Presentation on theme: "Sediment Quality Assessment and New York City Watersheds"— Presentation transcript:
1Sediment Quality Assessment and New York City Watersheds Stephen LewandowskiMajor, United States ArmyNYC Watershed/TifftScience & Technical SymposiumSeptember 19, 2013West Point, New YorkSession XIV: Supply Protection (Location: Bradley South)3:00 pm–3:30 pmSediments represent an important and dynamic compartment in aquatic ecosystems due to their ability toserve as a sink of many chemical contaminants. However, there is currently no single recommended sedimentregulatory framework available. This study reviews existing sediment quality guideline (SQG) approachesfor their application to watershed protection and examines US EPA National Sediment Inventory (NSI) datafrom the Catskill/Delaware and Croton watersheds. Key findings from the NSI analysis will be presented.
2Importance of Sediment Quality AGENDAImportance of Sediment QualityOverview of Sediment Quality Guidelines (SQGs) in New YorkU.S. EPA National Sediment Inventory Data for Catskill/Delaware watershedThis presentation is organized into 3 primary parts:- The importance of assessing sediment quality as part of a comprehensive ecosystem surveyProvide background on the guidelines used in New York state and the technical basis used for derivationInvestigate what the EPA NSI can tell us about the C-D watershed
3Sediment Quality Introduction Serve as “sink” for many chemicalsEcology and human health effectsSediment is comprised of all detrital, inorganic, or organic particles eventually settling on the bottom of a body of water (Power and Chapman 1992).
4Sediment ProcessesFigure 2.2. Sediment processes affecting the distribution and form ofcontaminants. CA EPA, July 18, 2008Sediment is a dynamic, complex material that plays an important role in aquatic ecosystems and provides habitat from a highly diverse community of organisms. Sediment-dwelling organisms range from microscopic bacteria and protozoans to large fish, amphibians, and reptiles that seek shelter and forage on the bottom, and may also hibernate there. Any and all of these organisms can potentially be put at risk from chemical contaminants in sediment, as well as wildlife that consume fish and invertebrates from the water body.Sediment comes in a large range of sizes.In flowing waters, streams and rivers, sediment is always on the move.Sediments at the bottom of lakes and ponds are considerably different than sediments in streams.Lakes can be described in terms of their trophic status, where trophy refers to the rate at which organic material is supplied by or to the lake per unit timeBecause sediment is such a complex material, it has a much more complex effect on contaminants that can cause toxicity. Sediment characteristics such as pH, cation exchange capacity (CEC), redox potential, oxic state, composition of the sediment (e.g., sand, clay, silt), amount and type of clay present (e.g., kaolin, bentonite, montmorillonite, etc.), grain size, pore size, the nature and volume of organic carbon present, and the presence of sulfides, nitrates, carbonates, and other organic and inorganic substances, can alter the chemical and biological activity of contaminants. There can be a high degree of variability in the concentration of a contaminant that causes toxicity in different sediments, and no single concentration of a contaminant in sediment can accurately represent a threshold of toxicity for benthic organisms in all sediments.Reduction potential (also known as redox potential, oxidation / reduction potential, ORP, pE, ε, or ) is a measure of the tendency of a chemical species to acquire electrons and thereby be reducedSediment processes affecting the distribution and form of contaminants (CA EPA)
6Sources and ReceptorsFigure 2.1. Principal sources, fates, and effects of sediment contaminantsin enclosed bays and estuaries. Adapted from Brides et alCA EPA, July 18, 2008Sources, fates, and effects of sediment contaminants (CA EPA)
8Equilibrium Partitioning (EqP) NY State ApproachesEquilibrium Partitioning (EqP)Consensus-based Sediment Quality Guidelines (freshwater sediments)Effects Range Low (ERL)/ Effects Range Medium (ELM) (marine/estuarine sediments)Image: USEPA, Procedures for the Derivation of Equilibrium Partitioning Sediment Benchmarks (ESBs)for the Protection of Benthic Organisms: EndrinNumerous efforts to develop suitable sediment quality guidelines for classifying sediment as toxic (contaminated) or non-toxic (relatively uncontaminated) have been published in the scientific literature.In order to best protect aquatic resources, the scientific literature was reviewed to identify existing sets of candidate sediment guidelines for use in New York State as numeric Sediment Guidance Values (SGVs), for the purpose of classifying sediments with respect to their potential for adverse impacts. As a result of that review, three methods were chosen for establishing New York State SGVs: equilibrium partitioning (EqP); consensus-based sediment quality guidelines for freshwater sediments (MacDonald, et al. 2000); and ERL/ERMs for marine/estuarine sediments (Long, et al. 1995).
