&EPA
United States'
Environmental Protection
Agency
Environmental Monuyrmg
Systems Laboratory
PO Box 15027
Las Vegas NV 89114-5027
EPA/600/4-86/020
April 1986
Research and Development
Guidelines for
Field Testing
Soil Fate and
Transport Models
Final Report
-------
EPA 600/4-86-020
April, 1986
GUIDELINES FOR FIELD TESTING SOIL FATE AND TRANSPORT MODELS
FINAL REPORT
by
Stephen C. Hern
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
Las Vegas, Nevada 89114
and
Susan M. Melancon
Environmental Research Center
University of Nevada, Las Vegas
Las Vegas, Nevada 89154
(Editors)
Cooperative Agreement Number
CR810550-01
Project Officer
Stephen C. Hern
Exposure Assessment Research Division
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada 89114
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
LAS VEGAS, NEVADA 89114
-------
NOTICE
The Information In this document has been funded wholly or In part by
the United States Environmental Protection Agency under cooperative agreement
CR810550-01 to the Environmental Research Center, University of Nevada,
Las Vegas. It has been subject to the peer and administrative review of the
Agency and It has been approved for publication as an EPA document.
11
-------
CONTRIBUTORS
Anthony S. Doniglan, Jr.
Aqua-Terra Consultants
Palo Alto, CA
Kenneth F. Hedden
Exposure Assessment Research Division
U.S. Environmental Protection Agency, Las Vegas, NV
Stephen C. Hern
Exposure Assessment Research Division
U.S. Environmental Protection Agency, Las Vegas, NV
William A. Jury
Department of Soil and Environmental Science
University of California—Riverside, Riverside, CA
Susan M. Melancon
Environmental Research Center
University of Nevada—Las Vegas, Las Vegas, NV
James E. Pollard
Environmental Research Center
University of Nevada—Las Vegas, Las Vegas, NV
P. Suresh Chandra Rao
Soil Science Department
University of Florida, Gainesville, FL
Jerald L. Schnoor
Department of Civil and Environmental Engineering
University of Iowa, Iowa City, IA
Richard L. Valentine
Department of Civil and Environmental Engineering
University of Iowa, Iowa City, IA
111
-------
1v
-------
CONTENTS
Figures vli
Tables 1x
Preface xfi
Chapter
1 Overview of Terrestrial Processes and Modeling
by A. S. Donigian, Jr., and P. S. C. Rao 1-1
2 Transport Mechanisms and Loss Pathways for Chemicals in Soils
by W. A. Jury and R. L. Valentine 2-1
3 Generic Steps in the Field Validation of Vadose Zone
Fate and Transport Models
by S. C. Hern, S. M. Melancon, and J. E. Pollard 3-1
4 Example Field Testing of Soil Fate and Transport Model,
PRZM, Dougherty Plain, Georgia
by K. F. Hedden 4-1
5 Example Model Testing Studies
by A. S. Donigian, Jr., and P. S. C. Rao 5-1
Appendices
A Chemical Movement Through Soil
by W. A. Jury A-l
B Volatilization from Soil
by W. A. Jury B-l
C Adsorption of Organic Chemicals onto Soil
by W. A. Jury C-l
D Mathematical Derivation of Chemical Transport Equations
by W. A. Jury D-l
E B ^transformation
by R. L. Valentine and J. L. Schnoor E-l
-------
CONTENTS (continued)
Appendices Continued
F Nonbiological Transformation
by R. L. Valentine F-l
G Spatial Variability of Soil Properties
by W. A. Jury G-l
vl
-------
FIGURES
Number Page
1-1 Chemicals in the soil environment 1-2
1-2 Hierarchy of model components 1-10
3-1 Lindane distributed through four soil columns compared
with SESOIL, PESTAN, and PRZM model predictions 3-15
4-1 Study site location 4-7
4-2 Dougherty Plain field site 4-11
4-3 Schematics of monitoring site 4-14
5-1 Comparison between SESOIL predicted and measured depth-
averaged soil-water contents for the Lake Hamilton site. . . . 5-9
5-2 Comparison of SESOIL predicted and measured percent
aldicarb remaining in the soil column using calibrated
parameter values for the Lake Hamilton, Florida, site 5-11
5-3 Comparison of SESOIL predicted and measured vertical
concentration distributions 30 and 20 days after
application, respectively, using calibrated parameter
values for the Lake Hamilton, Florida, site 5-13
5-4 Comparison of SESOIL predicted and measured vertical
concentration distributions 60 and 45 days after
application, respectively, using calibrated parameter
values for the Lake Hamilton, Florida, site 5-14
5-5 Comparison of SESOIL predicted and measured vertical
concentration distributions 90 and 75 days after
application, respectively, using calibrated parameter
values for the Lake Hamilton, Florida, site 5-15
5-6 Comparison of PRZM, PISTON, PESTAN, and SESOIL predicted
and measured percent aldicarb remaining in the soil
column using parameter values 5-17
5-7 Comparison of measured metolaxyl concentration profiles
at three sampling dates with those predicted by PRZM model . . 5-21
vii
-------
FIGURES (continued)
Number Page
5-8 Comparison of measured and predicted metolaxyl
concentrations in the 0-15 cm depth increment 5-23
Appendices
B-l Calculated vapor, liquid, and total effective diffusion
coefficients as a function of water content for
20 chemicals B-6
B-2 Volatilization fluxes predicted for 20 chemicals in soil
at same concentration, for three water evaporation
rates E=0, E=2.5 mm/day, E=5.0 mm/day B-10
viii
-------
TABLES
Number Page
1-1
1-2
1-3
1-4
2-1
2-2
2-3
2-4
2-5
3-1
3-2
3-3
4-1
4-2
4-3
4-4
4-5
Definition of Terms Used 1n ASTM Standard Practice
A Classification of Leaching Models
Simulation Capabilities of SESOIL Cycles
Summary of Significant Process-Parameter Relationships
In Chemical Transport and Loss to the Atmosphere
Methods of Measurement of Model Parameters or Soil
Properties Relevant to Modeling and Validation
Sources of Measured Values for Various Chemical Properties . . .
Major Factors Influencing Biological and
Nonbiological Transformations
General Approaches and Methodologies for the
Steps in Field Validation of Soil Fate and Transport Models. . .
Some Factors Influencing Selection of the Most
Appropriate Soil Sampling Gear for Vadose Zone Monitoring. . .
General Sample Collection and Logistical Considerations
for Field Validations
Input Data Requirements for PRZM
The Major Output Prediction Functions Currently
Chemical and Physical Properties of Aldicarb and
Its Sulfoxide and Sulfone Derivatives
1-11
1-13
1-17
1-24
2-3
2-9
2-12
2-14
2-17
3-2
3-11
3-12
4-3
4-4
4-5
4-8
4-9
1x
-------
TABLES (continued)
Number Page
4-6 lexicological Properties of Metolachlor 4-9
4-7 Area (ha) per Soil Series 1n Daugherty Plain, Georgia 4-10
4-8 Sample Allocation by Soil Type in Daugherty Plain, Georgia . . . 4-12
5-1 Initial and Calibrated Hydrologic Cycle Parameter
Values for the Lake Hamilton, Florida Site 5-8
5-2 Computed Statistics for Comparisons at the
Lake Hamilton Site 5-10
5-3 Initial and Calibrated Pollutant Cycle Parameter Values
for the Lake Hamilton, Florida, Site 5-10
5-4 Pollutant Cycle Parameters used in Testing PRZM, PISTON,
and PESTAN on the Lake Hamilton, Florida, Site 5-16
5-5 Comparison of Model Leaching Predictions for Lake Hamilton . . . 5-16
5-6 Revised Statistics for Model Comparisons at the Lake
Hamilton Site 5-18
5-7 Summary of Hydrologic and Soil Characteristics of the Two
Field Sites Used in the Metolaxyl Study 5-18
Appendices
A-l Soil, Environmental, and Management Parameters Influencing
Chemical Transport Through Soil A-3
A-2 Potential Problems Encountered in Field Validation of
Water and Chemical Transport Models A-10
A-3 Limiting Transport Model Assumptions Imposed by Inadequate
Measurements A-15
B-l Principal Soil, Environmental, and Management Parameters
Influencing Chemical Volatilization from Soil B-5
B-2 Parameters Required for Volatilization Estimates in
the Field B-12
C-l Intermolecular Interactions Involved in Adsorption C-2
C-2 Soil, Chemical, and Environmental Properties Influencing
Adsorption of Chemicals Onto Soil C-4
-------
TABLES (continued)
Number Page
E-l Monod Based Rate Expressions E-8
E-2 Major Environmental Factors Affecting Blotransformation E-12
E-3 Limitations in Applying Degradation Rate Expressions E-21
F-l Relative Oxidation States F-5
F-2 Factors Influencing Nonbiological Reactions F-9
G-l Field Studies of Soil Matrix and Water Retention Properties. . . 6-4
G-2 Field Studies of Water Transport Properties G-6
G-3 Field Studies of Chemical Concentrations G-9
G-4 Log Distribution of Solute Transport Velocity Parameters
in Field Experiments G-ll
G-5 Sample Sizes Required to Have a 95% Probability of
Detecting a Change of F% in the Mean Using a T-Test
with o=5% G-13
G-6 Lognormal Scaling Factor Parameters Measured in
Field Experiments G-15
G-7 Correlation Length Parameters Measured in Field
Experiments G-17
xi
-------
PREFACE
The present document was conceived to address the needs of a growing body
of individuals working with soil fate and transport models. A variety of
stochastic and deterministic soil leaching models have been developed in the
past decade, particularly to measure the transport and transformation of organic
pollutants moving through the vadose zone. The U.S. Environmental Protection
Agency (USEPA) has supported the development and testing of three such models,
largely for screening purposes: the Seasonal Soil Compartment Model (SESOIL),
the Pesticide Root Zone Model (PRZM), and the Pesticide Analytical Model
(PESTAN). This document seeks to provide the reader with a general overview of
the uses and limitations of vadose zone models and with a generic set of guide-
lines for field collection of data necessary to calibrate and evaluate their
predictive capabilities. The document will specifically examine the assump-
tions, data requirements, and processes underlying SESOIL, PRZM, and PESTAN.
These guidelines are intended for use by readers coming from a diversity
of scientific backgrounds. However, it is recommended that all users have a
working knowledge or at least a Bachelor's level training in mathematics, soil
science, engineering, modeling, or one of the physical sciences. The first
four chapters of the document provide background information to the general
model user. Chapter 1 includes an overview of the basic soil processes repre
sented in vadose zone models, including process definitions and common model
xi i
-------
assumptions. In particular, SESOIL, PRZM, and PESTAN are described and com-
pared. In Chapter 2, chemical transport and transformation processes which are
presumed operative in the vadose zone are examined more closely. This chapter
provides the reader with tables of references for process rate constants and
field collection methods for parameters commonly input to vadose zone models.
Also in this chapter, the reader is directed to the various appendices where
detailed process descriptions, mathematical derivations of significant equa-
tions underlying these processes, and information regarding the spatial vari-
ability of soil properties can be found. The reader is encouraged to be cog-
nizant that although the process mathematical derivations are separated from
the main text for ease in reading, an understanding of the material included in
the appendices is essential if vadose zone models are to be meaningfully used
for screening purposes or in field validation attempts.
In Chapter 3, the reader is introduced to a stepwise generic approach for
the implementation of a data acquisition strategy which can be used in field
model testing. This chapter provides guidelines on criteria to be used in
model selection and establishment of validation acceptance criteria, as well
as site and compound selection, and implementation of a field-sampling program.
Chapters 4 and 5 then present two scenarios describing actual field validation
attempts using the three models we have more closely scrutinized. In the first
of these examples (Chapter 4), the generic guidelines described in Chapter 3
are followed step-by-step to demonstrate how the guidelines can aid in the
design of a field study intended to yield data For use in subsequent model
validations. Thus, this scenario will demonstrate the steps that must be
taken to deal with such considerations as site and compound selection, model
xiii
-------
calibration and sensitivity testing, field data collection, chemical analyses,
and quality assurance. The second scenario (Chapter 5) 1s based on field data
which were originally collected for purposes other than model testing and which
were then applied to various model validation attempts. This scenario 1s the
more common 1n the environmental modeling literature and illustrates for the
reader the difficulties in post hoc comparisons of field data with model pre-
dictions when the original study design had a different data collection goal.
This document provides a unique compilation of many of the factors that
must be considered in the field testing and application of vadose zone fate and
transport models. It does not address models or processes specifically
applicable either to the soil saturated zone or to the movement and transforma-
tion of inorganic compounds. Furthermore, many multi-dimensional environmental
elements affecting the processes under consideration, such as the effects of
surface runoff, plant washoff, or distribution of plume emissions, are beyond
the scope of the present report. However, references are given throughout the
document to assist the user in the appropriate application of vadose zone
models. Although these guidelines are directed towards areas Impacted by
organic pollutants, they contain valuable information for the model user work-
ing with other vadose zone pollution problems as well.
xiv
-------
CHAPTER 1
OVERVIEW OF TERRESTRIAL PROCESSES AND MODELING
by
A. S. Donigian, Jr., and P. S. C. Rao
INTRODUCTION
The terrestrial environment which extends from the top of the growing
vegetation to the capillary fringe of ground water is the primary home for most
living things on earth. We deliberately introduce chemicals into this environ-
ment to grow and expand our food supply, to protect us and our crops from pests
and disease, and to dispose of our wastes; unintended entry also occurs through
transport accidents, inaccurate or inappropriate application procedures, and
leaking storage facilities. The joint occupancy and use of the terrestrial
environment by chemicals and other living organisms (humans, animals, and
plants) can lead to significant human and animal exposures to these often toxic
chemicals with resulting detrimental health impacts.
Since chemicals are necessary for maintenance of current life styles and
standards of living, a better understanding of the fate and migration of chem-
icals introduced to the terrestrial environment would allow us to better eval-
uate and control the resulting exposure and health risk. This environment is a
dynamic, interdependent system of abiotic and biotic factors that are linked by
physical, chemical, and biological processes. Chemicals introduced into this
system use these linkages to migrate within and between the various media and
are in turn transformed and degraded as they move.
This chapter presents an overview of these processes influencing the fate
and migration of chemicals in the terrestrial environment. This overview
provides the framework for discussing the underlying concepts of mathematical
models that have been developed both as research tools (to help us better
understand the terrestrial system) and as regulatory or management tools (to
help us assess and control the exposure and risks resulting from chemical use
and waste disposal). Model classifications are discussed, selected models are
described, model selection criteria and guidelines are presented, and model
limitations are explored in terms of their ability to represent chemical fate
and movement in the soil environment.
SOIL SYSTEM PROCESSES
Figure 1-1 schematically demonstrates the complex and dynamic interaction
of processes controlling the fate and transport of chemicals in the soil en-
vironment. In order to provide some order to the multitude of individual
1-1
-------
CHEMICAL SOURCES
DEGRADATION
VOLATILIZATION
VOLATILIZATION
RUNOFF & EROSION
TABLE
DEGRADATION
Figure 1-1. Chemicals In the soil environment.
-------
processes that can occur, we have grouped the terrestrial processes into the
following five categories, as suggested by Rao and Jessup (1982) and Wagenet
and Rao U555T:
Transport
Sorption
Transformati on/Degradati on
Volatilization
Plant Processes
Although we have listed these categories individually, it is the dynamic,
temporal, and spatial interaction among these groups of processes that deter-
mines the ultimate disposition of chemicals in the soil. In the paragraphs
below, each of these process groups is defined and discussed with specific
emphasis on their role in affecting chemical fate and transport and on their
mutual interactions. These processes are discussed more thoroughly in Chapter
2.
Transport
After chemicals are introduced into the terrestrial environment, they can
move by three separate pathways and mechanisms: by runoff and erosion to the
aquatic environment, by volatilization to the air environment, and by lateral
or vertical movement (or leaching) to ground water. The transport component is
critical to the environmental fate of a chemical; without it, environmental
contamination problems would be minimal. Chemicals applied for agricultural or
siIvicultural purposes would remain on target areas until they were degraded or
transformed. Chemical spills would impact only the immediate spill site, and
contamination from hazardous waste disposal and land treatment sites would be
limited to the facility confines. However, we are well aware that agricultural
chemical runoff does occur, that chemical applications and spills can move over
the land and through the soil to contaminate surface and ground-water supplies,
and that hazardous wastes often tend to migrate away from their storage, dis-
posal, and treatment facilities.
The emphasis in this manual is on the vertical movement of chemicals
through the soil to evaluate the potential for ground-water contamination.
However, it is important to recognize that the other transport processes can
occur depending on specific environmental, soil, chemical, and management
conditions and practices. Moreover, the relative significance of runoff/
erosion and volatilization will affect the amount of chemical that remains in
the soil and that can subsequently leach to ground water.
Sorption
Since movement of chemicals to ground water is primarily a liquid-phase
process involving water movement and associated dissolved solute, the parti-
tioning of the chemical between the sorbed and dissolved phases is a critical
factor in determining how rapidly the chemical will leach. Chemicals that sorb
readily to soil organic matter and clay particles will not migrate significant-
ly away from the region of the soil profile in which they are initially placed
or applied. Thus, highly-sorbed chemicals applied on soils will usually remain
1-3
-------
on or near the land surface unless extensive soil cracks, macropores, and other
preferential flow paths exist to allow for a rapid vertical movement of soil
particles and associated chemicals. For these chemicals, transport by surface
runoff and erosion may be of greater environmental significance than chemical
leaching.
On the other hand, chemicals that do not sorb will exist primarily in the
dissolved phase, and their movement to ground water will be controlled by the
relative timing and amount of water applications (either by precipitation or
irrigation), soil characteristics, and the "fate" processes discussed below.
Many chemicals of environmental interest are moderately sorbed and thus exist
in both the dissolved and sorbed phases. For these chemicals, the surface-
runoff and erosion components usually comprise a small fraction of the total
available chemical in the soil although the resulting aquatic concentrations
can be significant. For agricultural applications, the runoff and erosion
losses for pesticides are usually a small percentage of the total application
amount (Baker, 1980; Uauchope, 1978; Wauchope and Leonard, 1980). However,
land disposal and treatment options for highly sorbed waste constituents can
generate significant surface runoff loadings if not properly designed to pre-
vent or minimize surface runoff. Thus, for chemicals in the mid-range of
sorption properties, both runoff and leaching may be important, depending on
the specific chemical, environmental, soil, and management conditions.
Transformation/Degradation
Whereas transport processes specify the vehicles for chemical movement,
and sorption determines the relative importance of dissolved versus sorbed
phase transport, transformation and degradation processes ultimately determine
whether a chemical will persist long enough in the soil environment to be of
concern. We have lumped the transformation and degradation processes to include
the primary chemical and biological mechanisms that encompass the "fate" of a
chemical and determine its persistence in the soil. The key processes include
biotransformation, chemical hydrolysis, photolysis, and oxidation-reduction.
These individual processes are defined and summarized in Chapter 2 and discussed
in detail in Appendices A-G. Although volatilization and plant processes can
be considered as "fate" processes, they also include transport considerations
and other special characteristics that require individual discussions (below).
Non-persistent chemicals (i.e., half lives less than 15-20 days) that
remain in the soil at significant levels for only a few months after initial
placement or application are highly dependent on the relative timing of rain-
fall or irrigation events to demonstrate significant mobility and potential for
contamination. Thus, for these chemicals, the transport vehicles of surface
runoff and soil water movement must occur during the period when the chemical
is resident in the soil profile in order for chemical transport to occur. The
classic example of this is agricultural runoff of relatively non-persistent
pesticides. Numerous studies have shown that the highest pesticide runoff
concentrations consistently occur in the first few runoff-producing storm
events following a field application (Baker, 1983; Johnson and Baker, 1982;
Smith et al., 1978; Ellis et al., 1977; Wauchope, 1978).
1-4
-------
A somewhat analogous but different situation occurs for chemical leaching;
the highest concentrations with the deepest penetration Into the soil profile
will likely occur during the few months Immediately following placement or
application. However, superimposed on this vertical transport is a change in
the specific transformation/degradation processes and their rates, that may be
active at different depths of the soil profile. Processes such as photolysis
and volatilization occur primarily in the top few centimeters of the soil pro-
file, whereas biotransformation, hydrolysis, and oxidation-reduction can occur
throughout the crop root zone and below. Although the specific rates at which
these processes occur depend on individual chemical, soil, and environmental
characteristics, a common situation is that surface processes occur at faster
rates than subsurface processes. Thus, if a chemical is soil-incorporated or
buried, it would degrade at a slower rate than if it were surface applied, and
thus it would present greater potential for leaching to ground water.
For surface-applied chemicals that are non-persistent, the faster the
chemical moves to the subsurface zones by soil-water movement, the greater the
potential for reaching ground water because of the slower decay in the sub-
surface soil. This situation would be reversed if subsurface degradation
processes occurred at rates faster than surface processes.
For persistent chemicals that demonstrate half-lives in soils on the order
of 100 days or more, the relative timing of rainfall or irrigation events (i.e.,
with respect to placement and application) is less critical to chemical movement
since the chemical resides in the soil for a much longer time period. However,
the impact of surface and subsurface processes and of their rates is still a
primary consideration.
Included within this category of transformation and degradation processes
are those chemical or biological mechanisms which can transform a parent com-
pound into various metabolites or daughter products which may be of equal or
greater environmental concern due to mobility, persistence, or toxicity char-
acteristics. Moreover, these characteristics of the metabolites may not be the
same as those of the parent chemical. Experience with a number of agricultural
chemicals (Hornsby et al., 1983; Jones et al., 1983; Dean et al., 1984c;
Bilkert and Rao, 1985) has shown the need to consider transformation products
and their differing mobility and persistence when evaluating the environmental
fate of a chemical.
Volatilization
Volatilization is generally defined as the loss of chemical in vapor
form to the atmosphere from soil, plant, or water surfaces. Since our interest
here is in the soil environment, volatilization from soil and plant surfaces is
of primary concern. Since chemical applications in agriculture, silviculture,
and land application of wastes (i.e., treatment or disposal) can be made to
both bare soil and growing plants, volatilization from these types of surfaces
can occur and can be important. In addition, volatilization of wastes buried
in landfills and waste disposal sites can be a significant pathway for human
exposure.
1-5
-------
As a chemical fate process, volatilization can be a major loss mechanism
resulting in reduced amounts of chemical in the soil that would be susceptible
to leaching and surface runoff. The extent to which volatilization is important
depends on chemical characteristics such as vapor pressure and Henry's Law
constant. In addition, soil, environment, and management variables are impor-
tant (as discussed in Chapter 2). For buried and soil-incorporated chemicals,
volatilization controls the extent to which the chemical will exist in the
vapor phase within the soil and thus will move through the soil by vapor diffu-
sion. Consequently, volatilization includes aspects of both chemical fate and
transport.
Plant Processes
Vegetation is an integral part of the terrestrial ecosystem that encom-
passes the soil environment. Chemicals applied to the land will enter the
biological system of a plant and will undergo the full range of transport and
fate mechanisms discussed above. In effect, the plant is a subsystem within
the terrestrial ecosystem where chemicals enter, are transported, are sorbed
onto organic carbon, are transformed and degraded, and, specific to biological
systems, can accumulate in different portions of the plant. This bioaccumula-
tion process is especially important since it has implications for direct
exposure to animals and humans through the food chain.
Chemicals applied to the land surface, such as in agricultural or silvi-
cultural operations or land application of wastes, can be intercepted by grow-
ing vegetation so that a portion of the applied chemical remains on the plant.
On the plant leaf surfaces, fate processes such as photolysis, biodegradation,
and volatilization can occur in addition to direct uptake and absorption into
the body of the plant. During subsequent rainfall or irrigation events, the
chemical can be dislodged and washed off the leaf surfaces to the soil
(Donigian and Dean, 1985).
From the soil, the plant roots allow uptake of the chemical into the plant
where it can be transported or translocated among the roots, stems, and leaves.
Within the body of the plant, the translocation process is affected by the
partitioning of the chemical between the dissolved and sorbed phases (Briggs
et al., 1982). Chemicals transported to the leaves can be exuded onto the leaf
surface and can undergo the fate processes noted above. Accumulation of chemi-
cals within the plant will vary for different portions such as the root, stems,
leaves, fruit, etc.; subsequent potential exposure to animals and humans depends
on the extent and location of the accumulation. When plants die, drop their
leaves, or are incorporated into the soil, the remaining chemical may be returned
to the soil where it is subject to all the soil fate and transport processes.
Thus, plant processes include a complex set of interacting mechanisms that
control the chemical uptake, transformation, and fate within the plant system.
For agricultural and silvicultural operations, and land application and treat-
ment of wastes, plant processes can be a significant component of chemical fate
and exposure. For landfills and buried waste disposal sites, these processes
are likely to be negligible due to the lack of extensive vegetation and rooting
systems.
1-6
-------
Space and Time Variability of Soil Processes
A key aspect of the soil environment is its inherent variability in space
and time. Spatial variability refers to changes in a measured value (e.g.,
soil property, solute concentration) at a specified depth throughout the area
of interest (e.g., a field) as well as changes with depth at a specific loca-
tion. Temporal variability is the change in a value with time for any specific
location in the field. In effect, we are dealing with a fully three-dimensional
dynamic system resulting from the indigenous spatial variation of soil proper-
ties and impacted by the temporal variation of natural processes and human in-
tervention. These variations are comprised of both intrinsic factors, such as
natural variations in soil characteristics, and extrinsic factors, such as water
and chemical applications, tillage practices, etc. (Rao and Wagenet, 1985).
The processes described in the foregoing section demonstrate significant tempo-
ral and spatial changes as a result of this total variability, both intrinsic
and extrinsic. These variations have important implications for collection of
field measurements, interpretation of field data, and modeling of soil system
processes.
Spatial variations in chemical behavior in soil systems are primarily a
result of significant point-to-point changes in soil properties related to the
structure, texture, composition, mineral content, organic content, etc. These
properties have a direct impact on virtually all the parameters that character-
ize chemical transport and on many of the parameters that characterize chemical
fate processes. In addition, these properties will influence environmental
conditions - soil water content, temperature, pH - which in turn impact the
chemical transport and fate processes.
For agricultural systems, additional variability is introduced by the
specific agronomic practices that may impose a specific variation: for example,
banded, surface-applied, or soil-incorporated chemical application; planting
and tillage in rows; and furrow, drip or sprinkler irrigation (Rao and Wagenet,
1985). Natural, spatial variations in meteorologic conditions, primarily
rainfall, are subsequently superimposed as additional spatial variability.
For waste disposal systems, barriers to water and chemical movement, such
as clay and synthetic liners, are deliberately constructed to contain the
movement of the waste and prevent external contamination. Such barriers usually
have relatively uniform properties that will likely be much different from the
native soil (e.g., hydraulic conductivity). Appendix G provides a detailed
discussion of the spatial variability of soil properties including current
research and implications for field data collection and interpretation.
Temporal variations are also comprised of both intrinsic and extrinsic
components. The major intrinsic factor is the natural variation of rainfall;
although rainfall is also spatially variable (as noted above), its temporal
variation is the primary driving force behind the dynamic nature of the soil
environment and the resulting chemical movement. Other meteorologic inputs,
such as temperature, solar radiation, wind, etc., are also highly time-variable
and have significant impacts on chemical processes through their interdepend-
ent environmental conditions.
1-7
-------
All of the soil chemical fate and transport processes discussed above
demonstrate significant kinetic, or time-variable, behavior, often as a direct
result of the time-variable inputs. The transport and sorption processes are
thought to be relatively fast; they are initiated and occur during and imme-
diately following water applications. The fate processes of transformation,
degradation, and volatilization occur generally at a slower rate, but on a more
continual basis. Thus, for agricultural chemical applications of nonpersistent
chemicals, the time between chemical application and the first few rainfall
events is a major determinant of the amount of chemical runoff and leaching.
The longer this time period, the greater the opportunity for the chemical to
dissipate through various pathways; thus, less of the chemical is available to
runoff and leach. Chapter 2 further describes the kinetic nature of soil
processes affecting chemical fate and migration.
Plant processes demonstrate a definite temporal pattern related to the
natural crop growth cycle of planting, emergence, development, maturity, and
harvest. These phases are in turn affected by the temporal variation of envi-
ronmental conditions which jointly influence the chemical uptake, transforma-
tion, and fate processes within the plant.
The primary extrinsic factors that affect temporal variations in soil
processes are related to the nature, timing, and frequency of management prac-
tices and operations. These activities include water and chemical applica-
tions, and disturbances to the soil as tillage, clear-cutting, or landfill ing
operations. The timing of water and chemical applications relative to evapo-
transpiration rates and the occurrence of natural rainfall events impact the
movement of both water and chemicals. Soil disturbances cause changes to basic
soil properties that affect runoff and infiltration which in turn affect asso-
ciated chemical losses.
MODELING SOIL SYSTEM PROCESSES
Models, as used in this report, are simply representations or approxima-
tions of terrestrial environmental systems. They are not exact. They are
simplifications of a real system since no model can realistically represent in
detail the intricate workings and processes of real terrestrial systems describ-
ed above. Models are basically "cartoons" of reality, as illustrated by the
schematic representation in Figure 1-1; they attempt to represent the essential
characteristics and behavior of a real system.
In general, models are used to analyze system behavior under both current
(or past) conditions and anticipated (or future) conditions. Modeling requires
some basic knowledge of the system being analyzed; however, it also promotes an
improved understanding of the system through sensitivity analyses of system
characteristics and observations of the resulting system response as predicted
by the model and characterized by field data. This research-type function is
especially important for soil leaching models to help expand our current knowl-
edge of the complex terrestrial environment.
The most critical and cost-effective use of models is in the analysis of
proposed or alternative future conditions. That is, the model 1s used as a
management or decision making tool to help answer the "what if" type questions,
1-8
-------
such as "What if the application rate, timing, or method were changed?" or
"What if management practices, remedial actions, or control measures are insti-
tuted?" Attempting to answer such questions through data collection programs
would be enormously expensive and practically impossible in many situations.
Most mathematical models of environmental systems can be referred to as
"linked-process models" since they include mathematical representations of
individual processes and their interactions (or linkage) in order to approxi-
mate an entire environmental system. Figure 1-2 shows the hierarchy of model
components. The model is comprised of linked or integrated processes which are
mechanisms known to occur within the environmental system being modeled. An
individual process is represented by an algorithm (or series of algorithms)
which are in turn comprised of mathematical equations or expressions that
combine to represent the process. Finally, the equations and expressions
include variables and parameters. A variable is a time-dependent storage or
flux in an algorithm; it can be an environmental variable such as an input
meteorological condition (e.g., precipitation, air temperature, solar radia-
tion) or an output variable which describes the behavior of the system (e.g.,
flow, solute concentration, chemical storage). A parameter is a constant in an
equation that describes a characteristic or property of the system (e.g.,
saturated hydraulic conductivity, sorption coefficient, hydrolysis rate)
(Donigian and Dean, 1985).
The American Society for Testing and Materials (ASTM) has prepared a
Standard protocol for evaluating environmental chemical fate models, including
the definition of selected modeling terms shown in Table 1-1 (ASTM, 1984).
These terms are commonly used in the modeling literature although often with
different meanings; the ASTM definitions in Table 1-1 may help to improve
communications among modelers and with other professionals by providing more
order to the technical language.
Model Classifications
Models to represent the fate of chemicals in the soil environment have
proliferated during the past decade, primarily as a result of increased atten-
tion and concern for protecting ground-water supplies from contamination by
hazardous wastes, pesticides, and other toxic chemicals. Unfortunately, a
variety of often confusing terms has been used to categorize and differen-
tiate the capabilities and characteristics of the many models available. Terms
such as deterministic, empirical, stochastic, analytic, semi-analytic, numer-
ical, mechanistic, equilibrium, kinetic, etc. have been used sometimes differ-
ently by different authors in attempts to classify models. Moreover, the
models have been developed by scientists and engineers with different back-
grounds, perspectives, and for different purposes, (e.g., research versus
regulatory or management), and often can be described by a number of the terms
listed above. For example, a single model may take a deterministic approach
to process representation, use a numerical procedure for solving the governing
transport equation, assume equilibrium sorption, and describe biodegradation
with first-order kinetics. Thus, any model which is an assemblage of interac-
ting processes may include different approaches to best represent a specific
process.
1-9
-------
Model
A
Processes
4
Algorithms
A
Equations / Expressions
Variables
Parameters
Figure 1-2. Hierarchy of model components.
1-10
-------
TABLE 1-1. DEFINITION OF TERMS USED IN ASTM STANDARD PRACTICE FOR
EVALUATING ENVIRONMENTAL FATE MODELS OF CHEMICALS
Algorithm -
Calibration -
the numerical technique embodied in the computer code.
a test of a model with known input and output informa-
tion that is used to adjust or estimate factors for
which data are not available.
Compartmentalization - division of the environment into discrete locations in
time or space.
Computer Code
(computer program)
Model -
Sensitivity -
Validation -
Verification -
Source: ASTM, 1984.
the assembly of numerical techniques, bookkeeping, and
control language that represents the model from accept-
ance of input data and instructions to delivery of
output.
an assembly of concepts in the form of a mathematical
equation that portrays understanding of a natural
phenomenon.
the degree to which the model result is affected by
changes in a selected input parameter.
comparison of model results with numerical data inde-
pendently derived from experiments or observations of
the environment.
examination of the numerical technique in the computer
code to ascertain that it truly represents the concep-
tual model and that there are no inherent numerical
problems with obtaining a solution.
The primary differences between models are the level of detail at which
the fundamental soil processes discussed in Chapter 2 are treated, and the
conceptual completeness in terms of the number of processes of concern included
in the model (Rao et al., 1982). Inherent differences in models are also re-
lated to the manner in which temporal and spatial variations are or are not
considered. Rao et al. (1982) suggest that these primary differences in model
characteristics are determined in part by the following factors:
a. Current state of understanding of the system and its components
(subsystems) to be modeled.
b. The modeler's conceptualization of the system processes.
1-11
-------
c. The modeling approach and error bounds allowable In approximations
required to solve the problem.
d. The spatial and temporal scales of Intended model applications.
Although the terms used in classifying models leave much to be desired,
some discussion of the major model types is needed to provide a framework for
relating the models described in this report to the universe of models and
modeling approaches. One such classification proposed by Addiscott and Wagenet
(1985) is shown in Table 1-2 and includes references to models represented by
each category. The two major groups in this classification are deterministic
and stochastic with the differences described by Addiscott and Wagenet as
follows:
A key distinction in comparisons between models is that between deter-
ministic models, which presume that a system or process operates such that
the occurrence of a given set of events leads to a uniquely-definable
outcome, and stochastic models, which presuppose the outcome to be un-
certain and are structured to account for this uncertainty.
The vast majority of soil chemical fate models are deterministic since
they attempt to represent the primary physical, chemical, and biological proc-
esses that determine chemical fate and movement in soil systems. In fact, most
of the so-called stochastic models are not completely stochastic in that they
often utilize a deterministic representation of soil processes and derive their
stochastic nature from their representation of inputs and/or spatial variation
of soil characteristics and resulting chemical movement. While the determinis-
tic approach results in a specific value of a soil variable (e.g., solute con-
centration) at any point in a field, the stochastic approach provides the
probability (within a desired level of confidence) of a specific value occur-
ring at any point (Rao and Wagenet, 1985).
Development of stochastic models is a relatively recent endeavor that has
occurred as a result of the growing awareness of the importance of intrinsic
and extrinsic variability of the soil environment. Consequently, stochastic
models are still primarily research tools, although they may hold promise for
management purposes in the future (Addiscott and Wagenet, 1985).
Attempts to validate terrestrial leaching models must address the issue of
spatial variability when comparing model predictions with limited field ob-
servations. If sufficient field data are obtained to derive the probability
distribution of pesticide concentrations, the results of the stochastic model
can be compared directly. For a deterministic model, the traditional approach
has been to vary the input data within its expected range of variability (or
uncertainty) and determine whether the model results fall within the bounds of
field measured values. Given the lack of comprehensive field data sets that
adequately characterize spatial variability, the absence of widely accepted
statistical methods for judging the goodness-of-fit models to measured data,
and our inability to directly measure water and solute fluxes which are more
logical variables for model validation, it has been suggested that complete
validation of any simulation model may not be possible (Rao and Wagenet, 1985).
1-12
-------
TABLE 1-2. A CLASSIFICATION OF LEACHING MODELS
333S3SS3333=33S&3333333=33333333SSS3S333333SS333=====S=====:S==333333=3333=====
I. Deterministic models
A. Mechanistic (usually based on rate parameters)
1. Analytical* (e.g., Neilsen and Biggar, 1962; van Genuchten and
Wierenga, 1976).
2. Numerical* (e.g., Childs and Hanks, 1975; Robbins et al., 1980).
B. Functional (usually based on capacity parameters)
1. Partially analytical (e.g., De Smedt and Wierenga, 1978; Rose et
al., 1982a,b).
2. Layer and other simple (e.g., Bresler, 1967; Tanji et al., 1972;
Burns, 1974; Addiscott, 1977).
II. Stochastic Models
A. Mechanistic (e.g., Dagan and Bresler, 1979; Amoozegar-Fard et al.,
1982).
B. Non-mechanistic (transfer function) (Jury, 1982; Jury et al., 1982).
*Refers to the solution of the flow equations.
Source: Addiscott and Wagenet, 1985.
Such complete validation may not be necessary for certain model cases if model
limitations are adequately recognized.
The second level of classification in Table 1-2 is mechanistic vs. func-
tional for deterministic models and mechanistic vs. non-mechanistic for sto-
chastic models. The distinction between mechanistic and functional is really a
matter of the degree of process representation; mechanistic models incorporate
the most fundamental understanding of soil processes, such as the use of
Richard's equation (see Appendix A) to define soil water movement, whereas
functional models use more simplified representations, such as field capacity
concepts for soil moisture retention. Functional models are largely management-
oriented while mechanistic models are primarily research-oriented.
The stochastic-mechanistic models as noted above combine elements of both
approaches. The stochastic non-mechanistic models are purely statistical and
must rely on some degree of calibration for evaluation of selected model param-
eters. Although still developmental, this approach may hold promise in the
1-13
-------
future for management tools with consideration of variability and uncertainty
(Addiscott and Wagenet, 1985).
Another distinction between model types listed in Table 1-2 is analytical
versus numerical deterministic models. Although Table 1-2 lists this under
mechanistic-deterministic models, it really applies to the entire category of
deterministic models. Analytical models are based on a direct solution to the
governing convective-dispersive solute transport equation (see Appendix A)
under defined initial and boundary conditions and on other assumptions. Certain
conditions and assumptions are required in order to obtain a direct solution to
the equation. The most critical of these assumptions for analytical soil
leaching models are as follows:
1. Constant, uniform soil-water content and soil-water flux throughout
the soil profile (steady water flow).
2. Uniform, homogeneous soil properties.
3. Constant, pulse, or exponentially decaying chemical boundary condi-
tions at the soil surface.
In spite of these simplifying restrictions, analytical models have been
used extensively in a variety of environmental applications. Van Genuchten and
Alves (1982) have cited a number of example applications of analytical models
in soil science, chemical and environmental engineering, and water resources
and have provided a useful compendium of solutions under various combinations
of initial and boundary conditions.
Numerical models are designed to provide the flexibility to analyze dynam-
ically changing and heterogeneous soil conditions that are often found in field
applications. They utilize numerical solution techniques (e.g., finite differ-
ence, finite element) to solve the governing solute transport equation. Thus,
numerical models are designed to circumvent the restrictive assumptions of the
analytical models.
Review and Comparison of Selected Models
The primary models considered in this report are the following three
models developed or sponsored by various EPA offices: PESTAN, SESOIL, and
PRZM. Each of these models is described below and is followed by a general
comparison with other available models.
PESTAN
PESTAN is an interactive program developed by the EPA R. S. Kerr Environ-
mental Research Laboratory in Ada, Oklahoma, for estimating the movement of
organic chemicals through soil to ground water. The theory on which PESTAN is
based is described by Enfield et al. (1982), who discuss three simple models
derived as primarily analytical solutions to the convective-dispersive solute
transport equation. In addition to the vertical (i.e., one-dimensional)
1-14
-------
convective movement of a chemical, these three models Include the following
specific combinations of processes:
1. Linear sorption and first-order decay (transformation), no dispersion
2. Dispersion and linear sorption, no decay (transformation)
3. Non-linear (Freundlich) sorption and first-order decay (transformation)
no dispersion
Models 1 and 2 are complete analytical solutions similar to those tabu-
lated by van Genuchten and Alves (1982), while model 3 uses the method of
characteristics to obtain a numerical solution to the governing equation due
to the use of a non-linear isotherm. Note that none of the three models account
for vapor-phase partitioning and transport. Thus, PESTAN's use is limited to
nonvolatile chemicals.
The PESTAN program is the computer implementation of model 2 described by
Enfield et al. (1982), with the additional capability to consider first-order
decay processes within an interactive framework that simplifies model use and
application. The primary output of the model is the chemical concentration
distribution within the soil profile for any user-specified time after appli-
cation. The program is easy to use, requires evaluation of only 10 input
parameters, and can be run interactively from a computer terminal without the
need for a user manual. PESTAN has been used extensively by the EPA Office of
Pesticide Programs (OPP) for initial screening assessments to evaluate the
potential for ground-water contamination of pesticides submitted for registra-
tion and for currently registered compounds (M. Lorber, 1985, personal communi-
cation). It has also been tested against field data by the authors (Enfield et
al., 1982) and by other investigators (Jones and Back, 1984; Jones et al.,
1983; Melancon et al., 1986).
In spite of its simplicity, PESTAN can be a useful tool for preliminary
assessments of the type performed by OPP as long as the user is fully aware of
its assumptions and limitations. The primary limitations of PESTAN are direct-
ly related to its analytical nature which requires the assumption of one-
dimensional steady water flow in a homogeneous soil profile with constant
hydraulic, sorption, and decay parameters. TfTus, the temporal variability of
soil processes, especially leaching, due to the relative timing and intensity
of water and chemical application rates, and the spatial variability (primarily
vertical heterogeneities such as layering) of soil characteristics are largely
ignored. The combined result of these assumptions can lead to underestimation
of subsurface chemical concentrations and the resulting contaminant input to
ground water (Jones and Back, 1984; Jones et al., 1983; M. Lorber, 1985, per-
sonal communication). PESTAN may be most appropriate for the intermediate
unsaturated zone between the bottom of the root zone and ground water because
the steady flow assumption is most applicable in this region. However, it
should be noted that decay of the contaminant within the vadose zone is still
not accounted for in Model 2.
PESTAN users should perform sensitivity analysis on key parameters, such
as water recharge rate, decay rate, and sorption coefficient, in order to
1-15
-------
assess the potential Impact of significant variations in these parameters under
field conditions. The water recharge rate is a key parameter determining the
rate of chemical movement to ground water. Simple water balance (i.e., annual
precipitation and irrigation minus evaporation) calculations may significantly
underestimate this parameter (Jones et al., 1983). Also, users should be
careful in evaluating the input decay rate required by PESTAN; it is defined on
the basis of the sorbed chemical concentration as opposed to the total concen-
tration. Thus, total soil persistence or attenuation rates must be appropri-
ately adjusted prior to input (M. Lorber, 1985, personal communication).
SESOIL
The Seasonal SOIL Compartment (SESOIL) model was developed for EPA Office
of Toxic Substances (OTS) in response to the need for a model for long-term
simulation of chemical fate in the soil environment. The current version of
SESOIL is designed for "...environmental process simulations that can describe
simultaneously water transport, sediment transport, and pollutant fate (trans-
port/transformation)...in an unsaturated soil compartment" (Bonazountas and
Wagner, 1984). A unique feature of SESOIL is that the hydrology is based on a
statistical representation of the water balance components over a "season"
rather than accounting for changes in soil moisture through time.
According to Bonazountas and Wagner (1982), the intended uses or applica-
tions of SESOIL include "...long-term leaching studies from waste disposal
sites, acid rain, pesticide and sediment transport on watersheds, contaminant
exposure, pre-calibration runs for other simulation models, hydrologic cycles
and water balances of soil compartments, overall chemical fate assessments,
exposure, risk and other studies."
SESOIL is a seasonal model in that it calculates the pollutant distribu-
tion in the soil column and on the watershed at the end of a "season" (e.g.,
year or month). The model is compartmental in that the soil column can be
subdivided into at least four layers or compartments with each having uniform
properties. Although each layer (compartment) is assumed to have uniform
properties, the depth-varying nature of soil properties can be approximated by
assigning an average or weighted value to each layer.
SESOIL is structured for the integrated simulation of three cycles: the
hydrologic cycle, the sediment cycle, and the pollutant cycle. Each cycle
encompasses a multitude of processes as shown in Table 1-3; some of these
processes are not operational in the latest version of SESOIL.
The hydrologic cycle in SESOIL is based on a statistical dynamic formu-
lation of the vertical water budget at the land-atmosphere interface (Eagelson,
1978). Physically based dynamic and conservation equations express the infil-
tration, exfiltration, transpiration, percolation to ground water, and the
capillary rise from the water table during and between storms in terms of
independent variables of precipitation, potential evapotranspiration, soil
properties, and water table elevations. Uncertainty of the hydrologic cycle
simulation is introduced into these equations via the probability density
functions of the independent climatic variables and produces derived probability
1-16
-------
TABLE 1-3. SIMULATION CAPABILITIES OF SESOIL CYCLES
ssssss==2=ss=======sssss=sss==========ss=s=====s======ss====sa===r===s
Hydrology Sediment Pollutant
precipitation resuspension advection
by Mind
infiltration/ washload diffusion
exfiltration
soil moisture volatilization
evapotranspiration sorption
capillary rise degradation
ground-water runoff hydrolysis
surface runoff oxidation
snow pack/melt* cation exchange
interception* complexation chemistry
photolysis*
nutrient cycles*
biological uptake*
*Not operational in current version of SESOIL.
distributions of the dependent water balance elements: surface runoff, evapo-
transpi ration, and ground-water runoff. The mean value of these quantities
give a long-term (season) average water balance.
The sediment cycle includes simulation of both sediment washload due to
precipitation runoff and sediment (dust) resuspension due to wind. Options for
annual and monthly simulation of sediment washload are included. The annual
option is based on the Universal Soil Loss Equation (Wischmeier and Smith,
1978) while the monthly option uses modifications of procedures developed by
Foster et al. (1980), for CREAMS model (Knisel, 1980). Resuspension losses from
the surface compartment due to physical removal of particles and associated
pollutants as a function of particle characteristics (chemical composition,
diameter, etc.) and weather/soil conditions (e.g., wind speed, soil moisture)
are also considered.
1-17
-------
The pollutant cycle allows for simulation of the 12 chemical processes
noted in Table 1-3; some alternate simulation options are available for selec-
ted processes. Diffusion is described by Pick's first law to estimate upward
solute flow in the soil column to the air as volatilization. Volatilization
considers pollutants buried under a layer of clean soil, assumes Pick's first
law, and assumes vapor phase diffusion is rate controlling. Adsorption/desorp-
tion is represented by the Freundlich isotherm, and degradation is assumed to
follow first-order kinetics. Neutral and acid or base catalyzed hydrolysis is
also described as first-order decay. Cation exchange is assumed to be an
irreversible, instantaneous process which immobilizes pollutant up to the soil
cation exchange capacity.' Consideration is also given to complexation of metal
ion in solution with organic ligand(s).
SESOIL has undergone testing by a variety of organizations in addition to
that done by its developers. These efforts have included sensitivity analysis,
comparison with other models, and some limited comparisons with field data.
Arthur 0. Little, Inc., the developers of SESOIL, compared model predictions
with field data for selected metals at a site in Kansas, and for two organic
compounds (i.e., naphthalene and anthracene) at a site in Montana (Bonazountas
et a!., 1982). Sensitivity analyses on the volatilization and adsorption
routines were performed by Wagner et al. (1983), for six pollutants in three
climates with four different soil types. Hetrick (1984) compared SESOIL hydro-
logic predictions with some calibration to analogous results from a more
detailed hydro!ogic watershed model, AGTEHM (Hetrick et al., 1982), and to
observations from a watershed in Tennessee and South Carolina. He concluded
that SESOIL hydrologic predictions were in good agreement with both AGTEHM
predictions and observed data but that the month-to-month deviations indicated
that SESOIL should not be recommended for short-term predictions (Hetrick,
1984).
SESOIL was incorporated as the soil/land component of the screening level
multimedia model, TOX-SCREEN, (Hetrick and McDonald-Boyer, 1984) developed by
Oak Ridge National Laboratories for the EPA Office of Toxic Substances. In a
separate model evaluation study, Bicknell et al. (1984) attempted to test
TOX-SCREEN with multimedia data on benzo(a)pyrene from Southeastern Ohio.
However, no definitive conclusions on the performance of SESOIL could be made
because model predictions could not be directly compared with measured soil
concentrations.
Battelle Pacific Northwest Laboratories recently completed a model com-
parison study that included SESOIL along with a number of flow, transport, and
geochemical models for soil and ground-water systems (Kincaid et al., 19845).
SESOIL has also been compared in a laboratory column study with PESTAN and PRZM
(Melancon et al., 1986). A landfill test problem was designed for comparing
flow predictions for one-dimensional unsaturated zone models. This latter
study concluded that SESOIL was not useful for analyzing sites with a high
degree of vertical variation in soil properties (i.e., layering) because SESOIL
characterizes the soil column as homogeneous and isotropic for water balance
calculations.
1-18
-------
The most recent and comprehensive evaluation of SESOIL was performed by
Watson and Brown (1984). The primary conclusions derived from that study are
as follows:
1. The basic framework for a useful, screening-level, chemical migration
and fate model exists in SESOIL. However, there are several errors in
the pollutant cycle (i.e., relative to neutral hydrolysis and chemical
leaching) that require correction, and several modifications should be
considered to improve the representation of chemical leaching and
washoff from the land surface.
2. SESOIL can be applied to generic environmental conditions if model
limitations are observed and judgment is used in the estimation of
model parameters. As with most models, SESOIL cannot be applied on a
site-specific basis with only "limited calibration."
3. SESOIL cannot be applied to sites exhibiting large vertical variations
in soil properties. Although up to four soil layers can be used for
the pollutant cycles, a single homogeneous soil column is used in the
hydrologic cycle.
4. Consideration should be given to modifying the algorithm used to
determine when chemical will begin to leach to ground water. This
algorithm is based only on the estimated travel time for water; the
sorptlve properties of the chemical are not considered.
5. Although the SESOIL documentation report/user's manual provides an
extensive review of the literature in many areas, it is not explicit in
terms of the specific processes that are considered, the theory that
is incorporated into the model, and parameter estimation procedures.
6. There are limitations associated with using SESOIL as a watershed
model, with applying SESOIL to certain disposal facilities (e.g.,
landfills), and with applications of SESOIL to certain release sce-
narios (e.g., spills, buried drums and tank leaks).
7. Although the newest version of SESOIL (dated May, 1984) is an improve-
ment over the version tested by Watson and Brown (1984), the improve-
ments did not significantly change the testing conclusions.
SESOIL continues to be used and supported by the Office of Toxic Sub-
stances (OTS). Code corrections and enhancements are currently being performed
by Oak Ridge National Laboratory as part of its continuing work on the TOX-
SCREEN model for OTS (A. Nold, 1985, personal communication). New users of
SESOIL should review the testing report by Watson and Brown (1984) to get
further details on the model limitations noted above. A new, non-EPA version
of SESOIL is currently under development at the Center for Environmental Manage-
ment of the University of Athens in Greece. This new version, expected to be
released in late 1985, will be based on a multiple layering scheme for repre-
senting the soil column, will include entirely new procedures for the water
1-19
-------
balance calculations, and will improve the sediment washload and evapotranspi-
ration routines (M. Bonazountas, 1985, personal communication).
PRZM
The Pesticide Root Zone Model (PRZM), developed at the U.S. EPA Environ-
mental Research Laboratory in Athens, Georgia, (Carsel et al., 1984) is a
dynamic, compartmental model for use in simulating chemical movement within and
below the plant root zone. Time varying transport including advection and
dispersion are represented in the program.
PRZM has two major components: hydrology and chemical transport. The
hydrology component for calculating runoff and erosion is based on the Soil
Conservation Service curve number technique and the Universal Soil Loss Equa-
tion (Wischmeier and Smith, 1978). Evapotranspiration is estimated from pan
evaporation or by an empirical formula if input pan data are unavailable.
Evapotranspiration is comprised of evaporation from plant interception, evap-
oration from soil, and transpiration from the crop. Water movement is simu-
lated by the use of an empirical model based on soil-water capacity terms
including field capacity, wilting point, and saturation. To produce soil water
and solid phase concentrations, the chemical transport component calculates
pesticide uptake by plants, surface runoff, erosion, decay/transformation,
vertical movement (leaching), foliar loss, dispersion, and retardation.
PRZM allows the user to perform dynamic simulations of potentially toxic
organic chemicals, particularly pesticides, that are applied to the soil or to
plant foliage. Dynamic simulations allow the consideration of pulse loads, the
prediction of peak events, and the estimation of time-varying mass emission
or concentration profiles; thus dynamic simulations overcome limitations of the
more commonly used steady-state models.
To apply PRZM, the soil profile is divided into a number of soil layers or
compartments. For each compartment, the model solves the solute transport
equation (see Chapter 2) including advection, dispersion, adsorption, degrada-
tion, and plant uptake of the chemical. For the surface compartment, addi-
tional terms are included in the equation to allow for chemical losses associ-
ated with surface runoff and erosion. A numerical finite-difference solution
procedure is used to solve the system of equations under appropriate boundary
conditions.
Currently in PRZM, model soil parameters such as the sorption coeffient
and degradation rate coefficient can be specified separately for each user-
defined soil zone, e.g., surface zone, root zone, and below root zone. Each
zone is then divided into uniform layers or compartments. As many soil layers
and compartments can be utilized as are necessary to accurately represent the
characteristics of the soil profile.
PRZM uses the Soil Conservation Service runoff curve number technique to
distribute the daily rainfall into runoff and infiltration components. Because
of this, the timestep for the model is one day. Infiltrating water is assumed
to cascade downward to successively deeper layers as the soil water content of
1-20
-------
each compartment reaches and exceeds field capacity. Field capacity is usually
reported as the soil-water content that field soils attain after all excess
water is drained by gravity. For loose sandy soils, the water in excess of
field capacity is assumed to drain within the one day time-step of the models.
For tight clay soils and profiles with restrictive, low permeability layers, a
drainage rate parameter can be specified by the user to allow excess drainage
to occur over a longer time period (e.g., weeks or months). Using these pro-
cedures, the pore-water velocity is estimated and is not directly based on soil
hydraulic conductivity; field capacity and wilting point values are used
operationally to determine percolation and soil-water content in a compartment
based on the infiltrating water. Wilting point is defined as the soil-water
content below which plants are unable to extract water.
Evapotranspiration is calculated as a function of soil-water conditions
and total potential evapotranspiration. It is either input by the user (and
usually estimated from local pan evaporation data) or calculated from air
temperature and the number of daylight hours.
The daily evapotranspiration demand is divided among evaporation from
canopy, soil evaporation, and crop transpiration. Total demand is first
estimated and then extracted sequentially from crop canopy storage and from
each layer until wilting point is reached in each layer or until total demand
is met. Evaporation occurs down to a user-specified depth. The remaining
demand (crop transpiration) is met from the layers between this depth and
the active rooting depth. The root zone growth function is activated at crop
emergence and increases step-wise until maximum rooting depth is achieved at
crop maturity.
Snowmelt calculations are included in PRZM based on the "degree-day"
approach utilizing average daily air temperature. Runoff is then calculated
using both rainfall and snowmelt as inputs to the curve number procedure.
Based on the calculated daily runoff, PRZM estimates daily storm event erosion
using the Modified Universal Soil Loss Equation (MUSLE) developed by Williams
(1975). The peak storm runoff rate required by MUSLE is estimated by assuming
a trapezoidal hydrograph and a mean storm duration input by the user.
Degradation of the chemical is represented as a single first-order process
with the rate coefficient specified by the user for each defined soil zone.
Thus, representation includes all significant biochemical transformation and
decay processes that reduce the amount of chemical in the soil. Different
rates are commonly used for the surface zone, root zone, and below root zone
(partially) to account for the different transformation/degradation processes
occurring in the various portions of the profile. Plant uptake is represented
as a separate process and is calculated as a function of the amount of tran-
spiration in each compartment and as an uptake efficiency factor.
Like SESOIL and PESTAN, PRZM does not account for vapor-phase partition-
ing and transport (via vapor diffusion). Thus, all three models cannot be
used to evaluate data for volatile chemicals (e.g., EDB, DBCP, TCP, etc.).
PRZM has undergone a moderate degree of performance testing in comparison
with field data at selected sites across the country and has been used in
1-21
-------
exposure assessments and other applications. Because of Its agricultural
emphasis and the more general availability of field pesticide data, the primary
testing studies listed below are for pesticide leaching.
a. Aldicarb applied to citrus in Florida (Jones et al., 1983)
b. Aldicarb applied to potatoes in New York (Carsel et al., 1985a) and
Wisconsin (Jones, 1983)
c. Metalaxyl applied to tobacco in Florida and Maryland (Carsel et al.,
1985b)
d. Atrazine and chloride applied to corn in Georgia (Carsel et al.,
1985a)
The results of these testing efforts by both EPA (i.e., the model devel-
opers) and other investigators have been consistently positive when comparing
field average values to PRZM predictions of soil profile concentrations and
mass flux to ground water. The spatial variability problems noted earlier
preclude effective comparison between PRZM results and concentrations from
individual points within the field. Considering the relatively simplistic
nature of PRZM relative to the dynamic soil environment, the testing results
to date demonstrate that PRZM is effectively representing the primary pro-
cesses controlling pesticide movement to ground water (Carsel et al., 1985b).
The PRZM model has been requested by and distributed to over 150 users
nationwide and is currently being applied in a variety of locations. The EPA
Environmental Research Laboratory in Athens, Georgia, is heading a cooperative
research project with the U.S. Geological Survey to develop field data for
further testing and refinement of PRZM and for investigating the spatial vari-
ability of pesticide leaching in field soils (Carsel et al., 1985b; see Chapter
4 for a description of this project).
Since its initial release in "draft" form in 1982, PRZM has been used in
a wide range of studies. Both the EPA Office of Pesticide Programs (OPP) and
various chemical manufacturers have used PRZM to assess the leaching potential
of new and currently registered compounds (R. Carsel, 1984, personal communica-
tion). Dean et al. (1984b), used PRZM to develop a methodology allowing expedi-
tious use by OPP to screen and assess pesticides for potential contamination of
ground water on a national scale; running PRZM for 25 years at selected sites
across the country, probability distributions of pesticide leaching below the
root zone were developed as a function of pesticide characteristics and cropping
practices.
PRZM has also been used as a framework for a terrestrial ecosystem model
for pesticide exposure to wildfowl (Dean et al., 1984a), has been linked to
ground water, surface water, and air models for a multimedia study of the
effects of conservation tillage on pesticide concentrations (Donigian and
Carsel, 1985), and has provided the key element of a PCB spill exposure assess-
ment methodology (Brown and Boutwell, 1985).
1-22
-------
In conjunction with detailed multidimensional ground-water models, PRZM
was used as part of an exposure assessment for aldicarb application to citrus
In Florida (Dean and Atwood, 1985) and as part of an analysis of the effects
of facility design and locational factors on leachate migration from hazardous
waste sites (Donigian et al., 1983).
Model Comparisons
In order to provide background on the general types of models available,
Table 1-4 summarizes and compares the primary capabilities and characteristics
of PESTAN, SESOIL, and PRZM along with a few additional models of flow and
contaminant movement in the unsaturated zone.
CREAMS2 {Leonard and Ferreira, 1984) is a recent modification and improve-
ment of the USDA CREAMS model (Knisel, 1980) originally developed for analysis
of agricultural runoff. CREAMS2 includes capabilities to model chemical move-
ment through the crop root zone using a layered, compartmental approach (Leonard
and Knisel, 1984).
The HELP (Hydrologic Evaluation of Landfill Performance) model was devel-
oped for use by permit writers and landfill designers to assess water movement
across, into, through, and out of landfills with rapidity (Schroeder et al.,
1984). It was recently modified to consider contaminant movement by volatil-
ization and percolation of the leachate through the landfill (Bicknell, 1984).
The remaining models listed in Table 1-4 are representative of a large
group of detailed numerical models that have been developed for variably satu-
rated conditions (i.e., unsaturated and saturated) and for multi-dimensional
contaminant migration problems. This allows a more detailed representation
of the dynamically changing interface between the unsaturated zone and ground
water and requires more detailed information on the spatial variation of
soil properties for the specific system being analyzed. These models generally
include linear or Freundlich sorption isotherms, lumped first-order degrada-
tion, and simplified surface conditions for rainfall and evapotranspiration;
runoff, erosion, and plant processes are usually ignored.
The last five models listed in Table 1-4 were selected for further study
in a recent review of 55 flow and transport models for the unsaturated zone by
Oster (1982). Only these five models were selected due to their consideration
of both flow and transport, availability of code, documentation, and demon-
strated application. Battelle, Pacific Northwest Laboratories recently com-
pleted an extensive evaluation of hydrogeochemical models for solute migra-
tion in both the unsaturated and saturated zones (Kincaid et al., 1984a; 1984b);
the five models noted above, plus SESOIL and CREAMS, were included in this
evaluation. Interested readers are referred to these two reviews for more de-
tailed information on the extent of available soil leaching models.
Criteria for Model Selection and Use
As noted above, many models in addition to the 11 models listed in Table
1-4 are currently available for simulating organic pollutant fate and transport
in the soil environment. The level of detail and the conceptual completeness of
1-23
-------
TABLE 1-4. COMPARISON OF SELECTED SOIL LEACHING MODELS
TRANSPORT
TRANSFORMATION /
DEGRADATION /
PLANT CHEMICAL
PROCESSES
PRIMARY MODELS
PESTAN
SESOIL
PRZM
OTHER AVAILABLE MODELS
CREAMS 2
HELP/HELP-WQ
SEGOL
SUMATRA-1
TARGET
FEMWATER/FEMWASTE
TRUST/ML TRAN
*
*
*
*
*
*
*
*
*
•
*
*
*
*
*
*
*
•
*
*
w
*
*
•
•
*
*
*
*
*
V
*
*
*
*
*
*
*
*
a
*
•
*
*
*
*
a
1
1
1 c
1
1
2,3
1
1,2,3
2
1.2
1
4
user
leflned
7
9
-
.
-
-
-
-
monthly
dally
daily
A
N
N
N
variable N
variable N
variable N
variable N
variable N
I
r\»
a - These processes were Included In an earlier experimental version of PRZM (Deam et al, 19B4a).
-------
the available models vary considerably. As evident from the earlier sections
of this chapter, these models are generally quite complex and Include many
processes and parameters. In selecting a model, a prospective user needs first
to thoroughly understand the simplifications and assumptions made In the devel-
opment of the model. Without sufficient documentation of the model, reports of
adequate testing by Independent workers, and a certain amount of hands-on
experience, most prospective users and even many modelers would have consider-
able difficulty in appreciating all the intricacies and idiosyncracies of a
model or of the computer code itself. This makes it difficult for a user to
select among the models for his application. The following discussion, based
primarily on a paper by Rao et al. (1982), attempts to provide general guide-
lines for selection and use of simulation models. It is not intended as a set
of steps to follow, but it does provide some of the factors to be considered by
the user in model selection.
Rao et al. (1982), suggest that the following evaluation factors be em-
ployed by a prospective user:
1. The intended use and the spatial/temporal scales at which the model
simulations are to be used.
2. The availability of computational facilities (i.e., computer hardware,
computer programmers, etc.) required to implement the computer code.
3. The availability and the reliability of the required model input data.
4. The confidence region(s) associated with the model output.
Two model attributes -- the type and scale of model output -- determine
whether a particular model may be used without any modifications for a specific
application. Each model and its computer code have a characteristic scale at
which the model may be used. For example, a model designed for management
purposes (say at a river basin scale) will most likely not be useful for de-
tailed descriptions of pollutant fate at a small experimental field site. A
model designed for research applications can be used for simulation at larger
temporal and spatial scales. However, the prohibitive costs, the paucity of
data for a large number of parameters in the model, and the spatial /temporal
variability constraints would make such an application impractical if not
impossible.
In many cases, the values of several model parameters are "adjusted" or
"calibrated" when independent estimates are unavailable. It must be recognized
that the number of parameters that are "adjusted" increases, and so may the
uncertainty in their values. Whether the calibrated values of model parameters
are appropriate for another site, season, or scale needs to be evaluated
carefully.
Uncertainties in the model input data provided by the user may propagate
through the various computations in a model and will be reflected in the model
outputs. Such uncertainties, however, may not always have amplified effects
(i.e., additive or multiplicative) on the model output data and may cancel each
1-25
-------
other out. With increasing complexity of the model, the user's ability to
identify and rectify such problems, when present, is reduced.
Spatial and temporal variations in various processes and environmental
factors also introduce uncertainties in measured data to be used for com-
parison with model output. Since well-established statistical (or other)
procedures are presently unavailable, comparison of measured data with model
predictions remains somewhat qualitative, and judgment of the goodness-of-
fit is, in most cases, arbitrary. This, in our view, is probably the most
vexing problem in model development, testing, and use.
All three models (PESTAN, SESOIL, and PRZM) selected for detailed discus-
sions here were either developed by or sponsored by the USEPA. These models
in general are wel1-documented and have been tested by a number of independent
users. These models have been and will continue to be used by the EPA, state
regulatory agencies, consulting firms, and industry. The discussion of the
various processes and how they may be incorporated in a model as well as
issues pertaining to parameter estimation and the design of field studies are
intended to be "generic" and generally applicable to other simulation models.
1-26
-------
REFERENCES
Addiscott, T. M. 1977. A simple computer model for leaching in structured
soils. J. Soil Science 28:554-563.
Addiscot, T. M., and R. J. Wagenet. 1985. Concepts of solute leaching in
soils: A review of modeling approaches. J. Soil Sci. 36:411-424.
Amoozegar-Fard, A., D. R. Nielsen, and A. W. Warrick. 1982. Soil solute
concentration distributions for spatially-varying pore water velocities
and apparent diffusion coefficients. Soil Sci. Soc. Amer. J_. 46:3-9.
ASTM. 1984. Standard Practice for Evaluating Environmental Fate Models of
Chemicals. E978, American Society for Testing and Materials.
Baker, J. L. 1980. Agricultural Areas as Nonpoint Sources of Pollution.
In: M. R. Overcash and J. M. Davidson (Eds.), Environmental Impact of
Nonpoint Source Pollution. Ann Arbor Science, Ann Arbor, MI.pp. 275-310.
Baker, D. B. 1983. Studies of Sediment, Nutrient and Pesticide Loading in
Selected Lake Erie and Lake Ontario Tributaries. Draft Final Report,
EPA Grant No. R005708-01.
Bicknell, B. R. 1984. Modeling Chemical Emissions from Lagoons and Landfills,
Final Report. U.S. Environmental Protection Agency, Athens, GA.
Bicknell, B. R., S. H. Boutwell, and D. B. Watson. 1984. Testing and Evalua-
tion of the TOX-SCREEN Model. Prepared by Anderson-Nichols and Co.,
Palo Alto, CA, for the U.S. Environmental Protection Agency, Athens, GA.
Bilkert, J. N., and P. S. C. Rao. 1985. Sorption and leaching of three non-
fumigant nematicides in soils. J_. Environ. Sci. Health B20(l):l-26.
Bonazountas, M., and J. Wagner. 1984. SESOIL: A Seasonal Soil Compartment
Model. Prepared by Arthur D. Little, Inc., Cambridge, MA, for the U.S.
Environmental Protection Agency, Office of Toxic Substances, Washington,
D.C.
Bonazountas, M., and J. Wagner. 1982. Pollutant Transport in Soils via "SESOIL."
Presented at ASCE National Conference at Environmental Engineering,
Minneapolis, MN. July 14-16, 1982.
1-27
-------
Bonazountas, M., J. Wagner, and B. Goodwin. 1982. Evaluation of Seasonal Soil/
Groundwater Pollutant Pathways. Prepared by Arthur D. Little, Inc.,
Cambridge, MA, for the U.S. Environmental Protection Agency, Office of
Water Regulation and Standards, Washington, DC.
Bresler, E. 1967. A model for tracing salt distribution in the soil profile
and estimating the efficient combination of water quality and quantity
under varying field conditions. Soil Sc_u Soc. Amer. 104:227-233.
Briggs, G. G., R. H. Bromilow, and A. A. Evans. 1982. Relationships between
lipophilicity and root uptake and translocation of non-ionized chemicals
by barley. Pest. Sci. 13:495-504.
Brown, S. M., and S. H. Boutwell. 1985. PCB Spill Exposure Assessment
Methodology. Prepared by Anderson-Nichols and Co., Inc., Palo Alto,
CA, for Electric Power Research Institute, Palo Alto, CA.
Burns, I. G. 1974. A model for predicting the redistribution of salts
applied to fallow soils after excess rainfall or evaporation. J.
Soil SCK 25:165-178.
Carsel, R. F., C. N. Smith, L. A. Mulkey, J. D. Dean, and P. Jowise. 1984.
User's Manual for the Pesticide Root Zone Model (PRZM): Release 1.
EPA-600/3-84-109. U.S. Environmental Protection Agency, Athens, GA.
Carsel, R. F., L. A. Mulkey, M. N. Lorber, and L. B. Baskin. 1985a. The
Pesticide Root Zone Model (PRZM): A procedure for evaluating pesticide
leaching threats to groundwater. Ecol. Modeling (In Press).
Carsel R. F., W. B. Nixon, and L. G. Balentine. 1985b. Comparison of Pesticide
Root Zone Model predictions with observed concentrations for the tobacco
pesticide metalaxyl in unsaturated soils. Environ. Toxicol. Chem. (In
Press).
Childs, S. W., and R. J. Hanks. 1975. Model for soil salinity effects on crop
growth. Soil Sci. Soc. Amer. Proc. 39:617-622.
