1 Office of Research and DevelopmentNational Risk Management Research Laboratory, Ground Water and Ecosystems Restoration Division, Ada, Oklahoma

Flux-Based DNAPL SiteAssessment & Remediation:Overview of ConceptsSuresh Rao, School of Civil Engineering, Purdue University

Triad Conference – June 10, 2008

0 10 20 30 40 50 60 70 80 90

4660

4665

0 10 20 30 40 50 60 70 80 90

4660

4665

4670

0 10 20 30 40 50 60 70 80 90

4660

4665

0 10 20 30 40 50 60 70 80 90

4660

4665

4670

0 10 20 30 40 50 60 70 80 90

4660

4665

0 10 20 30 40 50 60 70 80 90

4660

4665

4670

Control Plane

z

y

x

nJ

q0

2

Without Whom this Wouldn’t be Possible!

• Nandita Basu & Irene Poyer, Purdue University• Michael Annable, Kirk Hatfield & James Jawitz

University of Florida• Lynn Wood & Michael Brooks, US EPA-Ada• Ronald Falta, Clemson University• J. Christ, US Air Force Academy• Linda Abriola, Tufts University• Charles Werth, University of Illinois-Urbana/Champaign• Greg Davis, Colin Johnston & Brad Patterson,

CSIRO-Perth• Ravi Naidu, Megh Mallavarupu & Subhas Nandy

CRC-CARE & University of South Australia

Hill AFB 1994-1996; Dover AFB 1997-1999Rao et al. (WRR, 1997); Jawitz et al. (EST, 1998); McCray and Brusseau (EST, 1998); Falta et al. (WRR, 1999); Brooks et al. (JCH, 2004); Childs et al. (JCH, 2006)Sages (FL) 1998; Bachman Road (MI) 2000Jawitz et al. (EST, 2000); Mravik et al. (EST, 2003); Ramsburg et al. (EST, 2005)

Partial source removal debate: TheorySale and McWhorter (WRR, 2001); Rao and Jawitz (WRR, 2003); Parker and Park (WRR, 2004); Jawitz et al. (WRR, 2005)Partial source removal and flux: Lab and field dataFure et al. (JCH, 2006); Basu et al. (WRR, 2008 in press); Basu et al. (JCH, 2008 in press); Kaye et al. (JCH, 2008 in press); Chen and Jawitz (EST, in review 2007); Brooks et al. (In review, JCH, 2007)

A Decade+ of ProgressDNAPL Site Investigations

1995

1997

2005

2001

1999

2007

2003

2009

Can’t find it!

Can’t deplete it!

No benefit!

Back Diffusion!

Flux-based Assessment & Management?

Frequently Asked Questions• How much data needed before remediation?• What types of data best serve CSM & design?• Are high costs justified in terms of reduced

uncertainty?• What are the short-term benefits of source clean

up? • What is the likely plume response to source

clean up?• How to select target (interim) endpoints?• How to determine long-term stewardship needs?• Is there a simplified modeling & decision

framework?• How does all this fit into the TRIAD framework?

Pump & Treat Wells orInterception Trench or

Stream etc.E(t)

Source Control PlaneMd (t)

Source MassM0 and Mnow

Plume Mass(Parent + Products)

Mp (t) & λ(t)

DNAPL Site Monitoring:Enhancing Archived Site Data

PlumeControl Plane

6

Contaminant Fluxes & Mass Discharge at Control Planes

Md = Σ Ji Ai

Ji = Local mass flux (ML2T-1)qi = Local Darcy flux (LT-1)Ci = Local conc. (ML-3)Ai = Area of element i (L2)Md = Source strength (MT-1)Ks = Satd. Hyd. Cond (LT-1)j = Hydraulic gradient (-)

Ji = qi Ciqi =-Ks j

i

Control Plane (CP)

Δx

Δz

Ai = Δx Δz

Control plane area should be just large enough to completely inscribe the dissolved plume width

7

Cleanup to the Extent Necessary(CUTEN or Q10)

Source StrengthSource LongevityDegradation Rate Receptor Loading

⎟⎠⎞

⎜⎝⎛ −=

vxMM CPCP

λexp)1()2(

Measured mass fluxesat control cross-sectionsPumping tests

ResidualNAPL-phase

Q, C(t)

mass flux

C

t

C

t

Mea

sure

men

ts(f

ield

scal

e)Fi

eld

site

Q, C(t)

Comparisson andInterpretation

Control

Plane I

flux 1

Control

Plane II

flux 2

Key measures for:Site characterizationRemediation design &Performance Assessment

8

Contaminant Mass Discharge Estimates

What We HaveTemporal Data

Concentration in select source zone monitoring wells over time

Spatial Data

1. Mass discharge at the source control plane at a point in time

2. Plume Mass

What We NeedSource Longevity (Source Strength Function)

- present source mass

- source depletion behavior

Source & Flux Distribution (Source Architecture)

- identification of hotspots, targeted treatment

Characterization of DNAPL Sources

10

Source Depletion

First-Order

Linear

Constant

Newell et al., 2006, Jour. Env. Eng.Suarez et al., 2004, Remediation.

