MEASUREMT AND PREDICTION OF HERBICIDE
TRANSPORT INTO SHALOW GMNDWATER
NATIHEW CIAY SMITH
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENT
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLODRIDA
To my mother,
Corrine L. Smith
She was taken from her family by illness during this research. Her
wisdom, courage, and strength continue to inspire me. Without that inner
strength, I would never have made it this far.
I love you Mom.
I would like to express my sincere appreciation to the following:
Dr. A. B. (Del) Bottcher, chairman of my advisory ccmittee, for
his friendship, patience, and guidance. He provided valuable advice and
philosophy during the highs and lows encountered in my Ph.D. program.
Dr. K. L. Campbell, cochairman of my advisory committee, for his
guidance and support. His calm and reasoned approach to problem solving
is a model which I need to emulate.
Dr. E. D. Threadgill, for serving on my advisory committee, for
making available the field site on which this research was conducted,
and for being the best engineer/scientist/administrator that I could
ever hope to be associated with.
Dr. W. C. Huber, for serving on my advisory coanittee, and being
one of the very best instructors that I have encountered during many
years of college. If all instructors were that good, my GPAs would have
Dr. P. S. C. Rao, for serving on my advisory committee, for showing
a sincere interest in my research, and for teaching me several very
valuable lessons during the qualifying exam.
Ms. M. W. Smith, for being a wonderful, loving, and supportive
wife. She assumed a greatly disproportionate share of child care and
household duties so that I could devote time to this research.
Nicholas and Sarah Smith, my children, for their love and for
providing meaning and purpose to my life and education.
Dr. G. W. Isaacs, for financial support in the form of a research
Dr. D. L. Thomas, for his friendship and support. He shielded me
from many distractions at work while I mcopleted this document.
Dr. W. B. Wheeler and Ms. S. J. Scherer, at the UF Pesticide
Residue Laboratory, for allowing me to use their equipment and guiding
me through the intricacies of pesticide residue analysis.
Mr. L. A. Asmussen, Dr. R. A. Leonard, Dr. W. G. Knisel, Ms. L. R.
Marti, and others at the USDA Southeast Watershed Research laboratory for
their advice, encouragement, and support.
USDA, Southern Region Pesticide Impact Assessment Program, for
financial support in the form of a grant.
Many others, too numerous to list here, in both Gainesville and
Tifton who contributed in many ways to the success of the project.
TABLE OF COENETS
LIST OF TABLES .................................................... vii
LIST OF FIGURES .................................................. viii
ABSTRACT' ........................................................... xiv
1 INTRODUCTION ................................................ 1
2 OBJECTIVES.................................................. 4
3 REVIEW OF THE LITERATURE.................................... 5
3.1 Evidence of Pesticide Residues in Groundwater......... 5
3.2 Factors Which Influence Pesticide Transport to
to Groundwater........................................ 7
3.3 Predicting Pesticide Transport....................... 15
3.4 Field Studies of Pesticide Transport................. 27
3.5 Summary .............................................. 36
4 EXPDERIMENTAL METHODS ....................................... 39
4.1 Field Site Description............................... 39
4.2 Site Instrumentation................................. 42
4.3 Chemical Applications................................ 49
4.4 Sample Collection and Storage........................ 52
4.5 Sample Analysis...................................... 55
5 MODELING THE EDPERIMENTAL SITE............................. 66
5.1 Selection of Ccommon Input Parameter Values........... 67
5.2 Parameters Unique to PRZM............................ 73
5.3 Parameters Unique to GLEAMS.......................... 77
6 RESULTS AND DISCUSSION..................................... 79
6.1 Data Collection ...................................... 79
6.2 Sample Analysis ...................................... 83
6.3 Chemical Applications................................ 85
6.4 Chemicals in the Unsaturated Zone.................... 91
6.5 Chemicals in the Saturated Zone...................... 107
6.6 Model Results and Comparisons....................... 142
7 SUMMARY AND CONCLUSICS ................................... 157
8 RECMMENDATICNS FOR IMF3OVEME3S AND FURTHER STUDY........ 160
A. MO41IORING WELL STATISTICS................................ 170
B. HOW TO GET COMPLETE DATA SET ............................. 173
C. WATER BAIANCE PROGRAMS.................................... 176
D. SURFACE PLOTS OF ATRAZINE CONCENTRATION IN GROUNDWATER.... 200
BIOGRAPHICAL SKETCH............................................... 213
LIST OF TABLES
5.1. Soil properties used in simulations........................ 69
5.2. Chemical properties used in simulations ..................... 70
5.3. Chemical applications summary.............................. 73
6.1. Chronological summary of field site activities..............80
6.2. Chemical application results............................... 88
6.3. Measured application rates and soil surface concentrations
of atrazine and alachlor................................... 91
6.4. Mean concentrations of atrazine (mg/kg) in soil samples..... 96
6.5. Mean concentrations of alachlor (mg/kg) in soil samples..... 97
6.6. Mean concentration of bromide (mg/L) at each sampling
following first application............................... 105
6.7. Mean concentration of bromide (mg/L) at each sampling
following second application............................... 106
6.8. Total mass of chemicals in the saturated zone............... 133
6.9. Simulated mass flux of chemicals........................... 142
A.1. Monitoring well statistics................................ 171
B.1. Example listing of the water sample data set............... 174
B.2. Example listing of the soil sample data set................ 175
C.i. Sample output from program ANALYZE........................ 191
C.2. Sample output from program FLUX........................... 198
IST OF FIGURES
4.1. Contour maps of soil surface and restricting layer
showing locations and ID labels of monitoring wells..........41
4.2. Cross-section of soil profile through application area
showing locations of monitoring wells and soil solution
6.1. Uniformity of atrazine application.......................... 86
6.2. Uniformity of alachlor application.......................... 87
6.3. Uniformity of bromide application on 4/27/87.................89
6.4. Comparison of alachlor and atrazine concentrations in