Dagan, G., and E. Bresler. 1979. Solute dispersion in unsaturated hetero-
geneous soil at field scale. I. Theory. Soil Sci. Amer. J_. 43:461-466.
Dean, J. D., A. S. Donigian, Jr., and J. E. Rafferty. 1984a. Development
of a Mathematical Model to Evaluate Pesticide Exposure to Birds in
Agricultural Ecosystems. Prepared by Anderson-Nichols and Co., Inc., Palo
Alto, CA for U.S. Environmental Protection Agency, Athens, GA.
Dean, J. D., P. P. Jowise, and A. S. Donigian, Jr. 1984b. Leaching Evaluation
of Agricultural Chemicals (LEACH) Handbook. EPA-600/3-84-0068.
U.S. Environmental Protection Agency, Athens, GA.
Dean, J. D., D. W. Meier, B. R. Bicknell, and A. S. Donigian, Jr. 1984c.
Simulation of DDT Transport and Fate in the Arroyo Colorado Watershed,
Texas. U.S. Environmental Protection Agency, Athens, GA.
1-28
-------
Dean, J. D., and D. F. Atwood. 1985. Exposure Assessment Modeling for Aldicarb
In Florida. U.S. Environmental Protection Agency, Athens, GA.
De Smedt, F., and P. J. Wierenga. 1978. Approximate analytical solution for
solute flow during infiltration and redistribution. Soil Sci. Soc. Amer.
J_. 42:407-412.
Donigian, A. S. Jr., and J. D. Dean. 1985. Nonpoint Source Pollution Models
for Chemicals. Iji: W. B. Neely and G. E. Blau (Eds.), Environmental
Exposure from Chemicals. Vol. 2. CRC Press, Inc., Boca Raton, PL.
Donigian, A. S. Jr., and R. F. Carsel. 1985. Impact of Conservation Tillage
on Pesticide Concentrations in Ground Water, Surface Water, and Air.
Draft Report. Prepared for the U.S. Environmental Protection Agency,
Office of Policy Analysis, Washington, DC.
Donigian, A. S. Jr., S. M. Brown, and S. B. Yabusaki. 1983. Ground-Water
Modeling of Selected Hydrogeologic Settings to Determine Leachate Fate and
Migration from Waste Facilities. Prepared for the Environmental Protection
Agency, Office of Solid Waste, Washington, DC.
Eagelson, P. S. 1978. Climate, soil, and vegetation. Water Resour. Res.
14:1-7. ~~
Ellis, B. G., A. E. Erickson, A. R. Wolcott, M. Zabik, and R. Leavitt. 1977.
Pesticide Runoff Losses from Small Watersheds in Great Lakes Basin.
EPA-600/3-77-112. U.S. Environmental Protection Agency, Athens, GA.
Enfield, C. G., R. F. Carsel, S. Z. Cohen, T. Phan, and D. M. Walters. 1982.
Approximating pollutant transport to ground water. Ground Water
20:711-722.
Foster, G. R., L. J. Lane, J. D. Nowlin, J. M. Laflen, and R. A. Young. 1980.
A Model to Estimate Sediment Yield from Field-Sized Areas: Development
of Model, Purdue Agricultural Experiment Station. Purdue Journal No.
7781.
Hetrick, D. M. 1984. Simulation of the Hydrologic Cycle for Watersheds.
Proceedings for the Applied Simulation and Modeling Conference, San
Francisco, CA.
Hetrick, D. M., J. T. Holdeman, and R. J. Luxmore. 1982. AGTHEM: Documenta-
tion of Modifications to the Terrestrial Ecosystem Model (TEHM) for
Agricultural Applications. ORNL/TM-7856. Oak Ridge National Laboratory,
Oak Ridge, TN.
Hetrick, D. M., and L. M. McDonald-Boyer. 1984. User's Manual for TOX-SCREEN:
Multimedia Screening-Level Program for Assessing Potential Fate of
Chemical Released to the Environment. Oak Ridge National Laboratory, Oak
Ridge, TN. (ORNL-6041) and U.S. Environmental Protection Agency, Office
of Toxic Substances, Washington, DC. (EPA-560/83-024).
1-29
-------
Hornsby, A. G., P. S. C. Rao, VI. B. Wheeler, P. Nkedi-Kizza, and R. L. Jones.
1983. Fate of Aldicarb in Florida Citrus Soils: 1. Field Laboratory
Studies. Proceedings of the NWWA/U.S. EPA Conference on Characterization
and Monitoring of the Vadose (Unsaturated) Zone, Las Vegas, NV,
December 8-10, 1983.
Jones, R. L., and R. C. Back. 1984. Monitoring aldicarb residues in Florida
soil and water. Environ. Toxicol. Chem. 3:9-20.
Jones, R. L., P. S. C. Rao, and A. G. Hornsby. 1983. Fate of Aldicarb in
Florida Citrus Soil: 2. Model Evaluation. Proceedings of the NWWA/
U.S. EPA Conference on Characterization and Monitoring of the Vadose
(Unsaturated Zone), Las Vegas, NV, December 8-10, 1983.
Jones, R. L. 1983. Movement and Degradation of Aldicarb Residues in Soil
and Groundwater. Presented at the Environmental Toxicology and Chem-
istry Conference on Multidisciplinary Approaches to Environmental
Problems. Crystal City, VA.
Johnson, H. P. and J. L. Baker. 1982. Field-to-Stream Transport of Agricultural
Chemicals and Sediment in an Iowa Watershed: Part I. Date Base for Model
Testing (1976-1978). EPA-600/3-82-032. U.S. Environmental Protection
Agency, Athens, GA.
Jury, W. A. 1982. Simulation of solute transport using a transfer function
model. Water Resour. Res. 18:363-368.
Jury, W. A., L. A. Stolzy, and P. Shouse. 1982. A field test of the transfer
function model for predicting solute transport. Water Resour. Res.
18:369-375.
Kincaid, C. T., J. R. Morery, and J. E. Rogers. 1984a. Geohydrochemical Models
for Solute Migration, Volume 1: Process Description and Computer Code
Selection. AEA-3417, Electric Power Research Institute., Palo Alto, CA.
Kincaid, C. T., J. R. Morery, S. B. Yabusaki, A. R. Felmy, and J. E. Rogers.
1984b. Geohydrochemical Models for Solute Migration. Volume 2, Preliminary
Evaluation of Selected Computer Codes for Modeling Aqueous Solutions and
Solute Migration in Soils and Geologic Media. EA-3417, Vol. 2, Electric
Power Research Institute, Palo Alto, CA.
Knisel, W. (ed.). 1980. CREAMS: A Field-Scale Model for Chemicals, Runoff,
and Erosion from Agricultural Management Systems. Conservation Research
Report No. 26. U.S. Department of Agriculture.
Leonard, R. A., and V. A. Ferreira. 1984. CREAMS2 - The Nutrient and Pesti-
cide Models. Proceedings of the Natural Resources Modeling Symposium.
U.S. Department of Agriculture. (In Press).
Leonard, R. A., and W. G. Knisel, Jr. 1984. Model Selection for Nonpoint
Source Pollution and Resource Conservation. Proceedings of the Interna-
tional Conference on Agriculture and Environment. Venice, Italy.
1-30
-------
Melancon, S. M., J. E. Pollard, and S. C. Hern. 1986. Evaluation of SESOIL,
PRZM, and PEST AN in a laboratory. Environ. Toxicol . Chem. (In Press).
Nielsen, D. R., and J. VI. Biggar. 1962. Miscible displacement. III.
Theoretical considerations. Soil Sci. Soc. Amer. Proc. 26:216-221.
Oster, C. A. 1982. Review of Ground Water Flow and Transport Models in the
Unsaturated Zone. PNL-4427. NUREG/CR-2917. Battelle, Pacific North-
west Laboratory, Richland, WA.
Rao, P. S. C., and R. E. Jessup. 1982. Development and verification of simu-
lation models for describing pesticide dynamics in soils. Ecol . Model i ng
16:67-75.
Rao, P. C. S., and R. J. Wagenet. 1985. Spatial variation of pesticides in
field soils: Methods for data analysis and consequences. Weed Sc_1_. 33:
(In Press).
Rao, P. S. C., R. E. Jessup, and A. G. Hornsby. 1982. Simulation of nitrogen
in agroecosystems: Criteria for model selection and use. Plant and
Soil 67 : 35-43 .
Robbins, C. W., R. J. Wagenet, and J. J. Jurinak. 1980. A combined salt
transport-chemical equilibrium model for calcareous and gypsiferous
Soils. Sor Sci. Soc. of Amer. J. 44:1191-1194.
Rose, C. W., F. W. Chichester, J. R. Williams, and J. T. Ritchie. 1982a. A
contribution to simplified models of field solute transport. »L
Environ. Qua!. 11:146-150.
Rose, C. W., F. W. Chichester, J. R. Williams, and J. T. Ritchie. 1982b.
Application of an approximate analytic method of computing solute profiles
with dispersion in soils. J_. Environ. Qua! . 11:151-155.
Schroeder, P. R., J. M. Morgan, T. M. Walbki, and A. C. Gibson. 1984. The
Hydrologic Evaluation of Landfill Performance (HELP) Model, Volume I.
User's Guide for Version I. EPA/530-SW-84-009. U.S. Environmental Protec-
tion Agency, Office of Solid Waste and Emergency Response, Washington DC.
Smith, C. N., R. A. Leonard, G. W. Langdale, and G. W. Bailey. 1978. Transport
of Agricultural Chemicals from Small Upland Piedmont Watersheds. EPA-600/
3-78-056. IAG No. IAG-D6-0381. U.S. Environmental Protection Agency,
Athens, GA and U.S. Department of Agriculture, Watkinsville, GA.
Tanji, K. K., J. D. Doneen, G. V. Ferry, and R. S. Ayers. 1972. Computer
simulation analysis on reclamation of salt-affected soils in San Joaquin
Valley, CA. £oil Sci. Soc. Amer. Proc. 36:127-133.
van Genuchten, M. Th., and P. J. Wierenga. 1976. Mass transfer studies in
porous sorbing media. I. Analytical solutions. Soil Sci. Soc. Amer. J.
40:473-480.
1-31
-------
van Genuchten, M. Th., and W. J. Alves. 1982. Analytical Solutions of the
One-Dimensional Convective Dispersive Solute Transport Equation. Bull.
1661. U.S. Department of Agriculture.
Wagenet, R. J., and P. S. C. Rao. 1985. Basic concepts of modeling pesticide
fate in the crop root zone. Weed Sci. 33.
Wagner, J., M. Bonazountas, and M. Alsterberg. 1983. Potential Fate of Buried
Halogenated Solvents via SESOIL. Prepared by Arthur D. Little, Inc.,
Cambridge, MA, for the Environmental Protection Agency, Office of Toxic
Substances, Washington, DC.
Watson, D. B., and S. M. Brown. 1984. Testing and Evaluation of the SESOIL
Model. Prepared by Anderson-Nichols and Co., Inc., Palo Alto, CA for U.S.
Environmental Protection Agency, Athens, 6A.
Wauchope, R. D. 1978. The pesticide content of surface water draining from
agricultural fields - a review. «K Environ. Qua!. 7:459-471.
Wauchope, R. D., and R. A. Leonard. 1980. Maximum pesticide concentrations
in agricultural runoff: A semiempirical prediction formula. J. Environ.
Qua!. 9:665-672.
Williams, J. R. 1975. Sediment Yield Prediction with Universal Equation
using Runoff Energy Factor. In: Present and Prospective Technology for
Predicting Sediment Yields and Sources. AARS-S-40 U.S. Department of
Agriculture.
Wischmeier, W. H., and D. D. Smith. 1978. Predicting Rainfall Erosion Losses
from Cropland - A Guide to Conservation Planning, Agricultural Handbook
537. U.S. Department of Agriculture.
1-32
-------
CHAPTER 2
TRANSPORT MECHANISMS AND LOSS PATHWAYS FOR CHEMICALS IN SOIL
by
W. A. Jury and R. L. Valentine
TRANSPORT MECHANISMS
Chemicals are transported through soil principally by three mechanisms:
mass flow of dissolved chemical within moving solution, liquid diffusion within
soil solution, and gaseous diffusion within soil air voids. The first mecha-
nism, mass flow, refers to the passive transport of dissolved solute within
moving soil water which is approximated as the product of the volume flux of
water times the dissolved solute concentration. Liquid diffusion refers to the
transport of the dissolved solutes within solution by intermolecular collisions
which move the solutes from regions of the higher solute density to lower
solute density. The diffusion flux is written as the product of the density or
concentration gradient and a coefficient of proportionality called the soil
liquid diffusion coefficient. Similarly, chemical vapor molecules in the soil
air spaces also undergo molecular collisions and spread out by vapor diffusion
which is expressed as the product of the vapor density or concentration gradient
and a proportionality coefficient called the soil vapor diffusion coefficient.
Since the soil water flux is represented as a continuous volume-averaged
quantity over many pores, the individual complicated water flow paths around
the soil grains are replaced by an equivalent one-dimensional water flow. When
this one-dimensional water flow is multiplied by the dissolved solute concen-
tration, the resulting mass flux does not take into account the extra solute
spreading which occurs by three-dimensional mass flow at the pore scale in the
actual system and is not represented in the volume-averaged mathematical treat-
ment. This apparent solute diffusion arising from the mass flux effects which
are obscured by mathematical volume averaging is called hydrodynamic dispersion.
Under certain conditions, this dispersion effect is mathematically equivalent
to liquid diffusion and may be included as a diffusion-like transport mechanism
by using an effective liquid diffusion-dispersion coefficient to account for
both dispersion and diffusion.
LOSS PATHWAYS FOR CHEMICALS IN SOIL
Chemicals added to a soil profile from the surface may leave the zone of
incorporation by one of three loss pathways. The first pathway, known as
leaching, takes place principally by mass flow and refers to the downward
movement of dissolved chemical. The second loss pathway, volatilization,
2-1
-------
refers to the loss of chemical vapor to the atmosphere through the soil surface.
The final loss pathway, degradation, refers to the biological or chemical
transformation of the chemical to a different form with properties distinct
from those of the chemical prior to transformation. Each of these loss path-
ways will be discussed in greater detail below.
PARAMETERS INFLUENCING TRANSPORT PROCESSES AND LOSS PATHWAYS
Table 2-1 summmarizes the significant soil, environmental, and manager-
ial parameters influencing the four transport processes and atmospheric vol-
atilization. Since mass flow is the product of water flux and dissolved
chemical concentration, it is most strongly influenced by those parameters
having a direct influence on these two variables. Thus, increases in the
amount of applied water cause increases in water flux and hence in mass flow
for a given amount of chemical. Furthermore, in structured soils, increases
in water application intensity for a given amount of applied water may increase
the probability of macropore flow (flow through cracks or channels) which can
move small amounts of chemical great distances.
For a given total amount of chemical, the fraction which exists in the
dissolved phase and hence is available for mass flow is strongly influenced by
the adsorptive properties of the soil matrix. For nonpolar organic molecules,
most of the adsorption takes place on organic matter surfaces. Thus, there is
a strong relationship between organic matter or organic carbon content and
total adsorption. For positively charged species such as inorganic cations and
a few pesticides like paraquat and diquat, the negatively charged clay mineral
surfaces will be the principal adsorption sites, unless the soil is strongly
acidic; if the soil is strongly acidic, negatively charged species may be
adsorbed as well. Unless the adsorption isotherm is strongly nonlinear, in-
creasing the total concentration of the chemical should give a nearly propor-
tional increase in the solution concentration and thus proportionally affect
mass flow.
When the soil is unsaturated, mass flow is principally regulated by the
amount of water applied at the surface. However, under saturated conditions
when the soil surface is ponded, the saturated hydraulic conductivity of the
soil will have a dominant effect on the amount of water entering the soil and
hence on mass flow. Under such conditions, it is also more likely that adsorp-
tion will be rate-limited and less effective, and the extent of chemical trans-
port by mass flow will be greater for a given amount of applied water than when
the soil is unsaturated.
Temperature principally Influences mass flow by increasing the total
fraction of chemical mass which is present in the dissolved phase. Increasing
temperature also lowers the viscosity of water which modestly increases the
saturated hydraulic conductivity (approximately 25 percent between 20°C and
30°C).
The principal factor influencing the amount of transport by vapor diffu-
sion for a given chemical is the fraction of the total chemical concentration
which is present in the vapor phase. In assessing this tendency, two indices
are important: adsorption and the Henry's constant K^. As a general rule,
2-2
-------
TABLE 2-1. SUMMARY OF SIGNIFICANT PROCESS-PARAMETER RELATIONSHIPS IN CHEMICAL TRANSPORT
AND LOSS TO THE ATMOSPHERE
Process
Parameter
Influence
1. Mass Flow
ro
i
a. Amount of
applied water
b. Adsorption
site density
c. Chemical
concentration
d. Water
application
intensity
e. Saturated
hydraulic
conductivity K
Mass flow of dissolved chemical is directly proportional
to water input.
Adsorption decreases dissolved chemical concentration and
decreases mass flow. Nonpolar organic chemicals bind
primarily to organic matter, whereas cationlc (posi-
tively charged) species adsorb to negatively-charged
mineral surfaces. Anionic (negatively charged) species
may also bind to soil minerals for soils of low pH.
For chemicals whose water solubility is not exceeded,
increased chemical concentration will increase dissolved
chemical concentration and increase mass flow. The
increase may be greater than proportional if the ad-
sorption isotherm of the chemical is nonlinear.
For a given amount of water application, increased water
application rate may increase mass flow by decreasing
rate-limited adsorption and may produce increased
macropore (pore bypass) flow thereby causing deeper
penetration of part of the chemical mass.
Under ponded conditions, mass flow is proportional to
saturated hydraulic conductivity. For subsaturated
water application, control is regulated by the surface
application, and the hydraulic conductivity has no
direct effect on mass flow. However, soils with large
K generally retain less water, and penetration is deeper
for a given water application.
-------
TABLE 2-1. (continued)
Process
Parameter
Influence
2. Vapor
Diffusion
no
i
f. Temperature T
a. Water content 0
b. Adsorption
site density
c. Chemical
concentration
d. Temperature
e. Henry's
constant
For many chemicals, increasing temperature decreases ad-
sorption and increases mass flow. Saturated hydraulic
conductivity increases with temperature.
Vapor diffusion decreases strongly with Increasing water
content. A frequently used model assumes that the soil
vapor diffusion coefficient Is proportional to
(0-0)10/3, where 0 is porosity.
Adsorption decreases gaseous chemical concentration and
decreases vapor diffusion. Most volatile adsorbed chem-
icals are nonpolar and adsorb primarily to organic
matter.
For chemicals whose vapor density is not saturated, in-
creasing chemical concentration will increase vapor
density and increase vapor diffusion. The increase may
be greater than proportional if the chemical vapor
adsorption isotherm is nonlinear.
Increasing temperature significantly increases vapor
density for a given amount of chemical in soil, thereby
increasing vapor diffusion. The vapor diffusion co-
efficient increases nonlinearly with increasing temper-
ature, proportional to T1-75 (Kelvin).
Henry's constant (ratio of saturated vapor density to
solubility) is an Index of the partitioning of a chem-
ical between dissolved and gaseous phases. The larger
the KH, the more likely the chemical is to move by
vapor diffusion as opposed to liquid diffusion.
(continued)
-------
TABLE 2-1. (continued)
Process
Parameter
Influence
3. Liquid
Diffusion
ro
en
4. Hydrodynamic
Dispersion
a. Water content 0
b. Adsorption
site density
c. Chemical
concentration
d. Henry's
constant K
e. Temperature
a. Mater
application
b. Scale of
problem
Liquid diffusion increases strongly with increasing water
content. A frequently used model assumes that the soil
liquid diffusion coefficient is proportional to O10/3.
Adsorption decreases dissolved chemical concentration
and decreases liquid diffusion (see Ib above).
For chemicals whose solubility is not exceeded, increasing
chemical concentration will increase dissolved chemical
concentration, thereby increasing liquid diffusion. The
increase may be greater than proportional if the adsorp-
tion isotherm is nonlinear.
Chemicals with low KH will primarily diffuse in the liquid
phase as opposed to the vapor phase.
Liquid diffusion increases with Increasing temperature
because adsorption generally decreases. The liquid dif-
fusion coefficient increases modestly with increasing T.
The dispersion coefficient Is usually assumed to be pro-
portional to water velocity which is proportional to
water application rate.
Dispersion is a pseudo-mechanism resulting from represent-
ing mass flow by an average one-dimensional model while
neglecting convective mixing which is represented as
dispersion. The dispersion coefficient is thus depend-
ent on the size of the simulation and may increase by
orders of magnitude between lab and field.
(continued)
-------
Process
TABLE 2-1. (continued)
:s=============================================:
Parameter
Influence
5. Volatilization
to the Atmosphere
PO
at
a. Henry's
constant
b. Concentration
c. Adsorption
site density
d. Temperature
e. Water content 9
f. Wind speed
g. Water
evaporation
Chemicals with large KH are generally more volatile than
those with small KH and tend to have volatilization
rates which are limited more by soil conditions than
atmospheric conditions.
For chemicals whose vapor density is not saturated,
increasing concentration will increase volatilization
(see 2c).
Volatilization decreases as adsorption increases (see
2b).
Volatilization increases significantly as temperature
increases (see 2d).
Decreasing water content increases vapor diffusion and
volatilization (see 2a). If the surface dries out com-
pletely, adsorption increases dramatically, and vol-
atilization may cease until surface is rewet.
Increased wind speed improves mixing with the atmosphere
and can increase volatilization especially of chemicals
with low KH-
Water evaporation aids in transporting chemical to the
surface thus increasing volatilization especially of
weakly adsorbed chemicals.
-------
increasing the total adsorption of the chemical to soil lowers the vapor con-
centration, and decreasing the value of Henry's constant KH lowers the vapor
concentration. Chemicals with Henry's constants lower than 10'4 in dimension-
less units tend to be found principally in the dissolved as opposed to the vapor
phase (Jury et al., 1984a). For chemicals with an appreciable vapor concentra-
tion, soil water content strongly influences the amount of transport by vapor
diffusion. One frequently-used model, called the Millington and Quirk tortuos-
ity model, expresses the ratio of the soil to air diffusion coefficient as the
10/3 power of the volumetric air content (Millington and Quirk, 1961).
For a given amount of chemical present in the soil, increasing temperature
can strongly enhance the relative amount of movement by vapor diffusion by
increasing the Henry's constant and also by raising the value of the vapor
diffusion coefficient.
The extent of solute transport by liquid diffusion in soil is moderated by
the amount of dissolved chemical and the space available for flow. Increases
in soil adsorption tend to decrease the concentration in the dissolved phase and
hence decrease liquid diffusion. Also, chemicals with low Henry's constant
tend to be found primarily in the liquid phase and hence have a higher affinity
for liquid as opposed to vapor diffusion. Increases in water content strongly
increase the tendency for liquid diffusion by increasing the accessible flow
volume. The Millington and Quirk model, also applicable to liquid diffusion,
predicts that the soil liquid diffusion coefficient is proportional to the 10/3
power of the volumetric water content. As is the case with vapor diffusion,
increases in the total amount of chemical applied to soil tend to increase the
concentration in the dissolved phase and hence proportionally increase the
amount of movement by liquid diffusion. Contrary to vapor diffusion, however,
temperature increases only modestly increase the amount of movement by liquid
diffusion.
Hydrodynamic dispersion is really a form of mass flow and hence is gov-
erned by the principles discussed above for mass flow. However, the most
dominant influence on hydrodynamic dispersion is the scale over which the water
flux is averaged. Hence, should one attempt to use a one-dimensional chemical
transport model over a large area such as a field, one would have to use a very
large value of hydrodynamic dispersion coefficient to take into account the
chemical transport by variable water velocities not included in the one-
dimensional average for the mass flux.
Volatilization of chemical vapor to the atmosphere is an extremely com-
plex process controlled by soil, chemical, and atmospheric influences. As a
general rule, however, chemicals with a large value of Henry's constant KH
are more volatile than those with low KH and tend to be relatively independ-
ent of atmospheric conditions. Since volatilization takes place in the vapor
phase, processes which limit vapor diffusion, such as increased adsorption
and increased water content, can decrease the amount of volatilization. Con-
versely, processes which increase vapor diffusion such as increased temper-
ature and increased total concentration will increase the amount of volatili-
zation. Water evaporation can also increase volatilization as upward water
flow tends to bring chemical to the surface to replace that which is lost by
volatilization. This enhancement of volatilization flux is most pronounced
2-7
-------
when the chemical Is weakly adsorbed. Finally, Increases In wind speed In-
crease mixing with the atmosphere at the soil surface and Increase the amount
of volatilization. This Increase Is most dramatic for chemicals which have
a low Henry's constant and may have only minimal volatilization under stagnant
surface conditions.
MEASUREMENT OF KEY PARAMETERS
Table 2-2 summarizes some of the published methods of measurement of
various model parameters or soil properties which are relevant to modeling and
validation. The first set of properties, classified as static soil properties,
are each measured from intact or disturbed samples in the laboratory by stand-
ard methods. Also, most of these properties tend to have only modest amounts
of variation across large soil areas and thus could be characterized from a few
replicates (see Appendix G).
Conversely, the water transport and retention functions tend to have
higher coefficients of variation and different properties when measured under
field conditions than when the measurements are made in the laboratory on
Intact or disturbed soil cores.
The third group, called basic chemical properties, contains all pure
chemical properties which are measured in the absence of soil. In many cases,
values of the parameters may already be available in the literature (see Table
2-3). For those situations where values are not available, the references
provide the accepted protocol for measurement.
The fourth group consists of those time-dependent parameters which require
periodic monitoring during an experiment. The references provide standard
methods for field measurement of these variables.
Soil adsorption parameters (group five), must be measured in the labora-
tory, although adsorption parameters may also be inferred from soil column
breakthrough curves obtained on undisturbed cores originally taken from the
field (El Abd, 1984). There is no method available for measuring soil adsorp-
tion jji situ.
More complex adsorption relationships, which are discussed in the Ap-
pendices, must be obtained in laboratory equilibrium or transport experiments
by curve fitting. It Is difficult to make these laboratory experiments
representative of the field environment.
The tortuosity functions in parameter group six which describe the reduc-
tion in diffusion coefficient when soil is present can only be obtained in
laboratory experiments in which the volumetric water content is known. Sub-
stantial experimentation over the last two decades has shown the Millington
and Quirk model to be an accurate representation of the influence of water
content on vapor and liquid diffusion. Thus, use of this model Is recommended
in lieu of experimental measurements.
2-8
-------
TABLE 2-2. METHODS OF MEASUREMENT OF MODEL PARAMETERS OR SOIL PROPERTIES RELEVANT
TO MODELING AND VALIDATION
Parameter
Locale of
Measurement
Methodology
Refer-
ence
1. Static Soil Properties
ro
vo
Porosity
Bulk Density
Particle size
(% Sand, etc.)
Specific
surface area
Organic carbon
Cation exchange
capacity
PH
Laboratory Water content at zero suction on undisturbed cores
Laboratory Coring into known volume or intact clod of soil
Laboratory Hydrometer or pipette method after sieving
Laboratory Soil adsorption of gases or polar liquids
Laboratory
Laboratory
Laboratory
2. Water Transport and
Retention Functions
Saturated
hydraulic conductivity
Matric potential-
water content function
Unsaturated
hydraulic conductivity
Laboratory
Field
Field
Laboratory
Field
Laboratory
Field
Field
Field
Walkley-Black chromic acid titration method
Sodium saturation followed by ammonium acetate
replacement
pH meter reading of paste extract or soil solution
Ponding of soil cores (steady or falling head
method)
Steady state infiltration while monitoring
pressure head
Air entry permeameter
Hanging water table and pressure plate
Simultaneous tensiometer-neutron probe
measurements
Steady flow in soil columns
Instantaneous profile method
Unit gradient methods
Air entry permeameter
(11)
( 4)
( 8)
(26)
( 1)
(30)
(30)
(22)
(28)
(32)
(34)
(17)
(30)
(31)
(23)
(32)
(continued)
-------
TABLE 2-2. (continued)
Parameter
Solute velocity and
dispersion coefficient
Water flux
Volatilization Flux
Locale of
Measurement
Laboratory
Field
Field
Field
Methodology
Parameter fitting to column breakthrough curve
Parameter fitting to soil core or solution
sampler data
Indirectly, by measuring solute travel time
Atmospheric monitoring
Refer-
ence
( 2)
(16)
( 3)
(13)
3. Basic Chemical Properties
ro
i
Vapor pressure
Water solubility
Henry's constant KH
Vapor diffusion
coefficient in air
Liquid diffusion
coefficient in water
Octanol-water
partition coefficient
Laboratory
Laboratory
Derived
Literature
Literature
Laboratory
Gas saturation
Dilution of solute
Ratio of vapor pressure (or density) to
solubility
Estimate from similar compounds tabulated
Estimate from similar compounds tabulated
Equilibration with octanol-water mix
(33)
( 5)
( 6)
( 7)
(20)
4. Time Dependent Parameters
Requiring Monitoring
Water content
Solute concentration
Potential evaporation
Volatilization
boundary layer thickness
Field to
Laboratory
Field
Field
Field
Derived
Gravimetric determination from soil core
Neutron probe
Solution samplers and soil cores
Various correlations or energy balance methods
Calculate from water evaporation rate
(12)
(12)
(30)
( 9)
(18)
(continued)
-------
TABLE 2-2. (continued)
Parameter
Locale of
Measurement
Methodology
Refer-
ence
5. Soil Adsorption Parameters
Distribution
coefficient
Freundlich isotherm
Organic carbon
partition coefficient
Laboratory Batch adsorption to equilibrium
Laboratory Batch adsorption to equilibrium
Derived Ratio of distribution coefficient to
organic carbon fraction
(15)
(15)
ro
6. Tortuosity Functions
Vapor diffusion
tortuosity
Liquid diffusion
tortuosity
Laboratory
Laboratory
(Both functions have been found to fit model
of Millington and Quirk)
(18)
-------
TABULATIONS OF MEASURED VALUES
Table 2-3 provides references for tabulated values of key parameters for a
large number of chemicals. These values frequently vary significantly between
studies and should be used only when experimentation Is not feasible. Further
detailed descriptions of the transport mechanisms and chemical pathways are
found in Appendices A-C; significant mathematical derivations of equations
related to these processes are in Appendix D.
TABLE 2-3. SOURCES OF MEASURED VALUES FOR VARIOUS CHEMICAL PROPERTIES
======S=S=SSS==S=======S===£===SS=SSS=S====S=£========S========S=£=S=S=S=======
Property References
Vapor diffusion coefficient in air ( 6)
Liquid diffusion coefficient in water ( 7)
Organic carbon partition coefficients (15, 19, 21, 24, 29)
Octanol-water partition coefficients (19, 20, 21, 29)
Vapor density or vapor pressure (19, 35)
Water solubility (19, 21, 35)
Henry's constant (19, 35)
Degradation rate constant (14, 19, 27, 29)
TRANSFORMATION PROCESSES
Chemicals added to a soil may be transformed to products having properties
distinct from those of the chemical prior to transformation. While all proc-
esses leading to structural changes in a chemical occur as the result of chem-
ical reactions, these processes can be categorized as being either biological
or nonbiological transformations, depending on the role of microorganisms.
Nonbiological transformations can be classified further into chemical and photo-
chemical transformations depending on the role of light. In general, the
transformation products of nonbiological and biological processes may be similar
and encompass a wide variety of specific compounds characteristic of many kinds
of chemical reactions. For example, hydrolysis and oxidation-reduction reac-
tions are two general classes of reactions which typically occur by biological
and nonbiological processes. Detailed discussions of biological and nonbio-
logical transformation processes can be found in Appendices E and F.
B ^TRANSFORMATION
Biological transformations occur as the result of the metabolic activity
of microorganisms, primarily attached bacteria, actinomycetes, and fungi,
through the action of enzymes which catalyze chemical reactions. Not all
chemicals are transformed by all organisms. Many chemicals are degraded only
by a select group of organisms. Additionally, the transformation may Involve
one or more species and enzymes and form several products through complex re-
action pathways. These reactions usually lead to the production of energy or
some essential nutrient. However, some chemicals are also transformed even
2-12
-------
though the transformation does not promote growth. Biotransformation of this
sort has been termed cometabol1sm and can occur when other substances are
present which support growth.
NONBIOLOGICAL TRANSFORMATION
Certain classes of organic compounds may undergo chemical hydrolysis.
These reactions result in the net exchange of a hydroxyl group, OH', with some
other group, including halides, alcohols, amines, and sulfur- and phosphorus-
containing moieties. Many compounds do not undergo hydrolysis reactions.
However, likely candidates can be ascertained from knowledge of their structure.
Additionally, hydrolysis rates may vary greatly even among compounds of similar
structure. Many chemicals may also be oxidized or reduced by substances in the
soil. Oxidation-reduction reactions constitute a very general class of reac-
tions characterized by a change in the oxidation state of a chemical. Many
specific changes in the identity of a chemical can be classified as having
occurred via redox reactions.
Photochemical transformations occur as the result of light being absorbed
by the specific chemical of interest or by some other substance which then
reacts with 1t. Photochemical transformations include a variety of common
reactions such as oxidation, reduction, substitution, elimination, etc. Since
light does not penetrate very deeply into the soil, photochemical transforma-
tions are potentially important only at the surface of the soil.
FACTORS INFLUENCING BIOLOGICAL AND NONBIOLOGICAL TRANSFORMATION
The biotransformation of a chemical depends primarily on factors which
govern the nature of the microbial population, including size and composition,
and the availability of the chemical to attack by the organisms or their en-
zymes. The nonbiological transformation of a chemical is determined primarily
by factors which govern the availability of reactants and the nature of the
specific chemical in the soil.
Many of the factors which affect biological transformation also affect
nonbiological transformation processes. Furthermore, many of these factors are
highly interrelated. Any factor which affects sorption can be expected to have
an effect on biological and nonbiological transformation processes. Processes
occurring at the soil-water interface may differ both in rate and in kind from
those occurring in solution. Table 2-4 lists some of the more important factors
that influence transformation processes.
Microbial and chemical concentration are the two most common parameters
which are incorporated into mathematical expressions describing the rate of
biotransformation. It is axiomatic that the rate of biotransformation should
increase with an increase in the population of actively degrading micro-
organisms. Likewise, the rate should also increase with increasing chemical
concentration (at least over some range in concentration) unless the chemical
is toxic to the degrading organisms. The rates of chemical reactions are
expected to be a function of those compounds reacting in the rate limiting
step(s).
2-13
-------
TABLE 2-4. MAJOR FACTORS INFLUENCING BIOLOGICAL AND NONBIOLOGICAL
TRANSFORMATIONS
=========================================================================
Primary Influence
Factor Biological
NonbioTogical
1. Mlcroblal
concentration
2. Chemical
concentration
3. Temperature
4. Oxygen
5. Nutrients
6. Inorganic and
organic
composition
7. pH
8. Soil water
content
9. Acclimation
10. Light
Rates increase with increas-
ing active population.
Rates usually increase unless
chemical 1s toxic.
Rates increase with relatively
small increases in temperature
over range tolerated by
organisms.
Determines types of organisms
and degradation pathways
(aerobic vs. anaerobic).
Needed for microbial
metabolism.
Organic carbon may support
microbial population;
affects bioavailability.
Affects population com-
position; affects bio-
availability.
Needed for growth; affects
bioavailability; may
limit oxygen content
causing anaerobic system.
Possible time-lag
in degradation while
some enzymes induced.
NA
2-14
Not applicable (NA)
Rates may increase but
depends on reaction
mechanism/process.
Rates increase with
increases in
temperature.
Potential oxidant.
NA
Potential reactants and
catalysts; affects
availability.
Major effect on hydro-
lysis, probable effect
on most chemical re-
actions; affects avail-
ability.
Needed for hydrolysis;
may affect soil-water
interface pH; affects
chemical availability;
may limit oxidation by
oxygen.
NA
Increases transfor-
mation of certain
chemicals at soil
surface.
=========================
-------
Increases in temperature usually cause increased biotransformation rates
at least over the range in temperatures tolerated by the degrading organisms.
Temperature may also affect species composition and the availability of the
chemical through sorption. Temperature-dependent biotransformation rate con-
stants have been incorporated into models often using some relationship that
can be rationalized in terms of the Arrhenius equation typically used to relate
chemical reaction rates to temperature.
Oxygen plays a major role in determining the types of organisms present
and the metabolic pathways used to degrade chemicals. Some chemicals are
degraded only by aerobic organisms which require oxygen as the terminal elec-
tron acceptor in their metabolic pathway. Others are degraded by anaerobic
organisms which exist only in the absence of oxygen. Some organisms can grow
both in the absence and presence of oxygen by switching metabolic pathways. It
is generally under aerobic conditions that a chemical can be mineralized, i.e.,
degraded to carbon dioxide and other inorganic products. Frequently, the pro-
ducts of anaerobic degradation are very similar to the original parent compound.
Oxygen may be an important oxidant in the nonbiological oxidation of a chemical.
A variety of substances are needed for growth of an organism and must be
available in order for organisms to degrade a chemical. For example, nitrogen,
phosphorus, and carbon are universally needed by all organisms along with many
trace inorganic substances. These are readily available in most soils. The
primary source of carbon available to the degrading population may be the
naturally occurring soil organic carbon, not the added chemical. Microbial
concentrations are frequently correlated with soil organic carbon content.
Certain anaerobic organisms require inorganic compounds such as sulfate and
nitrate which are used as electron acceptors in the degradation mechanism.
Naturally occurring organic and inorganic substances may act as oxidants or
reductants in nonbiological transformation pathways. Some metals may also act
as catalysts in these processes.
The hydrogen ion content of the soil as measured by pH also affects both
the types of organisms present and the rate at which chemicals may be degraded.
Some organisms can exist only over a narrow range in pH. The rate of hydrol-
ysis is frequently directly related to pH, and pH may affect the rate of redox
reactions.
Water is required by all microorganisms and serves as the media through
which a chemical is made available to the attached microorganisms. Microbial
activity has been correlated over limited ranges to moisture content. Mois-
ture content affects sorption and is therefore expected to affect the rates
of transformation of chemicals in the soil. Soil moisture content may also
govern the oxygen content in the soil by limiting the rate at which it diffuses
into the soil. Hence a waterlogged soil may become anaerobic. The pH of the
soil-water interface is frequently a function of moisture content.
Many enzymes used to degrade a particular chemical must be first induced
after the chemical comes into contact with the microorganism. This can lead
to long acclimation periods over which little chemical is transformed.
2-15
-------
MEASUREMENT OF KEY PARAMETERS
The mathematical treatment of transformation is frequently based on an
approach which "lumps" together all processes due to the difficulty in distin-
guishing between biological and nonbiological transformations without extensive
studies. Such models usually consider the transformation of a chemical as a
single first order degradative process and are inherently the most site spe-
cific. Models which treat each transformation mechanism independently are
potentially applicable to a wider variety of field conditions and therefore may
be of greater usefulness.
Biotransformation rate expressions can be rationalized based on the work of
Monod and Michaelis-Menton. First order rate expressions have been frequently
used to model biotransformation. The first order rate constant is an implicit
function of the concentration of the actively degrading microbial population
present in the soil sample used to obtain it. Second order models which are an
explicit function of both chemical concentration and concentration of the
actively degrading microbial population are potentially less site specific.
Hydrolysis rate expressions are expected to be first order in chemical concen-
tration. However, rate expressions for other chemical processes occurring in
soil have not been developed from first principles. Nonbiological transforma-
tions are frequently modeled as first order processes for lack of some better
rate expression. Regardless of the seemingly detailed considerations given to
the evaluation of rate constants, they should be used with caution because of
the many unknown and possibly variable factors which affect them.
Separate biotransformation and nonbiological transformation rate coeffi-
cients cannot easily be evaluated under field conditions because of concurrent
abiotic and biological processes which lead to chemical loss. Laboratory
studies using sterile controls must be conducted to resolve biological from
nonbiological transformation processes. Most commonly used rate coefficients
are based on the rate of loss of the parent chemical concentration in the soil
without differentiating sorbed chemical from that in true solution or consider-
ing the extent of change in the identity of the parent chemical. These coeffi-
cients can be determined under aerobic and anaerobic conditions. The potential
importance of photochemical transformations at the soil surface can be evaluated
using natural and artificial light. In general, methodologies for the evaluation
of rate constants and rate parameters in soil systems are not well developed.
Table 2-5 lists some references that should be of use in evaluating rate con-
stants and parameters.
2-16
-------
TABLE 2-5. GENERAL APPROACHES AND METHODOLOGIES FOR THE DETERMINATION
OF TRANSFORMATION PARAMETERS
===============5====================================================3=========
Methodology/Transformation Process Reference
1. General field and laboratory testing
2. Use of microcosms
3. Subsurface sampling
4. Determination of microbial activity
5. Biological
6. Chemical
7. Photochemical
(43, 44, 45, 50, 54, 58, 60)
(66)
(36, 37, 46, 53, 65)
(40, 41, 42, 47, 51, 52, 62, 63,
64)
(50, 54, 59)
(39, 47, 50, 54, 56, 57, 67)
(38, 48, 49, 50, 54, 55, 61)
2-17
-------
REFERENCES
Allison, L. E. Organic Carbon. 1965. In: C. A. Black (Ed.), Methods of Soil
Analysis. American Society of Agronomy Monograph 9, Amer. Soc. Agron.,
Madison, WI.
Atlas, R. M., and R. Bartha. 1981. Microbial Ecology. Addison-Wesley.
Biggar, J. W., and 0. R. Nielsen. 1967. Miscible displacement and leaching
phenomena. Agron. Monograph 11:254-274.
Biggar, J. W., and D. R. Nielsen. 1976. Spatial variability of the leaching
characteristics of a field soil. Water Resour. Res. 12:78-84.
Blake, 6. R. 1965. Bulk Density, j^: C. A. Black (Ed.), Methods of Soil
Analysis. American Society of Agronomy Monograph 9, Amer. Soc. Agron.,
Madison, WI.
Board, R. G., and D. W. Lovelock. 1973. Sampling-Microbiological Monitoring
of Environments. Academic Press, New York.
Bowman, B. T., and W. W. Sans. 1979. The aqueous solubility of 27 insecti-
cides and related compounds. J_. Env. Sci. Health B14(6):221-227.
Boynton, W. B., and W. H. Brattain. 1929. Interdiffusion of gases and vapors.
Int. Crit. Tables 5:62-63.
Bruins, R. 1929. Coefficients of diffusion in liquids. Int. Crit. Tables
5:63-72.
Burkhard, N., and J. A. Guth. 1979. Photolysis of organophosphorus insecti-
cides on soil surfaces. Pestic. Sci. 10:313-319.
Burkhard, N., and J. A. Guth. 1981. Chemical hydrolysis of 2-chloro-4,
6-bis(alkylamino)-l,3,5-triazine herbicides and their breakdown in soil
under the influence of adsorption. Pestic. Sci. 12:45-52.
Daley, R. J. 1979. J^: Cooterton and Colwell (Eds.), Native Aquatic
Bacteria: Enumeration, Activity, and Ecology. ASTM/STP 695:29.
Day, P. R. 1965. Particle Fractionation and Particle Size Analysis. In:
C. A. Black, (Ed.), Methods of Soil Analysis. American Society of
Agronomy Monograph 9, Amer. soc. Agron., Madison, WI.
2-18
-------
Doorenbos, J., and VI. 0. Pruitt. 1976. Crop water requirements. Irrigation
and Drainage Paper 24, FAO, Rome.
Eiland, F. 1979. An improved method for determination of adenosine triphos-
phate (ATP) in soil. Soil Biol. Biochem. 11:31-35
Eiland, F., and B. S. Nielsen. 1979. Influence of cation content on adenosine
triphosphate determinations in soil. Microb. Ecol. 5:129-137.
El Abd, H. 1984. Spatial variability of the pesticide distribution coeffi-
cient. Ph.D. thesis, University of California, Riverside, CA.
Everett, L. G., L. G. Wilson, and L. G. McMillion. 1982. Vadose zone
monitoring concepts at hazardous waste sites. Groundwater 20:312-324.
Everett, L. G., L. G. Wilson, and E. W. Hoy1 man. 1983. Vadose Zone
Monitoring for Hazardous Waste Sites. EPA-600/X-83-064. U.S. Environ-
mental Protection Agency, Las Vegas, NV.
Ford, P. J., P. J. Turina, and D. E. Seely. 1983. Characterization of
Hazardous Waste Sites - A Methods Manual. Volume II, 2nd edition.
EPA-600/4-83-040. U.S. Environmental Protection Agency, Las Vegas, NV.
Foth, H. D. 1978. Fundamentals of Soil Science. 6th Ed. J. Wiley and Sons,
New York.
Gardner, W. H. 1965. Water Content. In: C. A. Black (Ed.), Methods of Soil
Analysis. American Society of Agronomy Monograph 9, Amer. soc. Agron.,
Madison, WI.
Gilmore, A. E. 1959. A soil sampling tube for soil microbiology. Soil Sci.
87:95-99
Glotfelty, D. D. 1981. Atmospheric Dispersion of Pesticides from Treated
Fields. Ph.D. Dissertation, University of Maryland.
Hamaker, J. W. 1972. Decomposition: Quantitative aspects. In:
C. A. I. Goring and J. W. Hamaker (Eds.), Organic Chemicals in the Soil
Environment. Marcel Dekker, Inc., New York.
Hamaker, J. W., and J. M. Thompson. 1972. Adsorption. In: C. A. I. Goring
and J. W. Hamaker (Eds.), Organic Chemicals in the Soil Environment.
Marcel Dekker, Inc., New York~7
Harris, J. C. 1982a. Rate of Hydrolysis. In: W. J. Lyman, W. F. Reehl,
and D. H. Rosenblatt (Eds.), Handbook oT""Chem1ca1 Property Estimation
Methods; Environmental Behavior of Organic Compounds. McGraw-Hill.
New York.
2-19
-------
Harris, J. C. 1982b. Rate of Aqueous Photolysis. In: W. J. Lyman, W. F.
Reehl, and D. H. Rosenblatt (Eds.), Handbook of Chemical Property
Estimation Methods; Environmental Behavior of Organic Compounds.
McGraw-Hill, New York.
Hautala, R. R. 1978. Surfactant Effects on Pesticide Photochemistry in Water
and Soil. EPA-600/3-78-060.
Howard, P. H., J. Saxena, P. R. Durkin, and L. T. Ou. 1975. Review and Eval-
uation of Available Techniques for Determining Persistence and Routes of
Degradation of Chemical Substances in the Environment. EPA-560/5-75-006.
Jenkinson, D. S., and J. N. Ladd. 1981. In: E. A. Paul and J. N. Ladd (Eds.),
Soil Biochemistry. Marcel-Dekker, New York.
Jury, W. A., and G. Sposito. 1985. Field calibration and validation of solute
transport models for the unsaturated zone. Soil Sci. Soc. Am. J.
49:1331-1341.
Jury, W. A., H. Frenkel, H. Fluhler, D. Devitt, and L. H. Stolzy. 1978. Use
of saline irrigation waters and minimal leaching for crop production.
Hilgardia 46:169-192.
Jury, W. A., W. F. Spencer, and W. J. Farmer. 1983. Behavior assessment model
for trace organics in soil. I. Model Description. J_. Environ. Qua!.
12:558-564.
Jury, W. A., W. F. Spencer, and W. J. Farmer. 1984. Behavior assessment model
for trace organics in soil. III. Application of screening model.
J_. Environ. Qua!. 4:573-578.
Karickhoff, S. W., and D. S. Brown. 1979. Determination of octanol/water
distribution coefficients, water solubilities and sediment. Water
partition coefficients for hydrophobic organic pollutants. EPA-600/
4-79-032. U.S. Environmental Protection Agency.
Kenaga, E. E. 1980. Predicted bioconcentration factors and soil sorption
coefficients of pesticides and other chemicals. Ecotox. and Environ.
Safety 4:26-38.
Klute, A. 1965. Laboratory Methods of Hydraulic Conductivity of a Saturated
Soil. ln_: C. A. Black (Ed.), Methods of Soil Analysis. American Society
of Agronomy Monograph 9, Amer. Soc. Agron., Madison, WI.
Leach, F. R. 1984. Biochemical Indicators of Groundwater Pollution. In:
G. Britton and C. P. Geraba (Eds.), Groundwater Pollution Microbiology.
Wiley-Interscience.
Libardi, P. L., K. Reichart, D. R. Nielsen, and J. W. Biggar. 1980. Simple
field methods for estimating soil hydraulic conductivity. Soil Sci. Soc.
Amer. J_. 44:3-7.
2-20
-------
Lyman, W. J. 1982. Adsorption coefficient for soils and sediments. In:
Lyman, W. J., W. F. Reehl, and D. H. Rosenblatt (Eds.), Handbook of
Chemical Property Estimation Methods. McGraw-Hill, New York.
McNabb, J. F., and G. E. Mallard. 1984. Microbiological Sampling In the
Assessment of Groundwater Pollution. In: G. Britton and C. P. Gerba
(Eds.), Groundwater Pollution Microbiology. Wiley-Interscience.
Mill, T., W. R. Mabey, D. C. Bombwerger, T. W. Chou, D. C. Hendry, and
J. H. Smith. 1982. Laboratory Protocols for Evaluating the Fate of
Organic Chemicals in Air and Water. EPA-600/3-82-022.
Millington, R. J. and J. M. Quirk. 1961. Permeability of porous solids.
Trans. Faraday Soc. 57:1200-1207.
Mortland, M. M., and W. D. Kemper. 1965. Specific Surface. In: C. A.
Black (Ed.), Methods of Soil Analysis. American Society of Agronomy
Monograph 9, Amer. Soc. Agron., Madison, WI.
Nash, R. G. 1980. Dissipation rate of pesticides from soils. In: W. G.
Knisel (Ed.), CREAMS. Vol. 3. U.S. Department of Agriculture,
Washington, DC.
Nielsen, D. R., J. W. Biggar, and K. T. Erh. 1973. Spatial variability of
field-measured soil-water properties. Hilgardia 42:215-259.
Parochetti, J. V., and G. W. Dec, Jr. 1978. Photodecomposition of eleven
dinitroaniline herbicides. Weed Sci. 26:153-156.
Perdue, E. M. 1983. Association of Organic Pollutants with Humic Substances:
Partitioning Equilibria and Hydrolysis Effects in Aquatic and Terrestrial
Humic Materials. R. F. Christman and E. I. Gjessing (Eds.). Ann Arbor
Science, Ann Arbor, MI.
Perdue, E. M., and N. L. Wolfe. 1983. Prediction of buffer catalysis in field
and laboratory studies of pollutant hydrolysis reactions. Environ. Sci.
Techno!. 17:635-642.
Plimmer, J. R. 1978. Degradation Methodology: Chemical-Physical Effects.
EPA-600/9-79-012. In: A. W. Bourquin and P. H. Pritchard (Eds.),
Proceedings of the Workshop on Microbial Degradation of Pollutants in
Marine Environments.
Rao, P. S. C., and J. M. Davidson. 1980. Estimation of pesticide retention
and transformation parameters required in nonpoint source pollution
models. In: M. R. Overcash and J. M. Davison (Eds.), Environmental
Impact of Nonpoint Source Pollution. Ann Arbor Science, Ann Arbor, MI.
Richards, L. A. 1954. Diagnosis and Improvement of Saline and Alkali Soils.
Agriculture Handbook 60. U.S. Department of Agriculture.
2-21
-------
Rose, C. W., W. R. Stern, and J. E. Drummond. 1965. Determination of hydraulic
conductivity in situ. Aust. J_. Soil Res. 3:1-9.
Russo, D., and E. Bresler. 1980. Field determination of soil hydraulic
properties for statistical analyses. Soil Sci. Soc. Amer. J_. 44:696-702.
Scow, K. M. 1982. Rate of Biodegradation. In: W. J. Lyman, W. F. Reehl,
and D. H. Rosenblatt (Eds.), Handbook ofThemical Property Estimation
Methods; Environmental Behavior of Organic Compounds. McGraw-Hill.
New York.
Sherma, J. 1981. Manual of Analytical Quality control for Pesticides and
Related compounds. EAP-600/2-81-059. U.S. Environmental Protection
Agency, Research Triangle Park, NC.
Smith, C. A., Y. Iwata, and F. A. Gunther. 1978. Conversion and disappear-
ance of methidathion on thin layers of dry soil. J_. Agric. Fd. Chem.
26:959-962.
Spencer, W. F., and M. M. Cliath. 1983. Measurement of pesticide vapor
pressures. Residue Rev. 85:57-71.
Stevenson, L. H., T. H. Chrzanowski, and C. W. Erkenbrecher. 1979. The adeo-
sine triphosphate assay: conceptions and misconceptions. ASTM/STP
695:99-111.
Taylor, S. A., and G. L. Ashcroft. 1972. Physical Edaphology. W. H. Freeman
and Co., San Francisco, CA.
Thomas, R. G. 1982. Volatilization from water. In: Lyman, U. J.,
W. H. Reehl, and D. H. Rosenblatt (Eds.), HanHbook of Chemical Property
Estimation Methods. McGraw-Hill, New York.
Trolldenier, G. 1973. The Use of Epifluorescence Microscopy for Counting Soil
Microorganisms. In: R. Rosswal (Ed.), Modern Methods in the Study of
Microbial Ecology. Bulletin from the Ecological Research Committee,
Swedish Natural Science Research Council, Stockholm 17:55-59.
Webster, J. J., G. J. Hampton, J. T. Wilson, W. C. Chiorse, and F. R. Leach.
1985. Determination of microbial cell numbers in subsurface samples.
Groundwater 23:17-25.
Wilson, L. G. 1980. Monitoring in the Vadose Zone. A Review of Technical
Elements and Methods. EPA-600/7-80-134. U.S. Environmental Protection
Agency, Las Vegas, NV.
Wilson, J., and M. J. Noonan. 1984. Microbial Activity in Model Aquifier
Systems. In: C. P. Gerba and G. Britton (Eds.), Groundwater Pollution
Microbiology. John Wiley and Sons.
2-22
-------
Wolfe, N. L., R. G. Zepp, J. A. Gordon, and G. L. Baughman. 1977. Kinetics of
chemical degradation of malathion in water. Environ. Sci. Techno!.
11:88-93.
2-23
-------
REFERENCES
1. Allison, L. E. Organic Carbon. 1965. ^n: C. A. Black (Ed.), Methods
of Soil Analysis. American Society of Agronomy Monograph 9, Amer. Soc.
Agron., Madison, WI.
2. Biggar, J. W., and D. R. Nielsen. 1967. Miscible displacement and
leaching phenomena. Agron. Monograph 11:254-274.
3. Biggar, J. W., and D. R. Nielsen. 1976. Spatial variability of the
leaching characteristics of a field soil. Water Resour. Res.
12:78-84.
4. Blake, 6. R. 1965. Bulk Density. In: C. A. Black (Ed.), Methods of
Soil Analysis. American Society of Agronomy Monograph 9, Amer. Soc.
Agron., Madison, WI.
5. Bowman, B. T., and W. W. Sans. 1979. The aqueous solubility of 27
insecticides and related compounds. J_. Erw. Sci. Health B14(6);221-227.
6. Boynton, W. B., and W. H. Brattain. 1929. Interdiffusion of gases and
vapors. Int. Crit. Tables 5:62-63.
7. Bruins, R. 1929. Coefficients of diffusion in liquids. Int. Crit.
Tables 5:63-72.
8. Day, P. R. 1965. Particle Fractionation and Particle Size Analysis.
In; C. A. Black, (Ed.), Methods of Soil Analysis. American Society of
Agronomy Monograph 9, Amer. Soc. Agron., Madison, WI.
9. Doorenbos, J., and W. 0. Pruitt. 1976. Crop water requirements. Irri-
gation and Drainage Paper 24, FAO, Rome.
10. El Abd, H. 1984. Spatial variability of the pesticide distribution
coefficient. Ph.D. thesis, University of California, Riverside, CA.
11. Foth, H. D. 1978. Fundamentals of Soil Science. 6th Ed. J. Wiley and
Sons, New York.
12. Gardner, W. H. 1965. Water Content. In: C. A. Black (Ed.), Methods
of Soil Analysis. American Society of Agronomy Monograph 9, Amer. Soc.
Agron., Madison, WI.
2-24
-------
13. Glotfelty, D. D. 1981. Atmospheric Dispersion of Pesticides from
Treated Fields. Ph.D. Dissertation, University of Maryland.
14. Hamaker, J. W. 1972. Decomposition: Quantitative aspects. In:
C. A. I. Goring and J. W. Hamaker (Eds.), Organic Chemicals in the Soil
Environment. Marcel Dekker, Inc., New York.
15. Hamaker, J. W., and J. M. Thompson. 1972. Adsorption. In: C. A. I.
Goring and J. W. Hamaker (Eds.), Organic Chemicals in the Soil Environment.
Marcel Dekker, Inc., New York.
16. Jury, W. A., and G. Sposito. 1985. Field calibration and validation of
solute transport models for the unsaturated zone. Soil Sci. Soc. Am. J.
49:1331-1341.
17. Jury, W. A., H. Frenkel, H. Fluhler, D. Devitt, and L. H. Stolzy. 1978.
Use of saline irrigation waters and minimal leaching for crop production.
Hilgardia 46:169-192.
18. Jury, W. A., W. F. Spencer, and W. J. Fanner. 1983. Behavior assess-
ment model for trace organics in soil. I. Model Description. J_.
Environ. Qua!. 12:558-564.
19. Jury, W. A., W. F. Spencer, and W. J. Farmer. 1984. Behavior assess-
ment model for trace organics in soil. III. Application of screening
model. J_. Environ. Qua!. 4:573-578.
20. Karickhoff, S. W., and D. S. Brown. 1979. Determination of octanol/
water distribution coefficients, water solubilities and sediment. Water
partition coefficients for hydrophobic organic pollutants. EPA-600/
4-79-032. U.S. Environmental Protection Agency.
21. Kenaga, E. E. 1980. Predicted bioconcentration factors and soil sorption
coefficients of pesticides and other chemicals. Ecotox. and Environ.
Safety 4:26-38.
22. Klute, A. 1965. Laboratory Methods of Hydraulic Conductivity of a
Saturated Soil. In: C. A. Black (Ed.), Methods of Soil Analysis.
American Society ^f Agronomy Monograph 9, Amer. Soc. Agron., Madison,
WI.
23. Libardi, P. L., K. Reichart, D. R. Nielsen, and J. W. Biggar. 1980.
Simple field methods for estimating soil hydraulic conductivity. Soil
Sci. Soc. Amer. J_. 44:3-7.
24. Lyman, W. J. 1982. Adsorption coefficient for soils and sediments.
In; Lyman, W. J., W. F. Reehl, and D. H. Rosenblatt (Eds.), Handbook
of Chemical Property Estimation Methods. McGraw-Hill, New YoTTTi
25. Millington, R. J. and J. M. Quirk. 1961. Permeability of porous solids.
Trans. Faraday Soc. 57:1200-1207.
2-25
-------
26. Mortland, M. M., and W. D. Kemper. 1965. Specific Surface. In: C. A.
Black (Ed.), Methods of Soil Analysis. American Society of Agronomy
Monograph 9, Amer. Soc. Agron., Madison, HI.
27. Nash, R. G. 1980. Dissipation rate of pesticides from soils. In: W. G.
Knisel (Ed.), CREAMS. Vol. 3. U.S. Department of Agriculture, Washington,
DC.
28. Nielsen, D. R., J. W. Biggar, and K. T. Erh. 1973. Spatial variability
of field-measured soil-water properties. Hilgardia 42;215-259.
29. Rao, P. S. C., and J. M. Davidson. 1980. Estimation of pesticide reten-
tion and transformation parameters required in nonpoint source pollution
models. In; M. R. Overcash and J. M. Davison (Eds.), Environmental Impact
of Nonpoint Source Pollution. Ann Arbor Science, Ann Arbor, MI.
30. Richards, L. A. 1954. Diagnosis and Improvement of Saline and Alkali
Soils. Agriculture Handbook 60. U.S. Department of Agriculture.
31. Rose, C. W., W. R. Stern, and J. E. Drummond. 1965. Determination of
hydraulic conductivity in situ. Aust. J_. Soil Res. 3:1-9.
32. Russo, D., and E. Bresler. 1980. Field determination of soil hydraulic
properties for statistical analyses. Soil Sci. Soc. Amer. J_. 44:696-702.
33. Spencer, W. F., and M. M. Cliath. 1983. Measurement of pesticide vapor
pressures. Residue Rev. 85:57-71.
34. Taylor, S. A., and G. L. Ashcroft. 1972. Physical Edaphology. W. H.
Freeman and Co., San Francisco, CA.
35. Thomas, R. G. 1982. Volatilization from water. In: Lyman, W. J.,
W. H. Reehl, and D. H. Rosenblatt (Eds.), Handbook of Chemical Property
Estimation Methods. McGraw-Hill, New York.
36. Atlas, R. M., and R. Bartha. 1981. Microbial Ecology. Addison-Wesley.
37. Board, R. G., and D. W. Lovelock. 1973. Sampling-Microbiological Moni-
toring of Environments. Academic Press, New Vork.
38. Burkhard, N., and J. A. Guth. 1979. Photolysis of organophosphorus
insecticides on soil surfaces. Pestic. Sci. 10:313-319.
39. Burkhard, N., and J. A. Guth. 1981. Chemical hydrolysis of 2-chloro-4,
6-bis(alkylamino)-l,3,5-triazine herbicides and their breakdown In soil
under the influence of adsorption. Pestic. Sci. 12:45-52.
40. Daley, R. J. 1979. In; Cooterton and Co'lwell (Eds.), Native Aquatic
Bacteria; Enumeration, Activity, and Ecology. ASTM/STP 695:29.
2-26
-------
41. Eiland, F. 1979. An improved method for determination of adenosine
triphosphate (ATP) in soil. Soil Biol. Biochem. 11:31-35
42. Eiland, F., and B. S. Nielsen. 1979. Influence of cation content on
adenosine triphosphate determinations in soil. Microb. Ecol. 5:129-137.
43. Everett, L. 6., L. G. Wilson, and L. 6. McMillion. 1982. Vadose zone
monitoring concepts at hazardous waste sites. Groundwater 20:312-324.
44. Everett, L. G., L. G. Wilson, and E. W. Hoylman. 1983. Vadose Zone
Monitoring for Hazardous Waste Sites. EPA-600/X-83-064. U.S. Environ-
mental Protection Agency, Las Vegas, NV.
45. Ford, P. J., P. J. Turina, and D. E. Seely. 1983. Characterization of
Hazardous Waste Sites - A Methods Manual. Volume II, 2nd edition.
EPA-600/4-83-040. U.S. Environmental Protection Agency, Las Vegas, NV.
46. Gilmore, A. E. 1959. A soil sampling tube for soil microbiology. Soil
Sci. 87:95-99
47. Harris, J. C. 1982a. Rate of Hydrolysis. In: W. J. Lyman, W. F. Reehl,
and D. H. Rosenblatt (Eds.), Handbook of Chemical Property Estimation
Methods; Environmental Behavior of Organic Compounds. McGraw-Hill.
New York.
48. Harris, J. C. 1982b. Rate of Aqueous Photolysis. In: W. J. Lyman, W. F.
Reehl, and D. H. Rosenblatt (Eds.), Handbook of Chemical Property Estima-
tion Methods: Environmental Behavio~of Organic Compounds. McGraw-Hill,
New York.
49. Hautala, R. R. 1978. Surfactant Effects on Pesticide Photochemistry in
Water and Soil. EPA-600/3-78-060.
50. Howard, P. H., J. Saxena, P. R. Durkin, and L. T. Ou. 1975. Review and
Evaluation of Available Techniques for Determining Persistence and Routes
of Degradation of Chemical Substances in the Environment. EPA-560/5-75-006.
51. Jenkinson, D. S., and J. N. Ladd. 1981. In: E. A. Paul and J. N. Ladd
(Eds.), Soil Biochemistry. Marcel-Dekker, New York.
52. Leach, F. R. 1984. Biochemical Indicators of Groundwater Pollution.
In: G. Britton and C. P. Geraba (Eds.), Groundwater Pollution Micro-
biology. Wiley-Interscience.
53. McNabb, J. F., and G. E. Mallard. 1984. Microbiological Sampling in the
Assessment of Groundwater Pollution. In: G. Britton and C. P. Gerba
(Eds.), Groundwater Pollution Microbiology. Wiley-Interscience.
54. Mill, T., W. R. Mabey, D. C. Bombwerger, T. W. Chou, D. C. Hendry, and
J. H. Smith. 1982. Laboratory Protocols for Evaluating the Fate of
Organic Chemicals 1n Air and Water. EPA-600/3-82-022.
2-27
-------
55. Parochetti, J. V., and G. W. Dec, Jr. 1978. Photodecomposltion of
eleven dinitroaniline herbicides. Weed Sci. 26:153-156.
56. Perdue, E. M. 1983. Association of Organic Pollutants with Humic Sub-
stances: Partitioning Equilibria and Hydrolysis Effects in Aquatic and
Terrestrial Humic Materials. R. F. Christman and E. I. Gjessing (Eds.).
Ann Arbor Science, Ann Arbor, MI.
57. Perdue, E. M., and N. L. Wolfe. 1983. Prediction of buffer catalysis
in field and laboratory studies of pollutant hydrolysis reactions.
Environ. Sci. Techno!. 17:635-642.
58. Plimmer, J. R. 1978. Degradation Methodology: Chemical-Physical Effects.
EPA-600/9-79-012. UK A. W. Bourquin and P. H. Pritchard (Eds.),
Proceedings of the Workshop on Microbial Degradation of Pollutants in
Marine Environments.
59. Scow, K. M. 1982. Rate of Biodegradation. In: W. J. Lyman, W. F. Reehl,
and D. H. Rosenblatt (Eds.), Handbook of Chemical Property Estimation
Methods: Environmental Behavior of Organic Compounds. McGraw-Hill,
New York.
60. Sherma, J. 1981. Manual of Analytical Quality Control for Pesticides
and Related Compounds. EPA-600/2-81-059. U.S. Environmental Protection
Agency, Research Triangle Park, NC.
61. Smith, C. A., Y. Iwata, and F. A. Gunther. 1978. Conversion and dis-
appearance of methidathion on thin layers of dry soil. J_. Agric. Fd.
Chem. 26:959-962.
62. Stevenson, L. H., T. H. Chrzanowski, and C. W. Erkenbrecher. 1979.
The adeosine triphosphate assay: conceptions and misconceptions.
ASTM/STP 695:99-111.
63. Trolldenier, G. 1973. The Use of Epifluorescence Microscopy for Count-
ing Soil Microorganisms. In: R. Rosswal (Ed.), Modern Methods in the
Study of Microbial Ecology. Bulletin from the Ecological Research
tt
il
Committee, Swedish Natural Science Research Council, Stockholm 17:55-59.
64. Webster, J. J., G. J. Hampton, J. T. Wilson, W. C. Chiorse, and F. R.
Leach. 1985. Determination of microbial cell numbers in subsurface
samples. Groundwater 23:17-25.
65. Wilson, L. G. 1980. Monitoring in the Vadose Zone. A Review of Tech-
nical Elements and Methods. EPA-600/7-80-134. U.S. Environmental
Protection Agency, Las Vegas, NV.
66. Wilson, J., and M. J. Noonan. 1984. Microbial Activity 1n Model Aquifier
Systems. UK C. P. Gerba and G. Britton (Eds.), Groundwater Pollution
Microbiology. John Wiley and Sons.
2-28
-------
67. Wolfe, N. L., R. G. Zepp, 0. A. Gordon, and G. L. Baughman. 1977.
Kinetics of chemical degradation of malathion in water. Environ. Sci.
Technol. 11:88-93.
2-29
-------
CHAPTER 3
GENERIC STEPS IN THE FIELD VALIDATION OF VADOSE ZONE
FATE AND TRANSPORT MODELS
by
S. C. Hern, S. M. Melancon, and J. E. Pollard
The primary emphasis of this document is on the transport and fate of
organic chemicals in the vadose Zone, i.e., from the soil surface to the ground-
water table. Model validation is defined in this report as comparison of model
results with numerical environmental data collected in the field or in labora-
tory observations. Complete model validation requires testing over the full
range of conditions for which predictions are intended. At a minimum, this
requires a series of validations in various climates and soil types with
chemicals that typify the major fate and transport processes. In this chapter,
suggested generic approaches to model validation will be presented, but the
reader should be aware that many validation problems are specific to a partic-
ular site, compound, or model and must be dealt with on a case-by-case basis.
In addition, model development and subsequent validation is an evolutionary
process by its very nature. Future research into chemical fate and transport
processes will produce a refinement of fundamental understanding which will
ultimately be used in the update of existing vadose zone models. Methodologies
for measurement of model input and output parameters will also be changed and
will thus impact model updates and improvement.
The major steps which will be discussed for designing a model validation
test are presented in Table 3-1. The reader should remain aware that these
generic steps will not always be completed in the sequence that they are here
presented. Rather, the validation process may involve dynamic overlap between
steps, depending on the data requirements of the model and the selected valida-
tion scenario.
STEP 1: IDENTIFY MODEL USER'S NEED
The first step in model validation is to obtain a detailed understanding
of the environmental problem confronting a potential model user. This under-
standing should be acquired through direct discussion with the user. The model
validator should elicit from the user the nature of the proposed simulation,
how the simulation results will be used, what model input data will be available,
how such data will be acquired, and what are the expected model outputs
(Donigian, 1980).
3-1
-------
TABLE 3-1. STEPS IN FIELD VALIDATION OF SOIL FATE AND TRANSPORT MODELS
Step 1. Identify Model User's Need—The first step in field validation
is to obtain a clear understanding of the model user's need, i.e., how will the
model be used.
Step 2. Examine the Model--
Step 2a. Detailed examination of the model: The user must pre-
cisely define model input data requirements, output predictions, and model
assumptions.
Step 2b. Collect Preliminary Data and Performance of Sensitivity
Analysis: Preliminary data are required to conduct a sensitivity analysis and
determine the most important input variables.
Step 3. Evaluate the Feasibility of Field Validation—Some models cannot
be validated in the field, and the validator should consider this possibility.
