£EPA
United States
Environmental Protection
Agency
Environmental Research
Laboratory
Athens GA 30605
EPA 600/3 78-065
July 1978
Research and Development
Exposure
Assessment
Modeling
A State-of-the-Art
Review
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ECOLOGICAL RESEARCH series. This series
describes research on the effects of pollution on humans, plant and animal spe-
cies, and materials. Problems are assessed for their long- and short-term influ-
ences. Investigations include formation, transport, and pathway studies to deter-
mine the fate of pollutants and their effects. This work provides the technical basis
for setting standards to minimize undesirable changes in living organisms in the
aquatic, terrestrial, and atmospheric environments.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/3-78-065
July 1978
EXPOSURE ASSESSMENT MODELING:
A STATE-OF-THE-ART REVIEW
by
Catherine Miller
Kennedy School of Government
Harvard University
Cambridge, Massachusetts 02138
Grant No. 805647
Project Officer
Kenneth Hedden
Technology Development and Applications Branch
Environmental Research Laboratory
Athens, Georgia 30605
ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
ATHENS, GEORGIA 30605
-------
DISCLAIMER
This report has been reviewed by the Environmental Research
Laboratory.- U.S. Environmental Protection Agency, Athens,
Georgia, and approved for publication. Approval does not
signify that the contents necessarily reflect the views and
policies of the U.S. Environmental Protection Agency, nor does
mention of trade names or commercial products constitute endorse-
ment or recommendation for use.
XI
-------
FOREWORD
Environmental protection efforts are increasingly directed
towards preventing adverse health and ecological effects associ-
ated with specific compounds of natural or human origin. As part
of this Laboratory's research on the occurrence, movement, trans-
formation, impact, and control of environmental contaminants, the
Technology Development and Applications Branch develops and eval-
uates management or engineering tools for assessing and control-
ling adverse environmental effects of these substances.
Solutions to serious environmental pollution problems
increasingly involve analyzing and evaluating difficult trade-offs
between useful substances and products and risks posed to human
health and environmental quality from toxic materials. One use-
ful method for estimating the actual human and environmental ex-
posure to substances and their effects is the application of
mathematical models. This report investigates a number of multi-
media mathematical exposure assessment models now in use or under
development and provides recommendations for research priorities
in the further development of these techniques for the use of
environmental decision-makers.
David W. Duttweiler
Director
Environmental Research Laboratory
Athens, Georgia
-------
ABSTRACT
This report presents a state-of-the-art review of exposure
assessment modeling. It describes currently available models
that simulate the environmental fate of substances, the exposure
to such substances, and the effects of such exposure. The
focus is first on exposure and effects, where relatively little
work has been done, and then on models of environmental fate,
in particular on intermedia transfer processes. Single-medium
air and water quality transport models are not assessed, but
the possibility of approaching multi-media problems through a
combination of single-medium approaches is explored. The report
also describes several actual risk assessments that have been
made with limited data, and considers some secondary applications
of the models that have been suggested.
The results of this investigation show that the available
models do not cover all of the areas necessary for an exposure
assessment. More effort has been directed to modeling environ-
mental fate than to modeling exposure and effects.
In general the models are not operational from the decision-
maker's point of view because they have not been validated, lack
sensitivity analyses for key parameters, do not include cost
estimates and often call for data not readily available. There
are also organizational and administrative barriers to the
actual use of these models.
This report was submitted in fulfillment of Grant No.
68-13-0040 by the Kennedy School of Government, Harvard Univer-
sity, under the sponsorship of the U.S. Environmental Protection
Agency. This report covers the period October 20, 1977 to
April 19, 1978 and work was completed as of June 10, 1978.
IV
-------
CONTENTS
Foreword
Abstract iy
Acknowledgments vi
I. Introduction 1
II. Conclusions 4
III. Recommendations 7
IV- Modeling Exposure and Effects 9
Health Effects 9
Dose-response Relationships H
Ecological Modeling 13
Human Population Exposure 16
Summary 17
V. Modeling Interfaces between Media 19
Air and Soil/Water Interface 20
Source Depletion Model 21
Surface Depletion Model 21
Wet Deposition Rates 22
Dry Deposition Rates 23
Evaporation or Volatilization Rates • - 23
Soil and Water Interface 26
Adsorption-desorption 27
Transport from soil to aquatic systems 28
Uptake in Biological Systems 29
Summary 30
VI. Examples of Fate Models 32
Unified Transport Model 32
Conceptual Model 37
Mass Balance Approach 38
Environmental Rates Approach 40
Summary of Fate Model Findings 43
VII. Examples of Exposure Assessments 45
VIII. Other Uses for the Models 48
Integrated Exposure Assessment Monitoring • 48
Ranking Schemes 49
IX. References 51
v
-------
ACKNOWLEDGMENTS
The author gratefully acknowledges the assistance and
advice of Professor Richard Zeckhauser of the Kennedy School
of Government. Helpful and constructive comment was received
from Kenneth Johnson of the Environmental Protection Agency,
while he was at the Kennedy School of Government. Substantial
assistance in collecting the materials and information was
received from Dr. James Falco (Environmental Research Laboratory,
Athens, Georgia), Dr. Dale Huff (Oak Ridge National Laboratory)
and Dr. Shonh Lee (Stanford Research Institute).
VI
-------
SECTION I
INTRODUCTION
Recent events in the environmental protection field have
demonstrated that the first generation of control activity has
run its course. Solving the more serious problems that remain
will require a much greater degree of understanding and sophisti-
cation, particularly in analyzing and evaluating the desirability
of alternative actions. In the future, we will be faced with
difficult trade-offs between useful substances and products and
the concomitant risks posed to human health and environmental
quality.
In the past, EPA's approach has been to model sources of
pollutants and apply controls to these sources. Single-medium
models have been used to simulate changes in air or water
quality due to substances emitted or discharged directly into
the medium being modeled. Quality is described in terms of
concentrations, spatially and/or temporally determined. However,
simply knowing where and how a potential pollutant is being
released to the environment is not an adequate basis for making
the sort of trade-off decisions that will be required. It is
also necessary to estimate the actual human and environmental
exposures and the effects that result from these exposures.
In assessing exposure, the decision-maker needs information
on:
-transport, transformation and fate of the substances
through all the media
-exposure of humans, animals and plants to the substance
-effects on humans, animals and plants
-severity and likelihood of occurrence of the effects.
This information can be supplied in a variety of ways; mathe-
matical models may be one of the most useful.
A model is a tool for exploring complex phenomena and their
interactions. It is a simplified representation of the inter-
relationships among elements of the particular system to be
explored. A model can take various forms. It can be a purely
mental model or philosophical construct. Explicit models in-
clude schematic ones, such as blueprints; physical ones, such
as microcosms; and symbolic ones. The models we investigate in
-------
this paper are symbolic, and use equations to represent the
interrelationships. A mathematical model expresses symbolically
one or more elements of the system (dependent variables) in
terms of the others (independent variables). The decision-
maker uses the model to predict what will happen to certain
elements of the system when the independent variables are
varied or controlled. These predictions, e.g., severity and
likelihood of a particular effect from a given change in
pollutant emissions, can provide a basis for setting priorities
and making trade-off decisions. Basically, they detail the
range of outcomes that is available given alternative control
strategies.
As mentioned above, the mathematical model is only one of
a variety of ways to provide the necessary information. Other
methods include monitoring and field and laboratory experiments.
Whether the decision-maker will use a model and which model he
will use will depend on the accuracy of its representation of
the system being studied and on how easily, and at what cost,
it provides the needed information. More specifically, the
decision-maker's criteria for evaluating the usefulness of a
model would include:
-applicability (appropriate boundaries, outputs useful for
the particular problem)
-accuracy of estimates and validation of model
-whether necessary data inputs are available
—assumptions explicit and significance of results
explained
—sensitivity analyses
-cost compared to other studies or sources of information.
Clearly these criteria are interrelated. In many situations
they are competitive and trade-offs among them will have to be
made.
The application of these criteria raises several specific
issues that must be addressed in a state-of-the-art assessment
of this type of modeling. The first issue concerns applicabi-
lity- In exposure assessment, a useful model should represent
important chemical, physical and biological processes and
interactive (synergistic or antagonistic) effects for each
medium. In addition it should produce estimates of the environ-
mental fate of substances, exposure and effects across media.
The second issue concerns accuracy, comprehensibility and
other operational aspects of the criteria. It is important to
hold down the cost of collecting input data and running the
model, if it is to be used. Regardless of the model's simpli-
city or accuracy, the decision-maker will not use it if he does
not understand what its results mean.
The third issue concerns actual use. Given the complemen-
tary nature of the criteria, models of varying degrees of
-------
applicability and operational readiness will or will not be
used in particular situations. Thus, the actual model uses
and the circumstances surrounding their use are also an impor-
tant part of a state-of-the-art assessment.
This report investigates a number of mathematical exposure
assessment models now in use or under development, considering
the completeness with which they cover the phenomena of interest,
their practical operational qualities, and the particular uses
to which they have actually been put.
Because, fundamentally, we are most concerned with the
effect of substances on human beings, those models that incor-
porate exposure and effects are discussed first. These include
studies of health effects, ecological modeling and human popula-
tion exposure studies.
These models depend on estimates of pollutant concentration
levels. Because of the single-medium focus of control activities
mandated by U.S. environmental legislation and EPA's historical
approach to regulation, a large number of models exist to
predict such concentrations in particular media. Assessing
these single-medium models is beyond the scope of this report.
Such models could be useful in assessing exposure if linked
together in such a way as to account for intermedia fluxes,
however. This report characterizes the information needed to
link together single-medium transport and transformation models,
and gives examples of a sequentially linked model and other
approaches to studying environmental fate.
The final sections of the report present examples of
exposure assessments that have been made and discuss other uses
for the information provided by the models described in this
report.
-------
SECTION II
CONCLUSIONS
Three issues have been investigated in this state-of-the-
art assessment of exposure assessment modeling: applicability
or coverage, operational state, and actual use of the models
for decision-making. Specific conclusions regarding these three
issues are summarized here.
APPLICABILITY
Available models have not given equal attention to the
various factors that should be of concern in an exposure assess-
ment. Population exposure and interactive effects (synergistic
or antagonistic) are the least understood or studied. The vast
majority of models deal only with a single pollutant and do
not consider possible interactive effects with other pollutants.
Only rather primitive studies of population exposure have
been done so far. These studies have been limited by their
dependence on census data of residential living patterns for
the human population. In addition, no sound theoretical basis
has been produced for developing predictions based on actual
behavior patterns, including travel and work.
Ecological models for plant and animal populations also
suffer from a lack of unifying theories for biological inter-
actions and thus tend to be highly site-specific, not permitting
extrapolation. In addition, because they often fail to esti-
mate both the level and the duration of exposure, the quantita-
tive link from dose to response frequently cannot be made.
Information on effects comes mainly from experimental tests
or field (epidemiologic) studies. Theoretically, it can be
linked to environmental fate and exposure models through a
dose-response relationship. Extrapolating from tests done on
one species (animal) to another species (human) is inherently
problematic, however. Moreover, experimental doses frequently
do not correspond to low environmental concentrations. Finally,
epidemiologic studies record correlations but rarely are able
to specify cause and effect; thus findings from such studies
should not be accepted uncritically for use in exposure assess-
ment models.
-------
Models do exist for studying transport and environmental
fate, though their applications remain largely single-medium
oriented. Moreover, they are most appropriately used for
conservative or non-reactive substances, because models of
chemical and biological transformations are not as advanced
as models of physical transport. Advances have been made in
modeling chemical processes, which predominate in the air
medium, however. Without a theoretical knowledge of the under-
lying processes, water and soil models remain dependent on
empirical estimation of parameters describing the distribution
of pollutants.
This report focuses on models of the processes at the
interfaces between the media. Here again, little is known
of the fundamental mechanisms underlying these processes.
Accordingly, models characteristically rely on empirical esti-
mates of parameters and require large input data sets and
calibration for specific sites before use. Several conceptual
approaches to modeling environmental fate from a multi-media
perspective have been suggested, but only a few actual applica-
tions exist. Thus, verification or experience with which to
choose among the alternative approaches is lacking.
OPERATIONAL CAPABILITY
Far more models have been proposed than are currently
operational. The relative abundance of suggested approaches
and paucity of applications is due in part to irrelevant or
inadequate data. The models' data requirements are beyond
available resources or in some cases beyond measurement
capability. In order to use a model, the decision-maker must
also understand its assumptions and the significance of its
results. Generally, these foundations of a particular model
are well documented. However, in almost no case is the model's
accuracy, sensitivity to key parameters, or cost defined. This
information must be available to the decision-maker if the model
is to be considered truly operational.
