I 
I 
I 
f. 
THE ECONOMIC EVALUATION OF THE 
SOCIAL COSTS OF AGRICULTURAL 
GROUNDWATER POLLUTION 
by 
Jennifer Lee Brunsdon 
A thesis submitted in partial fulfillment 
of the requirements for the degree 
of 
Master of Science 
in 
Applied Economics 
MONTANA STATE UNIVERSITY 
Bozeman, Montana 
June 1989 
I' 
J 
ii 
APPROVAL 
of a thesis submitted by 
Jennifer Lee Brunsdon 
This thesis has been read by each member of the thesis committee 
and has been found to be satisfactory regarding content, English usage, 
format, citations, bibliographic style, and consistency, and is ready for 
submission to the College of Graduate Studies. 
Date Chairperson, Graduate Committee 
Approved for the Major Department 
Date Head, Major Department 
Approved for the College of Graduate Studies 
Date Graduate Dean 
iii 
STATEMENT OF PERMISSION TO USE 
In presenting this thesis in partial fulfillment of the require­
ments for a master's degree at Montana State University, I agree that the 
Library shall make it available to borrowers under rules of the Library. 
Brief quotations from this thesis are allowable without special 
permission, provided that accurate acknowledgement of source is made. 
Permission for extensive quotation from or reproduction of this 
thesis may be granted by my major professor or, in his absence, by the 
Dean of Libraries when, in the opinion of either, the proposed use of the 
material is for scholarly purposes. Any copying or use of the material 
in this thesis for financial gain shall not be allowed without my written 
permission. 
Signature ________________________ __ 
Date ____________________________ ___ 
I 
Figure 
1. 
2. 
3. 
4. 
5. 
6. 
7. 
8. 
9. 
vii 
LIST OF FIGURES 
Relationships among the major components in 
the evaluation of agricultural pollution 
externalities .•..•.•...••• 
Factors influencing the behavior and export 
of agricultural chemicals from an agricultural 
watershed • • • • . • • . . 
Subsurface moisture zones .••....••• 
Relationships between economic and environ­
mental models in the evaluation of ground­
water pollution •...•.•••••.•• 
The damage model as a function of environ­
mental characteristics and management 
practices . . . . • • • • . • • • .• 
Output-damage transformation surface • 
Output and damage relationships in the 
integrated log-linear physical-economic 
production model . . . . . • . . . • • . 
Risk ladder illustrating annual risks of dying . 
Risk ladder illustrating lower-level risks 
of dying . . . . . . . . . . . . . . . . . . . 
Page 
2 
21 
22 
42 
43 
45 
49 
103 
104 
! ) 
iv 
ACKNOWLEDGEMENTS 
I would like to extend my sincere thanks to the chairman of my 
graduate committee, Dr. John M. Antle, for his patience and expertise 
provided throughout the preparation of this thesis. 
Additional thanks are extended to the other members of my 
committee, Drs. Susan M. Capalbo and Terry L. Anderson, for their 
suggestions and criticisms. 
( I 
v 
TABLE OF CONTENTS 
APPROVAL . . . . . . . . . . . 
STATEMENT OF PERMISSION TO USE 
ACKNOWLEDGEMENTS . . . . . . . . . . . . . . 
TABLE OF CONTENTS 
LIST OF FIGURES . 
ABSTRACT 
CHAPTER: 
1. 
2. 
INTRODUCTION 
THE ECONOMIC PRODUCTION MODEL . . . . . . . . 
Page 
ii 
iii 
iv 
v 
vii 
viii 
1 
7 
The Neoclassical Production Model • • . . • . • . • 7 
3. 
Dynamics of Agricultural Production . . • . . • • • 8 
Risk in Agricultural Production . . . . . 14 
CHEMICAL FATE AND TRANSPORT MODELS 
Management-Oriented Chemical Transport 
Models .......•...•.•.• 
The Pesticide Root Zone Model (PRZM) ..•• 
The Groundwater Loading Effects of 
Agricultural Management Systems (GLEAMS) Model . . • . . . . . . • . . • • . 
The Hydrological Simulation Program-
FORTRAN (HSPF) Model . . . . . . . • . . • . 
Model Applications ...•.....•• 
Pesticide Root Zone Model (PRZM) 
Application ...•..•........•• 
Groundwater Loading Effects of Agri-
cultural Management Systems (GLEAMS) 
Model Application ...•..... 
Hydrological Simulation Program-
FORTRAN (HSPF) Model Application •.•• 
Conclusions ............. . 
19 
29 
29 
30 
31 
32 
32 
34 
36 
37 
4. 
5. 
6. 
7. 
8. 
9. 
vi 
TABLE OF CONTENTS--Contjnued 
LINKING THE ECONOMIC AND ENVIRONMENTAL MODELS • . . • 
VALUING HEALTH RISKS . 
HEALTH RISK ASSESSMENT 
DATA BASES . . . . • . 
. . . . . . . . . 
. . . . . . . . 
Environmental Quality Data Bases .•••.•• 
DRASTIC Data Base . • . • . • • • • • . • • . • 
Interactive Soils Information System (ISIS) 
Management Data Bases • • • . . • . . • . • • • 
Fertilizer Use Data Base . • • • • . •.• 
Pesticide Use Data Base .•.••• 
National Pesticide Use Inventory • 
Tillage Data ..••••••... 
Contamination Data Bases •••••.••• 
. . . 
. . . . 
Groundwater Contamination Data Base •••• 
Pesticides in Groundwater Data Base ••• 
WATSTORE . . . • • . . . . . 
National Pesticide Survey 
CASE STUDY . • • . . • 
. . . 
Page 
39 
52 
61 
70 
71 
71 
75 
76 
77 
77 
78 
79 
79 
79 
81 
82 
84 
86 
Pesticide Parameters • • • • . • • . . • . . 91 
Soil Parameters • . . • • . • . . . . . • • • . 93 
Hydrologic and Crop Characteristics . . • • • 93 
Valuation of the Social Costs . . • • • 97 
Conclusions • • • • • . • . . . . • • • • . 108 
CONCLUSIONS AND RECOMMENDATIONS . 
LITERATURE CITED . . . 
109 
112 
121 APPENDIX • • . • • . . . . 
viii 
ABSTRACT 
There is overwhelming evidence that agricultural chemicals make a 
positive contribution to U.S. agri cul tura 1 production. In order to 
determine the net benefit (cost) to society of agricultural chemicals, 
the social costs and benefits must be quantified and valued. One 
potential social cost of agricultural chemical use is the human health 
effects of chemically-contaminated groundwater. In this thesis a multi­
disciplinary framework, incorporating physical models and economic 
production models, is developed to value the health risks of polluted 
groundwater. This framework can also be used to determine the impacts 
of agricultural policy on groundwater quality. 
In the economic model, farmers jointly make input use, management 
and land use decisions. Land use decisions determine the environmental 
characteristics of the 1 and in production. The farmer's economic 
production model is linked with an environmental damage model (such as 
a chemical fate and transport model) to determine the amount of ground­
water pollution resulting from the use of agricultural chemicals on land 
with particular environmental characteristics. Toxicology and epidemi­
ology studies are used to estimate the human health risks presented by 
groundwater contamination, and a contingent valuation method is used to 
p 1 ace a va 1 ue on those risks. The contingent va 1 uat ion method uses 
survey techniques to elicit individuals' willingness to pay for a change 
in the level of groundwater contamination and the accompanying change in 
health risks. 
A case study is presented in order to evaluate the feasibility of 
linking the chemical fate and transport, economic, and human health 
models. Although the general physical models needed for this framework 
are currently available, most of these models are designed to be used 
by researchers within the respective discipline. Consequently, there are 
some important gaps in methods and data, including: (1) chemical fate 
and transport models that do not simulate chemical movement down to the 
groundwater zone, (2) lack of chemical-specific toxicity and epidemio­
logic data, and (3) lack of location-specific environmental data. This 
study illustrates the need for researchers to be aware of the implica­
tions and potential applications of their research, both within and 
outside their fields. 
I \.; 
1 
CHAPTER 1 
INTRODUCTION 
There is overwhelming evidence that agricultural chemicals 
(fertilizers and pesticides} make a positive contributi.on to U.S. 
agricultural production (Headley, 1968; Capalbo and Antle, 1988). There 
is also evidence that suggests there may be social costs associated with 
agricultural chemical use due to environmental and human health impacts 
(Sharp et al., 1986; Berteau and Spath, 1986; Moses, 1987; Coye, 1985}. 
In order to determine if a technology is making a positive net contribu­
tion to society, the private and social costs and benefits of that 
technology must be quantified and valued. The private costs and benefits 
of agricultural chemical use can be quantified and valued using existing 
data and methods (Pimentel, 1979), but the social costs and benefits are 
more difficult to quantify because these impacts involve non-market goods 
such as environmental amenities and human health. A multi-disciplinary 
approach is therefore necessary to quantify and value the social costs 
of agricultural chemical use. 
The soc i a 1 costs or extern a 1 it i es of an agri cul tura 1 production 
technology can be accounted for if they can be quantified and valued. 
This process i nvo 1 ves: (1) the quant ifi cation of the re 1 at i onshi p 
between agricultural production and the environmental sectors it impacts 
(both directly and indirectly), and (2) the valuation of these 
2 
externalities. Figure 1 presents the four major components (and the 
interaction of these components) that must be considered in the 
evaluation of an agricultural pollution externality. The interactions 
among the agricultural production sector, the human health sector, and 
the agro-ecosystem must be quantified and then valued to determine the 
extent of these impacts on soc i a 1 we 1 fare. Phys i ca 1 , chemica 1 , and 
biological models can be used to estimate the impacts of the agricultural 
production technology on the agro-ecosystem and on human health. 
Figure 1. Relationships among the major components in the 
eva 1 uat ion of agri cul tura 1 po 11 uti on extern a 1 it i es. 
As with an evaluation of any agricultural pollution externality, 
the evaluation of groundwater pollution resulting from agricultural 
chemical use would include an analysis of the impacts on the four 
sectors presented in Figure 1. The characteristics of the agro-eco­
system and the producer's management decisions jointly determine the 
amount of damage to the agro-ecosystem (e.g., groundwater contamin­
ation), which in turn can impact human health. To quantify the damage 
to the agro-ecosystem, it is first necessary to model the movement of 
the chemical within the agro-ecosystem. There is a variety of chemical 
fate and transport models that can be used to trace chemical movement. 
Once the distribution of the chemical within the agro-ecosystem is 
3 
established, one can begin to evaluate the impacts of the chemical. The 
evaluation of health risks requires information on the levels and 
duration of exposure to the chemical and the types and extent of health 
risks presented by that exposure. 
To conduct an economic evaluation of the net social benefits of 
agricultural chemicals, benefit-cost methodology can be used. Benefit­
cost analysis provides a useful framework for the evaluation of environ­
mental quality problems (Freeman, 1979; Lave and Seskin, 1977). It is 
based on neoclassical welfare theory; i.e., an individual's preferences 
are accepted as the basis for measuring welfare gains and losses associ­
ated with alternative pesticide use. For a benefit-cost analysis, the 
private and social costs and benefits of the technology must be iden­
tified, quantified, and valued. Benefit-cost analysis of agricultural 
chemical technology consists of the following steps. The first is to 
specify the economic production model. This step provides the informa­
tion that can be used to ca 1 cul ate private benefits; it wi 11 a 1 so 
determine the producer's use of chemical inputs, which is needed to 
estimate the amount of potential groundwater pollution. The second step 
is to use the physical and biological models to describe the dynamics of 
the agro-ecosystem. In the evaluation of groundwater pollution, this 
step involves tracing the movement of the chemical through the soil 
profile, estimating the extent of groundwater pollution, and determining 
the physical and biological effects of the groundwater pollution on the 
agro-ecosystem. 
The third step is to use a risk assessment model to quantify the 
human health risks of polluted groundwater. There are numerous ways to 
! ) 
I ' 
I [ 
4 
assess health risks and to measure and estimate exposure. The partic­
ular risk assessment model used in an analysis will depend on factors 
such as the chemical in the groundwater, the types of health hazards the 
chemical presents, and the type and length of exposure. 
The fourth step is to value the private and social costs and 
benefits quantified in the previous three steps. The valuation will 
require the use of market information (such as input and output prices) 
to value private costs and benefits and non-market techniques (hedonic 
pricing, contingent valuations, etc.) to value public costs and benefits 
such as health effects. 
The fifth step of the economic evaluation framework is to make a 
determination of net benefits (costs). The private and social costs and 
benefits are summed for each period and discounted to determine the net 
present value of the stream of costs and benefits; if the net present 
value is positive (negative), the use of agricultural chemicals results 
in positive benefits (costs) to society. If all the social and private 
costs and benefits were not valued in the fourth step, the total net 
present value of the costs and benefits cannot be determined. 
Determining the total cost to society of a technology is important 
for po 1 icy ana 1 ys is purposes. If a techno 1 ogy creates a net cost to 
society, the magnitude of the social costs can be a factor in the deter­
mination of society's optimal policy regarding agricultural chemicals. 
For example, two possible regulatory options concerning chemicals are a 
total ban on use of the chemical, or restrictions on its use. Banning 
the chemical would be optimal if total costs exceed total benefits for 
all quantities of chemical application. Restricting chemical use may be 
5 
optimal if total benefits exceed total costs over some range of chemical 
use. 
This thesis develops a methodology for evaluating the social costs 
of agricultural chemical pollution of groundwater and consists of three 
major components: 
{1) The development of a framework for evaluating agricultural chemical 
pollution of groundwater. This framework consists of multidisci­
plinary components needed to quantify and value social costs for a 
benefit-cost analysis. 
(2) The linking of the physical and biological models, human health 
hazard models, and economic models involved in (1) above. 
(3) A case study on the use of the herbicide Ro-Neet in sugar beet 
production in the Yellowstone Valley of Montana. An assessment of 
available methods, models, and data bases is presented and will 
indicate the feasibility of using this framework given the current 
state of knowledge. 
The outline of this thesis is as follows. Chapter 2 presents the 
necessary considerations for modeling the agricultural production 
process with explicit consideration given to incorporating the use of 
agricultural chemicals. Chapter 3 discusses chemical fate and transport 
models. Chapter 4 presents a method of linking an economic production 
mode 1 and a chemica 1 fate and transport mode 1 . Chapter 5 discusses 
techniques of valuing human health risks. Chapter 6 discusses current 
methods of estimating human exposure to agri cul tura 1 chemica 1 s and 
health risks of exposure. Chapter 7 provides an overview of data bases 
6 
that could be used in an application of the framework discussed in 
Chapter 4. Chapter 8 is a case study that illustrates the steps and 
considerations for a valuation of the social costs of groundwater 
pollution in the Yellowstone Valley of Montana. Chapter 9 presents some 
concluding remarks and discusses research recommendations. 
r 
/ 
7 
CHAPTER 2 
THE ECONOMIC PRODUCTION MODEL 
The neoclassical theory of production is commonly used to analyze 
a firm's production decisions. Neoclassical theory is centered on the 
assumption that a firm's short-run objective is to maximize profits under 
conditions of costless information and certainty; maximizing decisions 
are assumed to be made at the beginning of the production process. 
The Neoclassical Production Model 
The neoclassical model is a one period model in which prices and 
output are known with certainty. The firm's objective function can be 
expressed as 
max J(x,y,1), where J(x,y,1) = pq- wx 
X 
q = f(x,y,o) 
where J is returns to fixed factors, p is output price, q is output, w 
is the vector of variable factor prices, x = (x1, ••• , x0 ) is the vector 
of variable input factors, y is the vector of fixed factors, e is a 
vector of parameters of the production function, and 1 is a vector of 
input and output prices and production function parameters, 1 = (p,w,s). 
The first order condition is: 
t = 1, ... , n [2.1] 
8 
The first order conditions equate the value of the marginal product (VMP) 
of input ~ to the marginal factor cost (MFC) of ~, which in this model 
is equal to the price wt to xt. Solving the set of first order equations 
results in variable factor demands that are functions of prices, level 
of q, fixed factors, and parameters of the production function: 
x* = x*(p,w,y,o) [2.2] 
where the* denotes the optimizing input level. 
This static and deterministic model is not always the most realistic 
representation of a production process or of profit maximization, and so 
the relaxation of some of these assumptions may result in a more repre­
sentative model. Two modifications to the static production model which 
are important for the research in this thesis will be discussed: 
incorporating dynamics and incorporating risk. 
Dynamics of Agricultural Production 
In a static model all input decisions are made at the beginning of 
the production process, whereas in a dynamic or sequential decision model 
production decisions are made over time. In the case of agricultural 
production, input decisions are often made throughout the production 
process because. of the time dimension of biological processes. For 
example, a pest infestation that was not existent or apparent at the 
beginning of the process may become apparent at some point in the 
production season; the producer is then faced with the intra-seasonal 
decision of whether or not to apply a pesticide. In this example the 
producer is using information gathered over time to maximize profits. 
Because time is a factor in this producer's production decisions, a 
9 
dynamic production model is a better representation of this production 
process than a static model. 
Antle (1983b) presents the two general forms of dynamics that can 
be incorporated into a production mode 1 : input dynamics and output 
dynamics. A production process may exhibit both input and output 
dynamics. Input dynamics are characterized by output that is a function 
of a sequence of inputs: 
qt = ft(Xt, Xt-1' · • ·' Ut) 
where qt is output at time t, xt is the input vector, and ut is the random 
production disturbance in time t. Output dynamics involve a series of 
decisions made over time and are characterized by a final output that is 
dependent on a sequence of intermediate outputs: 
qt = f(Xt, qt-1' qt-2' • · •' Ut} · 
The production model can be formulated as a multistage or multi-
period decision problem. A multistage problem divides a production 
process into a sequence of production stages, whereas a multiperiod 
problem examines a sequence of temporally related production processes. 
For example, the production of a crop that reaches maturity in one season 
may exhibit output dynamics and the producer may have a multistage 
decision problem; the crop will produce a series of intermediate outputs 
during that season and because of these observable intermediate outputs, 
the producer may be faced with intra-seasonal decisions. The objective 
function of the multistage model (having T stages) is of the form: 
T 
max E ( p1q1 - :E gtwtxt) 
t=l 
10 
where E(•) is the expectation operator, qT is the final output, and gt is 
a discount factor. 
With a multiperiod production process, the producer may be faced 
with inter-seasonal decisions such as the type of crop rotation to 
implement. The multiperiod model objective function is of the form 
T 
max E ( ~ Jtgt) 
t=1 
where J = Ptqt - WtXt 
In both multiperiod and multistage dynamics, the producer's objective 
function is to maximize the present value of producer surplus (return to 
fixed factors). 
Solutions to these dynamic optimization problems can be obtained 
using the dynamic programming algorithm. This algorithm relies on the 
segmentation of a problem into a sequence of smaller subproblems, called 
stages, with each stage requiring some type of decision. Each stage has 
a number of states associated with it; a state describes the condition 
of the process at a particular stage. The decision made at a stage will 
determine the transformation of the current state into a state associated 
with the next stage. The solution to a dynamic programming problem 
consists of the sequence of optimal decision rules for each stage 
(Bellman, 1957). 
The dynamic programming algorithm starts at the end of the planning 
horizon and works backwards to the beginning by solving a series of 
recursive equations. The following example illustrates the form of these 
equations: 
T max J(xr) => Xr* 
Xr 
T -1 max J [xr-P Xr*(xr.1)] 
Xr-1 
11 
t max J [><t, ><t+1 (><t), ... , xr*(xr.1)] 
><t 
1 max J[xp x2*(x1), ... , ><t*(><t.1), ... , xr*(xr.1)] 
x1 
where ><t* is the decision rule for stage t and T is the number of stages. 
The optimal decision rule at stage t is conditional on decisions to 
be made in stages 1, ... , t-1 and on optimal decision rules for stages 
t+1, ... , T. (See Burt and Allison, 1963, for an application of the 
dynamic programming algorithm to agricultural production.) 
In a dynamic, stochastic model with a risk-neutral producer, the 
producer's objective is to maximize the present va 1 ue of profit by 
choosing the optimal input sequence. In the short-run, with fixed costs 
and the state of technology held constant, maximizing profit is equiv­
alent to maximizing producer surplus. This dynamic problem can be 
approached with the objective function for each stage being expressed as: 
max E (J !It) 
~ 
where It is the information set in stage t, and It = ( I0 , ••• , It.1 , qt_1 , wt, 
><t+1*, .•. , xr*), I0 is the initial information set, and I0 = (w1, .•. , WT' 
v0), and v0 is the vector of parameters of the producer's initial expecta­
tions of future prices and output. 
!'· . 
12 
The first order conditions of the dynamic model differ from those 
of the static model. To illustrate the differences, consider a two-stage 
dynamic model: 
qt = f 1 ( Xp U1 ) 
q2 = f2(x2, qp U2) 
where ~ is output at the end of stage t, ~ is the variable input at 
stage t, and ut is the random production disturbance occurring during 
stage t. 
In stage 2 the producer's maximization can be expressed as: 
max E[{pq2 - w1x1 - w2x2)lv2] 
where p is the price of q2, wt is the price of input xt, and Vt is the 
vector of parameters of the producer's expectations at time t of future 
prices and output. 