application samples......................................... 90
6.5. Correlation between atrazine concentrations in the top 5 cm
of soil and application solution concentrations.............. 93
6.6. Correlation between atrazine concentrations in the top 5 cm
of soil and application rate................................ 93
6.7. Correlation between alachlor concentrations in the top 5 ancm
of soil and application solution concentrations.............. 94
6.8. Correlation between alachlor concentrations in the top 5 ancm
of soil and application rate................................ 94
6.9. Bromide and atrazine concentrations in solution sampler
09N-2 as a function of total water applied since
6.10. Bromide concentration for six sampling locations following
the first application at a 61 cmn depth..................... 101
6.11. Bromide concentration for six sampling locations following
the first application at a 122 on depth ..................... 101
6.12. Bromide concentration for six sampling locations following
the first application at a 183 an depth.................... 102
6.13. Average bromide concentration for the three sampling depths
following the first application............................ 102
6.14. Bromide concentration for six sampling locations following
the second application at a 61 cm depth..................... 103
6.15. Bromide concentration for six sampling locations following
the second application at a 122 cmn depth..................... 103
6.16. Bramide concentration for six sampling locations following
the second application at a 183 cm depth.................... 104
6.17. Average bromide concentration for the three sampling depths
following the second application............................ 104
6.18. Bromide concentration in the groundwater on 12/15/86......... 108
6.29. Bromide concentration in the groundwater on 6/01/87......... 113
6.30. Atrazine concentration in the groundwater on 2/23/87.........115
6.31. Atrazine concentration in the groundwater on 3/02/87.........115
6.32. Atrazine concentration in the groundwater on 3/09/87.........116
6.33. Atrazine concentration in the groundwater on 3/16/87......... 116
6.34. Atrazine concentration in the groundwater on 3/23/87......... 117
6.35. Atrazine concentration in the groundwater on 3/31/87......... 117
6.36. Contour plot of water table elevation on 5/08/87 showing
direction of flow ............................................ 118
6.37. Concentration of atrazine and water table elevation in
well 09-11.................................................. 120
6.38. Nitrate concentration in groundwater on 5/01/87............. 122
6.39. Nitrate concentration in groundwater on 5/03/87............. 122
6.40. Nitrate concentration in groundwater on 5/05/87..............123
6.41. Nitrate concentration in groundwater on 5/08/87.............. 123
6.42. Nitrate concentration in groundwater on 5/13/87............. 124
6.43. Nitrate concentration in groundwater on 5/18/87............. 124
6.44. Nitrate concentration in groundwater on 5/25/87..............125
6.45. Nitrate concentration in groundwater on 6/01/87..............125
6.46. Chloride concentration in groundwater on 5/01/87............. 127
6.47. Chloride concentration in groundwater on 5/03/87............. 127
6.48. Chloride concentration in groundwater on 5/05/87............ 128
6.49. Chloride concentration in groundwater on 5/08/87............. 128
6.50. Chloride concentration in groundwater on 5/13/87............. 129
6.51. Chloride concentration in groundwater on 5/18/87............ 129
6.52. Chloride concentration in groundwater on 5/25/87............. 130
6.53. Chloride concentration in groundwater on 6/01/87.............130
Total mass of atrazine stored in the saturated zone......... 133
Subareas used in water balance............................. 135
Comparison of percolation volumes predicted by GLEAMS and
PRZM ........................................................ 144
6.57. Catparison of measured and PRZM predicted atrazine
concentrations in the soil
6.58. Comparison of measured and
concentrations in the soil
6.59. Comparison of measured and
concentrations in the soil
6.60. OCaparison of measured and
concentrations in the soil
6.61. Comparison of measured and
concentrations in the soil
6.62. Comparison of measured and
concentrations in the soil
6.63. Ccnparison of measured and
concentrations in the soil
6.64. Ocparison of measured and
concentrations in the soil
6.65. COoparlson of measured and
concentrations in the soil
6.66. Comparison of measured and
concentrations in the soil
6.67. Comparison of measured and
concentrations in the soil
on 11/18/86 ...................... 145
PRZM predicted atrazine
PRZM predicted atrazine
PRZM predicted atrazine
PRZM predicted atrazine
PRZM predicted atrazine
PRZM predicted alachlor
on 11/18/86...................... 148
PRZM predicted alachlor
PRZM predicted aladchlor
PRZM predicted alachlor
on 2/09/87....................... 149
PRZM predicted alachlor
6.68. Oumparison of measured and IRZM predicted alachlor
ccentraticns in the soil on 5/25/87 ................. ...... 150
6.69. Measured and RZM predicted bromide concentrations in the
soil solution at a 61 ncm depth following the first
6.70. Measured and FPRZM predicted bromide concentrations in the
soil solution at a 122 ncm depth following the first
6.71. Measured and PRZM predicted bromide concentrations in the
soil solution at a 183 cm depth following the first
6.72. Measured and PRZM predicted bromide concentrations in the
soil solution at a 61 cman depth following the second
6.73. Measured and PRZM predicted bromide concentrations in the
soil solution at a 122 cm depth following the second
6.74. Measured and FRZM predicted bromide concentrations in the
soil solution at a 183 cm depth following the second
6.75. Measured and PRZM predicted concentrations of atrazine in
the soil solution at a 61 cm depth ..........................155
C. 1. Listing of program to calculate water and chemical
fluxes and storages........................................ .177
C.2. Program to calculate mass balance between sampling periods.. 194
D.I. Atrazine concentration in the groundwater on 1/19/87.........201
Abstract of Dissertation Presented to the Graduate Sdchool
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
MEASUREMET AND PREDICTION OF HERBICIDE
TRANSPORT INTO SHALLOW GRNOWATER
Matthew Clay Smith
Chairman: Dr. A. B. Bottcher
Major Department: Agricultural Engineering
A field study was conducted to observe the movement of the
herbicides atrazine and alachlor within the soil profile and a shallow
water table aquifer following a surface application. Surface-applied
bromide was used as a nonadsorbed tracer of water movement. The movement
of nitrate from a fertilizer application to the site was also monitored.