Γ = 0

Γ=1

Γ = 0.5

where,

Cs(t) and C0 =

flux-avg. conc. at source CP at time = t; and t=0

M0 = initial source mass

Vd = Darcy flux

A = Source CP Area

Γ = empirical constant

Simple Cases

0( )sC t C=

Falta et al. 2005a

Source Mass Estimation

( ) ,0,0

0

exp DD D

MM t M t

M⎛ ⎞

= −⎜ ⎟⎝ ⎠

First-Order

Γ=1

Fit monitoring well data to standard functions to estimate a value of Γ

( ) ,0,0

00

expt

DP D

MM t M t dt

M⎛ ⎞

= −⎜ ⎟⎝ ⎠

∫

Two equations and two unknowns –solve for MD,0 and M0

Estimate present source mass using

,00

0

exp Dt now

MM M t

M=

⎛ ⎞= −⎜ ⎟

⎝ ⎠

Method A: Requires only MW data

1. Monitoring well data over time fitted with an exponential to estimate k.

2. Now,

3. Here, a = time from which sampling data available

4. Thus:

/t a d t aM V AC k= ==

Method B: Requires mass discharge (MD) and plume mass (MP)

( )expt now t a dM M kt= == −

Source Mass Reduction

Dover AFB Florida Study

0

0.2

0.4

0.6

0.8

1

Sages Drycleaner

Site

Hill AFB OU-2

Dover AFBClemson Study

Dover AFB OU

Sour

ce F

lux

Red

uctio

n0 0.2 0.4 0.6 0.8 1

Falta et al., 2005

Exponent Γ is a function of DNAPL source architecture, hydrogeologic heterogeneity & correlation between the two.

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

Source Mass Reduction

Sour

ce F

lux

Red

uctio

n

2ln 0.1kσ =

2ln 1kσ =2ln 3kσ =

Brusseau et al. 2008

Basu et al. 2008

UTCHEM Simulations

Source Mass & Source Strength

13

Source Zone ArchitectureEulerian Approach

Reactive Travel Times

•Source zone conceptualized as a network of stream tubes, each characterized by velocity (travel time) and NAPL saturation.

•The domain described by mean and variance of travel time and NAPL saturation distribution, measured by non-reactive and reactive tracers, respectively.

•The measured parameters are used to predict change in the mean contaminant flux in time over the source control plane

Source Zone ArchitectureLagrangian Approach

Source Depletion Dynamics:

Decreasing GTP

Increasing Dissolution

Suchomel & Pennell, ES&T, 2006

Surfactant Flushing in 2-D Flow Chambers

M0 = Initial Mass of NAPL, κ0

β2 = Mass Depletion ExponentDamkohler

Model(Parker and Park 2005)

M0 – Initial Mass of NAPLβ - Variability Index

Power Law Model

(Zhu and Sykes 2005)

μlnτ= Mean of Hydrodynamic Field and DNAPL Architectureσlnτ= Variability of Hydrodynamic Field and DNAPL Architecture

Streamtube Model

(Jawitz et al. 2005)

Simplified Source Depletion Models

Basu et al., 2006, JCH

ln

ln

( ) ln1 1 erf2 2 2

f

c s

C T Tf C

τ

τ

μσ

⎡ ⎤−= − ⎢ ⎥

⎣ ⎦

( ) ( )

s o

f

c

C T M Tf C M

β⎡ ⎤

= ⎢ ⎥⎣ ⎦

2( ) ( )1 exp o

s o

f

s

C T L M TC K M

βκ⎡ ⎤⎛ ⎞⎛ ⎞

⎢ ⎥= − −⎜ ⎟⎜ ⎟⎢ ⎥⎝ ⎠⎝ ⎠⎣ ⎦

Cf(T) = f (HS, DS)Cf(T) = Flux-avg. Concentration

at Source Control PlaneHS = Hydrodynamic StructureDS = DNAPL Structure

Dissolution Profile Fitted to Source Depletion Models

ConclusionAll source depletion models fitdissolution behavior effectively

BUT, can we estimate parameters of any of these models independently and predict the dissolution profile apriori?

- YES, the streamtube model can be parameterized using tracer tests –

Power Function Model

0

0.25

0.5

0 5 10 15 20 25 30

pore volumes

Cf/C

s

αNRMSD = 0.015

Damkohler Model

0

0.25

0.5

0 5 10 15 20 25 30

pore volumes

Cf/C

s

κ0, β2NRMSD = 0.017

Basu et al., JCH, 2008

Streamtube Model

0

0.25

0.5

0 5 10 15 20 25 30

pore volumes

Cf/C

s

UTCHEM output

Model Fit

μlnτ, σlnτNRMSD = 0.007

Predicting Source Depletion:Simplified Models

GTP can be measured in lab. Field methods need to be developed.Tracer tests for calibrating stream-tube modeling demonstrated at lab & field scales.Eulerian vs. Lagrangian approaches, $$$!!!Γ=1 may be an adequate approximation??