The first stage input, x1, is chosen based on the producer's 
expectations of future prices and output, and decision rules for future 
optimal inputs, in this case x2*. In stage 1 the producer's maximization 
can be expressed as: 
max E [ { pq2 - w1 x1 - w2x2 *) I v d 
The first order condition for stage 1 is: 
E1 [P aq2 + aq2 • ax2*] = Et [wt + w2 ax2*] 
ax1 ax1 ax1 ax1 
[2.3] 
Recall from equation [2.1] that the first order conditions of the 
neoclassical model equate the VMP at t to the ~, and so factor demands 
are a function of all variable factor costs, output price, and fixed 
factors at time t. The first order condition for this dynamic model 
13 
equates the expected value of marginal product (EVMP) to the expected 
marginal factor cost (EMFC). In the dynamic case, EVM~ (the left-hand­
side of equation [2.3]) equals the direct effect of x on output plus the 
indirect effect of x on output due to its effect on input requirements 
in future stages, and EMFC (the right-hand-side of equation [2.3]) equals 
the factor cost for that stage plus the effect on cost in the future 
stages. The production decisions in the dynamic model are functions of 
previous input decisions, and the effect of current decisions on future 
decisions {Antle, 1988). Therefore, because the producer updates his 
information set, factor demands are functions of output over time 
'Xt.* = 'Xt.*(wt, wt+P ... , Wr, qt-P qt-2' ... , It) [2.4] 
Imposing any type of restrictions on the production model has 
certain imp 1 i cations about the behavior of the firm {producer) . A 
dynamic production model implies the producer makes decisions over time 
and can use newly-acquired information to optimize. Imposing a dynamic 
structure on a production model also has implications regarding the 
separability of inputs. A production function exhibiting output dynamics 
can be expressed as 
qt = f[Xt., qt-1' ut] 
= f[Xt., ft-1 <Xt-1' qt-2' ut-1)' utl 
where 'Xt. is a vector of inputs and ut is a random production disturbance. 
Because of the time factor in this ex amp 1 e, the Xt..1 inputs are 
separable from the xt inputs, but the 'Xt. inputs are not separable from the 
Xt..1 inputs. Being able to assume separability of inputs can simplify the 
task of econometric estimation. 
I 1 
I ( 
14 
Risk in Agricultural Production 
Because agricultural production is not deterministic, it may be 
necessary to incorporate risk into the production model. The agricul­
tural production process is dynamic and production is not instantaneous 
so that output and all prices usually are not known with certainty until 
after the input decisions have been made. There are two general types 
of risk that apply to agricultural production: production risks, such 
as pests, weather, etc., and price risks due to changes in market 
conditions. 
Incorporating production risk into a production mode 1 can be 
accomplished by interpreting output as a random variable. The firm's 
revenue and profit are stochastic if output is a random variable. When 
profit is stochastic, the producer's objective function is often assumed 
to be to maximize expected utility of profit. In the case of risk 
neutrality, the marginal utility of profit is constant; utility is a 
monotonic transformation of profit and so maximizing expected utility of 
profit is equivalent to maximizing expected profit. In a static model 
where the producer is risk neutra 1 and a price taker, his objective 
function is 
max EU{J) = E{J) = E{pq - wx) 
X 
= E{p)E{q) + COV{p,q) - wx 
[2.5] 
[2.6] 
where E(•) is the expectation operator, J is profit, U(J) is the utility 
of J, and COV{p,q) is the covariance between p and q {Newberry and 
Stiglitz, 1981). Assuming that price and output are statistically 
independent, [COV(p,q) = 0], the first order condition is 
15 
aE(J)/ax = o 
which implies 
E(p)aE(q)/ax = w [2.7] 
The left-hand-side of equation [2. 7] is the expected value of the 
marginal product (EVMP) which is equated to the marginal factor cost 
(MFC), or factor price w. 
The utility function of a risk-averse producer exhibits diminishing 
marginal utility. Utility of profit is not a monotonic transformation 
of profit, and so maximizing expected utility of profit is not equivalent 
to maximizing expected profit. The risk-averse producer's objective 
function is: 
max EU(J) = EU(pq - wx) 
X 
The first order condition can be expressed as: 
E(p)aE(q)/ax + E(p)cov(U' ,aq/ax)/E(U') - w = 0 
where U' is marginal utility of income (Antle, 1988). 
[2.8] 
Equation [2.8] can also be interpreted with the use of the certainty 
equivalent (CE). The CE is the dollar amount that the firm could receive 
with certainty if it were insured against risk. CE is a measure of 
producer welfare. 
CE = E(J) - R 
where R is a risk premium (see Newberry and Stiglitz, 1981, and Pearce 
and Nash, 1981). 
In this model, maximizing profit (J) is equivalent to maximizing 
producer welfare (producer surplus): 
! 
' I 
I i 
I 
I I 
16 
aEU(J)jax = aCE/ax = aE(pq-wx)jax - aRjax = 0 
= E(p)oE(q)jax - aR;ax - w = o 
[2.9] 
[2.10] 
where aRjax is the marginal risk premium. Because equation [2.8] is 
equivalent to equation [2.10], the second term on the right-hand-side of 
both equations can be interpreted as the marginal risk premium. In a 
static risk-neutral model, R = 0. If in the risk-averse model aR/ax is 
not equal to zero, then the producer's risk attitude affects his optim­
izing decisions. If the marginal risk premium is equal to zero, the 
first order condition of the risk-averse case is equal to that of the 
risk-neutral case, and the factor demands in both cases will be equal. 
If the covariance is not equal to zero, then the marginal risk premium 
is the amount by which the risk-averse producer's input use differs from 
that of the risk-neutral producer. A negative covariance (positive 
marginal risk premium) indicates that the risk-averse producer will use 
less of the input than the risk-neutral producer and the input is said 
to be marginally risk-increasing; this negative covariance has an 
equivalent effect on input usage as an increase in MFC. If the covari­
ance is positive (negative marginal risk premium), the risk-averse 
producer will use more of the input than the risk-neutral producer and 
the input is said to be marginally risk-reducing. This positive covari­
ance has an equivalent effect on input usage as does a decrease in MFC. 
As Antle (1983b) demonstrates, in some cases it may not be necessary 
to incorporate risk into the production model. However, in models that 
involve dynamics or a risk-averse producer, it is important to consider 
risk. In a dynamic model, incorporating risk is important whether the 
producer is risk-neutral or risk-averse because in both cases risk 
I I 
. I 
17 
affects resource allocation and expected productivity. As stated 
previously, a dynamic model may be a multistage model or a multiperiod 
model. Although these models have different objective functions, the 
solutions of the first order conditions result in factor demands of the 
form 
'Xt* = Xt*('Xt.1' 'Xt-2' • • • , qt-1, qt-2' • • ·, wt, ut) 
where ut is the parameter vector defining the joint probability distribu-
tion of current and future output and future prices. In both the 
multistage and multiperiod models the factor demand is a function of the 
distribution of current and future output and future prices, indicating 
that optimal resource allocation is dependent on price and output risk. 
When a production function is stochastic, the producer's utility 
function can be expressed as a function of profit and risk attitudes: 
U1 = U(Jp g1) 
where u1 is the jth producer's uti 1 i ty function, J1 is profit (and is 
stochastic), and g1 is a vector of parameters that measures the jth 
producer's risk attitudes. Therefore, the producer's production 
decisions (determined by maximizing expected utility), are affected by 
the stochastic profit function, ~, and his attitudes towards risk, ~· 
Determining the risk attitudes of every individual in a population 
may be impossible due to large data requirements, and to assume that 
producers' risk attitudes are homogeneous may be an inappropriate 
simplification. Antle (1987) presents a method of estimating the 
distribution of producers' risk attitudes, assuming producers use similar 
production technology. A moment-based approximation of the profit 
distribution allows for the estimation of the derivatives of the expected 
I ': 
I I 
. I 
i 
18 
utility function; the sign of the derivatives of the utility function can 
then be inferred from the derivatives of the expected utility function. 
In summary, the neoclassical model is a static and nonstochastic 
model in which producer behavior is determined by prices and the produc­
tion function. However, agricultural production can involve risk and 
decision-making over time, and so the neoclassical production model may 
have to be modified to include these effects on producer behavior. 
Incorporating risk into the production model results in producer behavior 
that is a function of the producer's risk attitudes. Incorporating 
dynamics into the production model results in producer behavior that is 
a function of time. 
I I I 
19 
CHAPTER 3 
CHEMICAL FATE AND TRANSPORT MODELS 
Mathematical models of chemical fate and transport are tools with 
which to evaluate nonpoint source contamination of groundwater. Chemical 
fate and transport models simulate the movement of a chemical in the 
environment. In the case of water quality, the output of these models 
may be the effects of chemical use on surface water quality, groundwater 
quality, or both. Concern over groundwater contamination is a relatively 
recent development (relative to concern over surface water contamina­
tion), and models that predict chemical leaching to groundwater are less 
developed than those that predict chemical runoff to surface water 
(Mulkey et al., 1986). 
Chemical loading models can be divided into three broad categories: 
research models, screening models, and management models (Wagenet and 
Hutson, 1986). Research models provide quantitative estimates of water 
and solute movement and have extensive data requirements. Screening 
models can be used to evaluate the potential environmental effects of a 
pol icy that is implemented over a broad geographical area. They are 
designed to be simple, quick, relatively inexpensive to use, and 
supported by available data bases (Bailey and Swank, 1983). Management 
models are less data-intensive and less quantitative, and consequently 
the ability of a management model to predict water and solute movement 
i l 
I 
20 
is less precise than that of a research model. According to Wagenet and 
Hutson (1986, p. 316), "In the case of all management models, the ability 
of the model to quantitatively estimate pesticide fate under field 
conditions is questionable." Managerial models are usually designed 
to incorporate changes in management decisions (such as tillage 
practices and crop rotations). Many managerial models are field-scale 
models because that is the area appropriate to agricultural decision 
making. 
Pollution of the agro-ecosystem due to agricultural chemicals is 
dependent on hydrologic, ecologic, agronomic, social, and economic 
subsystems. Figure 2 illustrates the factors that affect the behavior 
of agricultural chemicals in an agricultural watershed. Natural inputs 
(e.g., climatic characteristics) and management inputs (e.g., chemical 
use, 1 and use, and cultura 1 or till age practices) combine with the 
environmental characteristics of the water shed to determine the fate and 
transport of the agricultural chemical. Figure 2 indicates that 
chemicals can be transported via air and water. The focus of this thesis 
is the transport of dissolved chemicals to groundwater (indicated as the 
subsurface flow of Figure 2). 
To model potential loading to groundwater, a chemical transport 
model is needed to trace the movement of the chemical from the applica­
tion site down through the unsaturated (or aerated) zone and into the 
saturated zone. Figure 3 illustrates the zones involved. The zone of 
saturation is the area in which a 11 the void spaces are fi 11 ed with 
water; in the aerated (or unsaturated) zone, the void spaces are filled 
with both air and water. The soil water zone lies below the ground 
I I 
I I 
. ) 
r 
r 
I I 
i I 
NATURAL INPUT 
Precipitation 
- Rain 
-Snow 
Temperature 
Radiation 
Pollutant -
Rc1inout 
- Nutrients 
- Pesticides 
- Particulates 
- Heavy Metals 
Figure 2. 
21 
MANAGEMENT INPUT 
Soils, Topography 
Geology, Vegetation 
Drainage Network 
AIRSHED OUTPUT-INPUT 
Chemical 
Spray 
Drill 
Dust -
Adsorbed 
Chemicals 
TO 
ESTUARY 
WATERFLOW OUTPUT 
INPUT 
TO 
ESTUARY 
Factors i nfl uenc i ng the behavior and export of 
agricultural chemicals from an agricultural water­
shed (Bailey and Swank, 1983, p. 30}. 
surface and extends down through the root zone; hence this zone is also 
called the root zone. The intermediate or vadose zone lies below the 
root zone and above the groundwater zone (Bear and Verruijt, 1987}. 
Some models simulate chemical transport only part way through the 
unsaturated zone (e.g., through the root zone). If the unsaturated zone 
model does not simulate chemical transport down to the unsaturated zone, 
the model may not give a reasonable estimate of groundwater loading. In 
order to estimate the potential groundwater concentration of a chemical, 
it would be necessary to 1 ink the chemical transport model with a . 
groundwater model. Some loading models may have this linkage built in, 
and some require the user to make this linkage. 
I I 
r I 
' I 
I I 
. I 
i ) 
~ i 
Ground surfa~.:e 
Soil water zone 
._ c 
0 0 
'!) ·;; 
c: "' 0 ... N~ 
22 
Pellicular and 
gravitational water 
-·-~~----~--tLJ.J..J..J.J ll ~llllll ilJ I~ I I Jlll I I I I lllllllll"llfULLJ.J..LU.U..W'-\UULW 
._g 
0·­
v'; 
c ... 
0 :s 
N'; 
Vl 
Groundwater 
zone 
..,. Capillary water 
Groundwater 
Impervious 
Figure 3. Subsurface moisture zones (Bear and Verruijt, 1987, 
p. 4). 
Before examining the chemical transport models, it is helpful to 
have an understanding of the behavior of chemicals in soil and water. 
Rao et al. (1983) provide an overview of the behavior of pesticides in 
soil and water; much of the following information is taken from their 
work. 
The fate of a pesticide applied to soil depends largely on two of 
the pesticide's properties: persistence and solubility. Persistence is 
a measure of a chemical's rate of degradation and is usually measured in 
terms of the chemical's half-life. 
Solubility and sorption determine how a compound partitions among 
soil, water, and air phases and affect whether the chemical is moved 
primarily with sediment or water. According to Rao et al. (1983, p. 2), 
"[Solubility is] probably the single most important property influencing 
a pesticide's movement with water .... " When a pesticide is applied, 
some of it will adhere to the organic carbon in the soil particles; this 
I I 
i 
, I 
23 
is called adsorption. Some of the pesticide will mix with soil water 
and may move down in the soil with the soil water. Desorption, the 
detachment of pesticide molecules from soil particles, may also occur as 
water moves through the soil (e.g., rain or irrigation). An inverse 
relationship exists between the solubility of a pesticide and its 
sorption to soil. A partition coefficient value (Kp) is used to describe 
the ratio of pesticide concentration in the adsorbed phase and the 
solution phase; the KP is determined by the chemical's properties such 
as solubility and melting point. The KP can be expressed as a ratio of 
the chemical concentration in the soil or solid phase (C8 ), and the 
concentration in solution (Cw): 
KP = CsfCw. 
The smaller the KP, the greater the concentration of pesticide in 
solution. The greatest threat to groundwater through leaching is a 
pesticide with a small KP and a long half-life. However, a pesticide 
with a small KP and a short half-life may also pose a threat of leaching 
into groundwater if heavy precipitation closely follows application. 
The transport of a chemical from the application site down through 
the unsaturated zone is dependent on hydrologic, soil, and chemical 
characteristics. In many chemical transport models these processes are 
incorporated into hydrologic, soil, and chemical components. Although 
the structure of a model may vary, most models contain some standard 
components. Novotny (1986) describes the basic components that are 
included in a nonpoint pollution model. The basic components needed to 
estimate potential groundwater loading from agricultural chemicals 
include: 
i I 
I i 
I 
I I 
; I 
I I 
! 
24 
(1) Surface Runoff Generation Component: Describes the transformation 
of precipitation into runoff. (Precipitation will become runoff or 
infiltration.) Surface runoff is a function of surface storage, 
evapotranspiration (the sum of plant and soil evaporation), and 
snow accumulation and melt. The USDA Soil Conservation Service 
Curve Number (SCSCN) model is commonly used to estimate runoff. 
This method relates direct runoff to daily rainfall as a function 
of a curve number representing soil type, soil drainage properties, 
crop type, and management practice. This SCSCN mode 1 uses the 
equation 
Q _ (P - 0.2S)2 
- (P + 0.8S) 
where Q is daily runoff, Pis daily rainfall, and Sis a retention 
parameter. The retention parameter is expressed as a function of 
the curve number (CN) (Williams and LaSeur, 1976): 
s = 1 ~~0 - 10 
(2) Soil and Groundwater Component: Describes chemical movement 
through the unsaturated soil zone and may also describe movement 
into the saturated zone. (As previously mentioned, not all models 
trace the movement all the way through the unsaturated zone to the 
saturated zone.) 
(3) Erosion Component: Estimates soil loss due to erosion. This is 
important when determining potential for groundwater contamination 
because soil sediment is a medium of transport for adsorbed chemi­
cals. A chemical that is transported off the field via eroded soil 
I I 
" 
! : 
25 
is not available for leaching to groundwater. The Universal Soil 
Loss Equation (USLE), or a modification of the USLE, is frequently 
used to model erosion. The USDA/Soil Conservation Service (1978) 
summarizes the USLE: 
A = RxKxLSxCxP 
where A is the computed soil loss in tons per acre per year, R is 
the rainfall factor, K is the soil erodibility factor, LS is the 
slope length ratio factor and slope gradient ratio factor, C is the 
cover-management (cropping-management) factor, and P is the erosion 
control practice factor. 
{4) Soil Adsorption and Desorption Component: Estimates the partition­
ing of a chemical between adsorbed particles and dissolved chemi­
cal. This is the KP mentioned earlier. This component estimates 
what portion of the chemical may be transported by soil sediment 
and what portion may be transported by soil water. It may also 
model volatilization and decay of the chemical. 
Of the chemical fate and transport models currently used, many are 
non-steady-state equilibrium, continuous simulation models. An equil­
ibrium model assumes the physical and biological processes rapidly 
achieve a state of equilibrium, and non-steady-state {or dynamic} models 
are designed so the ecosystem varies with time due to changing condi­
tions or due to changing inputs. Continuous simulation models are 
usually based on a series of differential equations and therefore model 
changes in chemical concentration in the soil profile over time. Long­
term simulations are used in chemical fate and transport models because 
simulating potential chemical loading for one year may not be indicative 
26 
of typical chemical loading since potential loading is dependent on 
rainfall patterns. Long-term simulations are better able to consider 
the typical variation in rainfall pattern as well as the impact of a 
crop rotation on potential chemical loading. The results of long-term 
simulations can be used to determine long-term average values (e.g., 
long-term average potential loading), or frequency distributions (e.g., 
of annual potential groundwater loading). 
Mulkey et al. (1986) have defined a system of general mass-balance 
equations that depict the changes in the chemical concentration profile 
over time. The following mathematical model is taken from that research: 
where 
n 
I: 
i=1 
a(C1V1) n a a(C1V1) n a (C1V1u1} 
+ 
= I: - Di - I: at i=1 ax ax i=1 ax [3.1] 
n n m 
I: APPi(t) - I: I: k. c. v. 
i=1 i=1 j=1 J 1 1 
= concentration of the pesticide in the illi environmental 
compartment, M/L3 
= volume of the illi environmental compartment, L3 
= coefficient of dispersion (includes diffusion), L2/T 
= velocity vector for the media of the ;lli compartment, 
L/T 
pesticide application rate for the illi compartment on 
day t, M/T 
~ = pesticide transformation rate for the jlli process, 1/T 
n = number of environmental compartments or media 
27 
m = number of transformation processes acting on pesticide 
within each compartment or media, assuming first order 
kinetics 
x = space dimension, L 
t = time, T 
where L is a measure of space, T is a measure of time, and M is a measure 
of mass. 
The left-hand-side term of equation [3.1] represents the change over 
time of the amount of chemical in a soil compartment. The first term in 
the right-hand side of this equation represents all of the chemical 
dispersion processes; dispersion is the mixing or spreading of solute 
caused primarily by variations in veJocity (Carsel et al., 1988b). 
Diffusion, one of the dispersion processes, is the spontaneous movement 
of a substance in solution or suspension; this movement is not a function 
of water flow. The second term represents the advection processes, the 
movement of the chemical with water. If surface runoff does occur, this 
advection term will model the transformation of precipitation into runoff 
and infiltration. (Both dissolved and adsorbed chemical can move with 
runoff water.) The third term represents the application of the 
chemica 1 , and the fourth term represents a 11 of the transformation 
processes acting on the chemical, including degradation. 
The management-oriented chemical transport models developed in 
recent years include the Groundwater Loading Effects of Agricultural 
Management Systems (GLEAMS), the Pesticide Root Zone Model (PRZM), and 
the Hydrological Simulation Program-FORTRAN (HSPF). These models have 
severa 1 shortcomings. One major disadvantage is that they cannot be 
used to model the movement of chemicals down to deep groundwater. For 
28 
ex amp 1 e, GLEAMS mode 1 s movement down through the root zone, but not 
through the vadose zone. HSPF models movement down to shallow ground­
water, but not to deep groundwater. PRZM models movement within and 
below the root zone and can simulate groundwater loading only if the 
depth to groundwater is sufficiently shallow. Although all three models 
can simulate the "potential loading" of a contaminant, this estimate is 
not always a meaningful estimate of the actual loading to groundwater. 
If there is a significant distance between the lowest soil zone modeled 
and the saturated zone, then at best the model will provide an upper 
bound on potential groundwater loading. Therefore, applying these 
models to get an estimate of potential chemical loading may be suitable 
in two circumstances: (1) if the depth to groundwater is sufficiently 
shallow that the pollutant loading estimated by the model feeds directly 
into the saturated zone, and (2) if shallow groundwater moves laterally 
within the unsaturated zone and then feeds into another water body, 
e.g., surface water. Research is currently underway to develop a vadose 
model that could be linked with a root zone model to simulate chemical 
movement down to the saturated zone. Another shortcoming of these three 
models is that none of them accurately predict levels of concentration; 
rather, they are used to compare the potential water loading effects of 
different management scenarios. That is, a simulation for a specific 
site can be run and then certain model parameters can be adjusted to 
reflect changes in management practices. Then a second simulation can 
be run and the results of the two simulations can be used to compare the 
relative potential loadings of the two management scenarios. 
29 
PRZM, GLEAMS, and HSPF all follow the same general steps in esti ­
mating chemical movement through the unsaturated zone. To provide a 
clearer view of some of the processes involved, all three models will be 
briefly discussed, with a more detailed description of PRZM provided in 
the appendix. To give the reader an idea of the input and output of 
these models, model applications are also presented. 