Step 4. Develop Acceptance Criteria for Validations—The model user must
provide criteria against which the model is to be judged.
Step 5. Determine Field Validation Scenario—Many different approaches to
field validation are possible. A scenario should be identified and approved by
the model user.
Step 6. Plan and Conduct Field Validations Which Should Include the Fol-
lowing Steps--
Step 6a. Select a Site and Compound(s): Consideration of model
input requirements, analytical methods, sources of contamination, and site soil
characteristics, etc. are among the many factors to consider in selecting a
site and compound(s).
Step 6b. Develop a Field Study Design: Development of a de-
tailed field sampling plan for the specific model compound and site.
Step 6c. Conduct Field Study: Implementation of the field plan
is not addressed in these guidelines.
Step 6d. Sample Analysis and Quality Assurance: Many analytical
procedures are available depending on the chemical and the matrix. Standard-
ized methods should be used together with a sound quality assurance program.
Step 6e. Compare Model Performance with Acceptance Criteria: A
comparison must be made between the performance of the model and the user's
acceptance criteria using either graphical or statistical techniques.
3-2
-------
This problem-identification step is important in defining whether a given
model is appropriate, i.e., whether it is capable of meeting the user's needs.
Thus, it is essential to have a thorough a priori understanding of the problem
to make an assessment regarding the utility of the model. If the model is to
be used in several ways or for several different purposes, each use or purpose
needs to be defined at the outset. Just because a model has been found valid
for one use does not mean it is appropriate for some other use. The validator
may discover after a detailed review of the model that the model cannot be used
as proposed because of the assumptions of the model the cost of obtaining the
input data, etc., and thus may save the user considerable time and expense
involved in attempting a field validation.
STEP 2: EXAMINE THE MODEL
This examination should concentrate on determining the required inputs to
the model (chemical, biological, and physical), the output predictions made by
the model, and the assumptions used in the model construction. The collection
of preliminary data, either field measured or obtained from the literature for
use in initial sensitivity analysis model runs, is also a part of this early
model examination stage.
A model input should be determined for each environmental fate process
which affects the compound of interest, e.g., oxidation, hydrolysis, etc. In
examining model inputs, it is important to define precisely the units of meas-
ure of each input. For example, "organic carbon content" may be expressed on a
dry or wet weight basis as g/kg, mg/kg, ug/kg, or as percent data by weight or
volume. Other considerations include determining whether the input values of
a model should represent a spatial average for some compartment, an isolated
point in space, or a soil horizon cross-section, etc., and whether these numer-
ical values represent one point in time or an average over an extended period.
Model outputs must be carefully defined by the same criteria as the input data.
For example, prediction of x ppm in soil could refer to the total concentration
in the soil and soil water matrix or just the soil portion of the matrix.
In addition to examination of model input and output parameters, the
relative sensitivity of the model to various input values must be evaluated.
Sensitivity is the rate of change of the output of the model caused by a change
in Input. If a change in an input causes a large change in the output, the
model is sensitive to that input; if a change in an input causes a small change
in the output, the model is less sensitive to that input. A sensitivity anal-
ysis identifies which inputs have a large influence on output and so should be
defined with better input accuracy and precision. Using model sensitivities to
plan the sampling design and to choose the number of samples can aid in obtain-
ing the appropriate accuracy and precision for the model users' needs.
Sensitivity analysis can range in complexity from a complete factorial
design using all combinations of different levels of input parameter values to
a simple high and low screening approach such as that described by Hoffman and
Gardner (1983). The level of complexity in a sensitivity analysis will depend
on the needs of the model user, the model being tested, and the number and
type of input parameters required by the model. For a complete factorial-
design sensitivity analysis to be practical, the model must have the capability
3-3
-------
of changing input parameters without extensive user Interaction. A simple
screening approach to model sensitivity may be the only practical means of
optimizing the experimental design of a validation when the model requires
user input to change input parameter values for each model run.
The type of input parameters required by the model may also have a pro-
found influence on the extent of a sensitivity analysis. For example, it is
possible for a given input parameter to be hypersensitive under one set of
environmental conditions and insensitive under another set of conditions.
Thus, for parameters of this type, it is necessary to have very good estimates
of the range of environmental conditions to be encountered during a validation
attempt.
Another major step in the model examination process is development of a
thorough understanding of the various assumptions upon which each model Is
based. Assumptions are commonly made to mathematically simplify models which
will affect output interpretations and conclusions drawn from the model simu-
lation results. The major assumptions used in vadose zone fate and transport
models are discussed in detail in Chapter 1 of this document. In general,
however, it is important for the user at this stage to realize that violation
of many assumptions are unavoidable in field validation attempts, especially in
the more uncontrolled scenarios. Whether or not violation of these assumptions
would result in an invalid test of the model will depend on the user's choice
of validation scenario, the sensitivity of model input or output data to the
violated assumption(s), and the user's a priori established validation accept-
ance criteria for the model.
STEP 3: EVALUATE THE FEASIBILITY OF FIELD VALIDATION
Field validation is probably the most credible test of a model. However,
inappropriate application of field-tested results can have a significant impact
upon the use and credibility of a model, and, in some cases, selection of an
alternative model may be the best course of action. The following are examples
of instances where the applicability of field validation may be questioned.
0 A model that assumes steady state or a dynamic equilibrium may be dif-
ficult to field validate fully. Steady state conditions typically do
not exist for long in the environment; temperature, infiltration rates,
pollutant loads, microbial populations, etc. are almost always changing.
0 In some models, input parameters may not be quantifiable. An example
of such an input is the active pollutant degrading fraction of the
total microbial population. No standard or routine procedure exists
for directly acquiring this data.
0 Model output data may be uncoilectable for purposes of comparison with
simulation results. For example, a model may predict the concentration
of volatile organic pollutants on clay. Currently, however, no field
methods exist for separating clay from sand and silt without affecting
the concentration of volatile organic compounds.
3-4
-------
0 It may be unproductive to field test a model due to a large input sam-
pling error. For example, a degradation rate constant may be measured
several times yielding results that vary by orders of magnitude. If
this input is sensitive for the chemical and model of interest (a small
change in the input results in a large change in the output), the
resultant prediction will also have a very large sampling error. This
error may exceed the user's acceptance criteria established a priori
for precision. However, the output extremes may also reflect the range
of conditions which the user could expect to observe in field condi-
tions and may not invalidate the test results.
In summary, it is important to realize that field validation is not always
practical and that caution must be used in drawing conclusions from some valida-
tion attempts. However, this impact must be weighed against the risks of
relying on models which have not been field validated and which may generate
output data that are used either with a false sense of security or with undue
caution. Persons involved in possible field tests must make this determination
on a case-by-case basis.
STEP 4: DEVELOP ACCEPTANCE CRITERIA FOR VALIDATIONS
Prior to attempting to validate a model, the user should develop and
provide to the validator those criteria which will be used to accept or reject
the mathematical model. After identifying the environmental problem, determin-
ing how the mathematical model may assist in resolving that problem, and con-
ducting sensitivity analyses on the model input and output parameters, the user
should have a good idea as to the accuracy and precision required of the model.
In developing acceptance criteria, the user should consider the predictive
ability of alternate physical methods available to solve the problem, e.g.,
physical construction of a miniature ecosystem by which the environmental
problem can be modeled in the laboratory.
The acceptance criteria for a validation should be given in terms of
required accuracy, precision, and confidence interval. For example, a certain
user may want to estimate the level of some pollutant in the soil within ± one
order of magnitude and be correct 95 percent of the time. On the other hand,
this same user may set criteria for estimating the migration of a pollutant
into ground water at ± a factor of two with the same confidence level, if
chemical concentrations in ground water are of greater concern in the scenario
of interest. Determination of these acceptance criteria, e.g., confidence
intervals about the predictions of the model must also include consideration of
the accuracy and confidence intervals associated with the field data to be
simulated.
A lack of defined acceptance criteria could result in attempts to validate
a model that could not possibly satisfy the user. It may be appropriate to
reject a model prior to any field testing after considering the effect of input
sampling errors on the predictions of the model. For example, sensitivity
analyses as discussed in Step 2 might indicate that the input rate constants
would have to be determined an infeasible number of times to achieve satisfac-
tory confidence in the output estimates.
3-5
-------
STEP 5: DETERMINE FIELD VALIDATION SCENARIO
There are many possible field validation scenarios that could be proposed.
Below, three possible scenarios with their various attributes and limitations
are discussed: (1) laboratory testing supplemented with field validations, (2)
field validation under controlled conditions, and (3) field validation under
natural conditions.
Laboratory Testing Supplemented with Field Validations
This approach involves extensive testing of a model and each of its
processes in controlled environments followed by one or a few field tests.
Compounds would be selected which undergo several environmental processes. The
controlled environment to be used may be a fairly simple laboratory microcosm
such as soil columns which offers a relatively high degree of environmental
control with varying degrees of complexity and "real world" simulation.
Laboratory testing in controlled environments would also permit detailed
study of processes at relatively low expense, and such testing can avoid some
problems associated with model assumptions. The limitations of the scenario
are that the results of tests in artificial environments lack credibility since
artificial environments only approximate the "real world". A model may perform
well under such restricted conditions yet perform poorly in the field since
field environments include many variables (e.g., multidimensional chemical
flow, soil-type discontinuities, decaying root holes, etc.) that models do
not consider in their predictions. The addition of limited followup field
testing will add credibility to the laboratory testing. However, this approach
may not completely test all processes in the model.
Field Validation: Controlled Conditions
Selection of the controlled field validation scenario would entail dosing
natural or semi natural soil systems with various compounds to test each chem-
ical and environmental fate process. In these situations, selected variables
such as the loading of test chemicals to the soil system or the delivery of
water to small plots or 1n situ columns (i.e., those planted in external sites)
are controlled. Each process in the model is then considered as a submodel and
is tested separately. It is possible for results from a test of one process to
exceed established criteria confidence levels while all other processes are
acceptably accurate.
For any given combination of major soil textural class (sandy loam, clay,
etc.), climate, and chemical process, a validation attempt should be conducted.
A soil plot could be concurrently dosed with low levels of relatively non-toxic
compounds each of which typify one environmental process (e.g., oxidation,
hydrolysis, etc.). In such a case, all the model processes relevant to such a
test environment could be simultaneously tested. Alternately, the plot and
a limited number of chemical processes could be tested. Similar studies would
then have to be conducted on other soil types and climates. A few such studies
could yield very credible tests at a relatively economical cost compared with
those studies based on uncontrolled field validations. However, finding afford-
able, non-toxic chemicals which represent a simple environmental process In the
3-6
-------
field is difficult. Furthermore, controlled testing of this type necessitating
the addition of contaminants to soil systems that may in some cases be scientif-
ically justifiable and environmentally acceptable yet politically impossible.
If appropriate test chemicals have been found, however, there are ways to avoid
or minimize problems associated with intentional dosing. For example, isolated
locations and particularly areas removed from underground drinking water sources
might be selected to avoid exposure to the public.
Field Validation: Natural Conditions
Selection of the uncontrolled field validation scenario would entail
testing fate and transport models under conditions simultaneously characterized
by a combination of soil types, climates, and chemical mixtures and pollutant
sources. Such sites would include agricultural fields, pastures, or bare soil
lots. These sites might receive contaminant loads from direct dumping of
industrial wastes, nonpoint sources (urban and agricultural runoff), atmospheric
deposition, or contaminated ground- or surface-water infiltration. The scen-
ario could also include cases where a known type and quantity of organic chem-
ical is distributed across a field, but no controls are placed on climatic or
soil conditions. The main advantage of field validation using this scenario is
one of credibility. Model validation under "real world" conditions is fre-
quently regarded as the ultimate test of a model. The more complex the environ-
mental conditions and types of chemical mixtures, the more credible is the
test. However, such increased complexity generally adds to the cost of testing,
increases the sources of errors, makes data interpretation more difficult, and
yields less precise predictions.
As in the previous scenario, this approach to model validation would
ideally consist of a matrix of validation tests using selected soil types,
climates, and compounds to typify each environment and process. A model that
satisfies the user's acceptance criteria for each test in the matrix would then
be considered valid. This approach, however, would require a large number of
field tests to fill in the matrix and thus is very resource intensive.
STEP 6: PLAN AND CONDUCT FIELD VALIDATIONS
This section addresses a number of applied factors that must be considered
in the final stages of designing field studies to test a process-type soil fate
and transport model. By now, the model has been examined, and all required
environmental and chemical inputs identified by process. The outputs of the
model are known and all the assumptions and design limitations thoroughly
understood. The model has been judged potentially capable of meeting the
user's needs, the user's acceptance criteria have been established, and a
preliminary decision has been made that field validation is feasible. Finally,
a validation scenario has been identified. With these important theoretical
preliminaries in mind, the user must consider a variety of specific factors
such as selection of field site and chemical compound(s), development and
implementation of a field sampling protocol, type of samples to be collected
and chemical analyses to be performed, and comparison of model performance with
a priori acceptance criteria.
3-7
-------
Step 6a: Select Site and Compound(s)
Site and compound selection must be considered together. It is most
likely that not all the selection criteria can be met at any one location, and
the importance of each factor must be weighed by the investigator.
Compound Selection Factors-
There are a number of compound-selection factors which are important to
consider at this stage, including cost and availability of standard chemical
analytical methods, knowledge of the chemical transformations selected compounds
undergo in various environmental conditions, compound toxicity, and source of
chemical contamination. For example, methods must exist for quantifying input
loadings to the model and for determining concentrations of the compounds of
interest throughout the soil column or section of vadose zone where field
validation is being conducted. In the field validation (controlled conditions)
and laboratory testing scenarios, the selection of a pollutant source is ob-
viously no problem since the investigator actually doses the experimental plot.
For the field validation under natural conditions, the investigator is restric-
ted to selecting compounds that regardless of source already exist in elevated
concentrations and can be tracked through the vadose zone. Precision, accuracy,
and compound detection limits within the soil medium need to be determined.
Analytical costs can be prohibitive depending on type of desired analyses and
required volume of samples.
Availability of reliable rate coefficient information must be considered.
In most fate and transport models, rate constants are coupled with site specif-
ic environmental parameters to produce site-specific rates. Ideally, develop-
ment of compound-specific inputs using actual soil and field conditions of the
present study should be attempted. If this is not possible, the investigator
will have to rely on rate constant information available in the literature.
These literature values are typically based on laboratory tests using only the
neutral (not ionic) species of the test compound, and, furthermore, different
investigator's results often vary over several orders of magnitude. In some
cases, rate constant information may not even exist, and an investigator may be
forced to select another compound or use rate constants for structurally
similar compounds. Excellent sources of rate constant or persistence informa-
tion are Lyman et al. (1982), Rao and Davidson (1980), Hamaker (1972), and
Verschueren (1983). If the latter option is exercised (using values for struc-
turally similar compounds), the user must consider the interpretive limitations
of his simulation results.
The environmental fate of the compound is important In all three valida-
tion scenarios. Unfortunately, soil chemical processes are highly complex and
often inherently linked (e.g., volatilization and water transport), so selection
of compounds dominated by a single process is difficult. In all cases, the
predicted half-life of the compound must be considered for each process relative
to the time during which the compound can be tracked. The movement rate of a
compound through a soil matrix is also of primary importance in evaluating
environmental fate. A number of physico-chemical factors, e.g., degradation
rates, soil-water content, pH, and adsorption coefficients, may affect the
half-life or persistence of a compound in the soil. In field validation
scenarios, the mode of compound application, watershed management practices,
3-8
-------
and episodic runoff events near the test area also will influence compound
persistence (Smith et al., 1978).
Compound toxicity and environmental hazard are very important factors when
organic chemicals are being applied to natural soil environments. If this
approach is pursued and a suitable site located, the least toxic compounds
representative of a process should be selected. Since the priority pollutants
are undesirable contaminants, other readily degradable pesticides or herbicides
that pose no significant threat to the environment should be considered. Com-
pound toxicity can also become a significant problem in laboratory tests in
which large quantities of contaminated soil (e.g., from replicated large column
tests) are generated. Solid wastes or effluents defined as hazardous by the
EPA (U.S. EPA 1983b) are subject to extensive federal disposal regulations
(U.S. EPA 1983a, 1983b) because of their long-term environmental risks. Com-
pound selection criteria may need to include consideration of new disposal
restrictions in a large scale laboratory procedure.
Site Selection Factors-
General ly, the simpler the site in terms of the amount of data that must
be obtained, the more cost effective the validation effort. As previously dis-
cussed, one important factor in site selection is that the compound of interest
must be present at sufficiently high levels so that it can be traced for a con-
siderable depth or time period. The quantity of pollutant load to the test
plot must be known in order to adequately test the model. The accuracy of
these data will depend upon the types and number of sources, their relative
loading, and their variability. A single uniform surface application is prob-
ably the simplest situation while multiple applications perhaps compounded by
capillary movement from contaminated ground water or atmospheric deposition
create much more complex situations to model. The additional complexity can
result in more sources of error, wider confidence intervals around the observed
and predicted values, and a less definitive test of the model. Other organic
chemicals naturally occurring in the test field may also interfere with chemi-
cal analyses of the compound(s) of interest. Soil and interstitial water
samples should be analyzed for the compound(s) of interest. Soil and inter-
stitial water samples should be analyzed prior to any field study to determine
background contaminant levels. Additional potential sources of contaminants
such as irrigation water should be defined and sampled. If interferences are
found, a new site must be selected unless clean-up procedures can be developed
prior to further sampling efforts and can be maintained throughout the study or
unless a way can be found to characterize the level of residues throughout the
various sites and depths.
The collection of model site-specific physical or climatic input data,
e.g., precipitation or relative humidity, may present no particular problem.
However, for many common input parameters, this is not the case (Jury, this
document; Philip, 1980; Parkes and O'Callaghan, 1980; Schmugge et al., 1980;
Sorooshian and Gupra, 1983). Site access, availability and limitations of
appropriate sampling gear, and sample replication limitations dictated by
financial or manpower contingencies are other variables of particular consid-
eration. Perhaps the most direct and obvious method of finding a study site is
to consult with persons engaged in model development and testing. Other sources
of information are the Surveillance and Analysis Divisions of the EPA Regional
3-9
-------
offices, and State and local waste disposal control personnel. These people
may know of sites that meet some or most of the site selection criteria. In
locating sites where contaminants can be added to a field, various government-
owned reservations, government-sponsored national laboratories, agricultural
extension services, private research contractors, and organizations equipped
for controlled hazardous waste disposal should be contacted.
Step 6b. Develop a Field Study Design
Type and Number of Samples—
The type of samples to be collected is a function of the input require-
ments of the model, output statements, and the selected test compound and site.
Through a detailed examination of the model, an investigator will determine
input data requirements by process (oxidation, physical transport, etc.),
output predictions, and model assumptions. Chemical tracers, e.g., chloride,
fluorescent dyes, etc., may be useful in defining local hydrologic conditions
prior to beginning chemical sampling for model validation. The relative number
of samples of each type to be collected will be determined by both sensitivity
analysis simulation runs which examine parameter sensitivity and variability
and by user definition of the desired statistical reliability. The optimum
desired number of samples to be collected may have to be tempered by cost and
time considerations to give the actual sample size. A detailed description of
the statistical rationale underlying decisions about type and number of samples
is presented in this document in Appendix A.
Sampling Equipment—
Efficient collection of soil samples in a field study requires a thorough
knowledge of the many types of gear available from which to select, character-
istics of the soil and field sites to be sampled, and one's ultimate research
goals. Technical equipment for unsaturated zone monitoring is frequently not
interchangeable from one soil type or environment to another; a few factors
influencing equipment selection are summarized for example purposes in Table
3-2.
There have been several excellent documents released in recent years that
provide detailed and highly usable information on vadose zone sampler types,
limitations, and appropriate application (Wilson, 1980; Everett et al., 1982;
Everett et al., 1983; Ford et al., 1983). These sources are recommended as
invaluable for field studies involving soil monitoring.
Sampling Location—
In selecting sampling stations for a field study, rate and the path of
movement of a pollutant through the soil compartment to be examined is the
first consideration. The investigator needs to determine over what distance
and time the compound of interest can be detected. This will be accomplished
through a knowledge of the pollutant loading, hydrological conditions, adsorp-
tion and chemical breakdown rates, and the detection limit for the compound of
interest. Sensitivity analyses, cost, and a knowledge of parameter variability
also dictate the number of stations to be included in the monitoring scheme.
Availability of appropriate sampling gear (determined by soil type and moisture
conditions at the various stations) is an important consideration since, as
previously stated, much soil sampling equipment does not have equal utility at
3-10
-------
TABLE 3-2. SOME FACTORS INFLUENCING SELECTION OF THE MOST APPROPRIATE
SOIL SAMPLING GEAR FOR VADOSE ZONE MONITORING
1. Soil type - sandy, high clay content, rocky, etc.
2. Soil moisture capabilities - constantly near saturation, cyclic wetting
and drying episodes, etc.
3. Site accessibility - roads available for heavy vehicle traffic, permit
restrictions, etc.
4. Sample size requirements - number of replicates, grams per replicate needed
for analyses, etc.
5. Labor and power requirements - hand-operated, need for electrical
generators, weight of sampler when full, etc.
6. Sample depth requirements - surface scrapes only, throughout vadose
zone, etc.
different sites or with different soil types. Other sampling criteria influ-
encing site selection include accessibility to the site and labor and power
requirements for the preferred soil sampling device.
Duration of Field Project—
In designing the length of a field study, one must once again consider the
fate and transport characteristics of the chosen chemical(s). Compounds with
either short or long half-lives can be used to test physical transport and
bioaccumulation processes in the field sampling scenarios. Consideration must
be given to the statistical design of the study since determination of the
minimum number of field replicates needed for a valid test of the model will
influence the duration of the sampling project. The user must decide whether
to sample throughout a variety of seasonal and climatic conditions any of which
will affect crop growth, hydrology, and type of precipitation (snow vs. rain)
or to limit sampling to an abbreviated period of time. Sampling throughout
only a single season will simplify the validation process but will also reduce
the overall predictive conclusions which can be drawn from validation results.
Step 6c: Conduct Field Study
The actual steps in conducting a field study are highly specific to each
individual project. It is impossible to give directions covering all field
conditions, so the choice in sampling techniques must often be left to the
analyst's judgment. For this reason, detailed instructions for conducting the
field study are not discussed here. However, Chapters 4 and 5 of this document
presents several example scenarios in which the reader can see how various
field projects were conducted and resultant data used for model comparisons.
Some of the generalized sample collection and logistic details the user should
be cognizant of during the actual field study are presented in Table 3-3.
3-11
-------
TABLE 3-3. GENERAL SAMPLE COLLECTION AND LOGISTICAL CONSIDERATIONS FOR
FIELD VALIDATIONS
=========================================================================
Sample Collection
sampling procedures
sample collection gear - soil cores, augers, etc.
sample containers - soil bags, moisture cans, etc.
sample container preparation - soap and water wash, acid rinse, etc.
sample preservation
sample coding
shipment of samples to laboratory
sample storage prior to analysis
Logistics
schedules
record keeping - field notebooks, data forms, etc.
vehicles - car, trucks, boats, etc.
maps
access - permission to sample on private or public lands, road egress,
etc.
notification of local, state, and federal authorities
personnel - competent team leader, reliable assistants
data management - how data will be reduced, etc.
================================================================:
It should also be noted that whenever samples are collected at field sites,
a variety of parameters important for interpretation and correlation of all
data must be recorded in addition to those required for model input. These
parameters may include: sampling site location (coordinates, site number),
sampling depth, date and time of day, meteorological conditions (air tempera-
ture, wind speed and direction, percent cloud cover, precipitation), vegetation
type and extent of cover, and surrounding land use. Particular attention
should be paid to record-keeping in the field collection portion of the study.
All sample containers should be sealed and labeled before they are shipped to
the laboratory. Pertinent information should be recorded on a sample tag,
e.g., sample number, date and time taken, source of sample, preservation,
analyses to be performed, and name of sample collector. The field investigator
is also responsible for assuring that adequate size (volume or mass) samples
are collected and that the samples are appropriately replicated for meaningful
statistical evaluation or quality assurance purposes. Safety procedures
(including proper handling protocols and emergency procedures) should be under-
stood and followed by all personnel.
Step 6d; Sample Analysis and Quality Assurance
Chemical Analysis-
It is beyond the scope of these guidelines to identify sample collection,
sample preservation, and analytical procedures for all organic chemicals that
3-12
-------
could be used In model evaluations. Detailed analytical procedures have been
proposed by the U.S. EPA (1979) for determining the concentration of 113 or-
ganic toxic pollutants in water. Methods 601-613 apply to the analysis of
individual compounds or groups of chemically similar compounds. Methods 624
and 625 are GC/MS procedures for the analyses of the same compounds. The
proposed methods cover calibration of instruments, quality control, sample
collection, preservation and handling, sample extraction and analyses, and
calculations.
Analytical procedures for measuring pesticides in soils are described in
detail in Sections 11A and C of a regularly updated EPA pesticide analytical
manual (Watts and Thompson, yearly revisions); discussion of general procedures
from this manual and a number of other relevant sources are summarized in
Sections 9A and D of Manual of Analytical Quality Control for Pesticides and
Related Compounds by Sherma (1981). New analytical procedures which have been
reviewed, tested, and validated by the Association of Official Analytical
Chemists (AOAC) are reported in Official Methods of Analyses (AOAC, in press).
A general description of pesticide residue analytical procedures is also pro-
vided by Sherma (1981) following a discussion on inter- and intra-laboratory
quality control. Covered in this document are gas chromatograph procedures and
troubleshooting, and procedures for analysis of samples including extraction,
isolation, and confirmation of pesticide residues. Although the manual deals
primarily with pesticides, many of the procedures and recommendations apply to
the analysis of any organic chemical. These are but a few of the many refer-
ences available for chemical analytical procedures in soils. Where possible,
established analytical methods should be used. Whatever method is selected, it
must be tested and verified by the analyzing laboratory.
Quality Assurance—
The purpose of field sampling in model validation is to provide quantita-
tive environmental input data to the model and to collect field residue data
that can be compared to the levels predicted by the model. Such field programs
usually include the following operational steps:
Planning and Conceptualization Sample Analysis
Sample Collection Data Acquisition
Sample Preservation Data Manipulation
Sample Transport Data Interpretation
Sample Storage Reporting
Sample Preparation
To obtain valid data, an overall Quality Assurance (QA) program must be a
part of any sampling project. In addition to the usual analytical and equip-
ment QA procedures, a comprehensive QA program should include details on the
reliability of the sampling program. Sampling schemes, data analysis strat-
egies, and the objectives of the sampling program must be well-defined for a
statistician to assist in the development of an efficient data collection
program. A recently published summary QA document for soil studies by Barth
and Mason (1984) is available; the EPA also has published several quality
assurance procedure manuals (U.S. EPA 1978, 1980a, 19805).
3-13
-------
It should be noted that part of the literature data used in the validation
process may have been created in the past by different researchers in a variety
of studies. While no rigid a priori criteria for acceptance or rejection of
these data can be imposed (Davis 1980), it should be stressed that they must be
closely scrutinized. In many instances, data may be of limited value or even
useless because the precision and accuracy were not reported or because other
key chemical or environmental parameters were not presented.
Step 6e; Compare Model Performance with Field Observations
Comparison of the distribution of organic chemicals present in soils with
model predicted values can be accomplished in a variety of ways. The descrip-
tive and statistical methodologies used for comparing model simulations to
observed data will be dependent on the characteristics of observed and simu-
lated data sets. For example, rigorous statistical testing for differences
between observed and predicted data may not be appropriate for chemical concen-
tration values obtained from the literature due to lack of variance estimates.
Similarly, distribution of observed and predicted data sets may be so different
that statistical testing of the data yields no more information than a simple
graphical presentation would provide (Figure 3-1). If, however, sufficient
overlap between observed and predicted data occurs or the shape of the observed
and predicted chemical profiles are similar, reasonably simple statistical
techniques exist for testing the two data sets for differences or similarities.
Correlation analysis (Sokal and Rohlf, 1981) can be used to determine the
similarity of the distribution of observed and predicted data. High correla-
tions between observed and simulated distribution curves indicate that the
differences among observed data are controlled by factors accounted for by
the model (Hoffman and Gardner, 1983). High correlations do not, however,
substantiate overlap between observed and predicted data sets. The absolute
values of model simulations and observed data may be very different in highly
correlated data with differences being a simple scale factor. Replication of
observed data at a given depth or time increment allows determination of the
significance of the correlation thereby providing additional substantiation
for conclusions about model performance.
If sufficient overlap between observed and simulated data sets exists
to render difference testing appropriate, a multivariate t-test (Hotel!ings
T2) can be used (Morrison, 1976). This test provides a means of testing the
overall null hypothesis of identical profiles in the populations from which
the two groups of data were drawn (Harris, 1975). Alternatively, confidence
intervals about means of observed data can be calculated for a chosen con-
fidence level (Sokal and Rohlf, 1981) and plotted with the degree of over-
lap between observed and predicted data sets being determined by the degree
of intersection of plotted confidence bars with the predicted curve.
Unreplicated observed data provide little information for comparison of
observed and simulated data. Simple descriptive graphical presentation can
be used to display the patterns of similarity or difference between unrep-
licated data sets, but the significance of this relationship is indeterminate.
3-14
-------
1500-n
i
H—
cn
Lindane
^GO-
D)
•Z 900-
2
*-*
0)
o
c
o
o
600-
300-
0 20 40 60 80 100 120 140 160 180 200
Depth in Centimeters
Figure 3-1. Lindane distributed through four soil columns (52-55), compared with SESOIL,
PESTAN, and PRZM model predictions (modified from Melancon et al., 1986).
-------
Similarly, correlations of unreplicated data sets may be high, but the sig-
nificance of this relationship is questionable due to a lack of variance esti-
mates for observed or predicted data.
In many cases, observed chemical profiles in soils may be extremely
"noisy" due to spatial and temporal variability. Underlying patterns within
the data can be made more obvious by the use of data-smoothing techniques
(Davis, 1973). Since the concentration of a chemical at one depth of soil is
often related to the concentration at a subsequent depth, it is possible to
remove a portion of the noise or random variation from this type of data using
moving averages. This technique may aid in visually displaying the underlying
pattern in a set of observed chemical concentrations in a soil profile which
can then be compared graphically with model simulations of that soil profile.
3-16
-------
REFERENCES
Association of Official Analytical Chemists. In press (1984). Official
Methods of Analysis. 14th edition. Arlington, VA.
Barth, D. S., and B. J. Mason. 1984. Soil Sampling Quality Assurance User's
Guide. EPA-600/S4-84-043. U.S. Environmental Protection Agency,
Las Vegas, NV.
Davis, J. C. 1973. Statistics and Data Analysis in Geology. John Wiley and
Sons, New York. 550 pp.
Davis, T. T. (Chairman). 1980. State-of-the-Art Report of the Hazardous Sub-
stances Committee. In; Workshop on Verification of Water Quality Models.
EPA-600/9-80-016. U.S. Environmental Protection Agency, Washington, DC.
Donigian, A. S. 1980. Recommendation to Improve the Use of Models in Decision
Making. In; T. T. Davis (Ed.), Workshop on Verification of Water Quality
Models. EPA-600/9-80-016. U.S. Environmental Protection Agency,
Washington, DC.
Everett, L. G., L. G. Wilson, and L. G. McMillion. 1982. Vadose zone moni-
toring concepts at hazardous waste sites. Groundwater 20:312-324.
Everett, L. G., L. G. Wilson, and E. W. Hoy!man. 1983. Vadose Zone Monitoring
for Hazardous Waste Sites. EPA-600/X-83-064. U.S. Environmental Protec-
tion Agency, Las Vegas, NV.
Ford, P. J., P. J. Turina, and D. E. Seely. 1983. Characterization of Hazar-
dous Waste Sites - A Methods Manual. Volume II, 2nd edition. EPA-600/
4-83-040. U.S. Environmental Protection Agency, Las Vegas, NV.
Hamaker, J. W. 1972. Decomposition: Quantitative Aspects. In^ C.A.I
Goring, and J. W. Hamaker (Eds.), Organic
ment, Volume I. Marcel-Dekker, New York.
Goring, and J. W. Hamaker (Eds.), Organic Chemicals in the Soil Environ-
Harris, R. J. 1975. A Primer of Multivan'ate Statistics. Academic Press, Inc.,
New York.
Hoffman, F. D., and R. H. Gardner. 1983. In: J. E. Till and H. R. Meyer,
Eds., Radiological Assessment. NRC No. NUREG/CR-3332, ORNL-5968. Nuclear
Regulatory Commission, Washington DC.
3-17
-------
Lyman, W. J., W. F. Reehl, and D. H. Rosenblatt (Eds.). 1982. Handbook of
Chemical Property Estimation Methods; Environmental Behavior of Organic
Compounds. McGraw-Hill, New York.
Melancon, S. M., J. E. Pollard, and S. C. Hern. 1986. Evaluation of SESOIL,
PRZM, and PESTAN in a laboratory. Environ. Toxicol. Chen. (In Press).
Morrison, D. F. 1976. Multivariate Statistical Methods. McGraw Hill,
New York.
Parkes, M. E., and J. R. O'Callaghan. 1980. Modeling soil water changes in a
well-structured, freely draining soil. Water Resour. Res. 16:755-761.
Phillip, J. R. 1980. Field heterogenicity: some basic issues. Water Resour.
Res. 16:433-448.
Rao, P. S. C., and J. M. Davidson. 1980. Estimation of Pesticide Retention
and Transformation Parameters. In; M. R. Overcash (Ed.), Environmental
Impact of Nonpoint Source Pollution. Ann Arbor Sci. Pub!., Ann Arbor, MI.
Schmugge, T. J., T. J. Jackson, and H. L. McKim. 1980. Survey of methods for
soil moisture determination. Water Resour. Res. 16:961-979.
Sherma, J. 1981. Manual of Analytical Quality Control for Pesticides and
Related Compounds. EPA-600/2-81-059. U.S. Environmental Protection
Agency, Research Triangle Park, NC.
Smith, C. N., G. W. Bailey, R. A. Leonard, and G. W. Langdale. 1978. Transport
of Agricultural Chemicals from Small Upland Piedmont Watersheds.
EPA-600/3/78-056. U.S. Environmental Protection Agency, Athens, GA.
•
Sokal, R. R., and J. Rohlf. 1981. Biometry. W. H. Freeman Press,
San Francisco, CA.
Sorooshian, S., and V. K. Gupta. 1983. Automatic calibration of conceptual
rainfall-runoff models: the question of parameter observability and
uniqueness. Water Resour. Res. 19:260-268.
U.S. EPA. 1978. Quality Assurance Guidelines for Biological Testing. EPA-
600/4-79-043. U.S. Environmental Protection Agency, Cincinatti, OH.
U.S. EPA. 1979. Guidelines establishing test procedures for the analysis of
pollutants. Fed. Reg. 44:69464-69675.
U.S. EPA. 1980a. Interim Guidelines and Specifications for Preparing Quality
Assurance Project Plans. QAMS-005/80. U.S. Environmental Protection
Agency, Office of Monitoring Systems and Quality Assurance, Washington, DC.
U.S. EPA. 1980b. The Quality Assurance Bibliography. EPA-600/4-80-009. U.S.
Environmental Protection Agency, Washington, DC.
3-18
-------
U.S. EPA. 1983a. Title 40-Protection of Environment. Part 260 - Hazardous
Waste Management System: General, pp. 344-357.
U.S. EPA. 1983b. Title 40 - Protecting Environment. Part 261 - Identification
and History of Hazardous Waste, pp. 360-392.
Verschueren, K. 1983. Handbook of Environmental Data on Organic Chemicals.
Van Nostrand Reinhold Co., New York.
Watts, R. R., and J. R. Thompson. Yearly Revisions. Analysis of Pesticide
Residues in Human and Environmental Samples. U.S. Environmental Protection
Agency, Research Triangle Park, NC.
Wilson, L. 6. 1980. Monitoring in the Vadose Zone: A Review of Technical
Elements and Methods. EPA-600/7-80-134. U.S. Environmental Protection
Agency, Las Vegas, NV.
3-19
-------
CHAPTER 4
EXAMPLE FIELD TESTING OF SOIL FATE AND TRANSPORT MODEL, PRZM,
DOUGHERTY PLAIN, GEORGIA1
by
K. F. Hedden
INTRODUCTION
The previous chapters have described the processes that are important to
the fate and transport of organic chemicals in the vadose zone, given an over-
view of models and modeling and suggested steps to follow in model validation
and testing. Chapters 4 and 5 will attempt to show by use of actual examples
how some attempts at model validation and testing have been done and will
hopefully alert the reader to some limitations and problems which he or she may
encounter. A major problem for any validation or testing effort is obtaining a
data set that has determined the needed parameters and has sampled the appro-
priate variables for an adequate period of time. The first example (Chapter 4)
describes an effort that was designed with the express purpose of testing a
particular vadose zone model, PRZM. In this example, the author will follow
insofar as possible the steps in field validation described in Chapter 3 and
outlined in Table 3-1. This effort is not yet complete, so only the initial
stages can be described, but these should serve to give the reader an idea of
what is involved in testing a model. The second example (Chapter 5) describes
two retrospective studies. It gives guidance on finding and selecting an
adequate data set and illustrates some of the problems encountered when apply-
ing retrospective studies to model validation attempts.
STEP 1: IDENTIFY NEEDS
The Dougherty Plain project is an ongoing five-year cooperative research
project between the U.S. EPA, Environmental Research Laboratory, Athens,
Georgia, and the U.S. Geological Survey (USGS) in Georgia. The project was
designed to develop a data base for testing mathematical models that evaluate
the potential for contamination of ground-water resources from increased pesti-
cide applications. The model being utilized for initial testing is the Pesti-
cide Root Zone Model (PRZM), developed by Carsel et al. (1984). PRZM simulates
'•This example is primarily based on "Pesticide Migration in the Unsaturated
and Saturated Soil Zones. Part I. A Field Study to Support Model Development
and Testing, Dougherty Plains, GA, 1983" by Sandra C. Cooper, Robert F. Carsel,
Charles N. Smith, and Rudolph S. Parrish. USEPA, ORD, ERL, Athens, Georgia.
4-1
-------
the transport of pesticides in the unsaturated zone by integrating hydrologic
and pesticide process data relating to the sorption, degradation, and leaching
properties of pesticides. The Dougherty Plain data will be used to calibrate,
test, and refine PRZM.
STEP 2: EXAMINE MODEL
Input requirements necessitate the collection of field data on the soil
chemical and physical properties and the existing ground-water conditions.
Certain climatic data are also required. For comparing model predictions with
field data, samples collected in the unsaturated and saturated zones after
pesticide application are also needed. Model input data requirements are
listed in Table 4-1. The PRZM user's manual (Carsel et al., 1984) provides
much guidance in obtaining required input data. The EPA, Athens Lab also has
developed a meteorological information data base that can provide all needed
meteorological data. The Soil Conservation Service (SCS) has a soils data base
that can provide most of the needed soil cropping information.
Model output predictions are listed in Table 4-2. The hard output may be
called by day, month, or year. Three files may also be called: (1) hydrologic,
(2) pesticide output, and (3) pesticide concentration. Values are printed by
time step and compartment with summary data, and a variety of plots may be
produced (see Carsel et al., 1984).
Several simplifying assumptions have been made in developing the basic
equations of PRZM. While these assumptions may not exactly describe the
process modeled, they greatly simplify the model. Sorption of pesticide on
soil particles is assumed to be instantaneous and reversible. Dispersion and
diffusion are combined and described using Pick's law, and they are assumed
constant. One first order decomposition rate is used to represent the sum of
all degradative processes in both the soil and water phases. Plant uptake of
pesticides is modeled by assuming that uptake of a pesticide by a plant is
directly related to the transpiration rate. The equations are written for
vertical movement at a single point, but field soils are quite variable, and
conditions at one point are not necessarily representative of conditions at
other points.
A sensitivity analysis was conducted to determine which input parameters
should be given attention, and to what extent variations in expected input
values would affect output values. The sensitivity analyses conducted at this
stage of the model examination were based on the range of possible values for
each input parameter. Table 4-3 lists some of the more sensitive parameters
identified during this analysis stage. Further sensitivity analyses were then
done using a range of probable values for more sensitive parameters which were
either taken from the literature or measured in analysis of preliminary samples.
The input parameters most sensitive with regard to leaching are the decay rate,
sorption coefficient, and depth of active layer (root zone).
4-2
-------
4-1. INPUT DATA REQUIREMENTS FOR PRZM
Meteorological Data
Daily precipitation, cm
Daily pan evaporation, cm j(both not necessarily needed)
Daily average temperature, C)
Hydrology Data
Pan factor, dimensionless
Snow factor, cm snowmelt/°C above freezing
Minimum depth to which evaporation is extracted, cm
Average daily hours of daylight for each month, hr/day
Universal soil loss equation parameters/factors dimensionless
K-Soil erodability; LS-Topographic/length-slope
P-Supporting/Conservation Practice; C-Cover and Management
Field area, ha
Average duration of runoff hydrograph from runoff producing storms, hr
Maximum interception storage for each crop, cm
Maximum active root depth for each crop, cm
Maximum areal coverage of each crop at full canopy, (percent)
Runoff curve number for antecedent soil water condition for fallow, crop,
and residue fractions of the growing season, dimensionless
Maximum dry foliage weight of each crop at full canopy, kg/m2
Number and dates of crop emergence, maturation, and harvest for each
cropping period, dimensionless
Pesticide Data
Number of applications, dimensionless
Date, amount, and depth of each application; dimensionless, kg/ha, cm
Type of application, dimensionless
For foliar application: decay rate on foliage, foliar extraction coeffi-
cient, and filtration parameter; days'1, cm'1, m2 kg'1
Pesticide solubility; may be used to estimate partition coefficient by any
of 3 methods, mole fraction or mg I'1 or nmoles I'1
Pesticide/soil sorption partition coefficient per horizon, cm3 g-1
Pesticide decay rate for each horizon, days-1
Soil Data
Depth of core, cm
Plant uptake efficiency, dimensionless
Total number of soil horizons and compartments
Thickness, bulk density, hydrodynamic dispersion,
initial soil water content, and drainage parameter per horizon; cm,
g cm'3, cm3, day1, cm~3, day -1)
Field capacity, wilting point, organic carbon content, percent sand and
clay in each soil horizon; cm3, cm'3, cm3, cm'3, percent, percent, percent
Initial pesticide level in each compartment, mg kg'1 or kg ha'1
==========================================================================
4-3
-------
TABLE 4-2. THE MAJOR OUTPUT PREDICTION FUNCTIONS CURRENTLY
PERFORMED BY PRZM
1. Calculation of soil-water characteristics based on textural properties
2. Calculations of Kjj based on water solubility models
3. Echo of inputs to output files
4. Determination of crop root growth
5. Specific time series data output
6. Crop interception of rainfall
7. Division of precipitation between rain and snow
8. Calculation of evapotranspiration
9. Snowmelt computation
10. Calculation of plant uptake factors
11. Determination of curve number from cropping period and soil moisture
12. Computation of runoff and infiltration
13. Calculation of soil hydraulics
14. Calculation of pesticide transport in soil
15. Total pesticide application
16. Water and pesticide mass balance computation
17. Output of fluxes, storages, etc
18. Input checking
19. Foliar pesticide application decay and washoff
20. Soil erosion and erosion pesticide loss
STEP 3: FIELD TESTING FEASIBILITY
PRZM was designed as a site specific model using field data (Carsel et al.,
1984). The user's manual and the model itself also lend themselves to determi-
nation of needed parameters by several alternate methods. For example, field
capacity, wilting point, and saturation soil-water contents may be calculated
from soil textural information. Also, the sorption coefficient, Kj, may be
calculated from chemical's aqueous solubility by any of three different methods.
The output variables that need to be measured for comparison with model output
are the soil-water content, pesticide content in the soil column, and pesticide
concentration in the soil water. These variables can be measured from soil
cores and soil-water samples collected with suction lysimeters (Wilson, 1980).
The amount of pesticide leaching past the root zone, which may be measured
using the above procedures, is that amount of pesticide which may potentially
contaminate ground water.
STEP 4: CRITERIA OF ACCEPTANCE
The participants of the Predictive Exposure Assessment Workshop sponsored
by the U.S. EPA in Atlanta, Georgia, on April 27-29, 1982, concluded that for
screening applications, a model should be able to replicate field data (concen-
tration profile, total mass, flux past root zone, soil-water content and storage,
etc.) within an order of magnitude and site specific applications within a
factor of 2. The Dougherty Plain project staff elected to use similar cri-
teria. Therefore, if the model simulates the field data outside more than a
4-4
-------
TABLE 4-3. SENSITIVE PRZM INPUT PARAMETERS
======5=====================================================
Category ' Parameter
Transport KD (Sorptlon Coefficient)
BD (Bulk Density)
THEFC (Field Capacity)
THEWP (Wilting Point)
CN (Curve Number)
Supply RA (Application Rate)
KS (Decay Rate)
AL (Active Layer-Root Zone)
factor of 2, the model will need to be modified to the extent needed to meet
this criterion.
STEP 5: DETERMINE FIELD VALIDATION SCENARIO
The example study is a field validation under natural conditions in the
Dougherty Plains area of southwest Georgia. PRZM has been tested a number of
times previously (see Donigian and Rao, Chapter 5; however, most of these
studies utilized data that had been collected for purposes other than model
testing. It was decided that a field validation under natural conditions would
offer maximum credibility and acceptance in the modeling community.
STEP 6a: PLAN AND CONDUCT FIELD TEST
Select Study Area and Compound(s)
The following criteria were used in the selection of a representative
field site for evaluation of PRZM. The study area should (1) be in a major
agricultural area with potential for leaching to ground water; (2) have rela-
tively easy access to the Athens lab, (3) should be composed of multilayered
soils in order to adequately test the model, (4) should be relatively flat
topography in order to minimize runoff, thus simplifying mass balance determin-
ations as well as making it less expensive to monitor, (5) should be a manage-
able size (less than 8 ha), (6) should have a shallow water table, (7) should
be isolated from domestic water supplies for safety reasons, (8) should be
located within close proximity to a ground-water divide to assure unidirec-
tional ground-water flow, and (9) should be available for lease for a period of
five years in order to obtain an adequate amount of data. An Interagency
Agreement was developed between the U.S. EPA in Athens, Georgia, and the USGS
District Office in Doraville, Georgia. The USGS has a field office in Albany,
Georgia, which had just completed a five-year irrigation study, so staff were
familiar with fields in the area and knew many of the land owners. Using the
criteria set up by EPA describing a desirable field site, the USGS identified
several potential study areas. These sites were then visited and ranked by
4-5
-------
EPA. A 3.9 ha field was selected In southeast Lee County (near Albany) In the
Dougherty Plain area of Georgia (Figure 4-1), and a five-year lease agreement
obtained with the land owner. The topography of the area is gently undulating
uplands and plains, and low-lying plains that are either well drained or swampy
(Owen, 1963). The relief of the Dougherty Plain is low to gently sloping with
maximum relief not exceeding 24 meters and slopes of less than 2 percent.
Agriculture accounts for 80 percent of the land use within Dougherty Plain
(Cooper et al., 1985).
In order to avoid legal complications, the pesticides selected for use
would need to be registered for land application to the selected crop. The
initial chemical chosen should be sufficiently mobile to assure that chemical
movement will be observed during the study period; if additional chemicals are
chosen they should span a range of lesser mobilities.
Two pesticides were chosen according to the previously established crite-
ria. They included the insecticide, aldicarb 2-methyl-2 (methylthio) prop-
ionaldehyde 0-(methyl-carbamoyl)oxine, and the less soluble less mobile herbi-
cide, metolachlor 2-chloro-N-(2-ethyl-6-methylphenyl)-N-(2-methoxy-l-methyl-
ethyl)acetamide. Both aldicarb and metolachlor are widely used in the Dougherty
Plain area. The selected crop was peanuts (Arachis hypogaea).
Aldicarb
Aldicarb is a systemic carbamate insecticide used in the control of soil
insects and nematodes in cotton, peanuts, potatoes, and other crops (Farm
Chemical, 1981). The registered use of aldicarb on peanuts is summarized in
Table 4-4. The chemical is applied to soil in the form of granules containing
10 percent or 15 percent active ingredient (AI). It is a noncorrosive, non-
flammable compound that is stable except in the presence of strong alkali
(Martin, 1971). Aldicarb is metabolized by both oxidative and hydrolytic
processes. Oxidation of aldicarb produces toxic derivatives of aldicarb sul-
foxide and aldicarb sulfone. Aldicarb sulfoxide and aldicarb sulfone are both
acetycholinesterase inhibiting compounds. The term "Total Toxic Residues"
(TTR) has been used to refer to aldicarb and its derivative compounds (Richey
et al., 1977). Aldicarb and the sulfoxide may be oxidized to the sulfone which
is then analyzed by HPLC as TTR (Cooper et al., 1985). Table 4-5 lists some of
the chemical and physical properties of aldicarb and its sulfoxide and sulfone
derivatives.
A number of laboratory studies have examined the fate of aldicarb residues
in soils (e.g. Coppedge et al., 1967, Bull et al., 1970, Campbell et al., 1971).
Supak, 1972, conducted laboratory investigations on the volatilization, deg-
radation, adsorption, and desorption characteristics of aldicarb in soils and
clays. Field studies on the fate and persistence of aldicarb and its metabo-
lites In soils have been conducted by Kearby et al. (1970); Andrawes et al.
(1971); and Quarishi (1972); Hornsby et al. (1983); Jones et al. (1983); and
Jones and Back (1984).
The Union Carbide Corporation has collected laboratory and field data
concerning the environmental impact of aldicarb. U.S. EPA (1975) reported that
aldicarb is oxidized in the soil to the toxic metabolite aldicarb sulfoxide.
4-6
-------
84C
85
Kilometers
0 10 20 3O 40 50 60 70
i i i i i i i i
field location
Kilometers
Figure 4-1. Study site location.
-------
TABLE 4-4. REGISTERED AGRICULTURAL USES OF ALDICARB ON PEANUTS
(from U.S. EPA, 1975)a
===============================================================================
Ounces of Formulation
per 1000 ft of rowb Pounds of Active
Pests Ingredient Per
Controlled 10% 15% Acre (range) Application
Trips
11 to 22
7.5 to 15
1.0 to 2.1
Nematodes
(root-knot
ring, lesion,
string, stunt,
spiral and
stubby-root)
22 to 33
15 to 22
2.0 to 3.0
Apply granules
in seed furrow
and cover with
soil. In south-
west use high
rate only.
Apply granules
in a 6- to 12-
inch band and
work into the
soil or cover
with soil to a
depth of 2 to
4 inches.
Plant seeds in
the treated
zone.
a0.05 ppm (peanuts), 0.5 ppm (peanut hulls), 90 day preharvest interval. Use
high rate of heavy organic or clay soils. Applied as soil application at plant-
ing. Granules should be worked into the soil to a depth of at least 2 inches
or covered with soil to a depth of at least 2 inches to provide maximum per-
formance and minimize the hazard to birds. Do not allow livestock to graze in
treated areas before harvest. Do not allow hogs to root in treated fields.
Do not feed peanut hay or vines to livestock. Do not plant any crop not
listed on label in treated soil within 100 days of last application. Do not
use in house or home gardens.
^Pesticide applied at planting using 36-inch row spacing.
The sulfoxide is largely hydrolyzed to nontoxic oximes and nitriles while a
small percentage is oxidized to aldicarb sulfone. The sulfone also degrades
into nontoxic nitriles.
Metolachlor
Metolachlor is a selective herbicide used for weed control in corn, soy-
beans, peanuts, and certain other crops (Farm Chemical, 1981). The acute
toxiclty of metolachlor to mammals is shown In Table 4-6.
4-8
-------
TABLE 4-5. CHEMICAL AND PHYSICAL PROPERTIES OF ALDICARB AND ITS
SULFOXIDE AND SULFONE DERIVATIVES3
Property
Melting point
CO
Vapor pressure
(25°C, mm, Hg)
Solubility (%, 27°C
• Water
Acetone
Trlchloromethane
Methyl alcohol
Oral LDsg, rats
(mg/kg)
=================
Aldicarb
98-100
io-4
)
0.6
28.0
38.0
NA
1.0
=======================
Aldicarb
Sulf oxide
108-110
10-4.15
25.0
27.0
39.0
56.0
< 1.0
Aldicarb
Sulfone
134-136
10-4.25
0.7
5.1
3.2
3.0
10.0
Clay reaction
Biological decay
Negative
montmorillonite
7-56 days,
first order
Negative
montmorillonite
7-56 days,
first order
positive koalinite
50-1900 days,
first order
===============================================================================
3Supak, 1972; Cooper et al., 1985.
TABLE 4-6. TOXICOLOGICAL PROPERTIES OF METOLACHLOR3
Metolachlor
Technical
Dual 8E
Acute oral LD^Q
Acute dermal LDo (Rabbit)
Acute aerosol inhalation LCso (Rat)
(4-hour exposure)
2780 mg/kg
>3100 mg/kg
>1.75 mg/1
2534 mg/kg
>3000 mg/kg
>0.94 mg/kg
technical Bulletin, Dual Herbicide Agricultural Division, Ciba-Geigy Corp.,
Greensboro, NC, 1980.
4-9
-------
Because metolachlor 1s a relatively new compound, limited information is
available. Several studies have shown, however, that it is readily sorbed to
clay and organic matter. The Chemistry Branch at Athens developed a GC method
for analyzing metolachlor in soil. Metolachlor will be used only in applica-
tion monitoring.
STEP 6b: DEVELOP A FIELD STUDY DESIGN
The Dougherty Plain project was designed to: (1) monitor leaching, (2)
evaluate spatial variability of pesticide application and soil properties, and
(3) establish sorption coefficients and degradation rate constants for the two
selected pesticides. The experimental area consisted of a 3.9 ha field with a
9.1 m (30 ft) border around the edge of the field to eliminate potential boundary
influence {Figure 4-2).
A soil survey of the field conducted by the local SCS office was used as
the basis for selecting the primary monitoring sites and additional soil sam-
pling sites. The total area in each soil series is shown in Table 4-7. The
sampling size of 30 has been determined to be a reasonably large number of sites
to adequately monitor a field (Bresler and Green, 1982). Due to management
practicality and resource limitations, the 20 sites shown in Figure 4-2 were
finally selected. The degree of precision lost in going from 30 to 20 sites
was within an acceptable range according to Bresler and Green (1982). Subjec-
tively, this number also seemed to provide reasonable coverage of the field
especially insofar as preliminary investigation was concerned. Additional infor-
mation on the Dougherty Plain field site is provided by Carsel et al., (1985).
Series
TABLE 4-7. AREA (ha) PER SOIL SERIES IN DAUGHERTY PLAIN, GEORGIA
=====================================================================
Total
Ardilla Clarendon Tifton Lucy Field
Area within
9.1-meter border
Area outside
0.90
1.43
0.65
0.14
3.12
9.1-meter border
Total acreage
0.22
1.12
0.31
1.74
0.21
0.86
0.04
0.18
0.78
3.90
The area of each of the three major soil series identified in the field
was divided by the total field area to calculate relative proportions of area
for each soil series in Table 4-8. The 20 sampling sites were allocated to the
three series according to these proportions. Four additional sites were selec-
ted for soil characterization so that a total of 24 sites were used for soil
characterization.
4-10
-------
Soil Series
Clarendon
Tifton
Ardilla
Lucy
Primary Monitoring Sites (20, ranked)
0 15 30 meters
16 17
18
Figure 4-2. Dougherty Plain field site.
-------
TABLE 4-8. SAMPLE ALLOCATION BY SOIL TYPE IN DAUGHERTY PLAIN, GEORGIA
No. Samples for No. Candidate No. Samples for
Soil Type Proportion Primary Sites Grid Points Additional Sites
Tifton
Ardilla
Clarendon
0.23
0.30
0.47
5
6
9
20
30
46
1
1
2
A 15.2 meter (50 foot) grid was used to select the array of sampling
sites. Grid points that were within 6.1 meters (20 feet) of a soil series
boundary were eliminated from consideration. This yielded a total of 96
candidate grid points out of the total of 134 grid points. The number of
candidate grid points in the three series is shown in Table 4-8. The samples
for each soil series were selected randomly from the candidate points for that
soil series. The resulting set of 20 selected sites became the primary sam-
pling sites; proportional allocation of the 20 random sites produced 5, 6, and
9 samples for the Tifton, Ardilla, and Clarendon series, respectively.
Alternate sites were then randomly selected from the remaining grid points.
These sites were to be used in the event that one of the primary sites proved
to be unacceptable. Permanent reference points were surveyed along the perime-
ter of the field in the area outside the 9.1 meter border. The reference
points were to be used in conjunction with a transit to relocate sampling
equipment after planting in the spring of each study year.
Soil core samples for physical characteristics (sand, silt, clay, organic
carbon, and pH) were taken at 24 sites at four depths to 1.2 m. The analysis -
was done by the University of Georgia Soil Test Laboratory. Background soil
core samples for aldicarb and metolachlor were taken at the 20 primary sites;
160 samples were taken in 15 cm increments to 1.2 m. Samples were analyzed by
the OPP Laboratory at Beltsville, Maryland, for aldicarb and at the Environ-
mental Research Laboratory, Athens, for metolachlor. Samples for degradation
and sorption studies were taken with well drilling equipment in duplicate at
four depths to 1.2 m using the method described by Wilson (1980). The sorp-
tion and degradation parameters for aldicarb and metolachlor using these soil
samples were measured by Rao et al. (1986). Undisturbed core samples were
taken from a minimum of seven sites on the field perimeter representing each
soil series for determination of bulk density and soil-water characteristic
curves. Pits were excavated to a depth of 1.2 m, and samples were taken from
the sidewalls of the pits. Three soil cores and one bulk soil sample were
collected from each horizon. The analysis was done by the R.S. Kerr Environ-
mental Research Laboratory in Ada, Oklahoma. Further background water samples
were taken (10 in the residium and 4 in the Ocala aquifer). Analysis was done
by U.S. EPA in Beltsville and Athens.
4-12
-------
Twenty-one permanent wells were installed at 15 of the 20 primary sites
selected by priority number. In the shallow saturated zone, these wells were
located at depths of 4.6 and 6.2 meters, and four additional wells were located
outside the perimeter in the Ocala aquifer at 53.3 meters. These wells were
established to characterize the hydrology of the site by providing information
on ground-water depth and direction of the flow as well as indicating if any
chemical leaches to ground water. The ground water data do not input to PRZM
but do further characterize the site. The use of suction lysimeters reduces
the time and manpower needed for sample collection; it also helps preserve the
integrity of the field. For monitoring by use of suction lysimeters, see
Wilson (1980). Figure 4-3 shows a schematic of a monitoring site.
A weather station equipped with a hygrothermograph, an evaporation pan, an
anemometer, and two rain gauges was installed on the north side of the study
field. This permitted measurement of air temperature, relative humidity,
evaporation, wind speed, and precipitation on a continuous basis.
Monitor Leaching
Soil cores were collected at each sampling site to a depth of 1.2 m (in
15-cm increments). Samples were collected from suction lysimeters located at
even intervals between the coring depth and the normal water tables at 1.5,
2.1, and 2.7 meters, and at wells located at depths of 4.6 and 6.2 meters.
Spatial Variability
The spatial variation of aldicarb (granular formulation) application at
the time of planting was measured by collecting soil samples on the same day
along the band where aldicarb was applied. The samples were collected by
pressing an aluminum can into the soil. Samples were systematically taken in a
three row square grid at each sampling site.
The spatial variation of metolachlor application (emulsifiable concentrate)
was measured by placing 200 (18.5 cm diameter) filter discs in duplicate at 100
sites at even intervals in the field prior to application using a systematic
sampling plan. The use of filter discs to monitor pesticide application using
boom-type spraying equipment has been successfully demonstrated in previous
field studies (Smith et al., 1978). Filter disc as well as other techniques for
pesticide application monitoring are discussed in greater detail by Smith et al.,
(1985). Metolachlor was limited to application monitoring due to resource
constraints.
Degradation and Sorption
Degradation and sorption studies for both aldicarb and metolachlor were
conducted under a cooperative agreement between the U.S. EPA and the University
of Florida. Other than the amount of precipitation, these two parameters are
the most important in determining if a pesticide will leach. Well-drilling
equipment was used to collect duplicate samples at each of the twenty primary
sites at four depths. Samples were taken before the start of the main study.
Cores were extracted using a Frazier sampler supplied by the Robert S. Kerr
Environmental Research Laboratory of the U.S. EPA in Ada, Oklahoma. Using
4-13
-------
Tensiometers Thermistors Lysimeters
4^£
_ 1.5-
o>
u
<0
•c
3
« 30-
5
o
45-
60-
Unsaturated
Zone
Saturated
Zone
Figure, 4-3. Schematics of monitoring site.
-------
aseptic procedures (Wilson, 1980), the first five cm of each soil core was
discarded in order to avoid contamination, and the center of the soil core was
extracted into sterile Mason jars for each depth sample Increment. Degradation
rates were determined as a function of depth. The outside parts of the soil
core were collected for sorption studies. Soil samples were used to determine
the field spatial variation of sorption coefficient, organic carbon content,
and pesticide degradation rate coefficients (Rao et al., 1986).
STEP 6c: CONDUCT FIELD STUDY
In addition to characterization and background work activities in the main
field during 1983, two 15.2 by 15.2 meter test plots (called SP1 and SP2) were
planted in an adjacent field and treated with aldicarb. The test plots were
used to evaluate the effectiveness of the soil solution samplers (lysimeters),
to develop soil sampling techniques, and to gather background data on aldicarb
movement in soils. These preliminary field activities will tend to modify and
fine tune the original field study design.
These two study plots were located in a 8.1 ha field directly north of the
main study area. Prior to planting, the field was treated with a mixture of
Balan and Vernam 7E. These are pre-emergent herbicides incorporated in the
soil for seasonal control of annual grasses and broadleaf weeds. On May 20,
1983, the farm owner planted the field with 1200 pounds of peanuts (Florunner
variety) and applied aldicarb at a rate of 0.56 kg ha~* active ingredient. The
peanut rows were spaced 0.9 meter apart. Shortly after planting, a mixture of
herbicides Alachlor and Dyanap was surface applied to the field. In addition,
Bravo, a fungicide, plus elemental sulfur were foliar-applied to the peanut
crop approximately one month later.
Soil core samples were collected for aldicarb residue analysis at 6, 20,
and 38 days after planting and application of aldicarb. Soil samples were
taken within two days after rainfall events occurred so as to allow time for
soil to reach field capacity soil-water content. Four soil cores were taken
from each plot to a depth of 1.2 meters. Soil samples were taken on two
adjacent rows, and additional samples were taken at 30-cm intervals across the
0.9-meter width perpendicular to the rows of peanuts. The same rows were
sampled in both plots, SP1 and SP2. A five cm diameter core barrel attached
to a Giddings hydraulic soil sampler mounted on the back of a four-wheel drive
truck was used to extract the soil cores. Two cores were collected on the row
of peanuts where aldicarb was previously applied, and two cores were collected
between the rows.
After each soil core was collected, a metal extracting rod was used to
push the core out into a stainless steel sample tray. Between samples, the
core barrel, the extracting rod, and the sample collection tray were scrubbed
and rinsed with acetone. The core was subdivided into 15-cm sample increments
and then transferred into aluminum cans and mixed thoroughly. The cans were
sealed with plastic tape. Labels on each can listed the sample location,
sample depth, and the date. The metal cans then were transferred into plastic
bags; the bags were sealed and placed in coolers of ice. All samples were
frozen prior to shipment to the analytical laboratory.
4-15
-------
Two types of lysimeters were considered for use In monitoring the unsat-
urated soil zone. These experimental lysimeters were installed in SP1 to test
and evaluate their respective effectiveness and efficiency. The lysimeters
under consideration included (1) a polyvinyl chloride (PVC) plastic pipe fitted
at one end with a porous ceramic cup and with a two-holed rubber stopper (for
polyethylene tubing) attached to the opposite end, and (2) an all Teflon® type
device including filter. Both lysimeters were found unacceptable for use in
this study. The PVC lysimeter sorbed chemical while the Teflon* lysimeter had
a very low air-entry value. Therefore, a stainless-steel lysimeter was de-
signed, tested, and then after proving acceptable installed in the main field
(Smith and Carsel, 1986).
During the 1984 growing season, samples were collected for aldicarb and
metolachlor consisting of (1) a field "shakedown" process for development of
sampling techniques, shipments/storage practices, calibration/equilibration of
monitoring equipment, and definition of team responsibilities between project
participants; and of (2) quantitatively monitoring the movement of pesticide
residue within the soil profile for total toxic residue (TTR) for aldicarb.
Samples were also taken to evaluate the spatial variability of a granular
pesticide application technique (aldicarb) and the spatial variability of a
surface application technique (metolachlor).
Soil and ground-water samples were collected in February 1984 from the
main (3.9 ha) site for background residue analysis. A soil core was taken to
a depth of 1.2 meters at each of the 20 monitoring sites. The soil core was
subdivided into 15-cm depth increments that were transferred into aluminum
cans, sealed, labeled, and placed in coolers of ice.
Ground-water samples were collected from 10 stainless steel monitoring
wells, selected by priority number, in the residue and from the four wells open
in the Ocala aquifer. The wells were evacuated with a small portable gasoline-
powered pump for approximately 15 minutes to remove at least one bore-volume of
water from the well. A one-liter volume of sample was collected by an all-
Teflon® point-source bailer. The samples were transferred into one-liter Nalgene*
bottles, and were then placed in coolers of ice.
After metolachlor application, the filter discs were collected and sealed
in Mason jars, placed in ice coolers, and sent to the analytical laboratory
for analysis.
All samples collected for degradation or sorption studies were stored in
ice chests and shipped to the University of Florida for analysis.
Soil cores were taken during the growing season for aldicarb analysis
within two days after major rainfall events occurred.
All soil core and ground-water samples were frozen prior to shipment in
order to preserve sample integrity. The frozen samples were shipped in ice to
the Beltsville Laboratory for background residue analysis of aldicarb.
4-16
-------
Two to three plants were taken for analysis from each sampling site for
analysis In order to evaluate pesticide loss by plant uptake. They were frozen
and shipped In Ice with the soil cores and water samples.
To effectively collect sufficient data that will provide a quantitative
statement of both model performance and pesticide leaching in Dougherty Plain
soils, a minimum of three more years of the present, ongoing, highly charac-
terized pesticide monitoring should be completed. The continuation of the
project through 1987 should provide an adequate amount of acceptable data for
model testing.
In addition to a continuation of the present sampling program, pesticide
leaching studies to resolve the following are anticipated in the 1986 and
1987 crop years:
° Soil - identification of aldicarb lateral plume movement from the
band of application.
0 Soil - evaluation of various size soil sampling probes for the recovery
of aldicarb residues.
0 Water - additional saturated zone sampling to characterize any aldicarb
movement in the ground water.
Unfortunately, the 1984 analytical results were not available before the
1985 cropping year started and, therefore, any needed modifications could not
be made. They are now available, and as soon as the data are processed, simu-
lations using these data will be made. It is hoped that the 1985 data will
also be available before the 1986 cropping year begins.
STEP 6d: SAMPLE ANALYSIS AND ANALYTICAL QUALITY ASSURANCE
The Beltsville Chemistry Laboratory of the Office of Pesticide Programs
of the EPA was responsible for screening water and soil samples for aldicarb *
during the 1984 cropping year. The Analytical Chemistry Branch of the Athens
Environmental Research Laboratory was responsible for the analysis of metol-
achlor on filter discs. A small number of peanut plant samples were also
considered to evaluate plant uptake. To assure accurate and reliable analyt-
ical data, a specific quality assurance protocol was established (see Cooper et
al., 1985) to be monitored by the Environmental Systems Branch of the Athens
laboratory.
Samples arriving at the Beltsville facility were shipped in ice. As sam-
ples were received, they were logged in a book dedicated to the project and
secured under lock and key. Samples remained frozen until time of analysis.
All reserve sample extracts were stored frozen until disposition was determined.
Sample number identification was assigned at Athens ERL. A common notebook was
provided in which all entries were made. All raw data (chromatograms, mass
spectra, etc.) were also collected and kept in a file dedicated to the project.
4-17
-------
STEP 6e: COMPARE MODEL PERFORMANCE WITH ACCEPTANCE CRITERIA
At the'time of this writing, the data from 1984 are still being processed,
and 1985 sampling is still in progress. Therefore, although the model has been
calibrated from the preliminary data, no testing or validation attempts have
been made at this point. Comparisons will be made between the hydrologic and
chemical outputs separately. The predicted and observed soil-water content
will be compared as will be evaporation losses. If output is outside acceptance
limits, an attempt at fine tuning the calibration will be made. If this does
not bring the predicted values within acceptance limits, the structure of the
model itself will need to be investigated. Similarly, the amount of pesticide
in the soil compartments and in the soil water will be compared with model
output. Also, if model predictions do not agree with the acceptance criteria
(factor of two), the model will be fine tuned, and, this failing, modification
in the model will need to be made.
4-18
-------
REFERENCES
Andrawes, N. R., W. P. Bagley, and R. A. Henett. 1971. Fate and carryover
properties of temik aldicarb pesticide. J_. Agrlc. Food Chem. 19:727-730.
Bresler, E., and R. E. Green. 1982. Soil Parameters and Sampling Scheme for
Characterizing Soil Hydraulic Properties of a Watershed. University of
Hawaii, Technical Report No. 148.
Ball, D. L., R. A. Stokes, J. R. Coppedge, and R. L. Rodgers. 1970. Further
studies of the fate of aldricarb in soil. J_. Econ. Entomol. 63:1283-1289.
Campbell, W. V., D. A. Mount, and B. S. Heming. 1971. Influence of organic
matter content of soils on insecticidial control of the wireworm. J_.
Econ. Entomol. 64:41-44.
Carsel, R. F., L. A. Mulkey, M. N. Lorber, and L. B. Baskin. 1985. The
Pesticide Root Zone Model (PRZM): A procedure for evaluating pesticide
leaching threats to ground water. Ecolog. Model. 30:49-69.
