:vv!

 

. 7v ..
.I‘

 

 

 

 

 

 

.1»

. o). 4%,.

, ifi‘fimfif

. 1.2.. ..

‘
.

 

MEG»...
’h r

 

 

 

{cf
.51

.7 Ivy: .
,.....l.. ‘1»?-
1‘yi.fit§xv

.4. :10... ‘ a.

.57. \c n 9..

I .. . a . ‘ ‘
“1.5.9.“.vidflmx 3.. a

«1:13) am. .3

tor... ct. .

. ii 6.! ..

If {3143;1‘2 I

lit.
. i:
.m

13;: )of 1.1:,- 002. . .inv 311).}.

”Emagfififi fiéfif

I ‘x . . II ‘

 

: .9! 1325. a , .
1.: 2‘ 4X: . . 33.1%

. I
r . A i .6335.
L... 4.. 4.. st... 4.: o .‘v ay
Hm. 55.3 .54». a th.‘«l.,
I . , ‘ 7-4. .

v u. v a
."I... 1.51%.”: a...
2 sari“... 15.43,.

.3- .
:rl' IV I. J.

: $155313

2
20%

 

‘LlBRARY
Michigan State
University

 

 

 

This is to certify that the
dissertation entitled

PLANT-ENHANCED REMEDIATION OF NAPHTHALENE

presented by

CHRIS M. SAFFRON

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Chemical Engineering and
Materials Science

 

 

(Hank/l - i/M/

- Major Professor's Signature

Maug‘r 7.3] 1.009

 

Date

MSU is an Affirmative Action/Equal Opportunity Institution

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

NOV 11"9 @008

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/ClRC/Date0ue.inddop.1

 

PLANT-ENHANCED REMEDIATION OF NAPHTHALENE

By

Chris M. Saffron

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTORATE OF PHILOSOPHY

Department of Chemical Engineering and Materials Science

2005

ABSTRACT

PLANT-ENHANCED REMEDIATION OF NAPHTHALENE
By
Chris M. Saffron

This study explored the effects of slow desorption on the rate of removal of
naphthalene in planted systems. The magnitude by which desorption limits the removal
of PAHs was assessed in different soils. The contaminant exhibited nonequilibrium
desorption in all of the soils studied. The desorption data were described using several
desorption models. The desorption models were also used to quantify the rate of mass
transfer from the soil solute to the soil solution. The mathematical models were ranked
using a construct called the Akaike information Criterion, which takes into account
accuracy and parameter variability. The effect that plants have on the rate of naphthalene
removal in soils was also assessed. The collected data was interpreted using a descriptive
model. Volatilization by gaseous diffusion, sorption to roots, transpirational uptake, fast
mass transfer from the soil solid, and slow mass transfer from the soil solid were included
in the model.

The lessons learned during the development of this model were used in positing a
decision-making methodology. This methodology contains a procedural approach for
quantifying the rate of mass transfer, and whether plants are able to overcome these
transfer limitations. A dimensional analysis was used to determine the efficacy of

planted systems, which leads to improved decision making regarding phytoremediation.

This work is dedicated to my wife, Danida.

iii

ACKNOWLEDGMENTS

I would like to acknowledge the effort and patience that my wife, family, and
friends have given me throughout this journey. My wife, Danida, has exhibited degrees
of patience, support, and love that I thought not possible. I thank my family for their
continuing encouragement and unwavering support. I thank my Mom, and my Grandma
and Grandpa Swanson for the work ethic that they’ve instilled in me since childhood. I
would not have done this without them. I thank God for the joy he has brought my wife
and I, especially since the birth of our daughter Annaka. I am grateful to the many
friends that I’ve made while in graduate school. I extend special thanks to Dr. Jeong-Hun
Park, for his insight and thoughtfulness. I thank Dr. Mazen Haydar for many stimulating
conversations. I thank Dr. Irfan Aslam for discussions about desorption in column
studies, and a unique way of making a lay-up. I thank Shawn McElmurry for discussions
concerning the environment and recreational activities. I thank Dr. Robert Dombrowski,
Dr. David Knop, and Dr. Prashant Srivastava for the many homework sessions and
discussions regarding chemical engineering. Thanks to Dr. Clayton Rugh and the
Phytolab crew for the lessons in sterile technique and the trips to Oodles of Noodles.
Finally, I would like to thank Professors Bruce Dale and Thomas Voice. Your
advisement has kept me on target through a very challenging portion of my life. And

now, the real challenge begins.

iv

TABLE OF CONTENTS

List of Tables ................................................................................. vii
List of Figures .............................................................................. viii
Chapter I ...................................................................................... 1
Introduction .......................................................................... 1
Objectives ............................................................................ 3
References ........................................................................... 5
Chapter 2 ...................................................................................... 6
Introduction .......................................................................... 6
Phytoremediation technologies ................................................... 7
Processes involved in phytoremediation ......................................... 8
Review of modeling efforts ........................................................ 16
Research 19
scope ..................................................................................
References ........................................................................... 23
Chapter 3 ...................................................................................... 29
Abstract ............................................................................... 29
Introduction .......................................................................... 30
Methods and Mathemtics .......................................................... 34
Results and Discussion ............................................................. 39
References ........................................................................... 59
Chapter 4 ...................................................................................... 63
Abstract .............................................................................. 63
Introduction .......................................................................... 64

Materials and Methods ............................................................. 68

Mathematics ......................................................................... 71
Results and Discussion ............................................................. 75
References ........................................................................... 95
Chapter 5 ...................................................................................... 102
Abstract .............................................................................. 102
Introduction .......................................................................... 1 03
Mathematics ......................................................................... 1 06
Results and Discussion ............................................................. 110
References ........................................................................... 126

vi

Table 2-1.

Table 3-1.

Table 3-2.

Table 3-3.

Table 4-1.

Table 4-2.

Table 5-1.

Table 5-2.

Table 5-3.

Table 5-4.

LIST OF TABLES

Types of phytoremediation (adapted from McCutcheon and
Schnoor[65]) .......................................................................

Selected properties of sorbents used in this study ............................

Model equations fit to desorption data sets ....................................

List of Symbols ...................................................................

Soil parameters for Spinks A and Kalkaska A soils ..........................

Estimated parameters for planted and unplanted soils .......................

A list of model variables and parameters for the planted and unplanted
mathematical models .............................................................

A list of dimensionless variables and parameters for the planted and
unplanted mathematical models ................................................

Base case parameter values used for naphthalene transport and reaction
in soil ...............................................................................

Dimensionless number and parameter values used to generate Figures
5-3 through 5-8 ..................................................................

vii

22

56

57

58

93

94

122

123

124

125

Figure 2-1.

Figure 3-1.

Figure 3-2.

Figure 3-3.

Figure 3-4.

Figure 3-5.

Figure 3-6.

Figure 3-7.

Figure 3-8.

Figure 3-9.

Figure 3-10.

Figure 4-11.

LIST OF FIGURES

A simplified diagram that shows some of the interactions that
occur in planted systems ................................................... 21

Best-fit models to atrazine-Capac A desorption data: (a) the

chemical two-site model, (b) the chemical three-site model, (c) the
two-parameter pore diffusion model, ((1) the three-parameter pore
diffusion model, (e) the gamma model, (I) the hybrid gamma/two-

site model, (g) the modified three-parameter kinetic model and (h)

the modified five-parameter kinetic model ................................ 46

. Best-fit models to naphthalene-Muck desorption data: (a) the

chemical two-site model, (b) the chemical three-site model, (c) the
two-parameter pore diffusion model, (d) the three-parameter pore
diffusion model, (e) the gamma model, (I) the hybrid gamma/two-

site model, (g) the modified three-parameter kinetic model and (h)

the modified five-parameter kinetic model ................................ 47

Comparison of the AICc for the nine models that were fit to Capac
A-atrazine desorption data 48

Comparison of the AICc for the nine models that were fit to Capac
A-naphthalene desorption data .............................................. 49

Comparison of the AICc for the nine models that were fit to
Colwood A-atrazine desorption data ....................................... 50

Comparison of the AICc for the nine models that were fit to
Colwood A-naphthalene desorption data .................................. 51

Comparison of the AICc for the nine models that were fit to
Hartsells-atrazine desorption data .......................................... 52

Comparison of the AICc for the nine models that were fit to
Hartsells-naphthalene desorption data ..................................... 53

Comparison of the AICc for the nine models that were fit to Muck-
atrazine desorption data ...................................................... 54

Comparison of the AICc for the nine models that were fit to Muck-
naphthalene desorption data ................................................. 55

The planted bioreactor ........................................................ 83

viii

Figure 4-12.

Figure 4-13.

Figure 4-14.

Figure 4-15.

Figure 4-16.

Figure 4-17.
Figure 4-18.

Figure 4-19.

Figure 4-20.

Figure 5-1.

The conceptual description of naphthalene transport in unplanted
soil that was developed upon interpretation of the methanol and
water extraction data .........................................................

Box model for the unplanted treatments ...................................

The conceptual description of naphthalene fate and transport in
planted soil that was developed upon interpretation of the methanol
and water extraction data .....................................................

Box model for the planted treatments. The dashed arrows represent
naphthalene transport that is attributed to the transpiration stream. . ..

Plots of methanol extraction data, water extraction data, and models
fits for the six treatments. The Y-axis is the naphthalene mass in pg.
The diamonds are the methanol extraction data points and the circles
are the water extraction data points. The solid line is the methanol
extraction best-fit line and the dashed line is the water extraction
best-fit line .....................................................................

Root surface area profile for A. gerardii and M. alba in SpAf soil.
Root surface area profile for A. gerardii and M alba in Kal A soil...

Plot of methanol extractable naphthalene mass and water extractable
naphthalene mass versus time for Kal A soil that is unplanted. The
diamonds (O) are the methanol extraction data and the circles (o) are
the water extraction data. Standard deviation error bars are included
on each point. The solid line is the model fit of the methanol and
water extraction data .........................................................

Plot of methanol extractable naphthalene mass and water extractable
naphthalene mass versus time for Kal A soil that is unplanted. The
diamonds (5‘2) are the methanol extraction data and the circles (o) are
the water extraction data. Standard deviation error bars are included
on each point. The solid line is the model fit of the methanol
extraction data and the dashed line is the model fit of the water
extraction data .................................................................

A diagram detailing some of the processes involved in the

phytoremediation of naphthalene. The hatched regions represent
soil aggregates .................................................................

ix

84

85

86

87

88

89

90

91

92

115

Figure 5-2.

Figure 5-3.

Figure 5-4.

Figure 5-5.

Figure 5-6.

Figure 5-7.

A box model of the transport and reactive processes occurring in
rhizosphere soil. The double arrows represent equilibrium processes
and the single arrows represent kinetic or transport processes. The
dashed arrows represent the kinetic process added to the
phytoremediation model ......................................................

A plot demonstrating the regimes that control the phytoremediation
of naphthalene in Spinks A-horizon soil ...................................

The effect of increasing the Damkohler number on the effectiveness
factor ............................................................................

The effect of increasing the soil-water partition coefficient on the
effectiveness factor ............................................................

The effect of the desorption rate coefficient on the effectiveness
factor ............................................................................

The effect of converting the equilibrium fraction to the
nonequilibrium fraction (feq = 0), converting the nonequilibrium
fraction to the equilibrium fraction (fneq = 0), and converting the

immobile fraction to the equilibrium fraction (fim = 0) ..................

116

117

118

119

120

121

Chapter 1. Introduction and objective
Introduction

Currently, the world oil consumption is estimated at 75 MM barrels of oil per
day[1]. This corresponds to filling a cube with sides that are roughly 1 mile wide
every year. This amount of production/consumption provides an ample driving force
for contamination of soils, waterways and air with organic contaminants. The
remediation of organic contaminants by plants has become a viable alternative to
more traditional schemes. Currently, phytoremediation is being used by academia,
industry, government and the military to remediate sites contaminated with
polyaromatic hydrocarbons (PAHs), polychlorinated biphenyls (PCBs),
trichloroethylene (TCE), perchloroethylene (PCB), trinitrotoluene (TNT), and many
others. When compared to competing technologies, such as excavation and
incineration, pump and treat, etc., phytoremediation has several inherent advantages.
In addition to being an effective technology for clean-up, advantages include: a
relatively low cost, a general social acceptance, an aesthetically pleasing appearance
over competing technologies and the support of a diverse and burgeoning population
of trained professionals. Like microbial bioremediation, phytoremediation is an in
situ technology that will not disturb the soil matrix. Phytoremediation takes
advantage of a plant’s innate ability to transpire water, which concentrates dilute
contaminants. The bacterial population in rhizosphere soils is also ten to 1000 times
greater than in unplanted bulk soil[2]. As much as 30 percent of a plant’s

photosynthate is exuded in the roots[2], and many of these exudates resemble organic

contarninants[2] (a feature of plants that may induce degradative enzymes or prompt
cometabolism). Plants are known to improve soil aeration, soil aggregation, reduce
erosion, and fix atmospheric carbon dioxide. Though planting, monitoring, amending
with nutrients, harvesting and disposal are costs associated with phytoremediation—
the energy required for growth is provided by sunlight and a plant’s ability to acquire
dilute nutrients by transpiration provides a concentrating effect in the root zone.

Cunningham and Berti have stated[3]:

“A green plant is a ‘solar-driven, pumping, and filtering system that
has measurable loading, degradative and fouling capacity’. Roots are
‘exploratory, liquid-phase extractors that can find, alter, and/or
translocate elements and compounds against large chemical

9”

gradients .

In truth, the complexities inherent in phytoremediation systems include many
mechanisms not included in the above description. Thus, given the obvious need for
remediation and clean-up, and including the advantages that phytoremediation has
over other technologies, further development in this area is warranted. The proper
selection of remediation technologies that are demonstrated effective, that are low-
cost, and that can add aesthetic value is of the utmost importance.

In 1999, the estimate for the total US. phytoremediation market exceeded 30
MM U.S. dollars[4]. This number will grow, as it is estimated that 1.7 trillion US.

dollars will be spent on clean-up in the US. over the next thirty years[5]. Cost

estimates for design and implementation ranged between 60,000 to 100,000 US.
dollars per acre in 1999. This is roughly one-fourth the cost of excavation followed
by landfilling[6]. Further evidence supporting the efficacy of phytoremediation of

organic pollutants in soils will expand the market for this technology.

Objectives

This study was designed to explore the effects of slow desorption on the rate
of remediation in planted systems. Four objectives were considered in this
investigation. The chapters of this dissertation are organized around each of these
objectives.

The first objective is to assess the magnitude that desorption limitations may
impose on planted systems. All of the studied soil and contaminant combinations
exhibited nonequilibrium desorption phenomena, meaning that desorption from the
soil matrix may limit the rate of remediation. Several mathematical models were fit
to each set of desorption data to describe the rate of desorption. The mathematical
models were ranked based on accuracy and parameter variability using a construct
called the Akaike Information Criterion. The knowledge that was gained concerning
the desorption limitations was used to design the experiments needed to satisfy the
second objective of this study.

The second objective was to determine the effect plants have on contaminated
soils that are limited by slow desorption. Completion of this objective would provide
evidence that phytoremediation is or is not limited by slow desorption from the soil

matrix.

The third objective was to formulate a descriptive model that fits the rate data
collected during the completion of the second objective. The model was developed as
a descriptive tool, i.e. a tool used to interpret the collected data. Several types of
mass transport were considered in the model development, including: volatilization
by gaseous diffusion, root uptake by sorption to lipophilic tissues, transpirational
uptake, fast mass transfer from the soil solid, and slow mass transfer from the soil
solid. Mechanisms that did not add to organic chemical mass transfer under the study
conditions of the second objective were not included in the model, e. g. the advection
and dispersion due to contaminant leaching. The lessons learned during the
development of this descriptive tool were used in positing a decision-making
methodology, which forms the basis of the fourth objective.

The fourth objective involves the use of a mathematical model, developed
using the lessons learned upon completion of the first three objectives, for developing
a decision-making methodology. This methodology contains a procedural approach
to determining whether organic mass transfer from the soil is limiting in batch
conditions, and whether plants are able to circumvent these transfer limitations. The
analysis will be conducted using the classical approach of dimensional analysis.

After the interpretation of the collected data using a constructed mathematical model,
plotting the dimensionless groups results in a division between a mass transfer limited
regime and a reaction rate limited regime. The decision regarding the use of

phytoremediation can be made by interpreting the plots of dimensionless groups.

References

1. Abraham, 8., Remarks by Energy Secretary Spencer Abraham Public Energy
Forum Lunch Conference of 08 Energy Ministers. May 2, 2002: Detroit,
Michigan.

2. Curl, EA. and B. Truelove, The Rhizosphere. 1986: Springer-Verlag Berlin
Heidelberg.

3. Cunningham, SD. and W.R. Berti, Remediation of contaminated soils with

green plants: an overview. In Vitro Cellular and Developmental Biology,
1993. 29: p. 207-212.

4. Glass, D., (1.5. and International Markets for Phytoremediation, 1999-2000.

5. Timian, SJ. and OM. Connely, The Regulation and Development of
Bioremediation. 1996.

6. Marmiroli, N. and SC. McCutcheon, Making Phytoremediation a Successful
Technology, in Phytoremediaiton: Transformation and Control of
Contaminants, S.C. McCutcheon and IL. Schnoor, Editors. 2003, John Wiley
and Sons, Inc.

Chapter 2. Literature Review

Introduction

The effort presented in this chapter reviews many of the published mechanisms
responsible for controlling the rate and extent of phytoremediation. The numerous
processes involved in phytoremediation are investigated to assess which may limit the
application of this technology. In actuality, phytoremediation encompasses several
technologies, and a short description of the various technologies is presented in this
chapter and summarized in Table 2-1. Many of the processes involved in the
phytoremediation technologies have been previously assembled into mathematical
models by other authors—several of these models are reviewed in this chapter. The
modeling investigations presented in the literature assume that equilibrium exists between
the soil matrix and the soil solution, while the investigation presented in Chapters 3 and 4
of this dissertation explores the kinetic limitations of contaminant desorption.
Furthermore, an obvious link between the effort in vitro and the effort in situ, regarding
the use of mathematical models describing phytoremediation, does not appear to exist.
Chapter 5 of this dissertation attempts to bridge the gap between a descriptive model and
a decision-making tool that would benefit the application of phytoremediation by the site
engineer

Phytoremediation is defined as the use of green plants to remove pollutants from
the environment or to render them harmless[l ]. This is a burgeoning technology that can
be used for a diverse array of contaminant-soil combinations. The number of
applications for phytoremediation has increased as the collective understanding,

concerning the root-soil-contaminant interaction, has improved. Plants have been

effectively demonstrated to remediate soils contaminated with several classes of
contaminants, including: organic contarninants[Z-l 1], heavy metal contaminants[12-17],
and radionuclide contaminants[18, 19]. Several phytoremediation technologies exist for
removing these contaminants, including those listed in Table 1. This review contains a
brief summary of these technologies, some of the mechanisms controlling these
technologies, and the models that are capable of describing the controlling mechanisms.
The last section of this review presents the scope of the research that is contained in the

following chapters.