Management-Oriented Chemical Transport Models 
The Pesticide Root Zone Model (PRZM) 
Carsel et al. (1984) provide a detailed description of the Pesti­
cide Root Zone Model. PRZM was designed by the Environmental Protection 
Agency to evaluate pesticide leaching threats to groundwater under 
differing agricultural crops and management practices and was developed 
to be used in pesticide registration. PRZM does not model nutrient 
movement. It is a dynamic, compartmental model for use in simulating 
pesticide movement in the unsaturated soil systems within and below the 
root zone. PRZM is divided into two main components: hydrology and 
chemical transport. The chemical fate and transport mechanisms repre­
sented include advection, dispersion, sorption, degradation, and plant 
uptake. Water balance equations are used to determine runoff, evapo­
transpiration, and infiltration. The PRZM soil profile is divided into 
as many soil layers or compartments as necessary to adequately represent 
the profile. For each compartment the model solves a finite difference 
advection-dispersion equation on a daily basis. 
In the PRZM model three basic components must be solved to estimate 
potential pesticide loading threats to groundwater: (1) water balance 
30 
in the soil profile, (2) erosion from the soil surface, and (3) chemical 
transport in the soil. These three components contain the four elements 
of groundwater contamination models de~cribed earlier (Novotny, 1986). 
PRZM's water balance component includes surface runoff generation 
(Novotny's first component) as well as the routing of water through the 
root zone. The erosion component in PRZM estimates soil loss due to 
erosion (Novotny's third component). PRZM's chemical transport component 
requires the input of the chemical's partition coefficient (Kp) and 
degradation rate (Novotny's fourth component) and simulates chemical 
movement through the root zone {Novotny's second component}. The three 
PRZM components are linked together in the advection-dispersion equa­
tions. A description of these equations is given in the appendix. 
The Groundwater Loading Effects of 
Agricultural Management Systems (GLEAMS) Model 
Groundwater loading Effects of Agricultural Management Systems 
{GLEAMS} is a mathematical model developed for field-size areas to 
evaluate the effects of agricultural management systems on the movement 
of agricultural chemicals within and through the root zone. GLEAMS, an 
extension of Chemicals, Runoff, and Erosion from Agricultural Management 
Systems {CREAMS} {Knisel, 1980}, was developed to utilize the management­
oriented, physically-based CREAMS model and considers the vertical move­
ment of agricultural chemicals. GLEAMS contains four basic components: 
{1) hydrology component, (2) erosion component, (3) pesticide component, 
and {4) nutrient component. 
31 
The Hydrological Simulation Program­
FORTRAN {HSPF) Model 
Barnwell and Johanson (1981) provide a brief description of the 
HSPF model. A more comprehensive description can be found in Johanson 
et al. (1984). HSPF, a basin-scale model, was developed for the 
Environmental Protection Agency and is a descendant of the Stanford 
Watershed Model. It is a comprehensive program for modeling runoff, 
sediment, pesticides, nutrients, and other water quality constituents 
from urban, agricultural, and other land uses. HSPF is a data-intensive 
model and requires large amounts of computer time for simulation. HSPF 
evaluates Best Management Practices (BMP) effects on runoff and pollutant 
loadings into field sized areas, and then links these runoff and pollu­
tant loadings with models to assess the quality in receiving waters 
resulting from BMPs used on the watershed. HSPF does model chemical 
transport through soil, but it was designed primarily to evaluate surface 
water qua 1 i ty, not groundwater qua 1 i ty. Chemica 1 1 oadi ng of sha 11 ow 
groundwater is included in this model because shallow groundwater can 
transport contaminants laterally to surface water. In HSPF, shallow 
groundwater is modeled as a medium of transport for the chemical, not as 
a receiving body of water; consequently, deep groundwater is not modeled. 
HSPF could be used to model secondary effects of shallow groundwater 
contamination such as surface water contamination. A 1 so, HSPF is of 
interest because it includes a risk assessment component. 
HSPF consists of three application modules: PERLND, IMPLND, and 
RCHRES. PERLND simulates a pervious land segment with homogenous 
hydrologic and climatic characteristics. IMPLND simulates impervious 
'· I 
I . 
I 'I 
32 
land segments with little or no water infiltration. RCHRES simulates 
the processes occurring in a sing 1 e reach of an open channe 1 or a 
completely mixed lake. Water movement is modeled along three flow paths: 
overland flow, interflow, and groundwater flow. To estimate the water 
quality effects of agricultural chemicals on a basin stream or river, the 
amount of loading from all three waterflow paths is summed. Because HSPF 
can model a larger variety of situations than the other models discussed, 
it is more complicated than GLEAMS and PRZM. However, HSPF models the 
groundwater loading processes in much the same way as GLEAMS and PRZM. 
The Chemical Migration and Risk Assessment Methodology (CMRA) 
developed by Battelle Pacific Northwest Laboratories has been incor­
porated into HSPF. The methodo 1 ogy inc 1 udes submode 1 s of sediment 
transport by particle size, pesticide routing and degradation kinetics, 
and sediment-pesticide interaction. A frequency analysis provides a 
statistical summarization of time-varying contaminant concentrations and 
provides the link between simulated instream toxicant concentrations and 
risk assessment. 
Model Applications 
Pesticide Root Zone Model (PRZM) 
Application 
The fo 11 owing information describing the PRZM app 1 i cation is derived 
from the work of Donigian and Carsel (1987), "Modeling the Impact of 
Conservation Tillage Practices on Pesticide Concentrations in Ground and 
Surface Waters." This PRZM application illustrates the linking of an 
unsaturated zone model with a saturated zone model. PRZM can model. 
/ 
33 
pesticide movement down to the saturated zone, providing the depth to 
the saturated zone is relatively shallow. To determine probable 
concentrations of a pesticide in groundwater, the Analytical Transient 
One-, Two-, ~nd Three-Dimensional Simulation of Water Transport in the 
Aquifer System (AT123D) is linked with PRZM. This application evaluates 
different pesticides and different tillage practices applied to a corn 
and soybean rotation in the Lake Erie Basin, and determines the probable 
concentrations of chemicals in groundwater and surface water under 
different tillage practices. 
In order to incorporate the effects of the tillage practices into 
the PRZM model, model parameters are adjusted to reflect these effects. 
The parameters selected to represent these impacts on soil hydrology, 
sediment erosion, and chemical processes are: the SCSCN, the soil field 
capacity, the soil organic carbon content, and the USLE. To estimate 
groundwater loading, the chemical transformation rate is needed. 
Donigian and Carsel point out that available data on transformation rates 
of pesticides in subsurface soil horizons and groundwater are limited, 
and that the transformation rate will depend on soil conditions (pH, 
temperature, etc. ) . Based on 1 i mi ted 1 aboratory data, the authors assume 
that the lower horizon decay rates of the pesticides examined are 50 
percent of the rate in the surface horizon. 
In this situation, the depth to groundwater is sufficiently shallow 
that PRZM can model chemical movement all the way through the unsaturated 
zone, and the chemical loading simulated can be fed directly into the 
groundwater model. The groundwater model can then calculate the ground­
water chemical concentration as a function of unsaturated zone loadings 
34 
(simulated by PRZM), pesticide properties, and hydrogeologic conditions. 
The outputs of th.e groundwater model are the mean estimates of the 
concentration of chemicals in groundwater resulting from the use of 
different chemicals and tillage practices. However, estimates of mean 
concentration may not be useful for a risk analysis. Carsel et al. 
(1988) use a Monte-Carlo simulation technique to assess potential 
groundwater loadings of a chemical. To use this technique, probability 
density functions are determined, a Monte-Carlo simulation is run, and 
the predicted amounts of chemical residue are used to construct a 
cumulative probability distribution which, in turn, can be used to 
predict the percentage of simulations that resulted in residue movement 
below a certain soil depth. 
Groundwater Loading Effects of Agricul­
tural Management Systems (GLEAMS) 
Model Appljcatjon 
The following information describing the GLEAMS model application 
is derived from the work of Leonard and Knisel (1988), "Evaluating the 
Groundwater Contamination Potential from Herbicide Use." This example 
application demonstrates some of the functional relationships that can 
be developed with a chemical fate and transport model. Because GLEAMS 
models chemical movement only through the root zone, it also illustrates 
the application of a model that does not simulate chemical movement down 
to the saturated zone (unless the saturated zone is sufficiently 
shallow). The goal of this application is not to estimate chemical 
concentrations in groundwater, but rather to estimate relationships 
between chemical and environmental characteristics and leaching loss 
35 
(where leaching loss is defined as chemical that moves through the soil). 
The application consists of a comparison of leaching loss on two fields 
with different soil types in the Georgia Coastal Plains. A 50-year 
continuous simulation is used to estimate herbicide leaching losses when 
different chemical partition coefficients and half-life values are used. 
A chemical's partition coefficient and half-life are functions of the 
chemical's properties and the environmental characteristics. In this 
application, Leonard and Knisel do not present the amount of chemical 
leaving the root zone or the depths to which leaching occurs, although 
GLEAMS can estimate this. Instead, the authors determine the distribu-
tion of simulated chemical leaching losses. Leaching losses are 
expressed in percent of the application applied (50-year means are 
simulated). 
According to Leonard and Knisel (1988), 
A difficulty encountered in applying continuous simula­
tion mode 1 s is the generation of so much specific 
detail that generalizations are difficult. A possible 
solution to this problem is to fit the specific model 
output to general relationships .... (p. 213) 
Therefore, to compare the results from the two fields, the authors 
estimate the relationship between the leaching losses (as a percent of 
application) and a half-life/adsorption ratio. Both a log-linear and a 
1 inear functional form are used. Again, 50-year means are used; however, 
Leonard and Kn i se 1 also acknowledge that mean va 1 ues may not be of 
interest in risk analysis. As an alternative to using a 50-year mean 
value, the authors use the annual pesticide leachate losses simulated for 
the 50-year period to express the percent of years that the amount of 
pesticide leached exceeds specified percentages of application. This 
36 
probabilistic approach may be more suitable when examining threshold­
exhibiting health hazards. 
Hydrological Simulation Program­
FORTRAN (HSPF) Model Application 
The following information describing the HSPF model application is 
derived from the work of Bicknell et al. {1985), "Modeling Water Quality 
and the Effects of Agricultural Best Management Practices in the Iowa 
River Basin." 
This HSPF application simulates water quality in the Iowa River 
Basin under a base scenario and a Best Management Practices {BMP) scen­
ario. The BMP scenario consists of: (1) a shift from moldboard plowing 
to chisel plowing and field cultivation as primary tillage, (2) one 
summer cultivation for weed control instead of two, and {3) allowing crop 
residue to remain on the field following harvest. This application 
evaluates the water quality of the receiving body, the Iowa River at 
Marengo, Iowa. Because groundwater flows into this river, the contam­
inant contribution of shallow groundwater is simulated. 
Alachlor, an herbicide, was used in this simulation; alachlor was 
chosen because it is used extensively on corn and soybean cropland in the 
basin. This simulation considered the following processes: application 
of the chemical to corn and soybean cropland, adsorption to soil, degrad­
ation, dissolved and adsorbed transport to streams, and instream adsorp­
tion, decay, and transport. The nutrient processes are also simulated 
including application of fertilizer and plant residue in chemical form 
to the land surface or incorporated into the soil, the soil nitrogen 
cycle, adsorption and desorption, nutrient movement with soil and water 
, I 
37 
to the stream, ammonia volatilization (loss of chemical into the atmos­
phere), and nitrification. The model parameters adjusted to reflect the 
effects of the BMP were soil moisture retention values, rainfall inter­
ception, surface roughness, land cover, and the sediment fines produced 
by tillage. 
The simulation estimated edge-of-stream loadings for the base scen­
ario and the BMP scenario. Edge-of-stream loadings of runoff, sediment, 
alachlor, nitrate, and ammonia were simulated. The total edge-of-stream 
loadings of runoff (mm), nitrate (kg/ha), and ammonia (kg/ha), is equal 
to the sum of surface runoff, interflow runoff, and groundwater runoff. 
The results of this simulation are used to assess the risk or 
exposure of aquatic organisms to various magnitudes and durations of 
chemical concentrations. The simulations of alachlor concentrations 
under the base scenario and the BMP scenario are examined to determine 
the percent of time acute, chronic, and sublethal concentrations exist. 
Conclusions 
This chapter discussed some of the basic features of chemical fate 
and transport models and also presented some applications of these 
models. Management models were emphasized because they are most useful 
for policy analysis. For example, if a policy restricts herbicide use, 
producers may substitute a different type of tillage practice for the 
restricted herbicide. A model that considers these changes in management 
practices (such as the three mode 1 s presented above) is necessary to 
estimate the change in potential groundwater pollution resulting from the 
implementation of an agricultural policy. 
38 
One major shortcoming of all three models is that only in certain 
situations can they be used to simulate chemical movement down to the 
groundwater. In situations where the groundwater zone is sufficiently 
deep, a gap exists between the lowest zone simulated (e.g., the bottom 
of the root zone) and the groundwater zone. Another shortcoming of these 
models is that they must be used to compare the relative effects of 
different management practices or scenarios; these models do not give 
absolute predictions of chemical loading (Leonard and Knisel, 1988). 
' I I 
39 
CHAPTER 4 
LINKING THE ECONOMIC AND 
ENVIRONMENTAL MODELS 
In the previous chapters, economic production and chemical fate and 
transport were viewed as separate processes. This chapter uses stylized 
versions of physical and economic models to illustrate the linkage 
between these processes. Key to the linkage is the dynamic nature of 
both processes. 
Consider first the dynamics of the environment. Damage-producing 
agents that are introduced into the environment eventually affect 
environmental quality. If the time elapsing between the introduction of 
these damage-producing agents and the manifestation of the damage is at 
least one production season, this relationship exhibits inter-seasonal 
dynamics and can be expressed as 
at = f(~.p Zt-t> 
where a is a vector of environmental characteristics and z is a measure 
of damage occurring {e.g., the amount of contamination leaching into 
groundwater). This assumption of inter-seasonal dynamics is a reasonable 
characterization of environmental dynamics because pollution is a 
cumulative process and usually becomes apparent in later seasons. 
The vector of environmental characteristics, a, can be divided 
into two subvectors: (1) a1 , a vector of characteristics that affect 
40 
agricultural productivity and chemical transport, and (2) a2 , a vector of 
characteristics that do not affect productivity but do affect chemical 
transport. 
The agricultural production process may exhibit intra-seasonal 
dynamics, inter-seasonal dynamics, or both, because a producer's 
optimizing decisions may include decisions made at the intensive margin 
and at the extensive margin. Decisions made at the extensive margin 
(e.g., what fields to put into production) are often inter-seasonal and 
therefore environmental characteristics may be considerations in these 
decisions. When inter-seasonal decisions are part of the agricultural 
production process, the producer's objective function can usually be 
specified as a multiperiod model. Decisions made at the intensive margin 
are often intra-seasonal (e.g., pesticide application rates) and the 
producer's objective function can usually be specified as a multistage 
model. Because changes in environmental quality resulting from 
agricultural chemical use may take a relatively long period of time to 
manifest themselves, environmental quality can be assumed to be constant 
over a production season. Therefore, for intra-seasonal decisions, 
environmental quality would be taken as given for that season. 
To illustrate intra-seasonal decisions in a model, let lilt be a 
vector of management decisions made during season t: 
"'t= g(m1,t' ~.t' ... , mn,t)· 
The management practices of a field interact with the environmental 
characteristics of that field to create environmental damage; this 
physical relationship can be expressed as a damage function. The damage 
function may be a relatively simple relationship. For example, Nielsen 
i ; 
i 
41 
and Lee (1987) matched areas of "high" agricultural chemical use with 
areas of "high" DRASTIC scores. The damage function may also be as 
complex as a chemical fate and transport model. 
The 1 i nkage between the agri cultura 1 production mode 1 and the 
environmental model requires some assumptions about the properties of the 
damage function (~) in relation to the vectors of environmental charac­
teristics (a1 ~ and a2~). Depending on the environmental characteristics 
being considered, zt may be a function of a1,t, or a2.t' or both. For 
example, if ~ represents potential chemical loading to groundwater, ~ 
represents chemica 1 app 1 i cations, a1.t represents soi 1 qua 1 i ty, and ~.t 
represents groundwater quality, then zt may be expressed as: 
zt = z(~, a1,t) 
This relationship assumes that current groundwater quality does not 
affect the amount of current groundwater loading. However, there may be 
cases where ~ is a function of a2,t. For example, if a2,t had a threshold 
level beyond which potential loadings are affected, then Zt would be a 
function of a2,t· 
To illustrate the linkage of economic and environmental models, 
consider the following set of relationships: 
~ = z(~, a1,t) [4.1] 
a1,t = a(~., a1,t·1) [4.2] 
a2,t = a<zt-1, a2,t·1) [4.3] 
Note that Zt is a function of~ and a1,t, and that a2,t'is a function of ~.1 • 
These relationships are illustrated in Figure 4. Although the damage 
function does not affect the productivity-related environmental charac­
teristics, management practices do. 
I r 
42 
t 
POLLUTION (z) MANAGEMENT (m) 
t 
SOIL QUALITY (a1) GROUNDWATER QUALITY (a2) 
t-1 t-1 
Figure 4. Relationships between economic and environmental 
models in the evaluation of groundwater pollution. 
The economic and en vi ronmenta 1 mode 1 s may be 1 inked through the 
production function. As Figure 4 illustrates, only those environmental 
characteristics that affect productivity (the vector a1) will be factors 
in the producer's optimizing decisions: 
qt = q (fit, a,,t) · 
Substituting equation [4.2] into equation [4.4] yields 
qt = q[fit, a,,t(fit.p a,,t_,)] 
qt = q (fit' fit.p • · • ' fit-n) · 
[4.4] 
[4.5] 
[4.6] 
When a1 ~ is a function of past management decisions, output is a function 
of past and present management decisions. Thus, in the nomenclature of 
Chapter 2, the production process exhibits input dynamics. 
If the vector a1.t is an argument of the damage function, the 
relationship between economic and environmental models can be expressed 
as a joint production function. Referring back to equation [4.1], the 
damage function can be expressed as: 
Zt = z (fit, a1,t) and 
dz = (az/aa1)da1 + (az/am)dm. 
43 
Assuming lilt represents chemical use and a1.t represents soil quality, one 
would expect the following relationships: 
aztf alllt > 0 and 
aztf aa1,t < 0. 
Assuming diminishing marginal product implies: 
a2zjam2 < 0 and 
a2zjaa/ < o. 
If the above relationships hold, the damage function can be presented as 
a series of positively sloped, concave isoquants as in Figure 5. This 
figure illustrates that while holding management practices constant, an 
increase in soil quality (an increase in a1) will result in a decrease in 
environmental damage. 
a1 
0~---------------------------------------- m 
Figure 5. The damage model as a function of environmental 
characteristics and management practices. 
44 
To demonstrate that the production process jointly produces pollu­
tion and agricultura1 output, equation [4.1] can be inverted: 
a1,t = z"
1(Zt, fit). [4.7] 
Substituting equation [4.7] into equation [4.4] yields: 
qt = Q[fit, z·1(Zt, fit)] 
[4.8] 
Equation [4.8] illustrates the joint production of agricultural output 
and environmental damage. This joint production function depicts the 
combinations of output of the agricultural good and the amount of 
chemical loading to groundwater possible by varying soil quality (Figure 
6). In Figure 6, m is being held constant and a is varied. As a 
changes, the resulting levels of q and z also change. This curve should 
be interpreted as an output-damage transformation surface and not as a 
production possibilities frontier because in this case variable inputs 
are held constant. 
The position of this transformation surface within the quadrant is 
determined by fit, the amount of chemica 1 used during period t, and the 
producer's position along the transformation surface is determined by 
the soil quality. The shape of the isoquant depends on properties of 
the production and environmental damage functions. Assume that the 
transformation surface is concave as in Figure 6. If two producers used 
identical types and amounts of chemicals (fit), the producer with rela­
tively poor soil quality (e.g., soil conducive to groundwater pollution) 
would be operating at point B, whereas the producer with relatively good 
soil quality would be operating at point A. The producer at point A 
I 
I 
I I 
I ! 
45 
obtains more output and less environmental damage than the producer at 
point B. 
q 
z 
Figure 6. Output-damage transformation surface. 
In this simplified model of economic and environmental interactions, 
the producer takes environmental damage into account only when a 
productivity-related environmental characteristic is being damaged. 
Assume: 
a1,t = a<Zt-1' a1,t-1> 
qt = q (ffit, a1 ,t) • 
Substituting equation [4.9] into equation [4.10] yields: 
qt = q[ffit, a(zt·P a1,t-d l 
qt = q (ffit, zt-P a1,t-1) • 
[4.9] 
[4.10] 
[4.11] 
i I 
46 
In this case the costs of the changes in a1 are being internalized by the 
producer, although any costs associated with changes in a2 are not. 
However, as equations [4.1] through [4.6] indicate, if the environmental 
damage resulting from chemical use does not affect productivity, then the 
producer wi 11 not consider that environmental damage when making his 
production decisions. Both equations [4.6] and [4.11] indicate that all 
of the damage to the a2 characteristics created by agricultural produc­
tion is an externality. 
One complication that can arise when 1 inking the production· and 
environmental models is the difference in time steps. The production 
model is constructed to express chemical use in terms of quantity of 
chemical applied at a point in time, whereas the chemical transport 
models are usually constructed to simulate changes over time. In order 
to reconcile the two models, the user may need to calculate the total 
chemical loading during a time interval meaningful to management 
decisions. For example, physical models can simulate change in chemical 
concentration per day or on a continuous basis (e.g., az;at), whereas the 
economic model examines the total change in chemical concentration over 
many days (e.g., Zt)· The physical model results can be used in an 
economic model by calculating zt as follo~s: 
t, 
I dz/dt dt = z(td - z(t0 ) = zt 
to 
where t 0 is time at the beginning of the chemical model simulation, t 1 is 
the point in time being evaluated, and Zt is the total potential loading 
at time t. 
47 
So far, the level of aggregation of this simplified model has not 
been addressed. Although a deterministic model may be useful for a 
re 1 at i ve 1 y homogeneous area of 1 and, for other areas (e.g. , 1 arger 
areas), variations in environmental characteristics and production 
practices may be significant and a statist i ca 11 y va 1 i d description of the 
region is needed. Following Antle and Just (1989), a producer makes 
intensive and extensive production decisions based on input and output 
prices, policy factors, and individual farm characteristics (i.e., 
technology and risk attitudes of the producer). The extensive decisions, 
such as what fields to put into production, determine the environmental 
characteristics of 1 and in production. The producer a 1 so makes intensive 
management decisions such as chemical use, tillage practices, etc. 