Measurements of chemical concentration were made using soil core samples,
samples of soil water fram the unsaturated zone, and water samples from
monitoring wells below the water table.
Atrazine was observed to move rapidly with both saturated and
unsaturated flows. Concentrations of atrazine exceeded 350 pq/L in soil
water samples at a depth of 61 can. Samples of shallow groundwater
contained atrazine residues as high as 90 jig/L. Measurable
concentrations of alachlor did not move below a depth of 45 ancm in the
Bromide and nitrate concentrations in soil water demonstrated large
variability between sampling locations. Nitrate concentrations in the
groundwater exceeded 40 mg/L after fertilization of the field.
Two root/vadose zone pesticide transport models, GLEAMS and
PRZM, were used to simulate the conditions on the site. Comparisons were
made between model simulation results and observed data. The
uncalibrated models predicted concentrations that were generally within
an order of magnitude, and often were within a factor of 2 to 3, of
observed values. Differences between the predictions of the models
appear to be due to the relative detail by which each model describes
soil properties as a function of depth.
Agricultural production systems have undergone many changes since
the 1940s. The capability of post-war industry to synthesize almost
unlimited varieties of chemicals in great quantities has provided
farmers with an arsenal with which to battle insects, weeds, and
disease. These chemicals can also provide plants with required
nutrients and regulate their growth. The result of using these chemical
tools along with improved machines for planting, cultivating, and
harvesting has been a dramatic increase in the productivity of the
However, scme serious problems associated with the use of agricul-
tural chemicals have been identified in recent years. Many of these
chemicals are toxic to non-target organisms such as birds, fish, and the
person who is applying them. Rachel Carson (1962) brought to the
attention of the public the potential harmful effects of using these
chemicals in large quantities over large land areas. Since the
publication of her book 'Silent Spring' there has been much research
conducted to evaluate the fate of agricultural chemicals in the
environment. IXDuring the 1960s and 1970s, great effort was initiated to
determine the quantities of soil, fertilizers and pesticides which were
entering our lakes and rivers. Through these efforts methods have been
developed (though only a few have been implemented) to minimize or
eliminate agricultural impacts on surface waters.
The discovery of abandoned waste sites such as the famous Love
Canal and the discovery of the pesticide aldicarb in the shallow
groundwater on long Island in New York generated concern for the
protection of groundwater. These concerns led to analyzing samples of
groundwater for toxic wastes and pesticides.
As more groundwater samples were analyzed, it became apparent that
agricultural chemicals had somehow entered groundwater aquifers in
several areas of the country. These results were publicized, and
citizens throughout the nation expressed alarm that they were drinking
potentially toxic chemicals. The federal and state governmnnts
responded to these concerns by intensifying monitoring efforts and
reviewing data on the many agricultural chemicals to determine potential
for leaching to groundwater axnd the extent of the current problem.
Several federal agencies, along with university and private industry
scientists, began developing computer models to help explain why these
chemicals were moving beyond their target locations and to screen
chemicals for their relative mobilities in soil. A limiting factor in
the development and utilization of these models was a lack of the
detailed data needed to test and validate model predictions.
To date, data on the presence of agricultural chemicals in
groundwater have been limited primarily to sampling of municipal and
domestic wells, providing only a large scale picture of contamination.
There have been only a few intensive monitoring studies completed which
provide sufficient data for model parameter estimation. Currently, there
are increased efforts at specific sites to provide more detailed
monitoring of chemical movement through the soil profile and within
The study described here is an effort to develop data for use in
the development and testing of models which describe the fate and
transport of pesticides and nutrients used in agriculture.
The overall objective of this research is to develop data on the
movement of two widely used herbicides through the crop root zone,
through the vadose zone (unsaturated zone between the crop root zone and
the top of the unconfined water table aquifer), and within the shallow
water table aquifer. These data will then be compared with the
predictions of two root/vadose zone pesticide transport models.
The specific objectives are to perform the following:
1. Instrument a field site for monitoring the movement of water and
chemicals through the soil profile and a shallow, water table
2. Observe the fate and transport of surface-applied herbicides
(atrazine and alachlor) over time.
3. Use surface-applied bromide as a nonadsorbed tracer of water
4. Observe transport of the nitrate ccuponent of fertilizers
applied to field.
5. Omrpile the observations of chemical movement into a data base
for use in testing pesticide transport models.
6. Simulate the field conditions present during the monitored
period with two pesticide transport models and compare
to observed results.
REVIEW OF THE LTERATURE
3.1 Evidence of Pesticide Residues in Groundwater
The U. S. Environmental Protection Agency estimates that at least 19
pesticides have been detected in groundwater in 24 states as a result of
agricultural practices (U.S. EPA, 1987a). The total number of
pesticides detected in groundwater is greater than nineteen. These
additional detections are the result of manufacturing, storage, and
loading activities which are not included in the above figures. The
number of reported cases of pesticides in groundwater is increasing.
Cohen et al. (1986) suggest that the increase is due to improvements in
the quality and quantity of studies instead of an increase in the
The insecticide/nematicide aldicarb was detected in the shallow
groundwater on Long Island, New York in 1979 (Wartenburg, 1988).
Aldicarb was used as an insecticide on potatoes. Aldicarb was detected
in 20 of 31 wells tested in 1978 and the presence was confirmed through a
repeat sampling of the wells in June, 1979. According to the U.S. EPA
(1987a), aldicarb residues have been detected in almost 2000 wells on
Long Island. Concentrations of aldicarb exceeded the state health
guideline of 7 pg/L (ppb) in nearly 50% of those wells. Concentrations
as high as 515 Ig/L were recorded in the Long Island wells as reported by
Ritter (1986). Aldicarb residues have been found in groundwater in 15
states (Cohen et al., 1986) with typical concentrations reported in the
range of 0.3 3 pg/L.