Carsel, R. F., C. N. Smith, L. A. Mulkey, J. D. Dean, and P. P. Jowise. 1984.
User's Manual for the Pesticide Root Zone Model (PRZM): Release 1.
EPA-600/3-84-109. U.S. Environmental Protection Agency, Athens, GA.
Cooper, S. C., R. F. Carsel, C. N. Smith, and R. S. Parrish. 1985. Pesticide
Migration in the Unsaturated and Saturated Soil Zones. Part I. A Field
Study to Support Model Development and Testing, Dougherty Plains, Georgia
(unpublished report). U.S. Environmental Protection Agency, Athens, GA.
Coppedge, J. R., D. A. Lindquist, D. L. Ball, and H. W. Dorough. 1967. Fate
of 2-methyl-2-(methylhics) propion-aldeleyde 0-methyl-carbonyl oxide
(Temik) in cotton plants and soil. J_. Agric. Food Chem. 15:902-910.
Farm Chemical Handbook. 1981. Meister Publishing Co., Willoughby, OH.
Hornsby, A. G., P. S. C. Rao, W. B. Wheeler, P. Nkedi-Kizza, and R. L. Jones.
1983. Fate of Aldicarb in Florida Citrus Soils: 1. Field and Laboratory
Studies, Proceedings of the NWWA/U.S. EPA Conference on Characterization
and Monitoring of the Vadose (Unsaturated) Zone, Las Vegas, NV,
December 8-10, 1983.
Jones, R. L., and R. C. Back. 1984. Monitoring aldicarb residues in Florida
soil and water. Environ. Toxicol. Chem. 3:9-20.
4-19
-------
Jones, R. L., P. S. C. Rao, and A. 6. Hornsby. 1983. Fate of Aldicarb in
Florida Citrus Soils: 2. Model Evaluation. Proceedings of the NWWA/
U.S. EPA Conference on Characterization and Monitoring of the Vadose
(Unsaturated) Zone. Las Vegas, NV. December 8-10, 1983.
Kearby, W. H., C. D. Ercogovich, and M. Bliss, Jr. 1970. Residue studies on
aldicarb in soil and scotch pine. J_. Econ. Entomol. 63:1317-1318.
Martin, H. (Ed.). 1971. Pesticide Manual, 2nd edition. British Crop
Protection Council.
Owen, V. Jr. 1963. Geology and Ground-Water Resources of Lee and Sumter
Counties, Southwest Georgia. U.S. Geological Survey, Mater Supply Paper
1666.
Quarishi, M. S. 1972. Edaphic and water relationship of aldicarb and its
metabolites. Canad. Entomol. 104:1191-1196.
Rao, P. S. C., K. S. V. Edwardson, L. T. Ou, R. E. Jessup. P. Nkedi-Kizza,
and A. G. Hornsby. 1986. Spatial Variability of Pesticide Sorption and
Degradation Parameters. In; R. Honeycutt (Ed.), ACS Symposium Series
American Chemical Society, Washington, D.C. (In Press).
Richey, F. A., Jr., W. J. Bartley, and K. P. Sheets. 1977. Laboratory studies
on the degradation of [the pesticide] aldicarb in soils. J. Agric. Food
Chem. 25:47-51.
Siegel, S. 1956. Non-Parametric Statistics for the Behavioral Sciences.
McGraw-Hill, New York.
Smith, C. N., and R. F. Carsel. 1986. A Stainless Steel Suction Lysimeter
for Monitoring Pesticides in the Unsaturated Zone. Soil Sci. Soc.
Amer. J_. 50:263-265.
Smith, C. N., D. S. Brown, J. D. Dean, R. S. Parrish, R. F. Carsel, and A. S.
Donigian, Jr. 1985. Field Agricultural Runoff Monitoring (FARM) Manual.
EPA-600/3-85-043. U.S. Environmental Protection Agency, Athens, GA.
Smith, C. N., R. A. Leonard, G. W. Langdale, G. W. Bailey. 1978. Transport of
Agricultural Chemicals from Small Upland Piedmont Watersheds. EPA-600/
3-78-056. U.S. Environmental Protection Agency, Athens, GA.
Supak, J. R. 1972. The Volatilization, Degredation, Adsorption and Desportion
Characteristics of Aldicarb [2-methyl-2-(methylthio propionaldehyde
0-methylcarbomyl oxime] in Soils and Clays. Ph.D. Thesis, Texas A and M
University.
U.S. EPA. 1975. Clinical Scientific and Microeconomic Review of Aldicarb.
Washington, DC.
4-20
-------
Wilson, L. G. 1980. Monitoring in the Vadose Zone: A Review of Technical
Elements and Methods. EPA-600/7-80-140. U.S. Environmental Protection
Agency, Las Vegas, NV.
4-21
-------
CHAPTER 5
EXAMPLE MODEL TESTING STUDIES
by
A. S. Donigian, Jr., and P. S. C. Rao
Most model testing efforts with pesticides have not had the advantage of a
comprehensive data collection program such as the Dougherty Plain Field Study
which was specifically designed to provide a.reasonably adequate data base for
model testing and validation. Although numerous field studies have provided
and are currently providing data on pesticide fate in the vadose zone, much of
the available data are of limited use for model evaluation. A frequent mis-
match occurs between the needs of the experimentalist who designs the field
studies and the needs of the modeler for model testing (Wagenet and Rao, 1985).
This manual is aimed at helping bridge this gap by informing and educating
these two groups regarding the data needs and problems of each group so that
future data collection efforts will more fully meet the needs of both the
experimentalist and the modeler.
At present most model testing and validation studies must rely on a lim-
ited data base of a few field studies designed to collect data for a variety of
purposes. This section discusses and summarizes a few selected model testing
studies of this type which relied on available field data. These studies
demonstrate the types of comparisons that are often made between field data and
model predictions, the procedures required for model testing, and the model
performance or acceptance criteria (or statistical tests) used to quantify
model performance. Since previously collected data are utilized, this section
demonstrates only steps 4 and 6e in the field validation procedures discussed
in Chapter 3.
The remainder of this chapter is based primarily on two recent studies
aimed at testing and evaluating the three models discussed in Chapter 1. The
first, by Watson and Brown (1984), involved a comprehensive evaluation of the
SESOIL model. They compared SESOIL model predictions with published field data
for aldicarb nematicide and with the predictions by Jones et al. (1983) of
three other models (PRZM, PESTAN, and PISTON). The second study reviewed here
is that of Carsel et al. (1985b), who evaluated the PRZM model using field data
for metalaxyl pesticide. We will summarize the findings of these two studies
here in order to illustrate the problems encountered in testing models using
field data, and those encountered in the testing methodologies used to evaluate
model performance. There are also a number of other on-going model testing
studies aimed at comparing PRZM with other, more research-oriented, models.
5-1
-------
Testing Methodology used by Watson and Brown
A two-tiered approach was used by Watson and Brown (1984) to test SESOIL.
The first tier involved making comparisons between SESOIL predictions and re-
sults from an analytical solution to the partial differential equation describ-
ing solute transport in the unsaturated zone; a closed-form solution to the
one-dimensional, convection-dispersion equation was used (van Genuchten and
Alves, 1982). The second tier of testing involved comparing SESOIL predictions
with field data and with predictions of other models tested with the same field
data sets. Comparisons were made with two types of data sets. The first type
was data collected on chemical leaching in the unsaturated zone. The second
type was data collected during pesticide runoff studies on field-scale water-
sheds. Only the procedures and results of the chemical leaching comparisons
will be presented in this section; the interested reader is referred to the
original report for the results of the first-tier SESOIL testing (i.e., compar-
isons with the analytical solution) and the pesticide runoff comparisons.
Chapter 1 summarized the primary conclusions of the entire SESOIL model testing
effort.
Jones et al. (1983), identified four major types of predictions that
should be provided by models designed for describing pesticide fate in the
unsaturated zone and for describing the potential for contamination of ground
water. These are listed as follows in decreasing order of importance:
1. the transit time for the pesticide within the unsaturated zone
2. pesticide mass emissions into the ground water
3. pesticide concentration in the ground water
4. pesticide concentration distribution within the soil profile
The testing by Watson and Brown emphasized the first two types of com-
parisons, transit time, and mass emissions (or loading) to ground water. They
also included comparisons of the pesticide mass remaining in the soil profile
at different times following application which are a useful aggregate measure -
of transit time, loading to ground water, and degradation processes. Compari-
sons of pesticide concentrations in ground water was beyond the scope of the
models and testing effort. Comparisons of pesticide concentrations in the soil
profile are also difficult tests due to spatial and temporal variations in soil
characteristics and processes as discussed in Chapter 1 and Appendix G. More-
over, the available field data were not sufficiently detailed to allow an
assessment of spatial variations.
The tests by Watson and Brown were initially conducted with limited cali-
bration of SESOIL parameters since Bonazountas and Wagner (1982, 1984) suggest
that the model can provide reasonable results with limited calibration. Tests
involving calibration and verification were also conducted with those data sets
having sufficiently long records. A split-sample calibration/verification
procedure (EPA, 1982) was used when sufficient data were available. A split-
sample procedure involves dividing the data set into two separate data sets:
one of which is used to calibrate model parameters, and the other is used to
test or verify model parameters.
5-2
-------
EPA-OTS currently uses SESOIL primarily to predict the extent to which a
chemical will leach to ground water. For this reason, the first tier of
testing by Watson and Brown focused entirely on examining the unsaturated zone
transport algorithms in SESOIL. An analytical solution to the one-dimensional,
convection-dispersion equation was selected for testing purposes because: (1)
model behavior for a range of hydrologic conditions and chemical character-
istics could be examined rapidly, and (2) model behavior under relatively
idealized conditions could be examined. The latter reason is perhaps the most
important. Analytical solutions are generally derived assuming idealized
(i.e., homogeneous and isotropic) soil properties, a constant pore-water veloc-
ity (i.e., steady water flow), and idealized chemical characteristics (e.g.,
equilibrium, linear, reversible adsorption, and first-order degradation).
Comparisons of model predictions with analytical solutions are often used to
examine and evaluate the numerical procedures and sensitivity of model param-
eters (ASTM, 1984). By parameterizing SESOIL and the analytical model to
represent the same conditions, similarities and dissimilarities in model behav-
ior were identified.
Field Data for Model Testing
Field data sets useful in testing a model, such as SESOIL, are limited
in number for the following reasons:
1. Most field studies focus on either chemical leaching in the unsatur-
ated zone or runoff losses from watersheds; few field studies have
measured both.
2. Because of the high costs associated with sampling and analysis, most
field studies of chemical leaching are of short duration.
3. Since most field studies are conducted for purposes other than for
generating data to test models, they are often incomplete.
Watson and Brown initially screened available data sets published in the
literature and identified seven candidate data sets for model performance
testing: (1) aldicarb leaching in Florida citrus groves (Hornsby et al., 1983);
(2) aldicarb pollution of ground water on Long Island, New York (INTERA, 1980);
(3) a field study of the transport of agricultural chemicals from the small
upland piedmont watersheds near Watkinsville, Georgia (Smith et al., 1978);
(4) a field study of agricultural chemical transport in the Four Mile Creek
watershed in Iowa (Johnson and Baker, 1980); (5) aldicarb leaching in Wisconsin
(Wyman et al., 1984); (6) a field study of bromide movement in California (Jury
et al., 1982); and (7) radionuclide movement in soils near Hanford, Washington
(Jones and Gee, 1984).
The above data sets were selected for consideration for several reasons.
First, all of the data sets are fairly complete in terms of both hydrologic
and chemical data. Second, measurements of chemical movement in the unsaturated
zone were made in all cases, and chemical losses due to runoff and soil erosion
were measured in two of the studies (i.e., Watkinsville and Four Mile Creek).
Third, the initial mass of chemical (i.e., the source term) is reasonably well
defined; this is particularly true for those field studies involving pesticides.
5-3
-------
Finally, other model tests have been conducted with all of the data sets. The
Florida and Wisconsin aldicarb data sets were chosen by Watson and Brown for
chemical leaching tests; only the Florida data set is briefly discussed below.
The original report by Watson and Brown (1984) includes summary descriptions of
the other data sets listed above.
The Florida field study was undertaken to assess the processes which
control the movement of aldicarb in citrus groves and to compare field and
laboratory-derived degradation and sorption parameters. Hornsby et al. (1983),
describe laboratory and field experiments conducted in two citrus groves, and
Jones et al. (1983), describe the model performance tests conducted with the
data. Field and modeling efforts concentrated on: (1) the transit time for
the aldicarb total toxic residues (TTR) within the unsaturated zone, (2) aldi-
carb TTR mass emissions from the vadose zone, (3) aldicarb concentrations in
the ground water, and (4) aldicarb distributions within the soil profile.
One field site was in Seminole County near Oviedo, Florida. The site is
a bedded grove composed primarily of Immokalee fine sand and Del ray fine
sand. Hornsby et al. (1984), note that the Immokalee series soils are poorly
drained, coarse textured, and strongly acid. The Del ray series soils are
deep, poorly drained or very poorly drained sands with a thick, highly organic
surface layer. The second field site is located in Polk County near Lake
Hamilton, Florida. The soil at this site is a well-drained, deep coarse sand,
typical of the central ridge area of Florida.
Aldicarb was applied to the Oviedo and Lake Hamilton sites on February
15 and 16, 1983, respectively. Soil samples were collected on March 3 and 4,
April 5 and 6, May 2 and 3, June 14 and 15, and August 23 and 24. Soil samples
at the field sites were collected every month throughout 1983; however, data
for a 4-month period following application were available for model testing.
Samples were collected in 30-cm intervals to a depth of 150 cm at the Oviedo
site, in 30-cm intervals to a depth of 60 cm, and in 60-cm intervals to a depth
of 300 cm at the Lake Hamilton site. Laboratory studies were conducted on un-
disturbed soil core samples to obtain soil-water characteristic curves, soil
bulk densities, and saturated hydraulic conductivities. Laboratory studies
were also conducted using bulk soil samples from different horizons to obtain
the data required to estimate sorption coefficients and degradation rate
coefficients.
Aldicarb concentrations in soil samples were found to be highly variable.
Despite this, the data did show detectable concentrations of aldicarb progress-
ing towards the water table with time (Hornsby et al., 1983). Sampling was
continuous throughout the soil profile, but reported TTR concentrations are the
result of composite sampling over 30 or 60 cm intervals. Jones et al. (1983),
tested three different models using these data: (1) PISTON (Rao et al., 1976),
(2) PESTAN (Enfield et al., 1984), and (3) PRZM (Carsel et al., 1984).
As noted above, the Florida and Wisconsin aldicarb data sets were selected
by Watson and Brown (1984) for the second tier of model testing because they
are the most complete of all the chemical leaching data sets reviewed. They
provided the primary test of unsaturated zone hydrology and pollutant transport
cycles in SESOIL. However, the short time frame covered by these data sets
5-4
-------
(i.e., four to six months) precluded the use of a split sample calibration and
verification procedure. Therefore, SESOIL was tested by Watson and Brown with
only limited calibration. Any additional parameter adjustments required to
bring SESOIL results in line with the field data were evaluated for reasonable-
ness.
Quantitative Measures of Performance
In evaluating the performance of a model relative to some field measure-
ment or another model, qualitative rather than quantitative statements are
often made. Statements like "good or reasonable agreement was achieved between
the model and the observed results" are commonplace. Such statements do not
provide a sound basis for judging model accuracy. Dissatisfaction with these
types of qualitative statements have prompted researchers to begin to develop
or adapt quantitative, statistical measures that can be used to generate more
meaningful statements.
Three general procedures have been identified that can be used to provide
quantitative measures of performance for a model (EPA, 1982). These procedures
include:
1. Paired-data performance: the comparison of simulated and observed
data for exact locations in time and space.
2. Time and space integrated, paired-data performance: the comparison
of spatially and temporally averaged simulated and observed data.
3. Frequency domain performance: the comparison of simulated and
observed frequency distributions.
In selecting one or more of these procedures, it is important to consider
the characteristics of the "simulated" and "observed" data. For example, in
this case SESOIL generates the simulated data by calculating seasonally aver-
aged fluxes of water and pollutant mass loading rates. It also calculates
compartmentally-averaged pollutant concentrations at the end of each season.
The observed field data tend to be for specific points in time and gen-
erally represent either points in space or regions (e.g., the upper 15 cm of
soil). Since it is impossible to spatially or temporally disaggregate the
simulated data (i.e., SESOIL results), the observed data have to be aggregated.
Another characteristic of both the simulated and observed data is that they
tend to cover relatively short periods of time (say 4 to 24 months). As a
result, it is difficult to develop a meaningful characterization for concentra-
tion variations with time (e.g., frequency distributions). Given these charac-
teristics, a time and space integrated, paired-data performance procedure by
Watson and Brown (1984) was selected and applied to all comparisons.
A number of statistical techniques can be used to obtain a quantitative
measure of the difference between simulated and observed data. Moore et al.
(1982), discuss different statistical measures and associated aggregate state-
ments useful for evaluating the performance of air quality models. Ambrose
5-5
-------
and Roesch (1982) discuss several measures used to evaluate an estuary water
quality model.
Basically, two classes of statistical models were selected to generate
quantitative measures of performance for the SESOIL tests. The first class is
standard linear regression statistics. As Ambrose and Roesch (1982) note,
regression statistics for observed and simulated data can be calculated by:
O1 = a P1 + b (5-1)
where 0^ = observed data
pi = simulated data
a = slope
b = intercept
The slope, a, provides a useful measure of the accuracy of model results;
values less than 1.0 imply over-prediction by the model, while values less
than 1.0 imply under-prediction. The intercept, b, is an indicator of statis-
tical errors. Perfect agreement between the observed data and model predic-
tions occurs when a = 1 and b = 0. The precision of model results can be
measured by r, the correlation coefficient. It should be recognized that.an
implicit assumption made in this analysis is that the predicted values (P1)
are "fixed" and are without error. This assumption arises from the deter-
ministic nature of the models used, where a single value is assigned for each
of the model parameters resulting in a unique predicted value (see Chapter 1).
Since most, if not all, model parameters are spatially-variable, predicted
values also have certain error. Thus, computation and interpretation of the
linear regression parameters, as shown in eq. (5-1), may not be strictly valid.
Under such conditions, to estimate parameters in linear models Halfon (1985)
has recently recommended the use of geometric mean functional regression
(GMF), which takes into account the errors in.both the dependent and the
independent variables (in this case, Q^ and P1).
The second class of methods are typically called estimation of ejrror
statistics. Ambrose and Roesch (1982) state that the average error, E, and
its associated relative error, RE, can be used to measure accuracy and
systematic errors:
E = I I (P1 -Qi) (5-2)
N 1=i
RE = F/0 (5-3)
where I) = observed data mean
N = number of data points
5-6
-------
The average error gives the absolute amount by which a given quantity is
over- or under-predicted, while the relative error gives the percentage of
over- or under-prediction.
Other estimation of error statistics that can be calculated include the
standard error of estimate, SE, and its coefficient of variation CV, as follows
l N • • Jl/2
E (P1 - O1)2 (5-4)
SE
N
SE
CV = — (5-5)
0
The standard error of estimate is the difference between the actual ob-
served and predicted values, while the coefficient of variation of standard
error gives the average relative difference.
Parameter Estimation and Model Calibration--
A two step approach was used to estimate and calibrate SESOIL model param-
eters. Data from the literature, guidance provided in the SESOIL manual, and
chemical property parameter estimation methods provided the basis for an
initial set of input parameters. SESOIL results based on this parameter set
were then compared with observations on depth-averaged soil -water contents,
percent aldicarb TTR leached to ground water, and percent aldicarb TTR remain-
ing in the soil profile over time.
The second step involved model calibration by adjusting selected input
parameters until reasonable agreement with field data was achieved. Soil -water
content was again the key calibration end point used for the hydrologic cycle.
Evaporation was not measured at either site, and runoff was minimal. Percent
aldicarb TTR leached to ground water was the end point used to calibrate the
pollutant cycle parameters; the high degree of variability in measured vertical"
distributions of aldicarb TTR concentrations in the soil profile precluded the
use of these data for model calibration. As will be discussed later, however,
these data were useful in a qualitative sense for interpreting the performance
Of SESOIL.
The initial hydrologic cycle parameter values for the Lake Hamilton site
are listed in Table 5-1. As was discovered by Watson and Brown (1984), the
intrinsic permeability and saturated hydraulic conductivity suggested in
Table ID-1 of the SESOIL user's manual were not consistent. Using the sug-
gested saturated hydraulic conductivity values, SESOIL predicted soil-water
contents that were too low compared to field measurements; using a conductivity
value derived from the suggested intrinsic permeability value produced soil-
water contents that were too high.
Agreement was improved by increasing the porosity and the disconnectedness
index (see Table 5-1) and by adjusting the saturated hydraulic conductivity to
a value of 0.01 cm/sec; this is equivalent to the laboratory results obtained
by Hornsby et al . (1983). Figure 5-1 shows the resultant SESOIL prediction
5-7
-------
TABLE 5-1. INITIAL AND CALIBRATED HYDROLOGIC CYCLE PARAMETER VALUES
FOR THE LAKE HAMILTON, FLORIDA SITE
Parameter Initial Value Calibrated Value
Hydraulic Conductivity, cm/sec 0.1, 0.001 0.01
Disconnectedness Index, c 3.7 4.1
Porosity, percent 35 43
compared to measured depth-averaged soil-water contents. The error bars
show the minimum and maximum average measured soil-water contents. The cal-
culated estimation of error statistics show that on the average SESOIL
over-predicted soil-water contents by 0.43 percent (E) with a relative error
(RE) of 0.07 (Table 5-2). The standard error (SE) between the predicted and
observed soil-water contents was 0.88 percent.
The same initial set of parameter values was used for the Oviedo site.
Attempts to calibrate SESOIL and to keep parameters in a reasonable range based
on literature data proved unsuccessful. In examining the measured data further
and in discussions with Dr. A. G. Hornsby of the University of Florida, it was
discovered that there was one or more low permeability layers in the soil
column. These layers restrict drainage and produce highly variable soil-water
distributions. Since SESOIL is unable to represent the effect of such layering
on water movement, no further attempts were made to calibrate the model on the
Oviedo site.
An initial set of pollutant cycle parameters was estimated using published
literature data on aldicarb migration and fate in soils and estimation tech-
niques in Lyman et al. (1982). The values are shown in Table 5-3.
Using this initial parameter set, SESOIL predicted that 65.6 percent of
the total mass of aldicarb TTR applied to the site would leach to ground water
during the growing season while Jones et al. (1983), reported that about 4-8
percent of the applied aldicarb TTR had leached to ground water at the Lake
Hamilton site. Runoff and volatilization losses were predicted to be minimal.
The predicted runoff losses are consistent with those observed by Hornsby et
al. (1983); low volatilization losses are also consistent given the chemical
properties of aldicarb.
Calibration of the pollutant cycle parameters focused on adjusting Koc and
hydrolysis rate constants until the predicted mass leached to ground water
decreased within the measured range of 4-8 percent. With a Koc of 240 and a
hydrolysis rate of 0.015/day, SESOIL predicted that 10 percent would reach
ground water. By increasing the hydrolysis rate to 0.0254/day, 4.3 percent was
predicted. The calibrated Koc parameter value is questionable given values
reported in the literature. Hornsby et al. (1983), found that Koc values for
the Lake Hamilton site were between 0 and 47 cnvYgm. With the exception of the
5-8
-------
FIELD
DATA
SESOIL
PREDICTION
in
10
VOLUMETRIC MOISTURE CONTENT (PERCENT)
14
11
10
9
6
7
e
5
4
3
2
1
0
-
-
-
-
f
1 I -r
-
-
-
-
I
-
-
_ ""
I
Ld
-
-
) fo
-
-
-
1 1
FEBRUARY
MARCH
APRIL
MONTH
MAY
JUNE
Figure 5-1. Comparison between SESOIL predicted and measured depth-averaged soil-water
contents for the Lake Hamilton site.
-------
TABLE 5-2. COMPUTED STATISTICS FOR COMPARISONS AT THE LAKE HAMILTON SITE
O1' = mP1 + b
Comparison m b r2 E RE SE CV
SESOIL/Observed
Volumetric Soil-Water
Content, Percent 0.58 2.02 0.38 0.43 0.07 0.88 0.17
SESOIL/Observed
Aldicarb TTR Residues,
Percent 0.88 -10.22 0.77 16.60 0.47 23.46 0.66
TABLE 5-3. INITIAL AND CALIBRATED POLLUTANT CYCLE PARAMETER VALUES FOR
THE LAKE HAMILTON, FLORIDA, SITE
Parameter
Solubility, mg/1
Koc, cm3/gm
Diffusion coefficient in air, cm2/sec
Biodegradation rate, /day
Henry's Law constant, M3atm/mole
Neutral hydrolysis rate, /day
Initial Value Cal
6000
39
0.06
0.0
0.33E-08
0.0051
ibrated Value
6000
240
0.06
0.0
0.33E-08 -
0.025
value found by Hough et al. (1975), for a clay soil, ranges published by
Bromilow et al. (1980), Supak (1972), and Hornsby et al. (1983), are roughly an
order of magnitude lower than the calibrated value. The calibrated hydrolysis
rate is comparable with data presented by Smelt et al. (1978a and 1978b), and
Bromilow et al. (1980), for similar soil, pH, and temperature conditions. It
is at the high end of the range reported by Hornsby et al. (1983), for the Lake
Hamilton site. Because of the extreme parameter values obtained in this "cali-
bration" exercise, it may be concluded that SESOIL does not adequately represent
the aldicarb behavior at this field site.
Figure 5-2 shows a comparison between predicted and observed total aldi-
carb residues in the soil profile using the calibrated parameter values in
Table 5-3. On the average, SESOIL predicts that 16.6 percent more of aldicarb
TTR remains in the soil column (E); the average difference (SE) in predicted
5-10
-------
MEASURED
SESOIL
en
i
% REMAINING
100
100
DAYS AFTER APPLICATION
150
200
Figure 5-2. Comparison of SESOIL predicted and measured percent aldicarb remaining in the soil
column using calibrated parameter values for the Lake Hamilton, Florida, site.
-------
and observed values is 23.5 percent. See Table 5-2 for values of other calcu-
lated performance measures.
Figures 5-3, 5-4, and 5-5 show predicted and observed vertical variations
in aldicarb TTR concentrations with time. Again, the high Koc value required
to limit the rate of leaching to ground water precludes the aldicarb TTR from
migrating vertically to any large extent. The field data, however, again show
a distinct downward progression of the aldicarb TTR pulse.
Comparison with Other Models—
Jones et al. (1983), report on a model evaluation study that was conducted
using the data collected by Hornsby et al. (1983). Three models, PESTAN,
PISTON, and PRZM, were run using equivalent model parameters, and the simula-
tion results were compared with the observed mass of aldicarb leached to ground
water and total residual aldicarb remaining in the soil column over time.
A comparison between SESOIL results and those of Jones et al. (1983), was ob-
tained by using an equivalent set of pollutant cycle parameters in SESOIL (see
Table 5-4). The hydrologic parameters were the same as those obtained in cali-
brating SESOIL against the measured soil-water content data.
Table 5-5 lists the percent of total applied aldicarb TTR leached to
ground water predicted by each of the models. The PISTON and PRZM results are
within the 4-8 percent range measured by Hornsby et al. (1983). The PESTAN
model under-predicted the mass leached to ground water while SESOIL over-
predicted by a factor of 4-10.
A comparison of predicted and measured aldicarb TTR residues in the soil
profile with time is given in Figure 5-6. This figure shows that PESTAN,
PISTON, and PRZM give comparable results while SESOIL initially predicts higher
residues and later predicts lower residues. A comparison of quantitative
measures of performance between PRZM and the observed residues, and between
SESOIL and the observed residues shows that PRZM had only slightly better
agreement than SESOIL (see Table 5-6). The estimation of error statistics are.
similar for both models.
The increase in the SESOIL-predicted rate of aldicarb TTR dissipation
around Day 30 corresponds to the time when aldicarb is initially released to
ground water. If this release had not occurred, SESOIL would have over-
predicted soil residues given the initial rate of dissipation predicted prior
to Day 30. In checking this rate of dissipation, it was discovered that it was
about two times smaller than the hydrolysis rate input to the model. An error
in the hydrolysis algorithm, discussed by Watson and Brown (1984), was later
found to be the source of the problem. Thus, while agreement between SESOIL
predicted and observed soil residues is qualitatively good, it is only because
of the over-prediction of aldicarb leaching to ground water.
5-12
-------
MEASURED
(20 DAYS)
SESOIL
(30 DAYS)
1000
CONCENTRATION PPB
100
01
i
10
I
4 6
DEPTH IN FEET
10
Figure 5-3. Comparison of SESOIL predicted and measured vertical concentration distributions 30 and
20 days after application, respectively, using calibrated parameter values for the
Lake Hamilton, Florida, site.
-------
MEASURED
(45 DAYS)
SESOIL
(60 DAYS)
1000
CONCENTRATION PPB
100
en
i
10
I
I i i— ill
4 6
DEPTH IN FEET
10
Figure 5-4. Comparison of SESOIL predicted and measured vertical concentration distributions
60 and 45 days after application, respectively, using calibrated parameter
values for the Lake Hamilton, Florida, site.
-------
MEASURED
(75 DAYS)
SESOIL
(90 DAYS)
1000
CONCENTRATION PPB
100
tn
!-•
O1
10
10
DEPTH IN' FEET
Figure 5-5. Comparison of SESOIL predicted and measured vertical concentration distributions
90 and 75 days after application, respectively, using calibrated parameter values
for the Lake Hamilton, Florida, site.
-------
TABLE 5-4. POLLUTANT CYCLE PARAMETERS USED IN TESTING PRZM, PISTON, AND
PESTAN ON THE LAKE HAMILTON, FLORIDA, SITE (Jones et al., 1983)
===============================================================================
Parameter Value
KQC, cm3/gm 26.7
Neutral hydrolysis rate, /day 0.019
TABLE 5-5. COMPARISON OF MODEL LEACHING PREDICTIONS FOR LAKE HAMILTON
PESTAN PISTON PRZM SESOIL
Percent of Applied Leaching 0-0.2* 7.4 3.5 35.4
below 10 feet
*Results depend on the value chosen for the curve coefficient. The lower
number represents the model value provided for sand; the higher number is
calculated based on the curve coefficient which gives a soil water content
approximately equal to the field capacity.
The new version of the SESOIL code was also tested using the Lake Hamilton
calibrated parameter values. The new version predicted that slightly less
aldicarb TTR would leach to ground water: 4.0 percent as opposed to 4.3
percent. The addition of a fourth compartment and modifying the model to
retain chemical in each compartment until the pollutant penetration time was
exceeded produced only a small reduction in the amount of chemical leached to
ground water.
The predicted chemical distribution in the soil column was also only
slightly improved. Aldicarb was still retained in the upper layers because of
the high Koc value needed to limit leaching to ground water. Aldicarb con-
centrations in the bottom compartment were slightly lower during the first
couple of months and slightly higher during the last two months.
PRZM Testing by Carsel et al.
Carsel et al. (1985b), evaluated the PRZM model performance using field
data for metalaxyl pesticide collected during the 1980-81 period at two field
sites, one each in the tobacco growing areas of Florida and Maryland. The
soils at the 3.5-ha field in Florida are classified as Blanton sand and are
characterized by high permeability, low water-holding capacity, and low soil
organic matter content. At the 0.6-ha Maryland site, the soils are classified
as Marl ton sandy loam. These soils are characterized by low permeability,
large water-holding capacity, and high soil organic matter content. Hydrologic
and soil characteristics of these two sites are summarized in Table 5-7.
5-16
-------
100
en
MEASURED
PRZM
PISTON
PESTAN
— SESOIL
100 150
DAYS AFTER APPLICATION
200
Figure 5-6. Comparison of PRZM, PISTON, PESTAN, and SESOIL predicted and measured percent
aldicarb remaining in the soil column using parameter values from
Jones et al. (1983).
-------
TABLE 5-6. REVISED STATISTICS FOR MODEL COMPARISONS AT THE
LAKE HAMILTON SITE
O1 = mP1 + b
Comparison m b r2 E RE SE CV
SESOIL/Observed
Aldicarb Residues,
Percent Remaining 0.78 -1.89 0.76 12.40 0.35 22.32 0.63
PRZM/Observed
Aldicarb Residues,
Percent Remaining 0.91 -6.50 0.86 10.80 0.30 17.03 0.48
TABLE 5-7. SUMMARY OF HYDROLOGIC AND SOIL CHARACTERISTICS OF THE TWO FIELD
SITES USED IN THE METALAXYL STUDY (Adapted from Carsel et al., 1985b)
Characteristic
Hydrologic Characteristics*
Precipitation, cm/yr
Runoff, cm/yr
Evapotranspiration, cm/yr
Recharge past root zone, cm/yr
Hydrologic soil group
Florida
110
10
70
30
A
Maryland
100
25
60
15
C
Soil Characteristics**
Soil Series Blanton Marl ton
Field-capacity, (%} 9.1, 9.1, 25.7 20.7, 33.9, 25.7
Wilting-point, (%) 3.3, 3.3, 14.8 9.5, 23.9, 14.8
Organic matter, (%} 1.5, 1.0, 0.5 2.5, 1.0, 1.5
Bulk density, g/cm3 1.45, 1.55, 1.65 1.25, 1.3, 1.3
*Average annual values.
**For each soil characteristic, three values shown are for surface soil,
subsurface soil, and substratum, respectively. Depths corresponding to
these zones were not specified by the authors.
5-18
-------
Subsequent to metalaxyl (Ridomil 2E) application at a rate of 2.2 kg
a.i./ha, tobacco was transplanted at both sites. Soil samples were taken at
the Florida site in 15-cm increments to a depth of 90 cm on the day of
application (0 days), and on 26, 55, 85, and 154 days following pesticide
application. At the Maryland site, soil samples were collected at the same
depth increments as in Florida but at 0, 15, 30, 44, 49, 61, 76, 91, 106, 121,
135, 152, 219, 261, and 287 days. For each sampling, four soil cores (one in
each of the quadrants of the field site) were collected and composited and
analyzed to obtain a "field-averaged" metalaxyl residue concentration with
depth and time. This method differs from that used by Hornsby et al. (1983),
who took 4 soil cores in each of the quadrants of the field plot, analyzed all
16 replicates for aldicarb, and then computed an average pesticide concentra-
tion. It is not known which of these two methods yields a statistically valid
value for the "field average" pesticide concentration.
Because soils are spatially-variable, both the numbers and locations of
soil sampling have a direct influence on the estimated average value of a soil
property. Compositing several soil samples only adds another element of uncer-
tainty to the estimated value because the soil samples taken "at random" may
not be statistically independent. The problems of compositing soil samples
and estimating a "field-average" value of a soil property have recently been
examined by Webster and Burgess (1984). They showed that the estimation var-
iances of a soil property (e.g., pollutant concentration) in the soil samples
will be essentially equal to the Kriged estimates when the following criteria
are met: (1) the property is additive; (2) the semivariogram, describing
spatial-dependence of the property, is either linear or spherical; (3) equal
portions of the soil samples are mixed in bulking; and (4) the samples to be
bulked are taken from an optimally configured sampling grid. Webster and
Burgess (1984) stated that the optimal configuration for sampling is to collect
samples at the nodes of a centrally-placed grid with its interval d=(L/n1/2)>
where L is the length of the block (or plot), and n is the number of sampling
points. To illustrate these criteria, Webster and Nortcliff (1984) reanalyzed
the data collected by Khan and Nortcliff (1982) for micronutrient (Cu, Fe,
Mn, and Zn) concentrations in soil samples from a 1-ha field.
Since pesticide concentration is an additive property, requirement (1)
above is satisfied. As long as soil samples taken over short distances are
bulked, condition (2) may also be valid. A common practice in bulking soils
is to mix nearly equal weights (or volumes) of soil; thus requirement (3)
is also satisfied. Condition (4) is probably not met in most studies, and
the pollutant concentration values estimated from bulk samples may be biased.
Thus, in the Carsel et al. (984) field experiment, an "optimum" sample and an
"unbiased" pesticide concentration estimate are obtained only if the four soil
samples were collected from the center of each quadrant and then equal portions
were mixed; otherwise the estimation variance would be larger. Apparently, the
field procedures were designed to meet this requirement (R. Carsel, personal
communication, 1986).
On the other hand, if the distance over which pesticide concentrations
were spatially-dependent was shorter than the quadrant size, i.e., the samples
were, in fact, spatially independent, only the number of sampling points and
not their locations in the field would determine the estimation variance
5-19
-------
(Webster and Burgess, 1984; Webster and Nortel iff, 1984). The more non-normal
the population frequency distribution of the soil property of interest, the
larger would be the number of samples needed to estimate the mean value with
a given level of confidence (Appendix G). From the foregoing discussion it
should be evident that unless the spatial structure of the variability in the
soil property is known, as determined by the semivariogram, the economic
benefits gained in bulking soil samples and reducing the costs of laboratory
analyses may come at the expense of increased estimation variance. Also,
all information on the nature of the soil spatial variability is lost by
compositing samples taken from different location within a field site. As we
have seen, such knowledge may be crucial in determining the statistical
validity of the sample collected and the pollutant concentration measured.
Carsel et al. (1985b), selected these two field sites because of the
availability of general hydrologic and soil characteristic data. However,
several site-specific hydrologic input data required in PRZM (e.g., daily
records of rainfall, pan evaporation, runoff, ground-water recharge, etc.) were
not collected and had to be estimated on the basis of model calibration using
historical climatological data from nearby meteorological stations. Unavail-
ability of site-specific data for model parameters is a major problem espe-
cially when data from field studies are used where monitoring is the primary
goal rather than model testing. A similar problem is encountered in estimating
pesticide-specific model parameters, such as sorption coefficient (Kg) and
degradation rate coefficient (k) or half-life (ti/2). Carsel et al. (1985b),
estimated the KQ values based on a Kom value of 40. However, the k values
for metalaxyl had to be estimated based on the decrease in total amount of
metalaxyl remaining in the 90-cm soil profile at various times during the
monitoring period. They noted that a first-order rate model with k = 0.014
day1 described well (R^ = 0.99) the metalaxyl losses at the Florida site.
A biphasic first-order model was needed to describe the metalaxyl data from
the Maryland site; the rate coefficient for the first phase (0-30 days) was
0.0455 day1, while that for the second phase (>30 days) was 0.00453 day1.
Carsel et al. (1985b), cite others who have noted a similar multiphasic pesti-..
cide degradation under field conditions with this degradation depending on the
timing of pesticide application and the occurrence of rainfall or irrigation
events. They attribute it to various processes (e.g., volatilization, photo-
degradation) that affect pesticide fate near or at the soil surface, but do not
occur once the pesticide has leached to subsurface depths. It should be
recognized that the rate parameters estimated in this manner are for pesticide
"loss" via all pathways except leaching and represent the "depth-averaged"
values for the entire soil profile sampled (i.e., 0-90 cm in this case).
Therefore, another likely explanation for the biphasic degradation is the
differences in degradation rates with soil depth. This points out the limita-
tions of assigning a single degradation rate coefficient for the entire soil
profile.
Observed and simulated (PRZM) metalaxyl concentration profiles for the
Florida site are compared in Figure 5-7. It is evident that the measured and
predicted concentration profiles agree for the 55 and 85-day sampling but do
not agree for the first sampling date (26 days post-application). A linear
regression of measured and simulated data yielded coefficients of determina-
tions (R2) of 0.33, 0.90, and 0.95 for the 26, 55, and 85 day sampling,
5-20
-------
Obuivad __
Predicted
Sampling dale
6/6/BO
Area undei curve
Obierved 1 0 X 10' l|ig kg - cm)
Predicted 1 2 X 10- (|ig kg •' cm)
10 20 30 40 60 60 70 80
10 20 3O 4O 50 BO 70 80
Soil Depth (cm)
Figure 5-7. Comparison of measured metalaxyl concentration profiles at
three sampling dates with those predicted by PRZM model (adapted from
Carsel et al., 1985b).
5-21
-------
respectively. Note that the areas under the predicted and measured curves
representing the total amount of metalaxyl in the 90-cm profile agree in all
cases. Since the degradation rate coefficient value was estimated by calibrat-
ing the model to measured data, this is an expected result. Thus, the failure
of the PRZM model to predict the 26-day post-application metalaxyl concentra-
tion profiles may be attributed to errors in the model input data (recall that
site-specific daily records for hydrologic and meteorological data were unavail-
able and were estimated), errors in measured data (in soil sampling and pesti-
cide analysis), and the inadequacy of the model to simulate water and pesticide
movement.
As a result of lower precipitation, a higher runoff (Table 5-7), and a
larger sorption coefficient compared to the Florida site, metalaxyl did not
leach past the 15-cm depth during the entire sampling period at the Maryland
site. Thus, PRZM model simulations were compared with the measured data only
for the 0-15 cm depth (Figure 5-8). A linear regression of measured and pre-
dicted values yielded an R2 of 0.75, indicating a reasonable agreement. How-
ever, it should be noted that since leaching past the 15-cm depth was not a
significant factor, the dominant loss pathways would be runoff, degradation,
and uptake. Of these, the model was calibrated to estimate the degradation
rate coefficient. Thus, agreement is expected. Based on the above results
from two field sites, Carsel et al. (1985b), concluded that "using the best
estimates of transport and transformation properties of metalaxyl and limited
calibration for water balance," the PRZM model was "effective in simulating the
important processes operating on the pesticide" under field conditions.
Closure
The SESOIL testing study by Watson and Brown (1984) is one of a number of
such model testing studies that have been performed. We have discussed it here
because it is a recent comprehensive study and includes comparisons of field
data with predictions by the three models discussed in this manual. Carsel et
al. (1985b), is a good example of the problems encountered in model performance
testing using environmental fate monitoring data collected as a part of the
pesticide registration process and is not specifically for model evaluation as
an objective.
5-22
-------
2.00 -,
0) 150-
O>
a
•> i-°°
X
_«
CO
:
0.50-
0.00
Predicted
-ff-
n
171
ro
0)
2.00 -i
1.50-
X* 1.00 -|
n
0.50-
0.00
Observed
Sampling
D
epth: 0-15 cm.
n n
((•
16 30 4449 61 75 91 106 121 136 162 219
Days After Application
n
261
267
Figure 5-8. Comparison of measured and predicted metalaxyl concentrations in the 0-15 cm
depth increment (taken from Carsel et al., 1985b).
-------
Chapter 1 noted a number of additional model testing studies that provide
valuable information and examples of model testing procedures and sensitivity
behavior of the specific models discussed herein. In addition to the specific
studies on SESOIL and PRZM noted above, the reader should review some of the
following studies for the particular model being applied:
a. PESTAN has been applied by Jones and Back (1984) to the aldicarb
data in Florida, and it has been used by the EPA Office of Pesticide
Programs for a variety of compounds in different regions (M. Lorber,
1985, personal communication).
b. SESOIL has been applied to leaching of selected metals in Kansas and
two organics in Montana (Bonazountas et al., 1982), has undergone
sensitivity analyses for different climates and soil types (Wagner
et al., 1983), has been tested for hydrologic prediction accuracy
(Hetrick, 1984), and has been compared to a variety of other models
for simulating vertical flow in a landfill (Kincaid et al., 1984).
c. PRZM has been tested against data on aldicarb in New York, and against
atrazine and chloride applied to corn in Georgia (Carsel et al.,
1985a).
The model validation and testing process has been discussed by Donigian
(1983) in terms of the most likely reasons for a difference between model
predictions and field observations. He identified and discussed four categories
of potential errors or discrepancies that often occur, including input data,
parameter values (including calibration), model representation, and output (or
observed data). Based on the discussion by Donigian (1983) and the SESOIL
testing described above, the following recommendations are provided to potential
model users.
a. Be sure to make conditions under which the model operates (e.g., input
meteorologic data, parameter values) as close as possible to actual
field conditions under which observed data were collected.
b. Be aware of model assumptions and limitations with respect to repre-
senting field conditions (e.g., the inability of SESOIL to accurately
represent low permeability soil layers). Biased parameter values can
result through calibration if model limitations are not respected.
c. Be especially wary of calibration efforts that result in parameter
values outside their normal range of expected values (e.g., Koc values
for aldicarb in the SESOIL testing) unless these values are justified
by local conditions.
d. Be sure that the model is operating correctly both in terms of algo-
rithm calculations, and proper usage and interpretation of input
values. Simple hand calculations to confirm decay rates and to check
mass conservation should be performed.
5-24
-------
e. Be aware of possible errors, omissions, or inaccuracies in the observed
data especially with regard to their possible impact on the calibra-
tion process (e.g., calibrating to inaccurate data). The problem of
spatial variation must be considered.
In summary, model users should develop an attitude of "informed skepticism"
when performing model testing. The user must be fully knowledgeable and informed
of model assumptions and limitations and of field conditions; this must be bal-
anced by a healthy skepticism or by a questioning approach, in the interpreta-
tion of both the model predictions and the observed data. In this way, models
can be used appropriately as tools to aid environmental decision-makers but
cannot be used as a crutch to support and defend pre-conceived policy and
regulations.
5-25
-------
REFERENCES
Ambrose, Jr., R. B., and S. E. Roesch. 1982. Dynamic Estuary Model Perform-
ance. Journal of the Environmental Engineering Division, ASCE, Vol. 108,
EE1.
ASTM. 1984. Standard Practice for Evaluating Environmental Fate Models of
Chemicals. E978, American Society for Testing and Materials.
Philadelphia, PA.
Bonazountas, M., and J. Wagner. 1982. Pollutant Transport in Soils via
"SESOIL." Presented at ASCE National Conference at Environmental Engineer-
ing, Minneapolis, MN, July 15-16, 1982.
Bonazountas, M., and J. Wagner. 1984. SESOIL: A Seasonal Soil Compartment
Model. Prepared by Arthur D. Little, Inc., Cambridge, MA, for the U.S.
Environmental Protection Agency, Office of Toxic Substances, Washington,
DC.
Bonazountas, M., J. Wagner, and B. Goodwin. 1982. Evaluation of Seasonal
Soil/Groundwater Pollutant Pathways. Prepared by Arthur D. Little, Inc.,
Cambridge, MA, for the Environmental Protection Agency, Office of Water
Regulation and Standards, Washington, DC.
Bromilow, R. H., R. J. Baker, M. A. H. Freeman, and K. Gorog. 1980. The
degradation of aldicarb and oxamyl in soil. Pesticide Sci. 11:371-378.
Carsel, R. F., C. N. Smith, L. A. Mulkey, J. D. Dean, and P. Jowise. 1984.
User's Manual for the Pesticide Root Zone Model (PRZM): Release 1.
EPA-600/3-84-109. U.S. Environmental Protection Agency, Athens, GA.
Carsel, R. F., L. A. Mulkey, M. N. Lorber, and L. B. Baskin. 1985a. The
Pesticide Root Zone Model (PRZM): A procedure for evaluating pesticide
leaching threats to groundwater. Ecol. Model. 30:49-69.
Carsel, R. F., W. B. Nixon, and L. G. Balentine. 1985b. Comparison of Pesti-
cide Root Zone Model Predictions with Observed Concentrations for the
Tobacco Pesticide Metolaxyl in Unsaturated Soils. Environ. Toxicol. Chem.
(In Press).
Donigian, Jr., A. S. 1983. Model Predictions vs. Field Observations: The
Model Validation/Testing Process. In: R. L. Swann and A. Eschenroeder
(Eds), Fate of Chemicals in the Environment. ACS Symposium Series 225.
American Chemical Society, Washington, DC.
5-26
-------
Donigian, Jr., A. S., J. C. Imhoff, and B. R. Bicknell. 1983. Modeling
Water Quality and the Effects of Agricultural Best Management Practices
in Four Mile Creek, Iowa. Prepared by Anderson-Nichols and Co., Inc.,
Palo Alto, CA, for U.S. Environmental Protection Agency, Athens, GA.
Enfield, C. G., R. F. Carsel, S. Z. Cohen, T. Phan, and D. M. Walters. 1982.
Approximating pollutant transport to ground water. Ground Water
20:711-722.
Halfon, E. 1985. Regression method in ecotoxicology: A better formulation
using the Geometric Mean Functional Regression. Environ. Sci. Tech.
19:747-749.
Hetrick, D. M. 1984. Simulation of the Hydrologic Cycle for Watersheds.
Proceedings for the Applied Simulation and Modeling Conference. San
Francisco, CA.
Hornsby, A. G., P. S. C. Rao, W. B. Wheeler, P. Nkedi-Kizza, and R. L. Jones.
1983. Fate of Aldicarb in Florida Citrus Soils: 1. Field and Laboratory
Studies, Proceedings of the NWWA/U.S. EPA Conference on Characterization
and Monitoring of the Vadose (Unsaturated) Zone, Las Vegas, NV, December
8-10, 1983. pp. 936-958.
INTERA Environmentaal Consultants, Inc. 1980. Mathematical Simulation of
Aldicarb Behavior on Long Island: Unsaturated Flow and Ground-Water
Transport. Prepared for South Carolina Pesticide Epidemiological Study
Center, Medical University of South Carolina.
Johnson, H. P. and J. L. Baker. 1982. Field-to-Stream Transport of Agricul-
tural Chemicals and Sediment in an Iowa Watershed: Part I. Data Base
for Model Testing (1976-1978). EPA-600/3-82-032. U.S. Environmental
Protection Agency, Athens, GA.
Jones, R. L., and R. C. Back. 1984. Monitoring aldicarb residues in Florida
soil and water. Environ. Toxicol. Chem. 3:9-20.
Jones, R. L., and G. W. Gee. 1984. Assessment of Unsaturated Zone Transport
for Shallow Land Burial of Radioactive Waste: Summary Report of Tech-
nology Needs, Model Verification, and Measurement Efforts (FY78-FY83),
PNL-4747, Battelle, Pacific Northwest Laboratory, Richland, WA.
Jones, R. L., P. S. C. Rao, and A. G. Hornsby. 1983. Fate of Aldicarb in
Florida Citrus Soils: 2. Model Evaluation. Proceedings of the NWWA/
U.S. EPA Conference on Characterization and Monitoring of the Vadose
(Unsaturated) Zone. Las Vegas, NV. December 8-10, 1983.
Jury, W. A., L. A. Stolzy, P. Shouse. 1982. A field test of the transfer
function model for predicting solute transport. Water Resour. Res.
18:369-375.
5-27
-------
Khan, M. A., and S. Nortel iff. 1982. Variability of selected soil micro-
nutrients in a single soil series in Berkshire, England. J. Soil Sci.
33:763-770. "
Kincaid, C. T., J. R. Morery, S. B. Yabusaki, A. R. Felmy, and J. E. Rogers.
1984. Geohydrochemical Models for Solute Migration, Volume 2, Preliminary
Evaluation of Selected Computer Codes for Modeling Aqueous Solutions
and Solute Migration in Soils and Geologic Media. EA-3417, Vol. 2.
Electric Power Research Institute, Palo Alto, CA.
Lyman, W. J., W. F. Reehl, and D. H. Rosenblatt. 1982. Handbook of Chemical
Property Estimation Methods. McGraw-Hill Book Company, New York.
Moore, G. E., T. E. Stoeckenius, and D. A. Stewart. 1982. A Survey of Statis-
tical Measures of Model Performance and Accuracy for Several Air Quality
Models. EPA-450/4-83-001. U.S. Environmental Protection Agency, Research
Triangle Park, NC.
Rao, P. S. C., J. M. Davidson, and L. C. Hammond. 1976. Estimation of Non-
Reactive and Reactive Solute Front Locations in Soils. EPA-600/9-76-015.
Proc. of Hazardous Wastes Research Symp., Tucson, AZ.
Smelt, J. H., M. Leistra, N. W. H. Houx, and A. Dekker. 1978a. Conversion
rates of aldicarb and its oxidation products in soils. I. Aldicarb
sulfone. Pesticide Sci. 9:293-300.
Smelt, J. H., M. Leistra, N. W. H. Houx, and A. Dekker. 1978a. Conversion
rates of aldicarb and its oxidation products in soils. II. Aldicarb
sulfoxide. Pesticide Sci. 9:286-292.
Smith, C. N., R. A. Leonard, G. W. Langdale, and G. W. Bailey. 1978. Trans-
port of Agricultural Chemicals from Small Upland Piedmont Watersheds.
EPA-600/3-78-056. IAG No. IAG-D6-0381. U.S. Environmental Protection
Agency, Athens, GA and U.S. Department of Agriculture, Watkinsville, GA.
Supak, J. R. 1972. The Volatilization, Degradation, Adsorption and Desorption
Characteristics of Aldicarb in Soils and Clays. Ph.D. Dissertation,
Texas A&M University.
U.S. EPA. 1982. Testing for Field Applicability of Chemical Exposure Models.
Results of the Workshop on Field Applicability Testing, Exposure Modeling
Committee Report, Athens, GA.
van Genuchten, M. Th. and W. J. Alves. 1982. Analytical Solutions of the
One-Dimensional Convective Dispersive Solute Transport Equation, U.S.
Department of Agriculture, Technical Bulletin No. 1661.
Wagenet, R. J. and P. S. C. Rao. 1985. Basic concepts of modeling pesticide
fate in the crop root zone. Weed Science 33: (In Press).
5-28
-------
Wagner, J., M. Bonazountas, D. M. Alsterberg. 1983. Potential Fate of Buried
Halogenated Solvents via SESOIL. Prepared by Arthur D. Little, Inc.,
Cambridge, MA, for U.S. Environmental Protection Agency, Washington, DC.
Watson, D. B. and S. M. Brown. 1984. Testing and Evaluation of the SESOIL
Model. Draft Report Prepared by Anderson-Nichols Co., Inc., Palo Alto,
CA, for U.S. Environmental Protection Agency, Athens, GA.
Webster, R. and T. M. Burgess. 1984. Sampling and bulking strategies for
estimating soil properties in small regions. J_. Soil Sci. 35:127-140.
Webster, R. and S. Nortel iff. 1984. Improved estimation of micronutrients in
hectare plots of the Soning series. J_. Soil Sci. 35:667-672.
Wyman, J. A., J. 0. Jensen, D. Curwen, R. L. Jones, and T. E. Marquardt. 1984.
Effects of application procedures and irrigation on degradation and
movement of aldicarb residues in soil. Submitted for publication.
5-29
-------
APPENDIX A
CHEMICAL MOVEMENT THROUGH SOIL
by
W. A. Jury
PROCESS DESCRIPTION
Transport Mechanisms
Chemicals are transported through soil principally by three mechanisms:
mass flow of dissolved chemical within moving soil solution, liquid diffusion
within soil solution, and gaseous diffusion within soil air-filled voids. The
first mechanism, mass flow, refers to the passive transport of dissolved solute
within moving soil water. To a first approximation, the solute within an arbi-
trary volume element of moving soil water is assumed to be uniformly distrib-
uted, and hence the mass flux of chemical is given by the product of the volume
flux of water times the dissolved solute concentration expressed in units of
mass of solute per volume of water (see Appendix D). Liquid diffusion refers
to the transport of the dissolved solutes by diffusion in response to molecular
scale collisions. The effect of large numbers of these collisions is to move
dissolved solutes from regions of higher solute density to lower solute density;
thus, the diffusion flux is proportional to the density or concentration gra-
dient. The coefficient of proportionality between the diffusion flux and the
concentration gradient is called the liquid diffusion coefficient (see Appendix
D). Chemical vapor molecules within the soil air spaces also undergo molecular
collisions and spread out by vapor diffusion. As in the case of liquid diffu-
sion, the flux of vapor molecules is proportional to the density or concentra-
tion gradient. The coefficient of proportionality is called the vapor diffu-
sion coefficient (see Appendix D).
An additional term is frequently required in the mathematical description
of chemical transport. Since the soil water flux is represented as a continu-
ous quantity which is volume-averaged over many pores, the individual compli-
cated water flow paths around soil grains are mathematically replaced by an
equivalent one-dimensional flow. When this one-dimensional flow of water is
multiplied by the dissolved solute concentration, the resulting solute mass
flux does not take into account the additional spreading of solute which occurs
by three-dimensional mass flow at the pore scale in the actual system but which
is not represented in the volume-averaged mathematical treatment. This apparent
solute diffusion arising from the mass flux effects which are obscured by
mathematical volume averaging is called hydrodynamic dispersion (Bear, 1972).
A-l
-------
Under certain conditions, this dispersion transport process is mathematically
equivalent to transport by liquid diffusion and may be included in the trans-
port equations (see Appendix D) by using an effective liquid diffusion-
dispersion coefficient to account for spreading of liquid solute molecules.
Because this effect depends on the size of the volume averaging, it is a func-
tion of the scale of approximation of the water flux. Thus, hydrodynamic
dispersion is considerably more important on the field scale than on the labora-
tory scale (Bear, 1972; Biggar and Nielsen, 1976).
Soil. Environmental, and Management Factors Influencing Chemical Transport
Through Soil
There are a host of factors influencing each of the transport mechanisms
mentioned above. In the mathematical derivation of the transport equations
and the boundary conditions given in Appendix D, the influence of many of these
parameters is expressed through the mathematical relationships. In this dis-
cussion, reference to these mathematical equations will be made where appropri-
ate; however, most of the discussion will be qualitative, emphasizing the
physical rather than mathematical connection between a parameter and a trans-
port mechanism.
Table A-l summarizes the various parameters influencing chemical transport
through soil and is somewhat arbitrarily divided into groups of soil parameters,
environmental parameters, and management parameters. In the discussion below,
the effect of each of these parameters on the four transport mechanisms of mass
flow, liquid diffusion, liquid dispersion, and gas diffusion will be discussed;
literature information will be brought in where appropriate.
Soil Parameters--
Soil water content—Volumetric soil water content (9) has a significant
influence on the liquid and gaseous diffusion transport mechanisms. The mass
of chemical moved per unit time from point A to point B by diffusion is in-
versely proportional to the distance between A and B (Nielsen et al., 1972).
Thus, the actual path length in soil followed by a vapor molecule or a dis-
solved solute molecule is strongly affected by water content. For liquid
diffusion, diffusive transport increases as water content increases because
the cross-sectional area available for flow increases, and the path length
decreases as liquid replaces air in the medium. Conversely, for vapor diffu-
sion, the transport by diffusion decreases with increasing water content because
the air space decreases. In fact, as shown in Appendix D, the volumetric air
content (a) which is equal to the porosity (0) minus the volumetric water
content (9) plays the same functional role in vapor diffusion as the liquid
water content plays in liquid diffusion (see equations D-9 and D-ll). Models
which describe the change in liquid or vapor diffusion coefficient as a func-
tion of water content are called tortuosity models (Nielsen et al., 1972).
Although there are a number of different kinds of tortuosity models which have
been proposed over the years, the most versatile appears to be the Millington
and Quirk model discussed in Appendix D (Millington and Quirk, 1961). In this
model, the liquid diffusion flux is proportional to the 10/3 power of the water
content, and the vapor diffusion flux is proportional to the 10/3 power of the
air content. This means that the amount of solute transported by each mechan-
ism will change dramatically as water content changes (Jury et al., 1983a).
A-2
-------
TABLE A-l. SOIL, ENVIRONMENTAL, AND MANAGEMENT PARAMETERS INFLUENCING
CHEMICAL TRANSPORT THROUGH SOIL
Environmental
Soil Parameters Parameters Management Parameters
Water Content Temperature Chemical Concentration
Bulk Density or Porosity Precipitation Irrigation Management
Permeability (Saturated) Evapotranspiration Crop Characteristics
Clay Content
Surface Area
Organic Matter
Content
Depth to Ground Water
Water Retention (Field Capacity)
Since mass flux of chemical is proportional to water flux multiplied by
dissolved solute concentration, it is not directly affected by water content.
Since increasing the water content of a given soil will result in higher mass
flux, some correlation between chemical movement and water content may be
found. Hydrodynamic dispersion is also proportional to water flux. Thus, it
is usually not considered to be a function of water content.
The partitioning of chemicals between gaseous, liquid, and solid phases
can obviously affect chemical transport by affecting the amount of chemical in
solution and in the gaseous phase. When soil water contents decrease below a
few monolayers of water (which can occur in the top few millimeters of soil
during intense drying) water molecules which preferentially occupy soil
adsorption sites are displaced from the surface, and the chemical adsorption
capacity of the soil is greatly increased (Spencer and Cliath, 1973). This
increased adsorption capacity causes liquid and gaseous concentrations to
decrease dramatically in this dry region, and all four of the above transport
mechanisms are reduced significantly as most of the chemical is displaced to
adsorption sites. However, this adsorption effect does appear to be reversible
when the soil surface layer rewets (Spencer and Cliath, 1973; Harper et al.,
1976). For this reason, it may be possible to ignore this aspect of the water
content dependence of adsorption provided that the time period during which
the surface layer is dry is short-lived. Should a permanently dry surface
layer (e.g., a desert soil) be part of a scenario, the effect should definitely
be considered.
A-3
-------
Bulk density or porosity—Soil porosity (0) is related to bulk density
by a linear relation
0=1- Pb/Pm (A-D
where pm is soil mineral density (g/cm3) which for most soils lies between
2.65 g/cm3 (clays) and 2.75 g/cm3 (sands). For this reason, it suffices to
discuss only the dependence of the soil porosity (0) on chemical transport
with the understanding that decreased porosity implies increased bulk density.
Decreasing soil porosity or increasing bulk density will generally
decrease chemical transport by each of the above mechanisms. In the case of
liquid or vapor diffusion, decreasing porosity generally decreases the cross-
sectional area available for flow and increases the path length by placing more
solid obstacles per unit volume in the way of the diffusing molecules.
Although the equations in Appendix D specifically contain a factor of 02
in the denominator of the tortuosity factor, this implied functional dependence
is misleading because the volumetric air and water contents themselves depend
on porosity. It is perhaps most instructive to examine the two equivalent
problems: vapor transport in very dry soil or liquid transport in very wet
soil. In each case, the tortuosity factor and hence the transport flux accord-
ing to the Millington and Quirk model is proportional to the 4/3 power of the
porosity. Thus as porosity decreases, both liquid and vapor diffusion decrease
somewhat faster than linearly.
Transport by mass flow is indirectly affected by porosity since regions of
low porosity are likely to have lower permeability to water transport. Al-
though no good structural models exist for the relationships between porosity
and permeability, permeability of a given soil type strongly decreases as
porosity decreases because the pore sizes contract. However, finer-textured
soils such as clays generally have a higher porosity and lower permeability
than sandy soils.
Hydrodynamic dispersion has not been quantitatively linked to porosity.
However, it is proportional to water flux (see equation D-13) which in turn is
decreased by decreases in porosity under most conditions.
Another influence of porosity on transport which equally affects all
mechanisms is that decreasing soil porosity increases the density of mineral
and organic adsorption sites and thus causes increased adsorption of chemical
with a corresponding decrease in solution and gaseous concentration. Since
transport only occurs in liquid or vapor phases, transport by all four mecha-
nisms decreases as porosity decreases.
Saturated hydraulic conductivity or permeability—Saturated hydraulic
conductivity (KS) is the proportionality coefficient between the saturated
water flux and the hydraulic head gradient (Hillel, 1971). An equivalent
index frequently used by engineers is the intrinsic permeability (k) which
is equal to K§g/pg where g is the viscosity of water, p is water density, and
g is the gravitational acceleration. The permeability k thus defined is
independent of any fluid properties. For any process where water is ponded on
A-4
-------
the soil surface, either through intense irrigation or rainfall or because of
a deliberate ponding management such as a waste holding pond, permeability will
have a dominant influence on the amount of water infiltrating into the soil
and hence will strongly influence mass flow and dispersion. When water flow
occurs at an unsaturated water content (such as when water is applied at the
soil surface at a constant rate less than the saturated hydraulic conductiv-
ity), the main influence of the soil hydraulic conductivity is to regulate at
what water content the flow will occur. There is no relationship between
hydraulic conductivity and liquid or vapor diffusive transport.
Clay content—The clay content of soil is a qualitative index which has
been correlated with a number of other soil properties such as water-holding
capacity, specific surface area, etc. One of the highest correlations is
between clay content and saturated hydraulic conductivity or permeability
since soils high in clay tend to be low in permeability (Cosby et al., 1984).
Thus, all of the arguments made above about permeability also apply to clay
content. Clay content is also highly correlated with ion exchange capacity and
therefore will strongly influence the adsorption of chemical ions. For most
non-polar organic chemicals, however, there is usually no correlation between
clay content and adsorption (Green, 1974). There is also a reasonably high
negative correlation between clay content and soil porosity. Therefore, the
comments made above about porosity also apply to clay content.
Adsorption site density—The density of adsorption sites for chemicals may
be represented by two other soil indices which are specific surface area (sur-
face area per soil volume or mass) and soil organic matter content.
Since clays have a large specific surface area, the comments made above
about clay content also apply to surface area. Thus, the correlation between
surface area and organic chemical adsorption is likely to be quite low. How-
ever, organic matter content has been found to be positively correlated with
organic chemical adsorption in a number of studies (Green, 1974; Hamaker and
Thompson, 1972). Thus, one immediate effect of increasing organic matter
content is to increase the extent of chemical adsorption and thus decrease
liquid and gaseous concentration thereby strongly decreasing the extent of
transport by each of the four transport mechanisms.
Depth to ground water—Depth to ground water can strongly influence the
extent of upward water flow occurring to a surface layer which has been evapo-
rating without water input for an extended period of time. It has been shown
both theoretically and experimentally (Gardner, 1959) that finer textured soils
can move water (and hence chemicals by mass flow) upward from much greater
depths than can coarse textured soils. For this reason, salinization of a
surface soil layer from a saline water table is an ever-present problem above
shallow saline water tables such as near the Salton Sea area of the Imperial
Valley of California.
Depth to ground water also obviously affects the travel time of a chemical
leached from the surface to ground water. However, depth to ground water has
no direct influence on downward transport processes of any of the four mech-
anisms other than to create a relatively thin region known as the capillary
fringe which is above the ground-water table and which has a high water content.
A-5
-------
Water retention (field capacity)--An extremely important property of soil
which can influence the above processes particularly when inputs of water to
the soil surface are infrequent is the amount of water remaining after drain-
age has become insignificant. This water content, known as the field capacity
is much higher in a finer textured soil than in a sandy soil (Millei, 1971).
Its influence on chemical transport is identical to that of soil water content
above.
Environmental Parameters--
Temperature--Temperature affects all rate processes in nature including
diffusive transport. The diffusion coefficient of a chemical moving through
pure water is an increasing function of temperature as is the gaseous diffusion
coefficient of a chemical moving through pure air. In addition, chemical vapor
pressure is a strongly increasing function of temperature, and hence as temper-
ature increases, the fraction of the total chemical present in the vapor phase
increases (see equation D-17). For this reason, vapor diffusion is much more
strongly enhanced by increases in temperature than is liquid diffusion. How-
ever, transport by both mechanisms will increase for the reasons given above.
There have been various models for describing the temperature dependence of the
vapor and liquid diffusion coefficients. Standard chemical engineering text-
books indicate that vapor diffusion coefficients have a temperature dependence
proportional to T1'7^ (Bird et al., 1960). They also indicate that liquid
diffusion coefficients increase with temperature, but no standard functional
relationship applies.
The temperature dependence of chemical mass flow depends on the tempera-
ture dependence of water flow for reasons given above. Liquid water transport
under temperature gradients is poorly understood at present and is likely to be
a relatively minor effect compared to the other factors influencing water
transport (Jury, 1973). Thus, the primary factor affecting the temperature
dependence of chemical mass flow is the temperature dependence of the liquid
concentration in a three-phase soil system. The amount of chemical present in
the liquid phase is strongly affected by absorption which generally decreases .
as temperature increases (Biggar and Cheung, 1973). Thus, increased liquid
concentrations resulting in increased mass flows would be expected as tempera-
ture increases.
Precipitation—The characteristics of the precipitation or rainfall events
(i.e., intensity and distribution) will have a critical influence on the extent
of chemical transport in all phases above. Precipitation will have a dominant
influence on mass flow because the rainfall rate is directly related to the
water flow rate in the soil. Thus soils with intense, frequent rainfall will
have high water fluxes and hence high chemical mass fluxes and dispersion
fluxes. Furthermore, an extremely intense rainfall event might induce satura-
tion which could result in a greatly enhanced mass transport through soils of
high permeability. The exact response of a soil to a rainfall event can be
determined only by solving the transport equations given in Appendix D. How-
ever, it is reasonable to conclude that for a given soil the effect of increas-
ing the rainfall intensity in a given time period will be to increase the mass
flux and also to increase the water content. The effect of the latter will be
to increase the extent of transport by liquid diffusion and to decrease the
extent of transport by vapor diffusion.
A-6
-------
Evapotranspi rati on—Evapotranspi rati on represents the amount of applied
water which is removed by plants or water loss from surfaces and hence is
unavailable for drainage. Thus the extent of evapotranspiration will strongly
affect the water flux below the root zone and hence the extent of chemical
leaching by mass flow. For soils not receiving water input by irrigation,
rainfall minus evapotranspiration determines the net amount of water leaching
beyond the crop root zone. This parameter, if positive, will imply long-term
drainage or downward movement, and, if negative, will imply drying processes
which will lower water content and induce upward flow from ground water.
Water loss by evapotranspiration (ET) may be regulated by external
meteorological conditions (potential ET) or by resistances in the soil or
plant. The latter generally dominates only when the soil is dry. The most
important meteorological conditions regulating potential ET are solar radia-
tion, wind, air temperature, and air humidity. These meteorological measure-
ments may be used in formulas such as the Penman combination equation to
predict potential ET for a crop or soil surface (Doorenbos and Pruitt, 1976).
Management Parameters—
Chemical concentration—To the extent that the amount of chemical applied
to soil is a management option, for example with pesticide application, this
variable has been characterized as a management parameter. The effect of
increasing chemical concentration on transport is always to increase the amount
of transport in each phase by increasing the concentration in each phase. In
fact, for systems represented by linear partition coefficients (see equations
D-27 through D-30), each of the four transport mechanisms are proportional to
total chemical concentration because a unit increase in chemical concentration
partitions proportionally into each phase. However, not all soils adsorb
solute linearly. In many cases, chemicals applied at higher concentrations
are less-efficiently adsorbed than are chemicals applied at low concentrations.
