ESTIMATION OF GROUNDWATER POLLUTION POTENTIAL BY
PESTICIDES IN MID-ATLANTIC COASTAL PLAIN WATERSHEDS
By Zhonglong Zhang1 and Mohamed M. Hantush2
ABSTRACT: A simple GIS-based transport model to estimate the potential for
groundwater pollution by pesticides has been developed within the Arc View GIS
environment. The pesticide leaching analytical model, which is based on one-
dimensional advective-dispersive-reactive (ADR) transport, has been directly
integrated into the GIS with menu interface and display tools to estimate the spatial
and temporal distribution of a potential pesticide's emission to groundwater. The
ADR model was chosen because it requires readily available data as compared to
other non-GIS models and it has the potential to handle multiple soil profile
descriptions. The ADR model has been used to assist with pollution assessment such
as the location and timing of pesticide spreading on watersheds, or choosing the most
effective "Best Management Practices". By embedding ADR modeling capabilities
into the GIS, one is able to evaluate the groundwater vulnerability to pesticide
contamination on a large scale where variable source areas are responsible for a part
or all of the groundwater contamination. To demonstrate the GIS-based contaminant
transport model and its capabilities, the program was applied to Mid-Atlantic coastal
plain agricultural watersheds, which are particularly vulnerable to agricultural
pesticide pollution.
INTRODUCTION
The fate of pesticides in the environment has been of great concern for several
decades because pesticides are a major source of nonpoint source (NPS) pollutants
and a major contamination threat to groundwater and surface water. A comprehensive
review of published information on the subsurface (groundwater and vadose zone),
conducted as part of the USGS' National Water Quality Assessment (NAWQA),
indicated that pesticides from every chemical class have been detected in ground
waters of the United States (Barbash, 1995). Many of these compounds are
commonly present at low concentrations in groundwater beneath agricultural land.
Protecting groundwater resources is especially important in agricultural areas where
most pesticides are used and where over 95 percent of the population relies upon
groundwater for drinking water. Therefore, there is an increasing need for
quantitative and objective assessment of environmental damage and water quality
impacts resulting from pesticide pollution. It is particularly important to be able to
identify the groundwater pollution potential due to pesticide leaching.
The description of pesticide pollution on a watershed scale is a complex
environmental problem because of the physical and chemical heterogeneities of the
subsurface. Pesticides are often spread over large areas, which makes it difficult to
determine their fates and to evaluate if they pose a threat to soil and groundwater
1.	ManTech Environmental, U.S. EPA NRMRL/SPRD, P. O. Box 1198, Ada, OK 74820
2.	U.S. EPA National Risk Management Research Lab., Subsurface Protection and Remediation Div.
(NRMRL/SPRD), P. O. Box 1198, Ada, OK 74820
1