In the state of Florida, residues of the nematicide EDB have been
detected in over 10% of the public and private drinking water wells which
have been sampled (U.S. EPA, 1987a). Approximately 1,200 wells require
treatment or have been closed as sources of drinking water. Marti et al.
(1984) detected EDB in the groundwater in southwest Georgia. EDB
residues have been detected in groundwater in at least 6 other states.
In California, residues of approximately 57 pesticides have been
detected in groundwater (U.S. EPA, 1987a). Most of these detections
were related to factors other than routine use on agricultural fields.
The nematicide DBCP has been detected at levels exceeding 1 Vg/L in
approximately 60% of the 2500 drinking water wells tested.
Residues of the widely used herbicide atrazine are commonly
identified in groundwater. Ritter (1986) reports that approximately
36,000,000 kg of atrazine were used in the United States in 1982.
Atrazine has been detected in PA, IA, NE, WI, MN, and MD (U.S. EPA,
1987; COhen et al., 1986; Ritter, 1986). According to Berteau and Spath
(1986), atrazine has also been detected in concentrations of up to 2 iq/L
in California groundwater. Pionke et al. (1988) reported atrazine
concentrations in groundwater at levels up to 1.1 gq/L in an agricultural
watershed in Pennsylvania. The wells sampled by Pionke et al. (1988)
were all located in unconfined aquifers with depths to the water table
ranging from 2 to 21 m. Atrazine concentrations of 10 J/L have been
reported in a karst aquifer in Iowa (Libra et al., 1986; Kelley et al.,
Residues of the herbicide alachlor have been found in groundwater in
the states listed for atrazine, excluding WI. According to Ritter
(1986), alachlor arnd atrazine together accounted for 25% of all
pesticides sold in the U. S. in 1982. Ritter (1986) reported that in
1982, approximately 38,600,000 kg of alachlor were applied to cropland in
the United States. Kslley et al. (1986) reported aladclor concentrations
in groundwater as high as 16 gg/L in Iowa.
There are many individual cases of pesticide residue detection in
groundwater that could be cited. The reader is referred to the
proceedings of the Agricultural Impacts on Groundwater Conferences
(National Water Well Assoc., 1986, 1988) for a variety of reports
concerning the detection of pesticide residues in groundwater. A book
edited by Garner et al. (1986) is also a good source of information
related to this subject.
3.2 Factors Which Influence Pesticide Transport To Groundwater
Donigian and Rao (1986a) list five processes which affect the fate
and movement of chemicals within the soil. These processes are
transport, sorption, transformation/degradation, volatilization, and
plant uptake. Other authors, e.g., Cheng and Koskinen (1986), use
slightly varied groupings which include the processes named above. The
interactions of these processes over time and space determine the fate of
chemicals in the soil. The spatial variability of soil properties and
other factors is an important consideration when interpreting field data
and model predictions (Donigian and Rao, 1986a). This section will
describe each of the processes and same of the factors which influence or
The chemicals can be transported in at least three different phases:
adsorbed to solid materials, in solution, and as a vapor. Surface runoff
can transport chemicals in both the adsorbed and solution phases.
Erosion and transport of soil particles during rainfall and irrigation
can result in significant transport of highly-adsorbed, low-solubility
pesticides (Waudcope, 1978). For most pesticides, the majority of
transport in runoff is in the solution phase (Wauchqpe, 1978; Rao and
Davidson, 1980). Chemicals in the solution phase can be transported
within the soil profile by saturated and unsaturated water flows.
Water percolating through the soil profile can transport chemicals
in the solution phase beneath the root zone and possibly into
groundwater. The most common methods used to describe the transport of
water and solutes in porous media are based on the assumptions that
Darcian flow conditions exist, and that solute movement is controlled by
advective and dispersive processes (Freeze and Cherry, 1979; Jury, 1986a;
Wagenet, 1986; Wang and Anderson, 1982). Darcian flow conditions imply
that the flow equations are representative of some volume of soil. All
soil properties such as pore size distribution and volumetric water
content are assumed to be uniform throughout the volume of soil (Jury,
The advective component of transport is usually based upon Richards'
equation and describes the average rate of water flow. The average rate
of solute flow is the product of the solute concentration and the average
pore-water velocity. Additional transport of solute can occur due to
mechanical mixing of water in adjoining pores during advective transport
and molecular diffusion of solute from pores with high concentrations to
adjoinirng pores with lower concentrations (Freeze and Cherry, 1979).
Molecular diffusion can be significant when average pore-water
velocities are low; otherwise, the processes involved in mechanical
dispersion usually dominate. Differences in pore-water velocities in the
direction of bulk flow cause a spreading out, and consequently a lowering
of peak concentrations, of the solute plume. Some of the solute will
arrive at a reference point earlier, and some will arrive later, than
would be predicted based upon the average linear flow velocity. This is
referred to as longitudinal dispersion. The tortuosity and
interconnection of pores in the soil will also cause the solute to move
perpendicularly to the bulk flow direction. This is referred to as
transverse dispersion. Freeze and Cherry (1979) and Rao et al. (1988)
provide detailed discussions of the dispersive-diffusive processes and
their mathematical formulations.
Transport predictions based upon the methods described above have
not caqpared well with field observations, e.g. Jury et al. (1986).
Numerous modifications to the general theory have been proposed in order
to help explain and reduce the differences between field observations and
theoretical predictions. Rao et al. (1980) proposed separating pore-
water in soils into two regions: inter-aggregate and intra-aggregate.