ACTUAL USE
Problems of organization and administration exacerbate
these technical problems and act as a barrier to the actual use
of the models. An exposure assessment requires a combination
of. skills traditionally taught and practiced separately.
Usually health effects researchers are not trained in chemistry
'and chemists know little biology, so that interdisciplinary
teams must be formed to develop the model. Within EPA, modelers
are not generally working with those responsible for monitoring
systems, so that coordination of these efforts must cross budget
lines as well as spheres of interest and academic training.
-------
A third barrier arises between researcher or modeler and the
user of the model. The user (e.g., in Washington, D.C.) is
often physically removed from the modeler (in EPA or contractor
laboratories) and may not be scientifically trained; it may be
difficult to communicate what is desired from the model and what
can reasonably be expected.
The modeler may also be separated from the user by atti-
tudes and objectives. A decision-malcer evaluates a particular
model to be used for predictive purposes, in terms of its
applicability, accuracy, availability of input data, comprehen-
sibility and cost. The research scientist, on the other hand,
is less interested in prediction than in furthering his under-
standing of the process or system being modeled." Thus, the
scientist's criteria for evaluating the usefulness of a model
as a research tool would include:
-theoretical foundation
-accuracy and validation
-availability of data inputs.
He would be less interested in its applicability, comprehensi-
bility and cost. Both parties must communicate their objectives
explicitly and try to understand the capability of the model in
relation to these different objectives.
It should also be noted that much work in this area is
being conducted by academic and private research institutions,
where the publication of new findings is important to individual
success. This provides an incentive to regard planned research
as proprietary rather than as information to be shared.
The imbalance of research effort noted above is in part
a result of these barriers. Moreover, available models appear
often to have been developed as a by-product of research for
other purposes. Consequently.- they are fragmentary and fre-
quently are not followed through to the application stage.
-------
SECTION III
RECOMMENDATIONS
The following recommendations concern the process by which
models are developed. Because exposure assessment modeling is
still not operational and because there is no consensus on what
approaches are likely to be most effective, specific recommenda-
tions for marginal changes in existing models are not appropriate
at this stage. Instead, the recommendations focus on steps that
could be taken to establish priorities for a research effort in
this area.
1. Statistical studies of factors affecting population exposure
should be conducted. The resulting data would be a vital
link in exposure assessment modeling and would advance the
understanding of which parameters may be significant for
estimating human behavior.
2. A careful, published record of environmental conditions pre-
vailing in experimental or field studies should be required.
A comparison of such records would provide clues to possible
synergistic or antagonistic effects.
3. All models should explicitly consider input data needs and
the feasibility of collection. For the environmental fate
models the emissions data were most often cited as inade-
quate. Because of the importance of such factors as control
equipment, quality of maintenance and age, an emissions
inventory should be based on observation of actual sources
rather than engineering estimates. The current lack of
technology to detect low concentrations of many substances
in the environment may be a limiting factor for this type
of modeling.
4. In order to improve communication between modeler and user,
meetings should be held which both research scientists and
policy makers attend. These meetings should cover both the
needs of the decision-maker and the capability of the models.
5. Improved reports describing the models are needed to facili-
tate communication among scientists and between scientists
and users. Because no consensus exists on how to do an
exposure assessment, a rapid dissemination of technical
information is required. It would be helpful if every EPA
-------
report of a research effort involving modeling were required
to include information on:
-input data required and how obtained
-theory and assumptions employed in the model
-output produced, its form and estimate of accuracy
-sensitivity analysis results for key parameters
-applications and validations attempted
-cost
-documentation on how to use.
Several points in this list deserve emphasis. It is impor-
tant to report gaps in the knowledge or assumptions made.
Modeling is an excellent way to discover such gaps but the
need to complete a model also serves to cover over those
gaps. To the user who must choose between alternative
models, cost estimates can be as important as output results.
6. Reports should be made for studies producing non-results.
It is understandable that research scientists find little
stimulation in conducting and documenting exposure studies
at concentrations or for substances that fail to produce
changes in the biologic system in which they are interested.
The conduct and reporting of such studies, however, is im-
portant for a total evaluation of risk. Knowledge of non-
results would assist in making trade-off decisions because
the lack of risk as well as benefit is an important factor
in such decisions.
-------
SECTION IV
MODELING EXPOSURE AND EFFECTS
Quantitative exposure assessment involves estimates of the
size of the exposed population, the concentration level and dura-
tion to which it is exposed, and the effects expected to result
from this exposure. The following sections present an account
of techniques by which exposure and effects can be estimated
for humans, a review of ecological modeling of plant and
animal populations, and a discussion of the few studies done
to estimate the size of the exposed population.
HEALTH EFFECTS
Exposure assessment modeling attempts to infer dose-response
relationships relating health effects to the level and duration
of exposure to the potentially toxic substance. Much of the
research being done in this field is directed toward developing
relatively quick and inexpensive tests that can be used to
screen substances for potential toxic effects. For example,
under the Toxic Substances Control Act of 1976 a hierarchy of
tests is likely to be required in determining whether a new
chemical can be produced. Such tests by themselves may not
yield the information necessary to establish a dose-response
relationship because the focus is on the effect and doses may
not vary. Some difficulties with this and other approaches to
measuring health effects are discussed below.
Several problems confront the researcher attempting to
establish a dose-response. The health effect of a single expo-
sure or repeated short-term exposures may—or may not—be dif-
ferent from the effect of long-term low-level exposure. Acute
adverse effects may be the cumulative result of long-term lower-
level exposures, or the effect of short-term peak exposures.
Conversely, chronic effects may follow short-term peak exposures,
as well as long-term low level exposures. In general, it is
easiest to obtain data on acute effects following short-term
exposures. Less information is available on acute and chronic
effects following,long-term exposures and very little is known
about measuring chronic effects of peak exposure.
Another complication is the possibility of synergism or
-------
interactions between substances in producing effects. When
studies are limited to a single pollutant or chemical, the
effects produced may not be representative of actual environ-
mental conditions. Another substance not present in the labora-
tory situation may enhance or inhibit the effect in the real
world.
A third problem is extrapolation. Tests made on animals
may not be applicable to humans because the animal may metabolize
the substance differently or because the animal's organs may be
differentially sensitive to the substance. The interpretation
of tests involving atypical levels of exposure is also prob-
lematic. Often a testing procedure will use high levels of the
substance over a short period of time in order to. produce a
measurable effect and lower the cost of the test. These levels
and durations may not correspond to any actually observed in
the environment; the problem is how to use such a test to pre-
dict effects at low environmental levels. This problem will be
more fully discussed in the next section of the report.
The problem of determining dose-response relationships has
been attacked through epidemiology, clinical research and
animal toxicology. Each of these approaches suffers to some
extent from the difficulties described above.
Epidemiological studies are set in the real world and allow
consideration of the effect of long- and short-term pollutant
exposures on different groups of the human population. Often,
however, only crude measurements of health effects are possible;
in addition, such studies frequently cannot adequately control
for all influencing factors and are restricted to a limited
range of exposures.
More sophisticated health measurements and controlled ex-
posures are possible in clinical studies. Human studies,
although expensive, have the obvious advantage of eliminating
the need to extrapolate the results from another species to
man. However, long-term exposures cannot be easily evaluated.
Toxicological testing provides an opportunity for carefully
controlled studies that may involve a wide range of pollutants
and actual examination of affected body tissues. These tests
often serve largely as the basis for judging whether the expo-
sure of a living system to a chemical presents a hazard.
However, differences among species and between animals and
humans limit the usefulness of such studies. This approach
is quite costly for tests of chronic effects.
Given their various imperfections, it may be appropriate
to use all three approaches for a particular problem. The
studies should be designed to complement each other and to
elucidate the biological mechanisms involved. Finklea's review
10
-------
(1977) of available research on exposures to certain air pollu-
tants shows that there has been an imbalance between the three
research approaches as well as a lack of quantitative data
about several effects of major concern. Atkeson (1976) attri-
butes this problem in part to the lack of effective coordination
of research on effects. Similar conclusions are summarized in
an EPA report (1978) on health effects of transportation-
related pollutants. Information on what research is planned
or in progress is not easily available, especially outside
government agencies. Also, because there is little interaction
between health scientists and modelers of transport and trans-
formation of pollutants, independent researchers may establish
priorities that greater shared understanding would show to be
inappropriate; duplication of effort is also a problem, and
there is the danger of overlooking a potential health hazard.
DOSE-RESPONSE RELATIONSHIPS
To estimate the health consequences of changes in environ-
mental quality, a dose-response relationship must be established.
Several difficulties may arise. There is little biological
theory on how a substance affects the body. In the field of
pharmokinetics, the metabolic model (Friberg, 1976) describes
processes of absorption, distribution, deposition, etc. at the
cell and organ level in order to estimate the critical concen-
tration or threshold level at which an effect is produced. This
procedure is potentially applicable to effects resulting from
long-term exposure to a substance with a long biological half-
life, but it requires an enormous amount of information, little
of which is currently available.
A more commonly used approach is to do animal experiments
or occupational exposure studies. However, the data collected
in such exercises may not be suitable for establishing a dose-
response relationship. An effect may be recorded and described
as to degree of severity. For a given population and given
dose the measure reported as effect is percentage showing the
effect. A different measure is response. Response enumerates
"reactors," percentage of the population all given the same dose
and exhibiting the same effect. Examples are LD5Q or LC ,
common statistics collected in toxicity studies. The statistic
LDj-f, is the dose causing death of 50% of the population; LCc-.
is the concentration causing 50% of the population to show the
given effect. Unfortunately, the tests may be conducted in such
a way that it is not possible to separate the two measures—
effect and response. The results may be due to both an increase
in the severity of the effect and an increase in response for
each effect as the exposure level increases.
Second, the commonly accepted concept of "dosage" may
obscure some important differences. Dosage is defined as the
11
-------
product of concentration level and exposure duration or time.
But equal dosages may not produce equal effects. Larsen (1977)
has shown that concentration is more important than exposure
duration in causing excess mortality in mice exposed to nitrogen
dioxide. This problem can be overcome, but it is often important
to give separate attention to the two factors of concentration
and duration when reporting and using experimental results.
Third, even when experiments are carefully conducted and
reported for the specific purpose of establishing dose-response
relationships, it is usually necessary to extrapolate from the
results to the much lower doses encountered in the environment.
Several approaches are taken in the literature. One is to
assume there is a threshold, a definable concentration below
which a chemical will not produce a toxic response in an ex-
posed subject. However, this assumption solves the problem
of extrapolation only if a study has been done at low enough
concentrations that no effects are observed. EPA's National
Ambient Air Quality Standards are an example of the use of
thresholds.
Another approach is a straight-line projection from the
experimental result to zero dose, which assumes there is no
threshold or safe level. This is called the linear or "one-
hit" model, since it assumes that each increment of exposure
has the same independent low probability of causing the effect,
regardless of the dosage level. This assumption is generally
accepted for radiation studies. However, such a model is not
likely to be realistic for chemicals, but rather would define
only an upper limit to the level of effects likely at very low
doses. Because physiological detoxification mechanisms would
render small doses less effective, there would probably be. some
sort of threshold or at least a smaller effect per dose incre-
ment at low levels.
A third approach, the log-probit model, is based on the
assumption that the observed changes in response are the result
of variations of susceptibility in the population. This varia-
tion is assumed to be log-normally distributed with dose. This
approach has been followed by Larsen (1976 and 1977) in assess-
ing both plant injury and animal mortality- It is based on the
finding by toxicologists and pharmacologists that an organism's
response to a drug dose is often proportional to the logarithm,
rather than the arithmetic value, of the dose. They have also
found that there is a range of doses so low that no response
is measured, and a maximum response that will not be exceeded
even if the dose is increased. Such data thus appear to be
log-normally distributed. The major problem with this model
is that the normal distribution may not be as reliable in the
tails of the distribution as in the central part. Unfortunately,
it is precisely the tails (very low environmental exposures or
high accidental exposures) in which we are interested and the
12
-------
probit function obtained from the observable range may over-
estimate the probability of a response at low doses.
A final recently suggested approach is based on kinetics
(Cornfield, 1977). It assumes a reaction in which the toxic
substance is deactivated by another substance, but which may be
reversed, disassociating back into toxic agent and deactivating
agent. This model produces a dose-response curve shaped like
a hockey stick, flat up to the threshold level and rising steeply
afterwards, once the protective or deactivating mechanisms are
saturated. The existence of such a threshold level depends on
the existence of at least one protective reaction.
Cornfield (1977) provides a good survey of the various
approaches to estimating dose-response relationships. All of
these approaches are still largely conjectural and much remains
to be learned from experimental tests for a wide range of
dosages.