Phytoremediation technologies

Several studies have validated the use of phytodegradation. Poplar trees were
found capable of transforming trichloroethylene (TCE) to trichloroethanol, trichloroacetic
acid, and dichloroacetic acid[20]. The ability of plant cell culture and plant cell extracts
were found capable of transforming glycerol trinitrate (nitroglycerin) to glycerol dinitrate
and later, glycerol mononitrate[21]. Nitroreductase and laccase enzymes, released by
plants, were shown to degrade 2,4,6-trinitrotoluene (TNT)[2]. Symbiotic bacteria were
found in poplar that are capable of degrading TNT, high melting explosive (HMX), and
royal demolition explosive (RDX)[22]. Enhanced microbial degradation was likely
responsible for the removal of petroleum hydrocarbons from the rhizosphere of
sorghum[23]. In general, the phytodegradation of contaminants can occur within the
plant tissue by transformation (either by enzymes or endophytes), or ex planta by plant-
released enzymes or the microbial consortia (known as rhizodegradation). Less well
demonstrated is the occurrence of phytostimulation, though a few studies have provided

interesting results. The mineralization of atrazine was promoted by the addition of root

exudates into silica sand and silt loam filled microcosms[24]. The stimulatory effect of
root exudates is a possible explanation for the enhanced microbial degradation of
pentachlorophenol in the rhizosphere of Hycrest crested wheat grass[25]. The removal of
polychlorinated biphenyls (PCBs) in the rhizosphere of red mulberry was attributed to the
stimulatory action of phenolics on the rhizosphere consortia[26]. Contrarily, the addition
of root extracts was shown to decrease the expression of nahG using Pseudomonas
fluorescens HK44 (this contains a nah-qu fusion), though the total amount of expression
was increased due to increased cell growth upon extract addition[27]. The use of
phytoremediation to stimulate the microbial consortia to degrade contaminants is an area
of ongoing research. Other phytoremediation technologies are being practiced in the
field and investigated in academia. The use of phytoextraction[28], both chelate-assisted
and unassisted is perhaps the most well documented mode of phytoremediation. It has
been mathematically determined that the evaporative flux from a stand of trees used in
phytocontainment of a TCE plume, must be much greater than the flux of the plume
itseltI29]. Phytovolatilization of methyl tert-butyl ether (MTBE) was found to occur
through the stem and leaves of poplar cuttings[30]. Phytostabilization may be especially
useful for controlling tailings from strip and open uranium mines[31]. Other
phytoremediation technologies exist in addition to those mentioned in this review, though
phytodegradation and phytostimulation are the most relevant phytoremediation

technologies for the effort presented in this dissertation.

Processes involved in phytoremediation

The effective use of phytoremediation depends upon an understanding of the

mechanisms that control the rate of removal in the rhizosphere. The word “rhizosphere”

is loosely defined as the soil zone influenced by the roots[32]. This zone contains the soil
solid (soil matrix), the soil solution, soil gas, the plant roots, and all rhizosphere microbes.
Many chemical, physical, and biological processes occur in the rhizosphere[32-34]—and
often these processes interact in a complex manner. A simplified depiction of the nature
of these interactions is as shown in Figure 1. As with bulk soil, the rhizosphere soil can
be characterized chemically, in terms of its soil organic matter (SOM) content, and
physically, in terms of its porosity and texture. Biologically, a multitude of
microorganisms, both in numbers and in types, exist in the rhizosphere[32, 34]. These
organisms can interact with plants by secreting plant growth promoting hormones[35].
Plants can influence the quantity and types of organisms, in the microbial community,
through root exudation[24, 36, 37] (a passive process) or root secretion (an active
process). Plants alter the chemistry of the soil solution through the process of
transpiration. The transpirational movement of water towards the root serves to
concentrate dilute nutrients in the rhizosphere. Through a combination of physical,
chemical, and biological processes that occur in the rhizosphere, the manner and degree
of soil aggregation can be affected[3 8]. The degree of aggregation can alter the level of
soil aeration and the level of moisture infiltration[3 8]. The growth of roots into the soil
matrix can also increase soil aeration and moisture infiltration by opening channels that
locally increases the soil conductivity. The complex interaction between biological,
chemical and physical processes is further demonstrated by lignin deposition. Lignin, a
component of plant cell walls, is an important component during the formation of soil
organic matter (SOM)[39]. Therefore the physicochemical properties of the soil matrix

are affected by root growth and necrosis. Essentially, a number of processes are

simultaneously occurring in the rhizosphere, which creates a highly complex, versatile,
and dynamic zone with the proposed potential for remediating hazardous compounds.

The transport and transformation of an organic contaminant in the rhizosphere is
of interest to scientists and engineers who want to explain the observed rates of
contaminant loss and apply the technology to contaminated sites. The rate of
contaminant loss can be controlled by: 1) mass transfer limited desorption of the
contaminant from the soil matrix, 2) the microbial biodegradation rate in the soil solution,
3) the rate of sorption to the plant root, 4) the rate of transpirational uptake, 5) the rate of
volatilization and 6) the rate of leaching. Other mechanisms that may allow planted
systems an advantage over unplanted systems include the release of oxidative enzymes
from the root, the exudation of molecules that compete for sorption sites with the
contaminant, or degradation by mycorhizal fungi. The aforementioned six mechanisms
are under consideration in this work, due to the preponderance of evidence that suggests
these mechanisms are universally important when considering the application of
phytoremediation.

The mechanism that is most likely to limit the rate and extent of a contaminant’s
remediation by plants is mass transfer limited desorption. The likely first step in a
contaminant’s removal is desorption. The rate of phytoremediation will equal the rate of
desorption, when no mechanism exists for enhancing the desorption rate. Conceptually,
the rate of desorption is limited by the mass transfer of the compound from the soil
matrix to the soil solution. The desorption rate depends on the path length, path
tortuosity, physicochemical properties of the matrix, and the physicochemical properties

of the contaminant[40, 41]. Furthermore, the desorption rate can change as organic

10

contaminant desorbs from the soil[42-46]. Desorption rate data has suggested that more
than one type of phenomena is responsible for controlling the rate and extent of HOC
desorption[47-51]. The desorption profile can be described by assigning the collected
desorption data into one, two or three conceptual regimes. Historically, a one-regime
model that assumes sorption-desorption reversibility, was used to describe the desorption
process in batch and flow-through systems. Conceptually, desorption was thought to be
an equilibrium process which occurred instantaneously upon addition of aqueous solution.
This equilibrium process is parameterized by the soil-water partition coefficient, which is
a thermodynamic construct that is measured by plotting a sorption isotherm. The use of
the soil-water partition coefficient assumes a negligible sorption-desorption hysteresis, an
assumption that often over-predicts the extent and rate of desorption. In reality,
desorption usually occurs at a slower rate than sorption. Thus the one-regime model of
contaminant desorption was found to be inadequate in predicting the extent and rate of
contaminant desorption for many systems. The advent of the two-regime model was
spurred by consideration of mass transfer limitations that may be inherent in soil matrices.
These models are constructed to include characteristics attributable to equilibrium
phenomena and characteristics attributable to kinetic phenomena. These models were
formulated to describe desorption data that are exemplified by fast desorption followed
by slow desorption. Many explanations exist for the kinetic phenomena that limit the rate
at which contaminant desorbs from the soil. Two classes of kinetic phenomena are
distinguished in the literature. These are the kinetic phenomena attributable to: 1) the
chemical characteristics of the contaminant and soil matrix, and 2) the physical

characteristics of the contaminant and soil matrix. The chemical two-regime models

11

assume that a chemical interaction between the contaminant and the solid matrix acts to
retain the contaminant beyond what is predicted by chemical equilibrium. The physical
two-regime models assume that a physical constraint retards the passage of contaminant
through the soil matrix. The use of a F ickian diffusive flux is normally used to describe
the transport of contaminant through a soil micropore, through the soil organic matter, or
through immobile water regimes. The main advantage of the two regime models is the
ability to describe desorption data that exhibits two distinct classes of behavior. However,
when the extent of contaminant desorption is less than one-hundred percent after
sequential water extraction (i.e. batch desorption) studies or flow-through (i.e. column
desorption) studies, the two regime models do not accurately describe the collected data.
In this circumstance, the formulation of three regime models was needed to accurately
describe the data[52]. These models were built using the two-regime models, with the
addition of a third regime that accounted for material that sorbed but did not desorb.
Irreversible chemical binding and pore clogging by mineral precipitates are two
explanations for desorption resistance. Regardless of the explanation, this regime
appears to retain organic contaminant for extensively long durations. The ability to
describe three distinct types of behavior is the advantage of using a three-regime model.
Other models for desorption exist in the literature, and are often of mixed physical and
chemical nature. For example, the multi-process nonequilibrium model divides the
kinetic behavior into a mobile and an immobile regime[53], as is the case in most
physical models. This model accounts for chemical nonequilibrium in each of the
physical regimes. The gamma distribution model[42] uses the gamma distribution

function from statistics to arbitrarily divide the soil into fractions that transfer mass at

12

different rates. This model, and the subsequent hybrid gamma/two-site model[47],
provide an empirical description of the desorption phenomena. To date, no general
consensus has been reached regarding the use of desorption models to describe
desorption data. The importance of applying state of the art desorption models lies in the
accuracy needed to describe phytoremediation. Models of contaminant desorption, and
the explanations that accompany these models, must be considered when formulating
descriptive models for phytoremediation.

Microbial biodegradation of hydrophobic organic contaminants may also be a
limiting process in the rhizosphere. Briefly, the ability of the consortia to degrade a
compound at an appreciable rate depends upon the number of degrading microbes, their
in vivo degradative capacity, and the bioavailability of the compound of interest. The
number of degrading organisms is likely to be increased in rhizosphere soils, as the plant
roots are known to exude and secrete molecules that may cause enzymatic induction or
prompt cometabolism[54]. For example, the exudation of phenolic materials is thought
to promote the bacteria that are capable of degrading PCBs[55]. Even though recent
evidence has suggested that the in vivo degradative capacity of the organisms is
debilitated[56], increased degradation remains the net effect in the rhizosphere[56]. A
net increase in rhizosphere soil over bulk soil is due to the comparative number of
organisms in the rhizosphere. The rhizosphere can support over two orders of magnitude
more microbes than unplanted soil. However, even with greater numbers, the
rhizosphere population may still be limited by constraints on bioavailability. Thus, the
bioavailability of hydrophobic organic contaminants is of great interest, as low

bioavailability will limit the rate and extent of degradation. Soil-contaminant

13

combinations, that exhibit slow desorption, are candidates for bioavailability limitations.
If the microbe, plant, or microbe-plant combination are not able to accelerate the
desorption rate, the rate of degradation will not exceed the rate of desorption.

The plant roots provide a sink for HOCs, as a portion of the root tissue is
composed of long-chain fatty acids and alcohols called suberin[57]. This waxy tissue is
named the Casparian strip, and it is chiefly responsible for sorbing material that enters via
the transpiration stream. The early effort into quantifying the potential for root sorption
was performed using pesticides[S 8]. More recent studies have been performed on poplar
sprigs placed in hydroponic media that is spiked with contaminants such as TCE[S9].
The soil solution is assumed to be the sole contaminant source (i.e. direct transfer from
the soil solid or soil gas is neglected). The amount of material present in the root tissue
versus the soil solution is assumed to follow an equilibrium-type mechanism. The
resulting equilibrium constant has been named the root concentration factor (RCF). The
RCF is commonly estimated as an empirical function of the octanol-water partition
coefficient. The total amount of contaminant sorbed to the root tissues is a function of
the aqueous concentration of the contaminant and the total mass of the root tissue. This
sorbed mass is likely small, as the root mass is relatively small compared to the mass of
rhizosphere soil. Furthermore, rate-limited desorption from the soil matrix may limit the
aqueous concentration, and thus the contaminant mass sorbed to the root.

Transpirational flow provides a significant source of dissolved nutrients and water
to the plant, which are needed for plant health. The transpiration stream will also carry
dissolved contaminant, carrying it to the stem and leaves. The development of a method

for quantifying the transpirational uptake parallels the quantification of root sorption, as

14

outlined in the pesticide literature[58]. Like the computation of the RCF, the
transpiration-stream concentration factor (TSCF) is computed using an equilibrium-type
relationship. This relationship can be empirically related to the octanol-water partition
coefficient. Therefore, the amount of contaminant in the transpiration stream is a
function of the contaminant concentration in the soil solution. The mass of contaminant
that is removed from the soil solution is a positive function of the transpiration stream
flow rate and the contaminant concentration in the soil solution. The mass of
contaminant that enters the transpiration stream may be limited by the rate of desorption
from the soil matrix to the soil solution.

Volatilization from the soil occurs when a volatile chemical (i.e. a chemical with
a high vapor pressure) enters the soil gas from either the soil solution or the soil solid.
Assuming that the soil moisture content is at an appreciable level, the amount
volatilization from the soil solid is usually neglected. For the case of dilute contaminant
concentrations in the soil solution, the concentration of a volatile contaminant in the gas
phase can be approximated using Henry’s law[40, 41]. The Henry’s law coefficient is an
equilibrium-type expression that relates the gas phase concentration to the soil solution
concentration. These values are extensively tabulated. In effect, the concentration in the
soil gas is limited by the soil solution concentration, which may be limited by desorption
from the soil matrix. Furthermore, if the soil harbors material that is not in equilibrium
with the soil gas, e. g. material that resides in domains associated with microporous water,
then this dissolved material is not to be included in the equilibrium expression. The
equilibrium concentration of the volatile contaminant provides a driving force for flux out

of the soil and into the above-ground atmosphere. The movement of contaminant in the

15

soil gas is assumed to occur primarily by Fickian diffusion. The Millingtion-Quirk
correction is normally applied to the molecular diffusivity to adjust for the soil
tortuosity[60]. The gaseous diffusion rate, is dependent upon the soil gas content, i.e. the
volume of the air-filled voids. Planted systems can increase the volatilization rate by
transpiring water, which increases the air-filled void volume and promotes a greater mass
of material to equilibrate with the soil gas. Increasing the air-filled void volume also
increases the soil gas diffirsivity, as the flow path becomes less tortuous. Consideration
of volatile and semi-volatile transport in the soil gas is important as much of this material
will exit the system in the vapor phase.

Contaminant leaching, particularly with low to moderately hydrophobic materials,
is a mechanism that occurs in planted and unplanted soils alike. Leaching is caused by
the infiltration of water[6l]—a process that removes contaminants from the rhizosphere.
The magnitude of leaching depends upon the amount of water that is infiltrating the
rhizosphere soil and the amount of contaminant dissolved in the mobile soil solution. As
the plant roots increase the soil porosity[33], the potential for contaminant leaching that
can occur during water infiltration is also increased. In this case, slow desorption from
the soil matrix is an advantage as contaminant leaching is minimized. Nevertheless, the
impact of leaching must be assessed when applying phytoremediation as a clean-up

technology.

Review of modeling efforts

Several conceptual models have been formulated to describe the action of plant

roots in the rhizosphere. Ultimately, these model formulations are based partly on

16

chemical equilibrium between compartments, certain kinetic interactions between
compartments, and mass flow between compartments. Figure 2-1 contains a simplified
model of the interactions between processes that are occurring in the rhizosphere. Often,
the concept of rate limited desorption has not been considered in the model formulations.
The following paragraphs contain a brief review of the models presented in the literature
for describing the effect of phytoremediation on organic contaminants.

An early model of phytoremediation proposed that the degradation of
contaminants in the rhizosphere is governed by a series of mass balances[62]. Balances
were included for the contaminant, the microbial biomass, the root exudates, and
dissolved oxygen. This resulted in the formulation of four, coupled partial differential
equations in time and in two spatial dimensions. Equilibrium—type formulations were
used to describe desorption from the soil, sorption to the roots, and partitioning into the
plant’s transpiration stream. Microbial growth was assumed to follow a Monod kinetic
model. Diffusion and advection in the soil solution occurred in two dimensions, as did
diffusion in the soil gas. The Henry’s law coefficient was used to describe the relative
amounts of contaminant in soil air and soil solution. A similar Henry’s law model was
used for soil oxygen. Though this model is relatively comprehensive—noticeably
missing is a mechanism that is responsible for rate-limited desorption from the soil
matrix.

A more recent phytoremediation model divides the planted system into several
compartments[63]. The compartments include: saturated and vadose zone soil; saturated
and vadose zone water; nonaqueous-phase liquid (N APL); bacterial metabolism in the

saturated zone, the vadose zone, and the root zone; plant metabolism of contaminants in

17

the root water, the stern water, and the leaf water; equilibrium-type sorption-desorption
between each solid phase and its associated liquid phase (e. g. the use of the soil-water
partition coefficient for determining the relative concentrations of contaminant in the
vadose zone water and the vadose zone soil); Henry’s partitioning between each liquid
phase and its associated gas phase (e. g. the root water and the root gas phase); and gas
diffusivity to account for the volatilization flux. Though this model is relatively
comprehensive, it does not include a mechanism for rate-limited desorption from the soil
matrix.

An investigation regarding the enhanced bioremediation of non-volatile
hydrocarbons by plants[64] considered many of the same mechanisms previously
described. An additional kinetic mechanism for contaminant degradation in a microbial
biofilm was included in this model formulation. These biofilms are claimed to surround
the roots and the simulation results suggested that enhanced degradation is due to biofilm
metabolism. The fact that mass transfer from the soil solid to the soil solution could limit
the rate of degradation in the biofilm was not addressed.

Contrary to many of the other models in the literature, Burken and Schnoor
developed a model that does include a mechanism accounting for slow mass transfer
from the soil[24]. This model was developed for assessing the rate processes that occur
in the planted bioreactors. These bioreactors contain soil that is contaminated with
atrazine. This model, relatively simple in formulation, contains compartments for slow
and fast atrazine desorption into the aqueous phase, microbial mineralization from a
bioavailable fraction of atrazine in soil, mineralization from the aqueous phase, and plant

uptake. Slow desorption was found to limit the mobility of contaminant in soils with an

18

appreciable organic matter content. However, no evidence of enhanced desorption from
desorption-resistant (i.e. non-desorbing regimes of soil) domains was explored.

The models developed for phytoremediation are relatively complex. It remains to
be seen whether much of this complexity is warranted for describing the data. The
development of a conceptual model, including mechanisms for rate limited desorption,
may be necessary to accurately describe the effect plants have on organic contaminants.
The experimental approach, presented in the following chapters of this dissertation, was
designed to assess the magnitude that desorption-resistance has on the overall rate of
phytoremediation. Contrary to the aforementioned models, and because of the inherent
complexity of the model formulation, a dimensional analysis will be developed to
simplify the interpretation of the resultant data. This dimensional analysis will be used to
develop a decision-making framework that can be used to aid the site engineer in

applying a phytoremediation technology.

Research scope

The scope of the following chapters is centered on the phytoremediation of
organic compounds, specifically the PAHs (polyaromatic hydrocarbons). As seen in
Table 2-1, the plant-assisted remediation of PAHs is thought to be accomplished by
phytostimulation. By releasing compounds with a PAH-like chemical nature, the
members of the microbial consortia that are capable of degrading PAHs are stimulated.
These microbes then degrade PAH contaminants by transformation or cometabolism.
The caveat exists in the presumption that the PAHs are bioavailable in the rhizosphere,

which may not be true. An analysis is needed to determine the rate of mass transfer that

19

may limit PAH removal in the rhizosphere. Slow mass transfer of PAHs from the soil
matrix may limit the bioavailability of PAHs. Slow mass transfer does not invalidate the
efficacy of phytoremediation as a clean-up technology, though it does prompt the
judicious selection of plant species for hastening desorption. A plant’s ability to enhance
the desorption of hydrophobic organic contaminants (HOCs), like the PAHs, would
provide evidence for an enhanced bioavailability. This would promote the use of

phytoremediation as a technique for removing hydrophobic organic contaminants.

20

 

plant root plant stem

 

 

 

 

soil-
solution
rhizosphere
consortia
soil-gas plant leaf

Figure 2-1. A simplified diagram that shows some of the interactions that occur in
planted systems.

21

Table 2-1. Types of phytoremediation (adapted from McCutcheon and Schnoor[65])

 

 

 

 

 

 

 

 

Type Definition Current use or potential use
phytodegradation the degradation or transformation chlorinated solvents, PCBs,
of organic contaminants to less energetic materials, Cl and
toxic forms by plants P based pesticides
E

phytostimulation the supply of a carbon source by organic compounds, e. g.
exudation, secretion and root BTEX, TPH, PAHs, PCBs,
necrosis that stimulates enzyme pesticides, etc.
induction or cometabolism by soil
microbes

phytostabilization stabilization of contaminants by metals, phenols,
inhibiting soil erosion, promoting chlorinated solvents
precipitation, enhancing sorption,
or causing irreversible binding of
contaminants to soil

phytocontainment the hydraulic control of water soluble contaminants
contaminants using plants to such as MTBE, chlorinated
transpire large amounts of water, solvents and energetic
thus minimizing contaminant materials
leaching

phytovolatilization volatile metals are taken up in the Sc, As, Hg, chlorinated
transpiration stream and transpired solvents
Phytoextraction contaminant uptake with the metals, radionuclides,
transpiration stream and transport relatively soluble organic
to the aerial tissues; compounds
hyperaccumulation occurs when
the compound or element exceeds
100 times its normal concentration
in the plant
R

 

 

 

 

22

 

References

10.

ll.