Therefore, the producers in an area are jointly determining management 
practices and environmental characteristics. The interactions of the 
management practices and the environmental characteristics then jointly 
determine the production of the agricultural output and of pollution for 
that area of land in production. Assume that 
m = m(p, w, r) [4.12] 
[4.13] 
where p is a vector of input and output prices, w is a vector of policy 
factors, and r is a vector of individual farm characteristics (i.e., 
technology and risk attitudes of the producer). Within a geographical 
region (and assuming perfect competition in factor and product markets), 
all producers will have identical price and pol icy vectors; only the 
individual farm characteristics will vary. Through producers' management 
and land use decisions, management practices and environmental 
' I 
I I 
48 
characteristics form a joint probability distribution. For this example, 
the joint probability distribution of management practices and environ­
mental characteristics can be expressed as 
¢(m, a: p, w, r). [4.14] 
Both output and environmental damage are functions of this joint 
distribution; the expected amount of damage can be expressed as 
E(z) = Jz(m, a) ¢(m, a: p, w, r) dm da. [4.15] 
Antle and Just (1989) present a log-linear production function and 
damage function that illustrate the effects of policy on the distribution 
of input use and resulting environmental damage. The damage function is 
expressed as 
a, ~ > 0 [4.16] 
and the agricultural production function is ~xpressed as 
0 < q < 1. [4.17] 
Assume that ln m and ln a are normally distributed and that z is log­
normally distributed. Antle and Just (1989) demonstrate that the shape 
of the joint production function of q and z depends on properties of the 
production and environmental damage functions. Given the model presented 
in equations [4.16] and [4.17], the joint production function can be 
expressed as 
ln y = (q-av/~)ln m + {v/~)ln z 
where a is varied and m is held constant. The shape of this joint 
production function depends on the value of v. 
Figure 7 presents possible shapes of this particular joint produc­
tion function whe~e a is varied and m is held constant. When v is 
positive, an increase in a results in an increase in both q and z for any 
49 
level of m. If 1 > p-v > 0, as a increases, z increases at a decreasing 
rate; if v > p 0, as a increases, z increases at an increasing rate; if 
p = v, as a increases, q and z increase at an equal rate. When v is 
negative, an increase in a will result in an increase in q and a decrease 
in z. This example illustrates that the amount of environmental damage 
occurring as a result of a policy will be determined in part by the 
relationship between output and environmental damage. 
q 
0 
Figure 7. 
----1 > p-v > 0 
v < 0 
z 
Output and damage relationships in the integrated 
log-linear physical-economic production model 
(Antle and Just, 1989). 
A policy can influence a producer's production decisions at the 
intensive margin, the extensive margin, or both, which in turn affects 
the joint distribution of management practices and environmental 
50 
characteristics expressed in equation [4.14]. The amount of expected 
damage that results from a policy will depend on the changes in this 
joint distribution as well as the relationship between the productivity­
and pollution-generating characteristics of a piece of land. For 
example, if highly productive land is more conducive to pollution than 
less productive land, a policy that encourages bringing highly produc­
tive land into production would result in an increase in expected 
damage. 
To apply this statistical model to the analysis of potential 
loading of agricultural chemicals to groundwater, the management vector 
could represent chemical use (e.g., types of chemicals and amounts 
applied) and the vector of environmental characteristics could represent 
those characteristics important in determining an area's vulnerability 
to groundwater contamination (e.g., DRASTIC scores). (A DRASTIC score 
is a composite estimate of an area's potential vulnerability to ground­
water pollution. The DRASTIC data base is discussed in more detail in 
Chapter 7.} Nielsen and Lee (1987} use a relatively simple application 
of this type of stochastic framework in a study that examines potential 
for groundwater pollution across the United States. The authors esti ­
mate the distribution of management practices (expressed in terms of 
"high" and "medium" levels of chemical application) and the distribution 
of environmental vulnerability to groundwater contamination (expressed 
in terms of "high" and "medium" DRASTIC scores). The joint distribution 
of management practices and environmental quality is estimated and the 
implied damage model is a simple relationship: an area with a high 
DRASTIC score and high chemical use would have a greater potential for 
51 
groundwater pollution than an area with a medium DRASTIC score and 
medium chemical use. This is a simple example of linking production 
practices (chemical use) and hydrogeologic characteristics (DRASTIC 
scores) to estimate groundwater pollution potential. Because Nielsen 
and Lee do not model the agricultural production process, the effects of 
the hydrogeologic characteristics and the damage function on producers' 
intensive and extensive production decisions are not examined. 
I I 
i 
' ' 
I \ 
i 
I 
I I 
52 
CHAPTER 5 
VALUING HEALTH RISKS 
An important component in the determination of the social costs of 
groundwater pollution is the valuation of human health risks. This step 
can be relatively complicated because it involves the valuation of a non­
market good, and because of the psychological considerations that 
accompany this type of valuation. A variety of methods have been used 
to determine the value of a human life. Mishan (1970) describes four 
commonly used. methods of estimating the economic value of a life. The 
"gross output" approach estimates the loss to the economy as a result of 
an individual's death. This estimation consists of the discounted 
present value of the individual's expected future earnings. The "net 
approach" is similar to the gross approach; however, this estimation 
calculates the present discounted value of the losses over time accruing 
to others as a result of the death of the person at age "t". The third 
method uses the value that policy makers implicitly assign a life in the 
creation of policy. The fourth method, the insurance principle, assumes 
than an individual determines the value of his life based on the 
insurance premium he pays. 
Mishan (1970) raises theoretical objections to all of the above 
methods, his strongest objection being that all four methods ignore the 
concept of potential Pareto improvement. Potential Pareto improvement 
'i 
' i I 
I 
53 
occurs when the sum of all individuals' compensating variations are 
greater than zero. Compensating variation is a measure of the change in 
an individual's welfare; it is the amount of money that can be taken away 
from an individual after an economic change, while leaving him as well 
off as he was before the change. For a welfare gain, it is the amount 
he would be willing to pay for the change. For a welfare loss, it is the 
amount he waul d be wi 11 i ng to accept as compensation for the change. 
Equivalent variation is another measure of welfare change; it is the 
amount of money we would need to give an individual if an economic change 
did not happen, to make him as well off as if it did. For a welfare 
gain, it is the compensation he would be willing to accept to forego the 
change. For a welfare loss, it is the amount he would be willing to pay 
to avoid the change. 
Potential Pareto improvement occurs when the sum of all individuals' 
measures of compensating variation is greater than zero. The problem 
with measuring an individual's change in welfare is that when a person's 
life is threatened, his willingness to pay to avoid death (a measure of 
equivalent variation) may be bound by his income. Alternatively, an 
individual may decide that the amount of compensation necessary to 
persuade him to accept death (a measure of compensating variation) may 
be unbounded. This problem of potentially unbounded compensating 
variation and income-bound equivalent variation can be avoided by using 
the concept of the "statistical life" that considers only small changes 
in health risk. A project that changes the number of statistical lives 
saved does not save specific individuals, but rather changes the health 
risks or probability of death faced by a group of individuals. Using the 
54 
concept of a statistical life, the compensating variation can be used to 
measure the willingness to pay for a reduction in probability of death. 
Mishan {1970) addresses another aspect that affects the valuation 
of life, the distinction between voluntary and involuntary risk. When 
evaluating the benefits of a voluntarily assumed service, risks 
associ a ted with that service do not have to be valued because the 
individual using the service is already taking those risks into account. 
However, involuntarily assumed risk must be included in a benefit-cost 
analysis. To clarify this distinction, consider the case of a pesticide 
application. According to Mishan, any health risks the pesticides have 
on the pesticide applicator do not have to be valued because the 
applicator has already accounted for those risks when he evaluated the 
costs and benefits of using the pesticides. (This assumes that the 
applicator is aware of the risks.) The applicator's willingness to pay 
to use pesticides {his compensating variation) is net of the value of the 
health risks; to include the value of the health risks in a benefit-cost 
analysis would be to double-count the risks. However, the health risks 
faced by other individuals (e.g., the health risks attributed to 
pesticide-contaminated groundwater) should be included in the benefit­
cost analysis because those individuals involuntarily exposed to the 
risks have not accounted for these costs in their private decisions. 
Mishan (1970) advocates that the valuation of life should be based 
on a method that is consistent with the Pareto base of existing 
allocation theory and benefit-cost analysis, such as summing individual 
compensating variation. However, applying such a measure becomes 
difficult because of the psychological considerations that can influence 
I 
i I 
! 
' I ! 
I I 
55 
an individual's compensating variation. Weinstein and Quinn (1983) 
examine the normative value of health risk reductions, and present some 
of the psychological motives for society not implementing least-cost 
policies of lifesaving. In keeping with Mishan's (1970) results, 
Weinstein and Quinn illustrate that involuntarily-assumed risks may be 
psychologically less acceptable than voluntarily-assumed risks. 
Weinstein and Quinn also point out that the concept of blame and ethical 
responsibility may influence policy making; ex ante identification of 
victims may cause society to feel more responsible, and ex post identi­
fication may cause decision-makers to feel more accountable. 
In the above discussion, health risks are perceived as being two 
separate commodities, voluntarily-assumed health risks and involuntarily­
assumed health risks. However, it is possible that the framing or 
wording of the description of a commodity can result in an individual 
assigning different values to one commodity. For example, if an 
individual is asked to value a change in the level of provision of a 
commodity, and that commodity is framed two different ways (e.g., in 
terms of an increase in lives saved versus a decrease in lives lost), the 
individual may have two different values for that commodity. Framing 
effects can result from descriptions of the change in the level of 
provision of the commodity, as well as the method of payment for that 
change. 
In normative decision theory, subjective probabilities of occurrence 
are supposed to be based only on beliefs about the likelihood of events; 
however, psychological research has found that these subjective proba­
bilities reflect beliefs about both the likelihood and the individual's 
( I 
i I 
56 
preferences regarding the occurrence (a "wishful thinking" effect whereby 
a positive relation exists between the value of a consequence and its 
perceived likelihood of occurrence). Weinstein and Quinn (1983) also 
explain that, according to normative decision theory, choices and actions 
should follow from beliefs and values, although behavioral research 
indicates that the reverse may sometimes be true. This "pseudo­
certainty" effect may also play a role in valuing health risk reductions; 
a project that gives the individual a feeling of certainty is given a 
greater value than the project that does not, even if the total 
(statistical) lives saved is the same for both projects. 
J 
Another important psychological determinant of an individual's 
valuation of a health risk reduction is the base level of risk involved. 
According to Weinstein and Quinn (1983) and Weinstein et al. (1980), the 
higher the base level of mortality probability, the higher the willing­
ness to pay for a decrease in the mort a 1 i ty probability. In a 1 ater 
study, Smith and Desvouges (1987) find the estimated valuation of a risk 
change decreases with increases in the level of risk. This result would 
appear to be consistent with the "pseudo-certainty" effect. 
The Smith and Desvouges (1987) study resulted in several important 
findings, including: (1) the estimated valuation of a change in the 
level of risk decreases with increases in the baseline level of risk, and 
(2) the valuation of a change in risk depends on the direction of the 
risk change. In the 1987 study, the individuals' mean bids to decrease 
risk were "uniformly greater" than the bids to avoid risk increases. 
As an alternative to measuring compensating variation, Thaler and 
Rosen (1976) use a hedonic method to determine the value of saving a 
I \ 
: I 
I 
57 
statistical life. They establish a relationship between job risk and 
wage rates in order to impute a set of implicit marginal prices for 
various levels of risk. Thaler and Rosen then estimate the market wage­
risk-personal characteristics equalizing difference function that matches 
workers and firms. The regression of this equation indicates a 
positively sloped wage-risk function, with the risk regression coeffi­
cient estimating the marginal wage rate paid to workers to assume another 
increment of risk. For example, if 1000 workers are employed at a job 
entailing an extra death risk of .001 per year, the yearly wage 
differential aggregated over all workers can be calculated, with this 
aggregate differential being the aggregate willingness to pay to save a 
statistical life. 
Household production theory can be used to illustrate how chemical­
related health risks affect producer decisions. Health can enter the 
economic production framework in several ways. Health effects may enter 
the production function directly if these effects decrease the farmer's 
1 abor productivity. For examp 1 e, agri cul tura 1 chemica 1 s may have an 
effect on the farmer's health that diminishes his ability to work; this 
in turn affects the farmer's labor, a factor in the production function. 
Household production theory assumes households to be both producers and 
consumers. Becker (1971) shows that the cost per unit of a household­
produced commodity, h1, is the sum of the costs of purchased goods per 
unit and the "shadow" or opportunity cost of time: 
1r1 = a1p1 + b1w 
where 1r1 is the total costs per unit of h1 produced, a1 and b1 are fixed 
input-output coefficients, p1 is the cost of purchased goods required to 
i I 
I l 
I I 
58 
produce a unit of h1, and w is the amount of time necessary to produce one 
unit of h1• 
A change in an environmental variable (i.e., age, health, education) 
would change the amounts of goods and time required to produce a unit of 
h1• The change in the environmental variable would change the efficiency 
of producing a unit of h1 by changing the input-output coefficients, a1 
and b1, thereby changing the marginal cost of h1• For example, a farmer's 
health can enter the household production function as an environmental 
variable and can be incorporated as some probability of, for example, 
cancer or death. Agricultural production decisions can affect the 
producer's health, in turn affecting available labor. Because health is 
a factor in the househo 1 d production function, he a 1 th wi 11 a 1 so be a 
factor in the farmer's utility function. 
This overview of methods and considerations of valuing human health 
risks illustrates that valuation is specific to the situation being 
considered. Smith and Desvouges (1987) showed that the level of baseline 
risk influenced valuation, as did the direction of change in the 
probability of health risks. Weinstein and Quinn (1983) explained some 
of the psychological factors affecting valuation. Their results indicate 
that extrapolation of survey results from one situation to another may 
not be appropriate. Therefore, to place a value on the health effects 
of groundwater pollution {using survey data), one would need to obtain 
data specific to that·case of groundwater pollution. 
Thaler and Rosen's (1976) method of determining the implicit value 
of saving a life could be modified to determine the value of the health 
risks resulting from agricultural chemicals. For example, one approach 
! 
' I 
I 
! I 
59 
would be to compare the wage differential of a pesticide applicator and 
a tractor operator, the differential being the value of the health risks 
of applying pesticides. A less precise approach would be to determine 
the upper bound on the value of the health risks. For example, if a 
chemical applicator could avoid the risk of health effects by using a gas 
mask, and if the applicator chooses not to wear one, then the price of 
the mask is the upper bound on the value of the risk; if the applicator 
does not use a mask, the value of the health risk must be less than the 
value of the mask. 
The above methods deal with the health risks associated with the 
application of chemicals. The valuation of health risks associated with 
polluted groundwater may be more difficult because using wage differen-
tials is no longer appropriate. One possible method of valuing 
groundwater health risks would be to use the value of a groundwater 
purification system as an upper bound on the value of health risks; this 
method was used by Nielsen and Lee (1987) in their valuation of the 
social costs of groundwater pollution. Alternatively, a hedonic approach 
may be used to determine a relationship between health effects and real 
estate values, the differential being the implied value of the risks. 
A problem with this approach is that one would expect land values to be 
measurably affected only when groundwater pollution levels are extremely 
high and land market participants know about the contamination and its 
effects. In the case of agricultural chemical pollution, these high 
levels may rarely occur. This technique may be useful in highly polluted 
situations, but referring back to the Smith and Desvouges (1987) study, 
I ; 
I I 
I I 
! 
I i 
I 
I I 
60 
the values calculated could not be extrapolated for areas with different 
base levels of pollution. 
As mentioned previously, a valuation method consistent with the 
concept of potential Pareto improvement involves determining individuals' 
compensating variations. These compensating variations (also called 
measures of willingness to pay, or WTP), can be elicited from individuals 
using survey techniques such as a contingent valuation (CV) survey. A 
CV is based on the concept that the elicited WTP bids are contingent on 
the situation presented to the survey respondents. An example of a CV 
survey will be presented in the case study in Chapter 8. 
61 
CHAPTER 6 
HEALTH RISK ASSESSMENT 
Merely determining the concentration of a groundwater contaminant 
may not be meaningful in itself. To make a level of pollution meaningful 
to people, the pollution can be expressed in terms of the health risks 
it presents. A health risk assessment can be used to place a value on 
a level of pollution or can be used for policy formulation; the 
Environmental Protection Agency uses risk assessment for chemical 
regulation. In order to quantify the human health risks presented by 
groundwater contamination, the types of he a 1 th effects and extent of 
exposure must be identified. The potential health effects from exposure 
to agricultural chemicals include a variety of acute and chronic effects. 
Much of the exposure assessment research has concentrated on the hazards 
related to pesticide exposure, possibly because the hazards associated 
with a pesticide application are relatively easy to observe and to link 
to the chemical. For example, it may be easier to trace a case of 
chemical· poisoning to a pesticide application than to groundwater 
contaminated by nitrogen fertilizer. 
Moses (1987) outlines potential pesticide-related health problems 
faced by farmworkers, including cancers, reproductive abnormalities, and 
mental and psychological changes. Moses notes that the difficulty 
involved in determining the health risks of a pesticide is heightened 
I i 
', I 
I ' 
I , 
62 
because of the "inert ingredients" in pesticides. Pesticides may contain 
both active ingredients and inert ingredients (ingredients that are not 
active against the target species); both types of ingredients may be 
toxic to humans, but the inert ingredients are not required to be tested 
for acute and chronic he a 1 th effects or to be 1 i sted by name on the 
pesticide label. This makes quantifying the health effects of agricul­
tural chemicals difficult because a complete analysis would consider all 
of the associated health effects, not just those caused by active 
ingredients. 
In order to value the effects of agricultural chemicals on human 
health, the risks the chemical presents must be assessed. The National 
Academy of Sciences defines risk assessment as "the scientific activity 
of evaluating the toxic properties of a chemical and the conditions of 
human exposure to it in order both to ascertain the 1 ike 1 i hood that 
exposed humans will be adversely affected, and to characterize the nature 
of the effects they may experience" (U.S. EPA, 1985, p. II-2). The 
Environmental Protection Agency (U.S. EPA, 1985) describes the four 
components used in a risk assessment: 
(1) Hazard Identification: This component consists of gathering and 
evaluating data on the types of health injury or disease that may 
be produced by a chemical and on the conditions of exposure under 
which injury or disease is produced. It is also important to 
determine whether it is scient i fica 11 y feas i b 1 e to extrapo 1 ate toxic 
effects from one setting to another setting (e.g., extrapolating 
effects that occur in laboratory animals to humans). Information 
on the toxic properties of a substance can be obtained from animal 
\ ; 
, I 
I 
(2) 
63 
toxicology studies, controlled epidemiologic investigations of 
exposed human populations, chemical studies or case reports of 
exposed humans, experimental studies on isolated animal organs, 
cells, etc., and the analysis of the molecular structure of the 
substance. Results from animal toxicology studies and epidemiology 
studies are the two principal sources of toxicity data. The use of 
animal toxicity data assumes the effects on humans can be inferred 
from the effects on animals; this EPA publication asserts that this 
assumption has been shown to be generally correct .. Also, Gough 
(1989a) demonstrates that estimates of human risk of cancer 
(associated with pollution) based on epidemiologic data and 
toxicologic data are similar. 
Dose-Response Evalyatjon: This evaluation determines the quantita­
tive relationship between the amount of exposure and the extent of 
toxic injury or disease. In most cases, dose-response relationships 
are estimated from animal studies. The noncarcinogenic effects of 
a substance are generally believed to exhibit threshold effects; 
that is, there is a dose below which adverse effects will not occur. 
The dose-response evaluation for these noncarcinogenic effects 
involves describing observed dose-response relationships and 
determining experimental threshold levels, also called "no observed 
effect levels" (NOEls). The NOEL can be divided by some specified 
(but arbitrary) safety factor in order to determine the Acceptable 
Daily Intake (ADI) of the substance for humans. 
Estimating dose-response relationships for substances that do not 
need to achieve threshold levels to cause adverse effects usually 
64 
requires the use of a mathematical model that estimates low-dose 
risks from high-dose risks. (It is generally thought that 
carcinogenic substances do not have thresholds, although there is 
some controversy regarding this point.) Estimating low-dose risks 
directly can be costly; research experiments can last years and a 
large number of observations is necessary. Regression analysis can 
be used to estimate a dose-response relationship by assuming the 
relationship has a particular functional form. Regulatory agencies 
currently use one-hit, multistage, and probit models, with 
regula tory decisions usually based on the one-hit or multi stage 
model results. Multihit, Weibull, and logit models may also be used 
for risk assessment. Mehlman et al. (1979), Lave (1982), and 
Tardiff and Rodricks (1987) all provide descriptions of the models 
used to describe the dose-response relationship. Brown {1987) notes 
that although these models yield similar results over the range of 
observable response rates (relatively high-dose levels), extrapolat­
ing results from high-dose levels to low-dose levels is dependent 
on the mathemat i ca 1 form of the mode 1 used; the further the 
extrapolation from the observable response range, the greater the 
divergence of model results. Brown {1987} feels that this 
divergence of results is the greatest limitation of this extrapola­
tion methodology and refers to mathematical dose response models as 
a "necessary evil" in the process of estimating low-level exposure 
effects. 
None of the above mentioned dose-response models is chemical­
specific; each is based on general theories of carcinogenesis rather 
65 
than on data for a specific chemical. It appears that modeling 
improvements will be brought about mainly through the incorporation 
of toxicological and biological information. "The contribution from 
statisticians and model-builders has reached an impasse, and more 
accurate extrapolations are not possible without additional 
information on the mechanisms of action of the toxic agents" (Brown, 
1987, p. 265). 