-------
resources. Most aquifer contamination is discovered only after a water-supply well
has been affected. To minimize the risk of groundwater contamination by pesticides
and to avoid the need for costly remediation efforts, it is essential to be able to predict
specific areas in a watershed that are at risk for pesticide leaching and to estimate the
amount of leaching that is possible. These efforts are the first step in improving
watershed management.
Because of the complexity of the processes involved, a recent trend in the
regulation of pesticide use has involved an increased reliance on solute-transport
models to predict the transport and fate of pesticides in the subsurface. Transport
simulation models have become a useful tool in understanding and analyzing NPS
pollution problems caused by the migration of pesticides through soil into
groundwater. These models can be used for the preliminary screening of the areas
susceptible to severe or high contamination as well as predicting the consequences of
management alternatives. While groundwater contamination by pesticides may be
predicted by several available leaching models, few models are currently available to
predict the spatial pattern of variable source areas. Some of the limitations of models
include the inability to handle and manage large amounts of model input data and to
account for the spatial variability and heterogeneity that are important in NPS
modeling processes. Moreover, improving model capability requires the management
of increasing volumes of spatially referenced data.
For these reasons, simulating pesticide movement through the subsurface has
been difficult. It has been recognized that the spatial-temporal variability in such
factors as topography, soil, climate, land use, and land management influences the
response of natural systems and limits the applicability of the models. GIS
(Geographic Information Systems) technology has gained widespread acceptance as a
data management and visualization tool for addressing spatially related environmental
problems. Modeling the fate and movement of pesticides in the subsurface is a spatial
problem well suited to the integration of a solute transport model with GIS. When
GIS and transport models are merged, they provide an efficient means for handling
the complex spatial heterogeneities of the surface and subsurface. The GIS-based
model has simple data requirements and is defined within a spatial context as
compared to more complex models that require routing information. The end products
of a GIS-based model of NPS pollutant fate and transport in the subsurface are maps
that show the spatial distribution of a solute within the unsaturated zone and solute
loading to groundwater. In both cases, an improvement with respect to the classical
non-GIS models has been accomplished (Zhang, 1998).
GIS-based models used to estimate NPS pollution range from simple empirical
models to complex physically based models (Corwin, et al. 1997). The pesticide
leaching analytical model used in this paper is a simple one-dimensional adveetive-
dispersive-reactive (ADR) transport model which estimates the potential leaching of
pesticides occurring at any point in the watershed. It was developed with the objective
of aiding management in identifying potential pollution source areas of watersheds.
By integrating the ADR model into a GIS, a tool with menu interface is created for
the prediction of pesticide leaching in the subsurface and making preliminary
groundwater contamination assessments within a GIS environment. This approach
also demonstrates the advantages of embedding subsurface transport modeling
capabilities into GIS to evaluate the groundwater vulnerability to pesticide
2

-------
contamination on a large scale (e.g. watershed) where a variable source area is
responsible for a part or all of the groundwater contamination.
In this study, the pesticide leaching analytical model was integrated within
Arc View GIS and its associated databases to estimate the spatial and temporal
distribution of pesticide leaching and potential vulnerability of the groundwater. The
ADR model is based on an analytical solution of the advective-dispersive-reactive
transport equation, which describes transport and fate of pesticides in the soil. The
model has also been tested under field and laboratory conditions. This paper describes
the model development within the Arc View environment and a field example
application. The example is used to demonstrate the capabilities of the GIS-based
pesticide leaching model.
PESTICIDE LEACHING ANALYTICAL MODEL
There are many pesticide leaching models available, ranging from simple to
sophisticated numerical ones. The models can be classified as either deterministic or
stochastic. Both deterministic and stochastic models can be further subdivided into
lumped or distributed models, depending on the treatment of space. A lumped model
represents the physical system as a spatially-homogeneous region which does not
account for the spatial variation of input parameters within the area. On the other
hand, a distributed model assumes that the physical system is made of discrete
subareas, each characterized by a uniform set of properties and input parameters. The
volume of information needed to temporally and spatially characterize the parameters
and variables in even the simplest functional models of solute transport in the vadose
zone is tremendous (Corwin, et al. 1997). Moreover, the most comprehensive models
do not always provide better results than simpler ones (Hutson and Wagenet, 1991).
The pesticide leaching analytical model used in this work is described in detail by
Ravi and Johnson (1992). The model is a one-dimensional advective-dispersive-
reactive transport equation. The vertical transport of a pollutant dissolved in water
through the soil can be described by the following principal governing equation:
ae ^d2c ac P- as
— - d— v—
5t dx dx
_ = DTrT-v—~K,C	(1)
where C is the liquid-phase pollutant concentration (M/L ); t is the time (T); x is the
distance along the flow path (L); D is the dispersion coefficient (L2/T); v is the
interstitial or pore-water velocity (L/T); pb is the bulk density (M/L3); 0 is the
volumetric water content (L3/L3); S is the solid-phase concentration (M/M); and K| is
the first-order decay coefficient in liquid phase (1/T).
The term dS/dt is the rate of loss of solute from liquid phase to solid phase due to
sorption. Under the assumption of linear, instantaneous sorption, dS/dt can be
estimated by:
— = K —	(2)
at " ax	w
3

