 is henceforth written alone. (The chemicalanalysesaswellastheobtainedresiduelevelsarepresentedin detail in the Supporting Information.) Missing Values.  When calculating the various biomag-nificationfactors,thelevelofquantification(LOQ)wasusedforafewsampleswithnonquantifiablelevelsofanOC.Whenpossible for those samples, LOQ values determined fromeachindividualchromatogramwereused,denotedasLOQ IND .For guillemot egg, this was not always possible, and then ageneral LOQ value was used. The LOQ was defined as 10times the standard deviation of the measurements, and inpractice, this was estimated by repetitive measurements of asamplecontaininganalytesataconcentrationclosetotheexpectedLOQ( 41 ).ThesegeneralLOQswerefor p , p ′ DDT50ng/g lw,  p , p ′ DDD 60 ng/g lw, R -HCH 30 ng/g lw,  γ -HCB 40ng/g lw, CB52 30 ng/g lw, and finally, CB101 35 ng/g lw. Statistics.  For basic statistics regarding the biologicalvariablesandconcentrationsofOCs,thesoftwareGraphPadPrism4.03( 42  )wasused.Multivariatedataanalyses(MVDA)(i.e.,principalcomponentanalysis(PCA),partialleast-squaresregression (PLS), and PLS discriminant analyses (PLS-DA)) were performed using the software SIMCA-P 10.0.4 ( 43 ). 8396  9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 21, 2005  PCAwasusedtoillustrateandtodiscerngroupingsinthedata, a significance level of 0.05 was used, and centered andscaled data (to variance 1) were fitted to a principalcomponents (PC) model ( 44 ). Values of the explainedvariation,  R 2 , and predicted variation,  Q 2 , were calculated,and R 2 values > 0.7 and Q 2 values > 0.4 denote an acceptablemodel when analyzing biological data ( 45  ).PLSwasusedtodeterminewhetherthereweresignificantdifferences in pollution patterns between the 3 years, forherring and guillemot separately. PLS is an extension of multiplelinearregressionscombinedwithPCAtomodeltherelationship between two matrixes,  Y   and  X  , that can bothbe multidimensional. This is performed by modeling   X   and Y  separatelyandthenrelatingtheirrespectivescorestoeachother ( 46  ,  47  ).PLS-DA was used to model contaminant patterns todiscerndifferencesbetweenthetwoherringlocationsaswellas between the two species. PLS-DA is used to determine whetheranobservationbelongstoapreestablishedclassby modeling and quantifying the eventual discriminating vari-ables that contribute to the class (e.g.,  X  / Y   matrixes)separation ( 48  ). Calculation of Biomagnification Factors (BMFs).  Tocalculate BMFs, different methods were used. By the firstmethod,themethodcurrentlyinuse,BMFswerecalculatedastheratiobetweenthegeometricmean(GM)aswellasthearithmetic mean (AM) concentration of a given OC inguillemot eggs for the 3 years 1997, 1998, and 1999 ( n ) 29)andtheGMaswellastheAMconcentrationofthesameOCin herring from both locations L and U for the 3 years 1996,1997,and1998( n ) 72).ThismethodgivessingleBMFvaluesfor each OC, which we denoted as BMF GM  and BMF  AM . With the new randomly sampled ratios (RSR) method,BMFs were calculated by randomly selecting one guillemotegg out of 29 and one herring out of 72, using the BMF RSR computerprogramespeciallydesignedforthetask( 49  ).Thisprogram uses a random sampling