In such cases, the amount of mass flow may increase more than linearly with in-
creases in chemical concentrations.
Irrigation management—Application of irrigation water to soil has an
effect similar to the rainfall application. However, since irrigation can be
more carefully controlled than rainfall, it is characterized as a management
parameter. For salinity control, crops must be irrigated at a sufficient
volume over the evapotranspiration rate in order to produce drainage. As a
practical matter, it is rarely possible to manage irrigation with less than a
leaching fraction (drainage divided by irrigation) of 0.2. While chemicals are
moving through a crop root zone, chemical concentrations are increased as water
is preferentially extracted and solute left behind by plant roots. Thus, one
effect of irrigation management is indirectly to raise chemical concentrations
in the root zone. However, at the same time, irrigation management can direc-
tly control mass fluxes by restricting the amount of salt added with irrigation
water and by decreasing drainage flux. Thus, it has a complicated effect on
mass transport. The other main influence of irrigation management on transport
is to affect the water content of the crop root zone. Irrigating at a higher
intensity will increase water content and thus affect the diffusive transport
mechanisms as indicated in the earlier discussion.
A-7
-------
Crop characteristics—General crop characteristics such as wilting point
and rooting depth will indirectly affect chemical transport. The major effect
of wilting point is to dramatically decrease evapotranspiration when water
application to the crop either by rainfall or irrigation becomes inadequate.
Rooting depth can somewhat affect the transport mechanisms by widening or
narrowing the zone of extraction. However, since the total amount of water
extracted is the same in each case, this effect is relatively minor and gener-
ally not given much importance in chemical transport.
Crop residues may influence chemical transport by increasing organic
matter in the soil, with the result that the number of adsorption sites will
also be increased. Also, channels left by decaying plant roots may induce
preferential flow of water and chemical at high rates through these macropores.
Chemical and water application variability—When one dimensional solute
transport models are used to describe downward movement below fields which have
received a common surface application of water and chemical, part of the vari-
ability of observed solute concentrations is caused by non-uniform application
of both water and solute. Water application methods such as furrow or flood
irrigation which pond water over the surface are highly variable and commonly
result in coefficients of uniformity (CU) of 50 percent or lower whereas
sprinkler or drip systems applied during periods of low wind may approach CUs
of 90 percent (Shouse et al., 1982).
Chemical application variability is also quite high particularly when the
chemical is spread or sprayed on the surface rather than added with irrigation
water (Rao and Vlagenet, 1985). Taylor and Klepper (1971) observed a CU of only
20 percent in dieldrin concentrations sprayed onto a 0.13 ha plot. Similarly,
Richter (1984) observed CUs of 50 percent on two fields sprayed with potassium
bromide. Finally, El Abd (1984) found a CU of 50 percent in recovered napropa-
mide sprayed on a 1.4 ha field. These large spatial variations must be taken
into account in sampling strategies for model validation.
MATHEMATICAL DESCRIPTION OF CHEMICAL MOVEMENT
The mathematical differential equations describing vertical flow of water
and chemicals in soil are derived in Appendix D. In this Appendix, it is shown
that the equations result from an application of conservation of mass (either
water or chemical) within a unit volume of soil together with an appropriate
expression for the flow of material per unit area per unit time using recog-
nized transport laws.
Appendix D also discusses the appropriate boundary conditions for use in
solving chemical and water transport processes and reviews standard methods of
solution of the mathematical equations and their boundary conditions.
LIMITATIONS OF THE PHYSICAL TRANSPORT MODELING APPROACH
Non-Equilibrium Effects
Chemical adsorption or reaction is ultimately a rate-limited process. In
soil solution, the approach to equilibrium is a function of, among other factors,
A-8
-------
soil geometry and water flow. At the present time, no model exists for describ-
ing the dependence of the rate coefficient in expressions such as equation
(D-26) of Appendix 0 on these variables. Rao et al. (1980), have made a promis-
ing first approach in describing this functional dependence for simple spherical
aggregates in a soil column, but their method has not been tested under more
natural conditions. In the interim, equilibrium models have usually been used
particularly under natural field conditions when geometry is poorly character-
ized. At the present time, such equilibrium models must be used but will be
most accurate when the time required to reach equilibrium is short compared to
the residence time of a chemical in the soil region of interest.
Measurement Limitations
There are a number of practical limitations to model calibration and test-
ting under field conditions. Table A-2 lists a number of potential problems
encountered in field validation of water and chemical transport models together
with the implications of the problem. These will be discussed in more detail
below.
Lateral and Vertical Variability of Transport and Retention Properties--
In recent years considerable attention has been focused on the spatial
variability of physical and chemical properties at the field scale (Nielsen et
al., 1973; Biggar and Nielsen, 1976; Wagenet and Jurinak, 1978; Russo and
Bresler, 1980; Gajem et al., 1981). This research, still more basic than
applied, has pointed out that in the field, physical and chemical properties
have a spatial correlation which persists over short length scales, as little
as 5 meters, and that properties associated with transport may have high co-
efficients of variation over field areas. Appendix G of this report provides
a review of some major experimental studies of spatial variability.
What these findings indicate is that a serious problem lies ahead in both
field characterization and model calibration when large areas are involved.
For example, Biggar and Nielsen (1976) estimated that for their 150 ha field,
hundreds of measurements of the effective chloride dispersion coefficient DE
would have to be taken to obtain the field-wide average value to within 50 per-
cent of the true mean. Such variability has led to new approaches for design-
ing measurement strategies and subsequently modeling chemical transport events
(Vlarrick and Nielsen, 1980). Even on scales smaller than 1 ha fields, however,
variability will likely limit the success of model calibration or testing.
Because of the innate variability, many replicates will be needed to character-
ize any of the parameters discussed above and summarized in Table A-l. Fur-
thermore, functional relations which are required to characterize hydraulic
conductivity as a function of water content, for example, are greatly obscured
by variability between replicate measurements and functional relations. In
principle, each location in the soil could have its own functional relationship,
but linking these together would require a complicated three-dimensional model
which is not currently in existence.
A-9
-------
TABLE A-2. POTENTIAL PROBLEMS ENCOUNTERED IN FIELD VALIDATION OF WATER AND CHEMICAL TRANSPORT MODELS
Problem
Implications
1. Lateral and vertical variability of
transport and retention properties
2. Macropores, cracks, plant and animal holes,
holes, interspersed with bulk structural
characteristics
3. Time dependent boundary conditions and soil
response characteristics
3>
»-•
O
4. Crop growth
5. Difficulty in sampling or measuring
important transport properties below surface
zone
6. Inability to measure water drainage flux
jm situ
7. Scale effects on transport
1. Many sample replicates are needed to characterize
mean values. Functional relations may be obscured
by variability.
2. Isolated soil structural features may have sig-
nificant influence on material transport which is
difficult to characterize by measurement.
3. Oscillations in input create cyclic water and
temperature regimes near surface. Hysteresis and
temperature effects on transport may influence
observations. Large water fluxes during storms
may introduce rate-limited adsorption and macro-
pore flow.
4. Growth will change rooting density and water
uptake patterns, and alter evaporation and
transpiration partition at canopy surface.
5. Surface-measured properties must be used to
represent greater depths, creating possible
errors.
6. Water and chemical mass fluxes cannot be inferred
from point measurements of concentration.
7. Field scale transport of chemicals may involve
new transport mechanisms not present in lab-scale
equations.
-------
Structural Effects
In any natural setting, inter-aggregate structural features such as soil
macropores, cracks, plant root holes, or animal burrows will be interspersed
with the bulk structural characteristics. These geometric features, although
very small in volume can have a significant effect on the transport: of chem-
icals and particularly on those which are highly adsorbed (Beven arid Germann,
1982). For soil containing large cracks and holes, the dominant flow mechan-
isms for water and bulk chemical flow may not even be those described in the
physical transport theory given in Appendix 0. Characterization of the in-
fluence of these structural voids on transport properties has not been at-
tempted on a field scale although proposals have been made to characterize
them in a probabilistic sense (Jury, 1982).
A major problem with such structural voids is that they are very diffi-
cult to detect through antecedent soil measurements or calibrations;. In the
future, it will likely require several detailed field experiments in order to
characterize their statistical probability in several representative soil
types. In the interim, it must suffice only to lump them with other causes of
variability in the field and to use the extent of measurement variability as a
guide in designing sample numbers (Warrick and Nielsen, 1980).
Time Dependent Boundary Conditions and Soil Response Characteristics
In any natural setting, surface boundary conditions are dominated by the
diurnal cycling of radiation at the soil surface. As a response to this
diurnal variation, temperature profiles near the surface fluctuate between
limits, and the soil profile near the surface undergoes wetting and drying
cycles. As mentioned above, both water and chemical transport and retention
parameters have been observed to be hysteretic. Therefore, in principle,
hysteresis can affect measurements of parameters of interest near the soil
surface when conducted under natural conditions. Further, and perhaps more
significantly, temperature variations in response to diurnal cycling are likely
to be extreme in the top 30 cm of soil. The influence of such temperature
variations on transport and retention is only partially understood and greatly
increases the requirements of a data base. With the exception of several
isolated research studies (Cary, 1965; Jury and Miller, 1974), the influence of
temperature variations on water transport in the liquid phase is not well
understood. Even less well understood is the influence of temperature varia-
tions on chemical transport. Thus, virtually all models assume isothermal
non-hysteretic conditions in the absence of any theoretical or experimental
information available. Nevertheless, these effects may influence observations
made near the soil surface.
Time-dependent water inputs due to intense storms may also result in large
infiltration fluxes near the soil surface. In addition to enhancing transport
by mass flow and dispersion of chemicals, such large water fluxes may increase
the influence of structural voids discussed above and may also enhance the
rate-limited non-equilibrium adsorption by decreasing the residence time in the
soil compartments near the soil surface. As mentioned above, none of these
responses to the effects of water flow variations are taken into account in
most models.
A-ll
-------
Crop Growth
Models of water and chemical movement through cropped fields require
experimental values of various crop growth parameters which are difficult to
obtain. Particularly difficult to measure is the rooting depth and root den-
sity which have critical influences on water uptake patterns. During root
growth, these density patterns may change dramatically over short periods of
time (Taylor and Klepper, 1971), and unless prohibitive numbers of replicated
measurements are made, such time variations may not be characterized by meas-
urement. Similarly, during crop growth, the extent of soil cover at the sur-
face is increasing, and the partitioning between evaporation and crop trans-
piration is continually changing. Modeling of this changing partition may be
required for certain transport scenarios.
Measurement Difficulties at Great Depths
In any practical field experiment, the bulk of soil measurements is taken
near the soil surface because of increased sampling difficulties at greater
depths. As a consequence, many important transport properties and much struc-
tural information about the soil are inadequately characterized at depths below
the soil surface. Nonetheless, many transport models and particularly those
involving hazardous chemicals are used to project chemical movement all the
way to ground water. In such cases, the data base measurements obtained for
calibrating and validating such models will necessarily be concentrated near
the soil surface, and the properties of the soil below such surface measure-
ments will be obtained by inference or extrapolation. This procedure could
have a critical effect on transport predictions far below the soil surface
particularly when a large horizon shift or bedrock layer is encountered.
Inability to Measure Water Drainage Flux
Unless a drainage lysimeter is installed, water drainage flux is not
measurable in the field. This greatly decreases the amount of validation and
testing that can be conducted on any chemical transport model since it pro-
hibits measuring chemical mass flux or water mass flux at various locations in
the field. Limiting such estimates to field-wide averages conducted over
long time intervals is likely to provide substantially less information about
model validity.
Spatial Variability and Volume Averaging
Recent research on leaching of non-adsorbed solutes at the field scale has
led to the tentative conclusion that the classical one-dimensional convection-
dispersion equation (D-35) which uses a field-wide average downward leaching
flux may not be applicable near the soil surface when substantial downward
movement due to mass flow is involved (Dagan and Bresler, 1979). As a result,
either a multi-dimensional form of equation (D-35) may have to be used, or some
sort of new representation based on a statistical model may have to be at-
tempted. The implications for this and chemical leaching studies are largely
unexplored at this time. However, Jury et al. (1982), conducted a field trial
over the top 3.5 meters of soil using a stochastic-convective model and showed
that it gave a superior description of observed chemical leaching to the
A-12
-------
conventional convection-dispersion equation above. Current approaches to
solute transport at the field scale have been recently reviewed in an article
by Jury (1984).
SAMPLING METHODOLOGY
The parameters listed in Table A-l and discussed in the text above should
ideally be measured in an undisturbed manner in the actual field setting where
the model will be applied. In most cases, this is not possible, and the values
must be inferred either from disturbed samples taken back to the laboratory or
perhaps even indirectly from behavior of other parameters which is interpreted
through a model. Measurement methods currently in use for the major parameters
and their limitations will be discussed below.
Static Soil Properties
The easiest of the parameters to measure is the static soil properties
which do not vary as a function of time, water content, or temperature at a
given location. For these properties it suffices to take a number of soil
cores at representative locations in the field and analyze them in the labora-
tory. Among the parameters which can be measured this way are: particle size
distribution or percent clay which is usually determined by a hydrometer or
pipette method (Day, 1965); specific surface area of soil which may be deter-
mined by a variety of methods including adsorption of ethylene glycol monoethyl
ether (Taylor and Ashcroft, 1977); and organic carbon fraction which may be
determined by C02 detection following combustion (Allison, 1965). In addition,
soil bulk density or porosity may be determined by measuring the mass of dry
soil in a known volume of a core sample (Blake, 1965). The major requirement
in this measurement is making sure that the soil core carves out art undisturbed
volume without compression of the soil inside. For this reason, measurements
of porosity or bulk density are usually in error when taken by compression core
measurements (such as pounding) deeper than a few meters below the surface.
Saturated hydraulic conductivity may be measured either by steady state
ponding on a large surface plot or by transient techniques such as the auger
hole or piezometer method (Boersma, 1965). These methods each take an average
saturated conductivity over a large volume and in addition display significant
variability across the field. For example, Nielsen et al. (1973), observed a
coefficient of variation of 133 percent in saturated hydraulic conductivity
measurements taken over a 150 ha field. They estimated that sever.il hundred
measurements would have to be taken to estimate the field mean saturated
hydraulic conductivity within 20 percent.
Transport and Retention Functions
Water and chemical transport coefficients which are functions of water
content or matric potential display significant variability in a field setting.
Even more importantly, however, it has been repeatedly shown that laboratory
measurements taken on either disturbed or undisturbed soil cores give values
different from measurements taken on the soil profile itself. Therefore, field
measurement of these transport parameters is essential. Hydraulic conductivity
as a function of water content or matric potential may be obtained dynamically
A-13
-------
by the instantaneous profile method which measures water contents and matric
potentials during drainage following saturation (Rose et al., 1965). However,
recent investigations have shown that useful information about hydraulic con-
ductivity as a function of water content may be obtained by using the far
simpler unit gradient method for measuring hydraulic conductivity (Libardi et
al., 1980). Matric potential (h) as a function of water content (9) may be
measured by measuring the water content near a tensiometer reading; this could
be done by use of a neutron probe or by taking soil cores. Each identifiable
soil horizon seems to possess a different matric potential water content rela-
tionship, h(0), so that measurements on depth increments of 30 or more centi-
meters often give different functional relationships between h and 0 even at a
single plot. Furthermore, variation among replicates at a given depth across
the field is substantial (Nielsen et al., 1973).
Hydrodynamic dispersion coefficients do not appear to be strongly depend-
ent on water content but are functions of the average water flux. The dispers-
ivity (see equation D-13) can be measured only by observing solute transport
and interpreting the measurements with the convection-dispersion equation
(D-35). Since dispersion arises from volume averaging, the extent of dispers-
ion depends on the size of the region being simulated. Therefore, dispersion
coefficients are much smaller on the laboratory scale than on the field scale.
At the present time there appears to be no a priori way of estimating the
dispersion coefficient, and it must be measured in situ by observing chemical
movement on the field scale.
Tortuosity Functions
The tortuosity functions which express the reduction in diffusion due to
soil (see Appendix D) have been fitted to many different models over the years
(see review by Sallam et al., 1984). However, few or no measurements of these
functions have taken place in field conditions because the volumetric water
content may vary in three dimensions. Therefore, tortuosity is unlikely to be
measurable in a natural setting and is likely to be replaced by an assumed
model form. At the present time the Millington and Quirk model (equation D-9)
appears to be the most versatile function for representing the effects of water
content on gas or liquid diffusion in soil.
Chemical Properties
The Henry's constant KH (equation D-22) describing the ratio of gaseous
and liquid concentrations at equilibrium is assumed to be the same in soil as
it is in air. For organic chemicals such as pesticides, Spencer and Cliath
(1970) have shown that for several chemicals, Henry's Law of proportionality
holds all the way up to saturation chemical concentrations. Therefore, for
these chemicals at least, the Henry's constant may be set equal to the ratio
between saturated vapor density and solubility. The major problem with the
estimation of Henry's constant at the present time appears to be agreement on a
standard protocol for measuring vapor pressure and solubility. As pointed out
by Spencer and Cliath (1981), improvement of protocol has led to a decrease in
the error among Henry's Law determinations for a single chemical by different
investigators.
A-14
-------
Chemical Adsorption Coefficient
At the present time, most chemical adsorption coefficient measurements are
made by batch equilibrium shaking of completely dispersed soil samples allowed
to come to equilibrium in the laboratory. The resulting functional relation-
ship between liquid concentration and adsorbed concentration is fitted to an
adsorption model such as the Freundlich relationship (equation D-24). However,
recent work has shown that adsorption under natural conditions on structured
soils is likely to be significantly less than that estimated from the batch
equilibrium measurement because the adsorption surfaces are incompletely ex-
posed during dynamic transport. For this reason, column flow-through methods
have also been used to measure adsorption behavior (Green and Corey, 1971), and
these methods do not always measure the same values as the batch methods. For
example, Jury et al. (1983b), showed that pesticide Kg values at 36 sites on a
field soil obtained by batch equilibrium methods were different frcm those
obtained by observing leaching patterns of undisturbed soil columns taken from
the same 36 sites. In addition, each measurement showed substantial variabil-
ity between replicates (coefficients of variation of 26 and 31 percent, respec-
tively). The disagreement between a batch equilibrium measurement of adsorp-
tion and one inferred from observing transport of an adsorbed chemical, a
so-called dynamic method, is due in part to rate-limited transport of the
organic chemical to isolated adsorption sites. Since these rate limitations
are likely to be more extreme under natural conditions than when the structure
is broken up, it seems essential to perform adsorption measurements: in the
field. However, to date no such adsorption measurements have been made, and
information on chemical adsorption continues to be inferred from laboratory
measurements and is usually based on the batch equilibrium method.
LIMITING MODEL ASSUMPTIONS
Because of the extreme limitations imposed by making measurements on the
field scale under natural conditions, soil models of water and chemical trans-
port are forced to assume a set of simplifying conditions which under certain .
circumstances may result in significant disagreement between prediction and
observation. These model assumptions are summarized in Table A-3 and will be
reiterated below.
TABLE A-3. LIMITING TRANSPORT MODEL ASSUMPTIONS IMPOSED BY
INADEQUATE MEASUREMENTS
= = == = = = = = = == = = = = = = = = = = = = = = == = = = =: = = = := = = = = = = == = = = = = = := = = = = = = := = = =
One-dimensional flow
Isothermal Soil
Non-Hysteretic Transport Functions
Homogeneous or Layer-Heterogeneous Soil
Equilibrium Non-Hysteretic Adsorption of Chemical
Complete Accessibility of Porous Medium (No immobile water)
Convective-Dispersive Transport of Chemical
A-15
-------
Virtually all models describing field-scale transport, except those with
a natural multi-dimensional symmetry (e.g., water flow from a furrow) assume
that water and chemical flow is one dimensional. This has the effect for
large-scale simulations of requiring field averaged representative transport
coefficients and a field-average retention curve. Although a one-dimensional
field-scale model has had some success in predicting upward water flow and
subsequent evapotranspiration (Nimah and Hanks, 1973), field-scale one-
dimensional transport has not been demonstrated for downward flow for either
water or chemical. In fact, Warrick et al. (1977), illustrated that use of
field-averaged transport coefficients in an average one-dimensional model
would not predict the true mean behavior of a variable field. Instead, what
must be done is to examine individual behavior at different parts of the field
and average these results to estimate field-scale drainage. The main limita-
tion to using a one-dimensional flow approximation is the size of the field and
the extent of the variability of the parameters within the field. Numerous
current research efforts are under way to improve model representation of this
problem (see review by Peck, 1983).
Virtually all water and chemical models assume isothermal soil conditions.
The theory for coupling effects between water and temperature movement was
worked out roughly in 1957 by Philip and de Vries. However, coupling effects
require a host of transfer coefficients which are extremely difficult to meas-
ure and whose functional dependence on water content or temperature is largely
unknown. For this reason, transport processes are usually considered isother-
mal, and a mean temperature value is used. Temperature effects are likely to
be most significant in simulations of transport of volatile chemicals whose
vapor pressure increases nonlinearly with temperature. Further, since the
volatilization site is the soil surface where temperature variations are ex-
treme, it is unclear the extent to which replacing such variations by an aver-
age value introduces error. A recent Ph.D. thesis (Streile, 1984) addresses
this problem.
Non-hysteretic transport functions are assumed because of the extreme
difficulty in obtaining a hysteretic representation of the relationship between
two variables. Although the problem still generates intense academic interest,
it seems unlikely in the future that hysteresis will be included in any trans-
port model. Even with the aid of computer simulation, it is at the present
time unclear the extent to which hysteresis is important in transport.
Because of the one-dimensional flow assumption, soil is necessarily
assumed to be laterally homogeneous. This allows, at most, a heterogeneous
layer representation for soil and does not permit lateral flow to occur. Since
at progressively larger depths from the soil surface it becomes more and more
difficult to calibrate transport coefficients, it seems likely in the future
that representative transport coefficients will be assigned based on whatever
geologic information is available about the strata being simulated.
Virtually all of the chemical transport models assume equilibrium non-
hysteretic adsorption of the chemical. An equilibrium model could possibly
be used in a rate-limited application provided that one obtained adsorption
measurements under conditions of transport similar to those which will be
simulated, i.e., at the same range of water fluxes. Such calibrations would
A-16
-------
automatically include the effect of rate-limited adsorption on the estimate of
the adsorption coefficient. Adsorption hysteresis may be quite important for
certain chemicals which desorb much less readily than they adsorb. This has
the effect of enhancing their retention in layers where they are deposited. It
seems likely that simple models of adsorption-desorption may be included in
transport models in the future.
In most chemical transport models, it is assumed that the entire wetted
pore space is available to the chemical, i.e., no immobile water. The major
effect of making such an approximation is to represent the pore water velocity
as the water flux divided by the total water content which could underestimate
water velocity when substantial pore bypass exists. This assumption is cer-
tainly very inaccurate in regions where macropore effects are extreme; in such
cases much of the water and chemical may be carried in a relatively small frac-
tion of the pore space. There exist a number of mobile-immobile water models
which include effects on chemical diffusion and adsorption that have been
tested under laboratory conditions (van Genuchten et al., 1974; van Genuchten
and Wierenga, 1976). However, these models introduced a number of new param-
eters which are not directly accessible to measurement under field conditions.
These parameters may be inferred only by applying the model to fie'ld data and
by fitting the model to observation by minimizing the difference between the
model and the data while varying parameters. Given the large number of factors
which contribute to the disagreement between a model and an observation on the
field scale, such minimization procedures are likely not to be useful in iden-
tifying subtle unmeasurable features of the transport process.
Analogous to the laboratory problem where most models have been developed
and tested, virtually all chemical transport models used in the field assume
convective-dispersive transport. Furthermore, most of these model-; are one
dimensional, representing downward movement by a mean one-dimensional water
flow and chemical mixing by an average dispersion coefficient. Field measure-
ment of this dispersion coefficient is very difficult and introduces consider-
able error. Furthermore, there are fundamental objections to its use at least
near the soil surface (Oagan and Bresler, 1979). A recent set of solute trans-
port models proposed for application near the soil surface (i.e., 0 to 3 m) do
not employ a dispersion coefficient but use a purely convective model based on
the presumption of substantial lateral variability in downward water flow
(Dagan and Bresler 1979; Jury et al., 1982). In the interim, convective-
dispersive models should probably be calibrated under conditions similar to
those in which the simulation will be made in order to minimize the disagree-
ment between the prediction and the observation.
RECENT IMPROVEMENT IN FIELD MEASUREMENT STRATEGY
Recently, the theory of regionalized variables (Journel and Hjijbregts,
1978) has been proposed as a method for maximizing the information content of a
finite set of discrete soil samples taken on the field scale. This method,
rather than using the classical statistical assumption of independence, inves-
tigates the spatial correlation between discrete measurements and constructs
field autocorrelation functions to study the relationship of a measurement made
at one place to another. The autocorrelation function is in turn used to
generate a minimum variance estimate of the parameter at a place in which it
A-17
-------
was not measured using the information obtained from the measurements them-
selves. This method is called kriging and shows promise for improving the
information content of a set of measurements. Illustrations of the kriging
method are given in a recent article by Vieira et al. (1983).
SUMMARY
The study of water and chemical transport on the field scale is a science
in its infancy. Nevertheless, at this time there have been identified a sig-
nificant number of soil, environmental, and management parameters which in-
fluence chemical transport through soil, and a large body of information has
been assembled for finding the relationship between these parameters and the
chemical transport. From years of study under controlled conditions, a sub-
stantial theoretical framework has been developed for describing water and
chemical transport. This theory uses differential equations developed from a
macroscopic mass balance and uses flux equations obtained as an extension to
equations which are valid in one phase systems. Chemical transport generally
involves liquid, vapor, and adsorbed phases so that a relationship between the
phases is also used in addition to the transport equation. Although rate
limitations can be expected to alter this equilibrium relation under general
conditions, models frequently simplify this using linear equilibrium partition
coefficients. For such systems, the three-phase transport may be replaced by
an equivalent representation in terms of total concentration using equivalent
three-phase transport velocities and dispersion coefficients.
Recent research in field-scale transport suggests that the extensive
spatial variability of soil physical and chemical properties may require a new
formulation for use when describing average transport across large cross-
sectional areas. Such research is only in the formulation stage right now and
should be further verified before general conclusions can be drawn. In the
interim, substantial field testing of existing and proposed models of transport
should be conducted together with substantial calibration and validation. Only
when a large body of experimental information is available will workers be able
to further refine the models.
A-18
-------
REFERENCES
Allison, L. E. 1965. Organic carbon. In: Black, C. A. (Ed.) Methods of
Soil Analysis. ASA Monograph 9, Amer. Soc. Agron. Madison, WI.
Bear, J. 1972. Dynamics of Fluids in Porous Media, American Elsevier, NY.
Seven, K., and P. Germann. 1982. Macropores and water flow in soils.
Water Resour. Res. 18:1311-1325.
Biggar, J., and Cheung. 1973. Adsorption of picloram on Panoche, Ephrata and
Palouse soils. Soil Sci. Soc. Am. Proc, 37:863-868.
Bigger, J. W., and D. R. Nielson. 1976. Spatial variability of the leaching
characteristics of a field soil. Water Resour. Res. 12:78-84.
Bird, R. B., W. E. Stewart, and E. N. Lightfoot. 1960. Transport Phenomena.
John Wiley and Sons, New York.
Blake, G. 1965. Bulk Density. Jn_: Black, C. A. (Ed.), Methods of Soil
Analysis. ASA Monograph 9, Amer. Soc. Agron., Madison, WI.
Boersma, L. 1965. Field measurement of hydraulic conductivity. I_n_: Black,
C. A. (Ed.), Methods of Soil Analysis. ASA Monograph 9, Amer" Soc.
Agron. Madison, WI.
Gary, J. W. 1965. Water flux in moist soil. Thermal versus suction gradients.
Soil Sci. 100:168-175.
Cosby, B. J., G. M. Hornberger, R. B. Clapp, and T. R. Ginn. 1984. A statis-
tical exploration of the relationships of soil moisture characteristics to
the physical properties of soils. Water Resour. Res. 6:682-690.
Oagan G., and E. Bresler. 1979. Solute dispersion in unsaturated soil at
field scale. Soil Sci. Soc. Amer. J_. 43:461-466.
Day, P. R. 1965. Particle fractionation and particle size analysis. Jn: C. A.
Black (Ed.), Methods of Soil Analysis. ASA Monograph 9, Amer.. Soc. Agron.
Madison, WI.
Doorenbos, J., and W. 0. Pruitt. 1976. Crop Water Requirements. Irrig. and
Drainage Paper 24, FAO Rome.
El Abd, H. 1984. Spatial Variability of the Pesticide Adsorption Coefficient.
Ph.D. Thesis, Univ. of Calif., Riverside.
A-19
-------
Gajem, Y. M., A. W. Warrick, and D. E. Myers. 1981. Spatial dependence
of physical properties of a typic torrifluvent soil. Soil Sci. Soc.
Amer. J_. 45:709-715.
Gardner, W. R. 1959. Solutions of the flow equation for the drying of soils
and other porous media. Soil Sci. Soc. Amer. Proc. 23:183-187.
Green, R. E. 1974. Pesticide-clay-water interactions. In: W. D. Guenzi
(Ed.), Pesticides in Soil and Water. Soil Sci. Soc. Amer.. Madison, WI.
Green, R. E., and J. C. Corey. 1971. Pesticide adsorption measurement by
flow equilibration and subsequent displacement. SSSA Proc. 35:561-565.
Hamaker, J. W., and J. M. Thompson. 1972. Adsorption. _Irr. C. A. I. Goring
and J. W. Hamaker (Eds.), Organic Chemicals in the Soil Environment.
Marcel Dekker, Inc., New York
Harper, L. A., A. W. White, R. R. Bruce, A. W. Thomas, and R. A. Leonard. 1976.
Soil and microclimate effects on trifluralin volatilization. J. Environ.
Qual. 5:236-242.
Hillel, D. 1971. Soil and Water. Academic Press, New York.
Journel, A. G., and C. J. Huijbregts. 1978. Mining Geostatistics. Academic
Press, London.
Jury, W. A. 1973. Simultaneous Transport of Heat and Moisture Through a
Medium Sand. Ph.D. Thesis. Univ. of Wisconsin.
Jury, W. A. 1982. Use of solute transport models to estimate salt balance
below irrigated cropland. Adv. Irrig. 1:87-104.
Jury, W. A. 1984. Field Scale Water and Solute Transport through Unsaturated-
Soil. Proc. Int. Conf. Soil Salinity Under Irrigation. Bet-Dagan,
Israel.
Jury, W. A., and E. E. Miller. 1974. Measurement of the transport coefficients
for coupled flow of heat and moisture in a medium sand. Soil Sci. Soc.
Amer. Proc. 38:551-557.
Jury, W. A., L. H. Stolzy, and P. Shouse. 1982. A field test of the transfer
function model for predicting solute transport. Water Resour. Res.
18:369-375.
Jury, W. A., W. F. Spencer, and W. J. Farmer. 1983a. Use of models for
predicting relative volatility presistence and mobility of pesticides
and other trace organics in soil systems. In: J. Saxena (Ed.), Hazard
Assessment of Chemicals, Vol. 2. Academic Press, New York.
A-20
-------
Jury, W. A., H. El Abd, and T. M. Collins. 1983b. Field Scale Transport of
Nonadsorbing and Adsorbing Chemicals Applied to the Soil Surface.
pp. 203-222. Proceedings of NWWA Symposium, Characterization and Monitor-
ing of the Vadose Zone, Las Vegas, NV.
Libardi, P. 1., K. Reichardt, D. R. Nielsen, and 0. W. Biggar. 1980. Simple
field methods for estimating soil hydraulic conductivity. Soil Sci. Soc.
Amer. J_. 44:3-6.
Millington, R. J., and J. M. Quirk. 1961. Permeability of porous solids.
Trans. Farady Soc. 57:1200-1207.
Nielsen, D. R., R. D. Jackson, J. W. Cary, and D. D. Evans. 1972. Soil Water.
Amer. Soc. Agron. Special Pub. Madison, WI.
Nielsen, D. R., J. W. Biggar, and K. T. Erh. 1973. Spatial variability of
field measure soil water properties. Hilgardia 42:215-259.
Nimah, M. N., and R. J. Hanks. 1973. Model for estimating soil water,
plant and atmospheric interrelations. 1. Description and sensitivity.
Soil Sci. Soc. Amer. Proc. 37:522-527.
Peck, A. J. 1983. In: Advances in Irrigation. Vol. 2. (In press).
Philip, J. R., and D. A. de Vries. 1957. Moisture movement in porous
materials under temperature gradients. Trans. Amer. Geophys. Union
38:222-232.
Richter, G. 1984. A Microlysimeter and Field Study of Water and Chemical
Movement Through Soil. MS Thesis, Univ. of California, Riverside.
Rao, P. S. C., D. E. Rolston, R. E. Jessup, and J. M. Davidson. 1980.
Solute transport in aggregated porous media: Soil Sci. Soc. Amer. J.
44:1139-1146.
Rao, P. S. C., and R. J. Wagenet. 1985. Spatial variability of pesticides in
field soils: methods for data analysis and consequences. Ueed Sci.
33: (In press).
Rose, C. W., W. R. Stern, and J. E. Drummond. 1965. Determination of hy-
draulic conductivity as a function of depth and water content for soil J^
situ. Aust. J_. Soil Res. 3:1-9.
Russo, D., and E. Bresler. 1980. Soil hydraulic processes as stochastic
processes. Soil Sci. Soc. Amer. J_. 45:682-687.
Sal lam, A., W. A. Jury, and J. Letey. 1984. Measurement of gas diffusion
coefficient under relatively low air-filled porosity. Soil Sci. Soc.
Amer. J. 48:3-6.
A-21
-------
Shouse, P., W. A. Jury, L. H. Stolzy, and S. Dasberg. 1982. Field measurement
and modeling of cowpea water use and yield under stressed and well watered
conditions. Hilgardia 50:1-24.
Spencer, W. F., and M. M. Cliath. 1970. Desorption of lindane from soil.
Soil Sci. Soc. Amer. Proc. 34:574-578.
Spencer, W. F., and M. M. Cliath. 1973. Desorption of lindane from soil
as related to vapor density. Soil Sci. Soc. Amer. Proc. 34:574-578.
Spencer, W. F., and M. M. Cliath. 1981. In: Test Protocols for Environmental
Fate and Movement of Toxicants. Assoc. of Analy. Chem., Arlington, VA.
Streile, G. 1984. Temperature Effects on Pesticide Transport. Ph.D. Thesis,
Univ. of California, Riverside.
Taylor, S. A., and G. L. Ashcroft. 1977. Physical Edaphology. Freeman and
Co., San Francisco, CA.
Taylor, H. H., and B. Klepper. 1971. Water uptake by cotton roots during an
irrigation cycle. Aust. ^. Biol. Sci. 24:853-859.
van Genuchten, M. Th., J. M. Davidson, and P. J. Wierenga. 1974. An evaluation
of kinetic and equilibrium equations for the prediction of pesticide
movement through porous media. Soil Sci. Soc. Amer. Proc. 38:29-34.
van Genuchten, M. Th., and P. J. Wierenga. 1976. Mass transfer in sorbing
porous media. Soil Sci. Soc. Amer. J_. 40:437-480.
Vieira, S. R., J. R. H. Hatfield, D. R. Nielsen, and J. W. Biggar. 1983.
Geostatistical theory and application to agronomy. Hilgardia 51:1-75.
Wagenet, R. J., and J. J. Jurinak. 1978. Spatial variability of solute salt
content in Mankos Shale watershed. Soil Sci. 126:342-349.
Warrick, A. W., G. J. Mullen, and D. R. Nielsen. 1977. Scaling field measured
soil hydraulic properties using a similar media concept. Water Resour.
Res. 13:355-362.
Warrick, A. W., and D. R. Nielsen. 1980. Spatial variability of soil physical
properties in the field. In: Applications of Soil Physics. Academic
Press, New York.
A-22
-------
APPENDIX B
VOLATILIZATION FROM SOIL
by
W. A. Jury
PROCESS DESCRIPTION
Definition
Volatilization is defined as the loss of chemicals in vapor form from soil
surfaces to the atmosphere. This process is ultimately limited by the chemical
vapor concentration which is maintained at the soil surface and by the rate at
which this vapor is carried away from the soil surface to the atmosphere. The
potential volatility of a chemical is related to its inherent saturated vapor
pressure, but the actual volatilization rate from soil in any specific circum-
stance will depend on all soil, atmospheric, or management factors which
influence the behavior of the chemical at the soil-air-water interface (Spencer
et al., 1982).
Rate Process Interactions at the Soil Surface
It is useful to picture the volatilization process from soil as depending
on the balance between two rate processes: the flux of chemical from the soil.
body to the soil surface and the flux of chemical vapor away from the soil
surface to the atmosphere (Spencer et al., 1973). The relative sizes of these
two rate processes determine whether or not chemical concentrations will build
up or deplete in the surface layer and hence affect vapor density at the site
of volatilization.
A number of soil, environmental, and management parameters influence the
volatilization process through their influence on these two rate processes and
the resulting soil vapor density at the surface. Prior to a detailed discussion
of this influence, however, it 1s helpful to develop the fundamental descrip-
tions of the soil chemical vapor density or concentration, the transport mecha-
nisms to the soil surface, and the transport mechanisms from the soil surface
to the air.
Soil Chemical Vapor Density--
When a given quantity of organic chemical is mixed with moist soil, it
ultimately partitions into adsorbed, solution, and vapor phases so that one may
write the total quantity of chemical per unit soil volume (Cj) as
B-l
-------
CT = Pb CA + QCL + a CG (B-I)
where CA is adsorbed chemical concentration (ng/g soil), Ci is dissolved chem-
ical concentration (ug/cm3 solution), CG is vapor density (ug/cnr/soil air),
pb is soil bulk density (g/orr), 0 is volumetric water content, and a is
volumetric air content. The units used in the individual phase concentrations
are conventional so as to correspond with the usual methods of measurement.
The relationship between vapor density and associated solution concentra-
tion at equilibrium is usually given by Henry's Law
CG • KHCL (B-z)
where KH is Henry's constant which in this system of units is dimensionless.
The applicability of Henry's law to soil systems has been confirmed for
a variety of chemicals and circumstances (Call, 1957; Goring, 1962; Leistra,
1970; Spencer and Cliath, 1970). Furthermore, Spencer and Cliath (1970)
showed that for the organic compounds they studied, the relationship equation
(B-2) persists from very trace concentration levels all the way to saturation.
The equilibrium relationship between the solution concentration C|_ and the
adsorbed concentration CA is called the adsorption isotherm
CA = f(CL) (B-3)
where a variety of functional forms have been used to describe this relation-
ship (see chapter on Adsorption). However, it is quite common, particularly
for pesticides and other trace organics, to assume a linear relationship be-
tween adsorbed and liquid concentration called a distribution coefficient as
defined in equation (B-4)
CA = KDCL (B-4)
where Kn, is the distribution coefficient which in this system of units has a
dimension of (cm3/g). Combination of equations (B-l) through (B-4) allows a
linear relationship to be formed between total organic chemical concentration
and vapor concentration
CT = [pbKQ/KH + Q/KH + a]CG = RGCG (B-5)
where RG is the vapor partition coefficient as defined by Jury et al . (1983a).
The utility of the formulation given in equation (B-5) is that all soil,
chemical, and environmental processes influencing the vapor concentration in
equilibrium with a given total amount of chemical are contained in the quanti-
ties defined by the partition coefficent. To further separate the soil and
chemical influences, the distribution coefficient KQ is sometimes expressed as
KD • f0cKoc (B-6)
B-2
-------
where KQC is the organic carbon partition coefficient, and foc is the organic
carbon fraction. Trace organics and pesticides show a much smaller coefficient
of variation in their Koc coefficient between different soils than with their
distribution coefficient KQ (Hamaker and Thompson, 1972). Thus, as an approxi-
mate index, one may take the organic carbon partition coefficient as a property
of the chemical irrespective of the soil. The limitations of this formulation
have been discussed in Rao and Davidson (1980) and in Jury et al. (1983a).
Vapor Movement from Soil to Atmosphere—
Although the motion of air in the atmosphere is generally turbulent, close
to the soil surface there exists a relatively stagnant boundary layer through
which gaseous vapor must move by molecular diffusion (Hartley, 1969). Jury
et al. (1983a), represented this vapor flux away from the soil surface using
Pick's law of diffusion through the boundary layer
AIR
J = ---- [CG(O) - CG(d)] (B-7)
d
AIR
where DC is the molecular diffusion coefficient of the gaseous chemical in
air, CGIO) is the vapor concentration at the soil surface discussed in the
previous section, and Cc(d) is the gaseous concentration in the free air above
the stagnant boundary layer. For most chemicals present in trace concentra-
tions in soil, this free air concentration is small enough to be neglected
compared to the surface concentration. The boundary layer thickness is an
idealization to represent the degree of mixing of the surface layer. An
alternative to equa tion (B-7) would be to use a transfer coefficient
AIR
hj = DQ /d between the surface and the atmosphere.
Chemical Movement from Soil to the Soil Surface--
Chemicals present in the soil generally move by three mechanisms: (1)
gaseous diffusion within air voids, (2) liquid diffusion within soil solution,
and (3) mass flow of dissolved chemical within moving soil solution. These
mechanisms were discussed in considerable detail in Appendix A and will be
briefly summarized.
Gaseous diffusion—Movement of gases through soil is described by an
extension of Pick's Law of diffusion
AIR acG
jv = _ Tio — (B-8)
G dz
where ^ is a tortuosity factor to account for reduced cross-sectional area and
increased pathlength in soil. An empirical model devised by Millington and
Quirk (1961)
n = a 10/3/^2 {B_gj
B-3
-------
where 4> equals porosity, has received substantial experimental verificati
on in soil over a large range of air contents, a (Letey and Farmer, 1974; Jury
et al., 1980; Fanner et al., 1980; Shearer et al., 1973).
Liquid diffusion—In a manner similar to vapor diffusion in soil, a
liquid diffusion equation is written as modification of Fick's Law
O10/3 8C,
j, = ...... DWATER -t (B-10)
-------
chemical is in category 2 if the stagnant boundary layer acts as a partial
barrier to transport which allows chemical concentrations to build up at the
soil surface. Chemicals which belong in category 1 are sometimes called "well
mixed."
Soil. Environmental, and Management Factors Influencing Chemical Volatilization
from Soil
Table B-l summarizes the various parameters influencing chemical vola-
tilization from soil divided, as in the case of the previous chapter, into
groups of soil parameters, environmental parameters, and management parameters.
This table is similar but not identical to Table A-l of the previous chapter.
Several parameters discussed there are not included here because they have a
negligible or undetermined influence on the volatilization process.. In the
following section, the effect of each of the parameters in Table B-l on the
volatilization process will be discussed with specific emphasis on the critical
transport mechanisms or partitioning processes involved in the rate process
interactions at the soil surface. It will be helpful to keep in mind the
volatilization classification developed in the previous section which entails
grouping chemicals into the well-mixed category 1 if their Henry's constant
KH > 1(T5.
TABLE B-l. PRINCIPAL SOIL, ENVIRONMENTAL, AND MANAGEMENT PARAMETERS
INFLUENCING CHEMICAL VOLATILIZATION FROM SOIL
Soil Parameters Environmental Parameters Management Parameters
Water content Temperature Chemical concentration
Bulk density or porosity Wind Depth of incorporation
Clay content Evaporation Irrigation management
Adsorption site density Precipitation Soil management
Soil structure
Soil Parameters--
Soil water content—Volumetric soil water content (0) influences the vola-
tilization process primarily by affecting the gaseous and liquid diffusion
coefficients which regulate transport of material upward to the soil surface.
As indicated in equations (B-8) and (B-9) and (B-10), the soil diffusion coef-
ficients depend non-1inearly on volumetric water content or air content. Hence
at higher water contents, transport by vapor diffusion drops dramatically while
transport by liquid diffusion increases. However, many chemicals have non-
negligible concentrations in both the liquid and vapor phases so that each
transport mechanism can contribute to the movement of material upward to the
volatilization surface. For this reason, Jury et al. (1983a), recommended
looking at the effective vapor-liquid diffusion coefficient describing the
total transport to the surface by diffusion in both phases. The mathematical
development of this effective diffusion coefficient is described in Appendix D.
Figure B-l, taken from Jury et al. (1983a), shows the effective diffusion
B-5
-------
00
Ot
0
O
o
o
x
*
- ETHOPROPHOS
0.2
0.4 0
0.2
0.4 0 0.2 0.4 0 0.2
WATER CONTENT 0
0.4 0
0.2
0.4
Figure B-l. Calculated vapor (dashed line), liquid (dotted line), and total (solid line)
effective diffusion coefficients as a function of water content for 20 chemicals.
The effective diffusion coefficient combines vapor and liquid transport,
reduced by adsorption (taken from Jury et al., 1983a).
-------
coefficient plotted as a function of water content for 20 organic pesticides
which contain representatives of both category 1 and category 2 behavior. This
figure reveals that many of the chemicals (in fact the category 2 chemicals)
have a negligible vapor diffusion coefficient and thus have a steadily increas-
ing effective diffusion coefficient with increasing water content (e.g.,
bromacil). For these chemicals, it is to be expected that transport by dif-
fusion to the soil surface will increase dramatically with increasing water
content. On the other hand, the category 1 chemicals with substantial vapor
diffusion coefficients (e.g., lindane) will compensate for decreasing vapor
transport by increasing liquid transport and hence will have less dependence
on water content.
A dramatic influence of water content on volatilization occurs when the
soil is dried to a point where only a few monolayers of water remain on the
soil solids. In this case, water molecules which preferentially occupy soil
adsorption sites are removed, and the chemical adsorption capacity of the soil
is greatly increased (Spencer and Cliath, 1973). This increased adsorption
capacity greatly decreases gaseous concentrations in the dry region, and vola-
tilization drops to insignificant levels. However, when the soil surface layer
rewets, the process is reversible, and much of the adsorbed solute is allowed
to later volatilize (Spencer and Cliath, 1973; Harper et al., 1976). For this
reason, it may be possible to ignore this aspect of the water content depend-
ence of adsorption provided that the time period during which the 'surface layer
is dry is short-lived. For prolonged drying processes, however, when the soil
surface is not allowed to rewet, the effect on volatilization could be sub-
stantial.
Bulk density or porosity—Decreasing soil porosity or increasing bulk
density will generally decrease volatilization because the diffusive transport
from the soil to the soil surface is reduced by decreasing the cross sectional
area available for flow and by increasing the path length. Transport of chem-
ical upward to the evaporating surface by mass flow may also be decreased if
the soil hydraulic properties are reduced. This dependence is likely to be
quite complex and lends itself only to a qualitative description. A secondary"
influence of decreasing soil porosity is to increase the density of soil min-
eral surfaces which could increase chemical adsorption and decrease transport
to the surface by all mechanisms.
Clay content—The principal direct influence of clay content on chemical
volatilization would be to increase chemical adsorption to mineral surfaces
thereby decreasing the concentration in liquid and vapor phases. However, most
non-polar organic chemicals adsorb primarily to soil organic matter so that no
direct guidelines relating clay content to soil adsorption capacity for organic
chemicals have been obtained in research studies. Also, ionic pesticides and
trace organics, both of which would be expected to adsorb strongly to soil
clays, generally have low vapor pressures and are considered to be non-volatile.
However, clay content is strongly correlated to many water transport and reten-
tion properties which influence the volatilization process. Thus, soils high
in clay tend to have high water contents and hence have an influence on the
volatilization process similar to that discussed in the previous section.
Also, soils high in clay which overlie relatively shallow ground-water tables
can move significant amounts of water and chemicals upward from ground water or
B-7
-------
from the soil to the soil surface. Thus, the combination of high clay content
and shallow depth to ground water could result in a significant transport
upward by mass flow.
Adsorption site density--The adsorption site density refers to the quan-
tity of adsorbing surfaces contributed from soil minerals or soil organic
matter. The influence of adsorption on a soil-incorporated chemical is to
increase the adsorbed and hence immobile phase concentration at the expense of
the liquid and vapor phases. Hence, increased adsorption always implies de-
creased volatilization because it decreases upward transport of chemicals from
the soil to the surface to replace the materials lost by volatilization and
because it also decreases the concentration of chemical vapor at the volatili-
zation site. Jury et al. (1983a), demonstrated that the volatilization of
category 2 chemicals, whose resistance to volatilization lies primarily in the
boundary layer above the soil surface, is more strongly affected by increases
in soil adsorption than is the volatilization of category 1 chemicals even
though the latter are also affected by adsorption.
Environmental Parameters—
Temperature—The influence of temperature on volatilization, although
undoubtedly quite important, has received very little attention in the past.
The discussion to follow, therefore, should be considered somewhat speculative
as it is based on inference from the temperature dependence of the various
processes influencing volatilization rather than on studies of the temperature
dependence of volatilization itself.
For virtually all organic chemicals, saturated vapor density increases
nonlinearly with increasing temperature (Spencer and Cliath, 1973). However,
the water solubility of certain chemicals will increase with increasing tem-
perature whereas for others it will decrease (Calvet, 1980). Thus, the Henry's
constant KH given by equation (B-2) will strongly increase with increasing
temperature thereby increasing the amount of vapor present for a given mass of
chemical as temperature increases. Since the theoretical study of Jury et al.-
(1983a), predicted that volatilization will increase as Henry's constant in-
creases, then it is to be expected that temperature increases will result in
volatilization increases. As a general rule, chemicals will partition propor-
tionally more of their total mass into the vapor phase at higher temperatures.
Thus, chemical transport in the solution phase by mass flow and liquid diffu-
sion will decrease compared to transport by vapor diffusion as temperature
increases.
Wind—The influence of wind above the soil surface on volatilization is
either to increase the mixing or equivalently to decrease the thickness of the
stagnant boundary layer limiting volatilization. Since category 1 chemicals
with large Henry's constants already act well mixed, wind speed will have only
a moderate effect on volatilization for these chemicals. However, category 2
chemicals will have their volatilization rates increase dramatically as wind
speed increases at the boundary layer. This phenomenon was demonstrated under
field conditions by Glotfelty (1981) who studied pesticide volatilization on
two days with dramatically different wind speeds for the same chemicals.
B-8
-------
Evaporation—The influence of water evaporation on organic chemical vola-
tilization is quite complex as demonstrated by Jury et al. (1983a). Figure B-2
shows predicted volatilization fluxes for 20 chemicals during water evaporation
rates of 0, 2.5, and 5 mm per day. As a general rule, the category 2 chemicals
such as atrazine, bromacil, etc. are more dramatically affected by water evap-
oration than are the category 1 chemicals such as DDT, dieldrin, etc., but
other category 1 chemicals such as lindane and phorate will also have enhanced
volatilization under water evaporation conditions. For all of the chemicals,
soil water evaporation has the effect of carrying dissolved organic chemicals
by mass flow to the soil surface where the volatilization process occurs. This
enhances volatilization to varying degrees, depending on the extent of upward
movement and the vapor pressure of the chemical at the volatilization site.
The only chemicals which are completely unaffected by water evaporation are
those such as DDT and trifluralin which are so strongly adsorbed that only
insignificant amounts of upward transfer occur by mass flow.
Precipitation—The effect of precipitation on volatilization is indirect
and serves to transport the chemical below the site of volatilization where
soil layers above it greatly decrease its volatilization. Since it has no
direct effect other than to transport the chemical away from the surface,
rainfall is likely to be a more important limiter of volatilization for chem-
icals which leach readily and have low adsorption.
Management Parameters—
Chemical concentration—Since the volatilization rate of a chemical is
proportional to its vapor concentration at the soil surface, increasing chem-
ical concentration creates higher vapor densities and hence higher volatiliza-
tion. Indeed, if the chemical partitions linearly, then volatilization is
proportional to chemical concentration.
Depth of incorporation—The volatilization rate of chemicals can be
greatly decreased by incorporating them below the surface particularly if they
are highly adsorbed (Cliath and Spencer, 1971). Volatile chemicals which are..
sprayed on the soil surface are particularly prone to losses and should be
plowed under soon after application. For a given mass of applied chemical, the
depth of incorporation also has the effect of reducing the concentration and
thus reducing volatilization as presented in the above argument.
Irrigation management—The effect of irrigation management is similar to
rainfall in the sense of leaching chemicals below the volatilization site. In
addition, the intensity and spacing between irrigations can affect the duration
of the evaporation cycle which can have the effect of redepositing chemicals at
the site of volatilization. Thus, high frequency irrigation would be likely to
have the most significant effect on decreasing volatilization of incorporated
chemicals.
Soil management—A significant amount of research has been conducted over
the last century to study the effect of soil management on soil aeration for
the purpose of maximizing diffusion of oxygen from the atmosphere down to the
soil (Baver et al., 1972). The same factors that increase diffusion of oxygen
downward into the plant root zone will also increase the volatilization of
chemicals from the soil to the atmosphere. Therefore, conditions of maximum
B-9
-------
in-'
IU
^ I0
^^
o
5 IO'5
o _t
\
o»
- io-3
X 10' 5
i irv '
_J IU
U_
0 I0'3
N ,0-5
-1 |n-l
Si"
-J ,
0 IO'3
-5
10
AT DA "7 1 MC
A 1 KA£IIMb
^, :
VARY E
1 1
- DIAZINON
~ :
LINDANE
^_.:r:.
*
PHORATE
nortM Ar*n
DnUMAVsIL
• •
1 1
^ — ^...^_:
DIELDRIN
• METHYL PARA. .
.PROMETRYNE .
• • •
Y' :
CARBOFURAN '
1 1
DIURON
•
£^ — • :
MONURON
•
1 i r
. SIMAZINE
f^, ""
'
1 \ —
EPTC
NAPROPAMIDE .
^— " ;
. **' -
TRIALLATE
r\r\^
\J\J \
<_
i — i —
ETHOPROPHOS "
PARATHION
•
,
S- — ^
'
TRIFLURALIN
iii
10 20 10 20 10 20
TIME (DAYS)
10
20
10
20
Figure B-2. Volatilization fluxes predicted for 20 chemicals in soil at same concentration
for three water evaporation rates E = 0 —, E = 2.5 mm/day , E = 5.0 mm/day...
(taken from Jury et al., 1983a).
-------
aeration, i.e., soil with good structure, surface tilling, etc., will also tend
to maximize the volatilization of chemicals.
PROCESS ROLE IN SOIL MODELS
Mayer et al. (1974), discussed the influence of the upper boundary condi-
tion in volatilization predictions using five different models. Fanner et al.
(1980), developed and tested a model for predicting vapor loss of hsxachloro-
benzene from a landfill covered with soil. The model is based upon the vapor
diffusion equations (B-8) and (B-9) applied in steady state through a soil
cover. Jury et al. (1980), developed and tested a volatilization model for
trial late on two different soils. This model was able to describe time-
dependent volatilization from a chamber initially filled to a uniform concentra-
tion with trial late and subsequently exposed to a moving air surface which
maintained chemical concentration at a low value. The model combinad the
effects of liquid and vapor diffusion and liquid mass flow with a three-phase
equilibrium model equation (B-l) using the convection-diffusion equation
(Nielsen et al., 1972). This model, when expressed in terms of total concen-
tration, is given by equation (B-13)
5?! = DE L£ - VE * (B-13)
a* azZ az
This model has recently been proposed for use in screening large numbers of
chemicals in Jury et al. (1983a). Figure B-2 shows predicted volatilization
fluxes resulting from uniform initial concentrations for 20 different chemicals.
Using the above theory, the influence of adsorption, water content, depth of
incorporation, air flow rate, and other variables on volatilization rates for
different chemicals are discussed in this document.
MEASUREMENT LIMITATIONS
All of the potential problems listed in Table A-2 under the Implications .
heading apply equally well to volatilization measurements or modeling valida-
tion. Particularly imposing are the problems presented by macropores, cracks,
etc. which are likely to create paths of low resistance for vapor transport to
the atmosphere and which would not be well described by bulk tortuosity models
such as those discussed above. Indeed, the influence of structural variability
on vapor movement may be substantial because the gradient of vapor concentra-
tion generally points toward the soil surface so that the concentrated vapor in
the soil would seek the path of least resistance to the atmosphere. The impli-
cations of this on field-scale transport are largely unexplored at this time
because of the small number of field volatilization studies which have been
conducted.
An additional problem to the ones discussed in the previous chapter is
posed by the upper boundary condition for volatilization which assumes a
stagnant boundary layer. This boundary layer is likely to be influenced by a
variety of poorly characterized processes such as wind speed, surface roughness,
and atmospheric turbulence above the zone of stagnation. The most severe
limitation to use of this model for the upper boundary is the fact that the
B-ll
-------
diffusion boundary layer thickness d in equation (B-7) is not directly measur-
able and must be inferred through an analogy. This will be discussed in the
next section. Failure to include some boundary-layer limitation on volatiliza-
tion, however, will likely result in a serious error when transport estimates
are made on category 2 chemicals (Jury et al., 1983a).
A potentially severe limitation to volatilization modeling under field
conditions is the difficulty in estimating the effective adsorption of the
chemical in a structured setting where many of the adsorption sites must be
reached by a relatively slow diffusion process. In such cases, laboratory
estimates of adsorption capacity are likely to poorly define the actual amount
of adsorption occurring in the system (Jury et al., 1983b).
Methods of Field collection of Parameters
Table B-2 lists the parameters required to make volatilization estimates
through modeling under field conditions divided into categories of soil, chem-
ical, and atmospheric parameters. In some cases, information required to make
these parameter estimates may be obtained under laboratory conditions. In
other cases, the methodology required for field estimates is not available.
These will be discussed below.
TABLE B-2. PARAMETERS REQUIRED FOR VOLATILIZATION ESTIMATES IN THE FIELD
r=======r======================================================================
Soil Parameters Chemical Parameters Atmospheric Parameters
Water flux Henry's constant Temperature
Liquid and vapor Diffusion coefficients Boundary layer
tortuosity functions in air and water thickness
Water content Adsorption parameters
Bulk density Degradation rate
Parameters Affecting Volatilization
Soil Parameters—
Water flux—Estimate of the water flux particularly for the case of upward
flow toward evaporating surfaces may be obtained only for bulk average esti-
mates valid for large areas. Meteorological methods such as those discussed by
Tanner (1968) may be used to estimate the potential evaporative loss when the
soil surface is wet. When the soil surface dries, a simple rate-limited evapo-
ration model may be used as discussed by Shouse et al. (1982). Models such as
these will give estimates of average evaporation rate for time periods in
between irrigations or rainfall and may serve as rough estimates of the poten-
tial for movement of dissolved chemical upward to the evaporating surface.
B-12
-------
Liquid and vapor tortuosity functions—The tortuosity functions which
model the reduction in diffusion of vapor or liquid due to the presence of the
porous medium (e.g., equations B-9, B-10) have been measured almost, exclusively
under laboratory conditions. Because of the extreme complexity of three-
dimensional chemical movement by diffusion, it is unlikely that reliable tor-
tuosity function estimates can be made in the field. Instead, what is recom-
mended is to use the Millington and Quirk (1961) tortuosity functions discussed
above (see equations B-9, B-10) which relate tortuosity to liquid or vapor-
filled pore space.
Water content—Water content is required to estimate the tortuosity func-
tions discussed above and is needed in the estimate of partitioning of chem-
icals into vapor and adsorbed phases. This parameter should probably be meas-
ured in the field setting by either neutron scattering methods or by gravi-
metric soil sampling.
Bulk density—Bulk density is needed to estimate the tortuosity function
and needed to make estimates of the chemical partitioning into liquid, vapor,
and adsorbed phases. It should be measured if possible in the field setting
by taking undisturbed field cores whose density is measured in the laboratory.
Porosity may be roughly calculated from bulk density.
Chemical Parameters—
Henry's constant—The Henry's constant KH may be set equal to the ratio of
saturated vapor density to chemical solubility each of which may be measured
by standard laboratory procedures (Freed, 1976). There is no known method for
directly measuring Henry's constant in the field, and at given temperature it
should be considered a benchmark property of the chemical irrespective of the
soil conditions. It is usually measured in the laboratory by setting it equal
to the ratio of the saturated vapor density and the water solubility (Spencer
and Cliath, 1970).
Diffusion coefficients in air and water—The pure chemical diffusion
coefficients in air and in water have been measured for a restricted number of~
organic chemicals (Jury et al., 1983a). However, for intermediate weight
molecular compounds, variation between chemicals is so slight that standard
values of pure diffusion coefficients may be assumed to apply for till chem-
icals. Jury et al. (1983a), recommended the values 0.43 and 4300 cm'/day for
liquid diffusion coefficients in water and vapor diffusion coefficients in air,
respectively.
Adsorption parameters—Considerable effort has been made in recent years
to simplify the adsorption characterization of chemicals by using distribution
coefficients KQ or even organic carbon coefficients Koc to characterize the
adsorption of a chemical. Information on these coefficients is generally
available for chemicals either by direct measurement or through structural
models or correlations with solubility. The more severe problem for field
scale transport seems to be non-equilibrium adsorption due to preferential flow
paths in the soil which bypass many of the adsorption sites which would be
characterized by measurement. Indirect adsorption coefficients under field
conditions could, in principle, be measured by comparing chemical leaching with
that of a water tracer such as chloride, but this has not been done except
B-13
-------
under laboratory conditions (Jury et al., 1983b). In the interim, all that may
be used to characterize the adsorption of a chemical is the standard adsorption
coefficient or, in cases where an adsorption isotherm has been characterized,
the Freundlich constants (see Appendix C).
Degradation rate—Degradation rate affects chemical volatilization in-
directly by decreasing the amount of chemical available for volatilization. In
most cases, degradation rates under field conditions have been measured by mass
balance estimates following intitial incorporation and therefore are likely to
be contaminated by volatilization losses (Jury et al., 1983a). Summaries of
effective degradation rate coefficients are available for a number of compounds
(Nash, 1980; Rao and Davidson, 1980), but these show large variation and merely
serve as a benchmark estimate of persistence.
Atmospheric Parameters--
Temperature--Soil temperature may be routinely measured by thermocouples
or thermistors. Furthermore, if only mean temperatures are to be used, the
average air temperature (max + min)/2 will give a reasonable estimate of soil
temperature near the surface. The spatial variability of mean temperature is
likely to be small and should not require extensive replication.
Boundary layer thickness—For chemicals which are not well mixed, the
stagnant air boundary layer can act to limit volatilization. However, the
boundary layer thickness which is required in several volatilization models
(Mayer et al., 1974; Jury et al., 1983a) is not directly measurable. It may be
inferred from atmospheric measurements of volatilization fluxes made by micro-
meteorological methods which measure the convection flux (Glotfelty, 1981)
provided that the soil surface concentration is known (see equation D-l of
Appendix D). Alternatively, as suggested by Jury et al. (1983a), it is assumed
that the boundary layer for water vapor is the same as for other compounds.
Then measurement of the evaporation rate above a wet surface and measurement of
the surface temperature and air humidity define the boundary layer in terms of
known parameters. This procedure leads to an estimate of a 0.5 cm boundary
layer thickness for an evaporation rate of 0.5 cm/day at T = 25°C and a rela- :.
tive humidity of 50 percent. This value could be used for a rough estimate
when no data are available.
B-14
-------
REFERENCES
Baver, L. D., W. H. Gardner, and W. R. Gardner. 1972. Soil Physics. 4th Ed.
John Wiley and Sons, Inc., New York.
Call, F. 1957. Soil fumigation. V. Diffusion of EDB through soils. J_. Sci.
Food Agr. 8:143.
Calvet, R. 1980. Adsorption-desorption phenomena. In: R. J. Hance (Ed.),
Interactions Between Herbicides and Soil. Academic Press, New York.
Cliath, M. M., and W. F. Spencer. 1971. Movement and persistence of dieldrin
and lindane in soil as influenced by placement and irrigation.. Soil Sci.
Soc. Amer. Proc. 35:791-795.
Farmer, W. J., M. S. Yang, J. Letey, and W. F. Spencer. 1980. Hexachloro-
benzene: Its vapor pressure and vapor phase diffusion in soil. Soil Sci.
Soc. Amer. J_. 44:676-680.
Freed, V. H. 1976. Literature Survey of Benchmark Pesticides. George Washington
University Medical Center, Washington, DC.
Glotfelty, D. D. 1981. Atmospheric Dispersion of Pesticides from Treated
Fields. Ph.D. Dissertation, Univ. of Maryland.
Goring, C. A. I. 1962. Theory and principles of soil fumigation. Adv. Pest '
Control Res. 5:47
Hamaker, J. W. and J. M. Thompson. 1972. Adsorption. In: C. A. I. Goring,
and J. W. Hamaker (Eds.), Organic Chemicals in the Soil Environment. Marcel
Dekker, Inc., New York.
Harper, L. A., A. W. White, R. R. Bruce, A. W. Thomas, and R. A. Leonard.
1976. Soil and microclimate effects on trifuluralin volatilization.
J_. Environ. Qua!. 5:236-242.
Hartley, G. S. 1969. Evaporation of pesticides. Adv. Chem. Series
86:115-134.
Jury, W. A., R. Grover, W. F. Spencer, and W. F. Farmer. 1980. Modeling
vapor losses of soil-incorporated triallate. Soil Sci. Soc. .Amer. J.
44:445-450.
B-15
-------
Jury, W. A., W. F. Spencer, and W. J. Farmer. 1983a. Use of models for assess-
ing relative volatility, mobility, and persistence of pesticides and other
trace organics in soil systems, pp. 1-43. In: J. Saxena (Ed.), Hazard
Assessment of Chemicals Vol. 2 Academic Press, New York.
Jury, W. A., H. El Abd, and T. M. Collins. 1983b. Field Scale Transport of
Nonadsorbing and Adsorbing Chemicals Applied to the Soil Surface. NWWA
Symposium, Las Vegas, NV. (In press)
Leistra, M. 1970. Distribution of 1,3-dichloropropene over the phases in
soil. J. Agr. Food Chem. 18:1124.
Letey, J., and W. J. Farmer. 1974. Movement of pesticides in soil. In:
W. D. Guenzi (Ed.), Pesticides in Soil and Water. Soil Sci. Soc. Amer.
Madison, WI.
Mayer, R. W., W. J. Fanner, and J. Letey. 1974. Models for predicting
volatilization of soil applied pesticides. Soil Sci. Soc. Amer. Proc.
38:563-568.
Millington, R. J., and J. M. Quirk. 1961. Permeability of porous solids.
Trans. Farday Soc. 57:1200-1207.
Nash, R. G. 1980. Dissipation rates of pesticides from soils. In: W. G.
Knisel (Ed.), CREAMS. Vol. 3. U.S. Department of Agriculture, Washington,
DC.
Nielsen, D. R., R. D. Jackson, J. W. Cary, and D. D. Evans. 1972. Soil Water.
Amer. Soc. Agron. Special Pub. Madison, WI.
Rao, P. S. C., and J. M. Davidson. 1980. Estimation of pesticide retention
and transformation parameters. In: M. R. Overcash and J. M. Davidson
(Ed). Environmental Impact of Nonpoint Source Pollution. Ann Arbor Sci.-
Publ., Ann Arbor, MI.
Shearer, R. C., J. Letey, W. J. Farmer, and A. Klute. 1973. Lindane diffusion
in soil. Soil Sci. Soc. Amer. Proc. 37:189-194.
Shouse, P., W. A. Jury, L. H. Stolzy, and S. Dasberg. 1982. Field measurement
and modeling of cowpea water use and yield under stressed and well-watered
growth conditions. Hilgardia 50:1-25.
Spencer, W. F., and M. M. Cliath. 1970. Desorption of lindane from soil.
Soil Sci. Soc. Amer. Proc. 34:574-578.
Spencer, W. F., and M. M. Cliath. 1973. Pesticide volatilization as related
to water loss from soil. J_. Environ. Qual. 2:284-289.
Spencer, W. F., W. J. Farmer, and M. M. Cliath. 1973. Pesticide volatiliza-
tion. Residue Rev. 49:1-47.
B-16
-------
Spencer, W. F., W. J. Farmer, and W. A. Jury. 1982. Review: Behavior of
organic chemicals at soil, air, water interfaces as related to predicting
the transport and volatilization of organic pollutants. Environ. Tpxicpl.
Chem. 1:17-26.
Tanner, C. B. 1968. Evaporation of water from plants and soil. l_n_: T. T.
Kozlowski (Ed.), Water Deficits and Plant Growth. Academic Press,
New York.
B-17
-------
APPENDIX C
ADSORPTION OF ORGANIC CHEMICALS ONTO SOIL
by
W. A. Jury
PROCESS DESCRIPTION
Definition
Adsorption refers to the bonding of a solute to adsorption sites on the
soil solids, either soil mineral surfaces or organic matter surfaces. The
effect of this bonding is to temporarily immobilize the molecule from trans-
port in either the solution or vapor phase. In most quantitative descriptions
of soil chemical transport processes in soil, the adsorbed molecules are
represented as a separate phase, i.e., distinct from vapor or solution phases
(see Appendix A).
Practical Importance of Adsorption
The adsorption process is an extremely important aspect of chemical move-
ment in soil because it acts to decrease chemical mobility in solution or vapor
phases. For this reason, chemical adsorptiye properties must be taken into
account when designing fertilizer or pesticide application procedures, hazard-'*
ous waste site management strategies, ground-water pollution potential esti- "
mates, etc. Since each chemical will, in principle, have a different adsorp-
tion affinity for solid surfaces in soil, it will move through soil in a unique
manner in response to driving forces such as gravity. For this reason, quan-
titative description of the adsorption process is essential if descriptive
models are to be developed for the transport of chemicals through the soil.
Published literature on adsorption of organic chemicals onto soil is quite
voluminous and has been the subject of a number of excellent reviews written in
recent years (Bailey and White, 1970; Green, 1974; Weed and Webber, 1974;
Hamaker and Thompson, 1972; Rao and Davidson, 1980; Karickhoff, 1981; Calvet,
1980; Mingelgrin and Gerstl, 1983). In spite of this widespread information,
however, much of the knowledge of the adsorption mechanisms which bind organic
chemicals in soil is empirical.