Basu et al., WRR 2008

A

Sn (x,y,z; t)

Spatially Distributed vs. Integrated Parameters

Transition from local parameters (Sn, K, C) to integrated system behavior [J(g/m2/day); MD (g/day)]

J = contaminant mass flux (g/m2/day)q = groundwater flux (cm/day)C = dissolved concentration (mg/L)

J (y,z;t) =

q(y,z;t) C(z,y;t)

MD (g/day)= mass discharge = J d A∫

Relationship between mean values:

Total DNAPL mass [m(t)] & Source Strength [MD(t)]

Source Mass & Flux Distributions

• How does source mass change with time? • How does the flux distribution change?

0

5

1 0

1 5

z

5

1 0

1 5

2 0

2 5

x1 0

2 0

y

XY

Zp e rm1 .0 E -1 04 .6 E -1 12 .2 E -1 11 .0 E -1 14 .6 E -1 22 .2 E -1 21 .0 E -1 24 .6 E -1 32 .2 E -1 31 .0 E -1 3

x

5

10

15

20

25

y

5

10

15

20

25

z

0

5

10

15

XY

ZSn

0.160.140.120.100.080.060.040.02

T2VOC Numerical Simulations

Basu et al. JCH, 2008

Flux Statistics at Source Control Plane

0

0.5

1

1.5

2

2.5

3

0 10 20 30 40 50

Time (years)

Flux

Mea

n an

d St

anda

rd D

evia

tion

(g/m

2/da

y) Mean Flux (Positive Correlation)

Mean Flux (Negative Correlation)

Sigma Flux (Positive Correlation)

Sigma Flux (Negative Correlation)

Both mean and standard deviation of contaminant flux distribution decrease with mass depletion from DNAPL source

Basu et al. JCH, 2008

22

Flux Statistics at Source Control Plane

Numerical simulations are for emplaced NAPL.

Will this be also valid for spills?

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

Reduction in Flux Mean

Red

uctio

n in

Flu

x St

d. D

evia

tion

Domain 1 (Case 1) - variance(ln k) = 1Domain 1 (Case 2) - variance(ln k) = 2.45Domain 1 (Case 3) - variance (ln k) = 1Domain 2 (Case 1) - variance (ln k) = 2.45Domain 2 (Case 2) - variance (ln k) = 2.45Hill AFB, Utah

coefficient of variation = σ/μ = constant

Basu et al. JCH 2008

DNAPL Spill & Dissolution Simulations:Evolution of Source Architecture

Simulation data provided courtesy of:

J A Christ (USAFCE) &

L M Abriola (Tufts University)

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

Reduction in Flux Mean

Red

uct

ion

in F

lux

Sta

nd

ard

Dev

iati

on

Red lines represent variance ln k =1Black lines represent variance ln k = 0.29

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

Reduction in Sn mean

Red

uct

ion

in S

n s

tan

dard

dev

iati

on Red lines represent variance ln k =1Black lines represent variance ln k = 0.29

DNAPL Mass Contaminant FluxBasu et al. JCH 2009?

24

Temporal Evolution OfSource & Flux Centroids

Flux Center of Mass

0

1

2

3

4

5

6

7

0 20 40 60 80 100

Mass Removed (%)

Cen

ter

of M

ass

(m

)

y_barz_bar

Saturation Center of Mass

0

1

2

3

4

5

6

7

0 20 40 60 80 100

Mass Removed (%)

Ce

nte

r of M

ass

(m

)

z_bar

x_bary_bar

Source Mass Contaminant FluxBasu et al. JCH 2009?

25

Lab Data: Flux Architecture Dynamics

Flow-cell top

Inlet chamber flush port

Water outlets

Water inlet

Brass screen

Inlet chamber

Outlet chamber

NAPL injection port

B1B2

B3M1

M2M3

T1T2

T3

Heterogeneous permeability field

x

yz

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.2 0.4 0.6 0.8 1

reduction in flux mean

redu

ctio

n in

flux

sta

ndar

d de

viat

ion

Exp1Exp2Exp3Exp4

Zhang et al. 2008, JCH

Implication of Results•Flux distribution is an important metric that can be used for design of optimal remedial system that targets “hotspots”

• Flux distribution more stable over time than source distribution

•The observed stability of flux distribution is an unexpected andinteresting result that warrants further investigation

•Stability of flux distribution suggests the ability to characterize flux distributions in time once initial distribution is known.

A

27

FLUX Characterization

28

TRIAD BenefitsImproved Site Decision Making:• Integration of archived site monitoring data with new data

collection for enhanced conceptual site model• Mass discharge at source & plume control planes enables

estimation of source mass, source longevity, and natural attenuationcapacity

• Mass discharge serves as a metric for site prioritization, remediation performance, and helps in setting interim cleanup goals

• Groundwater & contaminant flux distribution measured at source control planes allows targeted source treatment & helps formulate cost-effective of site monitoring strategies

Questions• How much data needed before remediation?• What types of data best serve CSM & design?• Are high costs justified in terms of reduced

uncertainty?• What are the short-term benefits of source clean

up? • What is the likely plume response to source

clean up?• How to select target (interim) endpoints?• How to determine long-term stewardship needs?• Is there a simplified modeling & decision

framework?• How does all this fit into the TRIAD framework?

30

Answers?