-------
where IQ is the linear Freundlich sorption coefficient, K Ks
(5)
where r is the rate of infiltration, Ks is the saturated hydraulic conductivity, and 0S is
the saturated water content.
Assuming the following initial and boundary conditions:.
0, -co oo
dx	11
the solution of (3) is obtained as follows:
erf
C(x,t) = ^-e-klt
rx + h-vt/Rs
2 y/Dt/R
erf
^ x-vt / R^
2 -jDt/R
(7)
(8)
where erf(z) is the error function which is defined as erf(z) = 2/4x\le y2dy. It
should be noted that the infinite boundary condition may cause prediction error at
short time and low vadose zone depth
4

-------
The model above simulates vertical solute transport over an area as if all
parameters within the area are uniform. It does not consider the effects of spatial
variability. Parameter values for the area as a whole are obtained by spatially-
averaging individual values. The model provides a unique output result for the whole
area, but it does not provide any information regarding the spatial behavior of the
outputs. So the model above is a lumped parameter model. The problem with this
lumped parameter model is that spatial variability can have significant impacts on
area response.
MODEL INTEGRATION INTO GIS
An integrated GIS system is comprised of three components: the model, the GIS
and the data. When working with vadose zone transport models, Tim (1996) and
Corwin et al. (1997) refer to three potential levels of integrating GIS with NPS
models. For the first level of integration, known as loose coupling integration, the
GIS and the model are developed separately and are executed independently. The GIS
serves only as a pre-processor of the input data for the model. The second level of
integration, partial integration, is the result of establishing an interactive interface
between the GIS and the model. In this level of integration, the GIS not only provides
input data to the model, but also accepts modeling results from the model for further
processing and/or presentation. The third level of integration is typically referred to as
complete integration or "modeling within GIS" which was used in this paper. For this
level of integration, the functionality of the model is implemented or programmed
directly into the GIS. With the equations programmed within the GIS, an interactive
and fully integrated contaminant transport modeling process can be performed within
the same environment, so that data pre-processing and analytical functions are
available under the same operating system. This level of integration is technically
preferred by most modelers, but is often difficult to implement due to
incompatibilities in the data structures of the model and the GIS, or due to proprietary
rights of commercial GIS software limiting the introduction of additional processing
routines. Fig. 1 shows a schematic illustration of the third level of integration for GIS
and models.
The NPS model can be expanded to act as a distributed parameter modeling
application through integration with the GIS. Distributed models break an area down
into a number of smaller homogeneous subareas or elements with uniform soils,
cropping, and topographic characteristics. Some distributed parameter models have a
cellular or grid structure, which simplifies database creation. A grid is placed over the
area of interest and then parameter values are obtained for each grid based on the
predominate parameter in the grid, or by area weighting parameter values within the
cell. In essence, averaging is done, but on a much smaller scale so that some spatial
variability is still maintained. The area is then modeled by solving the equations
describing the state of each individual element. The entire area response is obtained
by integrating all the elemental responses. The distributed parameter model also gives
information on what is happening within each element and can therefore be used to
identify critical locations within the modeled region. Lumped models do not have this
capability.
In order to run the model within ArcView, Avenue scripts were written to
calculate each of the transport processes described above and produce modeling
5