Convective-dispersive transport was assumed to be limited to the water in
the inter-aggregate region. The intra-aggregate region was assumed to
act as a source/sink of solute. Diffusion between the regions could
remove or add solute to the water in the inter-aggregate region. Jury
(1982) proposed a method of predicting the transport times of solutes to
specified depths which is not based on physical processes. Rather, the
transfer function theory uses measured values to develop a probability
density function of travel times and uses it to predict solute travel
times to depths greater than those that were measured. Gish (1987)
combined the ocnvective-diffusive transport theory with a stochastic
representation of water velocities to describe field-scale (average)
concentrations of bromide at various depths in the soil profile.
Camparisons between the average measured bromide concentrations and
predicted values showed good agreement.
Volatilization from soil and plant surfaces transports the chemicals
into the atmosphere and reduces the amount remaining for transport with
runoff or percolation. Diffusion of chemicals in the vapor phase can
also transport the chemical within the soil profile. Most transport
models do not consider vapor diffusion.
Many of the factors which are assumed to influence the transport of
chemicals through porous media are summarized by Jury (1986b). Some of
the soil properties influencing transport are water content, bulk
density, permeability, clay content, organic matter content, and water
retention (field capacity). Helling and Gish (1986) present results of a
simple modeling exercise to demonstrate the effect of several soil
properties on pesticide transport. Some of the environmental factors
which have been shown to influence transport are precipitation,
evapotranspiration, and temperature (Jury, 1986b).
The partitioning of solutes between the liquid and solid phases
(dissolved and adsorbed) is a major factor determining the mass of
solute available for advective-dispersive transport through the soil
profile. If the chemical of interest is strongly adsorbed, then only a
small fraction will exist in the solution phase at a given time and be
available for transport. Non-adsorbed chemicals, however, will exist
entirely in the solution phase and are available for transport with
Many pesticides are nonpolar and adsorption occurs primarily on
organic matter surfaces. Pesticides and chemicals which are polar will
adsorb primarily to clay surfaces (Jury, 1986c). Since the majority of
pesticides are nonpolar, the adsorption onto soils has been primarily
related to the organic matter or organic carbon content of the soil. A
detailed discussion of the factors influencing pesticide adsorption onto
soils is given by Jury (1986c). The relationship between concentrations
of a pesticide in the dissolved and sorbed phases is often represented by
the Freundlich equation:
where S = adsorbed concentration (pq/g of soil), C = solution
concentration (Ag/mL), and K and n are empirical constants for the soil-
pesticide system. The exponent n in equation 3.1 is often assumed to be
equal to 1.0 which results in a linear relationship of the form:
S = KC 3.2
where Kd = partition coefficient (miL/g of soil).
For most uses of equation 3.2 in transport modeling, the
partitioning between the solution and adsorbed phases is assumed to be
instantaneous and reversible. Errors associated with the assumptions of
linearity, instantaneos equilibrium, and reversibility are discussed by
Rao and Davidscn (1980).
The partition coefficient, Kd, is unique to a given pesticide-soil
combination. However, Rao and Davidson (1980) report that investigators
have shown that when Kd is normalized for the organic carbon content of
the soil, the resulting value is independent of the soil type and can be
considered a property of the pesticide. The normalized partition
coefficient is represented as Koc and is defined by:
Kc = Kd 100 / %OC 3.3
where %OC is the percent organic carbon content of the soil and Kd is the
measured partition coefficient. Methods for measurement of the
partition coefficient are reviewed by Rao and Davidson (1980).
Volatilization of pesticides from the soil and plant surfaces
reduces the amount of the pesticide available to be leached or
transported with runoff. Volatilization also determines the quantity of
the pesticide which exists in the vapor phase within the soil and thus
the amount available for diffusive vapor transport. Thus, as noted by
Donigian and Rao (1986a), volatilization affects both the fate and
transport of a pesticide.
Same of the factors which influence volatilization are summarized by
Jury and Valentine (1986). They are Henry's constant KH, chemical
concentration, adsorption site density, temperature, water content, wind
speed, and water evaporation. Henry's constant, KH is the ratio of
saturated vapor density to solubility and is an index of the partitioning
between the vapor and solution phases. A larger KH implies increased
volatilization. Increasing concentrations of the pesticide will increase
the volatilization as long as the vapor density is not saturated.
Adsorption of the pesticide will reduce volatilization. Volatilization
increases as temperature increases. As the soil water content decreases,
the rate of volatilization increases. Increasing wind speed can increase
the volatilization of pesticides, particularly those with low KH values.
Evaporation of water from the soil surface can transport pesticides from
within the soil profile to the surface where temperature and wind
effects can increase the volatilization. Each of these factors is
discussed in more detail by Jury (1986d).
Transformation of a pesticide is the change in structure or
composition of the original compound and degradation is the breakdown of
the compound into smaller fragments with eventual inorganic endproducts
such as H20 and CO2 (Cheng and Koskinen, 1986). The transformation and
degradation processes represent a loss of the original (or parent)
compound, thus reducing the amount remaining in the soil for transport by
surface runoff and percolation.
In general, transformation and degradation processes occur at faster
rates on plant surfaces and in the top few centimeters of soil than in
the deeper soil zones (Donigian and Rao, 1986a). Thus, a chemical that
is foliar or surface-applied will likely degrade more rapidly than if it
were incorporated. For non-persistent pesticides (half-lives less than
15-20 days), the timing of rainfall events or irrigation is important in
determining the fraction of the applied mass that will be available for
transport. Events occurring shortly after application will likely result
in the largest concentrations in runoff and percolation water (Wauchope,
1979; Donigian and Rao, 1986a). The timing of rainfall and irrigation
events is less critical for persistent (half-lives in excess of 100 days)
pesticides since they will reside in the soil for longer periods
(Donigian and Rao, 1986a).