(3) Human Exposure Evaluation: This component describes the nature and 
size of the population exposed, and the magnitude and duration of 
the exposure. The exposure ex ami ned can be past, current, or 
anticipated exposure. To estimate the human dose of a contaminant, 
information regarding the dosage from various contact sites must be 
available. In the case of a water contaminant, potential contact 
routes include direct ingestion through drinking, inhalation, skin 
absorption from water (e.g., from bathing) and from contaminated 
soil, and from ingestion of contaminated soil. In order to evaluate 
dosage from a water contaminant, standard simplifying assumptions 
are often made (e.g., an adult is assumed to ingest two liters of 
water per day). The total dose an individual receives is equal to 
the sum of the doses from all possible routes. However, the concept 
of absorption can complicate this estimation; the dose that enters 
the body is not always equal to the dose that is absorbed. If the 
absorption rate of a substance is not known, the rate is often 
assumed to be 100 percent. 
( 4) Risk Characterization: Using information from the above components, 
risk characterization determines the likelihood that humans will 
i 
I \ 
66 
experience any of the various forms of toxicity associated with the 
substance. For noncarcinogenic effects, the margin of exposure 
(MOE) is used as a surrogate for risk; it is the experimental NOEL 
divided by the estimated daily human dose. The 1 arger the MOE, the 
smaller the risk. For carcinogenic effects, risk per unit of dose 
is estimated from the dose-response modeling. (If a variety of 
assumptions and models is used, a range of risks might be produced.) 
To estimate total risk, the actual human dose is multiplied by the 
risk per unit of dose. 
A health risk assessment requires an understanding of the hazards 
a particular substance presents. As mentioned above, epidemiologic 
studies can be used to estimate potential health risks. An epidemiologic 
study examines the incidence, distribution, and control of disease in a 
population. Many of the epidemiologic studies on the toxicity of 
agricultural chemicals focus on the most easily observed health effects 
and the most direct route of human exposure. For this reason, there is 
a relatively large number of epidemiologic studies on acute pesticide 
poisoning, often due to direct contact with the pesticide (e.g., exposure 
while hand-applying the chemical). 
Loevinsohn (1987) presents an epidemiologic study that examines the 
relationship between insecticide use and increased mortality. His study 
area is Central luzon in the Philippines, where organophosphate and 
organochlorine insecticides have been widely used by farmers. Loevinsohn 
compared mortality rates for the rural population (the exposed group) and 
the urban population (the control group) in two periods, one of low use 
of insecticides {1961-71) and one of high use (1972-84). In Central 
I 
I 
, I 
I , 
67 
Luzon, pesticides are usually applied by men using backpack applicators; 
protective clothing is generally not worn. 
The study results indicate a possible linkage between mortality and 
occupational exposure to insecticides, with mortality increasing only in 
the age and sex class occupationally exposed (men aged 15-54 years). The 
overall increase in non-traumatic mortality among rural men between 
periods of low and high insecticide use was 27.4 percent, from 2.15 to 
2.74 per 1000 (a 0.59 per 1000 increase). Also, causes of mortality 
likely to be confused with insecticide poisoning (i.e., stroke) also 
increased in rural men (aged 15-54) between low-use and high-use periods. 
If it could be established that insecticide use does increase the 
applicator's risk of mortality by 0.00059, a hedonic approach similar to 
that used by Thaler and Rosen (1976) could be used to determine the 
aggregate willingness to pay to save a statistical life. Unfortunately, 
it is often difficult to establish causality from epidemiologic studi'es. 
One reason (as this study illustrates} is the difficulty in implementing 
controls in epidemiologic studies. Loevinsohn (1987) is able to 
establish that there may be some association between high levels of 
insecticide application and increased mortality, but no conclusive 
associations can be made because of the lack of an adequate control 
group. 
On a similar note, Berteau and Spath (1986) point out, "A study 
conducted by the California Department of Health Services (Jackson et 
al., 1982) is the only known actual epidemiological study that has been 
conducted on peop 1 e exposed to water contamination with pesticides" 
{p. 429}. The study indicates a possible link between a soil fumigant 
68 
(DBCP) in drinking water and stomach cancer, but once again the lack of 
an adequate control group makes the determination of conclusive 
associations impossible. 
As indicated by the above studies, health risk assessment can be 
complicated by a number of factors. A tradeoff exists in the determina­
tion of human health hazards. Human health hazards can be extrapolated 
from laboratory animal studies; in this case, control groups can be 
established but extrapolating human hazards from animal hazards may not 
be appropriate. Estimating human health hazards from epidemiologic data 
eliminates the extrapolation problem, but introduces the possibility of 
inadequate control groups. Also, synergistic effects between substances 
may be important in estimating health hazards. These are just a few of 
the factors that may add to the imprecision of a risk assessment of 
contaminated ground water. 
There is still much to be discovered about the human health effects 
of groundwater contaminated by agri cul tura 1 chemica 1 s. However, as Gough 
(1989b} points out, agricultural chemical contamination of groundwater 
is a relatively insignificant cause of human cancers. He states, "Only 
a small percentage of human cancers is caused by the environment in the 
sense of air, water, soil and their contaminants" (personal communica­
tion). Current sunlight exposures, environmental tobacco smoke, and 
indoor radon account for between 70 and 96 percent (depending on 
which approach is used to calculate total cancer risk) of all cancers 
associated with environmental exposures. This indicates that groundwater 
pollution by agricultural chemicals does not present a major cancer risk 
to the public in general. It should be noted that some groups of people 
I 
I , 
I 
I ' 
69 
are more susceptible to the health risks presented by agricultural 
chemicals; for these people, the health effects of contaminated 
groundwater may be significant. 
70 
CHAPTER 7 
DATA BASES 
In order to perform any type of empirical analysis of groundwater 
contamination, some combination of data on environmental quality {a), 
management practices (m), and groundwater contamination (z) is needed. 
The suitability of a data base can be evaluated only after the type of 
analysis to be performed is determined. In the evaluation of potential 
groundwater pollution, data can be used for three general levels of 
analysis: 
(1) Descriptive Analysis: This type of analysis can qualitatively 
evaluate pollution potential based on environmental characteristics 
and management practices. The data used may be relatively aggre­
gated. For example, Nielsen and Lee (1987) use this descriptive 
approach; in their study they employ DRASTIC scores (scores that are 
an aggregate measure of hydrogeologic quality), and combine them 
with agricultural chemical usage to qualitatively determine ground­
water pollution potential. The end result of their study is the 
classification of counties according to the level of chemical 
application ("high" or "medium") and the level of hydrogeologic 
quality {"high" or "medium"). 
(2) Ex Post Analysis: An ex post analysis requires data on the inci­
dence and levels of groundwater contamination, environmental 
71 
characteristics, and management practices. For example, a study 
could be undertaken to determine the relationship between the amount 
of groundwater contamination that has occurred and the properties 
of the chemicals applied (e.g., solubilities and degradation rates). 
(3) Ex Ante Analysis: Several of the chemical fate and transport models 
were designed to be used for ex ante analysis. For example, PRZM 
was designed to be used in pesticide registration. Of the three 
types of analysis, ex ante usually has the largest data requirement. 
As discussed in Chapter 3, the user is usually required to input 
data on hydrogeologic characteristics, soil characteristics, 
management practices, and chemical properties. Additionally, an ex 
ante analysis may require calibration of some parameter values, 
increasing the data requirement. Although a simulation of changes 
in groundwater quality will require a relatively large amount of 
data, it may be less costly to conduct (in terms of time, money, and 
health risks) than a survey of groundwater contamination (as would 
be required for an ex post analysis). 
This section surveys some of the data bases currently available, 
with a focus on national data bases. 
Environmental Quality Data Bases 
DRASTIC Data Base 
DRASTIC is a system developed by the National Water Well Association 
(NWWA) to evaluate the potential for groundwater pollution of hydrogeo­
logic sites in the United States. This system was designed to fulfill 
several requirements. The system must: (1) function as a management 
I ! 
72 
tool, (2) be simple and easy to use, (3) utilize available information, 
and (4) be able to be used by individuals with diverse backgrounds and 
levels of expertise (Aller et al., 1986). As with the chemical transport 
models discussed, the DRASTIC system is intended to be used on a compara­
tive basis; the DRASTIC county scores can be used to estimate a county's 
vulnerability relative to the vulnerability of another county. 
The DRASTIC system consists of two main sections: (1) the designa­
tion of hydrogeologic settings, and (2) the application of a system for 
relative ranking of hydrogeologic parameters to evaluate the relative 
potential for groundwater pollution of the hydrogeologic setting. A 
hydrogeologic setting is a composite description of all the major 
geologic and hydrologic factors which affect and control groundwater 
movement into, through, and out of an area. It is defined as a mappable 
unit with common hydrogeologic characteristics and therefore similar 
potential for groundwater pollution (Aller et al., 1986). DRASTIC uses 
a classification system that divides the United States into 13 ground­
water regions: (1) Western Mountain Ranges, (2) Alluvial Basins, 
(3) Columbia Lava Plateau, (4) Colorado Plateau and Wyoming Basin, 
(5) High Plains, (6) Nonglaciated Central Region, (7) Glaciated Central 
Region, (8) Piedmont and Blue Ridge, (9) Northeast and Superior Uplands, 
(10) Atlantic and Gulf Coastal Plain, (11) Southeast Coastal Plain, 
(12) Hawaiian Islands, and (13) Alaska. These groundwater regions are 
then classified into smaller hydrogeologic settings. Approximately 86 
hydrogeologic settings have been identified within the groundwater 
regions of the United States. 
I 
' ! 
I I 
! 
I ) 
73 
The second portion of this evaluation system involves the relative 
ranking of hydrogeologic parameters called DRASTIC factors. These 
factors are considered to be the most important mappable factors deter­
mining groundwater pollution potential. These seven factors form the 
acronym DRASTIC: D, depth to water; R, (net) recharge; A, aquifer media; 
S, soil media; T, topography (slope); I, impact of the vadose zone; C, 
(hydraulic) conductivity of the aquifer. 
To estimate the relative groundwater pollution potential of a 
hydrogeologic setting, these seven factors are rated. Because these 
factors have unequal impacts on groundwater vulnerability to pollution, 
the factors are weighted. There are two indices of weights: the generic 
DRASTIC index and the Agricultural DRASTIC index. The Agricultural 
DRASTIC index was developed to examine the application of agricultural 
pesticides, whereas the generic DRASTIC index is used for all other 
contaminants, including fertilizers. The agricultural index assigns 
higher weights to topography and soil media. The weighted DRASTIC score 
for a hydrogeologic setting is calculated by multiplying each factor 
rating by its respective weighting and then summing these products. 
The Environmental Protection Agency (EPA) incorporated the DRASTIC 
rating system into its National Pesticide Survey (NPS). NPS will be 
described in more detail later in this section. Stage I of the NPS 
consists of the classification of all U.S. counties in terms of ground­
water vulnerability. Research Triangle Institute (RTI), under contract 
to EPA, completed this first stage. A modification of the DRASTIC system 
was used to classify all 3,144 counties of the United States into three 
74 
categories of groundwater vulnerability. Alexander et al. (1986) 
describe the procedures and results of this study. 
After obtaining the DRASTIC factor data, the numerical DRASTIC 
scores were computed, the most important scores being the weighted score 
and the VARSCORE. The VARSCORE represents the weight score and an index 
of variability; because the weighted score does not account for hetero­
geneities in the characteristics of the seven DRASTIC factors, an "index 
of variability" can be used to reflect these heterogeneities. In terms 
of the Agricultural DRASTIC score, two factors, aquifer media and the 
impact of the vadose zone, are thought to be the primary factors where 
variability may apply (Alexander et al., 1986). 
In this RTI application of the DRASTIC system, the classifications 
of vulnerability are based on a 10/60/30 percent distribution of the data 
base of VARSCORES for low, medium, and high vulnerabilities, respec­
tively. This distribution was selected to minimize the likelihood of 
counties being misclassified as having either high or low vulnerability. 
The authors of this application note that the DRASTIC system has 
several shortcomings. Because .the area of study (the county) is 
relatively large, the averaging effect may result in areas of high 
vulnerability within a county being overshadowed by areas of low vulner­
ability within the same county. Also, rating DRASTIC factors is 
generally subjective and differences in opinion may occur when rating 
these hydrogeologic factors over county-sized areas. 
As mentioned earlier, Nielsen and Lee (1987) use weighted DRASTIC 
scores calculated by RTI to conduct a descriptive analysis of groundwater 
pollution potential from agricultural chemicals. In that study, the two 
i 
! I 
75 
main data bases used to determine distributions of groundwater vulner­
ability and agricultural chemical usage are the DRASTIC index and the 
Resources For the Future (RFF) National Pesticide Use Inventory data 
base. 
Interactive Soils Information 
System (ISIS) 
ISIS consists of a group of programs designed by the U.S. Army 
Construction Engineering Research Laboratory to access and analyze USDA/ 
Soil Conservation Service soil data. The two data bases in this system 
are the Soil Interpretive Records (SOI-5) data base and the Map Unit 
Records (SOI-6) data base. 
SOI-5 contains data on over 16,000 different soil series nationwide. 
A soil series is a group of soil horizons that are similar in differ-
entiating characteristics such as texture, color, structure, consistence, 
reaction, content of carbonates or other sa 1 ts, content of organic 
matter, and mineralogical composition. Each SOI-5 record contains 
information including estimates of soil properties such as soil texture, 
erosion, flooding, pH, organic matter, permeability, depth of bedrock, 
frequency and duration of flooding, yield estimates of crops, suitability 
or limitations of soils for specified land uses, and soil features 
affecting specified land uses. Data for SOI-5 is obtained from SCS Soil 
Interpretation Records that are completed by soil scientists. The data 
. base is accessed by three programs: {1) the Soils Information Retrieval 
System (SIRS), {2) the Line Printer Soils Information Retrieval System 
{LPSIRS), and {3) the Multiple Parameter Series Search (MPSS) {Thompson 
et al., 1987). 
! I 
i 
! ) 
, I 
' I 
i ! 
76 
SOI-6 contains data on more than 175,000 soil mapping units nation­
wide. [According to Thompson et al. (1987), a taxonomic unit is a named 
kind of soil (taxon) that has specific properties with defined limits or 
ranges in characteristics; a soil map unit is an area of soil(s) delin­
eated on a soil map that contains one or more taxonomic units. A soil 
mapping unit is the aggregate of the map units of a kind of soil (s) 
identified by the same name (symbol) on a soil map or in a soil survey 
area.] The data on the soil mapping units includes mapping unit charac­
teristics, critical phase criteria, and survey acreage by county. These 
data are obtained from SCS Map Unit Records. The programs that access 
the SOI-6 data base are the Map Unit Use File System (MUUFS) and the 
Computer-Aided land Evaluation System (CALES). 
One shortcoming of ISIS is that neither of the two data bases, SOI-5 
and SOI-6, provide information concerning where within a geographical 
area a map unit or soil series occurs. However, SCS state and local 
offices can provide exact locational information (Thompson et al., 1987). 
Management Data Bases 
A variety of management practices can affect the amount of ground­
water contamination occurring. Two of the practices that can have the 
greatest impacts on contamination potential are chemical use and tillage 
practices; therefore, this section focuses on chemical use and tillage 
practice data bases. 
' I 
i. 
! I 
77 
Fertilizer Use Data Base 
The Economic Research Service (ERS) Division of the USDA has 
developed a fertilizer use data base from survey data for the primary 
producing states of corn, cotton, soybeans, and wheat. The data is for 
the state level and the fertilizers considered are nitrogen (N), 
phosphate (P205), and potash (~0). The data information includes percent 
of acres fertilized of the acres surveyed in each state, the average 
application rate for those acres receiving fertilizer, and the percent 
of proportion fertilized at or before seeding, after seeding, and at both 
times. Detailed data on fertilizer use by other crops are not collected 
in any consistent manner. 
Pesticide Use Data Base 
The Economic Research Service (ERS) of the USDA has developed a 
pesticide use data base from survey responses. The data are for the 
regional level. (USDA has divided the U.S. into 10 production regions.) 
The crops inc 1 uded in the data base are: wheat, corn, sorghum, soybeans, 
cotton, peanuts, rice, tobacco, barley, oats, pasture and rangeland, 
alfalfa, and "other hay," and the data are expressed in pounds of active 
ingredients applied. This survey has been conducted several times in the 
past two decades, but in 1989 the survey format was changed. The new 
survey consists of state-level data for wheat, cotton, soybeans, and 
corn. Instead of data on application rates, data on numbers and percent­
ages of acres treated are included. According to Herman Delvo of the 
ERS (personal communication, February 1989), application rates do not 
I 
I I 
78 
need to be included because these rates do not vary significantly over 
time. 
National Pesticide Use Inventory 
This data base, developed by Resources For the Future (RFF), is a 
more complete pesticide data base than that of the ERS. The RFF data 
base includes estimates of the use of 15 soluble pesticides by state for 
10 major field crops. These 15 pesticides represent approximately 53 
percent of the total national usage of the 55 active ingredients 
monitored in EPA's National Pesticide Survey. The 15 pesticides 
represented are: al achl or, metol achl or, atrazi ne, cyanazine, 2,4-0, 
trifluralin, carbaryl, disulfoton, methyl parathion, carbofuran, 
diazinon, dinoseb, acifluorfen, chlorothalonil, and fluometuron 
(Gianessi, 1987). The 10 major field crops (representing approximately 
90 percent of the national volume of use of the 15 pesticides) are: 
corn, soybeans, wheat, barley, oats, sorghum, cotton, rice, sugarcane, 
and hay. Estimates of use of each pesticide are made on a state-by-state 
and crop-by-crop basis, with an assessment being made for each state that 
has a significant acreage of any of the 10 field crops (Gianessi, 1988). 
Although county-level estimates of pesticide use are contained in the 
data base, these estimates are extrapolated from state-level estimates; 
the assumption is that there is uniform use within a state for the same 
crop/pesticide combination. 
As discussed in an earlier section, the amount of active ingredient 
applied to a field may not be a good measure of the amount of potentially 
harmful chemicals applied to that field. "Inert" ingredients may present 
79 
more potential health risks than active ingredients. Therefore, when 
using a data base that measures chemical applications in terms of amount 
of active ingredient applied, it should be remembered that the amount of 
potentially hazardous chemical applied may be underestimated. 
Ii 11 age Data 
The Conservation Technology Information Center (CTIC) has developed 
the CTIC National Survey of Conservation Tillage Practices. Data have 
been collected each year from 1983 through the present. The survey 
elicits information on the use of five conservation tillage practices 
(no-till, ridge-till, strip-till, mulch-till, and reduced-till) for 12 
major crops for every county in all 50 states. Crops included are: corn 
(full season and double crop), small grains (spring seeded and fall 
seeded), soybeans (full season and double crop), cotton, grain sorghum 
(full season and double crop), newly established forage crops, newly 
established and renovated permanent pasture, and "other crops" (a catch­
all category for all other planted acres in a county). The information 
in this data base is aggregated to the county level. 
Contamination Data Bases 
Groundwater Contamination Data Base 
Algozin et al. (1988) have developed the Groundwater Contamination 
(GWC) data base to compare the groundwater contamination potential of 
various agricultural practices under various environmental conditions. 
This national data base includes both environmental quality information 
and management practices information. The data base represents the 
I 
I 
: I 
I 
80 
compilation of six files: the 1982 Natural Resource Inventory (NRI) 
file (land use information); the Soils-5 Interpretation file (soil and 
water table characteristics; thr.ee 1985 Resource Conservation Act (RCA) 
budget files (production practice and economic data); and the DRASTIC 
index file. 
! 
The data base can be separated into two macro files which are joined 
via several linkage variables. The macro file that describes the 
environmental characteristics of an area is the foundation of the data 
base and consists of the NRI file, the SOILS-5 file, and the DRASTIC 
file. The 1982 NRI consists of approximately 700,000 records of specific 
sample points where the data is collected; each record includes informa­
tion on the physical characteristics of the site, erosion potential, and 
vulnerability to groundwater contamination. In this data base NRI data 
are used for some point-specific data, such as soil characteristics, and 
DRASTIC scores (county averages) are used for other characteristics such 
as rainfall. The SOILS-5 data base has been described earlier. 
The second macro file provides the economic production information 
relating to each sample point using data from the three RCA budget files. 
This macro file is called the RCA Midset and was developed by the Center 
for Agricultural and Rural Development (CARD) at Iowa State University. 
The information contained in the RCA Midset includes: crop rotation, 
tillage, conservation practices, and input uses and yields for all crops 
included in the rotation. 
Algozin et al. (1988) use this GWC data base to link up areas of 
groundwater vulnerability with areas where agricultural practices 
conducive to groundwater contamination occur. Because their analysis 
! i 
i 
81 
incorporates site-specific information (from the NRI file) when computing 
DRASTIC scores, these scores are a significant improvement upon the 
county DRASTIC scores used by Nielsen and Lee (1987) that use only 
county-averaged data. 
Pesticides jn Groundwater Data Base 
The Pesticides in Groundwater Data Base, developed by the EPA Office 
of Pesticide Programs, is a compilation of individual groundwater 
monitoring studies for pesticides. EPA intends to use the data base to 
identify pesticides that are contaminating groundwater, to identify areas 
that are vulnerable to groundwater pollution, to develop a better under­
standing of the chemical migration pathways and environmental processes 
associated with groundwater contamination, and to identify gaps in 
current data. 
The data contained in this data base are from numerous studies 
conducted by federal and state agencies, chemical manufacturers, and 
universities. Approximately 130 reports have been entered into the data 
base. In addition to collecting and entering these studies, EPA has been 
conducting a program to assess the validity of the information gathered. 