ECOLOGICAL MODELING
Because direct experiments can be made on plants and
animals, there should theoretically be no need to extrapolate
from one species to another. However, in practice such experi-
ments are exceedingly expensive and choices must be made about
which studies to conduct and how to extrapolate from them.
Models of the processes of interaction of plants, animals and
the environment would aid in making these choices. Until
recently, however, ecologists have focused on qualitative
descriptions of the structure and dynamics of relatively un-
disturbed ecosystems. It is only in the past decade that
modeling has played an important role in ecological research.
Modeling can enhance fundamental understanding by organizing
the research effort, synthesizing existing knowledge, analyzing
hypotheses about a system's structure and dynamics, and identify-
ing fundamental constraints on a system's functioning and the
mechanisms to which a system's behavior is most sensitive. As
a predictive or management tool, modeling can also be used to
measure environmental changes caused by human-induced distur-
bances of natural ecosystems.
Management models include models of terrestrial grazing
systems, fisheries and pest control. These models have founda-
tions in population dynamics including basic processes of
predation, competition, birth, death and migration. To model
environmental factors, the concepts of flow, threshold and
interference must be included. The flow of materials or energy
between populations reflects the dynamics of interacting popula-
tions and feedback mechanisms that occur in nature. Threshold
effects include the appearance of behavioral or physiological
responses that would substantially alter the form of the rela-
13
-------
tionship between the population and its environment. Inter-
ference may be due to the scarcity of material resources as
opposed to the effects of direct competition for available
space. The few working models of the management type are
summarized in Wiegert (1975). In general they are confined to
a single species or link in the food chain and do not incor-
porate feedback or threshold effects. Wiegert (in press) has
proposed a method for incorporating these effects but the appli-
cations remain limited to single species.
Ecosystem models, which include more than a few component
species or environmental conditions, are primarily directed
toward an understanding of the natural system and the identifi-
cation of critical parameters and processes. These include
models dealing with aquatic ecosystems, forest, grassland,
tundra, desert and soil-litter. They focus on the integrated
dynamics of the entire community under study and describe the
flow of energy, biomass, and nutrients through the system.
Aquatic ecosystem models are probably the most numerous and
have grown out of a desire to simulate water quality changes as
well as quantities and flow. Patten (1976) presents some of
the more recent of these models which describe both aquatic and
terrestrial ecosystems.
Ecosystem models must relate physical, chemical and
biological processes. Although considerable progress has been
made in developing realistic mathematical models such as these,
several remaining major problems limit their predictive ability.
These problems include the requirement of a multi-disciplinary
approach, sparse data sets, biological variability, interactions
of processes operating at significantly different temporal and
spatial scales, and nonlinearity.
Because many disciplines are involved, most modeling
efforts require a team approach. There has not been a great
deal of experience in organizing and maintaining modeling teams.
Nor have fundamental unifying theories been developed that can
be used to deduce quantitative ecological behavior. Thus,
existing models often reflect the expertise of their builders
in being sophisticated in one respect and quite crude in others.
The lack of a unifying theory may also exacerbate the prob-
lem of data availability. Many data sets exist for a given
phenomenon or system but few contain measurements of all the
factors that are important in controlling the dynamics. Dif-
ferent parameters observed in nature may be indicators of the
same substances; some indicators may be redundant. It may be
difficult to apply a model because the data needed to quantify
all the model's parameters may not be available for a specific
situation. On the other hand, not all the substances that have
been quantified by analysts are of real ecological significance.
The problem is to select a set of both meaningful and measurable
14
-------
parameters that will reflect sufficiently the ecological changes
of interest. It should be noted that collecting the data needed
to estimate the parameters or validate an ecological model
usually takes a long time.
Biological variability or uncontrolled factors make eco-
systems much more difficult to model than physical transport
systems. Not all individuals of a species placed in identical
environments will respond identically, due to differences in
heredity and/or environmental preconditioning. This variability
poses special problems for deterministic models and may argue
for the incorporation of probabilistic elements.
The interactions between the various components of an
ecosystem are played out over an extremely broad temporal and
spatial range. Actions and reactions are delayed in time and
components are nonrandomly arranged in space. One solution is
to run many concurrent models, each appropriate to a given
spatial-time resolution. However- the problem remains if the
areas being modeled change in size or the spatial heterogeneity
itself changes over time.
Ecological phenomena are inherently nonlinear. Both
threshold and saturation reactions occur commonly in ecological
systems. Nonlinear interactions make mathematical analysis of
a model complex.
A large body of mathematical theory is concerned with
deriving analytical solutions to linear systems; this work gives
insight into the functioning of a system and can be used to
check the model for logical or conceptual inconsistencies. The
argument for the use of linear equations as a reasonable approxi-
mation is that behavior is linear if the state variables fluctu-
ate around some given point, such as an equilibrium point.
However, linear equations can only show how the system operates,
not why it operates as it does. If wide fluctuations or changes
in policies occur, more realistic nonlinear equations may be
needed. In addition, the use of nonlinear interactions may
increase the stability of a system simplifying the system's
response to the disturbance. Further discussion of this problem
is found in Bledsoe (1976).
It should be noted that the scope of this report has
prompted a somewhat narrow view of ecological modeling. Much
work is being done with physical models of systems in the form
of microcosms in the laboratory. They provide an overall
assessment of the complex interactions taking place but are not
a mathematical model in the sense we use the term in this
report.
15
-------
HUMAN POPULATION EXPOSURE
The main environmental routes through which humans are
exposed to pollutants are air, water, and food. Human behavior
patterns as well as pollutant characteristics will determine
which of these routes are important avenues of exposure. Trans-
port and transformation models predict concentration levels in
the various media on the basis of the specific characteristics
of the substance. In making an exposure assessment it is also
necessary to estimate the size of the population actually ex-
posed to these concentrations.
Air
Studies of air pollutants have generally estimated popula-
tion exposure from census data by determining how many people
live within a certain radius of the sources (Kuzmack and
McGaughy, 1976 and Mara and Lee, 1977). Several problems are
inherent in such a methodology. Census data concern residents
of an area, but people spend part of their time away from home.
If the substance is a localized air pollution problem, travel
patterns would affect exposure. Generally no allowance is made
for this. However, Horie and Stern (1976) have done an analysis
for the sub-populations of school-age and elderly (as well as
for the whole population), eliminating the working-age population
because of its greater daily mobility.
In addition, the length of time a person lives in the area
will be important. Thus, some account should be taken of
longer-term mobility into and out of the area. Horie and Stern
(1976) correlate ambient air concentrations for each year with
the census data for that year. EPA (1977) has incorporated
this factor by using an average figure for the entire U.S.
(6.7 percent of U.S. population moves out-of-county each year).
These methods yield an estimate of a level of concentration
that is assumed to be continuous. However, health effects may
vary according to duration as well as level of exposure. Horie
and Stern (1976) have estimated various long-term and short-
term exposures to obtain measures of population-at-risk, the
distribution of population associated with various dose levels
above the EPA standards.
Another problem in using census data is matching the loca-
tion of sources and concentration levels or air monitoring
stations with census statistical areas on a geographical basis.
Mara and Lee (1977) assume the source is at the city center and
use the population density for that city. EPA (1977) assumes
average population densities for urban areas. Horie and Stern
(1976) aggregate the population data to establish points within
an area indicating resident population and then interpolate air
16
-------
quality data from air quality monitoring stations nearest the
point.
Water
The other routes of exposure are even less well described
than air. Generally, if the source of a pollutant is drinking
water, some average intake of water per day is assumed, with
the average concentration in water determined through tests.
Food
Reliable estimates of exposure through food are especially
difficult to obtain. It is difficult in any case to analyze the
substances in food; in addition, the analysis is often done on
raw food whereas the actual exposure takes place through a mixed
diet or processed foods. Concentrations of the substance in
question may increase or decrease during processing. Thus,
assumptions about the types, the amounts and the processing of
food must be made.
Occupational Studies
Environmental exposures are not the only source of contact
with pollutants. Indeed, exposure at the workplace can be a
dominant source for some substances. Occupational studies
generally take the form of surveys of effects and exposure
levels. Because it is possible to link exposure and effect more
directly than in environmental studies, occupational investiga-
tions can provide a more accurate estimate of risk. However,
because the concentration levels are generally much higher than
environmental exposures, findings are not directly applicable
to those not occupationally exposed. This problem was discussed
in the sections on health effects and dose-response relationships,
SUMMARY
Table I summarizes the broad areas of exposure and effects
modeling according to the decision-maker's criteria presented
in the introduction to this report. Generally, health effects
are not modeled but are obtained from laboratory or field experi-
ments. Very little sensitivity analysis or cost information is
provided in any of the areas. The human exposure models have
not been well validated; moreover, data availability problems
often permit only highly aggregated estimates. Ecological
modeling is still very much a developing rather than an opera-
tional endeavor.
17
-------
TABLE I
EVALUATION OF EXPOSURE AND EFFECTS MODELS
Type of
Model
Health
Effects
Studies
Eco-
logical
Modeling
Human
Popula-
tion
Exposure
Studies
Applicability
Output units
vary widely,
making it
difficult
to combine
studies.
Methods
applicable
to many
types of
pollutants .
Output units
and species
included
vary widely.
Output rele-
vant though
highly
aggregated.
Accuracy,
Validation
Confidence
limits often
established.
Comparisons
between
studies
sometimes
done.
Usually not
known .
Not known.
Input Data
Available
Not in
model form
so not
require
•input data.
Require
large
amount of
site
specific
data.
Census
data most
often used.
Assumptions
Initial
conditions
and other
factors
not always
reported.
Initial
conditions
not always
reported.
Explained.
Sensitivity
Analysis
Not
usually
done.
Not
usually
done.
Not done .
Cost
Varies
but not
usually
reported
Not
reported
Not
reported
I-1
00
-------
SECTION V
MODELING INTERFACES BETWEEN MEDIA
Models of the environmental fate of substances and of their
transport are used to predict pollutant concentration—a neces-
sary input to the exposure and effects models discussed earlier.
Although relatively advanced, the fate and transport models
typically consider only a single pollutant and a single medium
of transport.
Assessing the single medium fate and transport models of
air and water quality is beyond the scope of this report.
Briefly, however,- models of urban scale (50 km) are the most
advanced of the air quality models. They model the dispersion
process for area and point sources and for reactive and un-
reactive gases. Some allow for variations in meteorology; most
are limited to non-complex terrain. These models are now being
validated with field data, and efforts are being made to simpli-
fy the required data input, depending on the application, and
to assess the models' accuracy and costs. This work is being
done at the EPA Environmental Sciences Research Laboratory,
Research Triangle Park, North Carolina.
Research has also begun on microscale (0 to 500 meters)
and regional scale models. Regional scale models must incor-
porate sink processes and chemical transformations that do not
appear for the urban scale.
Water quality models exist to model transport and trans-
formation of substances from point discharges and non-point
sources, including urban and non-urban runoff. The process
rates differ according to the type of water body (lake, river,
estuary). Most of these models incorporate a mass balance
equation describing flow into and out of a given section of
water. Processes such as chemical and biological transforma-
tions are accounted for as sink processes whose rate constants
are estimated from empirical data specific to the site being
modeled. The models predict concentration levels for water
quality parameters such as dissolved oxygen, nitrogen and phos-
phorus compounds. More recent models incorporate heavy metal
transport and pesticide dispersion.
The air and water quality models were developed in response
19
-------
to a perceived need to control direct emissions or discharges.
However, a source of contamination can also be indirect, as in
an intermedia transfer. Thus, our investigation focused on how
to model these intermedia fluxes in order to predict concentra-
tions in all media of the environment. This report describes
models that simulate the processes at the interfaces, including:
-for the air and water/soil interface
—wet deposition
—dry deposition
—evaporation or volatilization
-for the soil and water interface
—adsorption-desorption
—transport from soil to aquatic system (infiltration
and run-off)
-uptake by biological systems
AIR AND SOIL/WATER INTERFACE
Pollutants emitted directly into the atmosphere are dis-
persed through space over time. There are air quality models
that describe this process and estimate pollutant concentra-
tions in space and time as a function of the emissions distri-
bution and certain meteorological and geographic conditions,
In addition to the dispersion process there are sink mechanisms
which serve to remove the pollutant from the atmosphere al-
together. These removal processes include:
1) chemical reactions or transformations
2) gravitational sedimentation (for particles with
radii greater than 5-10 yum)
3) wet deposition
a) rainout (pollutant first absorbed by cloud then
brought to ground by rain)
b) washout (pollutants collected by falling raindrops)
4) dry deposition (impingement of particles onto elements
of earth's surface as well as absorption of gaseous
compounds)
Because the residence time of most pollutants is much
longer than the time it takes for winds to carry the pollutants
out of the local area under consideration, the effect of sink
mechanisms has not been included when modeling air quality on
an urban scale. Recent efforts to improve air quality models
and extend them to a regional scale (100 to 1000 km) have con-
centrated on including the transformation and deposition pro-
cesses, since sedimentation is negligible for gases and most
traces metals are emitted as air pollutants.