The major processes involved in transformation and degradation
include biotransformation, chemical hydrolysis, photolysis, and
oxidation-reduction (Donigian and Rao, 1986a). Factors influencing
transformation and degradation of pesticides, as summarized by Jury and
Valentine (1986), include: microbial populations, chemical concentration,
temperature, oxygen, pH, soil water content, and light. Detailed
descriptions of the many factors influencing these processes are
presented by Rao and Davidson (1980), Valentine and Schnoor (1986), and
3.2.5 Plant processes
Plant processes relating to pesticides are very complex. The
uptake, translocation, accumulation, and transformation of pesticides by
plants affect the availability of pesticides for transport processes and
the potential exposure to pesticide residues by the consumers of the
vegetation, fruits, etc. (Donigian and Rao, 1986a). Plant processes can
serve as both a sink and a source of pesticide residues available for
Pesticides applied to plant foliage will likely be transformed or
degraded more rapidly than if the pesticide were applied to the soil or
incorporated. Pesticides on the foliage may also be absorbed into the
plant. Pesticides in the solution phase may be taken up by plant roots
and translocated to various parts of the plant. These processes serve as
a loss mechanism reducing the availability of the pesticide for
Pesticide residues on plant foliage can be dislodged by rainfall and
irrigation and washed onto the soil surface, resulting in an increased
mass of pesticide available for transport in runoff and percolation
(Smith et al., 1981; Donigian and Rao, 1986a). Plant residues which are
left on the soil surface or incorporated can also serve as a source of
pesticide which may be available for transport.
3.3 Predicting Pesticide Transport
There are many methods that can be used to predict the mobility of a
given pesticide. These range fram simple indices to complex research
models. Each method has value when used for the purpose for which it was
designed and with recognition of the assumptions and limitations
associated with it.
3.3.1 Indices of contamination potential
Rao et al. (1985) compared several methods for computing indices of
the contamination potential of pesticides. Most of the indices are based
on one or more chemical properties such as Koc and solubility. Three
indices also included distance to groundwater and recharge rate. The
attenuation factor (AF) proposed by Rao et al. (1985) can also be used to
estimate the mass of pesticide which will leach below the root zone or
out of the vadose zone and into groundwater. The AF indr1ex is the only
one of the irxndices compared by Rao et al. (1985) which incorporated mass
Dean et al. (1984) developed a methodology in which soil
characteristics, crop type, chemical properties, and management practices
(e.g. tillage type) can be combined to find a cumulative frequency
distribution of the percentage of applied pesticide mass leadching below
the crop root zone. These frequency distributions were derived from
hundreds of 25-year simulations of pesticide leaching using the FpRZM
model (Carsel et al., 1984). The methodology was applied to four crop
types (corn, soybean, cotton, and wheat) and 19 representative growing
regions. The results are somewhat unique in that the user is presented
with a statistical probability of leaching based on a 25-year
Another index which is currently receiving much attention is the
DRASFTIC index developed by Aller et al. (1985). This irxndex does not
consider chemical properties; instead, it is an index of the
vulnerability of groundwater at a given location to contamination from
surface-applied chemicals. The DRASTIC index assigns numerical scores or
weights to seven factors which could influence pollution potential. The
factors are depth to water, net recharge, aquifer media, soil media,
topography, impact of vadose zone, and conductivity of the aquifer. The
final DRASTIC score is used to describe an area as having high, medium,
or low susceptibility to groundwater pollution.
ERASTIC has been used to map every county in the United States
(Alexander and Little, 1986) as part of the first stage of a national
survey of pesticide contamination of drinking water (U.S. EPA, 1987b).
In the first county-level assessment the Southeast Coastal Plain, which
covers all of Florida and southern Georgia, was shown to be the
groundwater region of highest vulnerability.
Indices like those mentioned above are beneficial to regulatory
agencies as a method to screen great numbers of compounds and locate
vulnerable areas. This information can aid in allocating funding for
further studies concentrating on the chemicals of interest in regions
which may be most vulnerable. These studies may involve detailed
modeling of pesticide transport to groundwater, modeling of surface and
groundwater hydrology, and collection of data from test locations.
3.3.2 Transport models
There are many models of pesticide transport reported in the
literature. These range from steady-state screening models, e.g. PESTAN
(Enfield et al., 1982), to very complex process-oriented research models,
e.g. LEACHMP (Wagenet and Hutson, 1986). Donigian and Rao (1986a), and
Shoemaker and Magette (1987) review a number of the models which are
available. Most models which are capable of simulating pesticide fate
and transport do not explicitly represent agricultural management and
cropping practice effects on runoff and leaching. Modifications of
management practices will be one method of reducing pesticide leaching to
groundwater. The models used for regulatory purposes should reflect the
trade-offs between surface water and groundwater quality. There is scme
evidence that no-till farming will increase pesticide transport to
groundwater while reducing surface runoff and thereby reduce pollution
of surface waters (Dick et al., 1986). Thus, solutions to one problem
may exacerbate another problem. The models discussed below are capable
of simulating, to sane degree, agricultural management practice effects
on pesticide fate and transport.
According to Donigian and Rao (1986a), SESOIL (The Seasonal Soil
Caqpartment model) is designed for long-term simulation of pesticide fate
in the soil environment. It was developed for the EPA and is used as a
screening model. SESOIL can simulate many ccnponents of the hydrologic
cycle including precipitation, evapotranspiration (ET) and surface
runoff. The model considers transport within the unsaturated zone
extending from the soil surface to the top of the saturated zone. The
hydrologic responses are determined using physically based equations in
which uncertainty has been included. The water balance used in the
model is a statistical representation of the hydrologic cCnponents over a
"season." A season is the time step of the model, e.g., month or year.
Erosion is simulated using the Universal Soil Loss Equation (Wischmeier
and Smith, 1978). Hetrick and Travis (1988) coupled SESOIL with EROS (a
submodel of the CREAMS (Knisel, 1980) watershed model) to predict surface
runoff and sediment yield from small watersheds. The pesticide ccanponent
of the model considers most of the processes and factors influencing
pesticide transport which were described in previous sections.
SESOIL requires calibration and represents long-term averages. Some
of the results of use and testing of SESOIL reported by Donigian and Rao
(1986a) indicate that it should not be used for short-term predictions.