To date, only 50 percent of the reports have been confirmed. Because the 
studies in the data base are from numerous sources, there is no consis­
tency in methodology. Also, these studies have a variety of focuses; 
some studies concentrate on more vulnerable (i.e., more shallow) ground­
water and other studies include wells that are not necessarily used for 
drinking water. (Consequently, this data base is not appropriate for 
estimating human exposure.) Two other shortcomings of the data base are 
I 
I I 
! 
I 
I 
I I 
I j 
I I 
I 
82 
that many agricultural areas of the U.S. have not been sampled and 
therefore are not represented in the data base, and that establishing the 
source of pesticide contamination is difficult. According to Williams 
et al. (1988}, determining the source was the most difficult parameter 
to confirm in the data base. Consequently, Williams et al. stress that 
the data base is to be used as a screening device and that "derived 
parameters contained in the report and in the appendices should be 
treated as indicators rather than absolute values" (p. 4-2}. 
The data base is separated into three data bases, with the Pesticide 
Monitoring Data Base containing the bulk of the pesticide information 
including: the pesticides tested for in the study, the suspected origin 
of the pesticide, number of wells tested/number of positive wells found, 
number of samples tested/number of positive samples found, lowest 
concentration found, highest concentration found, average concentration 
of positives, and a data confirmation rating (that indicates the quality 
of the data}. 
WATSTORE 
WATSTORE, a Water Data Storage and Retrieval System, was established 
by the U.S. Geological Survey in 1971 to modernize the USGS water-data 
processing and management. Currently, files are maintained for the 
storage of: (1} surface-water, quality-of-water, and groundwater data 
measured on a daily or a continuous basis; (2} annual peak values for 
streamflow situations; (3} chemical analyses for surface and groundwater 
sites; (4) water-data parameters measured more frequently than daily; 
(5} geologic and inventory data for groundwater sites; and (6) summary 
i 
, I 
I I 
I I 
. I 
83 
data on water use (Kilpatrick, 1981). Although WATSTORE contains six 
major files, the two files of primary interest for groundwater contamina­
tion are the Ground-Water Site-Inventory (GWSI) file and the water 
quality file. The GWSI file contains information on wells and springs 
at sites from all 50 states. As of February 1981, the file contained 
data for nearly 700,000 sites. The GWSI file is composed of hydrologic, 
geologic, and well inventory data for groundwater sites including site 
location and identification, geohydrologic characteristics, well 
construction history, and one-time field measurements such as water 
temperature. The data base is a combination of county-averaged data 
(e.g., climate), and point-specific data (e.g., soil properties); 
however, the data are statistically valid at the Major Land Resource Area 
(MLRA). There are approximately 169 MLRAs in the U.S. An MLRA is 
characterized by relatively homogeneous soils, crops, and production 
practices. Although the bulk of GWSI data is collected by USGS personnel 
in cooperation with federal, state, and local government agencies, the 
data may be biased in that those wells suspected of contamination are 
sampled. 
The water-quality file contains data on the analyses of water 
composition. The file contains a relatively large amount of data on 
nitrate analyses of water, but relatively little data on pesticide 
analyses. This is because nitrate levels are somewhat easily and cheaply 
determined and so nitrate analyses are routinely performed. Pesticide 
analyses, however, are more difficult and expensive to conduct and 
consequently are often performed only as part of a special project. 
Although some areas of the country have assembled more pesticide data 
I ( 
I I 
! 1 
I I 
I , 
I 
: l 
: I 
84 
than others, on the whole, nitrate information is more extensive than 
pesticide information. WATSTORE contains data on nitrate levels in 
samples collected from 87,000 wells throughout the U.S. during the past 
25 years. 
National Pesticide Survey 
EPA has developed a National Pesticide Survey (NPS) to obtain a 
better understanding and characterization of agricultural chemicals in 
drinking water wells. This survey focuses on the quality of drinking 
water in wells and will elicit data on the relationship between contam­
inated well water and groundwater. The two major objectives of the NPS 
are: 
(1) to characterize nationally the distribution of pesticide residues 
in community system and rural household water wells, and 
(2) to assess the associations between the agricultural use of 
pesticides, the distribution of agricultural pesticide residues in 
well water, and hydrogeologic factors that influence groundwater 
contamination (Alexander et al., 1986). 
The NPS is "the first nationwide survey of pesticide contamination 
in domestic and community water wells in the United States" (USEPA, 1987, 
p. 1). The survey involves sampling approximately 1,500 drinking water 
wells, and the survey results are expected to be statistically repre­
sentative of over 13 million domestic wells and some 51,000 community 
water systems. Fifty-five active ingredients will be monitored in this 
surv~y. Because a goal of the survey is to determine the relationships 
among agricultural pesticide use, environmental characteristics, and well 
I 
l 
! ! 
85 
water contamination, this survey involves more than simply testing well 
water. The survey includes interviews with householders and community 
water system operators, and data collection in the area surrounding each 
well. The information elicited will include the uses and characteristics 
of the well water, well construction data, and information on factors 
that could explain the source or route of contamination (e.g., chemical 
spills, abandoned wells). 
Prior to the survey, EPA developed health advisories for the 61 
pesticides with the highest leaching potential. These advisories were 
designed to be used by well owners, operators, and the general public to 
determine whether the contamination levels found in well water warrant 
further action. EPA also prepared non-technical summaries of the health 
advisories to explain the health risks presented by exposure to these 
pesticides. (These health advisories and summaries could be used in 
valuing the health risks of pesticides.) 
. I 
1 I 
I , 
I 
I 
86 
CHAPTER B 
CASE STUDY 
A case study is presented in this chapter in order to evaluate the 
feasibility of linking together the chemical fate and transport, econ­
omic, and human health models. Sugar beet production in the Yellowstone 
Valley of Montana was selected for this analysis because this producti~n 
process includes irrigation and a relatively high use of agricultural 
chemicals, two factors that may contribute to groundwater pollution. 
To begin this case study, knowledge of the production practices is 
necessary. Baquet and Johnson (1988) present a study on the costs and 
returns of irrigated crops in the Yellowstone Valley. The irrigated 
crops are sugar beets, malt barley, alfalfa, and corn silage; these four 
crops are usually used in some type of rotation. The counties included 
in this study are Treasure, Big Horn, Yellowstone, Carbon, Stillwater, 
and Rosebud. Baquet and Johnson (1988) conducted surveys of 15 randomly 
sampled producers in this six-county area who produce sugar beets, and 
determined production practices, costs, and returns. Survey results are 
grouped into three farm sizes: small (less than 100 acres of sugar beet 
allotment), medium (between 100 and 250 acres of sugar beet allotment}, 
and large (more than 250 acres of sugar beet allotment). Baquet and 
.Johnson formed a composite estimate of the production practices, costs, 
and returns for each of the three groups. The composite results of the 
;' 
, I. 
' \ 
87 
surveys of the medium-sized farms are presented in Table 1. The authors 
point out that the composite estimates "should not be interpreted as 
representative for any particular farm in the Yellowstone Valley, but 
rather a composite estimate of the producers surveyed" (Baquet and 
Johnson, 1988, p. 1). 
An understanding of the general production practices involved in 
sugar beet production is necessary to get an idea of some of the types 
of decisions facing the producer; also, production practices (e.g., 
chemical use and types and timing of tillage) are inputs in management­
oriented chemical fate and transport models. 
Seedbed preparation for sugar beet production starts in the late 
summer or in the fall with the plowing of fields. Fields must be leveled 
before planting if surface irrigation is to be used. In the Yellowstone 
Valley, planting takes place between late April and late May. Baquet and 
Johnson's composite estimates indicate that fields are cultivated about 
three or four times per season, and are irrigated about four or five 
times per season. Plants must be thinned to 10 to 12 inches apart in a 
row. Although thinning can be done both mechanically and by hand, most 
of the producers surveyed used hand-thinning methods. Harvesting sugar 
beets consists of three general steps: (1) topping the plant, (2) lift ­
ing the root, and (3) loading them. 
Because sugar beets are a relatively profitable cash crop, they are 
usually given the priority position in a crop rotation; that is, rotation 
away from sugar beets is done mainly to control sugar beet diseases and 
nematodes (Chapman and Carter, 1976). It is because of this necessary 
rotation that it would be difficult to isolate the potential chemical 
Table 1. Production operations and materials for sugar beets. 
EQUIPMENT TRACTOR OR LABOR AMOUNT PRICE 
OTHER POWER TIMES MATERIALS PER PER 
OPERATION TYPE SIZE UNIT OVER HIRED FAMILY USED ACRE UNIT UNIT 
Plow 4-16" 140 hp 1 
Roller Harrow 15' 140 hp 2 
Plane 12' 105 hp 2 
Fertilize Custom App. 1 140 II N 
90 II P 
30 II K 
Triple K 17!' 105 hp 1-2 
Plant 6 Row 75 hp 1 Seed 2 11/A. $15/cwt. 
-
Cultivate 6 Row 105 hp 4 Roneet 12 II 11/ A $1.10/1/ 
Spray 6 Row 75 hp 1 Betamix 1 qt/A. $6.6/gal. 
Poast 1 qt/A. $8/gal. 
Thin 1 X $40/A 
Irrigate 4 X 
Defoliate 6 Row 105 hp 1 X 
Dig 3 Row 140 hp 1 X 
Truck 1 X 
Truck 2 X 
Truck 3 X 
(SOURCE: Baquet and Johnson, 1988, p. 6.) 
89 
loading attributable directly to sugar beet production. For a long-term 
simulation of potential groundwater loading, the effects of all of the 
crops in the rotation must be considered. 
Sugar beet production is relatively chemical intensive, requiring 
both high levels of fertilizer and pesticide use. Moderate sugar beet 
yields such as those obtained in the Yellowstone Valley (about 20 tons 
per acre) , can remove 1 arge amounts of nitrogen, phosphorus, and 
potassium from the soil. Fertilizer can be applied before seeding, 
during seeding, or as topdressing during the growing season. Timing of 
applications is important for both fertilization and irrigation. 
Fertilizer applications should be timed so that soil nitrogen is depleted 
10 to 14 days prior to harvest (to encourage sugar production in the 
root). Table 1 indicates that fertilizer use per acre consists of: 150 
pounds nitrogen, 90 pounds phosphorus, and 30 pounds potassium. Timing 
of irrigation is also crucial; the final irrigation of the season should 
meet the plant's water requirements, but should also ensure that the soil 
is dry enough for harvesting, yet not so dry that it impedes digging the 
beets. 
The most common weeds in sugar beet fields are pigweeds, lambs­
quarters, barnyard-grasses, mustards, kochias, and wild oats (Chapman and 
Carter, 1976). Weed control is accomplished with the use of herbicides, 
mechanical cultivation, and hoeing. Mechanical cultivation can be 
substituted for chemical weed control up to a point; however, once the 
sugar beet plants reach a certain size, mechanical cultivation between 
the rows is no longer possible. The three pesticides included in Table 
1 (Ro-Neet, Betamix, and Poast) are all herbicides. It should be noted 
\ ! 
90 
that nematodes and rip-maggots can cause significant sugar beet damage 
in the Yellowstone Valley, although nematicides and insecticides are not 
mentioned in Table 1. Hallenbeck and Cummings-Burns {1985) provide a 
good description of the various classes of pesticides, their acute­
exposure human health effects (observed in humans), their chronic human 
health effects, and their expected human health effects (which may be 
extrapolated from animal studies). Much of the following information on 
the chemicals Ro-Neet and Betamix is taken from Hallenbeck and Cummings­
Burns {1985). Cycloate, a synonym for Ro-Neet, is in the thiocarbamate 
class of chemical compounds. Effects observed in humans following 
exposure to cycloate include: coughing, dermal irritation, eye irrita­
tion, and respiratory mucous membrane irritation. Suspected effects may 
include paralysis, dermal edema, and dermal hyperemia. Cycloate may also 
present significant chronic effects, although EPA has not yet made its 
studies available to the public. The active ingredients in Betamix are 
phenmedipham and desmedipham, which are in the carbamate class of 
pesticides. The potential chronic exposure effects of carbamates include 
anorexia, cholinesterase depression, muscle weakness, and renal damage. 
The chemical fate and transport model used in this application is 
PRZM. A 1 though a computer s i mul at ion was not run, a description of 
potential sources of key model parameters is included. It should be 
noted that the deve 1 opment of the input data set is no sma 11 feat. 
Carsel et al. {1984) divide a model simulation into five tasks and 
indicate the relative effort required for their completion. Of the five 
tasks, the development and input of the data set is estimated to require 
40 percent of the effort. The type of data necessary for the analysis 
91 
will depend on the level of aggregation. Although several of the data 
bases discussed in Chapter 7 may provide the necessary data for the use 
of simple screening methods, such as the calculation of a DRASTIC score, 
acquiring the necessary data for a more site-specific analysis using a 
numerical model is more complex. The PRZM User's Manual (Carsel et al., 
1984) does provide many estimates of model parameters as well as tech­
niques to calculate parameters, but the level of disaggregation may not 
be sufficient for a site-specific analysis. What follows is a discussion 
of some of the key model parameters and some of the potential data 
sources. 
The Montana Agricultural Potentials System (MAPS) is a data base of 
Montana hydrogeologic characteristics that contains information that can 
be used for this case study. This data base sections the entire state 
into eight-square-mile blocks; data on several dozen climatic and 
hydrogeologic characteristics are gathered for each block of land. The 
data include characteristics on the growing season, land use, potential 
evaporation, potential evapotranspiration, precipitation, regions, and 
soils. Precipitation characteristics include monthly averages and 24-
hour precipitation. Soil characteristics include general soils of 
Montana (136 map units), soil water holding capacity, soil temperature, 
soil depth classes, soil pH, and a rainfall intensity factor. 
Pesticide Parameters 
As discussed in Chapter 3, the fate of a pesticide applied to soil 
depends largely on two of the pesticide's properties: persistence and 
92 
so 1 ubi 1 i ty. These properties are taken into account in PRZM by the 
pesticide transformation rate and partition coefficient. 
(1) Pesticide Transformation Rate: Transformation rates are often 
obtained from chemical-specific studies. It should be noted that 
compared to surface transformation rates, relatively 1 ittle is known 
about chemical transformation rates below the soil surface. It is 
not uncommon to estimate a subsurface rate as a percentage of the 
surface rate. Based on 1 imited laboratory data, Donigian and Carsel 
(1987) estimated transformation rates in the horizon below the plant 
root zone and in groundwater to be 50 percent of the rate in the 
surface soil. 
(2) Chemical Partition Coefficient: As discussed in Chapter 3, a 
chemical partition coefficient (Kp) is a major determinant of the 
leaching ability of a chemical. The KP for each soil horizon can 
be expressed as a function of the chemical's adsorption coefficient 
for organic carbon, K00 , and the organic carbon fraction or organic 
matter in each soil horizon. The Soil Conservation Service collects 
some data on organic carbon content of soil horizons, and the Soils 
Information Retrieval System (SIRS) contains data on soil organic 
matter. According to Leonard et al. (1987), data on the K00 are not 
available for all pesticides, but significant work has been done to 
provide methods of estimation. These methods are based on a 
relationship between K00 and K0 w (octanol: water distribution 
coefficient), and between K00 and water solubility (see Kanaga and 
Goring, 1980, and Rao and Davidson, 1980). 
i 
j ! 
i 
93 
Soil Parameters 
Key soil parameters of PRZM include field capacity, wilting point, 
and the percent of sand or clay in the soil. 
(1) Field Capacjty and Wilting Point for All Horizons Modeled: Although 
the wilting point will depend on specific soil characteristics 
(e.g., organic matter content), most soils will have a similar 
wilting po1nt for all common plants. Carsel et al. (1988a) obtained 
data on wilting point, field capacity, bulk density, and organic 
matter from the Soil Conservation Service. Bulk density is the mass 
per unit of undisturbed soil, dried to constant weight at 1os•c; 
organic matter consists of the remains and decomposition products 
of both plants and animals (Fitzpatrick, 1983). MAPS contains soil 
field capacity data; also, Carse] et al. (1984) present three 
· j possible field capacity estimation techniques. 
(2) Percent Sand or Clay: SIRS contains these data. 
Hydrologic and Crop Characteristics 
The hydrologic and crop characteristics determine the amount of 
water avail ab 1 e for i nfi 1 trat ion and runoff. Key parameters inc 1 ude 
precipitation, erosion, SCS curve number, evapotranspiration, depth to 
groundwater, and recharge. 
(1) Precipitation: MAPS includes data on monthly averages~, peak 24-
hour precipitation rates, and a rainfall intensity factor, and also 
expresses monthly precipitation as a percentage of annual precipi­
tation. 
I 
' I 
i 
i ! . 
94 
(2) Erosion: A modified USLE is used to estimate soil erosion. This 
equation requires information including peak runoff rate and the 
soil erodibility factor. MAPS contains both peak 24-hour precipita­
tion rates and a rainfall intensity factor. The soil erodibility 
factor is a soil-specific parameter and specific values can be 
obtained from the SIRS data base or from the local Soil Conservation 
Service office. 
(3) SCS Curve Number: The SCS curve number technique is used to 
estimate runoff. This technique accounts for the interaction of the 
hydrologic soil group, land use, and treatment (cover crop). Carsel 
et al. (1984) provide a method for calculating the SCS curve numbers 
and include some of the data needed for this calculation; however, 
the data inc 1 uded are 1 i mi ted to the major agri cul tura 1 crops in the 
u.s. 
(4) Evapotranspiration: MAPS provides data on both potential evapora­
tion and potential evapotranspiration. 
(5) Depth to Groundwater: Although the ISIS does contain some data on 
the depth to the water table, many PRZM applications rely on region­
specific studies by the U.S. Geological Survey. 
(6) Recharge: Recharge is the amount of water per unit of land that 
penetrates the ground surface and reaches the water table (Nielsen 
and Lee, 1987). Sources of net recharge data include the U.S. 
Geological Survey, State Geological Surveys, Soil Conservation 
Service, and State Department of Natural/Water Resources {Alexander 
et al., 1986). 
' . 
' ! 
95 
One important note regarding data sources is that many of the 
chemical fate and transport model applications examined rely on studies 
of specific chemicals or specific sites for data, instead of on large, 
national data bases. 
PRZM was designed to simulate pesticide loading to a small area 
where soil and climatic characteristics are relatively homogeneous. 
Because this case study examines the pesticide loading potential of this 
six-county sugar beet producing area of the Yellowstone Valley, a Monte­
Carlo simulation technique can be used to account for the heterogeneity 
of soil and climatic characteristics in this area. In their study of 
aldicarb application to Ohio corn, Carsel et al. (1988a) incorporate 
several randomized parameters, including those for weather year, pesti­
cide degradation rate, and values of field capacity, wilting point, and 
organic matter for each soil horizon. (The distributions of field 
capacity, wilting point, and organic matter were identified using SCS 
soil series data.) Once the input data set is complete, a series of 
Monte-Carlo simulations can be run and the predicted amounts of chemical 
movement in the simulations can be used to construct a ·cumulative 
probability distribution that indicates the number of simulations that 
resulted in leaching below the root zone. The results of this series of 
Monte-Carlo simulations can be used to compare simulation results under 
different management scenarios. As noted in Chapter 3, many management 
models are designed to be used to compare the potential loadings of 
alternative management practices. More importantly, for policy analysis 
we are interested in evaluating the effects of a change in policy because 
I 
I 'r 
! 
I I 
96 
a change in policy can affect production decisions, possibly changing the 
distribution of potential loadings. 
As an example of a policy that affects management decisions at the 
intensive margin, assume that a policy is implemented that restricts the 
use of a herbicide, say Ro-Neet. This type of change could affect both 
chemical use and tillage practices. As discussed in Chapter 4, the types 
of weed control methods that will be substituted for Ro-Neet will be 
determined by the producer's economic production function. To determine 
the amount of potential loading resulting from the change in policy, the 
PRZM model parameters affected by the policy change must be identified 
and the parameters must be adjusted to reflect the changes in production 
practices. Assume that the restriction on the use of Ro-Neet results in 
producers using more mechanical cultivation. The change in chemical use 
can be incorporated into the PRZM input data set by adjusting the rate 
{and possibly timing) of chemical application. The change in tillage 
practice may involve a change in the amount of tillage or the use of a 
different type of tillage. 
Once the input data set has been adjusted to reflect the changes in 
management practices, a second series of Monte-Carlo simulations can be 
run to determine the effect of the change in policy on the distribution 
of potential loading. The simulation results can be used to construct 
a cumulative probability distribution similar to that of the pre-policy 
series of simulations. From the cumulative probability distribution, it 
is possible to establish the percentage of simulations for which the 
model predicted residue movement less than {or greater than) a specified 
amount and below a specified depth. {For this case study the relevant 
97 
depth is the bottom of the root zone.) It is also possible to determine 
the percentage of simulations that predicted residue movement (exceeding 
a specified amount of chemical) below the root zone. 
A shortcoming of PRZM is that it simulates pesticide movement only 
in the plant root zone. To determine the amount of pesticide that enters 
groundwater, it may be necessary to 1 ink up PRZM with a model that 
simulates pesticide movement and transformation down through the unsat­
urated zone. To determine the groundwater concentration of a chemical, 
a groundwater model must be linked to the saturated and unsaturated zone 
models. linking these models can be a difficult procedure that is 
complicated by a lack of data on some of the physical, biological, and 
chemical processes occurring below the root zone. Carsel et al. (1988b) 
provide a relatively simple method of converting a potential pesticide 
loading to a groundwater concentration, although more theoretical 
groundwater models like the Analytical Transient One-, Two, and Three­
Dimensional (AT1230) model (Yeh, 1981), are more complicated. If the 
unsaturated zone is sufficiently shallow, the potential loading simulated 
can be used directly in a groundwater model, and the potential ground­
water concentration of the chemical can be calculated. 