with a replacement tech-nique where a single specimen is randomly sampled andcan be sampled several times. The ratio between the con-centrations of a given OC in that sampled guillemot egg andthesameOCinthatsampledherringmusclewasdetermined.In this way, 50 000 iterations were performed, generating 50000 observations of BMF RSR  for each contaminant, andgeometric mean (GM, BMF RSRGM ) with geometric meanstandarddeviation(GMSD)aswellasarithmeticmean(AM,BMF RSRAM )witharithmeticmeanSD(AMSD)werecalculatedfor this new distribution of ratios. BMF RSRGM  as well asBMF RSRAM ranges( ( 1SD)werethencalculated.ForBMF RSRGM ,the lower limit is calculated as BMF RSRGM /GMSD and theupper limit as BMF RSRGM  * GMSD. For BMF RSRAM , the lowerlimit is calculated as BMF RSRAM -  AMSD and the upper limitasBMF RSRAM +  AMSD.Standarddeviationisameasureofthedegree of dispersion of the data from the mean value. Forapproximatelynormallydistributeddata,asinthiscaseafterlog transformation of the quotients, then about 68.27% of the values are within  ( 1 SD around the mean, and about95.45% of the values are within  ( 2 SD around the mean. Also, minimum and maximum BMFs for all organochlo-rines were determined. These BMFs were calculated as theratiobetweentheguillemoteggwiththelowestconcentrationof a given contaminant and the herring muscle with thehighest concentration of the same contaminant (BMF MIN )andbetweentheguillemoteggwiththehighestconcentrationand the herring muscle with the lowest concentration(BMF MAX   ) (Figure 1). Number of Iterations.  To discern variations in BMF RSR due to the numbers of iterations (number of randomly selected pairs), a series of different numbers of iterations (3,5, 10, 50, 100, 500, 1000, 5000, 10 000, 50 000, and 100 000) was performed 10 times (e.g., 10 runs with 3 iterations, 10runs with 5 iterations...10 runs with 100 000 iterations). This was done using the CB153 concentration in herring muscleandguillemotegg.Thecoefficientofvariation(CV%)betweenthe 10 resulting BMF RSRGM  from the 10 runs with the samenumber of iterations was calculated. This was also done inthe same way for BMF RSRAM . Results Biological Variables.  The herring ( n  )  72) had (GM with95% CI) a fat content (F%) of 2.0% (1.8 - 2.3%), body weight(BW) of 33.7 g (32.0 - 35.5 g), and age of 3.7 years (3.6 - 3.8 years). The guillemot egg ( n ) 29) had F% of 11.8% (11.3 - 12.3%) and total weight/egg of 110 g (107 - 113 g). (Themeasured biological variables are presented in detail in theSupporting Information.) Concentrations. TheOCwiththehighestlevelinherring musclewas p , p ′ DDEwithaconcentration(GMwith95%CI)of 370 ng/g lw (320 - 430 ng/g lw), and this OC also showedthe highest concentration in guillemot egg of 21 200 ng/g lw (19400 - 23200ng/glw).AmongtheOCsanalyzed,thelowestconcentration in herring muscle was found for CB28 that was 7.0 ng/g lw (6.2 - 8.0 ng/g lw), and here, 23 out of thetotalof72individualshadlevelsbelowtheLOQ.Inguillemotegg,  p , p ′ DDT concentrations were below the LOQ (50 ng/g lw) in all samples. (The concentrations of organochlorinesinherringmuscleandguillemoteggsarepresentedindetailin the Supporting Information.) Multivariate Data Analysis.  