The uncertainty in the literature on adsorption to soils stems in large
part from the diversity and complexity of the adsorption surfaces. Most or-
ganic chemicals are attracted to both clay mineral surfaces and organic matter
surfaces. For a given chemical, the nature of the bonding mechanisms to the
C-l
-------
accessible adsorption sites depends on a variety of factors which will be
discussed below.
DESCRIPTION OF BONDING MECHANISMS
Table C-l summarizes the principal molecular interactions involved in ad-
sorption of organic chemicals to soil solid Interfaces. Only a short descrip-
tion of each interaction will be given here. More details on the possible
bonds between organic molecules and organic and mineral adsorbents can be found
in Mortland (1970), Calvet and Chassin (1973), Theng (1974), and Hayes (1970).
TABLE C-l. INTERMOLECULAR INTERACTIONS INVOLVED IN ADSORPTION
===============================================================================
High Energy Bonds Low Energy Bonds
Ionic bonds Charge-dipole and dipole-dipole bonds
Ligand bonds Hydrogen bonds
Charge transfer bonds
van der Waal's-London bonds
Entropy generation
Magnetic bonds
Ionic Bonds
Adsorption by ionic bonds is due to ion exchange. Ionic bonds occur
between organic anions or cations and positive or negative electric charges
located at the adsorbent surfaces. Some organic molecules, such as certain
herbicide molecules (e.g., paraquat or diquat), are always present as cations
while others can be ionized with the extent of ionization depending on the
acidity or pH of the medium. These latter compounds are either weak acids
(e.g., phenoxyacetic acids) or weak bases (e.g., s-triazines).
Ligand Exchange
Ligand exchange interactions are possible in soil with the water molecule
usually acting as the exchange ligand. Infrared spectroscopy investigations
have verified this mechanism on montmorillonite clays (Mortland and Meggitt,
1966). Ligand exchange has also been postulated for the binding onto the
residual transition metals of humic acids (Hamaker and Thompson, 1972).
Pi pole Bonds
Charge-dipole bonds and dipole-dipole bonds are involved between polar
organic molecules and electrically charged or polar adsorbing surfaces. This
type of bonding has not been studied extensively for organic molecules in soil
(Calvet, 1980; Green, 1974). Thus, its relative importance in the adsorption
process is unknown.
C-2
-------
Hydrogen Bonds
Hydrogen bonds are mainly associated with NH or OH groups and N and 0
atoms. Thus, all organic molecules are able to establish hydrogen bonds. The
hydrogen bond is important in the "water bridge" between an exchanger cation on
a clay and a polar organic molecule (Green, 1974). This water bridge has been
illustrated by Bailey et al. (1968), and has been demonstrated to occur with
clays for a number of different organic compounds containing nitrogen and/or
oxygen (Mortland, 1970). At the present time, there is only limited evidence
for the importance of hydrogen bonds in bonding of organic molecules in soil
(Calvet, 1980).
Charge Transfer Bonds
Charge transfer bonds are due to the transfer of electrons across the sur-
face of the adsorbent or organic molecule. This mechanism has been postulated
for organic cations adsorbed on clays (Haque et al., 1970) and for the adsorp-
tion of s-triazines onto organic matter (Hayes, 1970).
van der Waal's Bonds
van der Waal's-London bonds are due to instantaneous dipoles established
by fluctuations in the electron distributions as the electrons circulate in
their orbitals. This interaction is weak and decreases as the sixth power of
the inter-molecular distance. However, it can be a significant bond under
certain circumstances. It has not been specifically studied in soil, but
probably it is present in most organic chemical interactions with soil adsorp-
tion sites (Calvet, 1980).
Entropy Generation
Entropy generation refers to the displacement of water molecules when
organic molecules interact with the surface of an adsorbent, which increases
the stability of the system (Hamaker and Thompson, 1972). It may be a contrib-
uting mechanism in the adsorption of nonpolar organic chemicals to hydrophobia
surfaces such as soil humin.
Magnetic Interactions
Magnetic interactions occur from ring currents and conjugated double bonds
and therefore may be a mechanism of adsorption for the larger organic molecules
(Hamaker and Thompson, 1972).
INFLUENCE OF SOIL, CHEMICAL, AND ENVIRONMENTAL PROPERTIES ON ADSORPTION
The complexity and diversity of the bonding mechanisms discussed in the
previous section has obscured much of the correlation between adsorption and
soil chemical or environmetal properties. Table C-2 summarizes the properties
for which quantitative information is available. These will be discussed in
greater detail below.
C-3
-------
TABLE C-2. SOIL, CHEMICAL, AND ENVIRONMENTAL PROPERTIES INFLUENCING
ADSORPTION OF CHEMICALS ONTO SOIL
Soil Properties Chemical Properties Environmental Properties
Clay composition and Electronic structure Soil temperature
exchange capacity
Organic matter content Water solubility
Soil water content Solution composition
Soil bulk density Solution concentration
pH
Properties Affecting Adsorption
Soil Properties--
Clay composition and exchange capacity—Because clay mineral surfaces have
a net negative charge, they will influence the adsorption reaction for virtu-
ally all chemicals. Their influence is strongest for compounds such as paraquat
and diquat which carry a permanent positive charge, but they may also have a
strong influence on chemicals which ionize in water. Thus, inorganic cations
such as calcium, magnesium, and potassium, as well as positively charged trace
metals such as cadmium, etc., are strongly adsorbed in soils with a large
cation-exchange capacity. However, because of the large preponderance of or-
ganic matter surfaces and the non-polar nature of many organic chemicals, there
is usually no correlation observed between adsorption and percent clay (Green,
1974; Calvet, 1980) unless the compound is permanently charged. A complete
review of pesticide-clay-water interactions is given by Green (1974).
Organic matter content—Soil organic matter is an extremely complex and
poorly defined substance. Because of this poor characterization, no exact
relationships have been established between its structure and its adsorbing
properties. However, it has been demonstrated that the adsorption of organic
chemicals varies greatly according to the nature of organic matter, as demon-
strated for trifuluralin by Grover (1974) and for atrazine by Dunigan and
Macintosh (1971). Because of this dependence on organic matter type, it is
likely that the adsorption capacity of a surface layer of soil will vary for a
given total organic matter content depending on the state of decomposition of
its residues (Calvet, 1980).
All of these complications notwithstanding, there has been a positive
linear relationship observed between soil organic matter content and adsorp-
tion of organic chemicals (Hamaker and Thompson, 1972). The correlation
coefficient is likely to be highest when the organic matter content is large,
but the correlation coefficient may still be significant even in soils with an
organic carbon content as low as 0.1 percent (Lyman, 1982). Because of the
C-4
-------
positive correlation between soil organic matter and adsorption, workers have
proposed defining an adsorption coefficient per unit of soil organic matter in
order to reduce the variability between soils. This amounts to defining an
organic carbon partition coefficient Koc according to equation (C-l)
KOC • KD/foc (C-l)
where foc is organic carbon fraction, and where KQ is the distribution coef-
ficient discussed in Appendix D. As shown by Hamaker and Thompson (1972), the
coefficient of variability of the organic carbon partition coefficient particu-
larly for nonpolar compounds is considerably less than the variation of the
distribution coefficient for a given chemical adsorbing on different soils.
However, as pointed out by Calvet (1980), the remaining variability in the
organic carbon partition coefficient is large enough to make it an unsuitable
constant for precise analysis. However, it has still gained favor as a bench-
mark property for soil organic chemicals (Kenaga and Goring, 1980; Rao and
Davidson, 1980; Jury et al., 1983). A good discussion of the limitations of
partition coefficients may be found in Mingelgrin and Gerstl (1983).
Soil water content—The soil water content can influence adsorption in two
ways: it can modify the solution pathway leading to the adsorption sites and
thus increase or decrease the accessibility of the surface to the solute
(Grover and Hance, 1970), and it may also affect the physical-chemical proper-
ties of the adsorbent by increasing or decreasing the hydrolysis of clay lat-
tices (Frenkel and Suarez, 1977). However, as shown by Spencer and Cliath
(1973) and Ehlers et al. (1973), the influence of soil water content on adsorp-
tion is slight until the soil is extremely dry. When dry, the preferential
coverage of water molecules on the soil's adsorbing surfaces is removed and
solute adsorption increases dramatically. There is also some evidence that
this effect is reversible when the soil rewets (Spencer and Cliath, 1973;
Harper et al., 1976).
Bulk density--The influence of increasing soil bulk density on adsorption..
is to increase the density of adsorption sites per unit volume which will
directly increase adsorption capacity. However, it is to be expected that the
correlation between adsorption and bulk density for a group of soils will be
small because clay and organic soils tend to be found at a lower bulk density
than coarser-textured soils which are low in organic matter. Thus, the effect
of increasing bulk density on adsorption refers to compressing a given soil
volume.
Chemical Properties—
Electronic structure—The electronic structure of the solute molecule has
been found to be extremely important in adsorption. For example, if the mol-
ecule bears a permanent charge, such as paraquat or diquat, a strong binding
mechanism of ion exchange to negatively charged mineral surfaces will be in-
volved. On the other hand, neutral non-polar molecules such as ureas may bind
to surfaces only by the van der Waal's interaction or perhaps through one of
the other minor mechanisms discussed above. Although the bonding mechanisms
discussed above and shown in Table C-l are quite complex, several integrated
properties of the solute have been used to characterize the electronic struc-
ture and in turn have been correlated against adsorption. Among these are the
C-5
-------
Hammett sigma function and the Taft function (Hamaker and Thompson, 1972) which
describe the free energy changes accompanying bonding. Brlggs (1969) found a
linear relationship between the organic carbon partition coefficient Koc and
the Hammett sigma function for 22 substituted phenyl-ureas.
Lambert (1967) established a relationship between the parachor (related
to the molecular volume) and the equilibrium adsorption coefficient for several
phenyl-ureas and dialkylamines. This relationship was later modified by Hance
(1969) to include hydrogen bonding.
Water solubility--Since solute molecules compete with water for the ad-
sorption sites, many workers have sought a relationship between adsorption and
solubility. Kenaga and Goring (1980) studied experimental data from 170 chem-
icals, mostly pesticides, and obtained regression coefficients between water
solubility and organic carbon partition coefficients. These obtained correla-
tion coefficients were all significant below the 1 percent level. Although, as
pointed out by Calvet (1980), a general relationship between water solubility
and adsorption has not been established for all chemicals, the correlation
appears to be strong enough within a class of compounds to provide useful
regression equations between solubility and adsorption.
Solution composition and concentration—The complexity of the soil solu-
tion is such that very few quantitative relationships have been established
between adsorption and macroscopic properties characterizing the soil solution
in terms of ionic strength, electrical conductivity, pH, etc. For inorganic
cations such as calcium, magnesium, and potassium, quantitative relationships
have been established between cation activity which is a function of soil or
ionic strength and exchange adsorption. It is to be expected that other pos-
itively charged compounds will also experience competitive adsorption effects
which depend on the composition and concentration of the soil solution. How-
ever, when nonionic chemicals are considered, the influence of bulk soil solu-
tion parameters is generally neglected and the distribution coefficient or
organic carbon partition coefficient is considered to be independent of these
effects. Organic solvents can also have a large influence on adsorption and
are frequently used to apply chemicals in a highly concentrated state which can
greatly exceed their solubility in pure water.
£H_—The pH of the soil system can have a marked effect on adsorption of
compounds in soil, particularly compounds which are weakly acidic or weakly
basic. Weak acids are in the free acid form when the pH of the soil solution
is low, and they are much more highly adsorbed in this form than when present
as an anion. Weak bases are converted to cationic forms at low pH which are
also more highly adsorbed than as free bases. Polar materials may not actually
form cationic or anionic forms but are capable of forming hydrogen bonds which
may depend somewhat on pH. On the other hand, highly neutral molecules such
as chlorinated hydrocarbons would be expected to show virtually no dependence
on pH (Hamaker and Thompson, 1972).
Despite these dependencies, however, the correlation of pH with adsorption
in soil has proved to be much lower than the correlation between organic matter
and adsorption or between cation exchange capacity and adsorption (Farmer and
Aochi, 1974). As a general rule, chemicals which dissociate in solution are
C-6
-------
likely to be strongly affected by pH whereas the rest will appear to be rel-
atively independent of pH (Bailey and White, 1970; Green, 1974).
Environmental Properties--
Soil temperature--Although some compounds may increase their adsorption
as temperature increases (Calvet, 1980), the majority decrease (Lyrnan, 1982).
Temperature variations have two distinct effects on the adsorption process:
solute-surface interaction effects and water-solute interaction effects. The
balance between these two effects will determine the direction of dependence of
the temperature. As a general rule, the effect of temperature on adsorption
equilibrium is a direct indication of the strength of adsorption so that the
weaker the bond the less the influence of temperature (Hamaker and Thompson,
1972). Summaries for the temperature effect of adsorption for a number of
chemicals are given in the review article by Calvet (1980).
PRACTICAL LIMITATIONS TO APPLYING ADSORPTION MODELS TO FIELD STUDIES
Because of the extreme complexity of the soil solution and the hetero-
geneous nature of the adsorption sites in soil, it seems unlikely that general
relationsips will be developed in the immediate future describing in detail the
dependence of adsorption on the variety of physical and chemical properties
discussed above. For this reason, engineering approximations such as the
distribution coefficient or the partition coefficient have gained acceptance as
approximate methods for describing the adsorption capacity of a given chemical.
There are certain complications which should be borne in mind for using these
coefficients in chemical modeling, the most severe of which are discussed
below.
Adsorption-Desorption Hysteresis
Virtually all organic chemicals studied and many inorganic compounds show
a non-singular adsorption-desorption isotherm and exhibit greater resistance to
desorbing than to sorbing. van Genuchten et al. (1974), attempted to take this
into account in modeling movement of pulses of chemical through soil columns. ;•
The major effect of this increased desorption resistance is to leave higher
residual concentrations of chemicals in soil during leaching than would be
predicted from the adsorption properties alone. However, the extent of this
effect is still unknown. Resolution of the influence of hysteresis on trans-
port is difficult because its effect on the shape of a column breakthrough
curve is similar to that caused by rate-limited adsorption-desorption without
hysteresis. Some different adsorption-desorption models are discussed in
Rao and Davidson (1980).
Influence of Soil Structure
The major method for characterizing soil adsorption is a batch equilibrium
determination of the adsorption isotherm. In this method, all of the organic
and mineral surfaces are exposed to the solute by shaking, and the change in
solution concentration is measured until equilibrium is reached. Under natural
conditions, however, soil is likely to be aggregated or cracked with certain of
the surfaces inaccessible to the solute except by the time-consuming diffusive
transport. For this reason, the amount of adsorption occurring under natural
C-7
-------
soil-aggregated conditions with flowing solution Is likely to be less than that
predicted from batch determination, and the adsorption process itself is likely
to be rate-limited (El Abd, 1984). Since the rate coefficient used to describe
the approach to equilibrium under rate-limited conditions is not at the present
time predictable from bulk soil properties, no a priori modeling predictions
can be made. However, recent attempts have been made to describe rate-limited
diffusive transport by using a spherical aggregate structural model (Rao et al.,
1979).
Influence of Spatial Variability
On the field scale, i.e., agricultural fields or large waste disposal
areas, order of magnitude lateral variation has been observed in the primary
soil properties influencing adsorption (Gajem et al., 1981; Nielsen et al.,
1973; El Abd, 1984). For this reason, it is expected that the distribution
coefficient KD will have a value which varies from point to point in a large
field either due to variations in soil organic matter or to a complex inter-
action between soil structure and transport. It thus should be concluded that
prediction of the field wide average and extreme behavior of a chemical moving
through natural soil is at this time an unresolved research problem (Jury,
1984). However, at any particular location in the field, measurement of the
distribution coefficient may be effective in modeling movement of the chemical
in the immediate vicinity of the measurement.
Nonequilibrium Adsorption
When chemicals are flowing through soil, the residence time of an adsorb-
ing solute may be too short to reach equilibrium with the adsorbing surface.
Under such conditions, adsorption is less than predicted from equilibrium
parameters such as KQ or KQC. Rate-limited models (see Appendix D) have been
proposed to account for this effect, but they contain parameters which cannot
be independently measured.
Linear Adsorption Models
As mentioned above, the complications of adsorption are frequently neg-
lected in favor of using the simple distribution or partition coefficient Kg
describing the relationship between adsorbed and solute concentrations. Be-
cause of the enormously large number of organic chemicals on the market or
under development, this simple index of adsorption has gained favor as a bench-
mark property for characterizing potential mobility of chemicals under leaching
(McCall et al., 1980). Large compendiums of KD or Koc values have been pub-
lished in the literature (Kenaga, 1980; Rao and Davidson, 1980; Hamaker and
Thompson, 1972) including in certain cases the coefficient of variation found
for the adsorption coefficient among different soils. As mentioned above, the
KQQ coefficient has been observed to have less variability among different
soils than the KQ in most cases.
As a further simplification in representing adsorption, it has been pro-
posed to express the adsorption of all chemicals in terms of an octanol-water
partition coefficient Kow. Highly significant correlations have been found
between Kow and Koc (Rao and Davidson, 1980; Karickhoff and Brown, 1978).
C-8
-------
However, because this relationship Is between the logarithms of the coeffi-
cients, prediction of adsorption potential from a Kow measurement may involve
order of magnitude error (Kenaga and Goring, 1980).
As shown in the chapter on chemical transport, use of a linear partition-
ing relationship between solution and adsorbed concentrations allows consider-
able simplification in mathematical modeling. For this reason, adsorption
isotherms are frequently linearized over the range expected to be found in a
given simulation. As demonstrated by Hamaker and Thompson (1972), no single KQ
value can describe adsorbed-solute partitioning from trace concentrations all
the way to very high concentrations. Hence, if a single KD value is used at
both low and high concentrations, models predicting transport will give erro-
neous descriptions (Rao and Davidson, 1980). However, Karickhoff et al.
(1979), showed that the linear representation was a good approximation to the
isotherm for trace concentrations of a chemical which are frequently at the
levels found in soil.
Briggs (1969) and others have developed a considerable amount of data
relating adsorption to chemical structure which involves correlations between
water solubility and adsorption and which involves molecular properties such as
melting point, etc. Although these regressions have been found to be useful
for a limited number of chemicals, they do not have a general validity and
should be used with caution.
FIELD MEASUREMENT OF ADSORPTION PARAMETERS
Virtually all adsorption studies have been conducted under laboratory
conditions. The only information obtained to date on field adsorption studies
consists of observing transport of an adsorbed chemical. As shown by El Abd
(1984), simultaneous observation of the movement of a mobile tracer such as
chloride or bromide and an adsorbed compound can be used to calculate an
effective adsorption coefficient for the compound. For this simple procedure,
all that is required is a measurement of the chemical concentration which may
be obtained from soil core measurements. In principal, soil core measurements*
may also be made of chemical concentrations for the purpose of calculating
adsorption by the method of El Abd (1984). However, adsorbed compounds are
frequently difficult to extract through solution samples particularly if
strongly adsorbed. Therefore, the soil coring method is frequently preferable.
The more complex adsorption functional relationships discussed in Appendix D
have never been characterized under field conditions. These methods require
precise laboratory procedures for preparation and extraction in order to give
the quantitative information required to determine the functional parameters.
In the future, it is likely that simple partitioning relationships will be used
under field conditions.
C-9
-------
REFERENCES
Bailey, G. W., and J. L. White, and T. Roghberg. 1968. Adsorption of organic
herbicides by montmorillonite: Role of pH and chemical character of the
adsorbate. Soil Sci. Soc. Amer. Proc. 32:222-234.
Bailey, 6. W., and J. L. White. 1970. Factors influencing the adsorption,
desorption and movement of pesticides in soil. Residue Rev. 32:29-92.
Briggs, 6. G. 1969. Molecular structure of herbicides and their sorption by
soils. Nature 223:1288-1289.
Calvet, R., and P. Chassin. 1973. Complexes organiques des argile.
Mechanismes de formation. Bui. Groupe Fr. Argile XXI pp. 87-113.
Calvet, R. 1980. Adsorption-desorption phenomena. In: R. J. Hance (Ed.),
Interactions Between Herbicides and Soil. Academic Press, New York.
Dum'gan, E. P. and T. H. Macintosh. 1971. Atrazine-soil organic matter inter-
action. Weed Sci. 19:279-282.
Ehlers, W., W. J. Farmer, W. F. Spencer, and J. Letey. 1973. Lindane diffusion
in soils. II. Water content, bulk density and temperature effects. Soil
Sci. Soc. Amer. Proc. 33:504-509.
El Abd, H. 1984. The Spatial Variability of the Pesticide Adsorption Co-
efficient. Ph.D. Thesis. University of California, Riverside.
Farmer, W. J., and Y. Aochi. 1972. Picloram sorption by soils. Soil Sci. Soc.
Amer. Proc. 38:418-423.
Frenkel, H., and D. Suarez. 1977. Hydrolysis and decomposition of calcium
montmorillonite. Soil Sci. Soc. Amer. Proc. 41:887-891.
Gajem, Y. M., A. W. Warrick, and D. E. Myers. 1981. Spatial dependence of
physical properties of typic torrifluven soil. Soil Sci. Soc. Amer. J.
45:709-715.
Green, R. E. 1974. Pesticide-Clay-Water Interactions. jn_: W. D. Guenzi
(Ed.), Pesticides in Soil and Water. Soil Sci. Soc. Amer., Madison, WI.
Grover, R. 1974. Adsorption and desorption of trifluralin, triallate, and
diallate by various adsorbents. Weed Sci. 22:405-408.
C-10
-------
Grover, R., and R. J. Hance. 1970. Effect of ratio of soil to water on
adsorption of linuron and atrazine. Soil Sci. 109:136-138.
Hamaker, J. W., and J. M. Thompson. 1972. Adsorption. In: C. A. I. Goring
and J. W. Hamaker (Eds.), Organic Chemicals in the Soil Environment.
Marcel Dekker, Inc., New York.
Hance, R. J. 1969. An empirical relation between chemical structure and the
sorption of some herbicides by soil. J_. Agr. Food Chem. 17:530-631.
Haque, R., S. Lilley, and W. R. Coshow. 1970. Mechanism of adsorption of
diquat and paraquat on montmorillonite surfaces. J_. Colloid Int. Sci.
3:185-188.
Harper, L. A., A. W. White, R. R. Bruce, A. W. Thomas, and R. A. Leonard.
1976. Soil and microclimate effects on trifuluralin volatilization.
J_. Environ. Qua!. 5:236-242.
Hayes, M. H. B. 1970. Adsorption of triazine herbicides on soil organic
matter. Residue Rev. 32:131-168.
Jury, W. A., W. F. Spencer and W. J. Farmer. 1983a. Use of models for assess-
ing relative volatility, mobility and persistence of pesticides and other
trace organics in soil systems. In: J. Saxena (Ed.), Hazard Assessment of
Chemicals. Vol. 2, Academic Press, New York.
Jury, W. A. 1984. Field scale water and solute transport through unsaturated
soil. Proc. Int. Conf. Soil Salinity Under Irrigation. Bet-Oagan, Israel.
March 1984.
Karickhoff, S. W. 1981. Semi-empirical estimation of sorption of hydrophobic
pollutants on natural sediments. Chemisphere 10:833-846.
Karickhoff, S. W., and D. S. Brown. 1978. Paraquat sorption as a function of-
particle size in natural sediments. J^. Environ. Qual. 7:246-252.
Karickhoff, S. W., D. S. Brown, and T. A. Scott. 1979. Sorption of hydro-
phobic pollutants on natural sediments and soils. Water Res. 13:241-248.
Kenaga, E. E. 1980. Predicted bioconcentration factors and soil sorption
coefficients of pesticides and other chemicals. Etox. and Environ.
Safety 4:26-38.
Kenaga, E. E., and C. A. I. Goring. 1980. Relationship between water solubility,
soil sorption, octanol-water partitioning, and bioconcentration of chemicals
in biota. Proceedings of third ASTM symposium on aquatic toxicology. ASTM
Special Technical Publication 707:78-115.
Lambert, S. M. 1967. Functional relationships between sorption in soil and
chemical structures. J_. Agr. Food Chem. 15:572-576.
C-ll
-------
Lyman, W. J. 1982. Adsorption coefficients in soil and sediments. In;
Lyman, W. 0. and others (Eds.), Handbook of chemical property estimations
methods. McGraw-Hill, New York.
McCall, P. J., R. L. Swann, E. A. Laskowski, S. M. Unger, S. A. Vrona, and H.
J. Dishburger. 1980. Estimation of chemical mobility in soil from HPLC
retention times. Bull. Env. Contam. Toxicol. 24:190-195.
Mingelgrin, U., and Z. Gerstl. 1983. Reevaluation of partitioning as a
mechanism of nonionic chemical adsorption in soil. J^. Environ. Qua!.
12:1-11.
Mortland, M. M. 1970. Clay organic complexes and interactions. Adv. Agron.
22:75-114.
Mortland, M. M., and W. F. Meggitt. 1966. Interaction of EPTC with montmor-
illonite. »J. Agr. Food Chem. 14:126-129.
Nielsen, D. R., J. W. Biggar, and K. T. Erh. 1973. Spatial variability of
field measure soil water properties. Hilgardia 42:215-259.
Rao, P. S. C., J. M. Davidson, R. E. Jessup, and H. M. Selim. 1979.
Evaluation of conceptual models for describing nonequilibrium adsorption-
desorption of pesticides. Soil Sci. Soc. Amer. J_. 43:22-28.
Rao, P. S. C., and J. M. Davidson. 1980. Estimation of pesticide retention
and transformation parameters. In: M. R. Overcash and J. M. Davidson
(Eds.). Environmental Impact of Nonpoint Source Pollution. Ann Arbor
Sci. Publ., Ann Arbor, MI.
Spencer, W. F., and M. M. Cliath. 1973. Desorption of lindane from soil
as related to vapor density. Soil Sci. Soc. Amer. Proc. 34:574-578.
Theng, B. K. 6. 1974. The Chemistry of Clay-Organic Reactions. Hilger,
London.
van Genuchten, M. Th., J. M. Davidson, and P. J. Wierenga. 1974. An
evaluation of kinetic and equilibrium equations for the prediction of
pesticide movement through porous media. SoiJ_ Sci. Soc. Amer. Proc.
38:29-34.
Weed, S. B., and B. Webber. 1974. Pesticide-organic matter interactions.
In: W. D. Guenzi (Ed.), Pesticides in Soil and Water. Soil Sci. Soc
., Madison, MI.
C-12
-------
APPENDIX D
MATHEMATICAL DERIVATION OF CHEMICAL TRANSPORT EQUATIONS
by
W. A. Jury
CONTINUITY EQUATIONS
Conservation Law
The differential equations describing water, heat, or chemical transport
through porous media are macroscopic and make no attempt to describe three-
dimensional flow at the scale of a soil pore or soil particle. Equations are
written and averaged over a volume element large enough to produce a meaningful
average value which is independent of the properties of the averaging volume
for such quantities as volumetric water content, bulk density, etc. This
scale, often called the Darcy scale, is itself a function of the degree of
heterogeneity of the porous medium (Bear, 1972). As a simplification, the
equations derived below will be one-dimensional with the direction assumed to
represent the vertical coordinate z. Although in principle, balance equations
can be written for three dimensions, in practice very little use has been made
of multi-dimensional transport equations except under very restrictive assump-
tions. The one-dimensional equations derived here will be appropriate for
simulation of many practical transport processes such as volatilization and
vertical chemical transport by leaching.
The mass balance of water, assumed to be flowing in the z direction, is
achieved by looking at a small cubic soil volume element of thickness AZ and
unit cross-sectional area AXAy = 1 for which the conceptual balance equation
given in equation (D-l) is written.
[rate of water flow into V] - [rate of water flow out of V]
= [rate of increase of water stored in V]
+ [rate of water uptake in V by plant roots] (D-l)
where V is the soil volume.
If we represent the volumetric water flux by Jw (cnr/cnr/day) and the
water storage per unit volume by the volumetric water content 0 (cnrYcirr) then
we may rewrite equation (D-l) as
D-l
-------
AO
Jw(z)AxAy - Jw (z+Az)AXAy = -- AxAyAz + SwAxAyAz (D-2)
At
where Sw (day"1) is the rate of water uptake per unit soil volume.
If we divide equation (D-2) through by the volume element AXAyAZ and take
the limit of small time and small volume, we obtain the water conservation
equation (D-3)
--- + -- Syy = 0 (D-3)
Chemical Conservation Law
In a similar manner, we may write a conceptual equation for any chemical
species of interest as
[rate of flow of chemical into V] - [rate of flow of chemical out of V]
= [rate of increase of chemical stored in V]
+ [rate of disappearance of chemical from V by reactions] (D-4)
If we represent the chemical transport flux as Js (g/cm2/day) and the mass
of chemical stored per soil volume (i.e., the concentration) as Cj (g/cm3), we
may rewrite equation (D-4) as
Js(z)AxAy - Js(z+Az)AxAy = — AxAyAz + SsAxAyAz (D-5)
At
where S. (g/cm3/day) is the term representing the rate of disappearance of chem-
ical per soil volume by chemical or biological reactions.
If we again divide equation (D-5) through by the volume element
and take the limit of small volume and small time, we obtain the chemical
species balance equation (D-6)
a*
(D-6)
FLUX EQUATIONS
Water Flux Jy/
The one-dimensional flux of water through unsaturated soil is given by the
Buckingham-Darcy flux law (Hillel, 1971)
D-2
-------
jw = -K(h) [-- + 1] (D-7)
5z
where h(cm) is matric potential head and K(h) (cm/day) is the unsaturated
hydraulic conductivity; this is a very non-linear function of the matric
potential head and may decrease up to seven orders of magnitude between very
moist (h * 0) and very dry (h * - •) soil (Hillel, 1971).
Chemical Flux Jq
Diffusion Fluxes—
Vapor diffusi
by the extended form of Fick
acc _cnT1 a_CQ
az
Vapor diffusion--The vapor diffusion flux in soil Jg (g/cm^/day) is given
ick's law of diffusion (Jury et al., 1983)
(a) DAIRG -5 = - DSOILG
when to Q is the gaseous diffusion coefficient of the chemical species in free
air, CG (g/cm3 soil air) is the mass of chemical vapor per volume of soil air,
a is volumetric air content (cnr/cnr), and TIQ is the tortuosity factor. Thi
s tortuosity factor accounts for the decreased cross-sectional area and in-
creased path length induced by a porous medium, which is a function of the
volumetric air space available for transport (Nielsen et al . , 1972).
There have been numerous structural models introduced to represent the
tortuosity factor n in soil. The formulation which has proved to be the most
useful over a large range of air content is the Millington and Quirk formula-
tion
nG (a) = a10/3/02 (D-9)
(Millington and Quirk, 1961) where 0 is soil porosity. This model has proven
to be very successful at representing the tortuosity effect on diffusion coef-
ficients of pesticides in soil (Letey and Farmer, 1974; Shearer et al., 1973; -
Farmer et al., 1980; Jury et al., 1980) even at water contents very close to
saturation (Sal lam et al . , 1983).
Liquid Diffusion—In analogy to the above diffusion model, the liquid dif-
fusion transport flux Jj_ (g/cnr/day) is given by equation (D-10)
j, = . r, (0) DWATER, -- = - DSOIL (D.10)
L L dz 5z
where CL (g/cm3 solution) is mass of liquid per volume of soil solution,
(jWATtR^ (cmZ/day) is tne molecular diffusion coefficient of the chwnical
species in pure water, and TIL is the tortuosity factor accounting for the de-
creased cros-sectional area and increased path length for molecular diffusion
in solution due to the porous medium. Research studies have also validated the
use of the Millington and Quirk (1961) equation for this tortuosity factor
D-3
-------
(Shearer et al., 1973; Jury et al., 1980). The Millington and Quirk model for
liquid water is achieved by replacing the air content a in equation (D-9) by
the water content 9
\ (9) « 910/3/02 (D-ll)
Dispersion Flux-
Chemical transport equations must include a term to account for hydro-
dynamic dispersion since pore-scale mass fluxes are replaced by a volume-
averaged mass flux. This dispersion flux JHD for one-dimensional flow is
given by equation (D-12)
acL
JHD = - DHD — (°-i2)
az
where DHD is tne hydrodynamic dispersion coefficient. The dependence of DHD on
soil and environmental properties is not well understood at present, but numer-
ous studies have shown that DHD increases as water flux Jw increases. In most
cases, this dependence is assumed to be linear, as in equation (D-13)
DHD = dHD Jw (D-13)
where dno is called the dispersivity (Bear, 1972). In three-dimensional flow,
DHD is a tensor which is usually modeled as having longitudinal and transverse
components (Bresler, 1975).
Because of the similarity in form, the liquid diffusive and hydrodynamic
dispersive fluxes equations (0-10) and (D-12) are usually lumped together as a
combined flux JLHD
JLHD - JHD + JL " - (°HD + DSOILL ™ - - °SOIL E ™ (D-14>-
oz oz
where D_ ILE is the effective diffusion-dispersion coefficient. This formula-"
tion will be used in subsequent analysis.
Mass Flux-
When water flow is occurring, substantial amounts of dissolved chemical
can be transported with the moving soil solution. This transport flux JM
(cm2/day) is given by equation (D-15)
JM = CL Jw (D-15)
Solute Flux—
The combined effects of liquid and vapor diffusion and mass flow are
summarized in the solute flux equation (D-16)
Js = JG + JLHD + JM - - DSOILG --- - DSOILE --- + CL Jw (o-ie)
az az
D-4
-------
STORAGE EQUATIONS
Since the amount of water adsorbed on soil minerals Is constant until the
soil reaches an extreme state of dryness and because the amount of water vapor
stored Is negligible, the water content per soil volume is adequately repre-
sented by the liquid water content per soil volume. In the case of chemicals,
however, particularly pesticides and other trace organics, substantial amounts
of material may be stored in all three phases. Thus, the total concentration
per soil volume Cj in equation (D-6) is expanded to give the contributions in
each phase
CT = p0 Cs + 9CL + aCG (D-17)
where Cc (g/g soil) is mass of chemical adsorbed per mass of soil and p& is
soil bulk density. These units are conventional and correspond to the usual
method of measurement of the concentrations in each phase.
TRANSPORT EQUATIONS
Water Transport
When the water transport flux equation (D-7) is plugged into the continu-
ity equation (D-3), the result is equation (D-18).
89 5 dh
.. + Sw = - [K(h)(- + 1)] (D-18)
at az az
In its present form, this cannot be solved because there are two dependent
variables: the matric potential head h and the volumetric water content 9.
The equilibrium relationship h(Q) used to relate these two variables is called
the water characteristic function or the matric potential-water content rela-
tion. This function is obtained experimentally. A problem, currently unre- '.
solved in research studies, is that a significant amount of hysteresis appears
between the wetting curve and the drying curve of this relationship h(9). For
this reason, experiments and simulations using equation (D-18) have generally
been monotonic wetting or drying studies. The water uptake function Sw is also
determined experimentally. The experimental determination is usual only for
fully developed crop root systems operating under a characteristic water regime,
and, as a result this function does not change appreciably with time. For
studies when no plant roots are present, the water flow equation may be written
as
ah a ah
C(h) [K(h)(- +1)] (D-19)
at az az
where
C(h) = a9/ah (D-20)
D-5
-------
is the water capacity function which is the slope of the water contentmatric
potential function for a monotonic process.
Equation (D-19) may also be recast with the volumetric water content as
the dependent variable if this is more convenient (Kirkham and Powers, 1972).
Solute Transport
When the solute transport flux equation (D-16) and storage equation (D-17)
are plugged into the continuity equation (D-6), the result is equation (D-21)
3 5 cmi aCL
-- (pb Cs + OCL + aCG) + Ss - - [DSOILE Jw CL] (D-21)
5t az az
A number of simplifications are needed in equation (D-21) before solution
is possible, since there are three dependent variables CQ, CL, Cs.
The relationship which is usually assumed to apply between the liquid
concentration and the gaseous concentration is given by Henry's Law
cG - KH CL (0-22)
where KH is Henry's constant. The applicability of Henry's Law to soil systems
has been confirmed for a variety of chemicals and circumstances (Call, 1957;
Goring, 1962; Leistra, 1970; Spencer and Cliath, 1973). Furthermore, Spencer
and Cliath (1973) showed that the relationship equation (D-22) persisted from
very trace concentration levels all the way to saturation levels of the organic
chemicals they studied. Thus, one may estimate Henry's constant KH as the
ratio of saturated vapor density to aqueous solubility (Spencer and Cliath,
1973).
The relationship which exists at equilibrium between adsorbed concentra-
tion and liquid concentration is called the adsorption isotherm.
Cs = f(CL) (D-23)
This relationship is obtained under equilibrium laboratory conditions. The
functional form which has most commonly been used to represent this relation-
ship for pesticides and other trace organics is the so-called Freundlich
relationship (Hamaker and Thompson, 1972).
Cs - KF(CL)N (D-24)
where Kp and N are constants. Karickhoff et al. (1979), and Karickhoff (1981)
have shown that for low aqueous concentrations of weakly polar compounds, a
linear isotherm
Cs = KDCL (D-25)
may be substituted for the non-linear relationship where KQ is called the
distribution coefficient. This representation is only an approximation
D-6
-------
which has been measured in the Cs - CL relation for a number of compounds
(van Genuchten et al . , 1974; O'Connor et al . , 1980).
Under natural conditions, equilibrium may not be completely reached
between the phases. For this reason, kinetic or rate-limited expressions,
particularly in chemical leaching studies, have been used to represent the
relation between Cs and Ci. Lindstrom et al . (1969), Oddson et al. (1970),
and van Genuchten et al . (1974), all used the rate-limited expression given
in equation (D-26) to represent the change in adsorbed concentration
dCs
— = r [f(CL) - Cs] (D-26)
at
where r is a rate coefficient. Equation (D-26) assumes that the change in
adsorbed concentration is proportional to the deviation from equilibrium. In
practice, no model has ever been able to predict the form of the rate coef-
ficient r for a given soil process, and it has to be measured from experimental
breakthrough data. Recently, however, Rao et al . (1980), have successfully
modeled the dependence of this factor r on certain geometric and flow proper-
ties of the porous medium.
PARTITIONING COEFFICIENTS
The large number of parameters in equations (D-21) through (D-26) have led
researchers to make simplifying assumptions in order to reduce the need for
calibration and repeated measurements. Jury et al . (1983), introduced parti-
tioning coefficients for each phase by assuming that Henry's Law, equation
(0-22), and the linear isotherm, equation (D-25), were valid and that equilib-
rium was achieved in a short enough period of time so that kinetic effects on
adsorption could be neglected. With these two assumptions, the storage equa-
tion (D-17) may be written as
CT = RSCS = RLCL = RGCG (0-27)-
where
RS = CT/CS = Pb + O/KD + aKH/KD (D-28)
RL = CT/CL = PbKD + o +aKH (0-29)
RG • CJ/CG = PI>KD/KH
are the adsorbed, liquid, and vapor partition coefficients, respectively.
The inverse of the R coefficients gives the fraction of total concentration
in each phase. The expressions in equations (D-27) through (D-30) may be used
to make a number of simplifying definitions in the above equations. For
example, the solute flux equation (D-16) may now be written in terms of total
concentration
8CT
js = _ OE — + VECT (0-31)
5z
D-7
-------
where
0E = (DSOILG KH + DSOILE)/RL (D-32)
is the effective diffusion-dispersion coefficient for both liquid and vapor
diffusion and
vE = JW/RL (D-33)
is the effective solute velocity which can be used to make estimates of resi-
dence time for a given chemical.
REACTION RATES
The generalized reaction term Ss in equation (D-21) is merely symbolic and
for a given chemical species may depend on a variety of factors. For multi-
electrolyte solutions, for example, the reaction term depends on chemical acti-
vity and thus involves all species in solution. Simple models of reaction such
as precipitation and dissolution have been achieved by combining chemical
equilibrium models with transport models (Jury, 1982). For pesticides and
other trace organics, reactions such as hydrolysis, photolysis, and microbial
degradation have been modeled (see chapter on these processes), but within
soil, all reactions are usually represented by an effective rate coefficient
v£. Although this rate coefficient depends on such factors as water content,
organic matter content, temperature, etc., it is generally represented as a
first-order constant rate coefficient for simplicity (Hamaker and Thompson,
1972; Rao et al . , 1980; Nash, 1980). The reason for this simplification is
that most measurements of degradation are made by observing residual concentra-
tions after a certain period of time and by fitting them to a model assuming
exponential depletion. With the partitioning coefficient formulation, the
first-order rate expression Ss is written as
(D-34J
where UE is the effective rate coefficient for all phases combined. With
these simplifications, equation (D-21) becomes
5 aCj
...... [DE ---- VECT] - uEcT (D-35)
dt az 3z
This equation is the so-called convection-dispersion equation, van
Genuchten has compiled a compendium of solutions to this equation for a variety
of surface boundary conditions appropriate to chemical leaching (van Genuchten
and Alves, 1982).
D-8
-------
BOUNDARY CONDITIONS FOR THE TRANSPORT EQUATIONS
WATER TRANSPORT
Normally when field problems are represented with the water flow equation
(D-19), the upper boundary condition is represented by specifying the water
evaporation rate. For bare soil conditions, when no crop is present, this is
either done explicitly by providing external evaporation data or implicitly
through a water evaporation model which shifts between potential and soil-
limited conditions. Two examples of such models are given by Ritchie (1972) or
Tanner and Jury (1976). A recent field study has demonstrated how to field-
calibrate completely a soil evaporation model for use in bare soil or partial
cover cropping conditions (Shouse et al., 1982).
Evapotranspiration Models
When a crop is present, both the upper boundary condition and the water
extraction patterns in the root zone must be obtained. This is usually done
through the aid of an evapotranspiration model. Unless the crop i«. undergoing
stress, water will be lost at a potential rate which is dictated by the external
meteorological conditions. In this case, the potential water loss., called the
potential evapotranspiration, may be estimated from a knowledge of the
appropriate external meteorological variables. The exact variables used to
make this estimation depend on the type of model. The most comprehensive model
currently in use for the prediction of evapotranspiration is the so-called
Penman evapotranspiration equation which requires the following for external
inputs: the solar radiation, air temperature, air relative humidity, and wind
speed as well as a knowledge of the crop reflectance coefficient or albedo.
From this information, an estimate is made of the net radiation and an energy
balance is conducted at the canopy surface from which the evapotranspiration is
determined as a latent heat deficit (Penman, 1948; Tanner, 1968). In recent
years, other models have been proposed which use less information than the
Penman equation. Among these are the Priestley and Taylor equation (Priestley
and Taylor, 1972) which uses net radiation and air temperature to predict
evapotranspiration, the solar radiation correlation which uses air temperature
and solar radiation to predict evapotranspiration (Jensen and Haiso, 1963), and
an advection modification of the Priestley and Taylor equation (Jury and Tanner,
1975). Effectiveness of these different models has been studied under field
conditions by Shouse et al. (1980).
Perhaps the widest known method for predicting evapotranspiration is the
correlation with the evaporation pan (Doorenbos and Pruitt, 1976). This method
is reasonably accurate when applied over large-scale areas the size of water-
sheds and for time resolutions of the order of months. It is not "likely to be
very accurate as a daily or weekly gauge of evapotranspiration from a small
field. For this reason, methods based on the solar radiation are usually
preferred when estimating evapotranspiration.
D-9
-------
Water Uptake Functions
As mentioned above, under growing conditions, the water extraction pat-
terns change dramatically for crops as the roots seek out water in new loca-
tions. Only under relatively stable conditions such as high frequency irriga-
tion of a fully-developed crop will the water extraction pattern become well-
defined as a function of depth. Under such conditions, it has been observed
that an exponential depth function adequately describes the extraction patterns
for a number of crops (Feddes et al., 1974). However, when chemical transport
to great depth is being considered, details of the water extraction patterns
within the root zone become relatively unimportant, and a simple uniform water
uptake function may be assumed without great loss of information.
For problems involving leaching processes at the column scale or in a
field-plot trial, a water flux rate representing the infiltration rate of water
is specified as the upper boundary condition. Or, in the case of ponded water
on the soil surface, a constant positive pressure equal to the ponding height
is specified. These conditions are used with equation (D-19). In a dynamic
simulation of wetting and drying processes, both water input and evaporation
rates must be included as part of the upper boundary condition. For this case,
the water flux is set equal to the net water input to the surface which is
equal to the difference between applied water and evaporated water.
The lower boundary condition for water flow through the unsaturated zone
is usually assumed to be the water table if a continuous transport model is
represented over the entire unsaturated zone. For simulations representing
transport through soil with deep water tables which are below the zone of
interest, the lower boundary condition is sometimes represented by assuming
gravity flow which means that the gradient of matric potential is set equal
to zero at some depth far below the surface.
CHEMICAL TRANSPORT
Unless a chemical is actually being added at the soil surface, the normal
upper boundary condition for a chemical is contingent on chemical volatili-
zation (Jury et al., 1983). Since transport to the atmosphere takes place
through the vapor phase, Jury et al. (1983), suggested that this boundary
condition be written as
DAIRr
Js (o,t) - [CQ (o,t) - CAIRG (d,t)] (0-36)
d
where the solute flux Js is given by equation (D-16), d is the (hypothetical)
thickness of the stagnant boundary layer above the soil surface, and CAIRG is
the chemical gas concentration in the well-stirred air above the boundary
layer.
In practice, this boundary condition is modified by defining an effective
transfer coefficient h = DA G/d when used under field conditions. Jury
et al. (1983), suggested that the hypothetical boundary layer thickness d could
be evaluated by assuming an analogy between water evaporation and chemical
D-10
-------
volatilization. With this analogy, the measured water evaporation rate could
be used to evaluate the boundary layer thickness.
For compounds which volatilize only slightly, the free air concentration
CAIRc may be set equal to zero. In analogy with the simplified version
of the transport equation above, one may use the linear equilibrium isotherm
and Henry's Law to construct the following form of equation (D-36).
aCj
- DE — (o.t) + VECT(o,t) = HECT(o,t) (D-37)
oz
where
HE = DAIRG/dCG (D-38)
is the effective transport coefficient. This approach was used in developing
the screening model of Jury et al. (1983).
The lower boundary condition for chemicals in soil is usually approxi-
mated by assuming no concentration of chemical at great depth. A finite
initial concentration incorporated over a layer of variable thickness can
be used to make realistic assessments of subsequent movement through soil
(van Genuchten and Alves, 1982; Jury et al., 1983).
METHODS OF SOLUTION OF THE TRANSPORT EQUATIONS
WATER FLOW EQUATION
Because of the non-linear relationship between hydraulic conductivity and
matric potential k(h) and between matrie potential and water content h(9), only
numerical solutions are available to the water transport equation (D-19).
There are currently two widely used numerical methods of solution to this
equation: the finite difference method and the finite element method. The
finite difference method appears to be more useful for one-dimensional problems
when moderate water fluxes Jw are involved; whereas the finite element method
has advantages for complicated multi-dimensional problems (Reeves and Duguid,
1975; van Genuchten, 1978).
CHEMICAL TRANSPORT EQUATION
The chemical transport equation (D-21) when simplified to the convection-
dispersion equation given in equation (D-35) may be solved by analytic methods
such as Laplace transform or fourier series when the soil and water transport
properties are uniform which occurs when the water flux Jw is zero or constant
and when the water content 0 is constant or nearly constant with depth. This
condition is commonly present in miscible displacement column breakthrough
studies which have been used to study a large number of chemical and soil
physical properties (Biggar and Nielsen, 1967). Analytic solutions have been
D-ll
-------
obtained to equation (D-35) for step function and pulse input boundary condi-
tions and for both zero and first-order chemical reactions occurring in the
soil (van Genuchten and Alves, 1982). The combined equations (D-35) and bound-
ary condition (D-37) have been proposed for use in a chemical environmental
screening model for large numbers of chemicals whose benchmark properties are
known. This model uses an analytic solution based on the Laplace transform
(Jury et al., 1983).
When the chemical transport equation is used in situations in which water
or soil transport properties are non-uniform or when transient water flow
occurs, a numerical method of solution must be used. As in the case of water
flow, this is either a finite difference or a finite element model. For multi-
dimensional simulations of chemical transport, the finite element model is
usually preferred.
ADSORPTION MODELS
EQUILIBRIUM RELATIONS
The relationship between the equilibrium concentration of chemical in
solution and the associated chemical adsorbed on the mineral or organic surface
is called the adsorption isotherm. Several mathematical relationships have
been used to describe this equilibrium relationship in soil water systems. In
cases where a maximum adsorbed concentration is reached, the relationship may
be fit to a Langmuir isotherm
MAX
Cs = KCS CL/(1 + KCL) (0-39)
where Cs is adsorbed concentration (g/g soil), and C[_ is solution concentation
(g/cm3 solution/g).
Although the Langmuir relationship has been occasionally used to represent
organic chemical adsorption in soil, the Freundlich isotherm, equation (D-40),
is much more commonly used to express the adsorption relation
Cs = Kf (CL) (D-40)
where Kf and N are constant.
Although the Langmuir relationship has a theoretical basis, it applies
only to a pure adsorber and adsorbent. Because organic molecules in solution
are competing with water molecules for the surface adsorption sites and because
the surface adsorbing sites are heterogeneous, the physical basis of the
Langmuir isotherm is obscured in a soil-water system. However, the adsorption
of organic cations by soil has been shown to fit a Langmuir isotherm (Weber
et al., 1965). The Freundlich relationship is empirical although it has been
related to a special form of the Langmuir relationship in certain cases.
D-12
-------
The Freundlich relationship is frequently able to describe adsorption both
at low and high chemical concentrations. At dilute concentrations, a linear
relationship is frequently used with N = 1
Cs = KDCL (D-41)
where KQ is the distribution coefficient. This relationship has considerable
advantages for mathematical modeling which are discussed in Appendix C.
Its applicability to soil-water systems has been shown by Karickhoff et al.
(1979). However, equation (D-41) should not be thought of as a general rela-
tionship for all chemicals. Its limitations are discussed in depth in the
review article by Mingelgrin and Gerstl (1983).
RATE-LIMITED EXPRESSIONS
In a soil-water system where solution is flowing particularly through
structured soils, it is unlikely that complete equilibrium is reached between
solution and adsorbed concentrations. Although equilibrium is usually assumed
in modeling, several researchers have investigated the approach to equilibrium
dynamically. Using non-linear kinetics, van Genuchten et al. (1974), describe
the approach to equilibrium using equation (D-42).
fiqo N 1
.... CL -cs
L K2Pb J
(D-42)
5t
where KI and Kg are the kinetic rate coefficients for forward and backward
reactions, 0 is soil volumetric water content, and p^ is soil dry bulk density.
The quantity in brackets in equation (D-42) is a form of the Freundlich iso-
therm, equation (D-40). Thus, by assumption, the rate of approach equilibrium
is proportional to the deviation from equilibrium. A problem with using an
expression like equation (D-42) to describe adsorption is that the rate co-
efficients KI, K2 are not known a priori and must be fitted to column break-
through data by regression. Since there are also other transport coefficients.,
such as the dispersion coefficient which must be fitted at the same time, this
procedure often masks the physical significance of the parameters obtained in -
the experiments. Lindstrom et al. (1970), proposed a modified model describing
the kinetics of adsorption and desorption. Their model differed from that of
van Genuchten et al. (1974), in that the probability of adsorbing to the sur-
face was allowed to vary with the degree of surface coverage. However, this
model required still another parameter which had to be obtained by fitting
procedures.
A major complication in describing adsorption of dissolved species in a
soil-water system is that all adsorption sites are not equally accessible.
Leenheer and Ahlrichs (1971), for example, reported that although 60 percent
of the adsorption reaction was completed in one minute, adsorption of carbaryl
and parathion onto organic matter continued at a slow rate for another two or
three hours, and these authors postulated that this was due to the increased
time required to reach complete equilibrium with the internal surfaces of
organic matter by diffusion. Other complications of rate-limited adsorption
modeling are discussed in Rao and Davidson (1980).
D-13
-------
REFERENCES
Bear, J. 1972. Dynamics of Fluids in Porous Media. American Elsevier,
New York.
Bigger, J. W., and D. R. Nielson. 1967. Miscible displacement and leaching
phenomenon. Agron. Monog. 11:254-274.
Bresler, E. 1975. Two-dimensional transport of solutes during nonsteady
infiltration from a trickle source. Soil Sci. Soc. Amer. J^. 39:604-612.
Call, F. 1957. Soil fumigation. V. Diffusion of EDB through soils.
J.. Sci. Food Agr. 8:143.
Doorenbos, J., and W. 0. Pruitt. 1976. Crop Water Requirements. Irrig. Dr.
Paper 24, FAO Rome.
Farmer, W. J., M. S. Yang, J. Letey, and W. F. Spencer. 1980. Hexachlor-
benzene: Its vapor pressure and vapor phase diffusion in soil. Soil Sci.
Soc. Amer. J_. 44:676-680.
Feddes, R. A., E. Bresler, and S. P. Neuman. 1974. Field test of a modified
numerical model for water uptake by root systems. Water Resour. Res.
10:1199-1206.
Goring, C. A. I. 1962. Theory and principles of soil fumigation. Adv. Pest
Control Res. 5:47
Hamaker, J. W., and J. M. Thompson. 1972. Adsorption. In: C. A. I. Goring
and J. W. Hamaker (Eds.), Organic Chemicals in the Soil Environment.
Marcel Dekker, Inc., New York.
Hillel, D. 1971. Soil and Water. Academic Press, New York.
Jensen, M.E., and H. R. Haise. 1963. Estimating evapotranspiration from solar
radiation. Amer. Soc. Civ. Eng. Proc. 89:15-41.
Jury, W. A. 1982. Use of solute transport models to estimate salt balance
below irrigated cropland. In: D. Hillel (Ed.), Advances in Irrigation,
Vol. 1. Academic Press, New York.
Jury, W. A., and C. B. Tanner. 1975. Advection modification of the Priestly
and Taylor evapotranspiration formula. Agron. th 67:840-842.
D-14
-------
Jury, W. A., R. Grover, W. F. Spencer, and W. F. Farmer. 1980. Modeling
vapor losses of soil-incorporated triallate. Soil Sci. Soc. Amer.
j_. 44:445-450.
Jury, W. A., W. F. Spencer, and W. J. Farmer. 1983. Use of models for
predicting relative volatility, persistence, and mobility of pesticides
and other trace organics in soil systems. In: J. Saxena (Ed.), Hazard
Assessment of Chemicals, Vol. 2. Academic Press, New York.
Karickhoff, S. W. 1981. Semi-empirical estimation of sorption of hydrophobic
pollutants on natural sediments. Chemisphere 10:833-846.
Karickhoff, S. W., D. S. Brown, and T. A. Scott. 1979. Sorption of hydro-
phobic pollutants on natural sediments and soils. Water Res. 13:241-248.
Kirkham, D., and W. L. Powers. 1972. Advanced Soil Physics. Wiley Inter-
science, New York.
Leenheer, J. A., and J. L. Ahlrichs. 1971. A kinetic and equilibrium study
of the adsorption of carbaryl and parathion upon soil organic matter
surfaces. Soil Sci. Soc. Amer. Proc. 35:700-704.
Leistra, M. 1970. Distribution of 1,3-dichloropropene over the phases
in soil. J.. Agr. Food Chem. 18:1124.
Letey, J., and W. J. Farmer. 1974. Movement of pesticides in soil. J_nj
W. D. Guenzi (Ed.), Pesticides in Soil and Water. Soil Sci. Soc. Amer.
Madison, WI.
Lindstrom, F. T., R. Haque, V. H. Freed, and L. Boersma. 1969. Theory on
the movement of some herbicides in soil. Env. Sci. Tech. l:Ei61-565.
Lindstrom, F. T., R. Haque, and W. R. Coshow. 1970. Adsorption from solution.
J. Phys. Chem. 74:495-502.
Millington, R. J., and J. M. Quirk. 1961. Permeability of porous solids.
Trans. Farady Soc. 57:1200-1207.
Mingelgrin, U., and Z. Gerstl. 1983. Reevaluation of partitioning as a
mechanism of nonionic chemical adsorption in soil. J_. Environ. Qua!.
12:1-11.
Nash, R. G. 1980. Dissipation rates of pesticides from soils. In: W. G.
Knisel (Ed.), CREAMS. Vol. 3. U.S. Department of Agriculture~ Washington,
D.C.
Nielsen, D. R., R. D. Jackson, J. W. Cary, and D. D. Evans. 1972. Soil Water.
Amer. Soc. Agron. Special Pub. Madison, WI.
O'Connor, G. A., P. J. Wierenga, H. H. Cheng, and K. G. Doxtader. 1980.
Movement of 2,4,5-T through large soil columns. Soil Sci. 130:157-162.
D-15
-------
Oddson, J. K., J. Letey, and L. V. Weeks. 1970. Predicted distribution
of organic chemicals in soil. Soil Sci. Soc. Amer. Proc. 34:412-417.
Penman, H. L. 1948. Natural evaporation from open water, bare soil and grass.
Proc. Roy. Soc.. (London). 193:120-146.
Priestly, C. H. B., and R. J. Taylor. 1972. On the assessment of surface
heat flux and evaporation using large-scale parameters. Monthly Weather
Review. 100:81-92.
Rao, P. S. C., and J. M. Davidson. 1980. Estimation of pesticide retention
and transformation parameters. In: M. R. Overcash, Ed., Environmental
Impact of Nonpoint Source Pollution. Ann Arbor Sci., Publ., Ann Arbor,
MI.
Rao, P. S. C., D. E. Ralston, R. E. Jessup, and J. M. Davidson. 1980.
Solute transport in aggregated porous media. Soil Sci. Soc. Amer. J.
44:1139-1146.
Reeves, M., and J. Duguid. 1975. Water movement through saturated-unsaturated
porous media — a finite element model. ORNC-4927. Oak Ridge National
Laboratory, Oak Ridge, TN.
Ritchie, J. T. 1972. Model for predicting evaporation from a row crop with
incomplete cover. Water Resour. Res. 8:1204-1210.
Sallam, A., W. A. Jury, and J. Letey. 1983. Measurement of gas diffusion
coefficient under relatively low air-filled porosity. Soil Sci. Soc.
Amer. J_. (In press).
Shearer, R. C., J. Letey, W. J. Farmer, and A. Klute. 1973. Lindane
diffusion in soil. Soil Sci. Soc. Amer. Proc. 37:189-194.
Shouse, P., W. A. Jury, and L. H. Stolzy. 1980. Use of deterministic and
empirical models to predict potential evapotranspiration in a advective
environment. Agron. ±. 72:994-998.
Shouse, P., W. A. Jury, and L. H. Stolzy. 1982. Field measurement and model-
ing of cowpea water use and yield under stressed and well watered growth
conditions. Hilgardia 50:1-25.
Spencer, W. F., and M. M. Cliath. 1973. Desorption of lindane from soil
as related to vapor density. Soil Sci. Soc. Amer. Proc. 34:574-578.
Tanner, C. B. 1968. Evaporation of water from plants and soil. In:
T. T. Kozlowski, Ed., Water Deficits and Plant Growth. Academic Press,
New York.
Tanner, C. B., and W. A. Jury. 1976. Estimating evaporation and transpira-
tion from a row crop during incomplete cover. Agron. J_. 68:239-244.
D-16
-------
van Genuchten, M. Th. 1978. Land Disposal of Hazardous Wastes. CPA-600/
9-78-016. Proceedings of the 4th Annual Research Symposium.
van Genuchten, M. Th., and W. J. Alvcs. 1982. Analytical Solutions of the
One-dimensional Convective Dispersive Solute Transport Equation. U.S.
Department of Agriculture Tech. Bull. 1661.
van Genuchten, M. Th., J. M. Davidson, and P. J. Wierenga. 1974. An
evaluation of kinetic and equilibrium equations for the prediction of
pesticide movement through porous media. Soil Sci. Soc. Amer., Proc.
38:29-34.
Weber, J. B., P. W. Perry, and R. P. Upchurch. 1965. The influence of temper-
ature on the adsorption of paraquat, diquat, 2,4-D, and prometone by clays,
charcoal, and an anion-exchange resin. Soil Sci. Soc. Amer. Proc.
29:678-688.
D-17
-------
APPENDIX E
BI (TRANSFORMATION
by
R. L. Valentine and J. L. Schnoor
PROCESS DESCRIPTION
Definition
The term biotransformation is a general term describing any alteration of
a compound affected by living organisms. Biodegradation is a more specific
term usually referring to a biologically mediated transformation of a chemical
into more simple products by the removal of one or more substituent groups.
Neither of the above terms expresses the extent of change in the identity of a
compound, the mechanisms involved in that change, the rate at which this change
occurs, or the species responsible for this change.
Other terminology is frequently used to describe biodegradation. Mineral-
ization refers to the degradation of a chemical to inorganic products such
as carbon dioxide, water, ammonia, sulfate, nitrate, and chloride. The term
ultimate degradation is frequently used interchangeably with mineralization.
Partial degradation~is commonly used to describe a level of degradation less
than complete mineralization. Compounds that are not easily degraded are
recalcitrant and persistent in the environment (Alexander, 1965, 1977).
Importance of Microorganisms
In the soil environment, biotransformation of a compound is a result of
chemical reactions catalyzed by enzymes which are produced as part of the
metabolic activity of living organisms. Biotransformation in the soil has
been primarily attributed to the action of microorganisms although exudates
from plants and excretia from higher life forms may contribute to the overall
degradation of a compound.
It is generally believed that bacteria, actinomycetes, and fungi are of
primary importance. Algae have also been implicated in the degradation of some
chemicals (Craigie et al., 1965; Swisher, 1970) although their importance is
probably small except near the soil surface. The species of organisms in the
soil are similar to those of an aquatic environment (Stotzky, 1974). The
collective sum of all microorganisms present is sometimes referred to as a
consortia which constitutes a mixed population.
E-l
-------
The bacteria are usually the most abundant group on a numerical basis.
However, because bacteria are small and actinomycetes and fungi are large,
frequently having extensive filaments, bacteria may not account for the major-
ity of microbiological mass in a soil (Alexander, 1977). The actinomycetes
which are classified as filamentous bacteria are a transitional group between
bacteria and fungi and are usually second to bacteria in abundance (Alexander,
1977). However the relative importance of each group and species within each
group depends on many factors such as oxygen and hydrogen ion concentration
(PH).
Nutritional and Metabolic Considerations
Microorganisms require an energy and carbon source in addition to various
nutrients to maintain themselves. Microorganisms are generally placed in one
of two broad categories. Organic compounds are biodegraded principally by
heterotrophic organisms which require organic compounds to serve as both a
source of energy and carbon for cell growth and biosynthetic reactions. Fungi,
actinomycetes, protozoans, and most bacteria are heterotrophic. Autotrophic
organisms obtain energy from the oxidation of inorganic compounds such as
ammonia, or from sunlight, and use carbon dioxide as a sole carbon source.
Algae are primarily autotrophic although several species can also oxidize
organic carbon to replace light (Alexander, 1977).
The type of terminal electron acceptor (hydrogen acceptor) used in the
electron transport system which extracts energy via oxidation of substances is
an important characteristic of microbial metabolism. Under aerobic conditions,
oxygen is used. Organisms growing in the absence of oxygen must use an organic
product of metabolism (metabolic intermediate) or some inorganic substance such
as nitrate or sulfate as an electron acceptor.
Microorganisms can also be classified based on their ability to grow in
the presence or absence of oxygen. Three distinct groups exist (Alexander,
1977): aerobes require oxygen and do not grow in its absence; anaerobes
can grow only in the absence of oxygen; and facultative anaerobes can grow
either in the absence or presence of oxygen by switching electron acceptors.
The ultimate product of aerobic respiration is carbon dioxide and water.
Anaerobic metabolism produces incompletely oxidized simple organic substances
such as organic acids in addition to several other products such as carbon
dioxide, water, methane, and hydrogen gas.
Microbial growth frequently requires several other minerals and organic
substances. Nitrogen, phosphorus, potassium, sulfur, magnesium, zinc, calcium,
manganese, copper, cobalt, iron, and molybdenum are typical elements found in
microorganisms. Some of these minerals such as molybdenum are required in
extremely small amounts. Microorganisms may also need externally supplied
growth factors which are organic molecules not synthesized by the organism but
required in trace quantities for growth. Growth factors are structural build-
ing blocks such as ami no acids and vitamins.
The biotransformation of a compound is frequently a stepwise process
involving many enzymes and many species of organisms. Enzymes show a marked
specificity in catalyzing a single type of reaction of one or a few related
E-2
-------
substrates. Therefore, extensive degradation of a complex substrate requires
a large number of enzymes. A single species of organism may degrade a compound
by the sequential breakdown of intermediate products. Several intermediates
may be produced, and some of these intermediate products may not be further
degraded. Consortia of bacteria may operate in a manner in which metabolites
produced by one species are used as substrates for another.
Biotransformations may occur either inside a microorganism via intracel-
lular enzymes or outside the organism by the action of extracellular enzymes.
Soil enzymes may also be released as a result of cell lysis, and these enzymes
may retain their activity in the soil. Although the role of extracellular
enzymes in the transformation of xenobiotic compounds is unclear, such reac-
tions are known to occur (Skujins, 1967). Getzin and Rosefield (1968, 1971)
have shown that an extracellular enzyme found in soil was capable of degrad-
ing malathion.
Enzymes always present in a microorganism and part of the "normal"
metabolic activity are called constitutive enzymes. Other enzymes may be
produced in response to a specific substrate and are called indueible enzymes.
If an inducible enzyme is required for the degradation of a substrate, then
significant degradation will occur only after a sufficient time period (time
lag or acclimation period) has elapsed during which the inducible enzyme has
been produced to an adequate concentration.
Frequently, the biodegradation of a substrate does not lead to energy
production or formation of an essential nutrient. In other words, the sub-
strate does not promote growth. Such substrates cannot be used as sole carbon
and energy sources. Biodegradation of this sort has been termed cometabolism
(Horvath, 1972) or co-oxidation if the transformation involves an oxidation
(Perry, 1979) and probably occurs widely in the environment (Horvath, 1972;
Jacobson et al., 1980). Cometabolism occurs when an enzyme produced by an
organism to degrade one substance also degrades another although the metabolic
intermediates cannot be used. The metabolic intermediates are usually similar
to the substance cometabolized and accumulate in the environment.
At low substrate concentrations, the organism may not get enough energy
to supply maintenance requirements even if the compound can be used as an
energy source. This could result in a persistent compound. However, low con-
centrations of a substrate may be degraded when another source of energy and
carbon is available. The phenomena is called secondary utilization since
the trace material does not significantly contribute to growth and maintenance
of the organism (McCarty et al., 1981). Secondary utilization of substrates
may be important in the degradation of trace concentrations of chenicals.
BIOLOGICAL REACTIONS
The pathways for the biological transformation of many anthropogenic
organic chemicals have been determined. These chemicals include hydrocarbons,
pesticides, herbicides, and fungicides. Higgins and Burns (1975) and Alexander
(1977) provide a good general overview of the degradation of some of these
compounds. Several more detailed reviews in papers and books are those of
Hill (1978), Matsumura and Benezet (1978), Goring et al. (1975), Bollag
E-3
-------
(1974), Kaufman and Kearney (1976), Helling et al. (1971), Woodcock (1978),
Cripps and Roberts (1978), Laveglia and Dahn (1977), Bollag (1972), Crosby
(1973), and Alexander (1981).
It is beyond the scope of this chapter to discuss in depth the specific
pathways for the biotransformation of specific compounds by specific organisms.
However, most biological transformations fit into a relatively small number of
categories. Biological reactivity has been categorized both according to the
general type of chemical or functional group involved, e.g., aromatic, alkyl,
etc. (Helling et al., 1971; Higgins and Burns, 1975), as well as by the general
type of reactions occurring, e.g., oxidation, reduction, etc. (Crosby, 1973;
Goring et al., 1975; Kaufman and Kearney, 1976).
Some typical reactions of some chemicals and the resulting products are
given below. The reactions are not balanced. They do not show reactants which
are either supplied from the environment (e.g., molecular oxygen) or by the
organism. Some listed reactions may fit into more than one category. It must
be stressed that the feasibility of any reaction depends on many environmental
factors in addition to the specific organism present.
Some typical biologically mediated reactions are as follows:
1. Oxidation of Alkyl Compounds
RCH3 q RCH2OH q RC02H
2. Oxidative Dealkylation
ROCH3 q ROH + HCHO
3. Decarboxylation
RCOOH q RH + C02
4. Aromatic Hydroxylation
Ar q ArOH
5. Ring Cleavage
Ar(OH)2 q CHOCHCHCHCOHCOOH
6. B-Oxidation
CH3CH2CH3COOH q CH3COOH + CH3COOH
7. Epoxidation
RC = CR q R C-C R
E-4
-------
8. Oxidation of Sulfur
R2$ -»• R2SO + R2S02
9. Oxidation of an Ami no Group
RNH2 > RN02
10. Hydrolytic Dehalogenation
RCHC1CH3 + RCHOHCH3 + CT
11. Reductive Dehalogenation
RCC12R > RCHC1R + CT
12. Dehydrohalogenation
RCH2CHC1CH3 + RHC = CHCH3
13. Nitro-Reduction
RN02 -»• RNH2
14. Hydrolysis
RCH2CN + RCHONH3
RC(0)OR -»• RC(0)OH + HOR
RC(0)NR2 + RC(0)OH + HNR2
KINETICS OF BIOTRANSFORMATION
Several rate expressions have been proposed to describe the kinetics of
biotransformation in the environment. Most modeling work to date has focused
on the rates of degradation in the aquatic system and not in soils. However,
it is logical that similarities should exist between the two systems. Funda-
mental differences exist because a soil system is inherently a two-phase sys-
tem with microorganisms primarily attached to surfaces (Marshall, 1976).
Both population growth and soluble substrate utilization have been de-
scribed by Monod kinetics (Monod, 1949) when the substrate is used to provide
energy for growth and is not limited by any other needed substance. Several
rate expressions commonly used to describe chemical degradation in the environ-
ment can be rationalized in terms of the equations proposed by Monod.