At the interface between air and water or land it is the
deposition process which is at work. Several models now being
developed include deposition.
20
-------
Source Depletion Model
The atmospheric dispersion process can be described by a
Gaussian plume model for a non-depositing material assuming
'reflection1 of the material at the ground to insure conserva-
tion of mass. A source depletion model can account for the loss
of airborne material due to deposition by reducing the source
strength as a function of downwind distance. The deposition
of airborne material is a little understood physical and
chemical phenomenon. In the absence of detailed microphysical
measurements, Chamberlain (1960) suggested the use of the con-
cepts of washout ratio and deposition velocity.
The washout ratio, W, is the ratio of the concentration of
the pollutant in air per kilogram of air and its concentration
in rain per kilogram of rain.c The amount of deposition per unit
area, D , is taken as D = W =• R, where C is the concentration
of the pollutant in airw P is the density of air and R is rain-
fall. Thus, wet deposition is assumed to be proportional to
the local air concentration for a given amount of rainfall.
Generally, washout ratios must be measured at the site in ques-
tion. Some representative washout ratios as a function of par-
ticle size can be found in Gatz (1975).
Similarly, the rate of dry deposition per unit area (D,),
is usually assumed to be directly proportional to the local
air concentration (C) evaluated at a reference height:
Dd = VdC
The constant of proportionality V, has the dimensions of velocity
and has been named the deposition velocity. Several theoretical
models have been developed to predict the exchange rate (or
deposition velocity) of gases at the air-water interface and
will be discussed below. However, measurements of the deposition
velocity are generally required. Gatz (1975) presents deposition
velocities for many trace metals as a function of particle size.
Prahm et al. (1976) summarizes the various measurements taken of
the deposition velocity for sulphur dioxide.
Several aspects of this interface problem have been
examined in more detail. These include the mechanism (source
or surface depletion) used to deplete the Gaussian plume, refine-
ments of the deposition rates to allow for dependence on factors
other than concentration, and attempts to model plant uptake or
the actual mechanisms involved in the deposition process.
Surface Depletion Model
As described above, the source depletion model reduces the
source strength according to the air concentration at a given
distance downwind. The reduction is in effect instantaneously
21
-------
— C(x',z ) D(x-x',z,0) dx'
distributed throughout the vertical extent of the plume. This
produces an artificial vertical mixing of the plume. A more
realistic model has been developed which selectively depletes
the Gaussian plume in the vicinity of the deposition surface
(Horst, 1977) . This surface depletion model allows deposition
only from the lowest layer, eliminating the artificial mixing
of the source depletion model. It predicts longer residence
times for airborne materials. This can be seen from the equa-
tions used in the different models.
Source depletion: C(x,z) = [Q - I VgC(x',z^) dx1] — (x,z,h)
Jo
Surface depletion:
C(x,z) = Q —(x,z,h) -
o u •
where C is the concentration at a distance downwind (x) and at
a height (z), Q is the source strength, h is the source height,
z, is the reference height near the surface, u is the wind speed,
v, is the constant of proportionality with the dimensions of
velocity, and D is the diffusion function with the dimensions of
reciprocal length. Comparing the variables upon which D is
functionally dependent shows that the source depletion model
assumes deposition is a loss at the source rather than at the
surface.
The source depletion model overpredicts the surface air
concentration and deposition close to a source and overestimates
total deposition between source and receptor. However, the
surface depletion model is computationally more complex.
Horst (1977) presents quantitative comparisons to aid in the
choice between the two methods.
Recently, in order to incorporate information on the ver-
tical mixing process (atmospheric stability), a finite dif-
ference box model was developed (Draxler and Elliott, 1977).
The model is Lagrangian in character since the deposition is
computed along a path rather than at a particular location.
Discrepancies similar to those presented by Horst (1977) were
found between the source depletion and the surface depletion
methods. The purpose of the model, however, was to test the
impact of atmospheric stability on deposition rates. Using
hypothetical data, it was found that atmospheric stability at
the time of emission release was one of the primary factors
determining the pollutant's residence time in the air.
Wet Deposition Rates
The application of the washout ratio described earlier de-
22
-------
pended on knowing the total rainfall occurring over the modeled
region and assumed it fell at a steady, uniform rate. Another
approach is assuming that the onset of a wet (or dry) period
occurs with a constant transition probability and to include
the washout ratio function only during a period of rainfall.
This method, proposed by Rodfee and Grandell (1972) and employed
by Fisher (1975), uses the same washout ratio but requires infor-
mation on the mean lengths of wet and dry periods. A sensitivity
analysis by Bolin and Persson (1975) shows that when a smaller
value for the washout ratio is chosen, the deposition due to
rainfall becomes more evenly distributed over a larger area.
Dry Deposition Rates
Using a constant value for the deposition velocity ignores
the dependency of deposition velocity on the height above the
surface, various surface characteristics and atmospheric con-
ditions. However, if a representative value for the deposition
velocity could be determined, the long-range transport models
described above could predict the rate of deposition solely by
predicting a surface layer concentration for the pollutant. For
particles, measurements have been taken to relate deposition
velocity to particle size (Gatz, 1975) and to particle size and
wind speed (Winchester and Duce, 1977).
For gases, models have been developed to estimate deposi-
tion rates using theoretical considerations and laboratory and
field measurements. They predict the rate of transport from the
air to both water and land surfaces (vegetation) and evaluate
some of the factors affecting deposition rates (Wesley and
Hicks, 1977; Bennett et al., 1973; O'Dell et al., 1977;
Shreffler, 1976). These studies are still limited and in the
development stage. Generally, they model mass transfer at the
microphysical stage using a series of resistances analogous
to electrical resistance (Spedding, 1977). A description of
these studies can be found in Table II.
Evaporation or Volatilization Rates
The deposition models we have been discussing are steady-
state models, requiring that the flux of the gas across both
boundary layers (e.g., from air to water and from water to air)
be constant. Thus work in this area has applications for the
measurement of evaporation rates as well as deposition rates.
The process by which a compound is transferred from soil
or vegetation to air is called volatilization and is modeled in
the same way as evaporation. For high-solubility and/or high-
reactivity substances such as SO2 the models predict deposition
rates and the water or vegetation acts as a sink. For low-
solubility compounds, such as pesticides, the models predict
evaporation or volatilization rates from water or land to air.
23
-------
to
TABLE II
STUDIES OF DEPOSITION AND EVAPORATION RATES
Reference
Whelpdale
and Shaw
(1974)
Liss and
Slater
(1974)
Mackay and
Leinonen
(1975)
Hicks and
Liss
(1976)
Smith
et al.
(1977)
Substance
S02 (depo-
sition)
gases
(evapora-
tion)
Low solu-
bility
gases
(evapora-
tion)
SO- and
other
reactive
gases
(deposi-
tion)
Soluble
organic
substances
(volatili-
zation)
Canopy
Type
grass
snow
water
water
water
water
water
Model
Type
bulk
transfer
analogues
to water
vapor
two-film
resistance
model
resistance
model
model of
flow-
gradient
relation-
ship
two-film
resistance
model
Input Data
bulk transfer co-
efficient of water
vapor, average wind
speed, concentration
gradient
empirical estimates
of atmospheric
transfer coeffi-
cients
lab measurements
of mass transfer
coefficients and
Henry ' s Law con-
stants
empirical esti-
mates of relation-
ships
molecular diameter
of substance, oxy-
gen reaeration rate
Results
V
-------
TABLE II - continued
Reference
Farmer
and Letey
(1974)
O'Dell
et al.
(1977).
Wesley
and Hicks
(1977)
SI inn
(1977)
Shref f ler
(1976)
Viebrock
(1977)
Neely
(1977)
Substance
Pesticides
(Volatili-
zation)
Gases
(deposi-
tion)
SC>2 and
similar
gases
(deposi-
tion)
Particles
and gases
(deposi-
tion)
Gases
(deposi-
tion)
Chemicals
Canopy
Type
Soil
Vegeta-
tion
vegeta-
tion
(dense
uniform
water
and
vegeta-
tion
vegeta-
tion
water
Model
Type
diffusion
and mass
flow equa-
tions
resistance
model
resistance
model
semi-
empirical
formulas
for wet
and dry
deposition
resistance
model
based on
results
from Liss
and Slater,
Mackay and
Leinonen
Input Data
vapor pressure of
pesticide, tempera-
ture & wind speed
aerodynamic, sto-
matal and meso-
phyllic resistances
aerodynamic, surface
and canopy stomatal
resistances
much input data
required but not
available
stomatal and aero-
dynamic resistances,
surface concentra-
tion
molecular weight,
temperature, solu-
bility, pressure
Results
volatilization rate
at high wind speeds aero-
dynamic resistance negli-
gible; for highly soluble
gases mesophyllic resis-
tance negligible
Vgj as function of height
of observation, wind
speed, roughness length,
canopy stomatal resis-
tance and Henry's Law
constant
deposition of gases dic-
tated by other than atmo-
spheric processes
flux depends on height,
zero-plane displacement,
roughness, leaf area
density
formula for evaporation
rate constant
to
Ul
-------
These models require measurement of such basic variables as
roughness, length, wind speed, canopy resistance, and Henry's Law
constant, among others. The values of these properties, however,
are rarely available for a large area. Precise values are
usually only obtained by micro-meteorological experiments. On
the macroscale, flux can be calculated directly from field
measurements and deposition velocity estimated from flux and
concentration values. The usefulness of these models lies in
the possibility of deriving generalizations about the likely
values of deposition velocity and the way various factors might
affect it.
Wesley and Hicks (1977) outline one procedure that could
be used in the case of a regional-scale model of S0_ or other
highly soluble gases. They include factors for atmospheric
stability, surface roughness (type of vegetation) and wetness
of exposed surfaces (crop moisture index). These values,
though imprecise, could be obtained from available resources
and might suffice in view of the large inaccuracies resulting
from other assumptions in such a regional model.
SOIL AND WATER INTERFACE
Pollutants deposited on the land, like those emitted to
the air, are subject to transformation and transport, though
the processes may differ. The following processes are known
to influence the fate and behavior of pollutants in soil systems
1) microbiological and chemical reactions (degradation)
2) adsorption-desorption
3) volatilization
4) movement or transport (infiltration, run-off)
5) plant or organism uptake
The last four processes, of course, affect both the original
compounds and the products of the degradation process.
Most studies of these processes have been concerned with
the transport and fate of pesticides or other organic compounds.
The following discussion will rely heavily on the organic com-
pounds literature.
Volatilization was discussed in the previous section and
is the major transfer process at the air-soil interface. The
phenomenon of adsorption-desorption directly or indirectly
influences the magnitude of the effect of the other processes.
It appears to be a major factor affecting the interactions of
pollutant and soil and transfers across the soil-water inter-
face. We will now describe the way pesticide transport models
incorporate the adsorption process.
26
-------
Adsorption-desorption
Adsorption is the attraction of pollutant molecules to soil
particles; the reverse process is known as desorption. How much
of an available substance is adsorbed determines how much can
be transported in solution and in sediment from the soil to the
water phase of a system. A number of factors determine the
extent of pesticide adsorption in a soil-water system. General-
ly, these are related to (1) the nature of the soil, (2) the
nature of the pesticide, and (3) the influence of environmental
conditions. Bailey and White (1970) and Browman and Chesters
(1977) have summarized current understanding and research on
these factors.
Because quantitative relationships describing the effects
of these factors on adsorption have not been developed, pesti-
cide transport models use empirically derived equations relating
equilibrium solution concentration to the amount adsorbed.
Several adsorption equations have been proposed and are de-
scribed by Bailey and White (1970) and by Browman and Chesters
(1977). The Freundlich equation is the one most frequently
applied to natural systems and takes the form — = KG 'n where
— is the amount adsorbed per unit soil, C is tne equilibrium
concentration in solution and K and n are empirically derived
constants. This describes an equilibrium state and assumes
the process is reversible.
It has been shown that some pesticides are so strongly
adsorbed that not all of them will desorb. Even under repeated
washings some portion will remain permanently fixed or ad-
sorbed. Crawford and Donigian (1973) have developed a model
to account for this^irreversible adsorption. Their adsorption
equation is ^ = KG + — where F/m is an empirical term for
the amount of the pesticide adsorbed in a permanent fixed state.
This term, F/m, is less than or equal to the permanent fixed
capacity of the soil for that pesticide. This maximum capacity
is approximated by the cation or anion exchange capacity for the
particular soil type. All available dissolved pesticide is
assumed to be adsorbed into the permanent fixed state until
the maximum is reached. The remaining dissolved pesticide is
then subject to reversible equilibrium adsorption according to
the Freundlich term, KG /n.