The PCA using the wholedata set, with guillemot egg and herring muscle chemicalvariables (analyzed OCs) as  X  , gave a model ( R 2  X  ) 0.83,  Q 2 ) 0.74, two components), indicating that the species makeuptwoseparategroups(Figure2a),meaningthatthecontam-inant pattern differed between the two species (Figure 2b).PLS-DA with herring versus guillemot ( R 2  X  ) 0.78,  R 2 Y  ) 0.95,  Q 2 ) 0.94, two components) showed complete separa-tion between the two species (score plot not shown, similartoFigure2a),andthecoefficientplot(Figure3)showedthat p , p ′ DDE,   HCH, HCB, CB28, CB118, CB138, CB153, andCB180 all had higher concentrations in guillemot egg, while p , p ′ DDD and CB101 had higher concentrations in herring muscle. DifferencesbetweenYears? For herring muscle, the PLSmodel with the respective years as  Y   and biological as wellaschemicalvariablesas  X  revealednosignificantdifferencesin the variables between the 3 years ( R 2  X  ) 0.55,  R 2 Y  ) 0.23, Q 2 )  0.02, two nonsignificant components). For guillemotegg, the corresponding PLS model ( R 2  X  ) 0.36,  R 2 Y  ) 0.87, FIGURE1. Concentrationonalogarithmicscale(ng/goflipidweight)ofCB153inherring( C.harengus  )muscle( n  ) 72)fromLandsortandUtla  1  nganandinguillemot( U.aalge  )egg( n  ) 29)fromStoraKarlso  1  , the Baltic Sea, 1996 - 1999. Biomagnification factors (BMFs)calculated as BMF MIN  (3.0), BMF GM  (24.7), and BMF MAX  (158) (seeMaterials and Methods and Table 1). VOL. 39, NO. 21, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY  9 8397  Q 2 )  0.64, two components) showed that there were sys-tematic and significant differences between the three years.The eggs from 1997 and 1998 had a significantly lower con-centration of HCB than the eggs from 1999 and also a sig-nificantly higher concentration of other OCs. (PLS plots notshown, for concentrations, see Supporting Information.) Differences between Herring from Landsort andUtla   1  ngan?  The PLS-DA model ( R 2  X  ) 0.66,  R 2 Y  ) 0.65,  Q 2 ) 0.55, three components) based on herring data (biologicalas well as chemical variables) from the two locations L andUshowedthatthepopulationsfromthetwolocationsoverlapto some small extent but that differences were discernible(Figure 4a). Herring from L were significantly older and hada lower body weight (BW) than fish from U (Figure 4b). TheOCs with significantly higher concentrations in fish from LthaninfishfromUwere R -HCH,   -HCH,CB101,andCB118(Figure 4b). The  p , p ′ DDT concentrations were significantly higher in fish from U than in fish from L (Figure 4b). Biomagnification Factors.  The BMFs calculated using different methods are presented in Table 1. The OCs thatbiomagnify,withBMF RSRGM significantlyhigherthan1,werein the following decreasing order:  p , p ′ DDE, CB118, HCB,CB138,CB180,CB153,   HCH,andCB28.TheOCsthatshowedvalues below 1 for the calculated BMF RSRGM  and did not bio-magnify were  p , p ′ DDD,  p , p ′ DDT, R HCH, CB52, and CB101. Number of Iterations.  