Monod's work resulted in the formulation of expressions relating the rate
of microbial growth per unit volume, rg CMT~1L"J], and the rate of substrate
utilization per unit volume, r$ [MT1 L~3], to both substrate concentration and
microbial biomass (equations E-l and E-2),
E-5
-------
(E-l)
(E-2)
where X is the microbial biomass per unit volume of liquid [ML~3], S is the
substrate concentration [ML"3], and ym, ks and Y are kinetic constants. ym
is defined as the maximum growth rate constant [T"1], KS is the half-saturation
constant [ML'3], and Y is the true yield coefficient (a factor describing the
efficiency of converting chemical mass into microbial mass). It should be
noted that relationships (E-l) and (E-2) do not consider processes leading to
the loss of biomass such as endogenous respiration. The term um/Y is often
replaced by k which is defined as the maximum rate of substrate utilization per
unit mass of microorganisms
I'm
k = - (E-3)
Y
to yield relationship (E-4)
k X S
rs (E-4)
Ks + S
Relationship (E-4) is a general biotransformation rate expression from
which others can be rationalized. This rate expression indicates that the
biotransformation rate is a function of properties at a particular point in
the environmental system and is an algebraic expression not a differential
expression. An expression of the form
dS -k X S
(E-S)
dt Ks + S
which is commonly written as a rate expression reflects a mass balance on a
batch (i.e., non-flow) system and does not in general describe the changing
concentration of substrate in space and time within an environmental system.
Substrate removal following a form given in equation (E-4) is also some-
times referred to as following Michaelis-Menton kinetics after the work of
Michael is and Menton (1913) who showed that the rate of substrate loss in an
enzyme catalyzed reaction is of the form
Vm S
rsE = (E-6)
E-6
-------
where PSE is the rate of substrate reaction, K^, is the Michaelis-Menton
constant, and Vm is the maximum reaction velocity which has a value directly
proportional to the enzyme concentration, E
Vm = k'E (E-7)
where k1 is a constant.
The similarity of equations (E-4) and (E-6) may be more than coincidental
if the enzymatic nature of microbial degradation is such that the microorgan-
isms may be treated as very small "packets" of enzymes. In any ca-se, both
Monod and Michaelis-Menton kinetics are frequently referred to when substrate
utilization models are presented (even though the Monod model is based on
empirical observation, and that of Michaelis-Menton is applicable only to
enzyme kinetics).
Two limiting conditions may be encountered in the use of equation (E-4).
At high substrate concentrations where S is much greater than Ks, the rate
of substrate utilization becomes zero order in substrate
rs = kX (E-8)
At low concentrations where Ks is much greater than S, the rate is first order
in both substrate and biomass and second order overall
k
rs = — XS = k2XS (E-9)
KS
where kg = k/Ks is a second order constant.
Further simplifications can be made if it is assumed that the microbial
concentration X is constant with time, such as in the case in which a dynamic
population equilibrium is established. New "constants" which depend on
microbial concentration can be defined
kx = kX (E-10)
^ = k2X (E-ll)
and substituted into equations (E-4), (E-8) and (E-9). The "exact" Monod
expression then becomes
Ks +S
At high substrate concentration the rate is a "constant"
rs = kx
and at low substrate concentration the rate is first order expression.
E-7
(E-12)
-------
rs = kiS (E-14)
Because the constant Iq depends on microbial concentration, relationship (E-14)
is sometimes referred to as a psuedo-first order relationship. A summary of
all Monod based rate expressions is given in Table E-l.
TABLE E-l. MONOD BASED RATE EXPRESSIONS
= = = = = = = = = = := = = = = = = = = = = = === = = === = = = = = = = = = === === = = = = := = === = = = === === = =
High S Low S
Assumptions Any X,S S > K$ S < Ks
k S X
Any X,S rs = ....... rs = kX rs = k2 SX
Ks + S
kxS
X = constant rs = ------ rs = kx rs = kj S
Ks + S
The direct applicability to soil systems of any rate expression based on
a rigorous interpretation of Monod or Michaelis-Menton kinetics would not be
expected since the microbial population in soil is diverse. These organisms
may derive energy from a large number of substrates some of which are at a
concentration too low to support growth. Monod kinetics also does not consider
the availability of substrates which may be limited by adsorption or mass
transfer.
Nevertheless, rate expressions having similar form to those based on
Monod kinetics have been used to describe the biotransformation of chemicals
in the environment. While these expressions are not on a firm theoretical
basis, they are logical. Logic dictates that the rate of biotransformation be
a function of both substrate concentration and of the actively-degrading
microbial population; it should approach zero as either X or S approaches
zero; and it should increase as X or S increases at least over some range in
concentration. Such a principle can be applied to both soluble and adsorbed
chemical concentration as well as to attached and dispersed microorganisms.
Biotransformation rate expressions specifically applied to soils have
generally not included microbial activity as a separate parameter. Two
general rate expression types have been considered (Hamaker, 1966; Goring et
al., 1975): A power rate law
rs = kSn (
where k is a constant, n is the reaction order, and S is the chemical concen
tration, and a "hyperbolic" expression
E-8
-------
rs = ------ (E-16)
C2 + S
where cj and eg are constants. Both expressions have been used to model over-
all chemical disappearance without regard to the relative Importance of abiotic
and blotic factors. In other words, they are general expressions describing
apparent chemical loss, not necessarily general biotransformation rate expres-
sions. Goring et al . (1975), compared the use of these rate expressions for
describing various degradation processes. The hyperbolic model is consistent
with Monod kinetics, yielding a simple first order degradation rate expression
at low chemical concentrations
(E-17)
where kj = cj/cg, and a zero order expression at high chemical concentrations.
rs = ci (E-18)
Rationalization of the hyperbolic expression with Monod kinetics requires the
constant ci to be a function of microbial concentration which implies that
expression (E-17) can be rewritten as a second order expression
rs = k2 S X (E-19)
where X is microbial concentration and k£ = kj/X.
More complex Monod type expressions while potentially describing the rate
of degradation more accurately are computationally more complex and require
difficult-to-measure parameters. Furthermore, the use of more simple rate
expressions may be rationalized at low chemical concentrations where a first
order dependence on chemical concentration may be reasonable (Hamaker, 1972).
The biotransformation of several pesticides in soils has been found to follow -
first order kinetics (Meikle et al., 1973; Walker and Stojanovik, 1973).
However, more complex rate expressions have been used to describe degradation
rates. McCarty et al . (1981), have developed a biofilm model incorporating the
more complex Monod type expression (E-4) (including microbial concentration as
a parameter) with its full complement of constants and have suggested that the
model may be applicable in predicting the fate of chemicals in the subsurface
environment.
The use of the power-rate model cannot be rationalized in terms of Monod
kinetics unless n equals zero or one (which corresponds to the high and low
substrate concentration assumptions used in the Monod formulation). Goring
et al . (1975), have suggested the use of the power rate model where the goal
is to develop an empirical relationship to fit degradation rate data without
regard to the nature of the degradation mechanism. This level of empiricism
should be avoided when using a fate model which presumably accounts for all
important phenomena as separate components.
E-9
-------
FACTORS INFLUENCING BIG-TRANSFORMATION
A variety of environmental and chemical factors limit the biotransforma-
tion of chemicals in the environment (Alexander, 1965).
Inaccessibility of the Chemical
The compound may be held in a microenvironment which precludes microbial
attack, e.g., it may be sorbed or entrapped in a manner which prevents the
organism or its enzyme from reaching the substrate.
Absence of Some Factor Essential for Growth
For example, no growth would occur in the absence of water, nitrogen,
or phosphorus.
Toxicity of the Environment
This could be the result of biologically generated inhibitors, extremes
in environmental temperature or pH, high salt concentration, high toxic levels
of the chemical itself, or other inhibitory conditions.
Inactivation of the Requisite Enzymes
Enzyme activity may be lost by adsorption or inhibited by other substances
(e.g., phenolic and polyaromatic substrates or products).
A Structural Characteristic of the Molecule
Some functional characteristics prohibit the formation of an enzyme-
substrate complex.
Inability of the Community of Microorganisms to Metabolize the Compound Because
of Some Physiological Inadequacy
An enzyme capable of degrading the compound may not exist, or the compound
may not be able to penetrate into the cells in which the appropriate enzymes
exist.
AVAILABILITY OF THE CHEMICAL
Availability of the chemical is discussed separately from major environ-
mental factors because of its general importance. Both macroscopic and micro-
scopic processes affect chemical availability. At the macroscopic level,
microorganisms may be distributed in a very patchy manner precluding signifi-
cant biotransformation where their numbers are relatively small. At the
microscopic level, phenomena occur which cause chemicals to partition them-
selves between the liquid and solid phases and may play a major role in deter-
mining the overall rate of biotransformation by determining chemical avail-
ability. Because of the relatively poor understanding of the causes of this
partitioning, the more general term sorption will be used in this discussion
instead of the frequently used term, adsorption.
E-10
-------
Upon sorption, biotransformation rates are frequently drastically changed.
Both retardation and acceleration effects have been observed (Weber and Coble,
1968; Steen et al., 1979; Kjellenberg et al., 1982). In most cases, sorption
appears to reduce degradation rates. For example, Steen et al. (1979), demon-
strated that some chemicals were not available to aquatic microorganisms when
sorbed onto sediments. However, growth of bacteria on surfaces to which sub-
strates can sorb (and concentrate) has been observed in solutions too dilute
to support a dispersed growth (Heukelekian and Heller, 1940; Kjellenberg et
al., 1982).
The effects of sorption on biodegradation may have several possible causes
(Burns, 1975):
1. A sorbed chemical may not be attacked by an enzyme or microorganism
because of a physical barrier or inability of the enzyme to form a
substrate-enzyme complex with the adsorbed chemical (rate decreases).
2. The chemical may not be sorbed in close enough proximity to a micro-
organism or enzyme so that interaction is delayed until the substrate
flux brings the reactants together (rate decreases).
3. The sorbed chemical may not be concentrated in an area where pro-
liferation of microbes can occur (rate decreases).
The effect of sorption on the mathematical expression of the degradation
rate can be handled several ways. The most general approach would be to con-
sider the biodegradation of the sorbed chemical as a separate rate term inde-
pendent of the rate of soluble chemical degradation. For example, if chemical
loss was assumed to be first order, then the rate of sorbed chemical loss per
mass of soil (rA) and the rate of soluble chemical loss per volume of liquid
(rw) could be expressed as
rA = kASA (E-20).
rw = kvA, (E-21)
where kA and kw are first order biodegradation rate constants [T-1i| specific
to each phase, and SA and Sy, are the sorbed and soluble chemical concentration
with units of mass sorbed per mass of soil and mass per liquid volume, respec-
tively. The adsorbed concentration and rate could also be expressed per volume
of soil by multiplying rA by soil density.
The rate of total chemical loss can be easily related to the rates of loss
from each phase for first order kinetics if rapid equilibrium is assumed. The
rate of total chemical loss rs can be expressed several ways: relative to soil
water volume, to soil mass, or to soil volume. Each yields a first order
expression of the same form,
rs - kTST (E-22)
where kj is a first order rate constant and Sj is the total chemic.il concentra-
tion. The rate constant kj is a function of the partition coefficient KQ
E-ll
-------
1], the ratio of soil mass to soil water volume, M, and the phase specific
biotransformation rate constants, k/\ and kw according to
kT = f!kw + f2kA (E-23)
where fi and fg are the fraction of total chemical mass in the water and on the
soil, respectively, and given by
fl - - ...... (E-24)
1 + KpM
KpM
*2 ........ (E-25)
1 + KpM
It should be stressed that while equations (E-20) and (E-21) may be re-
garded as "fundamental" biotransformation rate expressions, equation (E-22) is
not, since it incorporates characteristics of the soil system which are inde-
pendent of biotransformation. This should be kept in mind since biotransforma-
tion studies frequently measure a first order biotransformation constant that
is in actuality not a single fundamental constant. It would clearly be in
error to use a first order rate constant measured from simple observations of
the rate of total chemical loss as a fundamental biotransformation rate con-
stant in a model which also separately incorporates the effect of sorption on
the rate of biotransformation.
Mass transfer limitations could also govern degradation rate by control-
ling the chemical concentration actually available to microorganisms. Mass
transfer effects can be incorporated into the overall chemical mass balance.
McCarty et al. (1981), incorporated mass transfer limitations in a biofilm
model which was suggested as being applicable to modeling of the subsurface
environment.
MAJOR FACTORS AFFECTING BIODEGRADATION
Factors that influence the rate of biodegradation are listed in Table E-2.
These variables fall into three general categories: (1) those that affect
substrate availability, (2) those that directly affect the microbial population
size, composition, and activity (e.g., population interactions), and (3) those
that directly control the degradation rate itself (e.g., temperature).
TABLE E-2. MAJOR ENVIRONMENTAL FACTORS AFFECTING BIOTRANSFORMATION
1. pH 6. Oxygen
2. Temperature 7. Nutrients
3. Water content 8. Nature of microbial population
4. Carbon content 9. Acclimation
5. Clay content 10. Concentration
E-12
-------
Most factors are not independent but are highly interrelated. For exam-
ple, pH may affect both the availability of a substrate as well as the composi-
tion of the microbial community. The extent of the interrelation of factors
frequently leads to overall effects on degradation that are not always the same
since some factors may act to increase the rate of degradation while others may
decrease it. While some generalizations regarding the net effect of each
factor can be made, the literature is full of exceptions.
£H
Microbial activity is affected by the hydrogen ion concentration. Micro-
organisms can grow in a limited range of pH values. Optimum growth may occur
at different pH values for different organisms. Most soil bacteria grow
optimally near neutral conditions (pH 6.5 to 8.5). Actinomycetes are less
tolerant of low pH as a group and generally are not found below pH 5; but they
do grow relatively better than bacteria under alkaline conditions. Fungi can
develop over a wide range of pH that generally spans from highly acidic pH (as
low as 2 to 3) to alkaline conditions although individual species may show a
lesser tolerance (Alexander, 1977). Therefore, as a soil becomes more acidic,
the proportion of fungi to bacteria generally increases. In addition, individ-
ual enzymes produced by a specific microbe may be affected by pH and therefore
may affect reaction rates with substrates.
Hydrogen ion concentration may affect the availability of substrate by
influencing the amount of chemical sorbed. Hydrogen ions may compete for
sorption sites thereby decreasing the amount sorbed. If the chemical ionizes,
lowering the pH may either increase or decrease adsorption depending on whether
the protonated chemical or nonprotonated chemical is more readily adsorbed
(Burns, 1975). Enzymatic activity may also favor one form of an ionizable
chemical.
Temperature
Microbial activity is generally stimulated by temperature increases within
the range tolerated by the microorganisms, and some species tend to dominate
within certain temperature ranges (Edwards, 1964). Microorganisms may be
classified into three major groups depending on the temperature range for
optimum growth rate: mesophiles (optimum activity between 25 and 35°C), psy-
chrophiles (optimum activity below 20°C), and thermophiles (optimum activity
between 45 and 65°C). Many species of microorganism may have optimum growth
over narrower or wider ranges.
Increasing rates of degradation with increasing temperature are a conse-
quence of increased chemical reaction rates. This effect is sometimes expres-
sed in terms of a QIQ value which is the ratio of the rate at a temperature 10
degrees higher than a reference temperature to that of the rate at the refer-
ence temperature. For example, a reaction characterized by a Qjg of 2.0 would
double in rate for a 10 degree increase in temperature. QJQ values may be
interpreted as reflecting a change in rate constants used in degradation rate
expressions.
E-13
-------
Other empirical relationships have been used to calculate blodegradatlon
rate constants such as the use of empirical factors correlating temperature
effects on rate constants for specific reactions
(E.26)
where kj is the rate constant at temperature TI, k2 is the rate constant at
T2, and 9 is an empirical factor which depends on temperature (Grady and Lim,
1980).
Smith and Walker (1977) developed an empirical expression relating an
overall first order pesticide degradation rate constant to both soil temper-
ature and soil-water moisture content, M
iq = aMT + e (E-27)
when a and 0 are constants which depend on specific pesticide and soil type.
However, their equation was developed to correlate observed losses in pesticide
without discerning between biological and nonbiological transformations.
Rate data have also been correlated using the Arrhenius equation
k = A exp(-E/RT) (E-28)
where k is the reaction rate constant, A is a preexponential constant, R is the
gas constant, E is the activation energy (a constant), and T is the absolute
temperature (Lee and Ryan, 1979). While the Arrhenius equation may be useful,
it should be kept in mind that it is applicable to only single rate limiting
chemical reactions and not generally to reactions that may involve several
consecutive steps and that are common in microbial systems. Relationship
(E-26) may be derived from consideration of relationship (E-28).
Adsorption processes are exothermic, and, therefore, an increase in tem-
perature is expected to decrease the amount of chemical sorbed. This may
increase the concentration which is available for degradation and hence may
contribute to increased degradation. In general, decreased sorption and
increased microbial activity associated with higher temperatures usually en-
hance pesticide degradation (Rao and Davidson, 1980). However, exceptions have
been noted in which increased temperature has increased sorption possibly as a
result of structural change in the adsorbing surface (Burns, 1975). This could
somewhat offset any increase in microbial activity. The overall effect would
depend on the relative rates of biotransformation of sorbed and dissolved
chemical.
Water Content
Moisture is required for microbial growth, but an oversupply can reduce
gas exchange and limit oxygen which is depleted by microbial metabolism; thus,
an anaerobic environment is created. This is because the rate of diffusion of
oxygen in water is slow. An optimum level for aerobic organisms is about 50-75
percent of the soil moisture holding capacity (Alexander, 1977). Bacterial
E-14
-------
and fungi populations have been closely correlated with moisture content
(Alexander, 1977).
Walker (1978) correlated herbicide half-lives to soil moisture content, M
using
tl/2 = AM
"B
where A and B are constants.
Moisture content affects soluble chemical concentration. Decreasing
moisture causes increased sorption of chemical which may affect the degrada-
tion rate. Increased moisture could result in lower chemical concentrations
in the aqueous phase and decreased biotransformation rates. Increased
moisture could also dilute any potentially toxic chemical and thereby could
possibly increase its biotransformation rate.
Carbon Content
Soil organic carbon content is due to organic matter consisting of an
amorphous colloidal fraction or humus, in a macro or microscopically definable
component made up of plant and animal material at various stages of decay
(Burns, 1975). Organic carbon and associated nutrients are the major constit-
uents of microbial food. Community size in mineral soils has been directly
related to organic matter content (Alexander, 1977).
Organic matter strongly sorbs many chemicals, and both persistence and
sorption have been correlated with organic content (Upchurch and Pierce, 1958;
Hartley, 1964). However, while sorption may tend to decrease the rate of
degradation, the potential increase in microbial activity may counter this
effect. Verma et al . (1975), have also suggested that soluble humic polymers
can act as stabilizing agents making compounds less resistant to biodegrada-
tion. Sequestering of a chemical could also make it more resistant to attack. .
Additions of supplemental carbon such as sucrose or nutrient broth to
soils have resulted in both increased (McCormick and Hiltbold, 1966; Miyazaki
et al., 1969) and decreased rates of biodegradation of a specific organic
chemical (Kaufman et al . , 1968; McClure, 1970). A decreased rate may result
from avoidance of the specific organic chemical because of a greater ease
of utilization of the added carbon. An increased rate may be a result of
increased microbial activity and cometabolism. However, addition of readily
utilizable carbon sources generally increases degradation rates. These
observations may have significant bearing on the biodegradability of chemical
mixtures, particularly if one chemical is in great excess of another. Cometab
olism and secondary substrate utilization may occur during the degradation of
both the added as well as the naturally occurring carbon.
Clay Content
Clay contributes to the mineral fraction of soil. Many chemicals, includ
ing enzymes, sorb to clay. Clays are cation exchangers; so cationic chemicals
may be readily sorbed.
E-15
-------
The quantity and type of clay is important in determining the effect on
degradation. Weber and Coble (1968) added montmorillonite and kaolinite clays
to nutrient solutions containing diquat. Montmorillonite sorbed the diquat,
and biodegradation was reduced in direct proportion to the amount of clay
added. Kaolin did not sorb diquat, and biodegradation was not affected by the
quantity of clay added. Clay may also affect the oxygen content in a soil
because a clay soil tends to retain a higher moisture content (which in turn
may restrict oxygen diffusion).
Oxygen
Oxygen is needed as the terminal electron acceptor for some microorgan-
isms, and the presence or absence of oxygen determines the specific species as
well as the general type of organism present. Filamentous fungi and actinomy-
cetes are, as a group, strict aerobes. Common soil bacteria include aerobes,
strict anaerobes, and facultative aerobes.
Some compounds are transformed by microorganisms only under aerobic con-
ditions while some are transformed under either aerobic or anaerobic condi-
tions, or not at all (Guenzi and Beard, 1968; Sethunathan and MacRae, 1969; -
Bower et al., 1981). The presence of oxygen not only may affect the biotrans-
formation mechanism (and rate) but also the products formed. Pesticide miner-
alization rates are considerably lower, and substantial amounts of intermediate
metabolites may accumulate under anaerobic conditions (Rao and Davidson, 1980).
Soil environments are rarely totally aerobic. Soils tend to be more
anoxic with depth because oxygen is consumed rapidly relative to the rate it
diffuses in from the surface. Anaerobic oiicroenvironments may exist even in
well-aerated soil (Alexander, 1977).
Nutrients
Nutrient is a general term referring to any substance required for growth:.
Nutrients supply elements which become part of the protoplasm (biomass), pro-
duce energy for cell growth and biosynthetic reactions, and serve as electron
acceptors. The general nutritional requirements of microorganisms have been -•
previously discussed in this appendix and will not be repeated here.
The consequence of a nutrient limitation may be inhibition of microbial
proliferation or the inhibition of microbial respiration. The effect would be
to reduce the rate of biodegradation. The availability of carbon, phosphorus,
and nitrogen frequently affects microbial growth because these are the major
constituents of protoplasm. These nutrients may be supplied by the chemical
being degraded or by the environment.
Nature of the Microbial Population
The degradation of chemicals is characterized by microbial specificity,
i.e., not all chemicals are degraded by all organisms. For example, cellulose
degraders are predominantly fungal (Burns, 1975), and many hydrocarbons support
E-16
-------
few microbial species (Brock, 1970). Differences also exist in th« degradative
pathways and rates exhibited by different species of microorganisms;.
Population size and spatial distribution are important factors. Micro-
habitats may differ greatly in numbers of microorganisms, possibly because of
the distribution of organic matter or other environmental factors. It is
axiomatic that an increase in an active degrading population should increase
the overall rate of degradation.
Interactions among species may indirectly affect biodegradation rates.
Competition and predation affect microbial composition. Metabolic activity of
one species may affect that of others through the production of cofactors or
metabolic intermediates that can be used by another organism. Chemicals may be
degraded sequentially by a mixture of species. Focht (1972) showed that two
microbes in liquid culture were capable of together metabolizing DDT, but no
single species has been discovered which can utilize DDT as a sole carbon and
energy source. Healy et al. (1980), have shown that a consortia of bacteria
were responsible for the anaerobic transformation of ferulic and bunzoic acids.
Acclimation
The degradation of a chemical frequently does not occur at an appreciable
rate immediately after its introduction to an environment. This "lag" period
in which the organism and system become acclimated may have several causes.
Required enzymes may have to be induced, a change in the environment may allow
for preferential selection of the degrading organism, or the added chemical may
gradually increase the population of existing microbes capable of degrading the
chemical (Brock, 1966).
The significance of prior acclimation, particularly for more complex
compounds, is well known. For example, Whiteside and Alexander (1960) demon-
strated that 2,4-D was degraded in soil by microorganisms after a lag period.
Furthermore, a second application was degraded more rapidly without a lag
period. Presumably the required enzymes were still present after the first
application had been degraded.
Concentration
High concentrations of chemicals may be bactericidal or bacteriostatic,
thereby reducing the rate of degradation. Many pesticides have been shown to
be more slowly degraded when applied to soils at high concentrations (Ahmed and
Morrison, 1972; Davidson et al., 1980). On the other hand, low concentrations
may not be sufficient to initiate enzyme induction or support microbial growth.
ESTIMATION OF RATE PARAMETERS
Field vs. Laboratory Measurements
Evaluation of biotransformation rate constants under field conditions is
difficult because of limitations imposed by concurrent abiotic processes (dis-
cussed in Limitations Section). In most field studies, the area is divided
into plots, and the chemical is applied and appropriately incorporated to the
E-17
-------
desired depth. Inability to properly control the myriad of variables can
result in meaningless numbers. Effective soil sterilization under field
conditions is particularly difficult to achieve, if not impossible.
Laboratory tests using soil samples are required if the role of abiotic
factors in degradation is to be assessed. A sterile control is often used
along with a non-sterile soil sample to distinguish between abiotic and biotic
factors. The soils may be incubated under either aerobic or anaerobic con-
ditions. Anaerobic conditions can be maintained by flushing with nitrogen,
nitrogen-carbon dioxide mixtures, or other inert gas. Aerobic conditions can
be maintained by flushing with air or by leaving the sample open to the atmos-
phere. Redox potentials have been measured as an indicator of the relative
oxidation state of a soil (Quispel, 1949; Bohn, 1971) under field conditions
and have been suggested as a basis for classifying the oxygen levels in soils
(Patrick and Muhaptra, 1968). Flooded conditions are easily simulated in the
laboratory. Wilson and Noonan (1984) have recently discussed the use of labo-
ratory microcosms to evaluate the transformation of chemicals in the subsurface
environment. Methodologies for field and laboratory testing are available
(Howard et al., 1975). Methods used in sampling the subsurface environment and
obtaining a representative soil sample are available (Gilmore, 1959; Wilson, -
1980; Atlas and Bartha, 1981).
Measurement of Test Chemical Disappearance
No attempt is usually made to differentiate between biotransformation of a
sorbed chemical and that found in true solution. Measured biotransformation
rate constants are therefore usually based on total chemical concentration in
the soil and include sorption effects. The importance of phase possibly could
be resolved through experiments utilizing various soil to water ratios and dif-
ferentiation between dissolved and sorbed chemical.
Biotransformation of a chemical under field conditions has generally been
measured as the rate of disappearance of solvent-extractable parent compound.
Therefore, the extent of transformation of the structure of the parent chem-
ical is not considered, and accumulation of metabolic intermediates cannot be
ruled out. Chemical extraction and specific analytical methodologies for
quantification of chemical concentration are discussed elsewhere (Howard et
al., 1975; Sherma, 1981).
Mineralization rates which indicate complete destruction of the parent
compound can be measured under laboratory conditions by using ^C labeled
chemicals and collecting evolved 14C02 in a solution such as KOH. The rate of
mineralization can be determined from the rate at which 14C02 is produced.
Biotransformation rate constants based on ^COg formation are expected to be
smaller than those based on solvent extraction of the parent compound (Rao and
Davidson, 1980). However, abiotic reactions resulting in decarboxylation of
the chemical could cause an erroneously high apparent biotransformation rate.
Inability to extract all the parent compound would also lead to misleadingly
high rates.
E-18
-------
Biotransformation Rate Expressions and Constants
Biotransformation rate expressions In soil systems have been expressed
primarily as a first order rate relationship. Data analysis is simplified
since only chemical concentrations need to be measured. In principle, first
order rate constants can be obtained from batch experiments from the slope of
the relationship
^n S/S0 = -k^ (E-30)
where S is the substrate concentration at time t, S0 is the initial substrate
concentration, and kj is the first order rate constant. In addition, sub-
strate half-life ty2 's easily related to the first order rate constant by
ki = 4n(2)/t1/2 = 0.693/t1/2 (E-31)
This approach is frequently taken even if it is apparent that the time depend-
ence of chemical disappearance is more complex. Ou et al . (1982), and Rao and
Davidson (1980) have compiled an extensive list of first order transformation
rate constants and half-lives for selected pesticides based on laboratory or -
field studies under aerobic or anaerobic conditions. It should be noted that
in most studies, experimental conditions were not controlled to eliminate
abiotic transformation pathways so that the constants obtained cannot neces-
sarily be attributed to microbial action only.
Second order rate constants can be developed if, in addition to chemical
concentration, microbial biomass or some surrogate parameter is measured.
Second order rate constants are frequently expressed relative to biomass con-
centration or microbial numbers. Evaluation of a second order constant would
require experiments to be conducted over a range of microbial biomass. Second
order constants could then be related to pseudo-first order constants obtained
at a constant microbial concentration X by
kx = k2 X (E-32)
from which k2 can be obtained as the slope of a kj vs. X plot.
Measurement of Microbial Numbers and Biomass
Determination of second order rate constants requires that some population
parameter directly proportional to chemical biotransformation rate be measured
and converted to either equivalent biomass or numbers. Because the actively
degrading population is impossible to measure accurately, estimates of total
biomass and numbers are made.
Proper attention must be given to sampling, storage, and extraction
methods which are used to separate microorganisms from the soil matrix. Micro-
biological sampling methods have been discussed by McNabb and Mallard (1984),
Gilmore (1959), Williams and Gray (1973), Atlas and Barthar (1981), and Board
and Lovelock (1973). Many approaches have been used to estimate microbial
numbers and biomass in the subsurface environment. Leach (1984) has recently
provided an excellent review of commonly used methods as well as those that are
E-19
-------
of a more experimental nature. Methodologies reviewed included bioassays,
measurement of growth, staining and microscopy, metabolic and physiological
responses, measurement of enzyme activity, and the determination of cellular
components (Leach, 1984). Methods vary greatly in precision, ease, and lower
detection limit of microbial numbers. Several of the more commonly used meth-
ods are briefly discussed here.
Microbial Numbers--
Microbial numbers are frequently enumerated by plating methods and by
direct microscopic evaluation. Alexander (1977) has discussed general methods
applicable in enumerating microorganisms.
Plate count methods utilize a soil sample of known size diluted with water
and agitated to disperse microorganisms. Samples are withdrawn and either
mixed with a cooled agar media (pour plate method) or spread onto a pre-dried
agar plate (spread plate method) which is then incubated. Each colony formed
is assumed to have originated from a single microorganism. Plating methods may
rely on a general media on which a wide variety of heterotropic organisms are
known to grow. Different media may be used to differentiate fungi from bac-
teria and actinomycetes.
Epifluorescence microscopy utilizing acridine-orange (AO) as a stain has
been used to determine microbial numbers in soils (Trolldenier, 1973) and
subsurface samples (Webster et al., 1985). Both living and dead cells are
counted. Electron microscopy has also been utilized (Gray, 1967). Many more
bacteria are usually counted by direct epifluorescence than by conventional
plate counts in aquatic samples, and this suggests that epifluorescence may be
the better method in enumerating bacterial numbers in soils (Daley, 1979). The
results of cell counts can be converted to biomass by applying appropriate
volume and density factors (Dale, 1974), or the numbers can be reported as
direct counts.
Biomass--
Measurement of the concentration of a chemical that is a cellular compon-
ent can be used to estimate cell number and biomass if a relationship between
them is assumed. Adenosine triphosphate (ATP) is present in all living organ-
isms and can be extracted and measured by determining the amount of light pro-
duced by its reaction with added firefly luciferace. The use of ATP measure-
ments as a measure of biomass has been discussed by Leach (1981), Eiland (1979),
Jenkinson and Ladd (1981), and Eiland and Nielsen (1979). Stevenson et al.
(1979), summarized the advantages of using ATP as an indicator of microbial
populations:
1. Sparse communities are detectable.
2. All physiological types of microbes are included.
3. Only living cells respond.
4. The method is quick.
5. The method is precise.
Problems with using ATP are associated with the extraction efficiency,
presence of interfering substances, variability in the amount of ATP per unit
biomass (which depends on the particular organism species and its nutritional
E-20
-------
state), and the instability of ATP unless the sample is properly preserved.
However, even given these limitations, ATP measurements may be more reliable
as a biomass indicator than those based on plating methods (Cavari, 1976).
Webster et al. (1985), have recently compared the use of acridine-orange stain-
ing and ATP measurements to determine cell numbers in subsurface samples. The
measurement of chloroform extractable lipid phosphate is also a simple tech-
nique (White et al., 1979).
LIMITATIONS TO APPLYING DEGRADATION RATE EXPRESSIONS
There are a number of practical as well as theoretical limitations in the
use of degradation rate expressions which may result in discrepancies between
predicted and measured chemical concentrations. These limitations arise from
measurement difficulties and assumptions made in arriving at a rate expression,
and they are summarized in Table E-3.
TABLE E-3. LIMITATIONS IN APPLYING DEGRADATION RATE EXPRESSIONS
==== ========= ==== ========== ===== === ===== = === ===== === ==== = = === =====::=:= === = ===== =
Measurement Model
1. Abiotic vs. biotic processes 1. Specificity of constants
2. Microbial concentrations 2. Concentration dependence of constants
3. Bound chemical residues 3. Acclimation
Measurement Limitations
Abiotic vs. Biotic Processes--
Biological transformations may be difficult to distinguish from abiotic
processes without carefully controlled experiments. Abiotic processes include
strictly chemical reactions, photochemical reactions, and physical processes
such as leaching and volatilization. The effect of volatilization has been
previously discussed. Chemicals may undergo various transformations such as
hydrolysis, oxidation, and reduction in the absence of microorganisms or their
enzymes (Helling et al., 1971; Crosby, 1976). The products of nonbiological
degradation are frequently identical to those resulting from biological
reactions (Burns, 1975).
A common approach to resolving biotic and abiotic processes is to inhibit
microbial growth and physiological processes in the soil under laboratory con-
ditions that minimize physical losses such as volatilization. Soil steriliza-
tion has been accomplished by chemical addition (e.g., sodium azide), heat
treatment, and irradiation (Skujins, 1967).
Some sterilization procedures such as heat treatment may alter physical
and chemical properties of the soil which affect abiotic processes. Chemical
treatment may inactivate microorganisms but have little effect on extracellular
E-21
-------
enzymes. Furthermore, microorganisms may be rendered incapable of true growth
but still capable of respiration. Some methods of microbe inactivation may
release enzymes which still function outside the cell. Presumably the absence
of chemical transformation in a sterilized soil indicates the absence of abiotic
processes in non-sterilized soil. Furthermore, transformations occurring in a
sterilized soil are assumed totally attributable to abiotic processes. However,
neither assumption may be true.
Microbial Concentration--
Application of a degradation rate expression (e.g., a second order rela-
tionship) may require measurement of microbial biomass or numbers. Unfortu-
nately, microbial biomass and numbers may not always correlate well with ob-
served degradation rates (Wright, 1979; Nesbitt and Watson, 1980). This may
have several causes beyond those associated with imprecise or inappropriate
methodology. The total microbial population is not necessarily responsible for
chemical degradation, and the fraction responsible may change with location.
Ideally, only the "active" population should be measured. The action of extra-
cellular enzymes may also account for a significant fraction of chemical trans-
formation. Fundamental differences in metabolic activity could exist as a
result of environmental changes.
Estimates of microbial numbers and biomass depend on the method used to
obtain them. For example, plating methods may involve the use of a media on
which some microbial species will not grow. Several microorganisms may be
clumped together and appear as a single colony (Alexander, 1977), or incubation
times may not be long enough for some organisms to grow. Estimates of fungi
may also be in error because colonies appearing on agar may be derived from a
spore or a fragment of vegetative mycelium. Direct microscopic determination,
such as epifluorescence microscopy, cannot discern viable from nonviable cells.
Chemical methods may also suffer from this problem as well as having a variable
ratio of measured chemical to biomass (Karl, 1930).
Bound Chemical Residues--
Chemicals or their transformation products added to soil may react with
material in the soil and become "bound" (Hill, 1978). Kaufman et al. (1968),
defines a bound pesticide residue as "that unextractable and chemically uniden-
tifiable pesticide residue remaining in fulvic acid, humic acid, and humic
fractions after exhaustive sequential extraction with non-polar organic and
polar solvents." Therefore, inability to recover an added chemical from soil
may not be due entirely to biodegradation.
Model Limitations
All of the degradation rate expressions are empirical albeit logical and
capable of being rationalized in terms of Monod kinetics or Michaelis-Menton
kinetics. As a consequence, use of the results at conditions not used to
develop or evaluate the rate expressions must be done with caution.
Specificity of Constants--
Field and laboratory measurements of rate constants will only be comparable
when experimental conditions (especially temperature, moisture content, soil
E-22
-------
properties, nutrient concentrations, oxygen concentration, and chemical con-
centration) are similar. Use of such data requires appropriate caution
because of unknown factors. In general, transformation rates under field con-
ditions appear greater than those under laboratory incubation studies because
of multiple processes occurring in the field (Rao and Davidson, 1980, 1982).
In particular, first order constants are inherently more site specific than
second order constants since a first order constant does not expressly consider
changes in microbial activity. Constants based on the rate of disappearance of
total chemical concentration may be a function of the degree of chemical sorp-
tion.
Concentration Dependence of Constants--
Degradation constants are assumed to be independent of concentration.
However, this may not be true. For example, Boethling and Alexander (1979a,
1979b) have shown that the Michaelis-Menton constants determined at high con-
centrations may not be consistent with the observed degradation rates at low
chemical concentrations.
Additionally, use of an inadequate rate expression could give rise to an
apparent concentration dependence. For example, a first order dependence on ..
substrate is not expected at high substrate concentration. Evaluation of first
order constants at both low and high concentrations could yield two different
"constants." Rate constants may also appear to increase when evaluated at
lower concentrations of a chemical if a chemical exhibits some inhibitory
effect on growth at higher concentrations.
Acclimation-
Rate constants will depend on the extent of acclimation of the microbial
population to the test chemical. Microbial populations in the field may not be
exposed to the chemical long enough for acclimation to occur. Hence, the
observed degradation rates may be lower than those predicted from constants
obtained from acclimated populations. Acclimation times have been observed
to range from a few hours to a few months depending on the chemical structure
and its concentration.
E-23
-------
REFERENCES
Ahmed, N., and F. 0. Morrison. 1972. Longevity of residues of four organo-
phosphate insecticides in soil. Phytoprotection 53:71-74.
Alexander, M. 1965. Biodegradation: problems of molecular recalcitrance
and microbial infallibility. Adv. Appl. Microb. 7:35-80.
Alexander, M. 1977. Introduction to Soil Microbiology. Wiley, Second
Edition.
Alexander, M. 1981. Biodegradation of chemicals of environmental concern.
Science 211:132-138.
Atlas, R. M., and R. Bartha. 1981. Microbial Ecology. Addison-Wesley,
Chapter 4.
Board, R. G., and D. W. Lovelock. 1973. Sampling - Microbiological Monitoring
of Environments, Academic Press, New York.
Boethling, R. S., and M. Alexander. 1979a. Microbial degradation of organic
compounds at trace levels. Environ. Sci. Techno!. 13:889-911.
Boethling, R. S., and M. Alexander. 1979b. Effect of concentration of organic
chemicals on their biodegradation by natural microbial communities.
Appl. Environ. Microb. 37:1211-1216.
Bohn, H. L. 1971. Redox potentials. Soil Sci. 112:39-45.
Bollag, J. M. 1972. Biochemical transformation by soil fungi. Crit. Rev.
Microb. 2:35-58.
Bollag, J. M. 1974. Microbial transformations of pesticides. Adv. Appl.
Microb. 18:75-130.
Bower, E. J., B. E. Rittman, and P. L. McCarty. 1981. Anaerobic degradation
of halogenated 1- and 2-carbon organic compounds. Environ. Sci. Techn.
15:596.
Brock, T. D. 1966. Principles of Microbial Ecology. Prentice-Hall, Inc.
Englewood Cliffs, NJ.
Brock, T. D. 1970. Biology of Microorganisms. Prentice-Hall, Inc., Englewood
Cliffs, NJ.
E-24
-------
Burns, R. G. 1975. Factors affecting pesticide loss from soil. Jlru E. A.
Paul and A. D. McLaren (Eds.), Soil Biochemistry, Vol. 4. Marcel-Dekker,
New York.
Cavari, B. 1976. ATP in Lake Kinneret: indication of microbial biomass or of
phosphate deficiency. Limnol. Ocean. 21:231.
Craigie, J. S., M. McLachlan, and G. H. N. Towers. 1965. A note on the fis-
sion of an aromatic ring by algae. Can. J^. Bot. 43:1589-1590.
Cripps, R. E., and T. R. Roberts. 1978. Microbial degradation of herbicides.
JJK I. R. Hill and S. J. L. Wright (Eds.), Pesticide Microbiology.
Academic Press, London, New York.
Crosby, D. G. 1973. The fate of pesticides in environment. Ann. Rev. Plant
Physiol. 24:467-492.
Crosby, D. G. 1976. Non-biological degradation of herbicides in the soil.
In: L. J. Audus (Ed.), Herbicides. Physiology. Biochemistry, Ecology.
2nd edition. Academic Press, London.
Dale, N. G. 1974. Bacteria in intertidal sediments - factors related to their
distribution. Limnol. Ocean. 19:509-518.
Daley, R. J. 1979. In: Cooterton and Col well (Eds.), Native Aquatic Bacteria:
Enumeration. Activity, and Ecology. ASTM/STP 695:29.
Davidson, J. M., P. S. C. Rao, L. T. Ou, W. B. Wheeler, and D. F. Rothwell.
1980. Adsorption Movement and Biological Degradation of Large Concen-
trations of Selected Pesticides in Soil. EPA-600/2-80-124.
Edwards, C. A. 1964. Factors affecting the persistence of insecticides in
soil. Soils and Fert. 27:451.
Eiland, F. 1979. An improved method for determination of adenosine triphos-
phate (ATP) in soil. Soil Biol. Biochem. 11:31-35.
Eiland, F., and B. S. Nielsen. 1979. Influence of cation content on adenosine
triphosphate determinations in soil. Micnob. Ecol. 5:129-137.
Focht, D. D. 1972. Microbial degradation of DDT metabolites to carbon dioxide
water and chloride. Bull. Environ. Contam. Toxic. 7:52-56.
Getzin, L. W., and I. Rosefield. 1968. Organophosphorous insecticide degra-
dation by heat-labile substances in soil. J_. Agr. Food Chem. 16:598-601.
Getzin, L. W., and I. Rosefield. 1971. Partial purification and properties
of a soil enzyme that degrades the insecticide Malathion. Bioch. Bioph.
Acta. 235:422-53.
Gilmore, A. E. 1959. A soil sampling tube for soil microbiology. Soil Sci.
87:95-99
E-25
-------
Goring, C. A. I., D. A. Laskowski, J. Hamaker, and R. W. Miekle. 1975.
Principles of pesticide degradation in soil. In: R. Hague and
F. H. Freed (Eds.), Environmental Dynamics of Pesticides. Plenum Press.
Grady, C. P., and H. C. Lim. 1980. Biological Uastewater Treatment: Theory
and Application. Marcel-Dekker, Inc., New York and Basel.
Gray, T. R. G. 1967. Stereoscan electron microscopy of soil microorganisms.
Science. 155:1668-1670.
Guenzi, W. D., and W. E. Beard. 1968. Aerobic conversion of DDT to DDD and
aerobic stability of DDT in soil. Soil Sci. Soc. Amer. Proc. 32:522.
Hamaker, J. W. 1966. Mathematical predictions of cumulative levels of pes-
ticides in soil. Adv. Chem. Series. 60:122.
Hamaker, J. W. 1972. Decomposition: quantitative aspects. In: C. A. I.
Goring and J. W. Hamaker (Eds.), Organic Chemicals in the Soil Environ-
ment. Vol. 1. Marcel-Dekker.
Hartley, G. S. 1964. Physical behavior in the soils. In: L. J. Audus,
(Ed.), Herbicides: Physiology, Biochemistry, Ecology. Academic Press,
London.
Healy, J. B., L. Y. Young, and M. Reinhard. 1980. Methanogenic decomposition
of ferulic acid, a model lignin derivative. Appl. Env. Microb. 39:436.
Helling, C. J., P. C. Kearney, and M. Alexander. 1971. Behavior of pesticides
in soil. Adv. Agron. 23:147.
Heukelekian, H., and A. Heller. 1940. Relation between food concentration
and surface for bacterial growth. »L Bacteriol. 40:547.
Higgins, I. J., and R. G. Burns. 1975. The Chemistry and Microbiology of
Pollution. Academic Press, London, New York, San Francisco.
Hill, I. R. 1978. Microbial transformations of pesticides. ^n_: I. R. Hill
and S. J. L. Wright (Eds.), Pesticide Microbiology. Academic Press,
London, New York.
Horvath, R. S. 1972. Microbial co-metabolism and the degradation of organic
compounds in nature. Bacteriol. Rev. 36:146-155.
Howard, P. H., J. Saxena, P. R. Durkin, and L. T. Ou. 1975. Review and Eval-
uation of Available Techniques for Determining Persistence and Routes of
Degradation of Chemical Substances in the Environment. EPA-560/5-75-006.
Jacobson, S. N., N. C. O'Mara, and M. Alexander. 1980. Evidence for
cometabolism in sewage. Appl. Environ. Microb. 40:917-921.
E-26
-------
Jenkinson, D.S., and J. N. Ladd. 1981. In: E. A. Paul and J. N. Ladd (Eds.)
Soil Biochemistry. Dekker.
Karl, D. M. 1980. Cellular nucleotide measurements and applications in micro-
bial ecology. Microb. Rev. 44:739-796.
Kaufman, D. D., J. R. Plimmer, P. C. Kearney, J. Blake, and F. S. Guardia.
1968. Chemical versus microbial decomposition of amitrole in soil. Weed
Sci. 16:266.
Kaufman, D. D., and P. C. Kearney. 1976. Microbial transformations in the
soil. In: L. J. Audus (Ed.), Herbici des, Physiology, Biochernistry,
Ecology. Vol. 2. Academic Press, London, New York.
Kjellenberg, S., B. A. Humphrey, and K. C. Marshall. 1982. Effect of inter-
faces on small starved marine bacteria. Appl. Environ. Microb.
43:1166-1172.
Laveglia, J., and P. A. Dahn. 1977. Degradation of organophosphorus and
carbamate insecticides in the soil and by soil microorganisms. Ann. Rev."
Entmol. 22:483-513.
Leach, F. R. 1981. ATP determination with firefly luciferase. J^ Appl.
Biochem. 3:473.
Leach, F. R. 1984. Biochemical indicators of groundwater pollution. Jto:
G. Britton and C. P. Berba (Eds.), Groundwater Pollution Microbiology.
Wiley-Interscience.
Lee, R. F., and C. Ryan. 1979. Microbial degradation of organoch'lorine com-
pounds in estuarine waters and sediments. EPA-600/9-79-012. Proceedings
of the Workshop; Microbial Degradation of Pollutants in Marine Environ-
ments. Washington, DC.
Marshall, K. C. 1976. Interfaces in Microbial Ecology. Harvard Press,
Cambridge, MA.
Matsumura, F., and H. J. Benezet. 1978. Microbiol degradation of insecti-
cides. ln_: I. R. Hill and S. J. L. Wright (Eds.), Pesticide Micro-
biology. Academic Press, London, New York.
McCarty, P. L., M. Reinhard, and B. E. Rittman. 1981. Trace organics in
ground water. Environ. Sci. Tech. 15:40-51.
McClure, G. W. 1970. Accelerated degradation of herbicides in soil by the
application of microbial nutrient broths. Contrib. Boyce Thomson Inst.
24:235-240.
McCormick, L. L., and A. E. Hiltbold. 1966. Microbiological decomposition
of atrazine and diuron in soil. Weeds 14:77-82.
E-27
-------
McNabb, J. P., and G. E. Mallard. 1984. Microbiological sampling In the
assessment of groundwater pollution. In: 6. Britton and C. P. Gerba
(Eds.), Groundwater Pollution Microbiology. Wiley-Interscience.
Meikle, R. W., C. R. Younston, R. T. Hedlund, C. A. I. Goring, J. W. Hamaker,
and W. VI. Addington. 1973. Measurement and prediction of picloram
disappearance rates from soil. Weed Sci. 21:549.
Michaelis, L., and M. L. Menton. 1913. Die Kihetik der Invertinwirkung.
Biochemische Zietschrift. 49:333-369.
Miyazaki, S., G. M. Boush, and F. Matsumura. 1969. Metabolism of 14C-
chlorapropylate by Rhodotorula gracilis. Appl. Microb. 18:972-976.
Monod, J., 1949. The growth of bacterial cultures. Ann. Rev. Microb.
3:371-394.
Nesbitt, J. J., and J. R. Watson. 1980. Degradation of herbicide 2,4-0 in
river water - I. Description of the study area and survey of rate
performing factors. Water Res. 14:1683-1688.
Ou, L. T., J. M. Davidson, P. S. C. Rao, and W. B. Wheeler. 1982. Pesticide
transformations in soils. In: P. S. C. Rao and J. M. Davidson (Eds.),
Retention and Transformation of Selected Pesticides and Phosphorus in
Soil-Water Systems: A Critical Review. EPA-600/3-82-060.
Patrick, W. H., and J. R. Mahapatra. 1968. Transformation and availability
to rice of nitrogen and phosphorus in waterlogged soils. Advan. Agron.
20:323-359.
Perry, J. J. 1979. Microbial cooxidations involving hydrocarbons. Microb.
Rev. 43:59-72.
Quispel, A. 1949. Measurement of the oxidation-reduction potentials of normal
and inundated soils. Soil Sci. 63:265-275.
Rao, P. S. C., and J. M. Davidson. 1980. Estimation of pesticide retention
and transformation parameters required in non-point source pollution
models. In: M. R. Overcash and J. M. Davidson (Eds.), Environmental
Impact of Non-point Source Pollution. Ann Arbor Science Publishers,
Ann Arbor, Michigan.
Rao, P. S. C., and J. M. Davidson. 1982. Rentention and Transformation of
Selected Pesticides and Phosphorus in Soil-Water Systems: A Critical
Review. EPA-600/3-82-060.
Sethunathan, N., and I. C. MacRae. 1969. Persistence and biodegradation of
diazinon in submerged soils. »L Agr. Food Chem. 17:221-225.
Sherma, J. 1981. Manual of Analytical Quality Control for Pesticides and
Related Compounds. EPA-600/2-81-059. U.S. Environmental Protection
Agency, Research Triangle Park, NC.
E-28
-------
Skujins, J. J. 1967. Enzymes in soil. Jji: A. D. McClaren and G. H. Peterson
(Eds.), Soil Biochemistry. Dekker, New York.
Smith, A. E., and A. Walker. 1977. A quantitative study of asulam persistence
in soils. Pest. Sci. 8:449-456.
Steen, W. C., D. F. Paris, and G. L. Baughman. 1979. Effects of sediment
sorption on microbiol degradation of toxic substances. Proceedings of
the Symposium on Processes Involving Contaminants and Sediments. ACS,
Honolulu.(April).
Stevenson, L. H., T. H. Chrzanowski, and C. W. Erkenbrecher. 1979. The
adenosine triphosphate assay: conceptions and misconceptions. ASTM/STP
695:99-111.
Stotzky, G. 1974. Activity, ecology and population dynamics of microorganisms
in soil. In: A. I. Laskin and H. Lechevalier (Eds.), Microbiol Ecology.
CRC Press, Cleveland, Ohio.
Swisher, R. D. 1970. Surfactant Biodegradation. Marcel-Dekker, New York.
Trolldenier, G. 1973. The use of epifluorescence microscopy for counting soil
microorganisms. In: T. Rosswall (Ed.), Modern Methods in the Study of
Microbial Ecology. Bulletins from the Ecological Research Committee,
Swedish Natural Science Research Council, Stockholm. 17:55-59.
Upchurch, R. P., and W. C. Pierce. 1958. The leaching of monuron from lake-
land sand soil. II. The effect of soil temperature, organic matter, soil
moisture and herbicide. Weeds 6:24.
Verma, L., J. P. Martin, and K. Haider. 1975. Soil Sci. Soc. Amer. Proc.,
39:279.
Walker, W. W., and B. J. Stojanovic. 1973. Microbial vs. chemical degrad-
ation of malathion in soil.
-------
Williams, S. T., and T. R. G. Gray. 1973. _In: R. G. Board and D. W. Lovelock
(Eds.)t Sampling-Microbiological Monitoring of Environments. Academic
Press, New York.
Wilson, L. G. 1980. Monitoring in the Vadose Zone. A Review of Technical
Elements and Methods. EPA-600/7-80-134. U.S. Environmental Protection
Agency, Las Vegas, NV.
Wilson, J., and M. J. Noonan. 1984. Microbial activity in model aquifer
systems. Ij: C. P. Gerba and G. Bitton (Eds.), Groundwater Pollution
Microbiology. John Wiley and Sons.
Woodcock, D. 1978. Microbial degradation of fungicides, fumigants and
nematocides. In: I. R. Hill and S. J. L. Wright (Eds.), Pesticide
Microbiology. Icademic Press, London, New York.
Wright, R. T. 1979. Natural heterotrophic activity in estuarine and coastal
waters. EPA-600/9-79-012. ln_: Workshop: Microbial Degradation of
Pollutants in the Marine Environment.
E-30
-------
APPENDIX F
NONBIOLOGICAL TRANSFORMATION
by
R. I. Valentine
PROCESS DESCRIPTION
While all processes leading to structural changes in chemical can be
considered as occurring as the result of chemical reactions, these processes
can be categorized as being either biological, chemical, or photochemical
transformations. Biological transformations, i.e., transformations mediated
by microorganisms, are discussed elsewhere. Chemical transformations of
importance in the soil environment include hydrolysis and oxidation-reduction
reactions (Redox reactions). Photochemical transformations can occur only
in the presence of light and hence are expected to be important only at or
very near the soil surface.
Hydrolysis
Chemical Hydrolysis refers to a general class of chemical reactions
involving water which result in the net exchange of a hydroxyl group, OH",
with some other group, X (Mabey and Mill, 1978). This occurs principally
by cleavage of carbon-X and phosphorus-X bonds. The leaving group depends
on the type of chemical. Common leaving groups include halides (Cl~, Br~),
alcohols (R-0~), amines (Rj^N"), and sulfur (RS~) and phosphorus containing
moieties ([RO^POo")- Several organic functional groups are particularly
susceptible to hydrolysis (Harris, 1981). These include alkylhalides, amides,
amines, carbamates, carboxylic acid esters, epoxides, nitriles, phosphonic
acid esters, phosphoric acid esters, sulfonic acid esters, and sulfuric acid
esters. Many functional groups are not very susceptible to hydrolysis. These
include saturated and unsaturated alkanes, aromatic hydrocarbons and amines,
halogenated aromatics, aromatic nitro compounds, alcohols, ketones, glycols,
phenols, ethers, carboxylic acids, sulfonic acids, and polycyclic and hetero-
cyclic polycyclic aromatic hydrocarbons (Harris, 1981).
Hydrolysis reactions represent one class of reactions involving water.
Several others may occur which include acid-base reactions, hydration of
carbonyl compounds, elimination reactions, and addition to carbon-carbon double
bonds. The first two reactions are considered reversible while the latter
are not, and therefore could lead to permanent chemical transformation. A
discussion of these is beyond the scope of this chapter.
F-l
-------
Mabey and Mill (1978) have discussed the hydrolysis of 12 classes of
compounds of environmental importance. Many reactions yield a single product,
however, several products may be possible if more than one site for attack
exists as exemplified by the hydrolysis of malathion (Wolfe et al . , 1977).
Several typical general reactions are presented.
a.
Examples of Hydrolysis
R - Cl + H20 q ROH + H+ + C
an alkyl halide an alcohol
b.
H20 q
an ester a carboxylic acid
R2OH
an alcohol
c.
- CHR2 + H20 q
0 OH OH
an epoxide a diol
d.
ROC(0)NR1R2 + H20 q ROH + C02 +
a carbamate an alcohol
an amine
e. RC(ON)R!R2 + H20
an amide
q RCOOH +
a carboxylic acid
an amine
f.
RCH2CN + H20
a nitrile
RCH2COOH + NH3
a carboxylic acid ammonia
Mechanism and Kinetics-
Hydrolysis is the result of a nucleophilic substitution reaction in which
water or a hydroxide ion (a nucleophile) attacks electrophilic carbon or
phosphorus and displaces a leaving group. These nucleophilic substitution
reactions occur by either a unimolecular (Sfjl) or bimolecular (Sfj2) mechanism.
These general mechanisms are described in more detail elsewhere (Lowry and
Richardson, 1976; March, 1977).
Aqueous hydrolysis has been characterized as potentially involving three
distinct mechanistic processes. These are an acid catalyzed reaction, a
neutral reaction, and a base promoted reaction. In pure water, a hydrogen
ion (H+) may catalyze the hydrolysis of a compound by reacting with it in such
a way as to promote nucleophilic attack. Neutral hydrolysis processes involve
F-2
-------
attack by water alone. Hydroxide ion may also directly attack the electro-
philic carbon or phosphorus to displace the leaving group.
In water, each process can give rise to a separate reaction term first
order in chemical concentration, [S]. The general overall rate of hydrolysis
(rn) is the sum of these three terms:
rh = kH[H+][S] + kH0[H20][S] + kQH [OH'][S] (F-l)
where kH, O, and kgH are rate constants for the acid-catalyzed, the neutral,
the base-promoted process. Since the concentration of water is essentially
constant, its concentration can be incorporated into the neutral process
constant to define a new constant
ko • >
-------
(F-7)
where kns and knw are the overall phase specific rate constants, and [Ss] and
[SyJ are the concentrations of chemical on the soil and in the water,
respectively.
The rate of total chemical loss, r^j. can be shown to be first order in
total chemical concentration [Sj] if sorption is rapid.
HIT = khT [ST] (F-8)
The overall first order "rate" constant, knT, is a function of the partition
coefficient, Kp, soil to water ratio, M, and the phase specific overall
hydrolysis rate constants, kns and knw,
khT - flkhS + f2^hw (F-9)
1
where fi = — (F-10)
1 + KpM
KpM
f2 (F-ll)
(1 + KpM)
The presence of soils, however, may affect hydrolysis independently of
sorption. Burkhard and Guth (1981) observed that several herbicides are
hydrolyzed in soils according to a first order relationship. They found that
the rate of hydrolysis in soils was increased compared to that in pure aqueous
solution, but that the rates decreased with the extent of sorption; this indi-
cates that sorption "protected" the herbicides, and that the presence of soils
accelerated hydrolysis. Konrad and Chester (1969) also showed that the hydrolr
ysis of an organophosphate insecticide in soil was first order in chemical con--
centration and catalyzed in the presence of soil. In contrast to the results
of Burkhard and Guth, Konrad et al., observed that the first order rate con-
stants were directly related to sorption (i.e., rates were proportional to the
degree of sorption).
Chemical Oxidation-Reduction Reactions
The concept of chemical oxidation-reduction (redox) is more easily under-
stood for simple inorganic reactions than for redox reactions involving organ-
ics. In the soil environment, redox reactions involving both inorganic and
organic species are important.
Inorganic chemists define oxidation as the loss of electrons and increase
in oxidation number while reduction is the gain in electrons and decrease in
oxidation number (March 1977). The oxidation number of an atom represents the
hypothetical charge an atom would have if the ion or molecule were to dis-
sociate (Stumm and Morgan, 1981). Unfortunately when dealing with organic
redox reactions, these definitions are not easy to apply. While electrons are
directly transferred in some redox reactions, the mechanisms of most organic
F-4
-------
reactions do not involve direct transfer but transfer of atoms (March, 1977).
Organic oxidations frequently involve a gain in oxygen and a loss in hydrogen
atoms. The reverse is true for reductions.
Organic chemists have set up a series of functional groups arranged in
order of increasing oxidation state to facilitate the classification of reac-
tions as either oxidations or reductions. Table F-l provides a suirmary of the
relative oxidation states of several important functional groups. Conversion
of a functional group into one in a higher oxidation state category character-
izes oxidation of the original group. Reduction is defined by the conversion
to a group in a lower oxidation state (March, 1977).
TABLE F-l. RELATIVE OXIDATION STATES*
Increasing Oxidation State
Least Oxidized Most Oxidized
RH ROM RC(0)R RCOOH C02
RC1 (R)2CC12 RC(0)NH2 CC14
RNH2 -C=C- RCC13
=====================================================================
*Adapted from March (1977).
The importance of chemical redox reactions in the soil is not well docu-
mented although laboratory model systems have shown potential importance (Stone
and Morgan, 1984a,b). Soil does contain a wide variety of substances including
natural organics, metal oxides, and oxygen which may be capable of entering
into redox reactions with chemicals. Several good discussions of redox re-
actions in soil and aquatic environments are available (Helling et al., 1971;
Crosby, 1976; Plimmer, 1978a,b; Mill et al., 1982). Further research on the
significance of redox reactions in soil systems is needed particularly if
transformation pathways occurring in environments which do not contain sig-
nificant microbial populations are to be adequately characterized.
Mechanism and Kinetics
An oxidation cannot occur without a reduction. In simple inorganic reac-
tions, the mechanism involves the exchange of an equal number of electrons.
For example the oxidation of Fe+2 to Fe+3 by the reduction of Mn+^ to Mn+3
involves the direct exchange of a single electron
F-5
-------
Fe+2 -»• Fe+3 + e" an oxidation
Mn+4 + e" * Mn+3 a reduction
Each individual reaction is called a half-reaction (one is a reduction, the
other an oxidation), and the net balanced reaction is their summation.
The concept of the additivity of half reactions can be applied even to
complex organic and inorganic redox reactions using several "rules" as an aid
in balancing the half reactions (Sawyer and McCarty, 1978; Snoeyink and Jenkins,
1980; Stumm and Morgan, 1981). For example, the oxidation of glucose to carbon
dioxide and water by oxygen can be written as
C6 H12 06 + 6H20 * 6C02 + 24H+ + 24e" an oxidation
602 + 24H+ + 24e" * 12 H20 a reduction
C6 H12 °6 + 6H2° + 602 * 6C02 + 6H2°
In many cases, half reactions can be written that do not describe the real
mechanism but are only a convenient "bookkeeping" method to correctly balance
the net reaction as is the case for the oxidation of glucose cited above.
Frequently, many individual elementary reactions, transient intermediates such
as free radicals, and other commonly available species such as H+, OH", and
H20, are involved in the transformation of a chemical. Several different
mechanisms may also lead to more than one product type.
Some of the reactions listed in the section on biotransformation show
transformations of a chemical that are typical oxidation or reduction reactions
(e.g., oxidative dealkylation, decarboxylation, nitro-reduction, etc.). Such
reactions, of course, must be accompanied by the simultaneous reduction of some
other substance if the chemical of interest is oxidized or by the oxidation of
some other substance if the chemical of interest is reduced. For example,
oxygen may be reduced to H20 in an oxidation of a chemical.
Evidence for the existence of redox reactions in the soil comes from the
identification of reaction products that are characteristic of redox reactions.
While much is known about the mechanisms and kinetics describing the oxidation
or reduction of specific organic chemicals in the laboratory by added chemical
oxidants or reductants such as chlorine (Cl?)* potassium permanganate, and
oxygen (02), very little is known about the oxidants and reductants or reaction
mechanisms and kinetics of importance in soil.
No relationship has been derived from first principles to describe the
general oxidation or reduction of chemicals in the soil and been validated.
However, a simple second order relationship has been used to model the oxida-
tion of a chemical in aqueous phase reactions (March, 1977; Smith et al., 1977;
Mill et al., 1980; Mill et al., 1982),
F-6
-------
and kox is a second order rate constant under conditions where the oxidant
species ox was photochemically produced and involved in the rate limiting step
of the reaction mechanism. The general applicability of this or a similar rate
expression describing the oxidation or reduction of a compound has yet to be
shown in soil systems.
Photochemical Transformations
The term photochemical transformation is a general term referring to all
reactions mediated by light through the action of a photon absorbed either by
the specific chemical of interest or some naturally occurring substance.
Photochemical transformations include a variety of common reactions such as
oxidation, reduction, elimination, isomerization, substitution, addition,
fragmentation, and hydrolysis (Turro, 1965; Crosby, 1976). Photochemical
transformation of a compound may lead to one or more reaction products typical
of these reactions.
Photochemical reactions have been shown to play a significant role in the-
transformation of several chemicals at^ the soil surface (Lichtenstein et al.,
1970; Helling et al., 1971; Crosby, 1976; Burkhard and Guth, 1979). However,
since light does not penetrate very far into soils, photochemical reactions do
not play a significant role in the transformation of the bulk of a chemical if
it is incorporated into the soil (Hautala, 1978). Consideration of photo-
chemical transformations may be important in order to correctly ascertain the
flux of chemical from the soil surface.
The general topic of photochemistry is discussed in the books of Calvert
and Pitts (1966) and Turro (1965). Several good reviews describing photolysis
of environmental importance are available (Zifiriou, 1977; Baughman and
Lassiter, 1978; Zepp and Baughman, 1978; Harris, 1982; Mill et al., 1982;
Miller, 1983).
Mechanism and Kinetics—
Photochemical transformations may occur by one or more processes that
depend on chemical structure and the presence of other substances in the
environment. The direct photolysis of a chemical occurs when the chemical that
absorbs light undergoes reaction while in an excited state. Indirect photoly-
sis occurs when one chemical called a sensitizer absorbs light energy and
transfers it to another chemical which then undergoes a reaction. This type of
indirect photolysis is a photosensitized process. Energy transfer can also
reduce (quench) the photolysis of a target chemical if energy is transferred
from the target chemical before it can react. Chemicals which deactivate the
excited state of the target chemical are called quenchers. Many natural or-
ganic substances are sensitizers or quenchers.
A specific chemical may also react with an energetic intermediate produced
when a natural substance absorbs light. Singlet oxygen and oxyradicals are
examples of such intermediates identified as being important in aqueous systems
which can react to cause the transformation of a chemical (Zepp et al., 1977;
(Mill et al., 1980; Wolfe et al., 1981; Miller, 1983). Their existence may
F-7
-------
also be Important at the soil surface. Reactions Involving these Intermediates
are frequently referred to as photooxygenations or photoinitiated free radical
reactions (after reaction with singlet oxygen and oxyradicals, respectively).
However, the mechanisms involved in photochemical transformations are frequently
unclear and cannot be differentiated.
The kinetics of photochemical transformations on soils has not been exten-
sively studied since light does not penetrate very far into soil. However, the
rate of photochemical transformation, rp, of a chemical in aqueous solution has
been expressed as a first order relationship
rp - K[C] (F-13)
where C is the concentration of chemical, and K is a constant which depends on
the photon flux, the light absorption coefficient, and the reaction quantum
yield which expresses the efficiency of conversion of absorbed light into chem-
ical reaction (Zepp and Cline, 1977; Baughman and Lassiter, 1978). The first
order constant is also related to the concentration of sensitizer present, if
any (Zepp and Baughman, 1978). Hautala (1978) studied the photolysis of
several pesticides on very thin soil samples having particles which did not
block access to light and also obtained a first order rate of loss although
the rates were very small in comparison to those expected in aqueous solution.
Since light does not penetrate deeply into soil, first order kinetics would not
be expected in the bulk of the soil.
Sorption may increase or decrease photolysis possibly by involvement of
quenching agents, photosensitizing agents, or by changing the absorption spec-
tra of a chemical (Ivie and Casida, 1971; Rosen, 1972; Crosby, 1976; Miller and
Zepp, 1979). Burkhard and Guth (1979) observed a decrease in the rate of
photolysis of several organophosphorus insecticides on soil with decreasing
moisture content which they attributed to a protective effect of sorption.
Hydrogen ion concentration may also have an effect on rates and products formed
(Crosby and Leitis, 1973).
FACTORS INFLUENCING NONBIOLOGICAL TRANSFORMATIONS
A number of factors influence nonbiological reactions by either determin-
ing the availability of important reactants or modifying the reactivity.
Little is known about the actual importance of many of these factors beyond
conjecture based on laboratory studies involving conditions quite remote in
some cases from those of environmental importance. Many of the factors listed
in Table F-2 are interrelated in complex ways. For example, the general effect
on sorption needs to be considered.
Soil Moisture Content
Soil moisture commonly serves as a medium in which hydrolysis and redox
reactions occur. It also affects the chemical concentration and amount sorbed.
The rates of hydrolysis of some chemicals have been correlated with moisture
content. Saltzman et al. (1976), have shown that the rate of hydrolysis of
parathion on Kaolinite, a common soil constituent, increases with moisture up
to about 11 percent moisture content. At a moisture content beyond this, the
F-8
-------
TABLE F-2. FACTORS INFLUENCING NONBIOLOGICAL REACTIONS
1. Soil moisture content 5. Soil clay content
2. pH 6. Metal oxides/ions and
inorganic acids/bases
3. Oxygen 7. Temperature
4. Soil organic content 8. Sunlight
rate decreased. In addition, the high solubility of oxygen in water makes soil
moisture a good vehicle for transport of oxygen. However, oxidation of a
compound by oxygen may be decreased if soil moisture limits oxygen avail-
ability. Surface reactions with atmospheric oxygen may still readily occur at
low moisture levels.
EH
Hydrogen ion concentration may affect the rates of hydrolysis several
ways:
a) Directly, because the rate of the acid-promoted process is an explicit
function of hydrogen ion concentration
b) By affecting the iom'zation of the chemical
c) By affecting the amount of chemical sorbed
Hydrogen ion concentration is related to the hydroxide ion concentration
by the ionization product of water, Kw,
I/
[OH-] = ---- (F-14)
therefore, the generalized rate of hydrolysis (in the absence of general acid
base catalysis) may be expressed as
rh = KH[H+] + k0 + koHkw % [S] (
The exact dependence of the hydrolysis rate depends on the magnitude of each
rate constant. The pH will also affect the concentration of inorganic acids
and bases which may act as general catalysts.
Not all chemicals are hydrolyzed by mechanisms involving all three proc-
esses. For example, organic halide hydrolysis in aqueous solution is not
significantly acid-catalyzed (Mabey and Mill, 1978). In addition, rate con-
stants can differ greatly, and half-lives can vary from a few minutes to thou
sands of years depending on chemical type and pH. For example, Zepp et al.
F-9
-------
(1974), report that the hydrolysis half-life of a butoxyethyl ester of 2,4-D
decreases from one year at pH 5 to only 9 hours at pH 8. Mabey and Mill (1978)
report that the half-lives of the acid esters, ethyl acetate and methyl ben-
zoate, are 20 years and 38 days, respectively, at pH 7.0.