This model assumes, therefore, that reversible adsorption
is single-valued. Research has indicated that this assumption
is not valid for many pesticides. Tests of this model have
shown that it did not adequately simulate the division of
pesticides between the sediment and solution phases. This was
especially true for soluble pesticides which also adsorb onto
soil particles.
A newer model by Donigian and Crawford (1976) allows for
27
-------
non-single-valued-adsorption/desorption. In this model the
desorption equation is x/m = K'C ' + F/m. The user must
specify n1 and the model calculates K1 as a function of K, n and
n1, the other adsorption/desorption parameters. Comparison
of simulation results from single-valued and non-single-valued
adsorption/desorption functions with observed data is incon-
clusive. One form performs better for some storms, while the
other form performs better for other storms (Donigian et al.,
1977) .
So far we have been concerned with soil-pesticide inter-
actions at the soil-water interface. Adsorption also governs
sediment-pesticide interactions at the sediment-water interface.
However, in this case adsorption will depend on different fac-
tors, such as particulate matter already present in the aquatic
system, the aerobic-anaerobic cycle, and the degree of mixing
that occurs. However, since the mechanisms underlying the
adsorption process have not been modeled, the Freundlich equa-
tion and its empirically estimated constants are also used in
the case of sediment.
Transport from Soil to Aquatic Systems
During precipitation pollutant residues can be transported
vertically through the soil (infiltration) into the ground-
water or across the soil surface (runoff and sediment loss)
into rivers or lakes. Usually models represent infiltration
by a capacity curve in which the capacity is an exponential
function of time. This assumes that the supply rate (precipita-
tion minus interception plus surface detention) always exceeds
capacity. When this is not the case, variation in infiltration
capacity is controlled by accumulation of soil moisture as well
as time. Crawford and Donigian (1973) describe infiltration
equations that can be used to relate infiltration capacity to
soil properties and time but are still based on empirical
estimation of the soil parameters. They also apply the method
so as to account for variations in soil properties throughout
the watershed modeled.
A considerable number of hydrologic models have been
developed to simulate runoff. The more recent ones include the
infiltration process. Generally, they are intended for applica-
tion either in rural areas (e.g., non-point source pollution
from agricultural lands) or urban situations (e.g., stormwater
runoff to sewer systems). A description of these models will
not be attempted here since they vary widely in scope and
purpose, mathematical detail, data requirements and computer
output. Several papers have reviewed the various models. Perez
et al. (1974) gives a summary of runoff and groundwater models.
Brandstetter et al. (1976) reviews urban runoff models.
Donigian and Crawford (1976) describe an agricultural model of
pesticide transport.
28
-------
Our descriptions of the adsorption, infiltration and run-
off processes have been derived mainly from models concerned with
pesticides or organic compounds. Because these processes are
largely described by empirically determined relationships,
similar models are applied to heavy metals. One example is the
Unified Transport Model (Van Hook and Shults, 1976). In this
model experimentally determined solubility constants form the
basis for estimating losses from litter to infiltration and
surface runoff. Once within a soil layer, exchange of the heavy
metal is assumed to take place instantaneously and the amount
exchanged depends on a measured distribution coefficient. The
use of cation exchange capacity to describe adsorption, however,
is not universally recognized as the dominant process in en-
vironmental fixation of trace metals. Some researchers propose
instead that hydrous oxides of iron and manganese may control
the exchange process (Korte et al. , 1976) . Thus, as in the case
of pesticides, the process of adsorption is not well understood
and may not be well described in current models.
UPTAKE IN BIOLOGICAL SYSTEMS
Plants absorb gaseous pollutants and the process (deposi-
tion) has been modeled as a series of resistances to transfer
across the air boundary layer over the surface of the leaf.
The absorption process from soil and water is usually described
by the use of partition coefficients (concentration in organism/
concentration in soil or water) comparing distribution of the
chemical in one phase to that in another phase.
Such coefficients do not represent absolute distribution
coefficients actually occurring in nature since these are
governed by an enormous range of factors. Even as a relative
measure of distribution they are at best indicators of pollu-
tant behavior rather than precise tools for prediction.
How a compound partitions between water and hydrocarbon-
containing organisms such as aquatic plants and animals is
apparently related to whether it is polar (as judged by its
water solubility) or non-polar (as judged by its fat solubility).
The partition coefficient of the substance between octanol
(a compound midway in the range of fat/water solubilities) and
water is sometimes used as a predictor of relative partition
coefficients. Neely (1977) uses this method. If the octanol/
water partition coefficient is unknown, it is estimated from
the solubility.
Several methods for estimating the absolute value for up-
take of chemicals by plants and fish have been suggested.
Metcalf (1977) uses the pharmokinetics of drug absorption and
metabolism to express the concentration rate as a function of
concentration in water, absorption rate determined by the lipid/
29
-------
water partition coefficient, organism mass or volume and the
rate constant for clearance of the drug by the combination of the
processes of degradation, elimination and growth dilution.
Neely (1977) estimates the rate constant for uptake of a chemi-
cal by a fish as a function of the volume of water flowing past
the gills, the efficiency of pollutant transfer across the gills
and the weight of the fish. Fulkerson et al. (1974) estimates
the rate of solute uptake by a plant from the soil as a function
of root biomass.
SUMMARY
The available models for predicting fluxes at the inter-
faces are summarized in Table III according to the criteria pre-
sented in the Introduction. The models have potential applica-
bility for a wide variety of soils, vegetation, aquatic animals
and pollutants. Input data are most readily available for
modeling the interface with air, and only for this interface has
some validation and sensitivity analysis been done to assess
the models' actual applicability.
30
-------
TABLE III
EVALUATION OF INTERFACE MODELS
Type of Model
Air to Water/
Soil Interface
Soil to Water
Interface
Uptake by
Biological
Systems
Applicability
Covers many
types of
soil,
pollutants .
Organic
compounds ,
some models
for heavy
metals
Wide appli-
cability
possible.
Accuracy,
Validation
Some com-
parisons
with field
data.
Field data
used for
calibration,
not yet
enough for
validation.
Some com-
parisons
with field
data.
Input Data
Available
Meteorolo-
gical data
available.
Other fac-
tors usually
must be
measured.
Available
for simplest
models only.
Not readily
available .
Assumptions
Explained.
Explained.
Explained.
Sensitivity
Analysis
Done for
key
factors .
Not
extensive.
Not done .
Cost
Un-
known .
Un-
known .
Un-
known .
-------
SECTION VI
EXAMPLES OF FATE MODELS
An overall model of the environmental fate of a substance
could be constructed by sequentially linking single-medium
models using the equations describing the processes at the
interfaces that have been presented in the preceding sections.
An example of such a model is the Unified Transport Model
developed at the Oak Ridge National Laboratory. Several other
theoretical models of environmental fate have also been de-
scribed, although their applications have been limited to a
single medium.
UNIFIED TRANSPORT MODEL
A team of researchers at the Oak Ridge National Laboratory
has developed a system of models whose purpose is to simulate
the movement of a trace contaminant in and through a watershed.
In concept, submodels that simulate the transport and fate of
the substance in a single medium can be linked together using
other submodels which simulate activity at the interfaces. The
coupled submodels are called the Unified Transport Model.
Several examples of the models describing the interfaces have
been referred to previously in this report. The following
sections describe the overall model, its development and appli-
cations .
Theory
The Unified Transport Model (UTM) is a physically based
model designed to simulate the transport, accumulation and
redistribution of trace contaminants in the environment. The
air, land, and aquatic systems are represented by three major
model components, designed to be run in sequence. The focus of
the UTM has been on watershed phenomena that are important in
trace contaminant mobility and pathways including atmospheric
dispersion, deposition on land, entrainment and erosion in over-
land flow, movement and exchange of soluble chemical species
with the moisture flux through the soil matrix, transport in the
stream channel and exchange with the stream sediment. Recent
refinements of the model have included chemistry of sulfur in
the atmosphere, sediment transport and soil chemistry along with
32
-------
plant uptake and vegetative growth. These additions extend the
purely physical framework to include some ecological processes.
The development of the UTM began with a search for existing
models that could be modified to provide appropriate data to
be run sequentially. The process of linking is complicated by
differences in scale and accuracy between the various media
models. Differences in scale reflect not only the state-of-the-
art of modeling but also the inherent differences in processes
which are important in the various media. Considerable time was
spent making the necessary compromises.
Fulkerson et al. (1974) describes the sources of the models
and the modifications= No comprehensive models were available
for sediment transport, plant uptake and atmospheric chemistry;
the new models developed for these phenomena at Oak Ridge are
described by Van Hook and Shults (1976).
Designed for flexibility in a wide variety of applications,
the UTM is actually a collection of compatible but distinct,
stand-alone submodels. Figure 1 illustrates the suite of sub-
models that may be used in assembling a model to address a
specific problem. The aim has been to reduce the complexity
of the modeling system required for a particular problem,
thereby reducing resource costs and increasing user understanding
of the results.
As the UTM has become more detailed, individual submodels
have been divided into subroutines representing specific physi-
cal processes; users should thus be better able to understand
the functioning of the model and to make substitutions for
particular algorithms. In addition, this type of structure
can be modified as the state-of-the-art advances or reassembled
for special applications not necessarily foreseen in the
original formulation.
Applications and Validation
Two watersheds, Walker Branch and Crooked Creek, have been
the focus of applications of various versions of the UTM and
its submodels. Data from these watersheds have been collected
for the specific purpose of validating these models. The air
transport submodel (ATM) has been used to estimate deposition
and air concentration of some 19 trace metals present in the
Walker Branch watershed near Oak Ridge, Tennessee. For 16 of
the metals, the simulated results were within the measured con-
centration ranges. The hydrology transport model (WHTM)
estimated stream flows for the summers fairly well in this area
but underestimated the heavier winter and spring flows. The
hydrology transport model was also applied using the optimiza-
tion routine (OPTRM), which provided simulated results closer to
the measured values.
33
-------
ORNL-DWG 75-15808
PRECIP
PRECIPITATION
DATA MGMT.
ATM
ATMOSPHERIC
TRANSPORT MODEL
SULCAL: SULFUR CONCENTRATIONS AND DEPO-
SITIONS FROM STACK EMISSIONS
GENCRD: GRID OF DEPOSITION POINTS
FOR OBTAINING AREA WEIGHTED
DEPOSITION RATES
LAND
(UNIT AREA TERRESTRIAL RESPONSE)
SNOMLT: SNOW PACK ENERGY BALANCE AND MELT
HTM: PARAMETRIC RUNOFF MODEL ION EXCHANGE OF CONTAMINANTS,
SOIL AND CONTAMINANT EROSION
SOLAR: SLOPE AND ASPECT ADJUSTMENTS TO RADIATION
PROSPR: SOIL PLANT WATER ATMOSPHERE DYNAMICS, EVAPORATION
SCEHM: SOIL CHEMICAL EXCHANGE OF HEAVY METALS
CERES: PLANT GROWTH, TRANSLOCATION
DRYADS: LEAF AND ROOT UPTAKE
DIFMAS: MASS FLOW AND DIFFUSION TO ROOTS
ODMOD: ONE DIMENSIONAL ANALYTIC MODEL FOR SOIL WATER
POTENTIAL TRACE CONTAMINANT CONCENTRATION
SUBSRF: HYDROLOGIC SOURCE AREAS, SATURATED AND UNSATURATED
DRAINAGE AND GROUNDWATER FLOW
1HNSED:
SEDTRN:
PNTSRC:
CHANL
STREAMFLOW HYDRAULICS AND TRANSPORT
SEDIMENT AND CONTAMINANT EXCHANGE AND TRANSPORT
SUSPENDED AND BED-LOAD TRANSPORT AND COMPOSITION
POINT SOURCE DISCHARGE INPUTS
OPTRM
WHTM OPTIMAL PARAMETER
SET DETERMINATION
Figure 1. The submodels that may be linked to form a
unified transport model.
Source: Van Hook and Shults (1976), p. 16.
34
-------
In the Crooked Creek watershed in Missouri, the UTM has
been used to model the fate of various metals. Two different
versions of the model, both with the same air (ATM) and stream
channel routing systems, but with different submodels (WHTM or
TERM) for the land portion of the system, have been developed.
WHTM uses considerably less input data and simulates water
quality of streams and buildup of toxic substances in the soil.
TEHM provides additional information on the effects of toxic
metals on plant growth and their accumulation in plant materials
Runs of these two versions of UTM are compared to each other as
well as to measured data (Munro et al., 1976) in order to dis-
cover remaining problems in the formulation of the model.