-------
outputs at specified time steps. The GIS database in the project was created by
extracting soil data for soil series occurring in the area. These data included soil
classification, many chemical parameters, as well as soil-water parameters. Weather
data input is read from interactive menus containing precipitation and potential
evapotranspiration. Actual evapotranspiration is calculated for each land use using a
table of crop factors. All other non-spatial data are entered from the menus within
ArcView. Output maps display the pollutant concentration profile at any given time.
APPLICATION TO MID-ATLANTIC COASTAL PLAIN WATERSHEDS
As an example application, the GIS-based pesticide leaching analytical model was
used to assess the impact of pesticide leaching on groundwater quality within
agricultural watersheds located in Kent County, Maryland. This example application
is provided for preliminary assessment and is not intended to evaluate the water
quality problems resulting from applied pesticides in this watershed region. To further
this model as a highly effective pollution assessment tool would require
improvements in the level of detail of source data.
The study area shown in Figure 2 is approximately 120 km2. Agriculture
dominates the landscape in the study area and most of the land there is used to grow
corn and soybeans in an annual rotation with winter wheat. The soil in the study area
is predominantly silt loam and to a lesser extent ranged from loam to sandy and
gravely loams. The GIS database created for the pesticide leaching analytical model
focused on the information required to implement the integrated GIS modeling. The
primary data included the soil characteristics and other chemical parameters. The
pesticides chosen in this study were Altrazine and Bromacil. Selected pesticide
properties are listed in Table 1.
Groundwater recharge estimates are based on the monthly water balance between
the infiltration water from precipitation and irrigation, and the outflow of water from
surface runoff and evapotranspiration (Thornthwaite and Mather, 1957). The runoff
volume, or rainfall excess, is estimated using the curve number approach developed
by the Soil Conservation Services (SCS) (Haan et al., 1994). The SCS has classified
more than 400 soils into four hydrologic soil groups according to their minimum
infiltration rate obtained for bare soil after prolonged wetting. The hydrological soil
groups are built into the GIS database in this study. Evapotranspiration is estimated
using quasi-empirical methods which rely on energy balance and heat transfer
concepts and empirical crop factors due to lack of real-time measurements.
The results presented in Figure 3 represent the preliminary groundwater
vulnerability predictions to pesticides applied to soils that are suitable for agricultural
lands. A single application of 2.5 lb/acre is assumed in the predicted groundwater
vulnerability maps. The predicted results show that both Atrazine and Bromacil have
degraded less than the Maximum Concentration Limit (MCL) before leaching into
groundwater. Pesticides with short half-life and relatively high organic carbon
partition coefficient as altrazine showed low concentration level. Compared to
Atrazine, Bromacil is considered highly mobile since it has a relatively large half-life
and low organic carbon partition coefficient. While a model like ADR can be used to
identify problem areas in general, a more detailed examination may be necessary to
determine the validity of the spatial groundwater vulnerability predictions and suggest
conservation management practices which could lessen the problem. As more
6

-------
detailed data becomes available, comprehensive models are more suitable for these
purposes.
CONCLUSIONS
In this study, an one-dimensional advective-dispersive-reactive transport model
was programmed and embedded within the ArcView GIS environment to predict
pesticide leaching potential. This approach provides a full range of model
input/output data manipulation capabilities and an improved estimate for pesticide
contamination in groundwater. This GIS-based transport model provides the
descriptive framework to estimate a spatially variable pesticide leaching distribution
in the subsurface that cannot be achieved by the original model. The GIS format
allows data to retain its spatial relationships and keeps track of all modeling
parameters and state variables. GIS also creates a user friendly environment and
allows modification for simulating different scenarios.
The GIS-based ADR model can be used for pollution assessment such as the
location and timing of pesticide spreading on watersheds, or choosing the most
effective "Best Management Practices". The major advantage of a GIS-based
transport model is that the procedure automates and facilitates spatial modeling
assessment and considers the heterogeneities of land use and soil. However, like most
simulation models, the ADR model used in this paper is mathematically and
conceptually ideal. It therefore may not accurately simulate natural conditions since
the approach is a simplified one and does not incorporate processes such as two
dimensional convective-dispersive transport or adsorption reactions. In some cases
the influence of these processes cannot be neglected and may therefore constitute a
limitation to the practical use of the proposed model. Continued work on this topic
will focus on developing an infiltration component and adding a plotting routine in
ArcView.
ACKNOWLEDGEMENTS
The research leading to this paper is supported by the U.S. Environmental
Protection Agency through its Office of Research and Development. It has not been
subject to Agency review and therefore does not necessarily reflect the views of the
Agency, and no official endorsement should be inferred. The authors would like to
acknowledge the contributions of Center for Subsurface Modeling Support personnel.
REFERENCES
Barbash, J.E. (1995). "Pesticides in ground water: An overview of current
understanding." The Rubin Symposium on Soil Science, Solute Transport, and
National-Scale Water Quality Assessments, American Geophysical Union,
Baltimore, Maryland, EOS, 76(17), pp. S142.
Carsel, R.F., Mulkey, L.A., Lorbcr, M.N., and Baskin, L.B. (1985). "The Pesticide
Root Zone Model (PRZM): A procedure for evaluating pesticide leaching threats
to groundwater." Ecological Modeling, 30, 49-69.
7