Valuation of the Social Costs 
In this case study, a contingent valuation (CV) method is proposed 
to elicit information that can be used to value human health risks. A 
CV survey presents survey respondents with a hypothetical scenario and 
asks them to va 1 ue a change in the 1 eve 1 of the provision of some 
public good (or to avoid a change in the level of a public good). The 
i i 
I' 
i 
98 
hypothetical scenario allows respondents to consider both use and nonuse 
(existence) benefits in their valuations and, because the scenario 
presented is hypothet i ca 1 , a CV survey can be used for an ex ante 
valuation of a change in the provision of a public good. 
A CV elicitation can obtain the appropriate compensated measure 
(holding the individual's level of utility constant) associated with a 
change in the level of the provision of the public good. This allows the 
individual to make his own tradeoffs between money and the public good. 
The bids elicited from a CV survey can be expressed as the difference 
between two expenditure functions. The compensating surplus (CS) can be 
expressed as: 
[8.1] 
where Po is the vector of prices, y0 is the initial level of provision of 
a public good, U0 is the individual's initial level of utility, and y1 is 
the new level of provision of the public good (Mitchell and Carson, 
1989). 
The difference between the two expenditure functions (CS), is the 
amount of money the individual would be willing to pay to change from Yo 
to y1 , while holding his utility level constant at the initial level, U0 • 
An advantage of the CV method over other methods of valuing public goods 
(e.g., a hedonic price method) is that the CV method does not require the 
estimation of an implicit demand curve. However, a disadvantage of the 
CV method is that it is subject to a variety of biases that may result 
because of: (1) incentives for individuals to misrepresent responses, 
(2) values being implied in the survey, and (3) scenario misspecification 
I 
99 
by the researcher or mi spercept ion of the scenario by the respondent 
{Mitchell and Carson, 1989). 
In this analysis of agricultural chemical policy, we want to 
evaluate the changes in the social costs and benefits resulting from a 
change in agricultural chemical use and possibly a change in tillage 
practices. It is strongly suggested that this change in the social costs 
and benefits (a change in the level of human health risk) can be valued 
with a contingent valuation method (Mitchell and Carson, 1989). However, 
in this case study, the chemical fate and transport model used {PRZM) 
simulates chemical movement only down through the root zone and not down 
to groundwater. Clearly it would not be meaningful to ask survey 
participants to value a change in the concentration of the chemical at 
the bottom of the root zone. Therefore, a gap exists between the lowest 
sector modeled by PRZM (the root zone) and the horizon that would be 
relevant to survey respondents (groundwater). This problem could be 
resolved by linking PRZM with a model that simulates chemical fate and 
transport from the bottom of the root zone to the groundwater zone. As 
mentioned in Chapter 3, work is being performed to develop a vadose zone 
model that would provide this linkage. For the purposes of this case 
study, we will assume that the unsaturated zone is sufficiently shallow 
that PRZM can simulate potential groundwater loadings. These loadings 
can be input into a groundwater model to estimate the chemical concentra­
tions, and these concentrations can be used in a CV survey. Although the 
assumption that the depth to groundwater is sufficiently shallow may not 
be realistic for the Yellowstone Valley, it may be a realistic assumption 
for other areas in the U.S. (e.g., parts of Florida). 
I i I 
I ! 
100 
To use the CV method, a survey tailored to the situation being 
examined must be developed. As discussed in Chapter 5, each CV is 
particular to a specific situation, and so extrapolating CV results from 
one situation to another is not appropriate (see Smith and Desvouges, 
1987, and Weinstein and Quinn, 1983). Furthermore, in order to elicit 
meaningful responses, the scenario presented to the survey respondents 
must be meaningful. According to Mitchell and Carson (1989), 
•.. the principal challenge facing the designer of a 
CV study is to make the scenario sufficiently under­
standable, plausible, and meaningful to respondents so 
that they can and will give valid and reliable values 
despite their lack of experience with one or more of 
the scenario's dimensions. (p. 120) 
Nielsen and Lee (1987) state that the primary potential effects of 
agricultural chemicals in groundwater are human health risks. To illus­
trate the necessary components of the development and implementation of 
a CV survey, an example of a survey that elicits willingness to pay (WTP) 
values for reductions in groundwater contamination (of Ro-Neet) is 
presented. The survey will examine a reduction in the risk of dying of 
cancer (25 years in the future) as a result of exposure to this chemical 
via groundwater. 
Chapter 5 addressed some of the psychological motives that can 
influence a contingent valuation of a change in a health risk. Conduct­
ing a CV on a reduction of the risks of dying of cancer presents some 
additional risks. Whereas many health risk valuations present the 
potentia 1 risks in terms of annua 1 risks, the risks associ a ted with 
cancer are not readily classified as annual risks. Cancer usually 
manifests itself many years after exposure to the cancer-causing agent 
101 
and so the risks of dying of cancer should be presented to the survey 
respondents as risks to be incurred in the future. Also, the level of 
exposure necessary to cause cancer can be very low and, according to 
Mitchell and Carson (1986), "Research has shown that people have diffi ­
culty understanding abstract risk levels and that they tend to exaggerate 
low-level risks" (p. 7}. 
The goal of this CV survey is to elicit WTP bids for a reduction in 
the 1 eve 1 of groundwater contamination. The change in the 1 eve 1 of 
groundwater contamination may have little meaning in itself; it is the 
associated change in the health risks that is meaningful to respondents. 
Therefore, the risks associated with the various levels of chemical 
contamination to be valued by the respondents must also be presented. 
Chapter 6 discussed types of data that can be used to determine the risks 
presented by various 1 eve 1 s of chemica 1 contamination. (Data on the 
toxicologic effects of a specific pesticide may be available from the 
Health Effects Division, Office of Pesticide Programs, Environmental 
Protection Agency.} It was also mentioned in Chapter 6 that there are 
several routes of exposure to water-borne chemicals. Data on the poten­
tial health effects of agricultural chemicals are incomplete, making it 
difficult to specify a dose-response relationship for low-level exposure 
to a chemical. However, as Nielsen and Lee (1987) point out, although 
the risks associated with low-level exposure may be uncertain, the 
public's perception is that they are significant. An individual's 
willingness to pay (WTP) is dependent on perception of risk, not neces­
sarily actual risk. 
I 
i l 
I I 
i I 
' l 
I 
102 
Communicating the idea of low-level risks may require that the 
potential risks of the contaminated groundwater be compared with more 
familiar low-level risk (i.e., the risk of being struck by lightning). 
In their CV, Mitchell and Carson (1986) use two risk ladders to familiar­
; ze respondents with the 1 ow-1 eve 1 risks i nvo 1 ved. The first risk 1 adder 
illustrates a full range of risks including basic risks (e.g., the risk 
of dying associated with different age groups), and specific "special" 
risks (e.g., risks associated with some professions). A second risk 
ladder that provides more detailed information on the lower-level risks 
is also presented. In addition to presenting a variety of relatively 
familiar low-level risks, this ladder also compares the number of cigar­
ettes an individual would have to smoke in a lifetime to run the same 
risk of dying of heart disease or cancer as the individual would have of 
dying as a result of one of the low-level risks. Illustrations of these 
risk ladders are presented in Figures 8 And 9. 
A referendum situation can be used to elicit survey participants' 
WTP for a reduction of the chemical concentration in groundwater. A 
referendum has the advantage that people may be familiar with the refer­
endum method because decisions regarding the provision of public goods 
are often made this way. An open-ended referendum format can skirt the 
problem of starting point bias, but it can also produce a large number 
of nonresponses or protest bids. "Starting point bias occurs when the 
respondent's WTP amount is influenced by a va 1 ue introduced by the 
scenario" (Mitchell and Carson, 1989, p. 240). However, an open-ended 
format can work well if the concept of the CV scenario is familiar to the 
respondents. For example, a CV survey could ask respondents how much 
103 
ANNUAL RISKS OF DYING 
BASIC RISKS SPECIAL RISKS 
1000 {per 10 0,000 people each year) 
900 
BOO 
700 
600 
Age 45-54, all risks 584 
500 
400 
300 If Smoker (at least one pack/day) 
Age 35-44, all risks 229 
200 If Skydiver 
Age 25-34, all risks 137 
100 
80 If Fireman {professional) 
25 If Police Officer 
0.05 By Lightning 
-0-
Figure 8. Risk ladder illustrating annual risks of dying {Mitchell 
and Carson, 1986, p. 46). 
l 
I I 
-0-
104 
LOWER-LEVEL RISKS OF DYING 
(annual) 
22 If Police Officer 
21 In Auto Accident 
20 If Have Appendectomy Operation 
15 In Airliner Crash (150 trips) 
If Woman Having a Baby 
By Drunk Driver 
If Woman Contraceptive Pill User (age 25-34) 
In Home Fire 
As Pedestrian 
------lllotl.OO In Airliner Crash (10 trips) 
0.95 
(one in one million people) 
0.75 
0.50 In Airliner Crash (5 trips) 
0.25 
0.10 In Airliner Crash (1 trip) 
0.05 By Lightning 
-o-
Lifetime 
Total Cigarettes 
(for comparison} 
443 
422 
403 
221 
88 
56 
21 
15 
10 
2 
1 
Figure 9. Risk ladder illustrating lower-level risks of dying (Mitchell and Carson, 1986, p. 47). 
! I 
I 
I I 
105 
they would be willing to increase their water bills. If the respondents 
are familiar with the concept of paying water bills, then the number of 
nonresponses may be less than if respondents were not familiar with the 
concept of paying water bills. 
For the purposes of this example CV survey, assume that the Yellow­
stone Valley population relies solely on a public water supply that 
originates as groundwater. The survey will be designed to elicit WTP 
bids for a reduction in the groundwater concentration of Ro-Neet as a 
result of purifying the water supply. Survey participants will be shown 
two risk ladders similar to those of Mitchell and Carson (1986), but 
since cancer risks are not readily classified as annual risks, this case 
study's risk ladders would present a variety of risks of dying prema­
turely (e.g., an individual's risk of dying prematurely if his occupation 
is that of a race car driver) . As with Mi tche 11 and Carson's risk 
ladders, the bottom portion of the first risk ladder will be enlarged and 
will be used in a second risk ladder to illustrate the low-level risks 
involved in this case study scenario. This second risk ladder will 
present a variety of low-level risks of dying prematurely (e.g., risk of 
being killed by lightning, risk of dying during an appendectomy opera­
tion). The levels of risk of dying prematurely of cancer presented by 
the two relevant groundwater concentrations of Ro-1Neet (z1 and z2) will 
also be included in this risk ladder. Determining the human health risks 
associated with levels of groundwater concentration of a specific 
chemical may be difficult, as was the case with groundwater concentra­
tions of Ro-Neet. Gosselin et al. (1984) indicate that for a "moderately 
toxic" pesticide such as Ro-Neet, the probable oral lethal dose for a 70 
' I 
. I 
106 
kg. person is a dose of 0.5- 50 gm./kg.; this wide range of values of 
lethal dose may be meaningless on a risk ladder. This case study looks 
at the change in the risk of dying of cancer, but Gosselin et al. (1984) 
do not provide information on the dose-response relationship for chronic 
effects such as cancer. Although EPA may have this type of information, 
it is not currently available to the public. 
Assuming that the dose-response relationship was available, the CV 
survey could be conducted .as follows. One adult per household surveyed 
will be asked if they are willing to pay higher water bills (to cover the 
costs of water treatment} in order to reduce chemical concentrations in 
groundwater from z1 to z2 (where z1 > z2}. Those who respond positively 
will then be asked for the highest amount they would be willing to pay 
for this reduction in groundwater concentration. 
Once a CV survey has been conducted, the resulting data can be used 
to estimate the value of a statistical life {VSL). Mitchell and Carson 
{1986} present several methods of calculating the VSL; two of these 
methods are presented below: 
{1) Calculate the implied VSL from the mean and median WTP values. WTP 
can be expressed as: 
WTP = s x VSL 
where s is the probability of a health risk occurring. Therefore, 
VSL can be calculated by dividing each of the WTP bids by its 
respective risk reduction. Because in this case study WTP bids of 
households {not individuals} were elicited, WTP bids would be 
divided by the respective risk reduction and by the size of the 
household surveyed. 
I I 
ir 
1 
I ~ 
! 
. .I 
107 
{2) Use individual observations to estimate VSL. This method is based 
on the calculation of compensating surplus (CS) as expressed in 
equation [8.1]. To use this method, Mitchell and Carson express 
the initial expenditure function as: 
E0 = E[p0 ,qoiU0*, Ro(AS,HB,OS,SEX), o=O, 
T(AS,AT,ED), HS] = INCC, 
where Po is a vector of current prices; q0 is a vector of currently 
provided public goods; Ro(•) describes the household's initial level 
of risk as a function of the household's age structure (AS), habits 
(HB), occupational structure (OS), and SEX; T(•) describes taste as 
a function of the household's AS, attitudes (AT), and education 
(ED); and HS is a measure of household size. 
Changing the level of risk but holding all other arguments constant 
yields a new expenditure function, ~: 
E1 = E[p0 ,q0 IU0*, Ro(AS,HB,OS,SEX), o=s1, 
T(AS,AT, ED), HS] = INCJ. 
The compensating surplus is the difference between these two expen­
diture functions and can be expressed as: 
CS = f[INC0 , Ro(AS,HB,OS,SEX), s1, T(AS,AT,ED), HS]. 
CS can be estimated as a function of its arguments, and the VSL can be 
calculated by dividing the CS by the respective level of risk. It should 
be noted that these two methods generally produce different estimates of 
the VSL. 
108 
Conclusions 
Although the general models and data are available to link together 
the chemical fate and transport, economic, and human health models, an 
app 1 i cation of this framework may be difficult because of a 1 ack of 
chemical-specific and site-specific data. Site-specific data needed for 
the chemical simulation models is available for only certain areas in the 
U.S., hence the reliance on site-specific studies instead of on national 
data bases. 
Because most management-oriented chemical fate and transport models 
do not s i mu 1 ate chemica 1 movement down to the groundwater zone, the 
researcher may be forced to make assumptions about the fate of a chemical 
as it moves between the root zone and the groundwater zone; this could 
result in under- or over-estimation of potential groundwater loading. 
Although using a CV method to value human health risks is consis­
tent with the concept of potential Pareto improvement, it can be subject 
to a variety of biases. This problem may be further exacerbated by a 
lack of data needed for the explanation of the CV scenario. In partic­
ular, data that indicate levels of health risks associated with level of 
groundwater pollution may not be available. 
, I 
109 
CHAPTER 9 
CONCLUSIONS AND RECOMMENDATIONS 
This study has identified the major components of a framework for 
the measurement and valuation of the social costs of groundwater pollu­
tion. In general terms, the major components of this framework are 
currently available in the relevant disciplines. There are, however, 
some important gaps in methods and data. Some of these gaps are due to 
the multi-disciplinary nature of the problem. A multi-disciplinary 
approach is needed, yet most physical and economic models are designed 
to be used by researchers within the respective discipline. To close 
these gaps, researchers must consider the broad implications and poten­
tial applications of their research, both within and outside their 
fields. Some research recommendations necessary to close these gaps are 
discussed below. 
A major gap in this linkage of models occurs because most 
management-oriented chemical fate and transport models do not simulate 
potential groundwater concentration resulting from a chemical applica­
tion. Modeling chemical movement down into the groundwater zone is a 
considerable task because much research is needed on the physical, 
biological, and chemical processes occurring in the unsaturated and 
saturated zones. However, it is a task that must be completed in order 
to model chemical transport down to a zone that would be meaningful to 
110 
participants of a contingent valuation exercise {e.g., the zone that 
supplies drinking water). 
One potential hindrance to the analysis of human health risks 
resulting from exposure to agricultural chemicals is difficulty in 
establishing causality. This problem of establishing causality is 
compounded in the case of groundwater po 11 uti on because exposure to 
contaminated groundwater is less direct than other means of exposure to 
agricultural chemicals. For example, the source of poisoning resulting 
from applying chemicals with a backpack sprayer may be easier to pinpoint 
than poisoning resulting from exposure to contaminated groundwater. 
Also, many of the health risks presented by agricultural chemicals are 
chronic effects with a long period of latency; because of the time 
between exposure to the chemical and manifestation of the disease, 
determining the cause of the disease may be difficult. 
Much of the human health risk analysis currently relies on animal 
toxicity data or epidemiologic studies; because of inter-species differ­
ences or lack of an adequate control group, extrapolation of human risks 
from these types of studies may be questionable. However, as discussed 
in Chapter 6, Gough (1989a) presents evidence that suggests that estimat­
ing human health risks from epidemiologic and toxicologic data may be 
appropriate. Also as discussed in Chapter 6, the current mathematical 
dose-response models do not incorporate chemical-specific data; incor­
porating this type of information may make extrapolation from these 
mathematical models more reliable. 
Another problem encountered when attempting to determine the 
potential health risks of an agricultural chemical is present because 
111 
testing of the inert ingredients of an agricultural chemical for acute 
and chronic health effects is not required. A complete analysis of the 
health effects of a chemical would include the health effects of the 
inert ingredients. 
Although a substantial amount of research has investigated the 
health risks of pesticides, relatively little research has been conducted 
on the health risks of fertilizers. In some areas nitrate contamination 
of groundwater may be more of a problem than pesticide contamination; 
concern over the he a 1 th risks of pesticides may be overshadow; ng the 
potentially greater health risks presented by fertilizers. This may be 
because more is known about the health risks of pesticides due to the 
availability of data on acute human toxicity of pesticides (e.g., 
especially resulting from pesticide application). 
One shortcoming of the currently available data is the lack of data 
on location-specific environmental characteristics, such as the type of 
data needed for the chemical fate and transport models. Also, data on 
the distribution of these environmental characteristics are needed to 
evaluate policy impacts. Although agricultural policies may be uniformly 
applied throughout the nation, the impacts of the pol icy may not be 
uniform. As illustrated by Antle and Just (1989), the relationship 
between, and distribution of, productivity-related characteristics and 
damage-producing characteristics must be determined in order to estimate 
the effects of a po 1 icy on the en vi ronmenta 1 damage of a particular 
region. 
112 
LITERATURE CITED 
i I 
! I 
113 
LITERATURE CITED 
Alexander, W.J., S.K. Liddle, R.E. Mason, W.B. Yeager. Groundwater 
Vul nerabi 1; ty Assessment in Support of the First Stage of the 
National Pesticide Survey. USEPA Contract No. 68-01-6646. 
Washington, DC: Research Triangle Institute, 1986. 
Algozin, K.A., W.Y. Huang, K.F. Alt, and B.M. Crowder. "Application of 
a Groundwater Contamination Database for Natural Resources Policy 
Analysis." Paper presented at Annual Meetings of American Agricul­
tural Economics Association, Knoxville, TN, 31 July - 3 August, 
1988. 
Aller, linda, Truman Bennett, Jay H. lehr, and Rebecca Petty. "DRASTIC: 
A System to Evaluate the Pollution Potential of Hydrogeologic 
Settings by Pesticides." In Evaluation of Pesticides in Ground 
~. Eds. W.Y. Garner, R.C. Honeycutt, and H.N. Nigg. 
Washington, DC: National Well Water Association, American Chemical 
Society, 1986. 
Antle, John M. "Incorporating Risk in Production Analysis." American 
Journal of Agricultural Economics 65 {1983a): 1099-1106. 
Antle, John M. "Sequential Decision Making in Production Models." 
American Journal of Agricultural Economics 65 {1983b): 282-290. 
Antle, John M. "Econometric Estimation of Producers' Risk Attitudes." 
Aroerican Journal of Agricultural Economics 69 (1987}: 509-522. 
Antle, John M. Pesticide Policy. Production Risk. and Producer Welfare. 
Washington, DC: Resources for the Future, 1988. 
Antle, John M., and Richard E. Just. "Effects of Commodity Program 
Structure on Resource Use and the Environment." Paper prepared for 
University of Maryland Commercial Agricultural and Resource Policy 
Symposium, College Park, MD, May 1989. 
Bailey, G.W., and R.R. Swank, Jr. 0 Modeling Agricultural Nonpoint Source 
Pollution: A Research Perspective.n In Agricultural Management and 
Water Oualjty. Eds. F.W. Schaller and G.W. Bailey. Ames, lA: Iowa 
State University Press, 1983. 
Baquet, Alan E., and James B. Johnson. Costs and Returns for Irrigated 
Crops: Yellowstone Valley. Extension Bulletin 26. Extension 
Service, Montana State University, Bozeman, MT, April 1988. 
- i 
114 
Barnwell, T.O., and R.C. Johanson. "HSPF: A Comprehensive Package for 
Simulation of Watershed Hydrology and Water Quality." In Nonpojnt 
Pollution Control: Tools and Techniques for the Future, pp. 135-
153. Technical Publication No. 81-1. Interstate Commission on the 
Potomac River Basin, Rockville, MD, 1981. 
Bear, Jacob, and Arnold Verruijt. Modeling Groundwater Flow and Pollu­
tion. Dordrecht, Holland: D. Reidel Publishing Co., 1987. 
Becker, Gary A. Economic Theory. New York: Alfred A. Knopf, 1971. 
Bellman, Richard. Dynamic Programming. Princeton, NJ: Princeton 
University Press, 1957. 
Berteau, P.E., and D.P. Spath. "The Toxicological and Epidemiological 
Effects of Pesticide Contamination in California Ground Water." In 
Evaluation of Pesticides in Ground Water. Eds. W.Y. Garner, R.C. 
Honeycutt, and H.N. Nigg. Washington, DC: National Well Water 
Association, American Chemical Society, 1986. 
Bicknell, B.R., A.S. Donigian, Jr., and T.A. Barnwell. "Modeling Water Quality and the Effects of Agricultural Best Management Practices 
in the Iowa River Basin." Water Science Technology 17 (1985): 1141-
1153. 
Brown, Charles C. "Approaches to Intraspecies Dose Extrapolation." In 
Toxic Substances and Human Risk: Principles of Data Interpretation. 
Eds. R.G. Tardiff and J.V. Rodricks. New York: Plenum Press, 1987. 
Burt, Oscar R., and John R. Allison. "Farm Management Decisions with 
Dynamic Programming." Journal of Farm Economics XLV, no. 1 (1963): 
121-136. 