These problems include the estimated rate of deposition
(too low), the lack of differentiation between complexed and
free heavy metals (the model predicts best those metals re-
maining as free cations in the soil solution), and the lack of
sufficient input data to adequately characterize the system
(sensitive parameters must be estimated).
Input
The input requirements will, of course, depend on the
particular problem and set of models used. Table IV lists the
input data required for the two major components of the UTM.
Sensitivity Analysis
Some submodels have been subjected to various sensitivity
analyses. Usually a series of calculations is made to show how
various output quantities change as input quantities are varied
in ranges thought to be realistic. These sensitivity analyses
serve several functions. First, many input items are hard to
measure. One of the submodels (OPTRM) was developed to optimize
or search for the best parameter values for the particular appli-
cation. This calibrates the model to the specific watershed
being studied. The sensitivity of the simulated variable to
various parameters guides the choice of the set of parameters
to be optimized. Second, researchers often use the model in
developing and testing hypotheses. Changes in output values
point to which pathways or processes may be dominant for any
given situation. Third, when input data are unavailable for a
particular application, they may be estimated from the results
of the sensitivity studies.
Cost and Documentation
Each submodel has been well documented. The documentation
generally includes a description of the input data required,
where it was obtained by the researchers, the structural equa-
tions used, a flow diagram, the FORTRAN IV listing, an example
of the output produced and a description of applications of the
35
-------
TABLE IV
INPUT REQUIREMENTS AND OUTPUT DATA
FOR TWO COMPONENTS OF THE UNIFIED TRANSPORT MODEL
ATMOSPHERIC TRANSPORT MODEL
INPUT: type of material (particle or gas), density, size, wind speeds,
wind directions, atmospheric stability, geographic location,
type and strength of emission sources, height of stack,
plume rise factor, hourly precipitation
OUTPUT: wet and dry deposition amounts, surface concentration
WISCONSIN HYDROLOGIC TRANSPORT MODEL
INPUT: hourly precipitation, daily maximum and minimum temperatures,
daily pan evaporation rates, hourly temperature and humidity,
daily precipitation, solar radiation; upper and lower zone
soil storage, soil moisture evaporation index, time distri-
bution parameter for interflow, average length of overland
flow, average slope of overland flow plane, interflow re-
cession constant, groundwater recession parameter, impervious
area and other parameters describing land features; residual
toxicant mass distributed on land surface, ion-exchange dis-
tribution coefficient, deposition amounts, toxicant content
of ion exchange plates and other parameters describing toxi-
cant properties.
OUTPUT: daily stream flow, total metal transported, daily average
metal concentration
Abstracted from Munro (1976).
36
-------
model. In addition, Van Hook and Shults (1976) briefly de-
scribe some of the steps in implementing the model that may
prove costly and time-consuming.
CONCEPTUAL MODEL
Ecosystem modeling has been characterized by the collection
of voluminous but not always relevant data. Although the
environmental fate of some compounds (primarily pesticides,
e.g. DDT) has been successfully modeled in this way, it is a
highly inefficient hit-or-miss method. Thousands of chemicals
are now manufactured, used and ultimately disposed of in the
environment. Managing the dispersal of these chemicals would
be easier if there were simpler, more reliable ways of evalu-
ating their potential hazard. An explicit model derived from
the principles of chemistry, physics and biology and valid for
all chemicals and all environments is not possible given the
bounds of current knowledge.
Instead a procedure for deriving a conceptual model for
the movement of chemicals through the environment is proposed
by Gillett et al. (1974) . The purpose of this model is to
force rigorous thinking about what processes and components of
the environment are involved. The conceptual model starts with
a diagrammatic representation of the processes and components
and their possible interactions. For instance, if the flow of
a pesticide through a terrestrial system is being investigated,
the components would include biota and soil and the processes
would include volatilization from soil, adsorption-desorption
onto soil and loss to biota, among others.
The model is refined through an iterative process of reduc-
tion and expansion of components. In examining the effect of
root growth on the physical structure of the soil, for instance,
the biota component might be divided into root and stem systems
or the soil into several layers. If it turned out that this
expansion of the model did not alter the pattern of movement of
the pesticide through the soil, then the model could be reduced.
The root and stem components could be recombined into one biota
component or the number of soil layers could be reduced without
affecting the capacity of the model to predict pesticide trans-
port.
Another example of the iterative process of reduction and
expansion will be presented more fully below in the discussion
of the environmental rates approach. Briefly, this procedure
evaluates process-system interactions, providing criteria for
the elimination of components that have the least effect on
system behavior. The operation of the thus simplified model
may in turn point out other components that should be expanded
to model the particular problem more adequately.
37
-------
The aim is to put bounds on the types of measurements
needed, the accuracy required of the measurements and the fre-
quency of sampling. An example of this procedure is given by
Haefner and Gillett (1976). This approach is still in the
conceptual stage, with current efforts focused on deriving a
functional mathematical statement for simulation experiments
and better defining measurement needs.
MASS BALANCE APPROACH
A mass balance study essentially accounts for all sources,
pathways, transformations and sinks for a particular pollutant
in a given area. The mass balance approach traces the flow
of a pollutant through an area. Each environmental pathway is
characterized as a chemical, physical or ecological interaction
and by a mass flow rate. The assumption is that the flow rates
can be estimated separately. Then the analysis requires a
satisfactory mass balance; that is, the sum of input flows should
equal the sum of output flows and stored quantities. This
approach can be applied to all substances and media as long as
the important environmental processes and input and output
characteristics are known.
Huntzicker et al. (1975) reports on applications of this
methodology in the Los Angeles area. Applications have been
primarily limited to determining fate of airborne trace metal
pollutants. The input, flow rate and output data required for
this type of analysis are shown in Table V. Problems encoun-
tered in these examples include an inaccurate emissions inventory
resulting in uncertain input flows as well as a lack of theoreti-
cal or experimental understanding of particular removal processes
in the environment. Given these uncertainties and the large
geographical expanse of the study areas, the method does not
provide detailed predictions of pollutant dispersion. It does
provide estimates of the relative flow of mass for each source
and sink identified.
Most of the studies are site-specific (e.g., Los Angeles
air basin); most model substances currently being emitted into
the atmosphere and use data collected in the field. However,
one study forecasts the flow of manganese as a substitute for
lead in gasoline. Because field data for this new substance
were not available, input data were estimated from lead flows,
because the two substances have similar particle size distribu-
tions in automobile exhaust.
A similar approach to forecasting the fate of substances
not widely dispersed in the environment is described in the
next section. The materials balance equations are still used
but the source of flow rates differ. Also, the primary emphasis
38
-------
TABLE V
MASS BALANCE APPROACH
EXAMPLE OF INPUT AND OUTPUT DATA
FOR AIRBORNE TRACE METAL POLLUTANTS
Source or Input Data
1. Identity of sources
2. Mass emission rates for each source
3. Aerosol size distribution for each source
4. Deposition rate (dry)
5. Rainout-washout rate (wet)
Output Data
1. Atmospheric flushing (ventilation) rates
a. Receptor site atmospheric concentration
b. Volumetric flow through air basin
2. Inputs to coastal waters, lakes, rivers, etc.
a. Deposition (dry)
b. Rainout-washout
c. Runoff
d. Sewage
Source: J. Huntzicker, S. Friedlander and C. Davidson (1975), p. 5.
39
-------
has been on applications in aquatic environments rather than in
the atmosphere.
ENVIRONMENTAL RATES APPROACH
This approach to predicting the fate of a substance in the
environment uses the results of laboratory measurements of
specific physical, chemical and biological processes in a
computer model that integrates the data with transport para-
meters. The laboratory data are collected in the form of rate
constants. These rate constants are then used in the material
balance models to produce time-varying concentrations. All of
the rate processes are described by expressions involving a term
for concentration, and each rate is a function of some property
of the environment such as pH, turbulence, and light intensity-
This approach essentially considers what the environment does
to the substance rather than the reverse. Thus, the importance
of environmental pathways corresponding to a particular process
is determined by its effect on the net rate of loss or accumula-
tion of the substance.
The temptation in seeking to eliminate the least important
variables is to focus on very rapid processes (which cannot be
rate limiting) and on very slow processes (which cannot trans-
port must matter). However, the environmental rates approach
shows that the effect of a process is dependent on other factors
as well as the rate coefficient and provides a technique for
estimating this effect.
The fundamental assumptions on which this approach is based
are that
(1) the overall rate of disappearance of a substance
is controlled by the dominant transformation and
transport processes,
(2) these processes can be studied separately in the
laboratory, and
(3) the laboratory data can be extrapolated to environ-
mental conditions.
Unlike the site-specific transport models we have dis-
cussed earlier, this approach hypothesizes an idealized environ-
ment. It can evaluate the influence of properties of the
environment by varying these properties and recording the
resulting variation in the fate of the substance. Thus, the
input data, are usually specified rather than measured. This
approach permits generalizations based on both chemical and
environmental properties. It is not intended to predict detailed
fate in specific environments, but the results can be used to
40
-------
assess the need for more detailed studies. This lack of
specificity is not necessarily a disadvantage. Where new
substances have not yet been introduced into the environment,
it is still possible to study their probable fate or at least to
determine what further tests are needed. Also, for substances
that are not widely dispersed and for which the collection or
monitoring or field data may not be practical, this approach
could be used to allocate monitoring resources more efficiently.
The environmental rates approach predicts concentrations
as a function of time. The usual model ecosystem studies, in
contrast, report equilibrium or single time values. Although
microcosm studies yield an overall assessment of complex inter-
actions in an environment, they do not provide a basis for
extrapolation. The environmental rates method, by providing
relative rates of many processes, provides a basis for extra-
polation and a fundamental understanding of what happened to
the substance.
Output
Several kinds of results can be obtained from this type of
model. First, the distribution within the environment for
given source characteristics is given in the form of concentra-
tion or mass of the pollutant. Steady-state concentrations are
measures of exposure for organisms living in the aquatic environ-
ment. Second, dominant pathways are found as a function of the
environment's properties. If the products that result from a
particular transformation are known, then examination of the
dominant pathways will identify what products should be present.
Third, the model can be used for ranking chemicals from the
point of view of chronic exposure. Together with toxicity data
the predicted concentrations could be used to produce relative
hazard rankings. Fourth, the relative patterns of concentration
in different parts of the environment may suggest areas that
need further study- If a chemical largely dissipates into the
atmosphere, then photo-degradation studies are indicated. On
the other hand, if there seems to be an accumulation in fish,
fish metabolism might be studied.
Validation
Knowledge of environmental processes and methods for pre-
dicting their rates is not extensive. Research by Smith et al.
(1977) has focused on this problem for the aquatic environment.
The principles, however, are applicable for any compound or
medium. A potential weakness of this approach is that it may
not incorporate all important transport or transformation pro-
cesses that occur in a natural system. Thus, validation of
this model should be done by comparing concentrations in
laboratory microcosms or, if possible, natural systems. This
sort of validation cannot prove that the model includes all
41
-------
important processes. There is no way to test this mechanically.
However, it can show the model to be inadequate in comparison
with the experimental or field data and that further model
development is needed.
Operational models incorporating this approach are few.
One of the earliest was an application to air pollution
(Friedlander and Seinfeld, 1969). More recently, they have been
applied to transport and transformation of substances in water.
Hill et al. (1976) examines the behavior of vinyl chloride and
Lassiter et al. (1976) describes the fate of mercury. Neely
(1977) presents a computer program for predicting environmental
exposure concentrations of a neutral organic chemical. This
model has been constructed so as to require as a minimum data
base, the aqueous solubility, the vapor pressure "and molecular
weight of the chemical. It will then generate rate constants
for characterizing the various movements of the chemical from
water to soil, to air and to fish. Smith et al. (1977) de-
scribes a compartmental model of a freshwater environment
incorporating the transport and transformation processes studied
in the laboratory, the laboratory procedures they have developed
for measuring the rate constants and the results of using this
model to study eleven chemicals.
Applications of this method has not been extensive, and
there are few material balance studies in the field to be used
in verification of this approach. Also, some of the assumptions
explicit in the models are poorly substantiated or understood at
this time; further applications might provide insights into
their consequences or limitations.
Input
Rate constants for the processes included in any particular
application are estimated in the laboratory. Smith et al. (1977)
describes how this is done for physical (solubility in water,
absorption spectra, volatilization, sorption partition coeffi-
cients) , chemical (photolysis, free radical oxidation, hydro-
lysis) and biological (biodegradability) properties of an aquatic
environment. Neely (1977) describes how to estimate the evapora-
tion rate constant, soil uptake, soil release, fish uptake and
fish excretion from basic properties of the system if experimen-
tal data are not available.
Sensitivity Analysis
Branson (1977) presents results obtained by varying some
of the rate constants. The hydrolysis rate, bioconcentration
uptake rate and water solubility are varied by approximately
+10% and compared to the resulting variation in concentration
of the substance in fish and in water.