When the series with increasing numberofiterations(3,5...100000)wasperformed10timesusing the CB153 concentration in herring muscle andguillemot egg, the variation between the 10 resulting BMF RSRAM  as well BMF RSRGM  values from the 10 runs withsame number of iterations showed similar patterns. Thecoefficient of variation (CV%) rapidly declined with anincreaseinnumberofiterations,fromaCV%of57%betweenthe 10 runs with 3 iterations to 0.5% or below between the10 runs with 10 000 iterations or more (Figure 5a). Whenusing a higher number of iterations, the resulting BMF RSRGM converge to one consistent value (equal to BMF GM ). Accord-ingly,whenusingahighernumberofiterations,theresulting GMSD and AMSD converge to one consistent value (shownfor GMSD in Figure 5b). Discussion ModelingBiomagnificationProcesses. Oneimportantissueto consider when modeling biomagnification processes is FIGURE 2. Principal component analysis (PCA) based on concen- trations of contaminants in guillemot ( U. aalge  ) egg ( n  ) 29) fromStora Karlso  1   (S) and herring ( C. harengus  ) muscle ( n   )  72) fromLandsort and Utla  1  ngan (L and U) from the Baltic Sea, 1996 - 1999.PCA model ( R  2 X  ) 0.83 and  Q  2 ) 0.74, two components). (a) Scoreplot and (b) loading plot. For abbreviations of contaminants, seeMaterials and Methods.FIGURE 3. Coefficient plot with 95% CI for the respective variablesfrom PLS-discriminant analysis (PLS-DA) based on concentrationsof contaminants in guillemot ( U. aalge  ) egg ( n   )  29) from StoraKarlso  1  (S) and herring ( C. harengus  ) muscle ( n  ) 72) from Landsortand Utla  1  ngan from the Baltic Sea, 1996 - 1999. PLS-DA model forguillemot vs herring ( R  2 X   )  0.78,  R  2 Y   )  0.95,  Q  2 )  0.94, twocomponents). For abbreviations, see Materials and Methods.FIGURE4. PLS-discriminantanalysis(PLS-DA)basedonbiologicalmeasurements and concentrations of contaminants in herring ( C.harengus  ) muscle ( n   )  72) from Landsort and Utla  1  ngan (L and U)from the Baltic Sea, 1996 - 1998. PLS-DA model for L vs U ( R  2 X   ) 0.66,  R  2 Y  ) 0.65,  Q  2 ) 0.55, three components). (a) Score plot and(b) coefficient plot for herring from L with 95% CI for the respectivevariables, weight (BW), length (BL), condition factor (ConF), andlipid and dry matter content (F% and DM%, respectively). Forabbreviations of contaminants, see Materials and Methods. 8398  9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 21, 2005   t  h  ewi   l    d  ani   m al    s ’    d i    e t   a s  t  h  e ex a c  t   c  om  p o s i    t  i    oni    s  un k  n own .I  n t  h  e  pr  e s  en t   s  t   u d   y , t  h  eB MF  s  ar  e b  a s  e d  on d  a t   af   r  omh  er r i   n  g , al    t  h  o u  gh  t  h  e d i    e t   of     g ui   l   l    em o t  i    s  ami   x ,m ai   nl     y of   h  er r i   n  g an d  s   pr  a t   (   2   3    ,  3   5   -  3   7    )   .H er r i   n  g an d  s   pr  a t   ar  e s i   mi   l    ar i   nm or   ph  ol   - o  g  y ,f    e e d i   n  g b  eh  avi    or  , an d  e c  ol    o  g  y (   e .  