12.

Cunningham, SD. and W.R. Berti, Remediation of contaminated soils with green
plants: an overview. In Vitro Cellular and Developmental Biology, 1993. 29: p.
207-212.

Schnoor, J .L., et al., Phytoremediation of organic and nutrient contaminants.
Environmental Science and Technology, 1995. 29(7): p. 318A-323A.

Schwab, A.P., A.A. Al-Assi, and MK. Banks, Plant and environment interactions:
adsorption of naphthalene onto plant roots. Journal of Environmental Quality,
1998. 27: p. 220-224.

Newman, L., et al., Phytoremediation of organic contaminants: A review of
phytoremediation research at the University of Washington. Journal of Soil
Contamination, 1998. 7: p. 531-542.

Barac, T., et al., Engineered endophytic bacteria improve phytoremediation of
water-soluble, volatile, organic pollutants. Nature Biotechnology, 2004. 22: p.
583-5 88.

White, J .C ., et al., Role of organic acids in enhancing the desorption and uptake
of weathered p, p ’-DDE by C ucurbita pepo. Environmental Pollution, 2003. 124: p.
71-80.

Sung, K., et al., Plant contamination by organic pollutants in phytoremediation.
Environmental Quality, 2001 . 30: p. 2081 -2090.

Huang, X., et al., A multi-process phytoremediation system for removal of
polycyclic aromatic hydrocarbons from contaminated soils. Environmental
Pollution, 2004: p. 2-12.

Alkorta, I. and C. Garbisu, Phytoremediation of organic contaminants in soils.
Bioresource Technology, 2001. 79: p. 273-276.

Chekol, T., L. Vough, and R.L. Chaney, Plant-soil-contaminant specificity affects

phytoremediation of organic contaminants. International Journal of
Phytoremediation, 2002. 4(1): p. 17-26.

Gao, Y., L. Zhu, and W. Ling, Application of the partition-limited model for plant
untake of organic chemicals from soil and water. Science of the Total
Environment, 2004.

Sidolio'Connor, C., N. Lepp, and G. Sunderland, The combines use of
electrokinetic remediation and phytoremediation to decontaminate metal-polluted

soil: a laboratory-scale feasibility study. Environmental Monitoring and
Assessment, 2003. 84: p. 141-158.

23

13.

14.

15.

l6.

17.

18.

19.

20.

21.

22.

23.

24.

Gremion, F ., et al., Impacts of heavy metal contamination and phytoremediation
on a microbial community during a twelve month microcosm experiment. FEMS
Microbiology Ecology, 2004. 48: p. 273-283.

Podar, D., M. Ramsey, and M. Hutchings, Eflect of cadmium, zinc and substrate
heterogeneity on yirld, shoot metal concentration and metal uptake by Brassica

juncea: implications for human health risk assessment and phytoremediation.
New Phytologist, 2004. 163: p. 313-324.

Meharg, AA. and L. J ardine, Arsenite transport into paddy rice (Otyza saliva)
roots. New Phytologist, 2003. 157: p. 39-44.

Leonard, T., et al., Mercury and plants in contaminated soils: 1. uptake,
partitioning, and emission to the atmosphere. Environmental Toxicology and
Chemistry, 1998. 17(10): p. 2063-2071.

Gong, J .-M., D.A. Lee, and J .I. Schroeder, Long-distance root-to-shoot transport
of phytochelatins and cadmium in arabidopsis. PNAS, 2003. 100(17): p. 10118-
10123.

Absalom, J.P., et al., Predicting soil to plant transfer of radiocesium using soil
characteristics. Environmental Science and Technology, 1999. 33: p. 1218-1223.

Huang, J .W., et al., Phytoremediation of uranium-contaminated soils: role of
organic acids in triggering uranium hyperaccumulation in plants. Environmental
Science and Technology, 1998. 32: p. 2004-2008.

Newman, L., et al., Uptake and biotransformation of trichloroethylene by hybrid
polars. Environmental Science and Technology, 1997.31: p. 1062-1067.

Goel, A., et al., Plant cell biodegradation of a xenobiotic nitrate ester,
nitroglycerin. Nature Biotechnology, 1997. 15: p. 174-177.

Van Aken, B., J .M. Yoon, and J .L. Schnoor, Biodegradation of nitro-substituted
explosives 2, 4, 6—trinitrotoluene, hexahydro-1,3,5-trinitro-I,3,5-triazine, and
octahydro-I , 3, 5, 7-tetranitro-I , 3, 5-tetrazocine by a phytosymbiotic
Methylobacterium sp. associated with Poplar tissues(Populus deltoides X nigra
DN34). Applied and Environmental Microbiology, 2003. 70(1): p. 508-517.

Banks, M.K., et al., Degradation of crude oil in the rhizosphere of Sorghum
bicolor. International Journal of Phytoremediation, 2003. 5(3): p. 225-234.

Burken, J .G. and J .L. Schnoor, Phytoremediation: plant uptake of atrazine and
role of root exudates. Journal of Environmental Engineering, 1996. 122(11): p.
958-963.

24

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

F erro, A.M., R.C. Sims, and B. Bugbee, Hycrest crested wheatgrass accelerates

the degradation of pentachlorophenol in soil. Journal of Environmental Quality,
1994. 23: p. 272-279.

Hedge, RS. and J .S. Fletcher, Influence of Plant Growth Stage and Season on the
Release of Root Phenolics by Mulberry as Related to Development of
Phytoremediation Technology. Chemosphere, 1996. 32(12): p. 2471-2479.

Kamath, R., J. Schnoor, and P. Alvarez, Effect of root-derived substrates on the
expression of nah-lux genes in Pseudomonasfluorescens HK44: Implications for

PAH Biodegradation in the Rhizosphere. Environmental Science and Technology,
2004. 38: p. 1740-1745.

Salt, D.E., R.D. Smith, and I. Raskin, Phytoremediation. Annual Review of Plant
Physiology and Plant Molecular Biology, 1998. 49: p. 643-668.

Matthews, D.W., J. Massmann, and SE. Strand, Influence of aquifer properties
on phytoremediation effectiveness. Ground Water, 2003. 41(1): p. 41-47.

Ma, X., et al., Phytoremediation of MT BE with hybrid poplar trees. International
Journal of Phytoremediation, 2004. 6(2): p. 157-167.

Dushenkov, 8., Trends in phytoremediation of radionuclides. Plant and Soil, 2003.
249(1): p. 167-175.

Curl, EA. and B. Truelove, The Rhizosphere. 1986: Springer-Verlag Berlin
Heidelberg.

Waisel, Y., A. Eshel, and U. Katkafi, eds. Plant Roots: The Hidden Half 2002,
Marcel Dekker, Inc.: New York City. 1120.

Lynch, J .M., ed. The Rhizosphere. 1990, John Wiley and Sons, Inc.: New York
City.

Siciliano, SD. and J .J . Germida, Bacterial inoculants of forage grasses that

enhance degradation of 2-chlorobenzoic acid in soil. Environmental Toxicology
and Chemistry, 1997. 16(6): p. 1098-1104.

J ordahl, J ., et al., Effect of hybrid poplar trees on microbial populations important
to hazardous waste bioremediation. Environmental Toxicology and Chemistry,
1997. 16: p. 1318-1321.

Jones, D., Organic acids in the rhizosphere - a critical review. Plant and Soil,
1998. 205: p. 25-44.

Hillel, D., Environmental Soil Physics. 1998, San Diego: Academic Press.

25

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

Stevenson, F.J., Humus Chemistry: Genesis, Composition, Reactions. 2 ed. 1994:
John Wiley and Sons, Inc.

Thibodeaux, L.J., Environmental Chemodynamics. 1996, New York City: John
Wiley and Sons, Inc. 593.

Schwarzenbach, R.P., P.M. Gschwend, and D.M. Imboden, Environmental
Organic Chemistry. 2nd ed. 2003, Hoboken: John Wiley and Sons, Inc.

Connaughton, D.F., et al., Description of time-varying desorption kinetics:
release of naphthalene fiom contaminated soils. Environmental Science and
Technology, 1993. 27: p. 2397-2403.

Harmon, T.C. and RV. Roberts, Comparison of intraparticle sorption and
desorption rates for a halogenated alkene in a sandy aquifer material.
Environmental Science and Technology, 1994. 28: p. 1650-1660.

Johnson, MD. and J. Weber, W.J., Rapid prediction of long-term rates of

contaminant desorption from soils and sediments. Environmental Science and
Technology, 2001. 35: p. 427-433.

Lesan, HM. and A. Bhandari, Atrazine sorption on surface soils: time-dependent
phase distribution and apparent desorption hystersis. Water Research, 2003. 37:
p. 1644-1654.

Sharer, M., eta1., Time dependence of chlorobenzene sorption/desorption by soils.
Soil Science Society of America] Journal, 2003. 67: p. 1740-1745.

Ahn, I.-S., L.W. Lion, and ML. Shuler, Validation of a hybrid "two-site gamma "
model for naphthalene desorption kinetics. Environmental Science and
Technology, 1999. 33: p. 3241-3248.

Bayard, B., et al., Investigation of naphthalene in soils and soil fractions using
batch and column assays. Environmental Toxicology and Chemistry, 1998.
17(12): p. 2383-2390.

Comelissen, G., P.C.M. van Noort, and H.A.J. Govers, Mechanism of slow
desorption of organic compounds fiom sediments: A study using model sorbents.
Environmental Science and Technology, 1998. 32: p. 3125-3131.

Park, J.-H., X. Zhao, and T.C. Voice, Development of a kinetic basis for
bioavailability of sorbed naphthalene in soil slurries. Water Research, 2002. 36: p.
1620-1628.

Tabak, H.H., et a1. Determination of bioavailability and biodegradation kinetics
of polycyclic aromatic hydrocarbons in soil. in Emerging Technologies in
Hazardous Waste Management. 1995.

26

52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

Park, J .-H., et al., Kinetic modeling of bioavailability for sorbed-phase 2,4-

dichlorophenoxyacetic acid Journal of Environmental Quality, 2001. 30: p. 1523-
1527.

Brusseau, M.L., R.E. Jessup, and P.S.C. Rao, Modeling the transport of solutes
influenced by multiprocess nonequilibrium. Water Resources Research, 1989.

25(9): p. 1971-1988.

Siciliano, SD. and J .J . Germida, Mechanisms of phytoremediation: biochemical
and ecological interactions between plants and bacteria. Environmental Reviews,
1998. 6: p. 65-79.

Fletcher, J .S. and RS. Hegde, Release of phenols by perennial plant roots and
their potential importance in bioremediation. Chemosphere, 1995. 31(4): p. 3009-
3016.

Rentz, J ., P. Alvarez, and J. Schnoor, Repression of Pseudomonas putida
phenanthrene-degrading activity by plant root extracts and exudates.
Environmental Microbiology, 2004. 6: p. 574-583.

Hopkins, W.G., Introduction to Plant Physiology. 1995: John Wiley and Sons, Inc.

Briggs, G.G., R.H. Bromilow, and AA. Evans, Relationships between
lipophilicity and root uptake and translocation of non-ionised chemicals in barley.
Pesticide Science, 1982. 13: p. 495-504.

Burken, J .G. and J .L. Schnoor, Predictive relationships for uptake of organic
contaminants by hybrid poplar trees. Environmental Science and Technology,
1998.32: p. 3379-3385.

Millington, R. and J. Quirk, Permeability of porous solids. Transactions of the
Faraday Society, 1961. 57: p. 1200-1207.

Freeze, RA. and J .A. Cherry, Groundwater. 1979: Prentice-Hall, Inc.

Davis, L.C., et al., Modeling the eflects of plants on the bioremediation of
contaminated soil and ground water. Environmetanl Progress, 1993. 12(1): p. 67-
75.

Davis, L.C., eta1., Benefits of vegetation for soils with organic contaminants.
Critical Reviews in Plant Sciences, 2002. 21: p. 457-491.

Chang, Y.Y. and M.Y. Corapcioglu, Plant-enhanced subsurface bioremediation
of nonvolatile hydrocarbons. Journal of Environmental Engineering, 1998. 124(2):
p. 162-169.

McCutcheon, SC. and J .L. Schnoor, Overview of phytotransformation and
control of wastes, in Phytoremediation: Transformation and Control of

27

Contaminants, S.C. McCutcheon and J .L. Schnoor, Editors. 2003, John Wiley and
Sons, Inc.: Hoboken, New Jersey. p. 3-58.

28

Chapter 3. Describing HOC desorption profiles exhibiting distinct observational
regimes

Abstract

Desorption of organic contaminants from soil can be modeled by dividing the
desorption profile into three distinct regimes. These are an instantaneous (i.e. too fast to
measure) desorbing regime, a nonequilibrium (i.e. slow enough to measure) desorbing
regime, and a desorption-resistant (i.e. a non-desorbing) regime. Batch desorption curves
for atrazine and naphthalene on four soils were experimentally generated to demonstrate
the existence of discrete observational regimes. Nine mathematical models, each
containing mechanisms formulated to describe at least one of the three regimes, were fit
to each contaminant-soil combination using the Gauss-Newton method for parameter
estimation. Each of the nine models was ranked using the small-sample corrected Akaike
information criterion (AICc). By interpretation of the AICc values, the atrazine
desorption data was best described by three behavioral regimes. Mechanisms for fast,
slow and non-desorption were justified by the gathered data. However, AICc values
often justified the inclusion of only two regimes for naphthalene desorption data.
Estimation of an equilibrium fraction was not justified by the data because of increased
model variability, whereas models that contain slow and non-desorption mechanisms
were justified by the data. This is a result of the sparse number of data points in the slow
regime—therefore, only nonequilibrium and non-desorptive regimes are justified in

describing the data.

29

Introduction

Rationale and scope for the present study

There is widespread evidence from the field that hydrophobic organic
contaminant remediation is often limited by desorption. Desorption can be a limiting
factor for a number of remediation technologies, including: microbial bioremediation,
phytoremediation, landfarming, and pump and treat[1]. Models that accurately predict
desorption phenomena at long times are needed for design of site clean up, as desorption
is often regarded as the first step in remediation. Though in some cases the risk
associated with slow contaminant desorption may be deemed acceptable and requiring no
remediative action[2, 3], it remains important to accurately predict the amount of
contaminant sorbed to the soil using a descriptive model. Several researchers have
suggested that contaminant desorption profiles can be divided into distinct regimes,
predicated on fast, slow, and very slow desorption behavior. Several models have been
formulated with the necessary descriptive power to quantify contaminant transport within
these regimes. Though the magnitude of desorption is important for remediation design,
few models adequately describe desorption profiles in the very slow or non-desorption

regime.

Review of desorption data

It is widely accepted that contaminant desorption in situ may consist of a fast
regime, occurring at a rate too quick to measure, a dynamic regime, occurring at a rate
that can be measured, and a slow regime occurring at a rate too slow to measure. Fast

and slow desorption have been witnessed by a number of researchers. However,

30

contaminant desorption witnessed on-site is often very slow and frequently not
measurable, due to the microscale processes of sequestration[4]. An estimate of the
amount of contaminant sorbed to this very slowly desorbing fraction can be determined
via simple lab assays (i.e. desorption experiments). These desorption experiments are
typically conducted over time spans ranging from hours to months and the resultant

contaminant desorption profiles have been presented in the literature[S-l 1].

Review of modeling literature

Relatively few models include mechanisms to describe all three types of
observations (i.e. fast, slow and very slow desorption). However, rarely does observed
desorption of organic contaminants from soils reach one hundred percent, hence the need
to adequately describe slow and very slow desorption. As a significant fraction of the
contaminant remains sorbed to the soil in a recalcitrant manner, it has been hypothesized
that contaminants displaying this type of behavior are sorbed to non-desorption[l 1],
desorption-resistant[12], or irreversible[1] binding fractions of the soil organic matter.
The rate of desorption from these fractions is a function of the quantity and quality of the
soil organic matter[5]. Several models have been formulated to predict organic
contaminant desorption though few accurately and precisely describe desorption at long
times. The chemical two and three-site models presume that a nonequilibrium desorption
reaction explains the behavior of organic contaminants in soil matrices[13]. The
chemical two-site model is not formulated to describe non-desorption, while the chemical
three-site model contains an added mathematical compartment specifically for non-
desorption. The physical one[14], two and three-parameter pore-diffusion models

explain contaminant transport through soil particles using Fickian diffusive flux terms.

31

The one-parameter pore diffusion model cannot predict instantaneous desorption or
desorption at long time scales. Both the two-parameter pore diffusion model and the
three-parameter pore diffusion model can predict instantaneous desorption; though only
the three-parameter pore diffusion model can describe the desorption profile in the non-
desorption regime. Models based on the gamma distribution function abstractly partition
the soil into a number of mathematical domains, with each domain having a different
mass transfer coefficient[15]. The gamma model is not formulated to predict
instantaneous desorption while the hybrid gamma/two-site model contains a mechanism
allowing for equilibrium (instantaneous desorption) between the solid and liquid
phases[16]. These models can describe desorption at long times though both are
somewhat arduous to solve. Three and five-parameter kinetic models have also been
developed to more adequately reflect desorption behavior[7, 9]. These models are simple
in formulation, though mathematical convergence of parameter estimation algorithms
applied to desorption data with a significant non-desorption regime can be problematic.
Comparison of this approach with previous attempts at model selection
A number of models may be used to describe the contaminant desorption profile, as the
limiting mechanism involved in contaminant desorption is likely system-specific. This
work focuses on comparing these various models using atrazine and naphthalene
desorption data that was gathered from four soils. Previously, Johnson et al[17] have
compared six models using phenanthrene desorption data from three different soils.
They concluded that desorption profiles are at least biphasic and that models composed
of two regimes are good starting points for describing desorption. They further

concluded that model selection is system-specific, as the apparent desorption rate is likely

32

a function of different rate—limiting mechanisms from one soil-contaminant combination
to the next. The effort presented in this article extends Johnson’s findings by surveying
desorption data that can likely be described by models composed of three regimes.
Furthermore, this study uses the Akaike information criterion (AIC) for the purpose of
model ranking and model selection. The AIC couples the information inherent in the
data with the model formulations to optimize the balance of model accuracy and model
variability. This technique was applied to atrazine and naphthalene desorption profiles to
investigate the types of desorption data that justify the use of a two-regime models and
the types that justify three-regime models. Comparisons were also made amongst four
soil types to determine the effect the soil matrix has on the selection of two- or three-
regime models for a given contaminant. To investigate the effect that model structure has
on model selection, several different two- and three-regime models were included in this
study. Ultimately, in the case that contaminant desorption from several soils is best
described by a single model, this model becomes more than a mere description of the

data—it becomes an explanation for the data, or theory.

33

Methods and Mathematics

Contaminant desorption profiles, demonstrating the existence of distinct
behavioral regimes, were generated for both contaminants in four soils. The four soils
used in this study were: Hartsells, Capac A, Colwood A and Houghton muck. Soils were
air dried, ground, and passed through a 2-mm sieve. Soil organic carbon contents and
particle size distribution were determined by the Soil and Plant Nutrient Laboratory at
Michigan State University. Table 3-1 summarizes the properties of the soils. Soil
samples were sterilized by y-irradiation (1.29 Mrad/hr, 5 Mrad from a 60C0 source) and
stored in sealed containers at room temperature. Before each experiment using soils, 0.1
g of each soil was placed on a half-strength nutrient agar plate and incubated at 30 °C for
3 days to verify sterility. No colony-forming units (CF U) were observed.