42
-------
SUMMARY OF FATE MODEL FINDINGS
The environmental fate models described in this report
could theoretically be applied to a variety of environmental
conditions and types of pollutants. However, in practice their
applications are limited. Sensitivity analyses have been per-
formed for several models, prompted primarily by the lack of
available input data. Through sensitivity analysis it is
possible to determine which kinds of inaccuracy in the input
data significantly influence the output results and which are
less important. Table VI presents a summary description of
these fate models according to the criteria listed in the
introduction to this report.
No matter how sophisticated modeling becomes, an under-
standing of environmental fate and predominant pathways alone
cannot be a sufficient guide to potential risk. For instance,
it might be established that a substance volatilizes rapidly,
implying that the largest concentrations will be found in the
air medium. However, if investigation were limited to its
chemical transformations and transport in air, its highly toxic
properties at very low concentrations in water might be over-
looked.
43
-------
TABLE VI
EVALUATION OF FATE MODELS
Type of
Model
Unified
Transport
Model
Con-
ceptual
Model
Mass
Balance
Approach
Environ-
mental
Rates
Approach
Applicability
Transport of
heavy metals
in all media.
Wide applica-
bility pos-
sible.
Wide applica-
bility pos-
sible.
Wide applica-
bility pos-
sible.
Accuracy,
Validation
Two applica-
tions and
comparisons
with field
data.
No applica-
tions.
Applications
for trace
metals in
air; compari-
sons with
field data.
Applications
for organic
compounds in
aquatic
systems .
Input Data
Available
Large amount
of data re-
quired , some
not readily
available.
Unknown .
Available
but emission
rates not
very ac-
curate.
Available
from
laboratory
measure-
ments .
Assumptions
Well docu-
mented .
Would be
described
for each
application
Explained.
Well
documented.
Sensitivity
Analysis
Done for
key factors,
has optimi-
zation
routine.
None.
Done for
input
emission
rates.
, Not
extensive.
Cost
De-
scribed.
Unknown .
Unknown .
Unknown .
-------
SECTION VII
EXAMPLES OF EXPOSURE ASSESSMENTS
The previous sections have described the development of
models to estimate pollutant concentrations, exposure and
effects relationships and size of exposed population. However,
few quantitative exposure assessments have actually been made.
An ideal model applicable to most substances would include:
provision for multiple emission sources
conversion of emission rates to contaminant
concentrations in all media (air, water, soil, food)
provision for exposure to humans, plants and animals
provision for multiple routes of exposure to humans
(inhalation, ingestion, transpiration)
allowance for interactive effects (synergistic or
antagonistic) of multiple pollutants.
Our review of the models developed so far suggests that we cannot
expect to discover a well-developed, operational exposure assess-
ment model meeting all these conditions. Nevertheless, regula-
tory decisions must be made and exposure or risk assessments
have been done. Table VII summarizes several such risk assess-
ment studies.
These studies have concentrated on human health effects,
air pollutants, exposure through inhalation only, and one
pollutant without accounting for synergistic effects. Although
some do propose methodologies to cover multiple exposure routes
and media, their applications revert to single exposure routes
and media because the data necessary to extend the analysis are
not available. In addition, only a few of the studies address
the problems of accuracy, validation and sensitivity analysis.
No study states its cost.
However, each study does provide at a minimum estimates for
size of exposed population, exposure level and resultant risk.
Differences in the available input data and the purpose of the
analysis have led to different methods of calculating risk. It
appears possible that estimates of accuracy, even if only as
orders of magnitude, and sensitivity analyses of dose-response
assumptions could have been made in most instances.
45
-------
TABLE VII
RISK ASSESSMENT STUDIES
Reference
Kuzmack
and
McGaughy
(1976)
Albert
(in
draft)
Albert
(1977)
Model Type \
Assumptions
Links exposure,
concentrations
and effects.
Vinyl chloride.
Inhalation
route only.
Links studies
of exposure
and effects.
Benzene. In-
halation
route only.
Census study of
farm workers ex-
posed to herbi-
cides. Assumes
all inhaled is
absorbed. In-
halation route
only.
Input Data
Size of
population.
Air diffu-
sion model .
No . and
location of
emission
sources .
Linear, non-
threshold
dose-response
model. Pro-
duction data .
Dispersion
model . Popu-
lation den-
sity. Epi-
demiological
studies.
Monitoring
data on con-
centration
levels. Sur-
vey estimates
of exposure.
Animal stu-
dies for dose-
response .
Dutput Data
Cases of ill-
ness per year
of exposure
For various
sources of
Denzene :
lifetime risk,
expected no .
of cases per
year
No. of cases
of cancer
from life-
time exposure
for various
types of con-
trol programs.
1
Accuracy
-10% to
+55%
Several
orders
of mag-
nitude
Consider-
able un-
certainty.
Validation
Compares
animal
study
results
to human
data
Compares
results to
occupa-
tional
study.
None
Sensitivity
Analysis
Different
dose-re-
sponse
curves
used.
None
None
-------
TABLE VII - continued
Reference
EPA
(1977)
Finklea
et al.
(1977)
Walsh
et al.
(1977)
Model Type
Assumptions
Link emissions
to concentration
to exposure to
effects. For
cadmium and
asbestos. In-
cludes sources
from air, water,
food & tobacco .
Links census
data to air
monitoring
data for air
pollutants .
Assumes best
j udgment
threshold
for dose-
response
curves .
Sums over expo-
sure routes
(air, water, food)
for total body
dose. Theory
covers multiple
pollutants but
data does not.
Input Data
Emission
rate, back-
ground con-
centration
in food,
body burden
model, dose
response
curves ,
population
density and
mobility.
Census popu-
lation data.
Air monitor-
ing data.
Dose response
curves .
Transport
models .
Physiologic
models of
uptake .
Output Data
Decrease in
life ex-
pectancy.
Rate of
illness .
Excess deaths.
Excess days of
illness per
susceptible
population
groups due to
not meeting
standards .
Index of dose
divided by
safe limit.
Effluent
limit so as
not to exceed
safe limit.
Accuracy
Confi-
dence
limits
for in-
put data
& dose-
response
curves .
Qualita-
tive
state-
ments of
uncer-
tainty .
Depends
on input
models .
Validation
None
Compares
best judg-
ment
thresholds
to least
squares
fit.
Depends
on input
models .
Sensitivity
Analysis
None
None
None
-------
SECTION VIII
OTHER USES FOR THE MODELS
The risk assessments discussed above concern pollutants
that are present in the environment and many of the environ-
mental fate models depend on measurements done in" the field.
A major source of uncertainty in the applications of these models
is the lack of measurements for input data. These models can
be used, then, to indicate what input data are needed but not
now routinely collected. A conceptual framework, called Inte-
grated Exposure Assessment Monitoring, has been proposed which
in conjunction with these models would optimize collection of
these data.
Resources for monitoring and modeling, however, are not
unlimited. Ranking (screening, indexing) schemes have been
proposed to set priorities among substances needing EPA regula-
tory attention. Modeling can be useful in evaluating the
hazards posed by substances already dispersed in the environment
or the potential risks in new substances being proposed for use.
These two uses of exposure assessment data are discussed
briefly below.
INTEGRATED EXPOSURE ASSESSMENT MONITORING
Integrated Exposure Assessment Monitoring is a concept being
developed by EPA's Environmental Monitoring and Support Labora-
tory in Las Vegas. It provides a framework for the systematic
collection and coordination of pollution data in air, water,
land and food. The idea is to optimize the use of monitoring
resources by focusing on measurements that are directly related
to critical receptors (the sub-group of humans, animals or
plants showing effects at a dose lower than any other group) and
critical sources (major contributors to expsoure of the critical
receptors). This approach differs from past EPA monitoring
efforts by focusing on total exposure to a pollutant rather than
on exposure from a single medium. Current EPA monitoring systems
involve single-medium trend monitoring or source monitoring.
This type of data is not adequate to address the problems of
acute or chronic health effects that may result from short-term
peak exposure, repeated short-term high exposures or low-level,
48
-------
long-term exposures, or the problem of synergism or the latency
period between exposure and illness. Current monitoring systems
measure intensity in one medium only and thus cannot generally
be related to other media. Data identifying environmental
characteristics, pathways, chemical and physical form, frequency,
duration and intensity are needed for an exposure assessment.
The exposure assessment monitoring concept has been applied
to lead. Jenkins (1976) finds that in spite of the voluminous
data available on the distribution and effects of lead, the
information is generally not in a form that can be used to
quantify the contribution of sources and pathways. He then
investigates what mathematical models might be available to
translate measurable data into information appropriate for a
total exposure assessment. Such models can be of value in
helping identify the relative importance of various sources,
pathways and exposure routes. Pollutant transport and trans-
formation models as well as models for distribution of lead in
the body were investigated. The result of this analysis was
to show the use and limitations of currently available data
and the relative importance of obtaining various kinds of
required missing data.
This approach, then, provides a framework for logically or-
ganizing existing information and determining what kinds of
missing information are needed. However, the lack of appropriate
data and models hinders the actual operation of such a system.
The approach has several other pertinent limitations. Each
pollutant would potentially require a different monitoring sys-
tem. Control of the critical source, as here defined, may not
necessarily eliminate the effect in question. Finally, the
focus on critical receptors does not define an overall popula-
tion exposure or risk.
RANKING SCHEMES
Ranking schemes are used for priority ranking of environ-
mental pollutants needing EPA regulatory attention. The same
methodology used to rank suspected pollutants can also be used
to evaluate the potential hazard of a substance not yet intro-
duced into the environment, though in this case little informa-
tion may be available. Because the decision-maker must initially
decide j.ust what types of information are important to him,
every ranking scheme should explicitly list the information re-
quired.
Methods of organizing this information for use by the
decision-maker range from a list subject to expert judgment
through information screening and ordering to a formal model
defining hazard.
49
-------
A model-based system described by Brown and Holt (1976)
would consist of six compartments to be modeled:
1) source: human and natural production
2) distribution: uses, disposal, releases to air, water,
land
3) fate: transport, transformation, exposure
4) effects: dose-response relationships to estimate risk
5) valuation: assign value weights to effects
6) ranking: derive aggregate environmental hazard index
and rank substances with respect to other substances.
The models discussed in this report would provide informa-
tion for the fate and effects compartments.
50
-------
SECTION IX
REFERENCES
1. Albert, R., Carcinogen Assessment Group's Preliminary
Report on Population Risk to Ambient Benzene Exposures,
Environmental Protection Agency, Washington, D.C. (in
draft).
2. Albert, R., Carcinogenic Risks of Contamination of the
Herbicides Treflan, Trysben and Benzac, Environmental
Protection Agency, Washington, D.C., May 1977.
3. Atkeson, T., "Proof of Effects in Environmental Law"
working paper for Effects of a Polluted Environment;
Research and Development Needs, National Research Council,
September 1976.
4. Bailey, G. and White, J., "Factors Influencing the Adsorp-
tion, Desorption and Movement of Pesticides in Soil,"
Residue Review, Volume 32, (1970), 29-92.
5. Bennett, J., A. Hill and D. Gates, "A Model for Gaseous
Pollutant Sorption by Leaves," Journal of the Air Pollution
Control Association, 23 (1973), pp. 957-962.
6. Bledsoe, L., "Linear and Nonlinear Approaches for Eco-
system Dynamic Modeling" in B. Patten (ed.), Systems
Analysis and Simulation in Ecology Volume IV, Academic
Press, New York, 1976, pp. 283-298.
7. Bolin, B. and C. Persson, "Regional dispersion and deposi-
tion of atmospheric pollutants with particular application
to sulfur pollution over Western Europe," Tellus 27 (1975),
p. 281-310.
8. Brandstetter, A., R. Field and H.C. Torno, "Evaluation of
Mathematical Models for the Simulation of Time-varying
Runoff and Water Quality in Storm and Combined Sewerage
Systems," in W.R. Ott, Environmental Modeling and Simula-
tion, U.S. Environmental Protection Agency, EPA 600/9-76-
016, July 1976, pp. 548-552.
51
-------
9. Branson, b., Predicting the Fate of Chemicals in the Aquatic
Environment from Laboratory Data, Dow Chemical Co., Midland,
Michigan, 1977.
10. Browman, M. and G. Chesters, "The Solid-Water Interface:
Transfer of Organic Pollutants Across the Solid-Water Inter-
face" in I.H. Suffet (ed.), Fate of Pollutants in the Air
and Water Environments, Part I, 1977, pp. 49-106.
11. Brown, S. and B. Holt, Systems for Rapid Ranking of Environ-
mental Pollutants, Stanford Research Institute, August 1976.