-------
Corwin, D.L., Vaughan, P.J., and Loague, K. (1997), "Modeling nonpoint source
pollutants in the vadose zone with GIS." Environmental Science & Technology,
31(8), 2157-2175.
Haan, C.T., Bar field, B.J. and Hayes, J.C. (1994). Design Hydrology and Sediment
logy for Small Catchments, Academic Press, Inc., California.
Hantush, M.M., Islam, R., and Marno, M.A. (2000). "Models for soil-aquifer
vulnerability assessment and pollution prevention." J. Hydrol., 227(2000), 66-80.
Hutson, J.L., and Wagenet, R.J. (1991). "Simulating nitrogen dynamics in soils using
a deterministic model." Soil Use and Management 7(2), 74-78.
Kleveno, J. J., Loague, K., and Green, R. E. (1992). "Evaluation of a pesticide
mobility index: impact of recharge variation and soil profile heterogeneity." J,
Contam. Hydrol., 11, 83-99.
Ravi, V., and Johson, J. A. (1992). PESTAN Pesticide Analytical Model, USEPA
R. S. Kerr Environmental Research Laboratory, Ada, Oklahoma.
Thornthwaite, C.W., and Mather, J.R. (1957). "Instructions and tables for computing
potential evapotranspiration and the water balance." Publications in Climatology
Vol. X, No. 3, Centerton, New Jersey.
Tim, U. S., Jain, D., and Liao, H.H. (1996). "Coupling vadose zone models with GIS:
Emerging trends and potential bottlenecks." J. Environ. Qual., 25(15), 535-544.
Van, der Zee, S.E.A.T.M., and Boesten, J.J.T.I. (1991). "Effects of soil heterogeneity
on pesticide leaching to groundwater." Water Resour, Res., 27(12), 3051-3063.
Wagenet, R.J., and Huston, J.L. (1986). "Predicting the Fate of Nonvolatile Pesticides
in the Unsaturated Zone." J. Environ. Qual., 15(4), 315-322.
Zhang, Z. (1998). "Simulating agricultural non-point source (NPS) pollution in the
coastal plains using geographic information system (GIS) interfaced model
systems." Ph.D. Dissertation, Clemson University, South Carolina.
Table 1. Selected pesticide properties
Pesticide
Solubility
(Kg/m3)
Koc (m3/kg)
Kh!
X (days)2
MCL (ppb)3
Atrazine*
Bromacil*
32x10"3
82x10"2
0.16
7.2x10"2
2.5 xlO'7
3.7 xlO"8
71
350
3
90
* Adapted from Hantush et al. (2000)
1. Kh is a Henry's constant, 2. X is a half-life, and 3. MCL is a Maximum Concentration Limit.
8

-------
Hydrography
Climate

Topography



Field Boundary
Soils

Land Cover
isjjii'	i
Communication bit ,
>itfJSyS^l!ji,^i. . - ¦	"j	
ra
Monitoring Data
Soil Sampling Data
MODEL
GIS
GRAPHICAL USER INTERFACE
Fig 1. Schematic representation of GIS applications integrated with model (Tim, et
al., 1996)
>-v"