All forms of an iom'zable chemical are not hydrolyzed at the same rate.
Since pH determines speciation, pH could also indirectly affect hydrolysis and
give rise to rate vs. pH relationships that are complex. lonization effects
can be handled mathematically by using separate rate terms for each species and
calculating the true species concentration from knowledge of the acid-base
dissociation constant.
Rates of chemical redox reactions are frequently a function of hydrogen
ion concentration. Some are promoted while others inhibited at higher pH
values. Additionally, H+ and OH~ may be consumed or produced during a redox
reaction. Types of products formed could also be affected.
Oxygen
Oxygen (03) is probably the most important oxidant in soils. Atmospheric-
oxygen is capable of directly oxidizing many organic and inorganic compounds.
Chemical oxidation by atmospheric oxygen is usually slow and has been termed
autooxidation, a process that may involve trace quantities of free radicals
(March, 1977). Oxygen is not expected to affect hydrolysis reactions but could
be involved in photochemical processes occurring at the soil surface (photo-
oxygenations).
Soil Organic Matter Content
Soil organic matter sorbs many chemicals and would therefore be expected
to affect the rate of hydrolysis and redox reactions. Several components of
soil organic matter are known to affect hydrolysis. For example, basic amino
acids catalyze the hydrolysis of organophosphate esters (Crosby, 1976), and
fulvic acids catalyze the hydrolysis of atrazine in aqueous solution (Kahn,
1978). Perdue (1983) has pointed out that humic substances can modify hydrol-
ysis reaction rates by general acid/base catalysis and by partitioning equi-
librium that lead to a humic-bound substrate in solution with a reactivity
different from that of the unbound substrate. Both catalytic and retardation
effects are possible. The net effect on hydrolysis would depend on the cata-
lytic potential, the effect of sorption, and the formation of humic-bound
substrates.
The organic fraction of soil represents a potential storehouse of both
oxidizing and reducing agents. Steelink and Tollin (1967) have even shown that
free radicals are associated with soil organic matter although their exact
source and concentration remains unknown. Natural organic compounds have been
found to reduce a variety of inorganic species (Skogerboe and Wilson, 1981).
Reactions involving xenobiotics with natural organics could also be important
and should be further studied. The rate of oxidation or reduction of a spe-
cific chemical could be reduced due to competing reactions involving soil
organics.
F-10
-------
Organic matter may serve as a source of photosensitizers which effectively
catalyze photochemically mediated oxidations or as a source of quenchers which
inhibit such reactions. Humic material has been shown to act as a photosensi-
tizing agent in aqueous solution (Zepp et al., 1981). Hautala (1978) studied
the rates of photolysis of several pesticides in soils and suggested that the
slow rates observed were due to quenching of the photoexcited pesticide by soil
pigments.
The presence of organic matter also may affect oxygen concentration
through increased microbial activity and, therefore, indirectly may affect the
rate of oxidation or reduction of a chemical.
Soil Clay Content
Soil clay may sorb chemicals and provide acidic and basic sites that
catalyze hydrolysis. Because of interactions between water molecules and clay
minerals, water surface films may exhibit greatly changed pH values as the
water evaporates (Crosby, 1976) and thereby may affect hydrolysis rates. The
net effect of clay content would depend on the type of clay and chemical. The
effect of clays on redox reactions has not been studied.
Metal Oxides/Metal Ions and Inorganic Acids/Bases
Metals are ubiquitous in the soil environment. Several alkaline earth and
heavy metal ions catalyze hydrolysis. For example, Mortland and Raman (1967)
found that copper ions catalyzed the hydrolysis of organophosphates in clay
suspensions. However, little appears to be known about the importance of metal
catalysis in the environment. The concentrations of metal ions normally found
in water may be too low in concentration to significantly affect hydrolysis, or
they are complexed with organic substances in such a manner that they are not
effective catalysts (Mabey and Mill, 1978). Soil waters may also contain a
wide variety of weak inorganic acids and bases including carbonates, silicates,
phosphates, sulfides, and amines that could potentially affect hydrolysis by
acting as general acid or base catalysts (Perdue and Wolfe, 1983).
Metal ions can act as reactants in redox reactions. Fleck and Haller
(1945) demonstrated the conversion of DDT to DDE in the presence of FeCl3 and
A1C13. Glass (1972) investigated the reduction of DDT to DDD in soils and
proposed that the reaction involved transfer of electrons from reduced organic
matter to DDT via a Fe+2 - Fe+3 redox couple in the absence of oxygen. Stone
and Morgan (1984a,b) have shown that many organic compounds may be'oxidized by
manganese (III) and manganese (IV) oxides. Ferric salts have also been shown
to act as photosensitizing agents (Helling et al., 1971; Koster and. Asmur,
1973).
Temperature
The rates of all elementary chemical reactions increase as temperature
increases. However, since sorption may play a role in the overall rate of
reactions, the effect of temperature on sorption should also be considered.
F-ll
-------
Elementary rate constants are frequently correlated with temperature using
the Arrhenius expression
-EA/RT
k = A e (F-16)
where k is the rate constant, EA is the activation energy, R is the gas con-
stant, T is absolute temperature, and A is a pre-exponential factor. The
Arrhenius expression should be applied only to rate constants for individual
processes and not necessarily applied to the overall observed rate constant.
For example, only the individual acid, neutral, and base hydrolysis constants
(kH> kg, and kgn) should be correlated with temperature using the Arrhenius
equation and not by using the overall observed hydrolysis constant (Kn or
The Arrhenius expression can be linearized by taking the logarithm of both
sides of equation (F-16) to yield
log k = log A - EA/RT (F-17)
which provides a useful way to present temperature correlations graphically by-
plotting log k vs 1/T. Mabey and Mill (1978) suggested the correlation of
aqueous hydrolysis rate data using a form which more accurately describes the
effect of temperature
-A
log k = — + B log T + C (F-18)
T
where A, B, C are constants.
Increased temperature generally increases the rate of redox reactions.
However, since many redox reactions are not elementary and may involve compet-
ing reactions that consume oxidants, the exact effect cannot be stated a priori^
Free radical oxidations are not strongly temperature dependent. Temperature
may also affect the concentration of oxygen dissolved in soil water and there-
fore may affect reactions involving oxygen. The solubility of oxygen increases
with decreasing temperature.
Sunlight
Several characteristics of sunlight influence photochemical reactions
(Crosby, 1976; Zepp and Cline, 1977). The energy content of light depends on
its wavelength, and only photons having sufficient energy can lead to a reac-
tion. Additionally, not all wavelengths of light are absorbed by chemicals.
The rate of reaction may be a function of the rate at which photons impinge on
a surface (intensity). Both sunlight wavelength and intensity are a function
of time of day, season, and geographical location.
F-12
-------
ESTIMATION OF RATE PARAMETERS
Field vs. Laboratory Studies
Evaluation of the importance of nonbiological transformations under field
conditions is difficult because of concurrent biological processes leading to
chemical transformation and in some cases to products that may be identical to
those produced by a chemical reaction. Laboratory tests using sterile soil
samples are required if the importance of abiotic transformation pathways is to
be ascertained and if rate constants are to be estimated.
A minimum objective should be to determine if significant nonbiological
transformations are possible at all. For example, hydrolysis can be studied
in water with or without soil present. The pH can easily be modified by acid
or base addition. The relative stability of a chemical in the presence and
absence of oxygen will indicate the probable importance of oxygen in the field
environment. Oxygen concentration can be controlled by appropriate mixing of
nitrogen and oxygen (and C02 if needed). Redox potentials can be measured as
an indicator of the relative oxidation state of a soil (Bohn, 1971). Light/
dark experiments can be conducted to ascertain if photochemical transformations
can occur. Soil to which a chemical has been applied can be exposed to natural
or artificial light. General laboratory methods applicable to the study of
nonbiological transformations have been discussed by Howard et al. (1975).
Laboratory methods for the study of hydrolysis have been discussed by Mill et
al. (1982) and Wolfe et al. (1977, 1980). Several approaches to studying
photochemical reactions on soil have been used by various researchers (Howard
et al., 1975; Hautala, 1978; Parochetti and Dec, 1978; Plimmer, 1978a,b; Smith
et al., 1978; Burkhard and Guth, 1979).
Measurement at Test Chemical Disappearance
Ideally, not only should the concentration of test chemical be measured, ,
but also the identity and concentration of the products should be measured.
The rate of product formation should equal the rate of chemical transformation.
If hydrolysis is suspected, then the products formed should be characteristic
of this. The same is true for redox reactions. The possibility of multiple -
products also complicates interpretation of the results.
Measured transformation rates in the soil are based on total chemical
concentration, usually determined by solvent extraction. However, experimental
methodologies can be developed (discussed in the next section) which could be
used to determine the relative importance of the soil fraction in promoting or
inhibiting a reaction.
Nonbiological Transformation Rate Expressions and Constants
Goring et al. (1975), and Hamaker (1966, 1972) have discussed general
empirical rate expressions (a power rate law and a hyperbolic expression)
frequently used to describe various degradative processes occurring in the
soil (see discussion of biotransformation kinetics in appendix). However,
nonbiological transformation rate expressions in soil systems have been
expressed primarily as a first order rate relationship
F-13
-------
r = kC (F-19)
where k Is a first order rate constant, and C is the chemical concentration. In
a batch system, equation (F-19) may be integrated to yield an equation relating
concentration kC at time t to the initial concentration C0.
An C/C = -kt (F-20)
o
First order rate constants can be obtained from batch experiments by plotting
this integrated expression or from measurements of the test chemical half-life,
tl/2
k = .693/ti/2 (F-21)
Ou et al. (1982), and Rao and Davidson (1980) have provided an extensive
summary of first order rate constants and half-lives for chemicals in the soil
environment. However, much of the data was obtained in the presence of micro-
organisms and therefore should not be interpreted to include only chemical
transformations.
The rate of hydrolysis is expected to follow a first order relationship in
soils as previously discussed. Several hydrolysis "rate" constants can be
determined in soil systems. The utility of each would depend on the formula-
tion of the soil model being used. The simplest and most site specific constant
is the overall hydrolysis rate constant, knf, characterizing the first order
loss in total chemical concentration.
The pure aqueous phase constants k^, k^, k0, and kgn can be obtained from
batch laboratory experiments conducted in the absence of soil using buffered,
sterile water. However, Perdue and Wolfe (1983) have pointed out tha,t buffers
may accelerate the rate of hydrolysis. They have provided a model which can
be used as a guide in selecting appropriate buffers. The value of the first
order constant kn can be obtained in a manner similar to that of k^T- Using
three pH values, k^, k0, and kgx can be obtained from the simultaneous solution
of three linear equations each corresponding to equation (F-4). Constants
characterizing the effect of humic substances and inorganics can also be evalu-
ated (Perdue, 1983). Mabey and Mill (1978) have summarized rate constants for
a number of organic compounds in water. Perdue and Wolfe (1983) have presented
a model which can be used to estimate the potential catalytic effect of in-
organic acids and bases. Perdue (1983) has presented a model allowing the
estimation of the potential effect of humic material on the rate of hydrolysis.
The phase specific overall hydrolysis constants knw and k^s can (at least
in principle) be obtained from experiments at a fixed pH utilizing different
soil-to-water ratios (M values) if the partition coefficient (Kp) is known.
Two simultaneous linear equations derived from equation (F-9) using measured
khi values and calculated fi and f2 values can be solved to yield estimates of
kjjs and knw. The pH dependency could be further investigated as previously
discussed.
Chemical redox reactions have been studied primarily in model systems in
the laboratory where oxygen concentration and other parameters can be controlled
F-14
-------
and evaluated. Most laboratory studies utilize conditions greatly different
from those in the environment. Observations are generally made regarding
persistence.
First order rate constants can be extracted from the data even if the data
indicates that a more complex relationship exists. First order constants are
expected to be particularly site specific since the concentration of oxidant or
reductant is not explicitly accounted for. A second order relationship could
possibly be developed if the oxidant/reductant in the soil can be identified
and their concentration measured. However, at present this is beyond the
sophistication of any available soil model.
Photochemical transformation studies which encompass redox reactions have
been conducted both under both field and laboratory conditions. Field studies
have focused on the determination of sunlight effects on relative persistence.
No effort is usually made to quantify rates in terms of any model. Laboratory
studies have likewise not been aimed at modeling but have been aimed primarily
at determination of relative persistence and product formation.
LIMITATIONS
Many of the limitations encountered in the measurement and use of bio-
transformation constants also apply to those characterizing nonbiological
reactions.
Multiple Processes and Pathways
Physical phenomena such as volatilization, other chemical processes, and
biotransformation can simultaneously occur causing an overestimation of rate
constants. The use of soil sterilization, while eliminating biotic factors,
may affect some property that affects a chemical reaction. For example,
Konrad and Chesters (1969) observed a decrease in the rate of hydrolysis of an
organophosphate insecticide after electron-beam irradiation was used to steri-;.
lize the soil. This decrease was attributed to changes in the adsorption of
the insecticide, not from the retardation of microbial processes.
The inability to resolve nonbiological transformations into their compo-
nent reactions would lead to the determination of rate "constants" which are
especially empirical. Additionally, several independent reaction pathways
each producing different products could exist giving rise to a complex rate
dependence.
Bound Chemical Residues
Inability to measure all of the applied chemical could result in a calcu-
lated rate constant too high if it is based on loss of a parent compound that
forms a bound residual or could result in a low estimate of a rate constant if
it is based on formation of a known product that forms a bound residue.
F-15
-------
Specificity of Constants
As nearly as possible, constants characterizing transformation rates
should be used under those soil conditions and concentration ranges at which
they were obtained since so many unknown factors can affect chemical reactions.
Unfortunately, laboratory studies frequently are conducted under conditions
very different from those encountered in the field.
F-16
-------
REFERENCES
Baughman, G. L., and R. R. Lassiter. 1978. Prediction of Environmental
Pollutant Concentration. Special technical publication 657, American
Society for Testing and Materials, Philadelphia, PA.
Bell, R. P. 1941. Acid-Base Catalysis. Oxford at the Clarendon Press.
Bohn, H. L. 1971. Redox potentials. Soil Sci. 112:39-45.
Burkhard, N., and J. A. Guth. 1979. Photolysis of organophosphorus insec-
ticides on soil surfaces. Pest. Sci. 10:313-319.
Burkhard, N., and J. A. Guth. 1981. Chemical hydrolysis of 2-chloro-4,6-
bis(alkylamino)-l,3,5-triazine herbicides and their breakdown in soil
under the influence of adsorption. Pestic. Sci. 12:45-52.
Calvert, J. G. and J. N. Pitts. 1966. Photochemistry. John Wiley and Sons,
Inc.
Crosby, D. G., and Leitis. 1973. The photodecomposition of triafluralin in
water. Bull. Environ. Contarn. Toxic. 10:237.
Crosby, D. G. 1976. Nonbiological degradation of herbicides in the soil.
In: L. J. Audus (Ed.), Herbicides: Physiology, Biochemistry, Ecology,
pp 65-94, Academic Press, New York.
Fleck, E. E., and H. L. Haller. 1945. Compatibility of DDT with insecticides,
fungicides and fertilizers. Ind. Eng. Chem. 37:403-405.
Glass, B. L. 1972. Relation between degradation of DDT and the iron redox
system in soils. «L Agr. Food Chem. 20:324-327.
Goring, C. A. I., D. A. Laskowski, J. Hamaker, and R. W. Miekle. 1975. Prin-
ciples of pesticide degradation in soil. In; R. Hague and V. H. Freed
(Eds.), Environmental Dynamics of Pesticides. Plenum Press.
Hamaker, J. W. 1966. Mathematical predictions of cumulative levels of pesti-
cides in soil. Adv. Chem. Series. 60:122.
Hamaker, J. W. 1972. Decomposition: quantitative aspects. In; C. A. I.
Goring and J. W. Hamaker (Eds.), Organic Chemicals in the Soil Environ-
ment. Marcel-Dekker, New York.
F-17
-------
Harris, J. C. 1982. Rate of hydrolysis. Irr. W. J. Lyman, W. F. Reehl,
D. H. Rosenblatt (Eds.), Handbook of Chemical Property Estimation Methods:
Environmental Behavior of Organi'c' Compounds. McGraw-Hill Book Co.,
New York.
Hautala, R. R. 1978. Surfactant Effects on Pesticide Photochemistry in Water
and Soil. EPA-600/3-78-060.
Helling, C. J., P. C. Kearney, and M. Alexander. 1971. Behavior of pesticides
in soil. Adv. Agron. 23:147-240.
Howard, P. H., J. Saxena, P. R. Durkin, and L. T. Ou. 1975. Review and Eval-
uation of Available Techniques for Determining Persistence and Routes of
Degradation of Chemical Substances in the Environment. EPA-560/5-75-006.
Ivie, G. W., and J. E. Casida. 1971. Sensitized photodecomposition and photo-
sensitizer activity of pesticide chemicals exposed to sunlight on silica
gel chromatoplates. vh Agr. Food Chem. 19:405-409.
Khan, S. U. 1978. Kinetics of hydrolysis of atrazine in aqueous fulvic acid -
solution. Pest. Sci. 9:39-43.
Konrad, J. G., and G. Chesters. 1969. Degradation in soils of ciodrin, an
organophosphate insecticide. 0_. Agr. Food Chem. 17:226-230.
Koster, R., and K. D. Asmur. 1973. Reactions of fluorirated benzenes with
hydrated electrons and hydroxyl radicals in aqueous solution. J_. Phys.
Chem. 77:749-755.
Lichtenstein, E. P., K. R. Shulz, T. W. Fuhremann, and T. T. Liang. 1970.
Degradation of aldrin and heptachlor in field soils during a ten-year
period. Translocation into crops. ±. Agr. Food Chem. 18:100-106.
Lowry, T. H., and K. S. Richardson. 1976. Mechanism and Theory in Organic
Chemistry. Harper and Row, New York.
Mabey, W., and T. Mill. 1978. Critical review of hydrolysis of organic com-
pounds in water under environmental condition. J^. Phys. Chem. Ref. Data.
7:383-415.
March, J. 1977. Advanced Organic Chemistry. 2nd ed. McGraw-Hill Book Co.,
New York.
Mill, T., D. G. Hendry, and H. Richardson. 1980. Free-radical oxidants in
natural waters. Science 207:886-887.
Mill, T., W. R. Mabey, D. C. Bomberger, T. W. Chou, D. C. Hendry, and J. H.
Smith. 1982. Laboratory Protocols for Evaluating the Fate of Organic
Chemicals in Air and Water. EPA-600/3-82-022.
Miller, G. C., and R. G. Zepp. 1979. Photoreactivity of aquatic pollutants
sorbed on suspended sediments. Environ. Sci. Technol. 13:860-863.
F-18
-------
t)
Miller, S. 1983. Photochemistry and natural water systems. Environ. Sci.
Techno!. 17:568A-570A.
Mortland, M. M., and K. V. Raman. 1967. Catalytic hydrolysis of some organic
phospnate pesticides by copper (II). J_. Agr. Food Chem. 15:163-167.
Ou, L. T., J. M. Davidson, P. S. C. Rao, and W. B. Wheeler. 1982. Pesticide
transformations in soils. In: P. S. C. Rao and J. M. Davidson (Eds.),
Retention and Transformation of_ Selected Pesticides and Phosphorous jm
Soil-Mater Systems: £ Critical Review. EPA-600/3-82-060.
Parochetti, J. V., and 6. W. Dec., Jr. 1978. Photodecomposition of eleven
dinitroaniline herbicides. Weed Sci. 26:153-156.
Perdue, E. M. 1983. Association of organic pollutants with humic substances:
partitioning equilibria and hydrolysis effects in aquatic and terrestrial
humic materials. R. F. Christman and E. G. Gjessing (Eds.), Ann Arbor
Science, Ann Arbor, MI.
Perdue, E. M. and N. L. Wolfe. 1983. Prediction of buffer catalysis in field_
and laboratory studies of pollutant hydrolysis reactions. Environ. Sci.
Techno!. 17:635-642.
Plimmer, J. R. 1978a. Photolysis of TCDD and trifluralin on silica and soil.
Bull. Environ. Contam. Toxic. 20:87-92.
Plimmer, J. R. 1978b. Degradation methodology: chemical-physical effects.
Irr. A. W. Bourquin and P. H. Pritchard (Eds.), Proceedings of the Work-
shop: Microbial Degradation of Pollutants in. Marine Environments.
EPA-600/9-79-012.
Rao, P. S. C., and J. M. Davidson. 1980. Estimation of pesticide retention
and transformation parameters required in non-point source pollution
models. In: M. R. Overcash and J. M. Davidson (Eds.), Environmental
Impact of Nonpoint Source Pollution. Ann Arbor Science Publishers, Ann
Arbor, MI.
Rosen, J. D. 1972. The photochemistry of several pesticides. In:.
F. Matsumura (Ed.), Environmental Toxicology of Pesticides. Academic
Press, New York.
Saltzman, S., U. Mingelgrin, and B. Yaron. 1976. Role of Water in the
Hydrolysis of Parathion and Methylparathion on Kaolinite. J_. Agric. Food
Chem. 24:739-743.
Sawyer, C. N., and P. L. McCarty. 1978. Chemistry for Environmental Engineers,
3rd ed., McGraw-Hill.
Skogerboe, R., and S. A. Wilson. 1981. Reduction of ionic species by fuluric
acid. Anal. Chem. 53:228-232.
F-19
-------
Smith, 0. H., W. R. Mabey, N. Bohonos, B. R. Holt, S. S. Lee, T. W. Chou, D. C.
Bomberger, and T. Mill. 1977. Environmental pathways of selected chem-
icals in freshwater systems. Part I: background and experimental pro-
cedures. EPA-600/7-77-113.
Smith, C. A., Y. Iwata, and F. A. Gunther. 1978. Conversion and disappearance
of methidathion on thin layers of dry soil. J. Agric. Fd. Chem.
26:959-962. ~
Snoeyink, V. L., and P. Jenkins. 1980. Water Chemistry. John Wiley and Sons.
Steelink, C., and 6. Tollin. 1967. Free radicals in soil. Jjr A. D.
McLarenand, G. H. Peterson (Eds.), Soil Biochemistry. Marcel-Dekker, Inc.,
New York.
Stone, A. T., and J. J. Morgan. 1984a. Reduction and dissolution of manganese
(III) and manganese (IV) oxides by organics. 1. Reactions with hydro-
quione. Environ. Sci. Technol. 18:450-456.
Stone, A. T., and J. J. Morgan. 1984b. Reduction and dissolution of manganese
(III) and manganese (IV) oxides by organics. 2. Survey of the reactivity
of organics. Environ. Sci. Technol. 18:617-624.
Stumm, W., and J. J. Morgan. 1981. Aquatic Chemistry. 2nd ed., John Wiley
and Sons, New York.
Turro, N. J. 1965. Molecular Photochemistry. W. A. Benjamin, Amsterdam, NY.
Wolfe, N. L., R. G. Zepp, J. A. Gordon, G. L. Baughman, and D. M. Cline. 1977.
Kinetics of chemical degradation of malathion in water. Environ. Sci.
Technol. 11:88-93.
Wolfe, N. L., W. C. Steen, and L. A. Burns. 1980. Phthalate ester hydrolysis:
linear free energy relationships. Chemosphere 9:403-408.
Wolfe, C. J. M., M. T. H. Halmans, and H. V. Vander Heijde. 1981. The forma-.
tion of singlet oxygen in surface waters. Chemosphere 10:59-62.
Zepp, R. G., N. L. Wolfe, G. L. Baughman, and J. A. Gordon. 1974. Dynamics
of 2-4,0 esters in the aquatic environment: hydrolysis and photodegrada-
tion. 167th ACS Meeting, Los Angeles, CA.
Zepp, R. G., and D. M. Cline. 1977. Rates of direct photolysis in aquatic
environments. Environ. Sci. Technol 11:359-366.
Zepp, R. G., N. L. Wolfe, G. L. Baughman, and R. C. Hollis. 1977. Singlet
oxygen in natural waters. Nature 267:421-423.
Zepp, R. G., and G. L. Baughman. 1978. Prediction of photochemical transfor-
mation of pollutants in the aquatic environment. In: 0. Hutzinger, I. H.
Van Lelyveld, B. C. J. Zoetman (Eds.), Aquatic Pollutants: Transforma-
tions and Biological Effects. Pergamon Press, New York.
F-20
-------
Zepp, R. G., G. L. Baughman, and P. F. Schlotzhaver. 1981. Comparison of
photochemical behavior of various humic substances in water. II.
photosensitized oxygenations. Chemosphere 10:119-126.
Zifiriou, 0. C. 1977. Marine organic photochemistry previewed. Marine
Chemistry 5:497-522.
F-21
-------
APPENDIX 6
SPATIAL VARIABILITY OF SOIL PROPERTIES
by
W. A. Jury
INTRODUCTION
Soils vary significantly from point to point in their structural proper-
ties, textural composition, and mineralogical constituents. As a result,
virtually all of the parameters characterizing the transport processes discus-"
sed in this report vary both laterally and vertically in an undisturbed soil
profile of field size. Consequently, in the field setting, the task of model-
ing becomes considerably more complex than for laboratory scale processes.
However, even though the transport processes to be characterized will inevit-
ably display lateral and vertical variation, it is not reasonable to expect
that threedimensional models capable in principle of describing this point-to-
point variability could be calibrated by any conceivable combination of measure-
ments. As a result, field-scale models, at least for processes in which large
surface areas are treated relatively uniformly (i.e., agricultural operations,
large waste holding ponds), will have to be one dimensional. Thus., replicated
measurements of representative parameters for these one-dimensional models
where large field areas are involved must be substantial enough so that their
mean values give a representative average over the field. Furthermore, the
uncertainty of the measurement, represented by the sample variance or coef-
ficient of variation, should be recorded as well as the mean value to indicate
the precision of the mean estimate and to act as an indication of the extent to
which the assumption of one dimensionality is valid for the field in question.
A consequence of the transition from laboratory simulation and experimen-
tation to variable natural environments is that the scale of the averaging
process becomes very important. Simulation of processes occurring in soil tank
lysimeters, for example, will be more difficult and less exact than represen-
tation of transport through laboratory columns but will be considerably less
difficult than representation of the average downward movement from a large
field.
There are a number of implications of field variability and the admitted
futility of constructing a three-dimensional chemical transport model. Most
significantly, one must abandon any hope of attempting to estimate a continuous
spatial pattern of chemical emissions at each point in space and must rely
instead on prediction of characteristic flow indices (i.e., mean solute concen-
tration at depth z) which represent averages across the entire cross-sectional
G-l
-------
area of the field. A second significant implication is the possibility for
extreme deviations from average movement to occur over a relatively small
fraction of the total cross-sectional area. Thus, there may be plumes of
solute moving downward through cracks, crevices, or other natural structural
voids which could carry potentially harmful pollutants to great depths and yet
would be difficult to detect by a point measurement of concentration.
Dilemmas posed by the extent of observed variability have caused research-
ers recently to propose that transport be modeled by stochastic as opposed to
deterministic hypotheses. In a stochastic model, prediction of concentrations
at depths as a function of time are interpreted as an index of the relative
probability of finding chemical at the given depth (Jury, 1982). This stochas-
tic philosophy may be better suited to the task of assessing extreme movement
of hazardous chemicals than the deterministic models would be.
Characterization of the spatial variability of a parameter is not com-
pletely straightforward. The traditional approach is to assume that all
replicated measurements of the property are statistically independent and to
represent the property variation by specifying its sample mean and sample
variance or coefficient of variation. However, recent investigations have
shown that lateral correlations exist between transport properties measured
near to each other (Warrick and Nielsen, 1980; Peck, 1984; Vieira et al.,
1983), and that this statistical dependence should be taken into account when
calculating sampling strategy (McBratney and Webster, 1983) or when interpo-
lating between measurements (Journel and Huijbregts, 1978).
Two distinct philosophies are currently in evidence in the ongoing re-
search dealing with spatial variability. The first approach, called geometric
scaling theory (Peck et al., 1977; Warrick et al., 1977; Russo and Bresler,
1980), uses certain standardized variables to scale the differential equations
describing transport and relates the standardized variables to some measurable
or definable property of each local site of a heterogeneous field. When these
variables have been identified, the task of characterizing variability is
reduced to calculating the statistical and spatial distribution of these scal-
ing parameters. The only current application of scaling theory used so far is
the so-called geometric similitude scaling (Miller and Miller, 1956) which
assumes magnified scale heterogeneity in which each individual soil location is
regarded as a magnified version of a reference location which may be character-
ized by a single microscopic length parameter. This length parameter has a
predictable influence on transport properties such as hydraulic conductivity
and matric potential. Thus, the statistical distribution of the scaling fac-
tors combined with a single set of reference transport coefficient functions
allows one to make repeated calculations of flow properties from a single
computation at the reference site and also to simulate transport processes for
al_l locations at once by solving the scaled transport equations (Warrick and
Amoozegar Fard, 1979).
The second hypothesis for treating spatial variability is to regard the
various parameters relevant to a field-wide description of transport as random
variables characterized by a mean value and a randomly fluctuating stochastic
component. A sampling at a point thus reveals a single momentary snapshot of
these fluctuating properties which may be analyzed by statistical techniques
G-2
-------
designed to detect spatial correlations In order to yield Information about the
spatial distribution of the statistical fluctuations. The stochastic solute
transport models which treat solute concentration as a random variable then
use the global statistical properties calculated from this analysis to predict
large-scale macrodispersion and macrovelocity parameters which are used in the
asymptotic (long time) description of solute transport (Gelhar et al., 1979;
Gelhar and Axness, 1983).
The discussion to follow will begin with a description of the known experi-
mental information characterizing the extent of variability of key soil water
and solute parameters. Next, the experimental evidence in support of the two
key statistical models, scaling theory and regionalized variable analysis, will
be discussed and summarized. Finally, the implications of spatial variability
on parameter measurement will be analyzed using both the assumption of statis-
tical independence and the assumption of statistical dependence, f.uch of the
tabular material in this chapter is adapted from Jury (1985).
SUMMARY OF OBSERVED VARIABILITY
Study 1 - Soil Matrix and Water Retention Properties
Many of the parameters needed for transport models have properties which
are dominated by the bulk characteristics of the solid matrix of the soil. As
such, their lateral variability is relatively small, reflecting the uniformity
of the soil formation processes.
Table 6-1 summarizes the available information from field stucies of these
so-called matrix properties which is summarized by giving the field mean and
sample coefficient of variation for various parameters along with the soil
texture and field size where the measurements were made. Also included is the
method of measurement used in the study, the number of replicate measurements
used in the averaging process, and the literature citation.
For the most part, properties in this group show low to moderate variation
among replicates irrespective of field size or soil type. Also, there are no
obvious relationships between parameter value and soil texture or between
parameter variability and soil texture. The limited evidence presented here
suggests that one could associate an average coefficient of variation (e.g.,
10 percent for porosity) with a given parameter irrespective of the particular
field properties.
Study 2 - Mater Transport Parameters
Table G-2 summarizes observed variability in field experiments for water
transport parameters, including saturated hydraulic conductivity, infiltration
rate, and hydraulic conductivity-water content or hydraulic conductivity-matric
potential functions. In contrast to the static soil properties covered in
Table G-l, this group of properties is characterized by variability of the
order of 100 percent or greater in many cases and by significant differences in
the extent of variability of a given property between different investigations.
For example, saturated hydraulic conductivity coefficient of variation ranged
G-3
-------
TABLE G-l. FIELD STUDIES OF SOIL MATRIX AND WATER RETENTION PROPERTIES
CD
Parameter
Porosity
Porosity
Porosity
Porosity
Bulk Density
Bulk density
Bulk density
Bulk density
Bulk density
Bulk density
Bulk density
Bulk density
Bulk density
% sand/% clay
% sand/% clay
% sand/% clay
% sand/% clay
% sand/% clay
:=========
Mean
0.45
0.37
0.53
0.42
1.36
1.30
1.20
1.47
1.26
1.47
1.65
1.59
1.20
24/45
17/32
59/12
83/9
65/28
CV
11
11
7
10
7
7
26
9
6
6
3
6
15
15/33
32/16
37/53
3/34
8/18
Field
Size
(ha)
150
0.8
.03
0.4
150
15
3.8
1.3
0.5
0.5
0.34
91.6
40
150
85
15
91.4
0.28
:==================
l»
Soil F
Texture c
clay loam
sand
clay loam
loamy sand
clay loam
sandy loam
sandy loam
sandy loam
silty clay
silt loam
sand
sand
clay loam
clay
silty clay loam
sandy loam
loamy sand
sandy clay loam
lumber
of
tepli-
:ates
120
120
20
12
120
64
30
192
144
72
5
5
36
480
100
64
5
35
Method
of
Measurement
Water content at
zero suction
Not given
Water content at
zero suction
Water content at
zero suction
Undisturbed cores
Not given
Measure volume of
plastic-lined
hole
Undisturbed cores
Undisturbed cores
Undisturbed cores
Undisturbed cores
Undisturbed cores
Undisturbed cores
Hydrometer
Light scattering
Not given
Not given
Not given
Reference
Nielsen et al .,
1973
Russo and Bresler,
1981
Cameron, 1978
Cassel , 1983
Nielsen et al . ,
1973
Gumaa, 1978
Court in et al . ,
1983
Cassel and Bauer,
1975
Cassel and Bauer,
1975
Cassel and Bauer,
1975
Babalola, 1978
Babalola, 1978
Stockton and
Warrick, 1971
Nielsen et al . ,
1973
Gajem et al . , 1981
Gumaa, 1978
Babalola, 1978
Vauclin et al . ,
1983
(continued)
-------
sssssssszzszss
Parameter
.1 bar water
content
.09 bar water
content
.1 bar water
content (Og)
.1 bar water
content
15 bar water
content
15 bar water
content
cp 15 bar water
tn content
15 bar water
content
15 bar water
content (Og)
pH
pH
pH
pH ,
KD (cm /9)
SSSSSSSSS!
Mean
.37
.37
.27
.45
.166
.041
.193
.074
.095
6.1
6.4
5.8
8.2
2.01
CV
(%)
4
17.6
20
15
14.4
45
14
19
33
15
7
9
2
31
zzszzsz»
Field
Size
(ha)
85
150
15
40
85
1.3
0.5
3.3
15
.04
.02
.04
85
0.64
TABLE 6-1.
Soil
Texture
clay loam
clay loam
sandy loam
clay loam
clay loam
sandy loam
silty clay
silt laom
sandy loam
clay loam
loam
sandy loam
clay loam
loamy sand
(continued)
zzt=szs=::sszss~ =
Number
of
Repli-
cates
100
120
64
36
900
172
144
192
64
1,040
640
208
100
36
Method
of
Measurement
Hanging water
table
Pressure plate
Pressure plate
Pressure plate
Pressure plate
Pressure plate
Pressure plate
Pressure plate
Pressure plate
Pressure plate
Batch equlibrium
Reference
Gajem et al . ,
1981
Nielsen et al .,
1973
Gumaa, 1978
Stockton and
Warrick, 1971
Gajem et al . , 1981
Cassel and Bauer,
1975
Cassel and Bauer,
1975
Cassel and Bauer,
1975
Gumaa, 1978
Cameron et al . ,
1971
Cameron et al . ,
1971
Cameron et al . ,
1971
Gajem et al., 1981
El Abd, 1984
-------
TABLE 6-2. FIELD STUDIES OF WATER TRANSPORT PROPERTIES
=============
Parameter
(cm d'1)
Saturated K
Saturated K
Saturated K
Saturated K
Saturated K
Saturated K
Saturated K
Saturated K
Saturated K
Saturated K
Saturated K
Saturated K
Saturated K
Saturated K
Infiltration
rate
=======
Mean
20.6
168
316
84
3.6
18.9
11.0
6.9
28.1
55.6
71.2
98.5
24.1
203
14.6
=======
CV
120
190
69
69
48
103
118
92
320
118
105
81
178
50
94
Soil
Texture
Clay loam
Sandy loam
Sand
Loamy sand
Silty clay
loam
Coarse
Fine
Silty clay
Very coarse
Coarse
Loamy sand
Loamy sand
Sandy laom
Silt loam
Clay loam
Number Method
Field of of
Size Measure- Measure-
(ha) ments ment
150
15
0.8
0.4
Composite
SCS date
for a given
soil series
in Imperial
Valley, CA
91.6
91.6
9.6
150
120
64
90
12
33
330
287
339
36
352
121
5
5
26
20
Steady infiltra-
tion 20 plots x
6 depths lab
permeameter Jji
Situ air-entry
permeameter lab
permeameter
Lab permeameter
Lab permeameter
Lab permeameter
Lab permeameter
Lab permeameter
Lab permeameter
Lab permeameter
Lab permeameter
Lab permeameter
0-30 cm infiltration
(double ring)
Steady state
Reference
Nielsen et al .,
1973
Gumaa, 1978
Russo and Bresler,
1981
Cassel , 1983
Willardson and
Hurst, 1965
Willardson and
Hurst, 1965
Willardson and
Hurst, 1965
Willardson and
Hurst, 1965
Willardson and
Hurst, 1965
Willardson and
Hurst, 1965
Willardson and
Hurst, 1965
Babalola, 1978
Babalola, 1978
Sharma et al . , 1980
Nielsen et al . ,
1973
(continued)
-------
TABLE G-2. (continued)
Parameter
(cm d-1)
Infiltration
rate
Infiltration
rate
Infiltration
rate
Infiltration
rate
Infiltration
rate
<•> Infiltration
-4 rate
Unsaturated
K(0)
K(0)=K0exp
KQ ~
B
KO
B
KO
6
KO
B
Mean
16.8
6.6
8.5
8.5
47
263
22.5
14.6
4.6
89.1
9.6
65.4
4.0
32.9
CV
40
71
56
23
79
97
343
64
235
41
76
37
46
19
Field
Soil Size
Texture (ha)
Loam 0.9
Silty clay .004
loam
Silty clay .004
Silty clay .004
loam
7 series 100 ha
7 series 100 ha
Clay loam 150
Clay loam
Silt laom
Loam .66
Number
of
Measure-
ments
1,280
625
125
25
20
15
20
20
611
24
Method
of
Measure-
ment
Steady infiltration
(double ring)
Adjacent infiltro-
meters along
transect
Double ring infiltro-
meter
Inverse auger hole
method 150 cm
Instantaneous profile
method
Unit gradient method
Unit gradient method
Instantaneous profile
method 4 plots x 6
Reference
Vieira et al . ,
Sisson and
Wierenga, 198
Duffy et al .,
Duffy et al . ,
Libardi et al .
1980
Libardi et al .
1980
Wagenet and Ra
1983
Simmons et al .
1979
1981
1
1981
1981
9
9
0,
,
depths
-------
from a low of 48 percent to a high of 320 percent for different Soil Conserva-
tion Service investigations in the Imperial Valley of California (Willardson
and Hurst, 1965).
Infiltration rate in most of the studies varied less than saturated
hydraulic conductivity, and this is owing partly to the fact that most inves-
tigations measured infiltration rate over the early stage of infiltration when
matric potential gradients as well as gravity were assisting in the flow. This
may have acted as a homogenizing influence. The final group of properties in
Table G-2, the parameters for the unsaturated hydraulic conductivity-water
content functions and hydraulic conductivity-matric potential functions, showed
enormous variability in the investigations summarized. The large variation in
these functional parameters should act as a warning against using a single
function to represent a large field area. As a general rule, the individual
transport parameters measured at a particular location apply only to a small
region around that location. Thus, predictions of models using measured values
of the parameters also apply only to the vicinity of the calibration.
Study 3 - Solute Concentration Variations
Variability of solute concentration among replicates is critically impor-
tant for the problem of calibrating and validating solute transport models.
Table G-3 shows summaries of the extent of concentration variation among rep-
licates observed in 10 field experiments involving either analysis by soil
solution samplers or by soil coring. The first four studies refer to solute
transport experiments in which pulses of chemical were applied to the surface.
The coefficient of variation reported refers to variation around the peak
height maximum which contains the most critical information for model valida-
tion. These coefficients of variation lie between approximately 60 and 130
percent; in the absence of more comprehensive information these coefficients
could be used as an index to predict anticipated variation in future studies
as a means of designing monitoring networks. The other studies in the table
relate more to characterization of salinity and involve either assessment of
native conditions (Studies 5, 6, 8) or averaging over entire profiles for
assessment of nutrient loading below irrigated agricultural fields (studies 7,
9, 10). Because they represent native conditions or are the result of depth
averaging their coefficient of variation is not strictly comparable to that of
the transport experiments (studies 1-4).
Study 4 - Solute Velocity Variations
Since mass flow is the principal mechanism governing solute movement, the
key parameter to measure or assess under field conditions is the apparent solute
velocity. Not surprisingly, this parameter has been observed to vary substan-
tially when large area experiments are conducted. Table G-4 summarizes six
studies in which the statistical distribution of this property or of related
properties was obtained. The table summarizes either solute velocity, net
applied water required to reach a given depth, or the depth of peak height
solute concentration in a set of replicated soil cores. Although these param-
eters are not strictly comparable, it may be shown that when log transformed
and fitted to a log-normal distribution, their log variance s2 should be very
similar for a given study (Jury, 1983). Thus, values of s^ given in the table
G-8
-------
TABLE G-3. FIELD STUDIES OF CHEMICAL CONCENTRATIONS
Chemical
1. Chloride
2. Bromide
3. Bromide
-------
TABLE G-3. (continued)
Chemical
8. Chloride
===========
Origin
of
Chemical
Native
Measure-
ment
Depth (m)
Surface
to bedrock
x20 cm
Soil
Texture
Various
Number
Field of
Size Repli-
(ha) cates
100-200
Method
of
Measure-
CV ment
12-70 Summary of 9
catchment areas
Average concen-
tration in core
Reference
Johnston et
al., 1980
9. Chloride
10. Chloride
Manure
fertilized
fields
Irrigated-
fertilized
fields
1.5-6.3
1.5-6.3
1.5-6.3
.6-1.2
.9-1.5
.6-1.2
Sandy loam
Sandy loam
Clay loam
Loam
Fine sandy loam
Fine sandy loam
81
14
14
—
8
2
19
19
66
60
102
87
Average concentra- Pratt et al.,
tion in entire 1972
vertical profile
Soil samples
Soil samples
Soil samples
Oster and
Wood, 1977
-------
TABLE 6-4. LOG DISTRIBUTION OF SOLUTE TRANSPORT VELOCITY PARAMETERS IN FIELD EXPERIMENTS
=================================================================================================:
Water
Parameter Soil Area Application Measurements N 2 Reference
1. Velocity loamy sand 4.6 x 6.1m2 ponding solution samples 44 .48 Starr et al.,
(4 plots) to 240 cm 1978
2. Velocity clay
8m x 8m tricklers solution samples 24 .32
to 150 cm
Van de Pol et al.,
1977
3. Velocity loam
4. Net applied sand
water
5. Peak height sand
150 ponding solution samples 120 1.56 Biggar and
(20 plots) to 180 cm Nielsen, 1976
.64 ha rainfall solution samples 70 .32 Jury et al., 1982
(14 sites) to 180 cm
.64 ha sprinkler soil cores
(36 sites) to 300 cm
36 .12 Jury et al., 1983
6. Peak height loamy sand 3m x 3m sprinkler soil cores
(8 plots) to 80 cm
32 .45 Wild and Babiker,
1976
-------
are equivalent whether velocity, net applied water, or depth of leaching are
considered, and these values can be compared in different studies. It is
significant that studies 4 and 5 (Jury et al., 1982; 1983, respectively)
observed considerably different values of a2 and were conducted on the same
field. The larger value of a2 found in the solution sampler study 4 could
have been caused partly by transient conditions resulting from erratic rainfall
(Jury et al., 1982) which can cause phase lags between net applied water and
drainage rate at the solution sampler depth.
IMPLICATIONS OF SPATIAL VARIABILITY ON THE MEASUREMENT PROCESS
Table G-5 gives sample sizes required in order to have at least a 95
percent probability of detecting a relative change of 20, 40, and 100 percent
in the value of the mean when using a one-sample two-tailed t-test with prob-
ability of Type 1 error set at a=5 percent. These values were obtained using
an approximation formula (Guenther, 1981) which states that to have probability
(1-0) of detecting a shift of F percent in the value of the mean using a one-
sample two-tailed t-test with probability of Type 1 error set at a, one needs
N = {(Z0/2- + ZB)2(CV)2/F2} + 0.5(Za/2)2 (G-l)
where Zg is the upper 1000 percentile of the standard normal distribution and
Za/9 is similarly defined. The value of N given by the formula is not usually
an integer, and to obtain the sample size one should round up to the next
higher integer. As an example, consider the case where the CV is 10 percent
and one wants to have a 95 percent chance of detecting a 20 percent change in
the mean using a t-test with a=5 percent, then
N = (1.96 + 1.645)2(10)2/(20)2 + 0.5(1.96)2 = 5.17 (6-2)
Hence one would want to use at least 6 measurements in this case. (In using
this approximation formula one always rounds up to the next highest integer).
Table G-5 presents calculations of sample numbers using equation (G-l) •'
required to estimate parameter mean values within F = .5, .2, or .1 of the true
population mean at 5 percent precision using coefficient of variation values
averaged over all the studies summarized above. These numbers could be thought
of as rough guidelines for sampling information content based on the experi-
mental information available.
The large sample numbers required for transport properties such as satur-
ated hydraulic conductivity seem at first glance to pose an almost insurmount-
able barrier for field experimentation. However, it should be kept in mind
that the arithmetic average hydraulic conductivity for an entire field is not
necessarily a useful parameter. It may be more useful to obtain an estimate of
the variation in hydraulic conductivity and use this information to construct a
crude sample frequency diagram. Second, it may not always be necessary to
construct highly accurate representations of the field average particularly if
best- and worst-case scenarios are used. In this case, one might make a lim-
ited number of measurements and then run sensitivity analyses for ±100 percent
variation of the parameters; this would seem reasonable given the experimental
information summarized on the coefficient of variability of this parameter.
6-12
-------
TABLE G-5. SAMPLE SIZES REQUIRED TO HAVE A 95% PROBABILITY OF DETECTING
A CHANGE OF F% IN THE MEAN USING A T-TEST WITH a=5%
(Sample CVS are mean of all field studies.)
Parameter
Bulk density or
porosity
Percent sand or
clay
0.1 bar water
content
15 bar water
Number of
Studies
13
10
4
5
F: 20%
40% 100% Average
Number of Samples CV±SD
6
28
9
23
* * 10 ±
9 * 28 ±
* * 14 ±
7 * 25 ±
6
18
7
14
content
PH
Saturated K
Infiltration
rate
K0 in K(9)
Ponded solute
velocity
Unsaturated
solute velocity
4
13
8
4
1
5
4
4
502
135
997
1,225
127
119-551
127
36
251
308
33
32-140
Solute concen-
tration
*Sample size estimates are less than 5 and should not be used.
22
8
42
51
7
7-24
8 ± 5
124 ± 71
64 ± 26
175 ± 139
194
62 ± 9
60-130
(range)
Finally, significant research is being conducted at this moment trying to
connect the extensive variability information available through sources such as
the U.S. Soil Conservation Survey and values of parameters such as those des-
cribed above. Several promising efforts in this regard have already been ac-
complished. Duffy et al. (1981), used Soil Survey descriptions to predict
values for hydraulic conductivity within well-described mapping units. Compar-
ison of these predictions with measurements inside these mapping units produced
a correlation of rz = .93. Alexander (1980) reviewed 720 measurements of wet
bulk density at 1/3 bar applied suction which were measured on alluvial and
upland soils covering 7 soil orders. The measured mean values were tested
against multiple regression on a large number of soil properties available
G-13
-------
through Soil Survey information. For both the upland and alluvial soils a
strong correlation was found between bulk density and functions of organic
carbon and 15 bar water content. Cosby et al. (1984), analyzed over 1400
measurements from Agricultural Research Service reports and found significant
linear relationships between midpoint particle size fractions of the 12 tex-
tural groups of the USDA classification and each of the following quantities:
mean of the field porosity, standard deviation of the porosity, mean of the log
of saturated hydraulic conductivity, and standard deviation of the log of
saturated hydraulic conductivity. In addition, they fitted measured moisture
retention data to the model given in equation (G-3).
*(0) = *2 [9/92]b (G-3)
where ^2 and b are constant parameters. They also found significant linear
relationships between particle size and log ?s, b and standard deviation of b.
Results of this type could assist greatly in providing initial parameter
estimates for areas where only soil survey information is available. Further
study is necessary to determine the generality of these relationships found in
the individual studies and also to establish guidelines for the precision of
the correlation estimates.
STATISTICAL MODELS OF VARIABILITY
Statistical models for spatially variable quantities have been studied
only in soil science disciplines for a few years. Thus, the information sum-
marsummarized below is far from conclusive. Since the potential benefits for
using a statistical model to interpret spatially variable phenomena are high,
future research into both scaling theory and regionalized variable analysis are
strongly recommended.
Scaling Theory
Scaling theory, which is a form of similitude analysis, manipulates the ..
mathematical equations describing the transport process and forms dimensionless
groups with parameters chosen either on the basis of physical intuition or by
making specific assumptions about the geometric relationships between different
parts of the porous medium (Miller and Miller, 1956). The latter philosophy,
called similar media analysis, treats the porous medium as a scale heteroge-
neous body consisting locally of magnified versions of a reference porous
medium in which all parts of the geometry are magnified by the same scale
relation relative to this standard (Miller and Miller, 1956). This geometric
relationship to the standard porous medium can thus be represented by a single
scaling factor X giving the ratio of the length of a unit cell (e.g., a pore)
of the porous medium relative to the length of the same unit cell in the refer-
ence porous medium. Thus, the scale factor may be thought of as a magnifica-
tion factor. If the scaling theory hypothesis is reasonably well satisfied for
a field, then a number of simplifications for describing spatial variability
result immediately. First, as shown by Warrick and Amoozegar-Fard (1979), the
standard water transport equation may be scaled and thus needs to be be solved
only once to represent a simulation of any one-dimensional transport process
within the porous medium once the pro-medium is known. Since this microscopic
G-14
-------
scaling factor cannot be measured directly, it must be inferred from measure-
ments of soil physical parameters such as those described above. These param-
eters are then compared to the reference site measurement of the parameter in
order to yield the scaling factor. The relationships between various measurable
parameters and the corresponding scaling factor relationships are reviewed in
Miller (1980).
Table G-6 presents published parameters of the scaling factor distribu-
tions which were inferred from various experiments which measured soil physical
properties randomly or systematically across a field unit. In each of these
studies, the scaling factor values were fitted to log-normal distributions.
Furthermore, all scale factors have been normalized so that they have a sample
mean value equal to 1.0 while assigning a reference value of unity to the
reference location of the porous medium.
TABLE G-6. LOGNORMAL SCALING FACTOR PARAMETERS MEASURED IN FIELD EXPERIMENTS
Soil Field Measure- Scaling
Type Size M(Lnx) Sz(Lnx) CV(%) ments Method*
Reference
Panoche 150 ha
silty
clay
loam
Pima loam 87 ha
Teller <7.2 km
Silt loam 9.6 ha
-.13
-.63
-.14
-.13
-.46
-.05
Hamra sand 0.8 ha -.05
-.13
-.16
-.08
0.51
1.17
0.59
0.50
-.10 0.45
-.45 1.04
1.06
0.33
0.30
0.51
0.57
0.41
Yolo loam 0.7 ha -.38 0.77
55
170
64
53
139
143
34
31
55
62
43
90
120
120
20
120
48 180
64
26
26
120
120
120
120
72
A
B
C
D
E
F
A
B
G
H
Warrick et al., 1977a
Nielsen et al., 1973
Warrick et al., 1977a
Warrick et al., 1977a
Sharma et al., 1980
n n
Russo and Bresler, 1980
n n
n n
Simmons et al., 1979
*A Scale matric potential - saturation curve
B Scale hydraulic conductivity - saturation curve
C Scale steady state infiltration
D Calculate from lognormal saturated conductivity distribution
E Scale S in Philip infiltration equation
F Scale A in Philip infiltration equation
G Scale wetting front distance (Method 1)
H Scale wetting front distance (Method 2)
G-15
-------
Immediately obvious from this table is the fact that the distribution of
parameters is not related to soil type or field size. However, there is
apparently a relationship between the variance of the scaling factor distribu-
tion and the type of property measurement used to estimate the scaling factors.
In fact, the first four values of the scaling factor distribution given on
Table G-7 were all obtained from an analysis of the same redistribution experi-
ment published by Nielsen et al. (1973). The first two scaling factor distri-
bution parameters were obtained in the study by Warrick et al. (1977) who
scaled matric potential and hydraulic conductivity, respectively, to form two
distributions of the scaling factor. According to the similar medium hypoth-
esis, these two distributions should have been identical. Instead, the var-
iance of the hydraulic conductivity scaling factors was substantially higher
than that of the matric potential scaling factor distribution. Although a
rather high correlation (r* = .85) was observed between the two distributions,
the extra spreading found in the hydraulic conductivity distribution was un-
doubtedly due to the higher error involved in this measurement. The third set
of values from Table G-6 were obtained by directly scaling the 20 steady state
infiltration values on the plots from the Nielsen et al. (1973), study prior to
the redistribution experiment. The fourth value was obtained by scaling the
unsaturated hydraulic conductivity-water content relationship for these field
plots as presented in the study of Libardi et al. (1980). If this field
behaves as a similar media, and if each of the measurements have the same
relative precision, the scaling factor parameter should have been described by
the same distribution. The substantially different values found in Table G-6
are evidence of the failure of the similar media hypothesis for this field.
A similar lack of correspondence between different methods of measuring
the scaling factor parameters was observed by Sharma et al. (1980), who scaled
the S and A parameters in the Philip infiltration equation according to the
similar medium hypothesis. Although the model itself described infiltration
remarkably well at each of their 24 sites, the inferred scaling factor distri-
bution using those equations produced two distributions with widely differing
properties. As in the case of the Warrick et al. (1977), study, these two
parameter distributions were highly correlated but represented distributions
with quite different variance. A third evidence of the failure of the similar
media hypothesis was found in the study of Russo and Bresler (1981) in which
four methods were used to determine the scaling factor distributions. As in
the case of the Warrick study, hydraulic conductivity scaling produced a higher
variance than the matric potential scaling method even though the parameters
estimated by the two methods were correlated.
Although the studies conducted above indicate that the similar medium
hypothesis is not a good conceptual model for describing soil variability, this
represents but one type of scaling possibility to use for field heterogeneity.
Other options include separately scaling the hydraulic conductivity and matric
potential distribution as was done by Simmons et al. (1979), or even by using
kinematic or dynamic scaling which may involve scaling of boundary conditions
as well (Tillotson and Nielsen, 1984). In the latter philosophy, scaling
factors may be developed for a particular type of process (e.g. infiltration)
and the scaling factors represented from this would be applicable to a set of
similar processes. Since a major advantage of scaling theory would be a
G-16
-------
TABLE G-7. CORRELATION LENGTH PARAMETERS MEASURED IN FIELD EXPERIMENTS
Parameter
Saturated
K
Infiltra-
tion rate
Soil
sand
clay loam
weathered
shale sandy
Field
Size
(ha)
0.8
0.004
0.08
0.9
Sample Correlation
Spacing Length
(m) (m) Reference
i!o
.05
2.0
1.0
25
1.6
.13
<2.0
35
Russo and Bresler, 1981
Gelhar et al . , 1977
Sisson and Wierenga, 1981
Luxmoore et al . , 1981
Vieira et al., 1981
Percent
sand
pH
EC
loam
sandy clay
loam
sand
variable
clay loam
clay loam
clay loam
clay laom
clay loam
clay loam
clay loam
0.3
7.2
Hawaii
85
85
85
455
85
85
85
10
10
36
30
16
,500
0.2
2.0
20.0
196
2,3000
1.5
21.5
130
80
0.2
2.0
20.0
800
1.2
20.0
20.0
Vauclin et al., 1983
Campbell, 1978
McBratney and Webster,1981
Yost et al., 1982
Gajem et al., 1981
Hajrasuliha, et al., 1980
Gajem et al., 1981
reduction in the numbers of measurements, this general approach is deserving of
further study.
Regionalized Variable Analysis
An alternate approach for describing a spatially variable phenomena has
been to apply the theory of regionalized variables to the interpretation of
discrete measurements of spatially variable soil physical properties. In this
approach, a value of the parameter measured at a particular location is re-
garded as a member of a statistical set of such values described by probability
laws. In order to apply the theory, it is necessary to assume that the prop-
erty is stationary in space so that each location is described by the same
probability distribution and spatial covariances depend only on the separation
between measurements and not on the absolute location of the measurement itself
(Journel and Huijbreghts, 1978). Thus, when set of N measurements are taken on
a field, an analysis is made not only of the sample mean and sample variance,
but also on the spatial correlations between measurements as a function of the
distance between the measurements. This is accomplished in the following way.
An ordered spatial sampling pattern, usually a transect in one dimension or a
grid in two dimensions, is set up across the field. A special quantity called
G-17
-------
the semi-variance Is calculated as the mean square deviation of all replicate
measurements a given distance apart.
1 N
Y2[h] E [Z(x+h)-Z(x)]2 (6-4)
2N(h) J=l
where N(h) is the number of replicates separated by a lag distance h, and Z is
the value of the parameter. If this semi-variance increases to a plateau when
plotted as a function of lag distance or a separation, the physical interpreta-
tion is made that the parameter is spatially correlated with like measurements
of itself within a zone defined by the distance required to reach this plateau.
The plateau distance which may be measured by this method or by similar statis-
tical procedures is variously called the zone of correlation, the integral
scale, or the range. A review of methods used to calculate the length is given
in Peck (1984).
Analysis of spatial correlations offers a potentially powerful tool both
for sample design and for finding the optimum method of analyzing the informa-
tion yielded by a finite set of samples on a field. For example, if it were ••
known that the correlation length of hydraulic conductivity were, e.g., 20
meters, then samples spaced at precisely this distance would provide informa-
tion with no redundancy caused by overlap.
Table G-7 summarizes the correlation parameters which have been inferred
from experimental studies of the spatial correlation of soil physical proper-
ties. The small number of studies presented reflects both the relatively short
period of time that such procedures have been in operation and also the dif-
ficulty and expense involved in constructing experiments on the scale required
to yield such information. The four parameters indexed in this table represent
different ways of interpreting the spatial dependence parameters but are highly
correlated with each other for a given study and for all practical purposes may
be considered equivalent.
The information summarized in this table is rather ambiguous. Although
the static soil properties are generally correlated over greater distance than
the dynamic properties, variations of the correlation length of all properties
between studies seem to be much more substantial than the variations of a given
property between studies. Significantly, there is a strong correlation between
the inferred zone of influence and distance between discrete samples used to
construct the spatial correlation function. For example, the pH values taken
from the study of Gajem et al. (1981), in Table G-7 show an apparent correla-
tion length of 1.5 meters when a 20 cm transect spacing was used, a correlation
of 21.6 meters when a 2 meter spacing was used, and a correlation of 130 meters
when a spacing of 20 meters was used all on the same field. Furthermore, the
total variance of pH was similar in all three cases.
This ambiguity becomes quite obvious when studies conducted at dramat-
ically different sampling densities are compared. For example, the pH measure-
ments taken by Yost el al. (1982), using a sample spacing of 800 meters were
found to be correlated up to 2 kilometers whereas the same chemical parameter
G-18
-------
was apparently not correlated at distances greater than 1.5 meters in the Gajem
study.
There are various theoretical explanations offered for this so-called
scale effect. It is to be expected that the variance of a property will change
with the scale of observation so that larger spacing may involve detection of
a larger scale of correlation. However, in such a case, the total variance of
the sample set should be larger than the variance of a sample set analyzed
within a smaller scale. An obvious way in which this scale effect can come
about is by a violation of the stationarity hypothesis. If the underlying mean
value of the parameter has a functional drift with distance, this can bias the
interpretation of the underlying stochastic component in a manner 'similar to
that discussed above (Starks and Fang, 1982). Drift removal, however, presents
a particularly difficult problem since the functional form of the drift func-
tion is not initially known. At the present time, only somewhat subjective
procedures for drift removal have been recommended; for example, an inter-
active procedure in which candidate drift functions are subtracted from the
original sample set, and the statistical analysis is repeated until some opti-
mum criteria is satisfied (Journel and Huijbreghts, 1978).
Regionalized variable analysis is at the present time a science in its
infancy. Future investigations will undoubtedly develop some objective
statistical criteria for making the stochastic measurements and interpreting
the results which will remove much of the ambiguity found in the present study.
However, it should be cautioned that the zone of correlation or it:; equivalent
is a required property of several stochastic transport models (Gelhar et al.,
1979; Gelhar and Axness, 1983). If these properties cannot be objectively
measured, then the predictive capability of the stochastic models becomes
doubtful.
ACKNOWLEDGEMENT
Appreciation is expressed to Dr. Thomas H. Starks for assistance in
calculating Table G-5.
G-19
-------
REFERENCES
Alexander, E. B. 1980. Bulk densities of California soils in relation to
other soil properties. Soil Sci. Soc. Amer. J_. 44:689-692.
Babalola, 0. 1978. Spatial variability of soil water properties in tropical
soils of Nigeria. Soil Sci. 126:269-279.
Biggar, J. W., and D. R. Nielsen. 1976. Spatial variability of the leaching
characteristics of a field soil. Water Resour. Res. 12:78-84.
Cameron, D. R. 1978. Variability of soil water retention curves and predicted
hydraulic conductivities. Soil Sci. 126:364-371.4.
Cameron, D. R., M. Nyborg, J. A. Toogood, and D. H. Laverty. 1971. Accuracy
of field sampling for soil tests. Canad. »L Soil Sci. 51:166-175.
Campbell, J. B. 1978. Spatial variation of sand content and pH within single
continuous delineations of two soil mapping units. Soil Sci. Soc. Amer. J_.
42:460-464.
Cassel, D. K. 1983. Spatial and temporal variability of soil physical prop-
erties following tillage of Norfolk loamy sand. Soil Sci. Soc. Amer. J^.
47:196-201.
Cassel, D. K., and A. Bauer. 1975. Spatial variability in soils below depth
of tillage: bulk density and fifteen atmosphere percentage. Soil Sci.
Soc. Amer. Proc. 39:247-250.
Cosby, B. J., G. M. Hornberger, R. B. Clapp, and T. R. Ginn. 1984. A statis-
tical exploration of the relationships of soil moisture characteristics to
the physical properties of soils. Water Resour. Res. 6:682-690.
Courtin, P., M. C. Feller, and K. Klinka. 1983. Lateral variability in some
properties of disturbed forest soils in S.W. British Columbia. Canad. vh
Soil Sci. 63:529-539.
Duffy, C., P. J. Wierenga, and R. A. Kselik. 1931. Variations in infiltration
rate based on soil survey information and field measurements. Agriculture
Experimental Station Bulletin 60, New Mexico State Univ.
El Abd, H. 1984. Spatial Variability of the Pesticide Distribution Coeffi-
cient. Ph.D. Thesis, Univ. of California, Riverside.
G-20
-------
Gajem, Y. M., A. W. Warrick, and D. E. Myers. 1981. Spatial dependence of
physical properties of a typic torrifluvent soil. Soil Sci. Soc. Amer.
J_. 45:709715.
Gelhar, L. W., A. A. Baker, A. L. Gutjahr, and J. R. MacMillan. 1977. Com-
ments on "A stochastic-conceptual analysis of one dimensional ground-
water flow in nonuniform homogeneous media" by R. A. Freeze. Water Resour.
Res. 13:477-479.
Gelhar, L. W., A. L. Gutjahr, and R. L. Naff. 1979. Stochastic analysis of
macrodispersion in a stratified aquifer. Water Resour. Res. 15:1287-
1397.
Gelhar, L. W., and C. Axeness. 1983. Three dimensional stochastic analysis of
macrodispersion in aquifers. Water Resour. Res. 19:161-180.
Guenther, W. C. 1981. Sample size formulas for normal theory t tests.
American Statistician 35:243-244.
Gumaa, G. A. 1978. Spatial Variability of In Situ Available Water. Ph.D.
Dissertation, Univ. of Arizona, Tucson. (Available as 78-24365 from Xerox
University Microfilms, Ann Arbor, MI.)
Hajrasuliha, S., N. Baniabbassi, J. Metthey, and D. R. Nielsen. 1980. Spatial
variability of soil sampling for salinity studies in Southwest Iran.
Irrig. Sci. 1:197-108.
Johnston, C. D., W. M. McArthur and A. J. Peck. 1980. Distribution of soluble
salts in soils of the Manjimup Woodchip License Area, Western Australia.
Land Res. Mgmt. Tech. Paper 5, CSIRO, Australia.
Journel, A. G., and Ch. J. Huijbregts. 1978. Mining Geostatisties. Academic
Press, London.
Jury, W. A. 1982. Simulation of solute transport with a transfer function
model. Water Resour. Res. 18:363-368.
Jury, W. A. 1983. Chemical Transport Modeling: Current Research and Un-
resolved Problems. In: Chemical Mobility and Reactivity in Soil Systems.
Soil Sci. Soc. Amer. Special Publication II.
Jury, W. A. 1985. A Review of Published Studies of Field Measured Soil Water
and Chemical Properties. EPRI Topical Report (In press).
Jury, W. A., L. H. Stolzy, and P. Shouse. 1982. A field test of the transfer
function model for predicting solute transport. Water Resour. Res.
18:369-375.
Jury, W. A., H. El Abd and T. M. Collins. 1983. Field Scale Transport of
Nonadsorbing and Adsorbing Chemicals Applied to the Soil Surface. NWWA
Symposium, Las Vegas. National Water Well Assoc., Worthington, OH.
G-21
-------
Libardi, P. L., K. Reichardt, D. R. Nielsen, and J. W. Bigar. 1980. Simple
field methods for estimating soil hydraulic conductivity. Soil Sci. Soc.
Amer. J_. 44:3-6.
Lund, L. J. 1982. Variations in N03 and Cl concentrations below selected
agricultrual fields. Soil Sci. Soc. Amer. J_. 46:1062-1067.
Luxmoore, R. J., B. P. Spalding, and I. M. Munro. 1981. Areal variation and
themical modification of weathered shale infiltration characteristics
Soil Sci. Soc. Amer. J.. 45:687-691.
McBratney, A. B., and R. Webster. 1981. Spatial dependence and classification
of the soil along a transect in Northeast Scotland. Geoderma 26:63-82.
McBratney, A. B., and R. Webster. 1983. How many observations are needed for
regional estimation of soil properties Soil Sci. 135:177-183.
Miller, E. E. 1980. Similitude and scaling of soil water phenomena. J^:
Applications of Soil Physics. Academic Press, New York.
Miller, E. E., and R. 0. Miller. 1956. Physical theory for capillary flow
phenomena. J_. App1• Pnvs. 27:324-332.
Nielsen, D. R., J. W. Biggar, and K. T. Erh. 1973. Spatial variability of
field measured soil water properties. Hilgardia 42:215-259.
Oster, J. D., and J. D. Wood. 1977. Hydrosalinity Models: Sensitivity to
Input Variables. In: Proc. Nat. Conf. on Irrigation Return Flow Quality
Management. USEPA, Colorado State.
Peck, A. J. 1984. Field variability of soil physical properties. Adv. ln_
Irrig. 2:189-221.
Peck, A. J., R. J. Luxmoore, and L. J. Stolzy. 1977. Effects of spatial vari-
ability of soil hydraulic properties in water budget modeling. Water
Resour. Res. 13:348-354.
Pratt, P. F., W. W. Jones, and V. E. Hunsaker. 1972. Nitrate in deep soil
profiles in relation to fertilizer rates and leaching volume. ±. Environ.
Qual. 1:97-102.
Richter, G. 1984. Microlysimeter and Field Study of Water and Chemical Move-
ment Through Soil. M.S. Thesis, Univ. of California, Riverside.
Russo, D., and E. Bresler. 1980. Scaling soil hydraulic properties of a
heterogeneous field. Soil Sci. Soc. Am. J_. 44:681-683.
Russo, D., and E. Bresler. 1981. Soil hydraulic properties as stochastic
processes: 1. Analysis of field spatial variability. Soil Sci. Soc.
Amer. J. 45:682-687.
G-22
-------
Sharma, M. L., G. A. Gander, and C. G. Hunt. 1980. Spatial variability of
infiltration in a watershed. »h Hydro!. 45:101-122.
Simmons, C. S., D. R. Nielsen, and J. W. Biggar. 1979. Scaling of field-
measured soil-water properties. Hilgardia 47:77-174.
Sisson, J. B., and P. J. Wierenga. 1981. Spatial variability of steady state
infiltration rates as a stochastic process. Soil Sci. Soc. Amer. J^.
45:699-704.
Starks, T. H., and J. H. Fang. 1982. The effect of drift on the experimental
semivariogram. Mathematical Geol. 14:309-311.
Starr, J. L., H. C. DeRoo, C. R. Frink and J. Y. Parlange. 1978. Leaching
characteristics of layered field soil. Soil Sci. Soc. Amer. J.
42:386-391. "
Stockton, J. G., and A. W. Warrick. 1971. Spatial variability of unsaturated
hydraulic conductivity. Soil Sci. Soc. Amer. J.. 36:847-849.
Tillotson, P., and D. R. Nielsen. 1984. Scale factors in soil science. Soil
Sci. Soc. Amer.
-------
Warrick, A. W., and D. R. Nielsen. 1980. Applications of Soil Physics.
Academic Press, NY.
Wild, A., and I. A. Babiker. 1976. The asymmetric leaching pattern of N03 and
Cl in a loamy sand under field conditions. J^. Soil Sci. 27:460-466.
Willardson, L. S., and R. L. Hurst. 1965. Sample size estimates in perme-
ability studies. vh Irrig. Drain. Div., Amer. Soc. Civil Eng. 91:1-9.
Yost, R. S., G. Uehara, and R. L. Fox. 1982. Geostatistical analysis of soil
chemical properties of large land areas. Soil Sci. Soc. Amer. J.
46:1028-1032.
G-24
------- |