The desorption profiles were constructed using a batch apparatus and spiking the
soil and water mixture with MC labeled atrazine and naphthalene. Atrazine or
naphthalene was spiked into each of these soils to observe desorption. The desorption
assay of naphthalene utilized batch soil slurries. An aliquot of MC-naphthalene stock (in
methanol, 7.5 g/L) was spiked into 25 mL centrifuge tubes containing 24 mL of sterile
phosphate buffer (20 mM) and each sterile soil (1.5 g of Hartsells, 1.3 g of Capac, 0.3 g
of Colwood, and 0.3 g of Houghton muck) to get 2 mg/L of initial aqueous naphthalene
concentration. The tubes were capped with a Teflon-lined Mininert® valve and screw-
sealed with polypropylene caps. A control tube without soil was prepared in the same
fashion. Tubes were tumbled at 9 rpm for 2 days in the dark, then each tube was
centrifuged for 20 min at 1200 g to separate soil, and the supernatant was sampled. The

final concentration of naphthalene in the liquid phase was determined by liquid

34

scintillation counting (LSC), and the amount of sorbed naphthalene was calculated by
difference. The supernatant was then decanted to the extent possible, and the residual
water determined gravimetrically. Desorption was initiated by adding fresh naphthalene-
free soil-extract to make-up the original volume. The tubes were tumbled again at 9 rpm,
then removed periodically and the liquid phase sampled for analysis by LSC. Samples
were initially taken at one-hour increments as this is the most dynamic regime in the
desorption profile. As the desorption profile appeared to enter the non-desorptive regime,
the time increment between samples was increased. The entire desorption profile was
collected over a period of three days. The concentration of naphthalene in the final
desorption samples were determined by LSC and verified with high-performance liquid
chromatography (HPLC). After the final desorption samples, the soil was separated from
the supernatant and extracted with methanol. The concentration of naphthalene in the
extracts was determined by LSC and verified with HPLC.

The models presented in Table 3-2 were formulated to describe the resultant
desorption data. The chemical three-site model was constructed to model irreversible
desorption[l 1, 18, 19]. This model contains a reversible regime, described using local
equilibrium; a nonequilibrium regime, parameterized by a desorption rate coefficient; and
a non-desorption regime, used to accurately model irreversible desorption profiles. The
chemical two-site model is not formulated to describe the non-desorptive (or irreversible)
regime of the desorption profile, but is formulated with an equilibrium and a
nonequilibrium compartment[20-23]. The three-parameter pore diffusion model is
formulated using a reversible regime, described using local equilibrium; a nonequilibrium

regime, parameterized by an apparent diffusion coefficient; and a non-desorption regime,

35

used to accurately model irreversible desorption profiles. This model has not been
previously presented in the literature. The two-parameter pore diffusion model[17] is not
formulated to model the non-desorptive regime of the desorption profile, but is
formulated with an equilibrium and a diffusion limited compartment. The one-parameter
pore diffusion model[6, 10, 14, 24-26] only contains a diffusive compartment, and cannot
model instantaneous desorption or non-desorption. The gamma distribution model[lS]
uses a gamma distribution firnction to mathematically partition the soil into a number of
distinct fractions. This model is strictly empirical as it is parameterized by two
coefficients, a and B. Neither of these parameters has an easily discernable physical
meaning. The hybrid gamma/two-site model[l6] assumes that the desorption from each
of the sites, assigned by the gamma distribution function, proceeds in the same manner as
that would occur in the chemical two site model. The added parameter, feq, improves the
model accuracy for describing certain desorption profiles. Both gamma distribution
models were somewhat cumbersome to program and interpret. The five-parameter
kinetic model[7, 9, 17] consisted of a fast desorption regime, a slow desorption regime
and a very slow desorption regime. Each regime is quantified using first order kinetics
and the resultant solution is a summation of exponentials. This model was difficult to
solve as the very slow regime kinetic constant converges to a value near zero.
Application of this model to the data in this article proceeded by equating the
accumulation rate for this fraction to zero. Thus a modified five-parameter kinetic model
only uses four parameters. The three-parameter kinetic model[8, 9, 27, 28] is similar to
the five-parameter model with the exception that it does not contain a very slow

compartment. Again, in order to provide convergence, the slow compartment

36

accumulation rate was set to zero, and the modified three-parameter kinetic model only
uses two parameters.

Each of the models presented in Table 3-2 was solved analytically or numerically
for the purpose of parameter estimation. All of the solutions were programmed using the
Matlab software package and the output was generated using Matlab’s graphical user
interface. The error sum of squares between the data and the model fits were used as
objective functions for parameter estimation. These objective functions were minimized
using the Gauss-Newton method assuming a tolerance of 10'3 as the convergence
criterion[29]. A modified version of the Gauss-Newton method was used for problems
where the dependent variables were nonlinearly related to the independent variables, such
as for the gamma and hybrid gamma/two-site models. Stiffness was overcome for both
the gamma distribution and hybrid gamma distribution models by using reduced
sensitivity coefficients. Parameter standard errors were estimated from the resultant
sensitivity matrix.

Model inference was accomplished through application of the Akaike information
criterion (AIC)[30, 31]. This criterion uses estimates of accuracy and precision as a
means of model inference and subsequent selection. The ranking of models based on
accuracy alone is not sufficient because model variability is not included. Akaike
formulated the AIC by noticing a relationship between the Kullback-Leibler distance and
the maximized log-likelihood function. The leftmost term on the right-hand side of
equation 3-1 represents a penalty for under-fitting data and the right term is a penalty for

over-fitting data.

37

AIC : n10g(62) '1' 2K (Equation 3_1)

The AIC tells what inferences the data support, not what reality might be. In a sense, the
AIC is a quantitative Occam’s razor (a rule that states that the simplest of competing
descriptions is preferred) that can be used to select the most parsimonious model
supported by the collected data. The principle of parsimony insists that the best-fit model
is the model with an optimal combination of bias and variability (contrary to R2, which
only determines goodness of fit based on accuracy alone). Because the AIC is the sum of
two penalty terms (i.e. one for bias and one for uncertainty), the smaller AIC values
correspond to models that fit the data more parsimoniously. Models are then ranked

according to each AIC value.

The small-sample corrected AIC—given the acronym AICc (equation 3-2), is used in this

study, as warranted by the small sample size relative to the number of model parameters.

2K(K +1)

AICc = AIC +
n — K ‘1 (Equation 3-2)

The nine models chosen for this inquiry were ranked for each soil-contaminant

combination using the value of the AICc.

38

Results and Discussion

Desorption data for Capac A-atrazine and Muck-naphthalene were fit with nine
mathematical models using the Gauss-Newton technique. Eight of the nine models are
presented in Figure 3-1 (Capac A-atrazine) and Figure 3-2 (Muck-naphthalene). Note
that the atrazine profile approaches the maximum desorbed amount in a more gradual
manner than does naphthalene profile. Therefore, there are more data points in the
dynamic region of the atrazine desorption profile than in naphthalene desorption profile.
Consequently, the atrazine data supports the inclusion of an equilibrium site in the model
formulations, whereas the naphthalene data does not support the inclusion of an
equilibrium site. The details of each soil-contarninant combination are described and

discussed in the following sections.

C apac A

The AICc values for Capac A soil are presented in Figure 3-3 for both atrazine
and naphthalene. Increasing the number of regimes is beneficial for atrazine desorption
from Capac A, and the desorption data is best described by the chemical three-site model.
The three-regime models that contain an equilibrium site—such as the chemical three-site
model, the three-parameter pore diffusion model, and the hybrid gamma/two-site
model—are favorable to the three-regime model that does not include and equilibrium
site, namely the five parameter kinetic model. Likewise, the two-regime models that
contain an equilibrium site—such as the chemical two-site and the two-parameter pore
diffusion models, are better formulated for describing Capac A-atrazine desorption data

when compared to the models that do not contain an equilibrium site, namely the gamma

39

and the three-parameter kinetic models. Unlike the model ranking for the three-regime
models, the two-parameter pore diffusion model provides the best description amongst
the two-regime models, as the two-parameter pore diffusion model is more capable of
handling curvature than is the chemical two-site model. The one-parameter pore
diffusion provides a poor description of Capac A-atrazine desorption data.

As with the Capac A-atrazine data, the Capac A-naphthalene data warrant an
additional regime in the chemical site and the pore diffusion models (Figure 3-4). The
chemical three-site and the three-parameter pore diffusion models provide superior
descriptions of the Capac A-naphthalene data than do the chemical two-site and the two-
parameter pore diffusion models, respectively. Contrary to the Capac A-atrazine data,
the Capac A-naphthalene data does not warrant an additional regime for the gamma and
kinetic models. Adding an equilibrium site to the gamma model to form the hybrid
gamma/two-site model does not improve the gamma model’s description of the Capac A-
naphthalene data. This is a result of the sparseness of the data at short desorption times
for the Capac A-naphthalene desorption profiles. In other words, there is a lesser amount
of data in the curved portion of the desorption profile for Capac A-naphthalene than for
the Capac A-atrazine, and therefore the ability to estimate an equilibrium site fraction is
negatively impacted. The additional parameters, used to formulate the five-parameter
kinetic model from the three-parameter kinetic model, are not justified by the data. Also,
converse to the Capac A-atrazine data, the Capac A-naphthalene is best described by a
two-regime model, namely the three-parameter kinetic model. Again the sparseness of
data in the early portion of the desorption profile negates the inclusion of an equilibrium

site fraction, a fraction that is not present in the three-parameter kinetic model. The

40

accuracy gained by addition of a further desorption regime is not significant compared to

the added parameter uncertainty.

Colwood A

The Colwood A-atrazine data follows the same trend as the Capac A-atrazine
data—specifically, the three-regime models provide better descriptions of the data than
the two-regime models (Figure 3-5). Again, the chemical three-site model provides the
best description of the data. The two-regime models that can adequately model curvature
in the desorption profile, such as the two-parameter pore diffusion and the gamma
models, are superior to those that cannot, namely the chemical two-site model.

The Colwood A-naphthalene data is best modeled by the two-parameter pore
diffusion model and the gamma model (Figure 3-6). In the case of the pore diffusion
models, adding a regime does not change the AICc value, and in the case of the gamma
family of models, the addition of an equilibrium site fraction does not improve the
description of the data. Converse to the Capac A-naphthalene data, the kinetic models
are improved by the addition of a regime. Like Capac A-naphthalene, the two-regime

models provide the best description of the Colwood A-naphthalene desorption profile.

Hartsells

The Hartsells-atrazine data is best described by the three-regime models (Figure
3-7), as with Capac A-atrazine and Colwood A-atrazine. Again, the chemical three-site
model provides the best description of the desorption profile. Of the two-regime models,

the two-parameter pore diffusion model provides the best description of the Hartsells-

41

atrazine desorption profile. The one-parameter pore diffusion model is a poor description
of the Hartsells-atrazine data.

The Hartsells-naphthalene data follows the same trend as the Capac A-
naphthalene data (Figure 3-8). For the chemical site and pore diffusion models,
increasing the number of regimes provides a superior description of the data. However,
the converse is true for the gamma and kinetic models. As with Capac A-naphthalene,
the gamma and the three-parameter kinetic models provide the best description of the
Hartsells-naphthalene desorption profiles, a consequence of the sparseness of early

desorption data and the lack of improved fit, respectively.

Muck

The Muck-atrazine data is best described by the three-regime models (Figure 3-9),
as with Capac A-atrazine, Colwood A-atrazine, and Hartsells-atrazine. Again, the
chemical three-site model provides the best description of the desorption profile. Unlike
Capac A-atrazine, Colwood A-atrazine, and Hartsells-atrazine, the gamma model is the
best two-regime model for describing Muck-atrazine desorption data. The one-parameter
pore diffusion model is a poor description of the Muck-atrazine data.

Of note is the sparseness of data in the curved portion of the profile. The Muck-
naphthalene data follows the same trend as the Capac A-naphthalene and Hartsells-
naphthalene data (Figure 3-10). For the chemical site and pore diffusion models,
increasing the number of regimes provides a superior description of the data. However,
the converse is true for the gamma and kinetic models. As with Capac A-naphthalene

and Hartsells-naphthalene, the gamma and the three-parameter kinetic models provide the

42

best description of the Muck-naphthalene desorption profiles, a consequence of the

sparseness of early desorption data and the lack of improved fit, respectively.

Discussion of contaminants

Atrazine desorption data tends to be well described by models in the three-regime
category. In all cases, increasing the number of parameters is justified by the decrease in
the AICc value. Of the three regime models, the chemical three-site model tends to
provide the best description of the data. Further benefits from using the chemical three-
site model are: l) the relative ease of programming and 2) the relative ease of interpreting
the values of the estimated parameters. Of the three-regime models, the modified five-
parameter kinetic model provides the worst fit of the data. The five-paramenter kinetic
model lacks an equilibrium compartment, and is therefore unable to accurately describe
the early portion of the desorption data. Of the two-regime models, the gamma model
provides the best fit of the data. This model is better formulated to describe the
intermediate and late portions of the desorption data, as it is capable of more accurately
fitting the curvature present in the desorption profile. The one-parameter pore diffusion
model does not accurately describe the desorption data, as it is not formulated to describe
the early and late portions of the desorption profile.

The naphthalene desorption data provides an exception to the conclusions drawn
from the atrazine desorption data. The gamma model tends to be justified by the data
over the three-regime models. This is a result of the gamma model’s ability to describe
curvature in the intermediate and late portions of the desorption profile, and the nature of
the naphthalene desorption data. Relatively few data points are collected during the early

desorption profile as naphthalene desorption rapidly approaches the maximum fraction

43

desorbed. Models with an equilibrium compartment are not as justified for naphthalene
as for atrazine, because the estimation of the equilibrium compartment size is based on a
small number of data points. Therefore, a model that is capable of describing the
intermediate and the late portions of the desorption profile—while not making estimates
of an equilibrium compartment based on sparse data—will tend to have a lower AICc
value. As such, the reformulation of the gamma model to include an equilibrium
compartment, i.e. the hybrid gamma/two-site model, is not justified. Accuracy in
predicting the intermediate and the late portions of the desorption profile is still justified,
as the chemical three-site and the three-parameter pore diffusion models remain good

descriptors of these types of desorption profiles.

Discussion of models

The three regime-models fit the atrazine data better than the two-regime models.
For the chemical site models, the pore diffusion models, and the kinetic models, the
additional parameters provide an improved fit to the data. For the gamma models, the
fact that the additional parameterization involves an equilibrium site fraction that is
justified by the nature of the early portion of the atrazine desorption profiles is an added
benefit. Of the two regime models, the two-parameter pore diffusion and the gamma
models tend to provide the best description of the data, a result of each model’s ability to
fit curvature compared with the other two regime models. For naphthalene, the gamma
and the three-parameter kinetic models provide the best description of the desorption
profiles. This is an artifact of the sparseness of the early desorption data which leads to
an inability to precisely determine the equilibrium site fraction. Finally, the one-

parameter pore diffusion model consistently provides the poorest description of the

44

desorption data, as it predicts a lower desorbed fraction at early times, and greater

desorbed fractions at longer times.

Conclusions

For desorption data exhibiting profiles reminiscent of atrazine desorption, the
chemical three-site model provides the best description of the data. For desorption data
exhibiting profiles reminiscent of naphthalene desorption, the gamma and the three-
parameter kinetic models provide the best descriptions of the data. The three-parameter
kinetic model is recommended over the gamma model, because of the relative ease

involved in formulation and subsequent interpretation of estimated parameters.

45

2 regime models 3 regime models

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

100 100
<§
50 (a) 50 (b)
AICc = -73.52 Alcc = -108.50
0 0
0 50 0 50
100 100
A 50 (c) 50l(d)
,,\° AICc = -76.49 AICc = -104.87
.3 00 50 0O 50
E'
8
8 100 100
<§
50 (e) . 50 (f)
A'Cc = -73.26 AICc = -104.00
0 0
0 50 0 50
100
l 50 (h)
Alcc = -69.13 AICc = 403.27
0 0
0 50 0 50

Time (hours)

Figure 3-1. Best-fit models to atrazine-Capac A desorption data: (a) the chemical
two-site model, (b) the chemical three-site model, (c) the two-parameter pore
diffusion model, (d) the three-parameter pore diffusion model, (e) the gamma
model, (1') the hybrid gamma/two-site model, (g) the modified three-parameter

kinetic model and (h) the modified five-parameter kinetic model.

46

2 regime models 3 regime models

 

 

 

 

 

 

 

100 O O O 100
50 (a) 50 (b)
AICc = -50.5524 AICc = -67.9753
0 0
0 50 100 o 50 100
100

 

 

50 (c)
AICc = -60.7416 AICc = -70.9167

 

 

 

 

 

50 1 00 0 50 1 00

Desorption (%)
O

 

 

   

 

 

 

 

 

 

 

 

 

 

 

AICc = -82.6367 AICc = 63.2616
0 0 ,
0 50 100 0 50 100
100
50 (h)
Alcc = -83.3884 Arc:c = -719145
0 0
o 50 100 0 50 100

Time (hours)

Figure 3-2. Best-fit models to naphthalene-Muck desorption data: (a) the
chemical two-site model, (b) the chemical three-site model, (c) the two-
parameter pore diffusion model, ((1) the three-parameter pore diffusion model,
(e) the gamma model, (0 the hybrid gamma/two-site model, (g) the modified

three-parameter kinetic model and (h) the modified five-parameter kinetic

model.

47

      

7////////// m g 265%
7////// m a 2.2.5
7/////////// m ////////////,_ Ea;

,/////// n a mass.
7///////// m g are?

,x/////// m a 6666-...
a v. % 866%;
L

w//////////// m g .63

 

 

 

7////// u a .63

0

0
0
0
20

_ou_<_

Figure 3-3. Comparison of the AICc for the nine models that were fit to Capac A-

atrazine desorption data.

48

100

[AICc]
ON
C

 

2-srte
3-site

«3 v .2 2
E E 6 6
g >, s: s:
—: I— I"

co '3!- —¥
1”) V3

l-diffusion
2-diffusion
3-diffusion

Figure 3-4. Comparison of the AICc for the nine models that were fit to Capac A-
naphthalene desorption data.

49

     

90
80‘ 7
, / . /

70* 7 / V / I
§ 50. é / g r r
jg40/ 86 / 82 / 32 /

69 72 73 66 71

30“ 49

20*

mi aiiéiii

0 “A .1 . -2, L- L -_ m E

Figure 3-5. Comparison of the AICc for the nine models that were fit to Colwood A-
atrazine desorption data.

50

|A1Cc|

 

2-site if:

3-srte

hybrid
3-kinetic f;
S-kmetrc

2-diffusion
3-diffusion 2;.
gamma

c:

o
E
55
'?
fl

Figure 3-6. Comparison of the AICc for the nine models that were fit to Colwood A-
naphthalene desorption data.

51

120

      

 

 

 

 

 

100‘, y 7 V .
‘5 80‘ / g 7 g / g g
B i 7 / / / 7 / /

./ / / / / /
g 60 ¢ 102 % 101 100

40 77 g % 76 g 76 72

20% 36 4% g Q a

0 -g I. -4- an -2, / -L, -‘ w

Figure 3-7. Comparison of the AICc for the nine models that were fit to Hartsells-
atrazine desorption data.

52

 

 

|AICc|

 

 

 

c3 '0
E E
g >.
on .5:

3-k1netrc
S-kinetic

l -d1ffusron
2-diffusion
3-diffusion

Figure 3-8. Comparison of the AICc for the nine models that were fit to Hartsells-
naphthalene desorption data.

53

      
   

100+
90~ y 7 .
« / //
r . r r
/ / / /
70~ ’/ / / / , /
3 60* Z g Z Z % Z
:5 d / / d d
— 50* 7 / 91 / 91 / 38
4070 / / 74 79
30' g 51 / / Z 69
f3; Z a g ? Z a
is? 33, .5 .5 E E E {3’ {5,

Figure 3-9. Comparison of the AICc for the nine models that were fit to Muck-
atrazine desorption data.

54

|A1Cc|
U]
0

 

2-site . 1?? iiéitiiii‘i
3-51te

E .2 9.
a... o-a H
O.) Q)
‘3. .5 E
M W

l-diffusion if“?

2-diffusion :gj‘ji

3-diffusion
gamma

Figure 3-10. Comparison of the AICc for the nine models that were fit to Muck-
naphthalene desorption data.

55

Table 3-1. Selected properties of sorbents used in this study.

 

 

Soil % 0.0a % Sand % Silt % Clay pH CECE
[cmol(+)/kg]

Hartsells 1.29 59.1 32.1 8.78 5.3 7.10

Capac A 3.28 54.6 24.0 21.4 6.8 24.4

Colwood A 7.80 64.2 20.7 15.1 6.0 43.0

Houghton Muck 38.3 We ND ND 5.1 156

 

a , b . . C .
O.C.: organic carbon content; CEC: cation exchange capacrty; ND: not deterrnrned.