g . , t  h  e t  w o s   p e c i    e s  s h  ar  e t  h  e s  am ef   i    s h  s h  o al    s  ,h  av e s i   mi   l    ar f    e e d   pr  ef    er  en c  e s  a s w el   l    a s  s i   mi   l    ar  d i    ur n al   r h   y t  h m s i   nv er  t  i    c  al   mi     gr  a t  i    onwh  enf    e e d i   n  g (    3   0    ,  5   0   -  5  2    )   )   .A n um b  er  of    s  t   u d i    e s h  a s  s h  own t  h  a t   t  h  e c  on t   ami   n an t  l    ev el    s i   nh  er r i   n  g an d  s   pr  a t   ar  er  em ar  k   a b l     y s i   mi   l    ar  (    5   3   -  5   5    )   .H ow ev er  , d  u e t   o d i   f   f    er  en c  e s i   n  gr  ow t  h r  a t   e , a s   pr  a t   of    a  gi   v en s i   z  emi     gh  t   b  e ol    d  er  an d h  av eh i     gh  er  c  on c  en t  r  a t  i    on s  of    c  on t   ami   n an t   s  t  h  an ah  er r i   n  g of    t  h  e s  am e s i   z  e (    5   5    )   , an d  t  h  er  ef    or  e ,i    t  i    s   p o s  s i    b l    e t   o ov er  e s  t  i   m a t   eB MF  s f    or  t  h  e s  t   e  ph  er r i   n  g t   o  g ui   l   l    em o t  i   f    t  h  e  g ui   l   l    em o t   d i    e t  w a s  t   o al    ar   g e ex t   en t   c  om  pr i    s  e d  of    ol    d  s   pr  a t   .I   d  e al   l     y ,f    or f    u t   ur  e TABLE 1. Biomagnification Factors (BMF) (BMF GM , BMF AM , BMF RSRGM , BMF RSRAM , BMF MIN , and BMF MAX ) as the Ratio of Concentration of Organochlorines between Herring ( Clupea harengus ) (Ch)Muscle,  n  )  72, from Landsort and Utla  1  ngan, and Guillemot ( Uria aalge ) (Ua) Egg,  n  )  29, from Stora Karlso  1   in the Baltic Sea, See Materials and Methods  a Σ DDT  Σ PCB  p  , p  DDE  p  , p  DDD  p  , p  DDT  r HCH   HCH  γ HCH HCB CB28 CB52 CB101 CB118 CB138 CB153 CB180 BMF GM b  36.0 24.6 57.1 0.42 0.93 1.37 18.6 2.10 33.8 12.5 1.96 0.46 42.5 29.9 24.7 27.8BMF AM c  31.3 20.9 47.4 0.35 0.76 1.35 18.7 2.08 30.0 10.3 1.46 0.39 36.1 25.3 21.0 21.8BMF RSRGM d  36.0 24.6 57.1 0.42 0.93 1.38 18.6 2.10 33.8 12.5 1.96 0.46 42.5 30.0 24.7 27.8GMSD e  1.86 1.97 1.98 1.88 1.96 1.27 1.47 1.16 1.75 1.74 2.10 1.82 1.90 1.98 2.04 2.25BMF RSRGM  range f  19.4 - 66. 9 12.5 - 48.4 28.8 - 113.1 0.22 - 0.79 0.48 - 1.82 1.08 - 1.75 12.7 - 27.3 1.82 - 2.43 19.3 - 59.2 7.18 - 21.8 0.94 - 4.12 0.25 - 0.84 22.3 - 80.8 15.1 - 59.4 12.1 - 50.4 12.3 - 62.6BMF RSRAM g  43.7 30.5 71.3 0.52 1.21 1.42 20.1 2.12 40.1 14.2 2.55 0.55 51.6 37.1 31.2 37.4AMSD h 31.0 20.0 49.2 0.42 1.15 0.34 8.78 0.32 26.2 6.65 2.00 0.33 32.1 23.9 21.4 29.5BMF RSRAM  range i  12.7 - 74. 7 10.4 - 50.5 22.2 - 120.5 0.10 - 0.94 0.06 - 2.35 1.08 - 1.75 11.3 - 28.9 1.81 - 2.44 13.8 - 66.3 7.54 - 20.8 0.55 - 4.56 0.21 - 0.88 19.5 - 83.6 13.2 - 61.0 9.83 - 52.6 7.89 - 67.0BMF MIN  j  6.94 3.77 9.30 0.10 0.22 0.52 7.17 1.11 11.0 0.79 0.17 0.11 8.94 3.65 3.04 2.24BMF MAX k  328 156 403 3.22 8.09 2.63 84.4 3.24 166 47.3 12.5 2.44 205 166 158 287BM l  yes yes yes no no no yes  m  yes yes no no yes yes yes yes a  Theanimalmaterialwascollectedbetween1996and1999.  b  BMF GM ;calculatedwithgeometricmean(GM)concentrationsforrespectiveOCinUa/Ch.  c  BMF AM ;calculatedwitharithmeticmean(AM)concentrationfor respective OC in Ua/Ch.  d  BMF RSRGM ; calculated with the randomly sampled ratios (RSR) method using 50 000 iterations, resulting in GM ( GMSD.  e  GMSD; geometric mean standard deviation.  f  BMF RSRGM  range;lowerlimit(BMF RSRGM  /GMSD)andupperlimit(BMF RSRGM *GMSD).  