9Equilibrium Partitioning Mechanistic: uses fundamental knowledge of the interactions between process variables to define the model structureBasis: non-polar organic contaminants will partition between sediment pore water and the organic carbon content of sediment in a constant ratioratio of the concentration in water to the concentration in organic carbon is termed the organic carbon partition coefficient (KOC)Limitations:does not consider the antagonistic, additive or synergistic effects of other sediment contaminantsdoes not account for bioaccumulation and trophic transfer to aquatic life, wildlife or humansEquilibrium partitioning is a mechanistic methodology for deriving SGVs for nonpolar organic contaminants from their corresponding ambient water quality standards or guidance values (AWQS/GVs) and their affinity to adsorb to organic carbon in sediment.EqP theory holds that a nonionic chemical in sediment partitions between sediment organiccarbon, interstitial (pore) water and benthic organisms. At equilibrium, if the concentration inany one phase is known, then the concentrations in the others can be predicted. The ratio of theconcentration in water to the concentration in organic carbon is termed the organic carbonpartition coefficient (KOC), which is a constant for each chemical. The ESB Technical BasisDocument (U.S. EPA, 2003a) demonstrates that biological responses of benthic organisms tononionic organic chemicals in sediments are different across sediments when the sedimentconcentrations are expressed on a dry weight basis, but similar when expressed on a ug chemical/g organic carbon basis (ug/g OC). Similar responses were also observed across sediments when interstitial water concentrations were used to normalize biological availability. The Technical Basis Document further demonstrates that if the effect concentration in water is known, the effect concentration in sediments on a ug/g OC basis can be accurately predicted by multiplying the effect concentration in water by the chemical’s KOC..
10Consensus-based Guidelines Empirical: derived from field-collected dataBasis: relates measured concentrations of contaminants in sediments to observed biological effectsERL – Effects Range Low: the 10th percentile concentration in a range of sediment concentrations for a given contaminant wherein adverse biological effects were observedERM – Effects Range Median: the 50th percentile concentration in a range of sediment concentrations for a given contaminant wherein adverse biological effects were observedTEC – Threshold Effects Concentration : derived by taking the geometric mean of similar sediment quality guidelines for concentrations of contaminants that below which, no adverse impacts would be anticipatedPEC – Probable Effects Concentration: derived by taking the geometric mean of similar sediment quality guidelines for concentrations of contaminants that above which, adverse impacts would be expected to occur frequentlyERL: Effects Range Low; ERM: Effects Range MediumTEL: Threshold effects (5oth %ile no effects/ 15th effects); PEL: Probable effects (80th/50th)The major limitation of LRM and other empirical approaches is that they classify individual chemicals independent of the other constituents that make up the sample and are therefore restricted in their ability to describe cause-and-effect and additive relationships
11Sediment Classification Class A - No Appreciable Contamination (no toxicity to aquatic life)EqP: chronic AWQS/GVsempirically-based: threshold effects concentration (TEC) or Effects Range Low (ERL)Class B - Moderate Contamination (potential for chronic toxicity to aquatic life)contaminant concentrations found between the threshold concentrations which define Class A and Class CClass C - High Contamination (potential for acute toxicity to aquatic life)EqP: acute AWQS/GVsempirically-based: probable effects concentration (PEC) or Effects Range Medium (ERM)There is high variability in the concentration of contaminants in sediment that cause toxicity. When reviewing studies that compare sediment bulk chemistry data and toxicity, there is a typical pattern across the concentration gradient. At low concentrations, there is a range where toxicity does not occur. At higher concentrations, there is a range where toxicity consistently occurs. In between, concentration and toxicity results are mixed. A given contaminant concentration might be toxic in one sediment sample but not in another. Within this range, toxicity cannot be reliably predicted simply from the contaminant concentration in sediment.