56

Table 3-2. Model equations fit to desorption data sets.

Model
Name

Chemical
three-site
[l 1, 18, 19]

Chemical
two-site
[20-23]

Three-
parameter

pore
diffusion

Two-
parameter
pore
diffusion
[17]

One-
parameter
pore
diffusion [6,
10, 14, 24-
26]

Hybrid
gamma/two-
site [16]

Gamma
distribution
[15]

Five-
parameter
kinetic [7, 9,
l 7]

Three-
parameter
kinetic [8, 9,
27,28]

 

 

 

 

 

 

 

 

 

Sr :(l'fs)sT’ Ss :fsST

Equation
dSnc
dt " = «(Sm —f,,,quc)
Snd:fnd1<dCeq(sorp)
dSne
dt q : —k(Sneq _ fnqu<dC)
ST = Scq +Sneq +Snd
as",q = D 625,,q + _2_ as“,q
6% 682 r 68'
ST 2 Seq +Sncq
2
83mq ___ D a s",q + 3 as",q
6% .312 1' 51
9L1) 628 ____+ 2 6_s
éh éhTz r éh'
00 ka—IBae—Bki
__ clk 1 dk'
31%;"ng ki(Si (lf eq)Kd / 14(0) 1
00 ka—l (I —Bki
- e
“S—l_k i( ch) kl B dki
0 F(a)
dS dS dS
_dtL = ’krSr’ _—S _ “ksssa f = ‘kvssvs
Sr = erTa S5 = fSST’ Svs = (l‘fr 'fslsT
L; -1. 3,, 1.8. . 1,3,

th.# 11
(3-3)

3
(3-4)
(3-5) 2
(3-6)

3
(3-7)
(3-8)

2
(3-9)
(3-10) 1
(3-11) 3
(3-12) 2

(3-13),(3-14)
(3-15),(3-16) 4‘
(3-17),(3-18)

(3-19)
(3-20)
(3-21)
(3-22)

* kVS is set to zero in the five-parameter kinetic model and kS is set to zero in the three-
parameter kinetic model and therefore are not counted as parameters.

57

Table 3-3. List of Symbols.

AIC

AICc
c

CCQ(SOFP)
D

feq
fnd
fneq

wxgnpqh

W
o.

(h

CDC/)C/JHDWWR‘W
< '1'
m

Akaike information criterion
small-sample corrected Akaike information criterion

contaminant concentration is the aqueous phase
contaminant concentration in the aqueous phase at sorption equilibrium

apparent diffusion coefficient
equilibrium site fraction

non-desorption site fraction
Nonequilibrium site fraction
rapidly desorbing fraction
slowly desorbing fraction
very slowly desorbing fraction

number of estimated parameters
first-order rate coefficient
soil-water partition coefficient

. .th .
first-order rate coefficrent for the 1 5011 compartment
first-order rate coefficient for the rapid fraction
first-order rate coefficient for the slow fraction
first-order rate coefficient for the very slow fraction

number of data points

radial distance

contaminant concentration in soil

contaminant concentration in the equilibrium soil compartment

contaminant concentration in the ith soil compartment

contaminant concentration in the non-desorption soil compartment
contaminant concentration in the nonequilibrium soil compartment
contaminant concentration in the rapid compartment

contaminant concentration in the slow compartment

total contaminant concentration in all soil compartments

contaminant concentration in the very slow compartment
Time

gamma distribution function

shape parameter

scale parameter

sample variance

58

References

l. Kan, A.T., et al., Irreversible sorption of neutral hydrocarbons to sediments:
experimental observations and model predictions. Environmental Science and
Technology, 1998. 32(7): p. 892-902.

2. Alexander, M., Aging, bioavailability, and overestimation of risk fiom
environmental pollutants. Environmental Science and Technology, 2000. 34(20):
p. 4259-4265.

3. Pignatello, J .J . and B. Xing, Mechanisms of slow sorption of organic chemicals to

natural particles. Environmental Science and Technology, 1998. 30(1): p. 1-11.

4. Luthy, R., et al., Sequestration of hydrophobic organic contaminants by
geosorbents. Environmental Science and Technology, 1997. 31(12): p. 3341-
3347.

5. Kukkonen, J ., et al., Sediment characteristics aflecting desorption kinetics of

select PAH and PCB congeners for seven laboratory spiked sediments.
Environmental Science and Technology, 2003. 37 : p. 4656-4663.

6. Wu, S. and P. Gschwend, Sorption kinetics of hydrophobic organic compounds to
natural sediments and soils. Environmental Science Technology, 1986. 20: p.
717-725.

7. Comelissen, G., et al., Temperature dependence of slow adsorption and

desorption kinetics of organic compounds in sediments. Environmental Science
and Technology, 1997. 31: p. 454-460.

8. Comelissen, G., et al., Rapidly desorbingfiactions of PAHs in contaminated
sediments as a predictor of the extent of bioremediation. Environmental Science
and Technology, 1998. 32: p. 966-970.

9. Comelissen, G., P.C.M. van Noort, and H.A.J. Govers, Mechanism of slow
desorption of organic compounds from sediments: A study using model sorbents.
Environmental Science and Technology, 1998. 32: p. 3125-3131.

59

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

Grathwohl, P. and M. Reinhard, Desorption of trichloroethylene in aquifer
material: rate limitations at the grain scale. Environmental Science and
Technology, 1993. 27: p. 2360-2366.

Park, J .-H., X. Zhao, and T. Voice, Biodegradation of non-desorbable
naphthalene in soils. Environmental Science and Technology, 2001. 35(13): p.
2734-2740.

White, J .C. and M. Alexander, Reduced biodegradability of desorption-resistant

fractions of polycyclic aromatic hydrocarbons in soil and aquifer solids.
Environmental Toxicology and Chemistry, 1996. 15(11): p. 1973-1978.

Brusseau, ML. and P.S.C. Rao, Sorption nonideality during organic contaminant
transport in porous media. Critical Reviews in Environmental Control, 1989.
19(1): p. 33-99.

Crank, J ., The Mathematics of Diffusion. First ed. 1956: Oxford at the Clarendon
Press.

Connaughton, D.F., et al., Description of time-varying desorption kinetics:
release of naphthalene fiom contaminated soils. Environmental Science and
Technology, 1993. 27: p. 2397-2403.

Ahn, I.-S., L.W. Lion, and ML. Shuler, Validation of a hybrid "two-site gamma "
model for naphthalene desorption kinetics. Environmental Science and
Technology, 1999. 33: p. 3241-3248.

Johnson, M., T. Keinath, and W. Weber, A distributed reactivity model for
sorption by sails and sediments. I 4. Characterization and modeling of
phenanthrene desorption rates. Environmental Science and Technology, 2001.

35(8): p. 1688-1695.

Park, J .-H., Bioavailability of sorbed organic contaminants, in Department of
Civil and Environmental Engineering. 2000, Michigan State University.

Park, J .-H., X. Zhao, and T.C. Voice, Development of a kinetic basis for
bioavailability of sorbed naphthalene in soil slurries. Water Research, 2002. 36:
p. 1620-1628.

60

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

Rao, P.S.C., J .M. Davidson, and HM. Selim, Evaluation of conceptual models for
describing nonequilibrium adsorption-desorption of pesticides during steady-flow
in soils. Soil Science Society of Americal Journal, 1979. 43: p. 22-28.

Selim, H.M., J .M. Davidson, and P.S.C. Rao, Transport of reactive solutes
through multilayered soils. Soil Science Society of Americal Journal, 1977. 41: p.
3-10.

van Genuchten, M.T., Two-Site/Two-Region Models for Pesticide Transport and
Degradation: Theoretical Development and Analytical Solutions. Soil Science
Society of Americal Journal, 1989. 53(5): p. 1303-1310.

Cameron, DR. and A. Klute, C onvective-dispersive solute transport with a
combined equilibrium and kinetic adsorption model. Water Resources Research,
1977. 13(1): p. 183-188.

Farrel, J. and M. Reinhard, Desorption of halogenated organics fiom model
solids, sediments, and soils under unsaturated conditions. 2. Kinetics. Environ.
Sci. Technol, 1994. 28: p. 63-72.

Werth, OJ. and M. Reinhard, Eflects of temperature on trichloroethylene
desorption from silica gel and natural sediments. 2. kinetics. Environmental
Science and Technology, 1997. 31: p. 697-703.

Harmon, T.C. and RV. Roberts, Comparison of intraparticle sorption and
desorption rates for a halogenated alkene in a sandy aquifer material.
Environmental Science and Technology, 1994. 28: p. 1650-1660.

Johnson, MD. and J. Weber, W.J., Rapid prediction of long-term rates of
contaminant desorption fiom soils and sediments. Environmental Science and

Technology, 2001. 35: p. 427-433.

Karickhoff, SW. and KR. Morris, Sorption dynamics of hydrophobic pollutants
in sediment suspensions. Environmental Toxicology and Chemistry, 1985. 4: p.
469-479.

Englezos, P. and N. Kalogerakis, Applied Parameter Estimation for Chemical
Engineers. Chemical Industries, ed. H. Heinemann. 2001, New York: Marcel
Dekker, Inc.

61

30.

31.

Burnham, K.P. and DR. Anderson, Model Selection and Inference. First ed. 1998,
New York City: Springer-Verlag.

Ginzburg, LR. and CS]. Jensen, Rules of thumb for judging ecological theories.
Trends in Ecology and Evolution, 2004. 19(3): p. 121-126.

62

Chapter 4. Enhanced removal of naphthalene by two plant species used for
phytoremediation

Abstract

The enhanced removal of naphthalene from soil by a monocot, big bluestem
(Andropogon gerardii), and a dicot, white mulberry (Morus alba), has been demonstrated
over a period of 186 days using two different soils. Soil naphthalene concentrations
determined by solvent extraction were lower for the planted treatments than in the
unplanted treatments at the end of the trial. The amount of naphthalene that is water
extractable approaches a constant value for both unplanted treatments. This implies that
a fraction of naphthalene that is water extractable is not available for gaseous diffusion.
However, the water extractable portion of naphthalene continues to decrease in all
planted treatments. This result is significant as a fraction of naphthalene that is not

mobile by gaseous diffusion is available to the plant’s transpiration stream.

63

Introduction

Phytoremediation is an emerging and promising technique for the cleanup of
organic contaminants[l-S]. It has widely been perceived as a low-cost alternative to
other remediation technologies[6-10]. However, phytoremediation may be subject to the
same bioavailability constraints as other bioremediative technologies, e. g. microbial
biodegradation[l 1-17]. In microbial bioremediation, bioavailability can be limited by the
extent and rate of desorption from the soil matrix to the microorganism[l7], meaning that
mass transfer is a critical factor in most cases of bioremediation[12]. Similarly,
desorption may limit the bioavailability of organic contaminants for phytoremediation.
Low bioavailability limits the endpoint contaminant concentration achievable by
microbial bioremediation, and could limit the rate and degree of phytoremediation.
Because regulatory agencies can set endpoint contaminant criteria at soil concentrations
that are below the amount that is able to desorb, often a significant fraction of
contaminant is difficult—but necessary—to remove. As such, a demonstrated ability to
achieve the low-level endpoint criteria set by regulatory agencies would enhance
phytoremediation’s appeal.

Several studies have demonstrated the efficacy of plants for removing PAHs
(polyaromatic hydrocarbons) from soil. Aprill and Sims[1 8] showed that eight varieties
of prairie grass were found to enhance the removal of benzo(a)anthracene, chrysene,
benzo(a)pyrene and dibenzo(a,h)anthracene. Naphthalene removal from a field plot was
enhanced by the addition of prairie buffalo grass (Buchloe dactyloides var. prairie)[19].
Alfalfa (Medicago sativa L.), switchgrass (Panicum vergatum) and little bluestem

(Schizachyrium scoparius) removed 56, 57 and 47 percent, respectively, of the total

64

PAHs at a manufactured gas plant[20]. The presence of tall fescue (F estuca arundinacea)
enhanced the degradation of benzo(a)pyrene in soil[21]. A growth chamber study using
tall fescue and switchgrass degraded more pyrene than an unplanted control[22]. The
rhizosphere soils of tall fescue and wheat (Triticum aestivum) contained a greater amount
of phenanthrene and pyrene than did an unplanted treatment[23], demonstrating enhanced
PAH mobility. Alfalfa and reed (Phragmites australis) were shown to degrade 74.5 and
68.7 percent, respectively, of the 16 EPA priority pollutant PAHs in two years[24].
Pyrene degradation was greater in planted versus unplanted treatments[25] and levels of
pyrene declined from 100 ppm to 2.4 ppm in 24 weeks. A number of studies have
demonstrated the ability of plants to remove PAHs from soil, though incomplete removal
is a common result in many of these trials.

The final achievable contaminant level is often limited by the PAH bioavailability.
Low bioavailability can be attributed to slow desorption from the soil matrix. Several
studies have demonstrated the limitations of hydrophobic organic contaminant desorption.
The effort presented in Chapter 3 of this dissertation demonstrated the tendency of
organic contaminant desorption to be segregated into different behavioral regimes, i.e.
equilibrium, nonequilibrium, and non-desorption. The non-desorption regime is
characterized by very slow desorption (i.e. immeasurably slow) from the soil matrix. It
has been hypothesized that contaminant sorbed to non-desorption regimes is not
bioavailable. However, the concept of bioavailability is somewhat more complex, as
microbial degradation rates tend to be faster than expected and non-desorbable material
can be degraded. White and Alexander[26] found that phenanthrene and naphthalene

sorbed to desorption-resistant soil fractions were biodegraded, albeit slowly. Guerin and

65

Boyd demonstrated that Pseudomonas putida 17484 accessed naphthalene sorbed to
domains deemed unavailable for biodegradation, a result attributed to cell attachment[27].
Evidence supporting enhanced bioavailability by attachment has been gathered by Park et
al.[16, 28, 29] with Pseudomonas putida G7 and NCIB 9816-4. Enhanced liquid-phase
biodegradation rates were observed and attributed to solid-phase degradation. Park et al.
determined that naphthalene and atrazine[30] sorbed to non-desorbable regimes could be
degraded by microorganisms. A firrther study involving soil-sorbed biphenyl
demonstrated biodegradation by attached gram-positive and a gram-negative bacteria[15].
Organisms in close proximity with the sorbed contaminant, as is the case with cell
attachment, appear capable of enhancing bioavailability. Likewise, the contact that exists
between the plant root and the soil matrix may also enhance bioavailability.

The focus of this investigation is on understanding the relationship between the
contaminant removed in planted systems and the desorbability of the remaining
contaminant. White mulberry (Morus alba) and big bluestem (Andropogon gerardii)
were grown in two soils for assessing the removal of soil-sorbed naphthalene and
inferring the mechanisms responsible for limiting this removal. Desorption is thought to
be the first step, and a limiting step, in the phytoremediation of sorbed organic
contaminants. Contaminant desorption is followed by transport in the aqueous phase to
the root surface, where the contaminant can be taken into the root with the transpiration
stream water. This description implies that desorption limits the contaminant
bioavailability, and hence, the degree of phytoremediation. However, analogous to
microbial bioavailability, the existence of mechanisms responsible for enhancing

bioavailability would greatly add to phytoremediation’s appeal as a cleanup technology.

66

Plants are associated with a variety of mechanisms that may be able to remove
contaminants residing in non-desorption domains of soil. The secretion of enzymes[9,
31], such as monooxygenases and peroxidases, may degrade recalcitrant contaminants.
Secretion and exudation of compounds that stimulate the microbial consortia to degrade
contaminants is also a potential mechanism for enhancing desorption [9], as is the plant
promoted production of surfactants[9]. Any combination of these mechanisms may be
manifested as an enhancement of PAH bioavailability. The presence or absence of this
bioavailability enhancement effect will alter the application of plants for remediation of
contaminated soils. A plant’s tendency to enhance bioavailability, analogous to a
microbe’s tendency to enhance bioavailability, is important knowledge when applying

phytoremediation as a cleanup option and achieving the necessary regulatory endpoints.

67

Materials and Methods

Soil contamination

Two soils, Spinks A horizon (SpAf) and Kalkaska A horizon (Kal A), were
sterilized by exposure to 5 Mrads of gamma radiation at the Phoenix Memorial
Laboratory (University of Michigan, Ann Arbor, MI). The sterile soil was placed in a
one gallon steel can that was sterilized in an autoclave. 100 ml of a 1:1 mixture of
acetonezpentane was used as the carrier solvent for delivering 500 mg of naphthalene to 1
kg of soil. Naphthalene was added to air dry SpAf and Kal A soils to achieve a
concentration of 500 ppm. The soil filled cans were mixed by rotation at approximately
12 rpm for 12 hours. Table 1 contains the measured soil characteristics for SpAf and Kal

A.

Plant tissue culture

Two species of plant were used for this experiment, Andropogon gerardii and
Morus alba. A. gerardii is a monocot with a fibrous root system and M alba is a dicot.
A. gerardii seeds were germinated on moist, sterile filter paper placed in a Petri dish.
The Petri dishes were placed in a growth chamber set for a 16 hour to 8 hour light to dark
cycle and a temperature of 26°C. M alba was propagated by vegetative stem cuttings
from a single plant that was surface sterilized with a ten percent bleach solution. Nodal
explants were placed in shooting media for two weeks and then moved to rooting media
for an additional three weeks. Plants that were approximately two centimeters tall were

transplanted to soil.

68

Planted bioreactor treatments

Plants were placed in contaminated soil using aseptic technique in a laminar flow
hood. For the unplanted treatments, ten milliliter test tubes were filled with 10 grams of
sterile, contaminated soil. About six grams of soil were poured into the test tubes to be
used for the planted treatments, then the plant was placed in the tube and the remainder
(about four grams) of the soil was poured over the roots. Two milliliters of one-half
strength MS (Murashige and Skoog) media were added to each test tube. The test tubes
were placed in sterile, coupled, sand-filled Magenta boxes (Figure 4-1). Four test tubes,
each with one plant were placed in the coupled Magenta GA7 boxes. Bioreactor
assemblies were placed in a plant growth chamber set at a 16 hour to 8 hour light to dark
cycle and a temperature of 28°C. Sterile water was added to the sand to maintain high
humidity within the Magenta box. Plants were watered as needed throughout the course
of the experiment. Initially, two milliliters of water was added to each tube. Planted
treatments were watered again approximately two weeks later as the transpiration by the
plant became significant. Thereafter, plants were watered with 1/2X MS as needed.
Each watering event proceeded in a laminar flow hood using sterile technique. A needle,
attached to the syringe, was used to puncture a septum on top of each coupled box,
allowing for each planted replicate to be watered. The humidity in the Magenta box

served to slow the water loss from the soil environment by transpiration.

Sampling
The Magenta boxes were harvested at five different time points in a laminar flow
hood. The plant was removed from the soil using forceps and a metal spatula. The roots

were cut from the shoots and washed in purified water. Soil from the test tube was

69

subdivided into samples for solvent extraction, sequential water extraction, moisture
content, and sterility. Solvent extractions were performed placing approximately one
gram of soil in a glass 4 ml vial with a PTF E (Teflon) liner. The vials were filled to the
top (no head space) with methanol. The contents of the vials were mixed via rotation at
12 rpm for 3 days. After mixing, the vials were centrifuged for 5 minutes using a
benchtop centrifuge, and the supernatant was sampled for subsequent analysis.
Sequential water extractions were performed by placing approximately 1.5 grams of soil
in a 4 ml glass vial with a PTFE liner. The vials were filled to the top with 1/2X MS
(Murashige and Skoog media). The contents of the vials were mixed via rotation at 12
rpm for 3 days. The samples were centrifuged and the supernatant was collected for
subsequent analysis. Fresh 1/2X MS was added to the vial with the previously water
extracted soil, 8 process that is reiterated until naphthalene is not detected in the
supernatant. Both the solvent extraction and the water extraction samples were analyzed
by HPLC (10 um, C18 column) using UV and fluorescence detection. Water content was
measured gravimetrically by first placing one to ten grams of soil in an aluminum
weighing pan and then placing them in a convection oven overnight at 110°C. Sterility
was checked by placing approximately 0.1 gram of soil in a sterile Eppendorf tube and
adding one milliliter of cell extraction buffer (contains EGTA and Tween 20). The tubes
were placed on a laboratory shaker set at 120 rpm for ten minutes. One hundred 111 of the
slurry was sampled and spread on a Petri plate containing YEPG media (yeast extract,
polypeptone and glucose). If any colonies appeared, the plate was subjected to the indole
assay (a small amount of solid indole is added to the lid of the plate causing colonies that

express naphthalene dioxygenase to turn blue as indole is converted to indigo). Root

70

tissue was analyzed using the WinRhizo image analysis software. This software package
allows for measurement of root surface area, volume, and color. Subsequently, the roots

and shoots were dried and weighed to determine dry weight.