The combined SESOI/EROS model of Hetrick and Travis (1988) adequately
predicted long-term (several months) average runoff and sediment yields
when tested against data from three small watersheds. However, monthly
predictions were not good, especially when most of the runoff occurred
due to large single storm events.
MUSE (Method Of Underground Solute Evaluation) was developed at
Cornell University as a management model and a training tool for students
and professionals (Steenhuis et al., 1987). MNOSE is an interactive,
menu-driven program which can run on a IBM-PC or compatible computer.
The model can read daily historical weather files or generate synthetic
rainfall, air, and soil temperature patterns. Surface runoff is computed
using the SCS curve number method. Erosion is not considered. Solute
movement in the vadose zone considers advective and dispersive flux, as
well as degradation and adsorption. MOUSE will also simulate the
movement of a pesticide in a two-dimensional unconfined aquifer.
Extensive graphics illustrate movement as it is being simulated.
PRZM (Pesticide Root Zone Model) is a field-scale hydrology and
transport model developed by the EPA (Carsel et al., 1984). PRZM is a
continuous model capable of simulating water and chemical fluxes over
many years of historical daily weather records. Runoff is predicted
based on the SCS curve number equation, and erosion is simulated using a
modification of the Universal Soil Loss Equation for daily time steps.
The model can simulate the entire vadose zone (soil surface to
groundwater). The vadose zone can be characterized by several layers
with varying properties. For calculations, the vadose zone is divided
into many cmpartments of equal depth. The model simulates crop growth
(leaf area and rooting depth). Water in the root zone can be removed by
percolation, evaporation, or transpiration by the plant. Percolation is
calculated based on the water-holding capacity of the soil. When the
water content in a layer exceeds field capacity (1/10 1/3 bar tension)
the excess water drains into the next lower cnpartment. There is an
option in PRZM that allows the draining of the profile to ocur over a
longer period than one day.
Pesticide processes represented include advective and dispersive
flux, sorption, degradation in soil and on plant foliage, and plant
uptake. Volatilization and transport in the vapor phase are not
considered. Application of pesticides can be partitioned between the
soil surface and plant foliage. Applications to the soil can be
incorporated by tillage. Different degradation rates can be specified
for soil and foliar pesticide residues. The degradation rate within the
soil can also be varied by soil layer.
PRZM will simulate multiple applications of one pesticide each year
for many years of continuous climatic record. Thus it can predict
temporal variations in leaching and runoff. Effects of agricultural
management and cropping practices can be simulated. PRZM allows the user
to request time series results for many variables in the model. This
feature is useful for observing processes at intermediate locations
within the vadose zone over time.
GLEAMS (Groundwater Loading Effects of Agricultural Management
Systems) is also a field-scale hydrology and transport model (Leonard et
al., 1987). It is based upon the extensively documented and applied
CREAMS model (Knisel, 1980). CREAMS is a nonpoint source model for
predicting sediment, nutrient, and pesticide losses with surface runoff
from agricultural management systems. GIEAMS builds upon the foundations
in CREAMS by adding components to simulate movement of water and
chemicals within the crop root zone. Like CREAMS, GLEAMS is a
continuous, daily simulation model.
GLEAMS predicts runoff using the SCS curve number method. Erosion
is predicted based upon modifications to the Universal Soil Loss
Equation. Surface runoff and eroded sediment, with chemicals in both the
dissolved and sorbed phases, can be routed overland, in channels, and
GLEAMS divides the crop root zone into seven layers. The vadose
zone between the root zone and the water table is not considered. The
first layer has a thickness of 1 an. This layer is assumed to be the
portion of the soil that determines the mass of pesticide available for
extraction into surface runoff. The thickness of the second layer is
1/6 of the root zone depth minus 1 cm for the first layer. The remaining
5 layers are each 1/6 of the root zone in thickness. This layering
structure is fixed by the program. A strong correlation has been
observed between pesticide concentrations in runoff and the concentration
of pesticides in the top 1 cm of the soil profile (leonard, 1988). Thus
the authors of CRFAMS and GLEAMS believe that a 1 cm active zone at the
soil surface is required to maintain sensitivity of runoff concentrations
to soil concentrations (leonard and Knisel, 1987). Soil profile
properties can vary with depth. The model weights the input values to
establish average properties for each layer in the model. Percolation
through the profile is based on the water-holding capacity of the soil as
in PRZM. Water in excess of field capacity drains to the next lower
Pesticide transport within the root zone is by advection. No
dispersive flux ccponents are included. Volatilization is not
considered. Pesticide applications can be partitioned between plant
foliage and the soil surface. Different degradation rates can be
specified for chemicals on the foliage and within the soil. Soil-applied
chemicals can be incorporated to a specified depth.
The model will simulate up to ten chemicals simultaneously. This
feature makes possible the observation of the effect of changes in
chemical properties, e.g. partition coefficient, on resulting leaching
losses with one model run. The model can also simulate the formation,
fate, and transport of degradation products of the parent chemicals.
3.3.3 Coupled saturated/unsaturated zone transport models
The root/vadose zone models described above are important tools for
assessing the mobility of pesticides within the unsaturated zone. The
concern expressed by the public and governmental officials relate to
pesticide residues in grourndkwater supplies used for drinking water. Thus
the unsaturated zone models need to be linked to saturated zone transport
models so that the concentrations and transport of pesticides within
aquifers can be calculated.
As described above, MNUSE (Steenhuis et al., 1987) can predict the
transport of pesticides in a vertical, two-dimensional cross section of
a water table aquifer. Jones et al. (1987) coupled the PRZM model to a
two-dimensional saturated zone transport model to predict the movement of
aldicarb residues within a shallow water table aquifer. Dean and Carsel
(1988) reported on progress in linking PRZM to a two-dimensional
saturated transport model. The linked model will use PRZM for simulating
the root zone, a one-dimensional transport model for the vadose zone, and
a two-dimensional saturated zone model which can simulate confined,
unconfined, and leaky confined aquifers. The saturated zone model will
also simulate pumping from the aquifer(s). The linked model is expected
to be released for testing during the fall of 1988.