12Multiple Lines of Evidence Sediment quality triad (SQT) decision matrixChemical contaminationLaboratory toxicityBenthos alterationPossible conclusions+Strong evidence for pollution-induced degradation; management actions required.-Strong evidence against pollution-induced degradation; no management actions required.Contaminants are not bioavailable; no management actions required.Unmeasured contaminant(s) or condition(s) have the potential to cause degradation; no immediate management actions required.Benthos alteration is not due to toxic contamination; no toxic management actions required.Toxic contaminants are bioavailable but in situ effects are not demonstrable – need to determine reason(s) for sediment toxicity.Unmeasured toxic contaminants are causing degradation – need to determine reasons for sediment toxicity and benthos alteration.Contaminants are not bioavailable; alteration not due to toxic chemicals – need to determine reason(s) for benthos alteration.A weight of evidence approach can be very beneficial when evaluating risks from sediment contamination and is likely to result in more defensible sediment assessments. Any meaningful assessment of sediment quality needs to involve consideration of multiple lines of evidence, typically from sediment chemistry, ecotoxicology, and benthic ecologySediment guidance values are primarily useful as the initial step in a hierarchal approach for addressing a sediment contamination problem. They are a conservative tool for making an initial assessment of the potential risks that might be associated with contaminants in a sediment sample.The best-known example of a weight of evidence approach is the sediment quality triad (SQT) (Long and Chapman 1985; Chapman 1990). The SQT uses the results of three different types of sediment evaluations to ascertain risk: bulk sediment dry weight concentrations, sediment toxicity tests, and benthic community analysesA benthic community analysis, or macrobenthic community analysis, is a study that examines the characteristics of the benthic community that inhabits a possibly contaminated site. A benthic community analysis requires the use of a reference site or sites wherein the physical and chemical characteristics of the sediment are comparable to those at the site being evaluated. Several different biometrics have been proposed for evaluating the health of the resident benthic community. Typical metrics include species abundance and richness.
13Multiple Lines of Evidence (California) Multiple Lines of Evidence (CA Approach)Steven M. Bay and Stephen B. Weisberg, A framework for interpreting sediment quality triad data-Address two key elements of a risk assessment paradigm: 1) Is there biological degradation at the site?and 2) Is chemical exposure at the site high enough to potentially result in an adverse biological response?A combination of four benthic community condition indices was used to determine the magnitude of disturbance to the benthos at each site.Steven M. Bay and Stephen B. Weisberg, A framework for interpreting sediment quality triad data
14MLOE (CA Approach)Multiple Lines of Evidence (CA Approach), Steven M. Bay and Stephen B. Weisberg, A framework for interpreting sediment quality triad dataSteven M. Bay and Stephen B. Weisberg, A framework for interpreting sediment quality triad data
15National Sediments Inventory Data fromMore than 50,000 stations~ 4.6 million observationsRiver, lake, ocean, estuary sedimentsMandated by Water ResourcesDevelopment Act of 1992EPA reports to Congress: 1998 and 2004The NSI is an extensive database that contains approximately 4.6 million observations compiled from multiple studies between 1980 and 1999 from more than 50,000 stations throughout the United States. The EPA developed the NSI as part of a national sediment quality survey directed by the Water Resources Development Act (WRDA) of 1992 (EPA 2004). The NSI includes data on surface and subsurface sediment chemistry, fish tissue residue, and bioassay toxicity results collected between 1980 and 1999, of which this study utilizes its surface chemistry and bioassay components. NSI sample stations represent 5,695 river reaches with sediment samples from river, lake, ocean, and estuary bottoms, representing both freshwater and marine environments.
16Bioassay Toxicity Tests EPA: significant toxicity 20% difference in survival from controlMedium: Bulk sedimentEndpoint: Percent mortalityImages:Ampelisca abdita (marine amphipod)
17Human Health Screening Values (SV)a for Interpreting National Lake Fish Tissue Study Predator ResultsThe National Study of Chemical Residues in Lake Fish TissueEPA, September 2009The National Study of Chemical Residues in Lake Fish Tissue (EPA, 2009)
19n = 2,239 (NY) 6 C-D subbasins (HU-8), n = 278 NSI Stations in NY Blue: 6 C-D subbasins (HU-8): 278 of 2239 stationsn = 2,239 (NY)6 C-D subbasins (HU-8), n = 278
20Stations in C-D Watershed Boundary, n = 9 Schoharie Reservoir2Pepacton Reservoir3Cannonsville ReservoirAshokan ReservoirRoundoutReservoirNeversinkReservoirCatskill AqueductDelaware AqueductStations in C-D Watershed Boundary, n = 9
21Fish Tissue Species SMB: smallmouth bass BT: brown trout Majority of tissue samples collected in 1998.SMB: smallmouth bassBT: brown troutWS: white suckerRB: rock bass
22% lakes above in EPA study health endpointSV fish tissue concunits% lakes above in EPA studyMean(ppb)Confidence Level(95.0%)CountMercurynoncancer300ppb48.8484.2943.55217Chlordanecancer670.3ND19DDT691.7pp-DDE5.371.07pp-DDT12.795.69pp-DDDMercury 146 samples at or above SVThe National Study of Chemical Residues in Lake Fish Tissue (EPA-823-R ), U.S. Environmental Protection Agency, September 2009.
23Tissue concentrations (ppb) Units for PCBs wrong?
24Mercury Tissue Histogram Hg Screening value = 300 ppb
25Mercury Tissue by Station vic. Esopus Creek (Catskill)Roundout (Delaware)Pepacton (Delaware)Neversink (Delaware)Ashokan (Delaware)The distribution of mercury (Hg) and sites of greatest Hg methylation are poorly understood in Catskill Mountain watersheds. Although concentrations of Hg in the water column are low, high concentrations of Hg in smallmouth bass and walleye have led to consumption advisories in most large New York City reservoirs in the Catskill Mountains. Mercury in natural waters can exist in many forms, including gaseous elemental mercury (Hg0), dissolved and particulate inorganic forms (Hg(II)), and dissolved and particulate methylmercury (MeHg). Most Hg in living organisms is MeHg, a highly neurotoxic form that bioaccumulates in aquatic food webs. The production of MeHg by methylation of inorganic Hg in the environment is a key process affecting the quantity of MeHg accumulated in fish. Small quantities of MeHg in the diet can adversely affect wildlife and humans, which are exposed to MeHg almost entirely through the consumption of fish.The Neversink watershed is a high relief, forested ecosystem that supplies part of New York City's drinking water supply. This watershed is in the general vicinity of some of the highest suspected atmospheric Hg deposition zones in the coterminous US, and as such the Hg problem here is likely not limited to the current supply of Hg in the watershed.Ashokan (Delaware)Screening value = 300 ppb
26Sediments are an important component of watershed ecosystems SummarySediments are an important component of watershed ecosystemsNew York State applies screening guidelines derived from both mechanistic and empirical models to classify contamination and potential for toxicityNational sediments database is useful for a historical perspective on contaminants and development of guidelinesImage:
27Environmental Program (Dr. Marie Johnson) AcknowledgementsUnited States Military Academy, Dept. of Geography & Environmental EngineeringEnvironmental Program (Dr. Marie Johnson)Geospatial Lab (COL Michael Hendricks)Harvard School of Public HealthProfessor Jim ShineProfessors Francine Laden and Bob HerrickImage:
28ReferencesScreening and Assessment of Contaminated Sediment. New York State Department of Environmental Conservation, Division of Fish, Wildlife and Marine Resources, Bureau of Habitat, January 24, 2013 (Draft Version 4.0)Technical Guidance for Screening Contaminated Sediments. New York State Department of Environmental Conservation, Division of Fish, Wildlife and Marine Resources, January 25, 1999.The National Study of Chemical Residues in Lake Fish Tissue (EPA-823-R ), U.S. Environmental Protection Agency, September 2009.The Incidence and Severity of Sediment Contamination in Surface Waters of the United States, U.S. Environmental Protection Agency, Office of Science and Technology, 1997.The Incidence and Severity of Sediment Contamination in Surface Waters of the United States, National Sediment Quality Survey: Second Edition, US EPA, 2004.
29BACK-UPBack-Up SlidesHSPH Practicum Multivariable Regression Sediment-Toxicity ModelAdditional GIS MapsC-D Fish Tissue Data Analysis
30Site on fish gill (or other receptor) is a ligand too Biotic Ligand ModelSite on fish gill (or other receptor) is a ligand tooThe biotic ligand model is actually a water quality model, used to predict the toxicity of metals in water (Paquin, et al. 2002). The most toxic form of a metal in water is the divalent metal ion (M++) or ionic hydroxide species (MOH+). Various organic and inorganic ligands in water can bind ionic species of metal, and limit their availability for uptake by organisms. The model uses a number of water quality characteristics (temperature, pH, alkalinity, and concentration of dissolved organic carbon (DOC), major cations (Ca, Mg, Na, K), major anions (SO4, Cl, S)) to predict the availability and toxicity of the metal. The biotic ligand model is the basis for the U.S. EPA water quality criteria for copper (U.S. EPA 2007).Shine (2010)Gill is primary site of toxic action for most metals, especiallyfor freshwater organisms and acute toxicity
33Data Analysis Binary dependant variable (toxicity) Continuous predictor variables (concentrations)Pr(tox=1) = F(β0 + β1chem1 + β2chem2 + β3chem3… βnchemnWe built the multivariable logistic regression model using toxicity as a binary dependant variable and chemical concentrations as continuous predictor variables. The model is fit by the equation : Pr(tox=1) = F(β0 + β1chem1 + β2chem2 + β3chem3… βnchemn. StationToxic EffectMetal 1Metal 2Metal 3PAHsPCBsDDTTOCA1conc.percentageBCD
34Surface Chemistry Dataset Data ManagementBioassay DatasetB1. Include all species or sortB2. Compress from sample-level to station levelB3. Threshold for station tox based on mean sample toxB4. Merge with surface chemistry dataset by stationSurface Chemistry DatasetC1. Select analytes to retainC2. Drop duplicate entriesC4. Merge with bioassay dataset by stationC3. Compress to station-level with mean sample concentrationsPaired DatasetP1. Drop unmatched observationsP2. Reshape data from long to wideP3. Remove observations with missing chemical concentrationsP4. Apply MLRM
35Results Model 1 2 3 4 Description All bioassay species, 23 chemicals + 10th root TOCAmpelisca abdita, 23 chemicals + 10th root TOCAmpelisca abdita, sigma PAH + 11 chemicals + 10th root TOCStepwise backward, Model 3: As, Cd, Cu, Hg, pyrenen1,7891,557Significant positive variables (α=0.05)Cu, Hg, 10th root TOCCd, Cu, NiCd, Cu, HgSignificant negative variables (α=0.05)acenaphthylene, dibenz(a,h)anthracene, napthalene, PCBsAsBIC-11,024-9,687-9,756-9,809HL GOF χ2 (8),(p-value)15.41 (0.052)4.15 (0.843)4.72 (0.787)5.59 (0.6925)Area under ROC curve0.720.6950.6960.678Toxicity distribution (% stations coded as toxic)51.726.1BIC: Bayesian Information Criterion (more negative values indicate better model fit)HL GOF: Hosmer-Lemeshow goodness-of-fit test (small p-values indicate a lack of fit)ROC: receiver operating characteristic, plot of sensitivity vs. false positive rate, closer to 1 indicates better accuracy
36Model Evaluation Bayesian Information Criterion (BIC) more negative values indicate better model fitHosmer-Lemeshow goodness-of-fit testsmall p-values indicate a lack of fitReceiver operating characteristic (ROC)plot of sensitivity vs. false positive ratecloser to 1 indicates better accuracy
38Discussion Decent overall model fit and predictive value High specificity, but low sensitivityScientific plausibilityCadmium, copper, nickel as positive indicatorsArsenic as a negative indicatorSpecies could be adaptive to As in seawater convert to arsenobetaineSuggestive of oxidized conditions: As(V) vs As (III)Competition for binding sites on sediment particles and biotic ligand receptorsHg, Pb not significant may be tightly bound with low bioavailabilityLarge standard error for DDT