/ 'V
(a)
So ils
IP Gravel (sand)
Gra vellyloam
Loam
Loamysand
Sandyclayloam
:"-i Sandyloam
Sittyloam
(b)
Water table
(feet)
0
1	• 13
Pfj 13 • 20
—| 20 • 28
— 28 • 37
37 • 46
46 • 04
Fig. 2. Study area (a) soil characteristics and (b) groundwater table
9

-------
mM-*

Atrazine (ppb)
0
Mm- 0,004
¦1 water
Atra2ine (ppb)
0
0- 0.006
att 0.008 • 0.077
HI 0.077 • 0.382
¦Hi water
X
}
Bromacil (ppb)
0- 0,002
IS 0.002 - 0.012
mm 0.012 - 0.246
{¦ 0.240 - 1 08?
¦¦ mater
(c)
Bromacil (ppb)
0- 1.821
1.821 - 4.021
41151 . 43 03?
PMf 43.032 - 88.738
¦¦ water
Fig. 3. Simulated contaminant concentration in groundwater (a) atrazine (3 months),
(b) atrazine (6 months), (c) bromacil (3 months), and (d) bromacil (6 months).
10

-------
TECHNICAL REPORT DATA
NRMRL ADA uuo44 (Please read Instructions on the reverse before compleV
1. REPORT NO. 2.
EPA/600/A—01/021
3. RECI
4. TITLE AND SUBTITLE
Estimation of Groundwater Pollution Potential by Pesticides in Mid-
Atlantic Coeastal Plain Watersheds
S. REPORT DATE
6. PERFORMING ORGANIZATION CODE
7. AUTHORS
Zhonglong Zhang* and Mohamed M. Hantush**
8. PERFORMING ORGANIZATION REPORT NO
9. PERFORMING ORGANIZATION NAME AND ADDRESS
*ManTech Environmental Research Services Corporation,
NRMRL/SPRD, R.S. Kerr Environmental Research Center, Ada, OK.
**NRMRL/SPRD, R.S. Kerr Environmental Research Center, Ada, OK.
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-C-98-138
12 SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
NRMRL/SPRD
R.S. Kerr Environmental Research Center
13. TYPE OF REPORT AND PERIOD COVERED
Symposium Paper
14. SPONSORING AGENCY CODE
EPA/600/15
15T&PPLEMtNTARY NOTES
16. ABSTRACT
A simple GIS-based transport model to estimate the potential for groundwater pollution by pesticides has been
developed within the ArcView GIS environment. The pesticide leaching analytical model, which is based on one-
dimensional advective-dispersive-reactive (ADR) transport, has been directly integrated into the GIS with menu
interface and display tools to estimate the spatial and temporal distribution of a potential pesticide's emission to
groundwater. The ADR model was chosen because it requires readily available data as compared to other non-GIS
models and it has the potential to handle multiple soil profile descriptions. The ADR model has been used to assist
with pollution asssessment such as the location and timing of pesticide spreading on watersheds, or choosing the most
effective "Best Management Practices". By embedding ADR modeling capabilities into the GIS, one is able to
evaluate the groundwater vulnerability to pesticide contamination on a large scale where variable source areas are
responsible for a part or all of the groundwater contamination. To demonstrate the GIS-based contaminant transport
model and its capabilities, the program was applied to Mid-Atlantic coastal plain agricultural watersheds, which are
particularly vulnerable to agricultural pesticide pollution.
17. KEY WORDS AND DOCUMENT ANALYSIS
a DESCRIPTORS
b. IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
ground water
pesticides
advective-sispersive-reactive (ADR) transport
spatial
watersheds


18. DISTRIBUTION STATEMENT
Release to the public.
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
20. SECURITY CLASS (This Page)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (Rev. 4-7?) PREVIOUS EDITION IS OBSOLETE	forms/admin/techrptfrm 7/8/99

-------