Mathematics

The mathematical model used to describe the data was formulated by performing
an overall mass balance on naphthalene, a mass balance on a nonequilibrium site fraction,
and a mass balance on an immobile-water site fraction. Figure 4-2 is a depiction of the
soil as described by the model equations formulated for this study and Figure 4-3 is a box
diagram of this model. The model equations for the unplanted treatments are presented

in equations 4-1 through 4-3.

 

 

 

 

 

 

a as’” 65;," 65"" ac 62C
VW CW + m, 9" + m, 8" + m, M + Vg —g— = og Vg ——£- Equation (41)
at a: at at a: 522
6S,’,’f,q m ,
at = —km_,q( neq — K d fnequ) Equatron (4-2)
asg’; .
at = 0 Equation (4-3)

VW is the volume of soil solution in the test tube and Cw is the concentration of

naphthalene in the soil solution. The mass of soil is given the symbol m, and the soil

71

itself is divided into three site fractions. 86?] is the concentration of naphthalene in the
soil solid that is connected and in equilibrium with the mobile soil solution, where
mobility is defined as contaminant that can be transported in the gas phase. Sc’g‘ is the
concentration of naphthalene in the soil solid that is in equilibrium with the immobile soil
solution (i.e. contaminant that cannot be transported in the gas phase). Snag is the
concentration of naphthalene that is not in equilibrium with the mobile soil solution. The

three soil concentrations sum to equal the total soil concentration. V,g is the volume of
soil gas and C g is the concentration of naphthalene that is present in the soil gas. Dg is the
diffusivity of naphthalene in air. The parameter km,q is the first-order desorption rate

coefficient for the nonequilibrium domain. The parameters fcq and fneq are the site

fractions for the mobile-equilibrium domain and the mobile-nonequilibrium, respectively.
The variable t is time and z is the vertical dimension.

Naphthalene is assumed to exit the soil by gaseous diffusion only. Transport of
naphthalene is dynamically limited by desorption from the nonequilibrium site fraction.
The extent of naphthalene removal from the soil is limited by the amount sorbed to soil
domains in equilibrium with immobile water. This fraction is not in contact with the
mobile soil air and therefore does not diffuse out of the soil. However, the immobile
domain is water extractable, whereas the nonequilibrium domain is not measurably

extracted with water. All of the soil domains are assumed to be extractable with
methanol. The model parameters (i.e. feq, fneq. and kncq) were estimated assuming that the

naphthalene loss from the unplanted soils was solely attributed to gaseous diffusion.

72

The model formulation for the planted treatments is modified to contain an added
term describing the uptake of contaminant by the transpiration stream. Figure 4-4 is a
depiction of the rhizosphere soil as described by the model equations formulated for this
study and Figure 4-5 is a box diagram of this model. The model is presented in equations

4-4 through 4-6.

 

 

 

 

85'" 65,7; asgy’ 8C 62C
VwaCW+ms “’+ms———"+ms '+Vg—£=Dng—£—kawa
at a: at at at 622
Equation (4-4)
as'" m _
j?- = —km,q( ”a, — K d fneqC w) Equatron (4-5)
85;}; k pr, . .
= _ _ St”: Equatron (4-6)
at ms Kd '1' VW

An additional first-order mechanism is used to describe the effect that plant roots have on

the mass of naphthalene in the soil. The additional first-order mechanism is
parameterized by kp, the first-order naphthalene removal coefficient.

The added mechanism is assumed to act on both equilibrium domains, as large
roots grow in the soil macropores (assumed to comprise the mobile region) and root hairs
are likely able to access water in the soil micropores (i.e. the immobile region). Plants

are not assumed to enhance the rate at which contaminant desorbs from the

nonequilibrium domain. The parameter kp is a lumped parameter with contributions from

73

the plant root density, the plant transpiration stream concentration factor and the average
daily water uptake rate.

Parameter estimation was accomplished using the Gauss-Newton method for
coupled-partial differential equations. The state variables that were fit are the methanol
extractable naphthalene mass and the sequential water extractable naphthalene mass. The
adjustable parameters are: feq, fncq, and kneq for the unplanted treatments, and kp for the
planted treatments. The accepted tolerance for the convergence criterion was assumed to
be equal to 10". All model equations and the partial differential equations for the
sensitivity coefficients were solved using the finite element method. After the equations

were cast in semi-discrete form, a finite difference time-stepping scheme was used to

generate model profiles for each state variable.

74

Results and Discussion

Description of data

The data is plotted along with the model fits in Figure 4-6. For the unplanted
treatments, the data collected from the methanol extractions and the sequential water
extractions are observed to generate two distinct profiles. The profiles for the unplanted
treatments appear to change rapidly at short times, before approaching a limiting value at
long times. At short to intermediate times, a significant deviation exists between the
methanol and water extraction data. A fraction of material that was methanol extractable
was not water extractable. At later times, both the methanol and water extraction data
sets appear to converge to nearly the same value. This mass is larger than zero, and does
not appear to approach zero at an appreciable rate.

Similar profiles exist for the planted treatments, though the final converged value
is lower than the unplanted treatments. Plant growth, as shown in plots of the root
surface area in Figures 4-7 and 4-8, correlates with the enhanced removal of naphthalene

from the test soils.

Discussion of the data collected for the unplanted treatments

As seen in the plots of the unplanted treatments present in Figure 4-6, both the
methanol and water extraction data can be divided into two distinct regimes—each
regime characterized by a different rate of naphthalene loss. In both data sets, the first
regime is characterized by a much faster loss rate than is the loss rate in the second
regime. Possible mechanisms for the loss of naphthalene from the unplanted test tube
soils are: volatilization[32, 33], biodegradation[34-3 7] and binding to non-extractable

domains[3 8-41]. Because of the rate of loss in the first regime, volatilization is the

75

suspected mechanism. The loss of naphthalene during the first 20 days, as seen in Figure
4-6, occurs at a rate that can be parameterized by the gaseous diffusivity for naphthalene
in the soil atmosphere, a value traditionally estimated using the Millington-Quirk
formulation[42]. This suggests that naphthalene loss can be attributed to volatilization,
not biodegradation or binding to non-extractable domains of soil. Furthermore, y-
irradiation was used to sterilize all soils—for the purpose of minimizing the
biodegradation potential in the soil. To determine the extent of any microbial
contamination, all soils were plated at each sampling event, as the release of microbial
endophytes into the soil was anticipated[43-46]. Each plate was also subjected to the
indole assay to test for naphthalene dioxygenase activity. Direct plate counts for the
unplanted treatments always resulted in less than 103 CF U per gram of soil. The absence
of indole positive colonies—an indole positive colony is a colony that is actively
converting indole to indigo——is consistent with no expression of the naphthalene
dioxygenase enzyme. Though bacteria are present, they are not actively expressing this
enzyme. Naphthalene may also bind to soil constituents to form non-extractable
residues[47]. The formation of these residues was not observed in previous studies using
either of these test soils[48]. From a toxicological perspective, naphthalene that forms a
non-extractable residue may not be a risk to human health[11].

The two regimes that Figure 4-6 can be divided to are characterized by a rapid
initial loss of naphthalene that is followed by a much slower rate of loss—an observation
supported by published literature[33, 49]. The decreased loss rate in the second regime
can be characterized by slow mass transfer from the soil matrix[50-55]. In a previous

study, both soil types have demonstrated significant mass transfer limitations in batch soil

76

studies[28, 29, 48, 56], so biphasic behavior was expected in this study. A difference
exists between the methanol extractable and water extractable component of the sorbed
naphthalene. This difference is likely due to naphthalene sorbed to a hydrophobic
component of the soil in which desorption to the soil solution is kinetically limited[55,
57-61]. This regime of the soil is responsible for the slow loss of naphthalene observed
in the methanol extraction profile. Though desorption from this regime can be witnessed
over the experimental time-course of 186 days, significant desorption was not witnessed
over the water extraction time-course of 3 days. However, much of the naphthalene
remaining in the soil after 186 days remains water extractable, as seen in Figure 4-6.
Several explanations exist for the water extractability of naphthalene at long times. The
simplest explanation is that naphthalene is released from aggregate micropores after the
abrasion of the aggregate that occurs during successive water extractions. Naphthalene,
residing in domains that are not in equilibrium with the soil gas, is released upon
breaking the aggregate by mixing, but this naphthalene is not released during the course
of the experiment. Further possibilities for this behavior exist, as perhaps naphthalene is
retained in micropores that have been blocked by precipitated inorganic materials[62].
This entrapment of naphthalene is expected to hinder volatilization, but the naphthalene
would likely be water extractable as the precipitate dissolves. Another explanation for
restricted naphthalene volatilization is the slow desorption through water that is arranged
in a highly organized. Water has been found to behave in this manner in the soil
micrOpores that contain hydrophobic regions[63]. Regardless of the manner in which
naphthalene is retained in micropores, contaminant in these domains of soil may not be in

ecluilibrium with the soil gas, i.e. a fraction of the total naphthalene is not governed by

77

Henry’s law. Again, regardless of the retaining mechanism, the explanation for the
significant pool of water extractable naphthalene resides in the manner in which the water
extractions were conducted—namely by mixing the soil slurry and sequentially decanting
and re-adding fresh solution. Mixing of the slurry may have resulted in a disaggregation
of the soil structure, a dissolution of precipitated material, or a disruption of the
micropore solution, all of which lead to naphthalene desorption into the bulk solution.
Transfer from this region under the experimental conditions is likely unmeasurable when
compared to transfer from the other limiting regimes present in the soil. Though transfer
from this regime is too slow to measure in the test tubes soils, the material contained in

this regime remains water extractable.

Discussion of data collected for planted treatments
Both planted soils contained decreased concentrations at long times when
compared to the unplanted treatments, as seen by comparing the treatments in Figure 4-6.
Clearly, the addition of plants facilitates an increase in naphthalene loss from the test tube
soil. In addition to a removal mechanism attributed to the plant, the mechanisms
responsible for naphthalene removal in unplanted soils are likely occurring.
Volatilization remains a mechanism for contaminant removal throughout the run time for
planted and unplanted treatments alike. Transpiration is likely responsible for further
naphthalene removal[64]. Mechanistically, naphthalene in the soil solution can be taken
into the plant through the transpiration stream and removed from the soil[64]. A
decreased water extractable mass, when compared to the unplanted treatments, suggests
that the transpiration stream is accessing naphthalene associated with the soil micropores.

Naphthalene residing in both macro- and micropores can enter the transpiration stream as

78

plants are capable of transpiring water from both macro- and micropores non-
preferentially[65]. Removal of water from these regimes by the roots can also promote
increased volatilization as micropore water is no longer restricting the access of soil gas.
The combined effect of increased volatilization and transpiration amounts to the
increased loss observed in the planted treatments. Thus plants are likely able to enhance

desorption from regions which are not in equilibrium with the soil gas, and in so doing,

increase the rate of mass transfer from the soil.

Model formulation for the unplanted treatments

Examination of the collected data with an equilibrium model fails to predict the
deviation that exists between the methanol extraction and water extraction data sets, as
shown in Figure 4-9. Though the initial loss rate of naphthalene is consistent with the
volatilization component of an equilibrium model—i.e. a model that uses Henry’s
partitioning and Fickian diffusion in the gas phase—the intermediate and late rates of loss
are not predictable by an equilibrium formulation alone. Also, the amount of material
that is extractable by sequential water extraction is not accurately predicted. A deviation
exists between the methanol and water extraction data sets. The magnitude of this
deviation can be described by a mass transfer expression. This mechanism states that a
fraction of contaminant that is available for volatilization, albeit slowly, is not
appreciably water extractable. Essentially, there is a longer time period for contaminant

loss by volatilization than is the time period used for the sequential water extractions.
The addition of a mass transfer mechanism accounts for the difference observed between

the methanol and water extractable fractions as depicted in Figure 4-10. Though the

79

methanol extraction data is adequately modeled, the water extraction data significantly
deviates from the model fit at intermediate and long times. An additional mechanism is
needed to account for naphthalene that is resistant to volatilization, though water
extractable—as the methanol and water extraction data sets converge to nearly the same
value at the 186 days. Water in immobile regions of soil, such as the aggregate
micropores, may retain naphthalene that is not in equilibrium with the soil gas. This
material is considered to be water extractable throughout the experimental time course,
but not volatile. The best-fit lines, presented with the data in Figures 4-6 and 4-9, are
generated using the three-regime model. This model was formulated using an
equilibrium regime, a mass transfer limited regime, and an immobile regime. This model
accurately fits the acquired data as the coefficient of determination is 87 and 88 percent

for the SpAf and KalA soils respectively.

Model formulation for the planted soil

Plants were observed to decrease the amount of naphthalene that remained in the
soil at the end of the experiment. The loss of naphthalene from the planted treatments is
consistent with the removal of microporous water via transpiration, a phenomena that has
been demonstrated previously[65]. An additional mechanism is needed to account for
this enhanced removal, and was done so assuming first-order kinetics. The model fits
presented in Figure 4-6 include best-fit lines that were generated using the three regime
model with the addition of a first-order removal mechanism attributed to the plant. The

values of the coefficients of determination range from 90 to 92 percent, and thus the data

80

are accurately described. Material that is associated with immobile domains, and not in

equilibrium with the soil gas, may be removed by phytoremediation.

Discussion of parameters
Table 4-2 includes the parameters that were fit for both unplanted treatments

using the Gauss-Newton method. The equilibrium regime was larger in the SpAf soil
than in the Kal A soil as the value for fc'g is larger for SpAf than for the Kal A soil. The

nonequilibrium regimes were of comparable size in both treatments, while the size of the
immobile regime was larger in the Kal A soil. This implies that a greater amount of

naphthalene will be present in the micropores of Kal A soil than in SpAf soil. The

desorption rate coefficient, kncq": is larger in the Kal A soil than in the SpAf soil, though

this difference is not significant. Interestingly, the value of kp does not depend upon the

species of plant as much as the soil type.

Conclusion

These results are significant as evidence that plants are able to enhance the
removal of contaminants from soil by transpiring water containing naphthalene in
micropores. The conceptual model that results from the analysis of unplanted soil is
presented in Figure 4-3. Mass transfer of naphthalene in soil is thought to occur by a
number of mechanisms, including: 1) instantaneous mass transfer due to equilibrium
partitioning between the soil solid and the soil solution, 2) rate limited transfer from the
soil solid to the soil solution, and 3) Henry’s partitioning from the soil solution into the

soil gas followed by gaseous diffusion through the air-filled void spaces. Naphthalene

81

residing in the micropores of the soil aggregates is not included as a transfer process, as
this material is not in equilibrium with the soil gas. Conversely, the depiction of the
planted system, as shown in Figure 4-5, depicts the ability of the plant to access
naphthalene in aggregate micropores. The authors speculate that this removal is due to
converting the immobile water in the aggregate pores to mobile (i.e. advective) water.
This process would result in transporting the naphthalene to the macroporous soil
solution, where the processes of volatilization or plant uptake could remove the
naphthalene from the test soil. The conceptual model, formulated by interpreting the data
gathered from methanol and sequential water extractions, justifies the use of plants for
removing contaminants that may reside in microporous regimes of soil. Finally, the
results of this study further advocate the selection of plants as a remediation technology,

as the extent and rate of removal are enhanced in planted systems.

82

Rubber

 

 

 

 

septa Magenta GA-7
box
Plant
14—- Box coupler
Test tubes ”7‘ —
containing

 

 

 

soil 1 }

 

 

 

 

U Support sand

 

Figure 4-1. The planted bioreactor.

83

l

Gaseous
diffusion

OD
Instantaneous
mass transfer

    

Instantaneous
mass transfer

Rate-limited
mass transfer

Figure 4-2. The conceptual description of naphthalene transport in unplanted soil
that was developed upon interpretation of the methanol and water extraction data.

84

 

Soil gas (contaminants mobile by diffusion)

 

+1

Micropore soil gas
(contaminants
immobile)

 

I .1

Soil solution

41 1

ii
Micropore soil
solution

4|

 

N l

Equilibrium soil Nonequilibrium soil
domain domain

 

 

 

N

Equilibrium soil-
micropore domain

 

Figure 4-3. Box model for the unplanted treatments.

85

 

i

Gaseous
diffusion

    

00
Root Instantaneous
S mass transfer

 
   

Instantaneous
mass transfer

 
   
     
 

Rate-limited
mass transfer

 
 

Figure 4-4. The conceptual description of naphthalene fate and transport in planted
soil that was developed upon interpretation of the methanol and water extraction

data.

86

 

Soil gas (contaminants mobile by diffusion) Micropore soil gas
(contaminants
‘ immobile)

 

 

4 l A I
Soil solution Micropore soil
( '— — solution — -—>

41 t M
N | N

Equilibrium soil Nonequilibrium soil Equilibrium soil-
domain domain micropore domain

 

 

 

 

 

 

Figure 4-5. Box model for the planted treatments. The dashed arrows represent
naphthalene transport that is attributed to the transpiration stream.

87

SpAf Kal A

unplanted

 

 

A. gerardii

 

 

M. alba

 

 

 

Time (days) Time (days)

Figure 4-6. Plots of methanol extraction data, water extraction data, and
models fits for the six treatments. The Y-axis is the naphthalene mass in pg.
The diamonds are the methanol extraction data points and the circles are
the water extraction data points. The solid line is the methanol extraction
best-fit line and the dashed line is the water extraction best-fit line.

88

w + A. gerardii

at" 50 ,—->K--M.alba
E

.9.

m 40

2

I!

8 30

i

m 20 - x’
*6

o

t: 10 *

 

 

 

   

0 50 100 150 200
Time (days)

Figure 4-7. Root surface area profile for A. gerardii and M. alba in SpAf soil.

89

+ A. gerardii
. --x-— M. alba

0'1
0

.h
C

 

 

Root surface area (cmz)
0)
O

 

20
10
O .2
0 50 100 150 200
Time (days)

Figure 4-8. Root surface area profile for A. gerardii and M. alba in Kal A soil.

90

5000 -

4500

4000

l

3500 t

I

3000

2500

2000

Ur fit

Naphthalene mass (ug)

—I
U"
0
o
I

1000 r

8

500 1.

 

l i;

60 80 100 120 140 160 180 200
Time (days)

 

Figure 4-9. Plot of methanol extractable naphthalene mass and water extractable
naphthalene mass versus time for Kal A soil that is unplanted. The diamonds (O)
are the methanol extraction data and the circles (O) are the water extraction data.
Standard deviation error bars are included on each point. The solid line is the
model fit of the methanol and water extraction data.

91

(’1
O
O
O
—11

A
OI
O
O
r

N w w
01 o 01
o o o
o o g

N
O
O

Naphthalene mass (ug)

1500
1
10001

500 '

 

 

0 20 4O 60 80 100 120 140 160 180 200
Time (days)

Figure 4-10. Plot of methanol extractable naphthalene mass and water extractable
naphthalene mass versus time for Kal A soil that is unplanted. The diamonds (O)
are the methanol extraction data and the circles (C) are the water extraction data.
Standard deviation error bars are included on each point. The solid line is the
model fit of the methanol extraction data and the dashed line is the model fit of the
water extraction data.