g  BMF RSRAM ;calculatedwithRSRmethodusing50000iterations,resultinginAM ( AMSD.  h AMSD;arithmeticmeanstandarddeviation.  i  BMF RSRAM range; lower limit (BMF RSRAM - AMSD) and upper limit (BMF RSRAM + AMSD).  j  BMF MIN ; minimum BMF, see Materials and Methods.  k  BMF MAX ; maximum BMF, see Materials and Methods.  l  BM; yes ) biomagnification(BM) of contaminant with BMF significantly ( p   <  0.05) higher than 1 and no  )  no BM of contaminant, BMF not significantly higher than 1.  m Eventual BM cannot be determined, too high a level of quantification(LOQ) for  γ -HCH in guillemot egg. F  I    G  U R  E   5  .E  f   f    e  c  t    o f   i   n  c r   e  a  s i   n   g  t   h   e n  u m b   e r   o f   i    t    e r   a  t   i    o n  s  o n  t   h   e  c  a l    c  u l    a  t    e  d   b  i    o m a   g n i   f   i    c  a  t   i    o n f    a  c  t    o r   (   B MF   R   S  R   G M  )    a  s w e l   l    a  s  o n  t   h   e  s i   z  e  o f    t   h   e  s  t    a n  d   a r   d   d   e v i    a  t   i    o n  (    G M S  D  )    b   a  s  e  d   o n  C  B 1   5   3   c  o n  c  e n  t   r   a  t   i    o n i   n h   e r  r  i   n   g  (     C  .h   a r   e n   g  u s   )   m u  s  c l    e  a n  d    g  u i   l   l    e m o  t    (     U . a  a l      g  e   )    e   g   g f   r   o m t   h   e B  a l    t   i    c  S   e  a  ,1   9   9   6   - 1   9   9   9  . (    a  )   E   a  c h   d   o  t   r   e   p r   e  s  e n  t    s  t   h   e r   e  s  u l    t   i   n   g v  a l    u  e  ;  B MF   R   S  R   G M  a f    t    e r   o n  e r   u n wi    t   h   3   , 5   ,1   0  ...1   0   0   0   0  i    t    e r   a  t   i    o n  s  /   r   u n .H  o r  i   z  o n  t    a l   l   i   n  e r   e   p r   e  s  e n  t    t   h   e B MF   R   S  R   G M f    o r   C  B 1   5   3   a f    t    e r   5   0   0   0   0  i    t    e r   a  t   i    o n  s .I   n  c r   e  a  s i   n   g n  u m b   e r   o f   i    t    e r   a  t   i    o n  s l    e  a  d   s  t    o  d   e  c r   e  a  s i   n   g  c  o  e f   f   i    c i    e n  t    s  o f   v  a r  i    a  t   i    o n  ,f   r   o m c  a . 6   0   % (   wi    t   h   3  i    t    e r   a  t   i    o n  s  )    t    o  0  . 5   % (   wi    t   h  1   0   0   0   0  i    t    e r   a  t   i    o n  s  )    , s  e  e M a  t    e r  i    a l    s  a n  d  M e  t   h   o  d   s . (    b   )   E   a  c h   t   r  i    a n   g l    e r   e   p r   e  s  e n  t    s  t   h   e r   e  s  u l    t   i   n   g  s  t    a n  d   a r   d   d   e v i    a  t   i    o n f    o r  B MF   R   S  R   G M  a f    t    e r   o n  e r   u n wi    t   h   3   , 5   ,1   0  ...1   0   0   0   0  i    t    e r   a  t   i    o n  s  /   r   u n .H  o r  i   z  o n  t    a l   l   i   n  e r   e   p r   e  s  e n  t    s  t   h   e   g  e  o m e  t   r  i    c m e  a n  s  t    a n  d   a r   d   d   e v i    a  t   i    o n  (    G M S  D  )   f    o r   C  B 1   5   3   a f    t    e r   5   0   0   0   0  i    t    e r   a  t   i    o n  s . V  OL . 3  9  ,N O.2 1  ,2  0  0  5  /   E NV I   R  ONME NT AL  S  C I   E N C E  &T E  C HN OL  O GY   9  8  3  9  9 
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks