Estimating the Reactive Component of Natural Attenuation of Dioxins in Sediments

Noémi Barabás, Peter Adriaens, Pierre Goovaerts (UM)

and Tim Dekker, Joe DePinto

Limno-Tech, Inc.Ann Arbor, Michigan

Technology Benchmarking Workshop:Technology Benchmarking Workshop:

Sediment and Floodplain RemediationSediment and Floodplain Remediation

March 25March 25--26, 200426, 2004

Presentation Overview

• Definition of Natural Attenuation (NA)• The components of NA• The role of reactive processes• A multivariate analysis technique for

estimating reactive contributions to field patterns

• Dioxin dechlorination patterns in sediments of the Passaic River, NJ

• Uncertainties• Practical considerations for the technique

Water

Algae

Buried Sediment

Benthos

Define Natural Attenuation in Contaminated Sediments

• Decline of contaminant concentrations in important receptors (fish, etc.)

• Decline of contaminant concentrations in sediments contributing to exposure:

Deep (when exposed):

surficial

Conceptual Model of Natural Attenuation

Water

Algae

Buried Sediment

Mixed Layer (~5-10cm)

Upstream Loading

UpstreamFlow

Runoff Loading TributariesAir-Water Exchange

Particle-boundchemical

Settling Resuspension

Particle-boundchemical

Burial

Partitioning

Dissolvedchemical

Partitioning

Benthos

Flow

Dispersion

AdvectionDiffusion

Dissolvedchemical

Chemical Decay or Biodegradation

Primary modes of attenuation:

settling/burial, resuspension/advection, chemical decay/biodegradation, porewater

Diffusion

Porewater Flow

Porewater Flow

Current Focus of Natural Attenuation Applications

Observed Media: • Surficial sediments (remediation studies and fate

modeling/forecasting)• Sediment cores (studies of national trends)• Fish (remediation studies and studies of national

trends)NA processes:• Burrial: settling and resuspensionConclusions:• Mix of some decreasing trends (half-time ~10 years)

and some stabilization in sediments and fish (PCBs, DDT)

Knowldege Gaps in NA Modeling and Application

• Bioavailability is often ignored due to insufficient information

• Biogeochemistry is important determinant of bioavailability:– Biogeochemical reactions determine partitioning

characteristics and thus, bioavailability.– relative importance increases for residual contamination

after remedial action. – Exposure through extreme events can lead to

increased/decreased risk depending on nature of reactions– Often assumed negligible for PCBs, Dioxins, persistent

chemicals

• How are long-term risks modified by reactive processes during NA? – How prevalent are reactive processes in sediments?

Conceptual Model of Natural Attenuation

Water

Algae

Buried Sediment

Mixed Layer (~5-10cm)

Upstream Loading

UpstreamFlow

Runoff Loading TributariesAir-Water Exchange

Particle-boundchemical

Settling Resuspension

Particle-boundchemical

Burial

Partitioning

Dissolvedchemical

Partitioning

Benthos

Flow

Dispersion

AdvectionDiffusion

Dissolvedchemical

Chemical Decay or Biodegradation

Primary modes of attenuation:

settling/burial, resuspension/advection, chemical decay/biodegradation, porewater

Diffusion

Porewater Flow

Porewater Flow

Role of Reactive Processes in NA

Water

Algae

Buried Sediment

Mixed Layer (~5-10cm)

Partitioning

Benthos

Chemical Decay or Biodegradation

Primary modes of attenuation: chemical decay/biodegradation

Particle-boundchemical

Dissolvedchemical

Limited quantitative knowledge about presence, nature and role of reactive processes in NA, especially for dioxins, and especially in the field.

Diffusion

Porewater Flow

Porewater Flow

John Pardue: “presence of starting halorespirers”?

Mike Dybas: “where are the environments that support halorespirers”?

Water

Algae

Buried Sediment

Benthos

Chemical Decay or Biodegradation

Primary modes of attenuation: chemical decay/biodegradation

Limited quantitative knowledge about presence, nature and role of reactive processes in NA, especially for dioxins, and especially in the field.

Polytopic Vector Analysis (PVA) as a tool to detect and estimate dechlorinationreactions involving dioxins in sediments.

Role of Reactive Processes in NA

PVA-Conceptual Model

FATE:FATE:Dechlorination

shifts initial ratios

2378T/PCDD = 0.3

Sediment PCDD Patterns

SOURCES:SOURCES:linear mixing

0

2

+0

1

0

+-

ClCl

Cl

Cl

O

O

Mixed source pattern

00.51

Hx4Hx6Hx9 Hp Pe T

Some Congeners increase others

decrease

00.51

Hx4Hx6Hx9 Hp Pe T

Sample

2378T/PCDD = 0.6 Source patterns are modified by

reactive processes

How do we infer this?

From this?

PVA

Traditionally PVA used to model source patterns

Modified PVA to model reactive patterns in dioxins/furans(M-PVA)

Algae

Benthos

Chemical Decay or Biodegradation

PVA in the Scientific Literature on Sediments

• Most applications model PCB sources (Arochlors)

• Some work on the impact of biotic reactions on the ability of PVA to identify Arochlor sources in PCB mixtures

• Some work on identifying dechlorinationrelated PCB signatures

• Fewer applications of PVA to dioxin/furan source patterns.

• No PVA adaptations for estimating dechlorination of dioxins/furans

Modified PVA X = AF

0

2 43.78%

-1

1 40.67%

-1

1 9.00%

-1

1 -0.09%

-1

0

1

T Pe Hx4 Hx6 Hx9

Hp0.42%

How many end-members

(k)?

What is their pattern?

How important is each in each sample?

k

XA

F

40% 15%…

Modified PVA X = AF

0

2 43.78%

-1

1 40.67%

-1

1 9.00%

-1

1 -0.09%

-1

0

1

T Pe Hx4 Hx6 Hx9

Hp0.42%

1. Multivariate field data2. Principal Components

Analysis – PCA3. Outliers? Number of

end-members?4. Rotation of PC axes

until all elements of matrices A and F are positive except for elements of Fdechlorin

5. Compare PVA patternswith known source and dechlorination patterns

XA

F

40% 15%…

Basic Approach for Modeling Dechlorination

(Variance-based)

Atlantic OceanRaritanBay

HudsonRiver

NewarkBay

N.YHarbor

New York

New Jersey

PassaicRiver

Dioxins in Passaic River SedimentsPetroleumCSOChemical manuf.Paint245T outfall area

Adapted from Huntley 1997 and Wenning 1993 1 mi

M-PVA: Dechlorination EMs

• If variability overestimated by factor of 2, dechlorinationcontributes at least 1.5% overall.

-1.0

0.0

1.0

2.0

T 2 3 4 5 Hp 7 8 9 10 11 12 13 14 15 16

2.88%

-2.0-1.00.01.02.03.0

T 2 3 4 5 Hp 7 8 9 10 11 12 13 14 15 16

3.53%

-2.0-1.00.01.02.03.0

T 2 3 4 5 Hp 7 8 9 10 11 12 13 14 15 16

2.52%

10 + 2:

10 + 3:

10 + 4:

-1.500

-1.000

-0.500

0.000

0.500

1.000

1.500

2.000

2.500

T Pe Hx4

Hx6

Hx9 Hp

TCD

F

Pe1F

Pe4F

Hx4

F

Hx6

F

Hx9

F

Hx4

6F

Hp6

F

Hp9

F

OCD

F

3 ± 1%

Model(pos+neg):

Pattern: Uncertainty:

Map of Dechlorination Loadings

0.0%

0.8%

1.6%

2.4%

3.2%

4.0%

0-2.5-5.0

Dep

th

651900652900

653900654900

655900

Easting

214433

213433

212433

211433

210433Northing

North

Site-wide contribution to variability: 2.88%

0.00.01 0.03 0.05 0.070.0

0.04

0.08

0.12Mean 0.03Std. dev. 0.01Maximum 0.07Minimum 0.00

• 3% means a 3% net change in dioxin and furan concentrations in a given sample due to dechlorination (distributed among the different kinds of dioxins).

Presence?!

Validation

• Convergence of laboratory and field methods:– Ratios above 0.5 are indicative of dechlorination activity as

indicated by laboratory experiments – As such, they correlate well with dechlorination loading

derived from field data.

Dechlorination Loading vs. Ratio

Fiel

d:D

echl

orin

atio

nLo

adin

g

Laboratory: Ratio0.00 0.20 0.40 0.60 0.80

0.0

0.01

0.03

0.05

0.07Correlation: 0.542

Correl 0.712

Consistent with Laboratory Results

How Has Dechlorination Affected Concentrations??

• On average dechlorinationcontributed 770 ng/kg to TCDD concentrations

• The proportion, can be as high as 100% relative to other sources of TCDD, in samples with low total concentration.

• At 33 ± 25%, dechlorination is the second most important contributor to 2,3,7,8-TCDD concentrations (after 2,4,5-T production, 60 ± 30%).

• Dechlorination is inversely proportional to total dioxin concentration.

0 2000 4000 6000 8000-30.48

-60.96

-91.44

-121.92

-152.4

Dep

th

Concentration (ng/kg)

Fractional contribution by dechlorination

DECHL proportion

TCDD concentrationDECHL contribution2,4,5-T contribution

0 0.5 1

Uncertainty Maps of Dechlorination

• There are three areas where dechlorination is very important (both maps) and these overlap with “contaminated” and “clean” locations

• Intermediate to high contributions cover about 30% of sediments.

• In the most contaminated areas dechlorination is least important

• “Contaminated” areas with high dechlorinationcontribution could be candidates for enhancement

Estimate0.850.460.07

Probability10.50

Estimate of contribution to 2,3,7,8-TCDD concentrations

Probability that contribution exceeds 50%

Demonstration of Uncertaintieswith Bootstrap Analysis

• 4 patterns!• How unique is the

dechlorinationsignature?

• Can the other patterns be interpreted?

• Re-partitioning during transport? Other pathways?

-1

0

1

2

7% of realizat ions3% exp lained

Prop

ortio

n

-1

0

1

2

40% of realizations5% exp lained

Prop

ortio

n

-1

0

1

243% of realizations4% exp lained

Prop

ortio

n

-1

0

1

2

8% of realizat ions

Prop

ortio

n

2% exp lained

T Pe Hx4

Hx6

Hx9

Hp

TCD

FPe

1FPe

4FH

x4F

Hx6

FH

x9F

Hx4

6FH

p6F

Hp9

FO

CD

F

-1

0

1

2

7% of realizat ions3% exp lained

Prop

ortio

n

-1

0

1

2

40% of realizations5% exp lained

Prop

ortio

n

-1

0

1

243% of realizations4% exp lained

Prop

ortio

n

-1

0

1

2

8% of realizat ions

Prop

ortio

n

2% exp lained

T Pe Hx4

Hx6

Hx9

Hp

TCD

FPe

1FPe

4FH

x4F

Hx6

FH

x9F

Hx4

6FH

p6F

Hp9

FO

CD

F

Conditions of Applicability of M-PVA

• Multivariate data set available (multiple congeners, multiple metals etc.)– Data-rich situations as opposed to finding similar

answer with satelite imagery.

• Candidate source/reactive patterns available for identification (fingerprints)

• Current method requires that sources dominate overall variability pattern (true for persistent contaminants)

Implementation Considerations for M-PVA

• Can only resolve patterns with differences in variability/patterns => similar patterns are lumped into single categories.

• Does not give information about reaction rates• Can we distinguish internal from external sources?• Variance-based approach makes pattern contribution to

individual samples most uncertain• Uncertainty analysis is important component (e.g.

Bootstrap, Monte-Carlo)• To assess performance efficacy, more research needed

with artificial data and laboratory experiments to determine:

• limits of pattern resolution, pattern uncertainty• effect of varying levels of dechlorination contribution on

uncertainty

Obstacles for Further Development/Use

• Requires implementation by experts familiar with multivariate statistics and reactive processes, due to:– Computational complexity of method

– Multiple levels of decisionmaking (statistical and interpretive)

• Availability of code/software

• Application is limited by uncertainties in the types of dechlorination/reactive patterns that can occur.

Conceptual Model of Natural Attenuation

Water

Algae

Buried Sediment

Mixed Layer (~5-10cm)

Upstream Loading

UpstreamFlow

Runoff Loading TributariesAir-Water Exchange

Particle-boundchemical

Settling Resuspension

Particle-boundchemical

Burial

Partitioning

Dissolvedchemical

Partitioning

Benthos

Flow

Dispersion

AdvectionDiffusion

Dissolvedchemical

Chemical Decay or Biodegradation

PVA gives partial answer to question about role of reactive processes in NA.

Diffusion

Porewater Flow

Porewater Flow

Integration of PVA in Site Assessment

Should be integrated with other methods and lines of evidence:

Statistics: indirect evidence

Interpolation

Sampling: direct evidence

Design: remediation

units

InterpolatedDechlorination

Map

H2, pε

RU-s

Transformation contribution

Sampling: no evidenceConcentrations

Uncertainty Uncertainty Uncertainty

Statistics: indirect evidence

Interpolation

Sampling: direct evidence

Design: remediation

units

InterpolatedDechlorination

Map

H2, pε

RU-s

Transformation contribution

Sampling: no evidenceConcentrations

Uncertainty Uncertainty Uncertainty