92

Table 4-1. Soil parameters for Spinks A and Kalkaska A soils

 

 

 

 

Soil type 11 0w Pb Kd KH Dg
SpAf 0.5 0.15 1.25 5.0 1.74~ 10'2 22.4
Kal A 0.5 0.15 1.54 13.64 1.74-10‘2 185

 

 

 

 

 

 

 

93

 

Table 4-2. Estimated parameters for planted and unplanted soils.

 

 

 

 

Unplanted Unplanted Unplanted A. gerardii M alba

Soil type 6:; fat: kne'z,‘ 0n") kp an") kp <hr")
SpAf 0.357:1:0.009 0.583zt0.009 000101200003 0.142i0.010 O.156:i:0.011
Kal A 0.272zt0.004 0.564:t0.008 0.0013i0.0002 0.072zl:0.024 0.056i0.022

 

 

 

 

 

 

94

 

References

10.

Mougin, C., Bioremediation and phytoremediation of industrial PA H—polluted
soils. Polycyclic Aromatic Compounds, 2002. 22: p. 1011-1043.

Schnoor, J.L., et al., Phytoremediation of organic and nutrient contaminants.
Environmental Science and Technology, 1995. 29(7): p. 318A-323A.

Reilley, K.A., M.K. Banks, and AP. Schwab, Dissipation of polycyclic aromatic
hydrocarbons in the rhizosphere. Journal of Environmental Quality, 1996. 25: p.
212-219.

Burken, J .G. and J .L. Schnoor, Phytoremediation: plant uptake of atrazine and
role of root exudates. Journal of Environmental Engineering, 1996. 122(11): p.
95 8-963.

McCutcheon, SC. and J .L. Schnoor, Overview of phytotransformation and
control of wastes, in Phytoremediation: Transformation and Control of

Contaminants, S.C. McCutcheon and J .L. Schnoor, Editors. 2003, John Wiley and
Sons, Inc.: Hoboken, New Jersey. p. 3-58.

Cunningham, SD. and W.R. Berti, Remediation of contaminated soils with green
plants: an overview. In Vitro Cellular and Developmental Biology, 1993. 29: p.
207-212.

Cunningham, S.D., W.R. Berti, and J .W. Huang, Phytoremediation of
contaminated soils. Trends in Biotechnology, 1995. 13: p. 393-397.

Davis, L.C., et al., Modeling the effects of plants on the bioremediation of
contaminated soil and ground water. Environmetanl Progress, 1993. 12(1): p. 67-
75.

Siciliano, SD. and J .J . Germida, Mechanisms of phytoremediation: biochemical

and ecological interactions between plants and bacteria. Environmental Reviews,
1998. 6: p. 65-79.

Shimp, J .F ., et al., Beneficial eflects of plants in the remediation of soil and
groundwater contaminanted with organic materials. Critical Reviews of
Environmental Science and Technology, 1993. 23(1): p. 41-77.

95

ll.

12.

13.

14.

15.

16.

17.

18.

19.

20.

Alexander, M., Aging, bioavailability, and overestimation of risk fiom
environmental pollutants. Environmental Science and Technology, 2000. 34(20):
p. 4259-4265.

Bosma, T.N.P., et al., Mass transfer limitation of biotransformation: quantifizing
bioavailability. Environmental Science and Technology, 1997. 31(1): p. 248-252.

Chung, N. and M. Alexander, Diflerences in sequestration and bioavailability of

organic compounds aged in dissimilar soils. Environmental Science and
Technology, 1998. 32(7): p. 855-860.

Cuypers, C., et al., The estimation of PAH bioavailability in contaminated
sediments using hydroxypropyl -B-cyclodextrin and Triton X-100 extration
techniques. Chemosphere, 2002. 46: p. 1235-1245.

F eng, Y., et al., Bioavailability of soil-sorbed biphenyl to bacteria. Environmental
Science Technology, 2000. 34: p. 1977-1984.

Park, J .-H., Bioavailability of sorbed organic contaminants, in Department of
Civil and Environmental Engineering. 2000, Michigan State University.

Zhang, W., E. Bouwer, and W. Ball, Bioavailability of hydrophobic organic
contaminants: eflects and implications if sorption-related mass transfer on
bioremediation. Ground Water Monitoring and Remediation, 1998: p. 126-138.

Aprill, W. and RC. Sims, Evaluation of the use of prairie grasses for stimulating
polycyclic aromatic hydrocarbon treatment in soil. Chemosphere, 1990. 20(1-2):
p. 253-265.

Qui, X., et al., Grass-enhanced bioremediation for clay soils contaminated with
polynuclear aromatic hydrocarbons, in Bioremediation Through Rhizosphere
Technology, T.A. Anderson and J .R. Coates, Editors. 1994, American Chemical
Society: Washington, DC. p. 142-157.

Pradhan, S.P., et al., Potential of phytoremediation for treatment of PA Hs in soil
at MGP sites. Journal of Soil Contaminantion, 1998. 7: p. 467-480.

96

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

Banks, M.K., E. Lee, and AP. Schwab, Evaluation of dissipation mechanisms for
benzo a pyrene in the rhizosphere of tall fescue. Journal of Environmental Quality,
1999. 28(1): p. 294-298.

Chen, Y., M. Banks, and A. Schwab, Pyrene degradation in the rhizosphere of
tall fescue (F estuca arundinacea) and switchgrass (Panicum virgatum L. ).
Environmental Science & Technology, 2003. 37 : p. 5778-5782.

Liste, H.-H. and M. Alexander, Accumulation of phenanthrene and pyrene in
rhizosphere soil. Chemosphere, 2000. 40: p. 11-14.

Muratova, A., et al., Plant-rhizosphere-microflora association during
phytoremediation of PAH-contaminated soil. International Journal of
Phytoremediation, 2003. 5(2): p. 137-151.

Schwab, AP. and M.K. Banks, Biologically mediated dissipation of polyaromatic
hydrocarbons in the root zone, in Bioremediation Through Rhizosphere
Technology, T.A. Anderson and J .R. Coates, Editors. 1994, American Chemical
Society: Washington, DC. p. 132-141.

White, J .C. and M. Alexander, Reduced biodegradability of desorption-resistant
fractions of polycyclic aromatic hydrocarbons in soil and aquifer solids.
Environmental Toxicology and Chemistry, 1996. 15(11): p. 1973-1978.

Guerin, W.F. and SA. Boyd, Bioavailability of naphthalene associated with
natural and synthetic sorbents. Water Resources, 1997. 31(6): p. 1504-1512.

Park, J .-H., X. Zhao, and T.C. Voice, Development of a kinetic basis for
bioavailability of sorbed naphthalene in soil slurries. Water Research, 2002. 36: p.
1 620- 1 628.

Park, J .-H., X. Zhao, and T.C. Voice, Biodegradation of non-desorbable
naphthalene in soils. Environmental Science and Technology, 2001. 35(13): p.
2734-2740.

Park, J .-H., et al., Assessment of bioavailability of soil-sorbed atrazine. Applied
and Environmental Microbiology, 2003. 69(6): p. 3288-3298.

97

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

Olson, P.E., K.F. Reardon, and E.A.H. Pilon-Smits, Ecology of rhizosphere
bioremediation, in Phytoremediation: Transformation and Control of
Contaminants, S.C. McCutcheon and J .L. Schnoor, Editors. 2003, John Wiley and
Sons, Inc.: Hoboken, New Jersey. p. 317-353.

Park, KS, et al., Fate of PAH compounds in two soil types: influence of

volatilization, abiotic loss and biological activity. Environmental Toxicology and
Chemistry, 1990. 9: p. 187-195.

Hickey, W.J. and J. Paek, Ventilation eflects on mineralization and volatilization
of naphthalene in a gasoline-contaminanted subsurface soil. Chemosphere, 1996.

32(8): p. 1655-1667.

Ahn, I.-S., L.W. Lion, and ML. Shuler, Microscale-based modeling of
polynuclear aromatic hydrocarbon transport and biodegradation in soil.
Biotechnology and Bioengineering, 1996. 51: p. 1-14.

Gunther, T., U. Domberger, and W. Fritsche, Effects of ryegrass on
biodegradation of hydrocarbons in soil. Chemosphere, 1996. 33(2): p. 203-215.

Heitkamp, M.A., J .P. Freeman, and CE. Cemiglia, Naphthalene biodegradation
in environmental microcosms: estimates of degradation rates and
characterization of metabolites. Applied and Environmental Microbiology, 1987.
53(1): p. 126-136.

Yolcubal, I., et al., Biodegradation during contaminant transport inporous media:
v. teh influence of growth and cell elution of microbial distribution. Journal of
Environmental Quality, 2002. 31: p. 1824-1830.

Hatzinger, P. and M. Alexander, Effect of aging of chemicals in soil on their
biodegadability and extractability. Environmental Science and Technology, 1995.
29(2): p. 537-545.

Kilbane II, J ., Extractability and subsequent biodegradation of PAHs from
contaminated soil. Water, Air, and Soil Pollution, 1998. 104: p. 285-304.

Northcott, G.L. and KC. Jones, Partitioning, extractability, and formation of
nonextractable PAH residues in soil. I . Compound diflerences in aging and
sequestration. Environmental Science and Technology, 2001. 35(6): p. 1103-1110.

98

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

Northcott, G.L. and KC. Jones, Partitioning, extractability, and formation of
nonextractable PAH residues in soil. 2. Eflects on compound dissolution behavior.
Environmental Science and Technology, 2001. 35(6): p. 1111-1117.

Millington, R. and J. Quirk, Permeability of porous solids. Transactions of the
Faraday Society, 1961.57: p. 1200-1207.

Hallmann, J ., et al., Bacterial endophytes in agriculutural crops. Canadian
Journal of Microbiology, 1997. 43: p. 895-914.

Lamb, T.G., D.W. Tonkyn, and DA. Kluepfel, Movement of Pseudomonas
aureofaciensfiom the rhizosphere to aerial plant tissue. Canadian Journal of
Microbiology, 1996. 42: p. 1112-1120.

Bell, C.R., et al., Endophytic bacteria in grapevine. Canadian Journal of
Microbiology, 1995. 41: p. 46-53.

Zinniel, D.K., et al., Isolation and characterization of endophytic colonizing

bacteria from agronomic crops and prairie plants. Applied and Environmental
Microbiology, 2002. 68(5): p. 2198-2208.

Macleod, C. and K. Semple, Sequential extraction of low concetrations of pyrene
and formation of non-extractable residues in sterile and non-sterile soils. Soil
Biology & Biochemistry, 2003. 35: p. 1443-1450.

Park, J .-H., X. Zhao, and T. Voice, Biodegradation of non-desorbable
naphthalene in soils. Environmental Science and Technology, 2001. 35(13): p.
2734-2740.

Lindhardt, B. and TH. Christensen, Measured and estimated volatilisation of
naphthalene from a sandy soil. Chemosphere, 1994. 29(7): p. 1407-1419.

Brusseau, M.L., R.E. Jessup, and P.S.C. Rao, Modeling the transport of solutes
influenced by multiprocess nonequilibrium. Water Resources Research, 1989.
25(9): p. 1971-1988.

Brusseau, M.L., R.E. Jessup, and P.S.C. Rao, Nonequilibrium sorption of organic

chemicals: elucidation of rate-limiting processes. Environmental Science and
Technolog)’, 1991. 25(1): p. 134-142.

99

52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

Brusseau, ML. and P.S.C. Rao, Influence of sorbate structure on nonequilibrium
sorption of organic compounds. Environmental Science and Technology, 1991.
25(8): p. 1501-1506.

Hwang, S. and T. Cutright, Evidence of underestimation in PAH
sorption/desorption due to system nonequilibrium and interaction with soil
constituents. Journal of environmental Science and Health, 2004. 39: p. 1147-
1162.

Rahman, M., F. Amiri, and E. Worch, Application of the mass transfer model for
describing nonequilibrium transport of HOCs through natural geosorbents.
Water Research, 2003. 37: p. 4673-4684.

Rao, P.S.C., J .M. Davidson, and HM. Selim, Evaluation of conceptual models for
describing nonequilibrium adsorption-desorption of pesticides during steady-flow
in soils. Soil Science Society of Americal Journal, 1979. 43: p. 22-28.

Park, J .-H., X. Zhao, and T.C. Voice, Comparison of biodegradation kinetic
parameters for naphthalene in batch and sand column systems by Pseudomonas
putida. Environmental Progress, 2001. 20(2): p. 93-102.

Comelissen, G., P.C.M. van Noort, and H.A.J. Govers, Mechanism of slow
desorption of organic compounds fi'om sediments: A study using model sorbents.
Environmental Science and Technology, 1998. 32: p. 3125-3131.

Comelissen, G., et al., Temperature dependence of slow adsorption and
desorption kinetics of organic compounds in sediments. Environmental Science
and Technology, 1997. 31: p. 454-460.

Farrell, J ., D. Grassian, and M. Jones, Investigation of mechanisms contributing to
slow desorption of hydrophobic organic compounds from mineral solids.
Environmental Science and Technology, 1999. 33: p. 1237-1243.

Reeves, W., et al., Partitioning ans desorption behavior of polycyclic aromatic
hydocarcon from disparate aources. Science of the Total Environment, 2004.

Sharer, M., et al., Aging effects on the sorption-desorption characteristics of

anthropogenic organic compounds in soil. Journal of Environmental Quality,
2003. 32: p. 1385-1392.

100

62.

63.

64.

65.

Steinberg, S., J .J . Pignatello, and B. Sawhney, Persistence of 1,2-Dibromoethane
in soils: entrapment in intraparticle micropores. Environmental Science and
Technology, 1997. 31: p. 1201-1208.

Li, J. and OJ. Werth, Slow desorption mechanisms of volatile organic chemical
mixtures in soil and sediment micropores. Environmental Science and
Technology, 2004. 38: p. 440-448.

Briggs, G.G., R.H. Bromilow, and A.A. Evans, Relationships between
Iipophilicity and root uptake and translocation of non-ionised chemicals in barley.
Pesticide Science, 1982. 13: p. 495-504.

Emerman, SH. and TE. Dawson, Experiments using split-root chambers on
water uptake from soil macropores by sunflowers. Plant and Soil, 1997. 189: p.
57-63.

101

Chapter 5. Using an effectiveness factor to improve decision making regarding
phytoremediation

Abstract

The development of an effectiveness factor for improving decision making in
phytoremediation is addressed. This effectiveness factor compares the remediation rates
in planted and unplanted systems. A plot of the effectiveness factor versus the Thiele
modulus (i.e. a dimensionless ratio of reactive processes to mass transfer processes) was
divided into three regimes. These regimes reveal the efficacy of phytoremediation and
the mechanisms that limit the rate of remediation in planted systems. At low values of
the Thiele modulus, phytoremediation does not significantly enhance the remediation
rate. At moderate values, phytoremediation improves the remediation rate, but is limited
by the kinetics of plant uptake or plant facilitated degradation. At large values of the
Thiele modulus, the use of plants remains beneficial, though improving the remediation
rate relies on enhancing the contaminant desorption rate. Systems that are completely
desorption limited benefit less from being planted, as desorption is regarded as a
necessary first step in a contaminant’s remediation. To overcome this limitation, further
exploration is needed regarding plant species that are capable of enhancing the
contaminant desorption rate or enhancing contaminant bioavailability. Desorption
limitations impact the decision making process by directing the site engineer to

technologies that are capable of overcoming these limitations.

102

Introduction

This chapter presents a method to determine whether phytoremediation provides
enhanced contaminant removal when compared to other remediation strategies. The
result of this comparison is improved decision making regarding the use of plants to
enhance contaminant removal. Increased use of phytoremediation in the field has made
the pursuit of improved decision making important. The decision to use
phytoremediation depends upon the cost, extent, and rate of removal using plants versus
the cost, extent, and rate of removal using other technologies. Cost, extent, and rate of
removal are at least partially dependent upon the physicochemical properties of the
contaminant and soil, and the biological properties of the selected technology (i.e.
phytoremediation, bioaugmentation, natural attenuation, etc.). These properties are
measured, either in vitro or in situ, and used to formulate mechanisms, either descriptive
(empirical) or explanatory (theoretical), that can be used to construct mathematical
models. These mathematical models are typically used in a predictive mode after the
accuracy and uncertainty of the models have been assessed by comparing experimental
data with model simulations. Once the robustness of a model has been accepted, a
method is needed for interpreting the model output to make decisions regarding the use of
a clean-up technology (e. g. phytoremediation). The method presented here is borrowed
from the discipline of heterogeneous catalysis. In heterogeneous catalysis—particularly
catalysis involving diffusion and reaction inside catalyst particles—chemical reactants
diffuse through the catalyst pore structure, adsorb to the catalyst surface, react, and the
resulting products desorb and subsequently diffuse through the pore structure and into the

surrounding fluid. Briefly, the mathematical model that describes/explains this process is

103

appropriately nondimensionalized to form a dimensionless group named the Thiele
modulus. A “rating factor”, or effectiveness factor, is defined to compare the rate of
reaction with pore diffusion resistance to the rate of reaction at the surface conditions[1].
By plotting the effectiveness factor versus the Thiele modulus, a profile is generated that
can be divided into two regimes: l) a reaction rate limited regime at low values of the
Thiele modulus, and 2) a mass transfer limited regime at high values of the Thiele
modulus. Improving the overall reaction rate by catalyst selection is predicated on which
regime characterizes the catalyzed reaction. By analogy, decision making in
phytoremediation can benefit from the same strategy used in heterogeneous catalysis.
What is needed is the selection of a suitable effectiveness factor, the division of limiting
behavior into discrete regimes, and a means of determining the regime in which a given
phytoremediation systems resides. The use or nonuse of phytoremediation can then be
determined. The aim of this paper is to use this methodology to improve decision making
regarding phytoremediation.

The mechanisms thought to be important for phytoremediation of PAHs include
plant uptake, sorption to roots, and degradation by rhizosphere microbes (depicted in
Figure 5-1). Plant uptake[2, 3] is more important for the lower molecular weight PAHs,
as these tend to be more soluble and less hydrophobic. The amount of contaminant
sorbed to roots depends on the relative hydrophobicities of the root lipid fraction and the
contaminant[4]. The promotion of degrading organisms in the rhizosphere by root
exudation or secretion is called phytostimulation. Phytostimulation[5, 6] is a proposed
mechanism for the enhanced remediation of PAHs—especially the higher molecular

weight PAHs.

104

 

 

As several mechanisms are responsible for a plant’s ability to remove or affect the
removal of contaminants, several limiting phenomena may ultimately be responsible for
slowing cleanup. Greatest among these is the contaminant’s desorption from the soil
matrix. The processes responsible for increased degradation by plants act after the
contaminant is thought to have desorbed. For example, microbial biodegradation has
been demonstrated to occur after the contaminant desorbs. Also, plant uptake of organic
contaminants occurs after the contaminant desorbs from the soil matrix. Volatilization
and leaching of the contaminant are also dependent upon prior desorption. Determining
the rate and extent of the mass transfer limitations is a necessary first step before
applying phytoremediation. If the contaminant is not limited by desorption, then
comparing the rates of the various processes remains an important matter. This
comparison may lead to the best method for enhancing the rate, whether degradation can
be enhanced by nutrient amendment, or whether a plant species that transpires more
water can be selected to hasten uptake. The formulation of a suitable effectiveness factor
would encompass the mechanisms present in planted and unplanted systems. Classical
plots of the effectiveness factor can be used to compare the rates of removal in planted
and unplanted treatments. Furthermore, these plots can be divided into regimes that
correspond to the various removal controlling mechanisms. Decision making is guided

by determining the regime a given planted system resides.

105

Mathematics

The model presented in Equations 5-1 through 5-3 was used to describe the
unplanted scenario for this investigation. A list of model variables and model parameters
is presented in Table 5-1. This model was formulated to describe the transport and

reactions of naphthalene in the Spinks A-horizon soil.

 

 

      

 

ac q BS __q as. 6C 62C .
L + m + __In+¢9 g :1) ,9 f E uatron 5-1
”’16 6: 10,7, phat “at ““62" q ()
aSm: °
797‘]— : _kmq(S Sneq— dequ CW) Equat10n(5‘2)
2%. z Equation (53)

A second model, presented as Equations 5-4 through 5-6, was formulated to describe the

action of the plant in a first-order kinetic manner.

 

 

 

 

5C BS, 65 , 6S. 5C 52 Cg
6 ———“-+ “’ + "”’ + ——’-’1+6 "' =D 0— -kp 6 C E uation 5-4
w at pl) at pb at pl) at g at g e g 622 q ( )
a’Sm'q -
at : _kneq(S Sneq— Kt! neq C» C) Equatlon (5-5)
. k 9 .
6S”, = ———£— Sm Equation (5-6)
at pbKd + 6,.

Both models divide the soil into three domains. The equilibrium domain is parameterized

by the soil-water partition coefficient, assuming that transfer from the soil to the water

106

occurs instantaneously. The first-order desorption rate coefficient parameterizes the rate
of transport from the mass transfer limited domain (a.k.a. the nonequilibrium domain) of
the soil. The models consider the volatilization of naphthalene to occur by Henry’s
partitioning into the soil gas, followed by gaseous diffusion out of the system boundary.
An immobile soil domain is constructed to represent material that does not volatilize.
The phytoremediation model, presented in Equations 5-4 through 5-6, considers that the
remediation of naphthalene is described by first-order kinetics. This description includes
phytoremediation in the immobile soil solution and the solution that is in equilibrium
with the soil gas. A box model depicting the transport and reactive processes included in
both models is presented in Figure 5-2.

The nondimensionalization of the unplanted model results in Equations 5-7

through 5-9.
’ dS', , .’ 2 ’
RI —a—C—‘i + R2 -——'11 + R3 % = R4Daa—C,i Equation (5-7)
or at at 6Z“
aShet ' ‘ -
——’— = —(Sm,q — C W) Equatron (5-8)
62‘
S ’
h = 0 Equation (5-9)
at

Likewise, the nondimensionalization of the phytoremediation model gives

Equations 5-10 through 5-12.

0

 

 

ac‘ 68' 6s, .. azc’ .
R——1+R ”“"+R ——1L""i=RDa ._ 2C E uation 5-10

107

 

_...._, = 45;“ _ q.) Equation (5-11)
at

6:," = 4,555" Equation (5-12)
2'

A list of dimensionless model variables and parameters is presented in Table 5-2.
Comparisons between planted and unplanted remediation were made using an
effectiveness factor. Historically, the effectiveness factor approach was used to describe
heterogeneous catalysis. This construct is defined as the ratio of the rate of reaction that
includes a diffusive resistance to the rate of reaction at the solid catalyst surface. An
effectiveness factor equal to one corresponds to a system that is limited by the reaction
rate. An effectiveness factor that is lower than one corresponds to a system that is limited
by mass transfer inside the catalyst pellet. Determining whether mass transfer or reaction
is limiting the rate of product formation defines how reaction can be increased. By
analogy, this effectiveness factor approach can be used to compare the rate of
contaminant remediation between planted and unplanted systems. The rate of
contaminant removal is compared by selecting an arbitrary growing season of 200 days.

Mathematically, this effectiveness factor is presented in Equation 5-13.

((mass of contaminant at t = 0)— (mass of contaminant at t = 200))plmd

— ((mass of contaminant at t = 0)— (mass of contaminant at t = 200))

 

unplantcd

Equation (5-13)

108

The effectiveness factor is typically plotted against a dimensionless number called the
Thiele modulus. In the field of heterogeneous catalysis, this modulus is the ratio of
reaction processes to mass transfer processes. The reaction processes are typically
parameterized by reaction rate coefficients and the mass transfer processes are typically
parameterized by diffusivities. The Thiele modulus, presented in Equation 5-14, is

formulated in an analogous manner for this study.

¢ = J— Equation (5-14)

In this description, the kinetic forces are parameterized by kp, which encompasses the
combined effect of transpirational uptake and phytostimulation. The mass transfer forces
are parameterized by kncq, which parameterizes the rate of contaminant desorption from
the soil matrix. Plots of the effectiveness factor versus the Thiele modulus can be divided
into discrete regimes of behavior. An interpretation of these regimes reveals whether
phytoremediation is an effective technology when compared to remediation in an

unplanted system.

109

Results and Discussion

Figure 5-3 contains a plot of the effectiveness factor versus the Thiele modulus
using a parameter set determined for naphthalene which is presented in Table 5-3. At
small values of the Thiele modulus, the effectiveness factor approaches a value of one.
In this situation, the magnitude of the plant’s effect on remediation is small, thus an
unplanted treatment remediates the system as well as the planted treatment. At
intermediate values of the Thiele modulus, the slope of the curve becomes positive. The
planted system enhances remediation in this transition regime. At relatively large values
of the Thiele modulus, the curve begins to approach a zero slope. This situation arises
when contaminant mass transfer is slow compared to the potential remediation rate of the
planted system. At large values of the Thiele modulus, there is no benefit in increasing
the transpirational flow, because slow mass transfer eventually limits remediation. The
slope of the curve approaches zero as rate-limited mass transfer controls the remediation
rate in planted and unplanted systems alike. Regime-one behavior corresponds to no
benefit to remediation by plants, a situation that occurs at low values of the Thiele
modulus. Regime two is characterized by plant enhanced remediation, though the
remediation rate is limited by some activity associated with the plant. Regime three also
displays enhanced remediation, though the remediation rate has become limited by
contaminant mass transfer from the soil. It is important to note that enhanced mass
transfer is not included in the model formulation (i.e. enhanced remediation is defined as

the increased remediation rate in planted systems when compared to unplanted systems).

110

Figure 5-4 is a plot showing the effect of increasing the Damkohler number on the
effectiveness factor. The parameter values used to plot Figures 5-3 through 5-7 are
contained in Table 5-4. Increasing the Damkohler number decreases the value of the
effectiveness factor. Mechanistically, increased Damkohler numbers are the results of
increasing the effective diffusivity relative to the desorption rate coefficient. Increasing
the effective soil gas diffusivity increases the rate of contaminant volatilization. Systems
with high volatilization rates are not likely to be improved by planting. However, soils
with significant tortuosities (e. g. soils that require the Millington-Quirk correction[7]) are
likely to be amenable to phytoremediation as the volatilization rate of semi-volatiles will
be reduced. Furthermore, increasing the effective diffusivity does alter the shape of the
profile. Specifically, the presence of the third regime suggests that rate-limited mass
transfer limits contaminant volatilization. Rate-limited transfer, present in the third
regime, is responsible for limiting both volatilization and phytoremediation.

Figure 5-5 is a plot showing the effect of increasing the value of K, on the
effectiveness factor. This is equivalent to selecting another contaminant with a larger
octanol-water coefficient or selecting a soil with a larger amount of organic matter.
Increasing Kd, results in increasing the effectiveness factor, as seen from Figure 5-5. An
increased effectiveness factor corresponds to a beneficial circumstance in planted
systems. Three-ring PAHs, such as phenanthrene and anthracene, are more likely to
benefit from phytoremediation than a two-ring PAH. Increased hydrophobicity leads to a
larger amount of contaminant in the soil matrix, particularly the soil organic matter. Thus
the amount of contaminant in the soil solution is decreased, which decreases the amount

in equilibrium with the soil gas. Consequently, the amount of contaminant available for

111

volatilization is decreased. Again, decreased volatilization leads to increased
effectiveness factors. The greater the effectiveness factor, the more responsive is the
system to the plant.

Figure 5-6 is a plot comparing different values of the desorption rate coefficient.
The effectiveness factor for phytoremediation increases with decreasing values of the
desorption rate coefficient. As the rate of mass transfer from the soil matrix eventually
limits the rate of volatilization, a technology that increases the driving force for mass
transfer will benefit the removal of contaminant. Planted systems are superior to
unplanted systems in this respect as transpirational uptake and microbial degradation
serve to increase the driving force for mass transfer.

Figure 5-7 is a plot comparing systems in which the site fractions have been
altered. The topmost curve (qu = 0) results when the equilibrium sites are converted into
nonequilibrium sites. The effectiveness of phytoremediation increases in this scenario, as
volatilization becomes limited by mass transfer, and contaminant removal occurs at a
slower rate in the unplanted system. Conversly, converting the nonequilibrium sites into
equilibrium sites reduces the effectiveness of phytoremediation when compared to the
base case. As volatilization becomes a more significant mechanism for naphthalene
removal, the effectiveness of phytoremediation decreases. Finally, converting the
immobile site fraction into equilibrium sites resulted in a significant decrease in the
effectiveness of planted systems. Though the immobile fraction only comprises 6% of
the total soil domain, removing this fraction had a profound impact on the effectiveness
factor. Essentially, naphthalene that is available to plant remediative processes, but is not

available for volatilization, has a significant role in the effectiveness of phytoremediation

112

of naphthalene in SpAf soil. Simply put, increasing the amount of naphthalene available
for phytoremediation that is not mobile in unplanted soils will increase the effectiveness
factor for phytoremediation. A water extractable amount of contaminant, that is not
removable by the mechanisms available in the unplanted system (e. g. volatilization), will

tend to favor technologies that can “mobilize” this immobile component.

Decision making

From the previous analysis, determining the regime that controls contaminant
remediation is useful. Contaminant remediation limited by the first regime will require
an improved approach if phytoremediation is to be beneficial. Several reasons exist for
remediation controlled in the first regime. The rate of transpiration relative to the rate of
volatilization may be too small. This can be ameliorated by using a different plant
species with a greater transpirational flow, such as Salix spp. (Willow). When
phytoremediation is used in the phytostimulation mode, the degrading portion of the
consortia may not be effectively stimulated. This limitation can be overcome through use
of plant species that exude molecules chemically resembling the contaminant. The use of
Morus rubra (Red Mulberry)l8] is an example of the stimulation of PCB degrading
bacteria by exudation of phenolic compounds. These phenolic compounds induce
degradative activity or promote cometabolism. The goal of increasing transpiration or
enhancing phytostimulation is to shift the control of contaminant remediation from
regime one to regime two or three.

Contaminant remediation controlled in the second regime is characterized by the
beneficial application of phytoremediation. Unlike the first regime, plants do enhance the

remediation of contaminant in the second regime. Like the first regime, the rate that

113

contaminants are remediated from the soil is limited by reactive processes, not mass
transfer processes. Accelerating the rate of transpiration or enhancing the mechanisms of
phytostimulation would benefit the overall remediation. Proper plant selection and the
optimal application of nutrient amendments are options for increasing the
phytoremediation rate. The effect of either of these actions is to shift the control of
contaminant remediation from regime two to regime three.

Rate-limited mass transfer is the controlling mechanism in the third regime. Like
the second regime, the use of plants remains beneficial when compared to an unplanted
treatment. However, the rate at which phytoremediation is occurring is limited by the
rate at which contaminant mass transfer is occurring. Accelerating the rate of
phytoremediation depends upon a plant’s ability to enhance the desorption rate from the
soil matrix. This is an area that warrants further study, as plants that enhance the rate of

desorption from mass transfer limited soil regimes have yet to be identified.

Approach

The general approach developed in this work does not favor the use of a particular
model. The effectiveness factor can be used to compare the rate of remediation between
planted and unplanted systems, while using any descriptive model (preferably a model
with a degree of experimental validation). Plotting the effectiveness factor versus the
Thiele modulus will result in a profile that can be divided into separate regimes. The
mechanisms that govern these regimes can be illuminated by carefirl analysis. Decision
making regarding the proper use of phytoremediation can then be made in a more

informed manner.

114

transpiration

 

 

'-.'_",...-c~u
.

l.-
‘-

-----
......
------
oooooooooooooooo
Con.
00....
......
a...
I...
-
'-
u
‘o
a
a
o

, .-"""'volatilizatioi‘r-.._
desorption '

   
  

......

    
  

microbial
degradation

0
0..
a.
c, ,o
h ,-
‘ o
"no". -----

leaching

transpirational

Figure 5-1. A diagram detailing some of the processes involved in the
phytoremediation of naphthalene. The hatched regions represent soil aggregates.

115

 

Micropore soil

Soil gas (contaminants mobile by gas
<—_ diffusion) (contaminants
immobile)
It |

 

A |
| l l l
Micropore soil
<— "’ Soil solution solution “ 9

It I ’1‘ M
l l I I l
Equilibrium soil Nonequilibrium Equilibrium soil-
domain soil domain micropore
domain

 

 

 

 

 

 

Figure 5-2. A box model of the transport and reactive processes occurring in
rhizosphere soil. The double arrows represent equilibrium processes and the single
arrows represent kinetic or transport processes. The dashed arrows represent the
kinetic process added to the phytoremediation model.

116

 

1.1 - . -.--wfi,-fi.,-

 

1.09 - .

 

 

 

 

1.08 - .
1.07 " -
.5
E 1.06 - -
§ 1.05 » -
.2 - - -
’5 Regime l Regime ll Regime Ill
33:) 1.04 - .
1.11
1.03 r -
1.02 - 4
1.01 r -
l . . ..--.l . . - ---..l - . A-A.,‘l . - . .---.l . n . --.-.l . . . -.-.n
-2 -1 0 1 2 3 4 5
10 10 10 10 10 10 10 10

Thiele modulus

Figure 5-3. A plot demonstrating the regimes that control the phytoremediation of
naphthalene in Spinks A-horizon soil.

117

 

 

 

 

 

 

 

 

2.5 . . . fix a, c a,
Da=50
21-
.5 1
Q
62
(I)
8
t:
0
.2
6
a—‘j Da= 100
L1.)
1.5- .
Da=500
l A AAAAAI A A AAAAAAl A A AAAAAAI A A AAAAAAI A A AAAAAA
-l 0 l 2 3 4 S
10 10 10 10 10 10 10

Thiele modulus

Figure 5-4. The effect of increasing the Damkohler number on the effectiveness
factor.

118

 

 

1.8'

1.75

 

1.4-

Effectiveness factor

1.3"

1.2-

 

1.1'

 

 

 

] A A AAAAAA. A A AAAAAAl LA AAA-AA. A A AAAAAL4 A AAAAAA|
-| 0 l

10 10 10 102 103 10‘ 10
'Ihiele modulus

5

Figure 5-5. The effect of increasing the soil-water partition coefficient on the
effectiveness factor.

119

 

1.15 . . . . -

1.1r

 

 

Effectiveness factor

1.05 ’

 

’0
.
‘0’
o"—.
A.""‘

L ALLA-AA. A A ‘A-ALLI A A AAA-AL. J A AAAAAA

-1 0 l 2

10 10 10 10 103 10’
Thiele modulus

 

 

Figure 5-6. The effect of the desorption rate coefficient on the effectiveness factor.

120

 

 

 

 

 

 

 

 

1.2 . - . n- - , -
1.18- .
1.16- 56:0 -
1.14- 1
.5
g 1.12- -
g 1.1- Basecase -
.2
8
t 1.08” "'
LLI
1.06- f =0 '
neq
1.04- -
1.02- fim=0 .
1-l d- 04 “““"I - --AA“-.2 A - --‘“J3 LLJ“““4 A “‘L“‘ 5
10 10 10 10 10 10 10

Thiele modulus

Figure 5-7. The effect of converting the equilibrium fraction to the nonequilibrium
fraction (f.eq = 0), converting the nonequilibrium fraction to the equilibrium fraction
(fneq = 0), and converting the immobile fraction to the equilibrium fraction (f;In = 0).

121

Table 5-1. A list of model variables and parameters for the planted and unplanted
mathematical models.

 

 

 

 

 

Variable Description Units

Cg contaminant concentration jig/(ml of soil gas)
in the soil gas

Cw contaminant concentration jig/(ml of soil solution)
in the soil solution

Seq contaminant concentration pg/(g of soil solid)
in the equilibrium-soil solid

S. contaminant concentration 118/(g of soil solid)

in the microporous
equilibrium soil solid

 

 

 

 

 

 

 

 

Sncq contaminant concentration pg/(g of soil solid)
in the nonequilibrium soil
solid
t time hrs
2 - vertical distance ' cm
Parameter
Dgg effective soil gas diffusivity cmz/hrs
feq fraction of soil that is in --
equilibrium with the soil
solution
fneq fraction of soil that is in --
nonequilibrium with the soil
solution
fim fraction of soil that is in --

equilibrium with the
immobile soil solution

 

 

 

 

Kd soil-water partition (ml of soil solution)/
coefficient (g of soil solid)
h rhizosphere depth cm
kneq first-order desorption rate l/hrs
coefficient
kp first-order kinetic 1/hrs

coefficient attributed to
plant action

 

 

 

 

9g soil gas content (ml of soil gas)/
(ml of total soil volume)
9,, soil moisture content (ml of soil moisture)/
(ml of total soil volume)
pb soil bulk density (g of dry soil)/

 

 

 

(ml of total soil volume)

 

122

Table 5-2. A list of dimensionless variables and parameters for the planted and
unplanted mathematical models.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Dimensionless Description Structure
variable
C; dimensionless soil solution Cw
concentration C l
51;. dimensionless microporous soil Sm.
solid concentration S
5;“, dimensionless nonequilibrium Sm
concentration ‘—
Sncqo
77200 effectiveness factor at 200 days (amount degraded in
unplanted treatments)200 days/
(amount degraded n planted
treatments)zoo day,
T dimensionless time t ' kneq
Z dimensionless vertical distance z
h
Dimensionless groups
Da Damkohler number D, e
kneq . h 2
R1 Dimensronless parameter 1 1+ Pt fa, K d + gg H
9. 9...
R2 Drmensronless parameter 2 Pt fiqud
9..
R3 Drmensronless parameter 3 101. 1;»th
9..
R4 Drmensronless parameter 4 (fig + fm )9g H
49..
R5 Dimensionless parameter 5 6,,
pl) Kd + 6w
(25 Thiele modulus kp
m'q

 

 

 

’r The symbol (0) denotes an initial value.

123

Table 5-3. Base case parameter values used for naphthalene transport and reaction
in soil.

kp kneq feq fncq I(d D g.c H 6w n

__ (ml soln)/ 2 (ml soln)/ (ml soln)/ (ml void)/
“h “h (g solid) cm (ml total) (ml total) (ml total)
0.15 0.0010 0.36 0.58 5.0 22.4 0.02 0.15 0.5

 

124

Table 5-4. Dimensionless number and parameter values used to generate Figures 3
through 8.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

K,
Figure # Da fcq f“,q flm (ml soln/ kncq
g solid) (l/hr)
3 552 0.36 0.58 0.06 5 0.001
4 50, 100, 500 0.36 0.58 0.06 5 0.001
5 552 0.36 0.58 0.06 5, 50, 500 0.001
6 552 0 0.94 0.06 5 0.01, 0.001, 0.0001
7 (fcq=0) 552 0 0.94 0.06 5 0.001
7 (fncq=0) 552 0.94 0 0.06 5 0.001
7 (fun-=0) 552 0.42 0.58 0 5 0.001

 

125

 

References

1.

F roment, G.F. and KB. Bischoff, Chemical Reactor Analysis and Design. 2nd ed.
1990: John Wiley and Sons, Inc.

Briggs, G.G., R.H. Bromilow, and A.A. Evans, Relationships between
Iipophilicity and root uptake and translocation of non-ionised chemicals in
barley. Pesticide Science, 1982. 13: p. 495-504.

Trapp, S. and J .C. McFarlane, eds. Plant Contamination: Modeling and
Simulation of Organic Chemical Processes. 1 ed. 1995, Lewis Publishers: Boca
Raton. 254.

Chiou, C.T., G. Sheng, and M. Manes, A partition-limited model for the plant

uptake of organic contaminants fiom soil and water. Environmental Science and
Technology, 2001. 35: p. 1437-1444.

Cunningham, S.D., W.R. Berti, and J .W. Huang, Phytoremediation of
contaminated soils. Trends in Biotechnology, 1995. 13: p. 393-397.

Siciliano, SD. and J .J . Germida, Mechanisms of phytoremediation: biochemical

and ecological interactions between plants and bacteria. Environmental Reviews,
1998. 6: p. 65-79.

Millington, R. and J. Quirk, Permeability of porous solids. Transactions of the
Faraday Society, 1961. 57: p. 1200-1207.

Leigh, M., et al., Root turnover: an important source of microbial substrates in
rhizosphere remediation of recalcitrant contaminants. Environmental Science and
Technology, 2002. 36: p. 1579-1583.

126

u 111
11 ll 1

 

Ill
3 1