39Chemical Analysis Exposure Misclassification LimitationsImpacts confidence and generalizablilityChemical Analysis Exposure MisclassificationDifferent methods by study and over time fromHandling of detection limits/ low concentrationsBioassaySpecies appropriateConsistent methods and endpoint determination (EPA toxicity classification)Data setSpatial resolution: sample vs. station identificationData input errorsModel: limited in number of parameters; trade-offs in selection of species and chemical predictors; care not to over-fit model Maintain large n with a complete representation of chemicals
40ConclusionsAble to develop a reasonable MLRM with decent goodness of fit and predictive valuefor A. abdita toxic effects from surface sediment chemical concentrationsBig limitations and uncertainty from the data set structure, chemical analysis, bioassays and the statistical model reduce overall confidenceMethodology adds value for investigating data and physical and chemical relationshipsMultiple lines of evidence with knowledge of local area should be examined to assess sediment quality
50 2 metals (mercury and 5 forms of arsenic) 17 dioxins and furans > summary(smptiss9)X siteid studyid stationid sampleid fieldrep labrep sampdate speciesMin. : Min. : Min. : AS001 :59 A0 : 8 #: #:236 Min. : SMB :511st Qu.: st Qu.: st Qu.: AS002 :57 A1 : st Qu.: BT :40Median : Median : Median : RR001 :50 A2 : Median : WS :29Mean : Mean : Mean : NV001 :47 A3 : Mean : RB :273rd Qu.: rd Qu.: rd Qu.: TR001 :12 A4 : rd Qu.: YP :16Max. : Max. : Max. : HE001 : 7 A5 : Max. : RT :13(Other): 4 (Other): (Other):60tissue noincomp length weight sex pctlipid exsampidHV: 22 Min. : Min. : Min. : F: 1 Min. : Mode:logicalSF: st Qu.: st Qu.: st Qu.: M: st Qu.: NA's:236WH: 7 Median : Median : Median : U:233 Median :2.850Mean : Mean : Mean : Mean :2.9683rd Qu.: rd Qu.: rd Qu.: rd Qu.:3.480Max. : Max. : Max. : Max. :6.860NA's :215SPECIESValue # of Cases % Cumulative %1 BB2 BDACE3 BLC4 BRT5 BT6 CARP7 CCHUB8 CMSH9 LLS10 LMB11 LT12 PICK13 RB14 RT15 SMB16 WEYE17 WS18 YB19 YPCase SummaryValid Missing Total# of cases$tissueFrequenciesValue # of Cases % Cumulative %1 HV2 SF3 WHCase SummaryValid Missing Total# of casesTissue residue data include detailed analytical results, analyte sampled, species,sex, tissue type, length, qualifier for concentration, field and laboratory replication identifier, andweight.The laboratories under contract to EPA analyzed the fish tissue samples for the following target chemicals: 2 metals (mercury and 5 forms of arsenic) 17 dioxins and furans159 PCB congener measurements (representing results for 209 congeners) 46 pesticides 40 other semi-volatile organics (e.g., phenols)
51> frequencies(smptiss9[c("stationid")] , r.digits = 1) FrequenciesValue # of Cases % Cumulative %1 AS2 AS3 HE4 NV5 PP6 RR7 TR8 UESMB: small-mouth bassBT: brown troutWS: white suckerRB: rock bassYP: yellow perchYB:RT: rainbow trout
52Recommended Target Species for Lakes and Reservoirs The National Study of Chemical Residues in Lake Fish TissueEPA, September 2009- The species is abundant and commonly consumed in the study area.- It may potentially accumulate high concentrations of chemicals.- The species is easy to identify, and it has a wide geographic distribution.- Adult specimens are large enough to provide adequate tissue for analysis.
53% lakes above in EPA study health endpointSV fish tissue concunits% lakes above in EPA studyMeanStandard ErrorStandard DeviationMinimumMaximumCountConfidence Level(95.0%)Mercurynoncancer300ppb48.8484.2922.09325.4750.00217.0043.55PCBscancer1216.8538.98204.0019.00Chlordane670.3NDDDT691.7pp-DDE5.370.512.222.0012.001.07pp-DDT12.792.7111.8043.005.69pp-DDDMercury 146 samples at or above SV