Capalbo, Susan M., and John M. Antle, eds. Agricultural Productivity: 
Measurement and Explanation. Washington, DC: Resources for the 
Future, 1988. 
Carsel, R.F., L.A. Mulkey, M.N. Lorber, and L.B. Baskin. "The Pesticide 
Root Zone Model (PRZM): A Procedure for Evaluating Pesticide 
Leaching Threats to Groundwater." Ecological Modelling 30 (1985): 
49-69. 
Carsel, R.F., R.S. Parrish, R.L. Jones, J.L. Hansen, and R.L. Lamb. 
"Characterizing the Uncertainty of Pesticide Leaching in Agricul­
tural Soils." Journal of Contaminant Hydrology I (1988a): 111-124. 
Carsel, R.F., R.L. Jones, J.L. Hansen, R.L. Lamb, and M.P. Anderson. "A 
Simulation Procedure for Groundwater Quality Assessments of Pesti­
cides." Journal of Contaminant Hydrology 2 (1988b): 125-138. 
115 
Carsel, R.F., C.N. Smith, L.A. Mulkey, J.D. Dean, and P.P. Jowise. 
User's Manual for the Pesticide Root Zone Model (PRZM): Release 1. 
USEPA No. EPA-600/3-84-109. U.S. Environmental Protection Agency, 
Athens, GA, 1984. 
Chapman, S.R., and L.P. Carter. Crop Production. Practices and Prin­
ciples. San Francisco, CA: W.H. Freeman and Co., 1976. 
Coye, Molly J. "The Health Effects of Agricultural Production: I. The 
Health of Agricultural Workers." Journal of Public Health Poljcy 
6 (1985): 349-370. 
Cummings, Ronald G., DavidS. Brookshire, and William D. Schulze, eds. 
Valuing Environmental Goods. Totowa, NJ: Rowman and Allanheld, 
1986. 
Donigian, A.S., Jr., and R.F. Carsel. "Modeling the Impact of Conserva­
tion Tillage Practices on Pesticide Concentrations in Ground and 
Surface Waters." Environmental Toxicology and Chemistry 6 (1987): 
241-250. 
Dowd, Richard M. "EPA Revisits Dioxin Risk." Environmental Science and 
Technology 22 (1988): 373. 
Fitzpatrick, E.A. ~. New York: Longman, Inc., 1983. 
Franklin, C.A., N.I. Muir, and R. Greenhalgh. "The Assessment of 
Potential Health Hazards to Orchardists Spraying Pesticides." In 
Pesticide Residue and Exposure. Ed. J.R. Plimmer. Washington, DC: 
American Chemical Society, 1982. 
Freeman, A.M. Ihe Benefits of Environmental Improvement. Baltimore, MD: 
The Johns Hopkins University Press, 1979. 
Gianessi, Leonard P. Use of Selected Pesticides in Agricultural Crop 
Production by State. USEPA Report No. CR-813719-01-0. Report for 
the Office of Policy, Planning, and Evaluation, U.S. Environmental 
Protection Agency, Washington, DC, December 1987. 
Gianessi, Leonard P. Use of Selected Soluble Pesticides jn Agricultural 
Crop Production in Three States. USEPA Report No. CR-813719-02-0. 
Report for the Office of Policy, Planning, and Evaluation, U.S. 
Environmental Protection Agency, Washington, DC, February 1988. 
Gosselin, R.E., R.P. Smith, and H.C. Hodge. Clinical Toxicology of 
Commercial Products. Baltimore, MD: Williams and Wilkins, 1984. 
Gough, Michael. 1'Epidemiologic and Toxicologic-Based Produce Similar 
Estimates for Cancer Mortality Associated with ! Environmental' 
Exposures." Envj ronmentaJ Science and Techno] ogy (1989a, in press). 
' I 
116 
Gough, Michael. "Estimates of 'Environmental Cancers' and Regulatory 
Opportunities to Reduce Those Numbers." Personal communication, 
Resources for the Future, Washington, DC, 11 May 1989b. 
Hallberg, G.A. "Nitrates in Iowa Groundwater." In Rural Groundwater 
Contamination. Eds. F.M. D'Itri and L.G. Wolfson. Chelsea, MI: 
Lewis Publishers, Inc., 1988. 
Hallenbeck, W.H., and K.M. Cunningham-Burns. Pesticides and Human 
Health. New York: Springer-Verlag, 1985. 
Headley, J.D. "Estimating the Productivity of AGricultural Pesticides." 
American Journal of Agricultural Economics 50, no. 1 (February 
1968): 13-23. 
Jackson, R.J., C.J. Greene, J.T. Thomas, E.L. Murphy, and J. Kaldor. 
"Literature Review on the Toxicological Aspects of DBCP and an 
Epidemiological Comparison of Patterns of DBCP Drinking Water 
Contamination with Mortality Rates from Selected Cancers in Fresno 
County, 1970-1979." Unpub 1 i shed report to the Ca 11 forni a Department 
of Food and Agriculture, California Department of Health Services, 
Fresno, CA, 1 June 1982. 
James, A., ed. An Introduction to Water Quality Modelling. New York: 
John Wiley and Sons, Ltd., 1984. 
Johanson, R.C., J.C. Imhoff, J.L. Kittle, Jr., and A.S. Donigian, Jr. 
Hydrological Simulation Program-Fortran (HSPF): User's Manual for 
Release 8.0. USEPA Report No. EPA-600/3-84-066. U.S. Environmental 
Protection Agency, Athens, GA, 1984. 
Kamrin, M.A. "Health implications of Groundwater Contaminants." In 
Rural Groundwater Contamination. Eds. F.M. D'Itri and L.G. Wolfson. 
Chelsea, MI: Lewis Publishers, Inc., 1988. 
Kenaga, E.E., and C.A.I. Goring. "Relationship between Water Solubility, 
Soil Sorption, Octanol-Water Partitioning and Concentration of 
Chemicals in Giota." In Aquatic Toxicology. Eds. J.G. Eaton, P.R. 
Parrish, and A.C. Hendricks. Philadelphia, PA: American Society for 
Testing Materials, 1980. (ASTM SIP 707) 
Kilpatrick, Mary C. WATSTORE: A Water Data Storage and Retrieval System. 
USDI/GS Ref. No. 1981-341-618:52. U.S. Department of the Interior/ 
Geological Survey. Washington, DC: U.S. Government Printing Office, 
1981. 
Knisel, W.G., ed. CREAMS: A Field Scale Model for Chemicals. Runoff. and 
Erosion from Agricultural Management Systems. Conservation Research 
Report No. 26, Agricultural Research Service/USDA. Washington, DC: 
U.S. Government Printing Office, 1980. 
i I 
I 
I 
i 
< I 
117 
Knisel, W.G., G.R. Foster, and R.A. leonard. "CREAMS: A System for 
Evaluating Management Practices." In Agricultural Management and 
Water Quality. Eds. F.W. Schaller and G.W. Bailey. Ames, IA: Iowa 
State University Press, 1983. 
Lave, Lester B., and Eugene P. Seskin. Air Pollution and Human Health. 
Baltimore, MD: Johns Hopkins University Press/Resources for the 
Future, 1977. 
Leonard, R.A., and W.G. Knisel. "Evaluating Groundwater Contamination 
Potential from Herbicide Use." Weed Technology 2 (1988): 207-216. 
Leonard, R.A., W.G. Knisel, and D.A. Still. "GLEAMS: Groundwater Loading 
Effects of AGricultural Management Systems." Transactions of the 
.8SAE. 30 (1987): 1403-1418. American Society of Agricultural 
Engineers, St. Joseph, Michigan. 
Libra, R.D., G.R. Hallberg, and B.E. Hoyer. "Impacts of Agricultural 
Chemicals on Ground Water Quality in Iowa." In Ground Water Quality 
and Agricultural Practices. Ed. D.M. Fairchild. Chelsea, MI: Lewis 
Publishers, Inc., 1987. 
Loevinsohn, Michael E. "Insecticide Use and Increased Mortality in Rural 
Central Luzon, Philippines." The lancet (June 1987): 1359-1362. 
McElroy, A.D., S.Y. Chiu, J.W. Nebgen, A. Aleti, and F.W. Bennett. 
Loading Functions for Assessment of Water Pollution from Nonpoint 
Sources. USEPA Report No. EPA-600/2-76-151. U.S. Environmental 
Protection Agency, Washington, DC, May 1976. 
Mercer, M.W., and C.O. Morgan. Storage and Retrieval and Ground-Water 
Data at the U.S. Geological Survey. Geological Survey Circular No. 
856. Washington, DC: U.S. Government Printing Office, 1982. 
Mishan, E.J. "Evaluation of Life and Limb: A Theoretical Approach." 
Journal of Political Economy 70 (1970): 687-705. 
Mitchell, Robert C., and RichardT. Carson. Using Survey to Value Public 
Goods: The Contingent Valuation Method. Washington, DC: Resources 
for the Future, 1989. · 
Mitchell, Robert C., and RichardT. Carson. "Valuing Drinking Water Risk 
Reduction Using the Contingent Valuation Method: A Methodological 
Study of Risks from IHM and Giardia." Working draft. Resources for 
the Future, Washington, DC, 1986. 
Moses, Marion. "Pesticide-Related Health Problems and Farmworkers." 
Paper presented to the Hispanic Health Status Symposium, University 
of Texas Health Science Center, San Antonio, IX, 6 November 1987. 
118 
Mulkey, L.A., R.F. Carsel, and C.N. Smith. "Development, Testing, and 
Applications of Nonpoint Source Models for Evaluation of Pesticides 
Risk to the Environment." In Agricultural Nonpoint Source Pollution 
Model Selection and Applications. Eds. Aldo Giorgini and Franco 
Zingales. New York: Elsevier Science Publishing Co., Inc., 1986. 
Nazario, Sonia L. "EPA Under Fire for Pesticide Standards." The Wall 
Street Journal. 17 February 1989. 
Nielsen, E.G., and L.K. Lee. The Magnitude and Costs of Groundwater 
Contamination from Agricultural Chemicals. Agricultural Economics 
Report No. 576, USDA/ERS, Washington, DC, October 1987. 
Newberry, D.M.G., and J.E. Stiglitz. The Theory of Commodity Price 
Stabilization: A Study in the Economics of Risk. Oxford: Clarendon 
Press, 1981. 
Novotny, Vladimir. "A Review of Hydrologic and Water Quality Models Used 
for Simulation of Agricultural Pollution." In Agricultural Nonpojnt 
Source Pollution Model Selection and Applications. Eds. Aldo 
Giorgini and Franco Zingales. New York: Elsevier Science Publishing 
Co., Inc., 1986. 
Pearce, D.W., and C.A. Nash. The Social Appraisal of Projects. New 
York: John Wiley and Sons, 1981. 
Pimentel, David. "A Cost-Benefit Analysis of Pesticide Use in U.S. Food 
Production." In Pest; ci des: Contemporary Roles in Agrj culture, 
Health. and Environment. Eds. T.J. Sheets and David Pimentel. 
Clifton, NJ: The Humana Press, Inc., 1979. 
Rao, P.S.C., and J.M. Davidson. "Estimation of Pesticide Retention and 
Transformation Parameters Required in Non-Point Source Pollution 
Models." In Environmental Impact of Non-Point Source Pollution. 
Eds. M.R. Overcash and J.M. Davidson. Ann Arbor, MI: Ann Arbor 
Science Publishers, Inc., 1980. 
Rao, P.S.C., R.S. Mansell, L.B. Baldwin, and M.F. Laurent. "Pesticides 
and Their Behavior in Soil and Water." Soil Science Fact Sheet SL 
40. Institute of Food and Agricultural Sciences, Gainesville, FL, 
September 1983. 
Reed, P .A. "National Pesticides in Well Water Survey. n In Rural 
Groundwater Contamination. Eds. F.M. D'Itri and L.G. Wolfson. 
Chelsea, MI: Lewis Publishers, Inc., 1988. 
Sharp, D.S., B. Eskenazi, R. Harrison, P. Callas, and A. H. Smith. 
"Delayed Health Hazards of Pesticide Exposure." Annual Revjew of 
Public Health 7 (1986): 441-471. 
I 
I 
i i 
' i 
119 
Smith, V. Kerry, and William H. Oesvouges. "An Empirical Analysis of the 
Economic Value of Risk Changes." Journal of Political Economy 95 (1987): 89-114. 
Thaler, R., and S. Rosen. "The Value of Saving a life: Evidence from the 
labor Market." In Household Production and Consumption. Ed. N.E. 
Terleckyj. New York: Columbia University Press, 1976. 
Thompson, P.J., K. Young, W.O. Goran, .and A. Moy. An Interactive Sojls 
Information System User's Manual. Technical Report No. N-87/18. 
U.S. Army Corps of Engineers, Construction Engineering Research 
laboratory, Champaign, Il, July 1987. 
U.S. Department of Agriculture/Soil Conservation Service (Columbus, OH). 
Water Management and Sediment Control for Urbanizing Areas. 
Washington, DC: U.S. Government Printing Office, June 1978. 
U.S. Environmental Protection Agency. "Principles of Risk Assessment: 
A Non-Technical Review." Unpublished EPA paper cited at EPA 
Regional Risk Assessment Conference, Atlanta, GA, 29-30 May 1985. 
U.S. Environmental Protection Agency, Office of Drinking Water, Office 
of Pesticide Programs. National Pesticide Survey: Pilot Study 
Evaluation Summary Report. Washington, DC: U.S. Environmental 
Protection Agency, 1987. 
Wagenet, R.J., and J.l. Hutson. "Predicting the Fate of Nonvolatile 
Pesticides in the Unsaturated Zone." Journal of Environmental Quality 15, no. 4 (1986): 315-322. 
Wagenet, R.J., and J.L. Hutson. "lEACHM." Continuum 2 (August 1987): 
1-80. Water Resources Institute, Center for Environmental Research, 
Cornell University. 
Weinstein, Milton C., and Robert J. Quinn. "Psychological Considerations 
in Valuing Health Risk Reductions." Natural Resources Journal 23 
(1983): 659-673. 
Weinstein, M.C., O.S. Shepard, and J.S. Pliskin. "The Economic Value of 
Changing Mortality Probabilities: A Decision-Theoretic Approach." Quarterly Journal of Economics 94 (1980): 373-396. 
Williams, J.R., and W.V. LaSeur. "Water Yield Model Using SCS Curve 
Numbers." Journal of the Hydraulics Division, American Society of 
Civil Engineers 102 {1976): 1241-1253. 
Williams, W.M., P.W. Holden, D.W. Parsons, and M.N. Lorber. Pestjcjdes 
in Ground Water Data Base: 1988 Interim Report. Washington, DC: 
U.S. Environmental Protection Agency, December 1988. 
120 
Yeh, G.T. AT123D: Analytical Transient One-, Two-, and Three-Dimensional 
Simulation of Waste Transport in an Aquifer System. Report No. 
ORNL-5602. Oak Ridge National Laboratory, Oak Ridge, TN, 1981. 
Yoshida, K., T. Shigeoka, and F. Yamauchi. "Non-Steady-State Equilibrium 
Model for the Preliminary Prediction of the Fate of Chemicals in the 
Environment." Ecotoxi col ogy and Environmental Safety 7 (1983): 179-
190. 
Zaki, M.H., D. Moran, and D. Harris. "Pesticides in Groundwater: The 
Aldicarb Story in Suffolk County, NY." American Journal of Public 
Health 72 (1982): 1391-1395. 
121 
APPENDIX 
I 
. I 
I I 
. i 
I I 
i ! 
i I 
I I 
I 
122 
APPENDIX 
This appendix provides an overview of the PRZM model. Mulkey et al. 
(1986) present the mathematical advection-dispersion equations used in 
PRZM to estimate the changes over time of the chemical concentration in 
the compartments of the soil profile. The equation variables are: 
Cw = dissolved pesticide concentration in soil water, M/L3 
C6 = sorbed pesticide concentration, M/M 
o = ~oil volumetric water content, L3/L3 
V = compartment volume, L3 
p = soil bulk density, M/L3 
D = coefficient of dispersion and diffusion, L2/T 
v = water pore velocity, L/T 
ku = pesticide plant uptake efficiency, 0 ~ ku ~ 1.0 
T = volumetric plant transpiration rate, L3/T 
RO = volumetric rate of surface water runoff, L3/T 
~kw = summation of all first-order rate constants for trans-
formation of dissolved pesticides, 1/T 
Fd = fraction of applied pesticide in dissolved form 
Jap = pesticide application rate, M/T 
t = time, T 
x = vertical distance, L 
DPR = depth of surface runoff zone, L 
ERO = surface soil erosion rate, M/T 
~ks =·summation of all first-order rate constants for trans­
formation of sorbed pesticides, 1/T 
I 
i I 
i j 
123 
where l is a one-dimensional measure of space, T is a measure of time, 
and M is a measure of mass. 
For the surface zone where 0 ~ x ~ DPR, the equations are: 
a{cwoV) = JL 0 a{cwoV) 
at ax ax [A.l] 
[A.2] 
Equation [A.l] determines the change in the amount of dissolved 
chemical in the surface zone. The processes modeled include dispersion 
and diffusion of dissolved chemical, advection, plant uptake of the 
dissolved chemical, surface water runoff of dissolved chemical, trans­
formation processes acting on the dissolved chemical, and the amount of 
dissolved chemical applied to the area. Equation [A.2] determines 
the change in the amount of sorbed chemical in the surface zone. The 
processes modeled include erosion of sorbed chemical, transformation 
processes acting on the sorbed chemical, and the application of undis­
solved (solid) chemical. According to leonard et al. (1987), PRZM is 
somewhat unresponsive to rainfall for surface runoff and erosion; there­
fore, the amount of chemical carried off the field may be understated, 
causing the model to estimate "worst case" leaching potential. 
The equations representing the changes in the 1 ower subsurface 
zones (x ~ DPR), differ from those of the surface zone because chemical 
applications and runoff are not modeled in the lower subsurface zones: 
124 
a(CwoV) 
= a D a( CWOV) a(vCwoV) - kuTCw - ~~cwov 
-
at ax ax ax 
a(C5 pV) = 
- ~ksC5pV, for X 2.. DPR at 
Water Balance Equations 
It is necessary to estimate the movement of water through the soil 
because water transports dissolved chemical. Water balance equations are 
separately developed for: {1) the surface zone, (2) horizons comprising 
the active root zones, and {3) the remaining lower horizons within the 
unsaturated zone. Depending on the zone, these water balance equations 
may be functions of precipitation, snow melt, evaporation, transpiration, 
percolation, and soil water contained in the soil layers. Carse] et al. 
(1985) present the water balance equations: 
(1) for the surface zone: 
SW1,t+1 = SW1,t + Pt + S~ - INTt - ~ - E1,t - PERu 
{2) for the root zones: 
SW1,t+1 = SW1,t + PER1,t - E1,t - T1 - PER1.t 
{3) for the lower horizon zone: 
sw1,t+1 = sw1,t + PER1.t - PER1,t 
where 
SW1,t = soi 1 water in 1 ayer i for day t {em) 
INTt = interception of precipitation by plants for day t (em) 
E = evaporation for day t (em) 
P = precipitation as rainfall for day t {em) 
I I 
125 
R = surface runoff for day t {em) 
SM = snowmelt for day t {em) 
T = transpiration for day t (em) 
PER1 =percolation of soil water from zone i for day t (em). 
A modification of the SCSCN technique is used to distribute daily 
rainfall between runoff and infiltration. According to Carsel et al. 
(1984), this SCSCN method was chosen because it is a reliable procedure 
used for many years and the required inputs are generally available. The 
SCSCNs are determined on a daily basis. 
The daily evapotranspiration demand is divided among evaporation 
from plant canopy, soil evaporation, and plant transpiration. Total 
demand is first estimated and then extracted sequentially from crop 
canopy storage and from each layer until wilting point is reached in each 
layer or until total demand is met. (The moisture content in a soil 
layer at which plant survival cannot be achieved and the plant perman­
ently wilts is called the wilting point.) 
The perco 1 at ion component of the water ba 1 ance equation can be 
calculated using either of two options. With Option 1, percolation is 
defined in terms of field capacity and wilting point. [Bear and Verruijt 
(1987) define field capacity as the amount of moisture remaining in the 
soil after an extended period of gravity drainage without an additional 
supp 1 y of water at the soi 1 surface.] This option is based on the 
assumption that any soil water in excess of field capacity will percolate 
or drain freely into the next (lower) layer. Option 2 is provided to 
accommodate soils having low permeability layers that restrict the "free 
drainage" assumed in Option 1. 
126 
Erosion Equations 
To calculate the amount of sorbed pesticide carried away on eroded 
sediments, an estimate of soil erosion is required. Soil loss is esti ­
mated using a modified Universal Soil loss Equation (Williams, 1977) that 
is expressed as {Carsel et al., 1984): 
where 
Xe = a (Vrqp)0'56 KLSCP 
Vr = volume of event (daily) runoff {m3) 
qP = peak storm runoff (m3/sec) 
K = soil erodibility factor 
LS = length-slope factor 
C = soil cover factor 
P = conservation practice factor 
a = units conversion factor. 
Chemical Transport Equations 
To produce soil water and solid phase pesticide concentrations (Cw 
and C8 , respectively), the chemical transport component calculations 
include pesticide uptake by plants, erosion and surface runoff losses, 
decay, vertical movement, foliar loss, and dispersion. Plant uptak~ of 
pesticides is modeled by assuming uptake is directly related to the 
plant's transpiration rate. Degradation of the pesticide is represented 
as a first-order process; this representation includes all significant 
biochemical transformation and decay processes such as hydrolysis 
I I 
I 
I 
; I 
i 
I 
I I 
127 
(chemical decomposition in the presence of water), photolysis (chemical 
decomposition in the presence of 1 i ght) , and microbia 1 decay. The 
pesticide application rate is a simple input rate, but must be parti ­
tioned between the plant canopy and the soil surface.