12. Chamberlain, A.C., "Aspects of the Deposition of Radio-
active and other Gases and Particles," in E.G. Richardson
(ed.), Aerodynamic Capture of Particles, pp. 63-88,
Pergamon Press, Oxford, 1960.
13. Cornfield, J. "Carcinogenic Risk Assessment," Science, 198
(1977), pp. 693-699.
14. Crawford, N. and A. Donigian, Jr. Pesticide Transport and
Runoff Model for Agricultural Lands, U.S. Environmental
Protection Agency, EPA 660/2-74-013, December 1973.
15. Donigian, A., Jr., D. Beyerlein, H. Davis, Jr., N. Crawford,
Agricultural Runoff Management (ARM) Model, U.S. Environ-
mental Protection Agency, EPA 600/3-77-098, August 1977.
16. Donigian, A., Jr. and N. Crawford, Modeling Pesticides and
Nutrients on Agricultural Lands, U.S. Environmental Protec-
tion Agency, EPA 600/2-76-043, February 1976.
17. Draxler, R. and W. Elliott, "Long-Range Travel of Airborne
Material Subjected to Dry Deposition," Atmospheric Environ-
ment, 11 (1977), pp. 35-40.
18. Environmental Protection Agency, Hazardous Wastes: A Risk-
Benefit Framework Applied to Cadmium and Asbestos, EPA
600-5-77-002, February 1977.
19. Environmental Protection Agency, Workshop on Health Effects
of Transportation-Related Pollutants, EPA 600/1-78-011,
January, 1978.
20. Farmer, W. and J. Letey, Volatilization Losses of Pesticides
From Soils, Southeast Environmental Research Laboratory,
Athens, Georgia, EPA 660/2-74-054, August, 1974.
52
-------
21. Finklea, J., C. Shy, J. Moran, W. Nelson, R. Larsen and
G. Akland, "The Role of Environmental Health Assessment,"
in J.W. Pitts, Jr. and R.L. Metcalf (eds.) Advances in
Environmental Science and Technology, Volume 7, 1977,
pp. 315-389.
22. Fisher, B.E.A., "The Long Range Transport of Sulphur
Dioxide," Atmospheric Environment 9 (1975), pp. 1063-1070.
23. Friberg, L. (ed.), Toxicology of Metals, Volume I, U.S.
Environmental Protection Agency, EPA 600/1-76-018, 1976.
24. Friedlander, S. and J. Seinfeld, "A Dynamic Model of
Photochemical Smog," Environmental Science and Technology,
3 (1969) pp. 1175-1181.
25. Fulkerson, W., W. Shults, and R. Van Hook, Ecology and
Analysis of Trace Contaminants, Progress Report October
1973-September 1974, Oak Ridge National Laboratory, Oak
Ridge, Tennessee, ORNL-NSF-EATC-11, December, 1974.
26. Gatz, D.F., "Pollutant Aerosol Deposition into Southern
Lake Michigan," Water, Air, and Soil Pollution, 5 (1975),
pp. 239-251.
27. Gillett, J., J. Hill IV, A. Jarvinen, and W.P. Schoor,
A Conceptual Model for the Movement of Pesticides through
the Environment, National Environmental Research Center,
Corvallis, Oregon, EPA-660/3-74-024, December, 1974.
28. Haefner, J. and J. Gillett, "Aspects of Mathematical Models
and Microcosm Research," in W.R. Ott (ed.) Environmental
Modeling and Simulation, U.S. Environmental Protection
Agency, EPA 600/9-76-016, July, 1976, pp. 624-628.
29. Hicks, B.B. and Liss, P.S., "Transfer of S02 and other
reactive gases across the air-sea interface." Tellus, 28
(1976), p. 348-354.
30. Hill IV, J., H. Kollig, D. Paris, N. Wolfe and R. Zepp,
Dynamic Behavior of Vinyl Chloride in Aquatic Ecosystems,
Environmental Research Laboratory, Athens, Georgia, EPA
600/3-76-001, January, 1976.
31. Horie, Y. and A. Stern, Analysis of Population Exposure
to Air Pollution in New York-New Jersey-Connecticut Tri-
State Region, U.S. Environmental Protection Agency, EPA
450/3-76-027, March 1976.
53
-------
32. Horst, T.W., "A Surface Depletion Model for Deposition
from a Gaussian Plume," Atmospheric Environment, 11 (1977),
pp. 41-46.
33. Huntzicker, J., S. Friedlander and S. Davidson (eds.),
Air-Water Land Relationships for Selected Pollutants in
Southern California, California Institute of Technology,
Pasadena, California, August, 1975.
34. Jenkins, D. (ed.), Design of Pollutant-Oriented Integrated
Monitoring Systems, U.S. Environmental Protection Agency,
EPA 600/4-76-018, April, 1976.
35. Korte, N. , J. Skopp, W. Fuller, E. Niebla and B. Alesh,
"Trace Element Movement in Soils: Influence of Soil
Physical and Chemical Properties," Soil Science, 122 (1976),
pp. 350-359-
36. Kuzmack, A. and R. McGaughy, "Quantitative Risk Assessment
for Community Exposure to Vinyl Chloride," in W.R. Ott (ed.)
Environmental Modeling and Simulation, U.S. Environmental
Protection Agency, EPA 600/9-76-016, July 1976, pp. 736-739.
37. Larsen, R., "An Air Quality Data Analysis System for Inter-
relating Effects, Standards and Needed Source Reductions,"
APCA Journal, 26 (1976): 325-333, 27 (1977): 455-459.
38. Lassiter, R., J. Malanchuk and G. Baughman, "Comparison
of Processes Determining the Fate of Mercury in Aquatic
Systems," in W.R. Ott (ed.) Environmental Modeling and
Simulation, U.S. Environmental Protection Agency, EPA
600/9-76-016, July 1976, pp. 619-623.
39. Liss, P. and P- Slater, "Flux of Gases across the Air-Sea
Interface," Nature, 247 (1974), 181-184.
40. Mackay, D. and P- Leinonen, "Rate of Evaporation of Low-
Solubility Contaminants from Water Bodies to Atmosphere,
Environmental Science and Technology, 9 (1975), 1178-1180.
41. Mara, S. and Lee, Shonh, Human Exposures to Atmospheric
Benzene, Stanford Research Institute, Menlo Park,
California, October, 1977.
42. Metcalf, R.L., "Biological Fate and Transformation of
Pollutants in Water," in I.H. Suffet (ed.) Fate of Pollu-
tants in the Air and Water Environments, Part II, 1977,
pp. 195-221.
54
-------
43. Munro, J., R. Luxmoore, C. Begovich, K. Dixon, A. Watson,
M. Patterson, and D. Jackson, Application of the Unified
Transport Model to the Movement of Pb, Cd, Zn, Cu and S
Through the Crooked Creek Watershed, Oak Ridge National
Laboratory, Oak Ridge, Tennessee, ORNL-NSF-EATC-28,
September, 1976.
44. Neely, W.B., A Preliminary Assessment of the Environmental
Exposure to be Expected from the Addition of a Chemical to
a Simulated Aquatic Ecosystem, Dow Chemical Co., Midland,
Michigan, 1977.
45. O'Dell, R., M. Taheri and R. Kabel, "A Model for the Uptake
of Pollutants by Vegetation," Journal of the Air Pollution
Control Association, 27 (1977), pp. 1104-1109.
46. Patten, B. (ed.), Systems Analysis and Simulation in
Ecology, Volume IV, Academic Press, New York, 1976.
47. Perez, A., W. Huber, J. Heaney and E. Pyatt, A Water Quality
Model for a Conjunctive Surface-Groundwater System, U.S.
Environmental Protection Agency, EPA 600/5-74-013, May 1974.
48. Prahm, L.P., U. Torp and R.M. Stern, "Deposition and
Transformation Rates of Sulfur Oxides during Atmospheric
Transport over the Atlantic," Tellus, 28 (1976), p. 355-372.
49. Rodhe, H. and J. Grandell, "On the Removal Time of Aerosol
Particles from the Atmosphere by Precipitation Scavenging,"
Tellus, 24 (1972), pp. 442-454.
50. Shreffler, J.H., "A Model for the Transfer of Gaseous
Pollutants to a Vegetational Surface," Journal of Applied
Meteorology, 15 (1976), pp. 744-746.
51. Slinn, W., "Some Approximations for the Wet and Dry Removal
of Particles and Gases from the Atmosphere," Water, Air,
and Soil Pollution, 7 (1977), pp. 513-543.
52. Smith, J., R. Makey, N. Bohonos, B. Holt, S. Lee, T.W. Chou,
D. .Bomberger and T. Mill, Environmental Pathways of Selected
Chemicals in Freshwater Systems, U.S. Environmental Protec-
tion Agency, EPA 600/7-77-113, October, 1977.
53. Spedding, D.J., "The Interaction of Gaseous Pollutants with
Materials at the Surface of the Earth," in J. Bockris (ed.),
Environmental Chemistry, Plenum Press, New York, 1977, pp.
213-241.
55
-------
54. Van Hook, R. and W. Shults, Ecology and Analysis of Trace
Contaminants, Progress Report October 1974-December 1975,
Oak Ridge National Laboratory, Oak Ridge, Tennessee, ORNL-
NSF-EATC-22, February, 1976.
55. Viebrock, H., ed., Fiscal Year 1976 Summary Report of NOAA
Meteorological Laboratory Support to the Environmental
Protection Agency, Air Resources Laboratories, Silver
Spring, Md., NOAA Technical Memorandum ERL ARL-67, July,
1977.
56. Walsh, P., G. Killough, D. Parzyck, P. Rohwer, E. Rupp,
B. Whitfield, R. Booth, R. Raridon, CUMEX - a Cumulative
Hazard Index for. Assessing Limiting Exposures to Environ-
mental Pollutants, Oak Ridge National Laboratory, Oak
Ridge, Tennessee, ORNL-5263, April, 1977.
57. Wesley, M.L. and B.B. Hicks, "Some Factors that Affect the
Deposition Rates of Sulfur Dioxide and Similar Gases on
Vegetation," Journal of the Air Pollution Control Associa-
tion, 27 (1977), pp. 1110-1116.
58. Whelpdale, D.M. and R.W. Shaw, "Sulfur dioxide removal by
turbulent transfer over grass, snow and water surfaces,"
Tellus, 26 (1974), p. 196-205.
59. Wiegert, R., Population Models: Experimental Tools for
Analysis of Ecosystems, University of Georgia, Athens,
Georgia (in press).
60. Wiegert, R., "Simulation Models of Ecosystems," Annual
Review of Ecology and Systematics, 6 (1975), 311-338.
61. Winchester, J. and R. Duce, "The Air-Water Interface:
Particular Matter Exchange Across the Air-Water Interface,"
in I.H. Suffet (ed.) Fate of Pollutants in the Air and
Water Environments, Part I, 1977, pp. 27-47.
56
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/3-78-065
2.
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Exposure Assessment Modeling:
the-Art Review
A State-of-
5. REPORT DATE
July 1978 issuing date
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Catherine Miller
8. PERFORMING ORGANIZATION REPORT NO
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Kennedy School of Government
Harvard University
Cambridge, MA 02138
10. PROGRAM ELEMENT NO.
1BA609
11. CDNXBACT/GRANT NO.
R 805647
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Research Laboratory—Athens, GA
Office of Research and Development
U.S. Environmental Protection Agency
Athens, GA 30605
13. TYPE OF REPORT AND PERIOD COVERED
Final, 10/77 to 4/78
14. SPONSORING AGENCY CODE
EPA/600/01
15. SUPPLEMENTARY NOTES
Project Officer:
Kenneth Hedden,ERL-Athens, GA
16. ABSTRACT
This state-of-the-art review of exposure assessment modeling de-
scribes currently available models that simulate the environmental fate
of substances, the exposure to such substances, and the effects of such
exposure. The focus is first on exposure and effects, where relatively
little work has been done, and then on models of environmental fate, in
particular on intermedia transfer processes. Single-medium air and
water quality transport models are not assessed, but the possibility of
approaching multi-media problems through a combination or single-medium
approaches is explored. The report also describes several actual risk
assessments that have been made using limited data and considers some
secondary applications of the models.
The investigation shows that the available models do not cover all
of the areas necessary for an exposure assessment. More effort has been
directed to modeling environmental fate than to modeling exposure and
effects.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COS AT I Field/Group
Mathematical Models
Models
Pollution
Health Effects
Toxic Substances
06F
12A
68D
68G
18. DISTRIBUTION STATEMENT
Release to Public
19. S.ECURJTY CLASS.(ThisReport)
unclassified. v '
21. NO. OF PAGES
63
20. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
57
i U.i GOVERNMENT PRINTING OFFICE; 1378—757 -140 / 1408
------- |