3.3.4 Model testing and uncertainty
Methods (such as described above) to predict the transport of the
chemicals used by agriculture are needed by regulators in order to assess
the impacts on surface water and groundwater quality from the use of
these chemicals. Field testing of the contamination potential of all
chemicals registered for use by agriculture would be infeasible due to
both cost and time constraints. Since decisions regarding chemical use
will be made based upon the results of transport models, it is important
to understand the errors and uncertainty associated with the models.
Each model or method will have errors associated with the assumptions
relating to the description of the system, choice of equations to
represent the system, and input parameters required by the models. There
are some general points that can be made regarding the uncertainties
inherent in deterministic, physically based models such as PWZM and
GIEAMS which were described previously. An excellent discussion of the
uncertainty in transport models is presented by leonard and Knisel
Limitations of the conceptual formulation of the system that is
described by a model lead to uncertainty in the results. Most
deterministic models assume that same of the physical properties of the
system are uniform with respect to location. That is, soil properties
such as porosity, water-holding capacity, and organic matter content are
constant across the field, although most models will allow these types of
properties to vary with depth in the soil profile. Furthermore,
variations in pore sizes within the soil, which can cause uneven water
velocity distributions, are usually not considered. Each model will
represent a system (ie. the crop root zone) in varying levels of detail.
The equations which are used to represent the fate and transport
processes have limitations and uncertainties. For example, PRZM and
GLEAMS both use the SCS curve number method for partitioning rainfall
between runoff and infiltration. This method utilizes daily rainfall
records and does not consider the effects of rainfall intensity on the
runoff process. Testing of the CREAMS model (Knisel, 1980), frcm which
GLEAMS was derived, showed that the daily hydrology option represented
average annual runoff volumes well, but did not do as well with daily and
monthly runoff volumes (Smith and Williams, 1980). PRZM and GLEAMS both
utilize modified versions of the USLE (Wischmeier and Smith, 1978) for
prediction of daily soil erosion. The USLE was developed for long-term
(20-yr) average annual erosion rates. Although these modifications have
been tested, the basis of these methods is the USLE which is an empirical
equation derived for long-term average predictions. Many transport
models utilize the linear form of the Freundlich equation (Equation 3.1)
to describe the partitioning of a chemical between the dissolved and
adsorbed phases. Rao and Davidson (1980) indicate that the assumption of
linearity can lead to errors on the order of a factor of two to three.
The previous examples illustrate that the mathematical representations of
processes used in models often have inherent inaccuracies and
The parameters which are required by the models to solve the
equations which describe the system also have inherent uncertainties
associated with them. Jury, (1985), in a review of the literature,
reported coefficients of variation (CV) from field measurements of
several soil physical properties. Soil porosity was found to have a mean
CV of 10%. Bulk density measurement had a CV of 9%. Determinations of
particle size fractions had an average CV of 28%, with a range of 3 to
55%. The water-holding capacity of the soil at matric potentials of 0.1
and 15 bars had a CV of 15 and 25%, respectively. Measurements of
saturated hydraulic conductivity and infiltration rates showed a CV of
124 and 72%, respectively. In a study by Jury et al. (1986), the
partition coefficient, Kd, of napropamide measured on 36 samples from a
1.44-ha field had a CV of 31%. Rao and Davidson (1980) reported that the
normalized partition coefficient, Koc, reported from individual studies
generally had a CV of 40-60%. The variability of the properties reported
above are not necessarily due to the fact the samples were taken over
large areas. Jury (1985) reported that one author had reported that up
to 50% of the variation in a parameter over a field occurred within a 1
Given the uncertainties associated with models and the parameters
required by the models as outlined briefly above, how closely must a
model match field observations in order to be considered validated
(determined to accurately represent the system)? Leonard and Knisel
(1988) indicate that there are no standard criteria for model validation.
Since the spatial distribution of variations in soil properties is
usually unknown, it would be unreasonable to expect any model to exactly
match point measurements from a field study. Hedden (1986) reported that
participants of the Predictive Exposure Assessment Workshop sponsored by
the U. S. EPA in Atlanta, GA, on April 27-29, 1982, agreed on two
criteria for model acceptance. For screening applications of the model
(limited site-specific data and not calibrated to previous data from the
site), the model should be able to replicate measured field data within
an order of magnitude. For site-specific applications (parameters
measured on-site and with the model calibrated to the site), the model
should be able to match the field observations within a factor of two.
The screening level criteria seems to be quite reasonable when all of the
sources of error and uncertainty are considered. The site-specific
criteria, however, may be difficult to meet by even the best models using
carefully measured site-specific parameters.
The fact that models can not be expected to exactly replicate field
measured values does not mean that they have no value. Models that have
been verified (ie. their response to changes in parameters reflect what
would be expected based on knowledge of the system) can be used with some
confidence to evaluate the differences in predicted leaching of different
chemicals or, perhaps, differences in leaching of a single chemical due
to different climatic or management conditions.
3.4 Field Studies of Pesticide Transport
Field-scale studies of pesticide runoff and leaching have, until
recently, been very limited. Studies of sufficient detail to be used for
the validation and testing of transport models have been limited due to
three factors, according to Donigian and Rao (1986b):
1. Most field studies have focused on either chemical leaching or
surface runoff from watersheds; few studies have examined both.
2. Since sampling and analysis costs associated with pesticide
leaching are high, most studies are of short duration.
3. Most studies were designed for purposes other than developing
data sets for model testing and therefore are incomplete with
regard to measurement of all the required model input parameters.
In this section, four field studies are reviewed. Methods of
sampling and monitoring will be emphasized. Some of the studies were
conducted specifically for model validation and testing purposes. A
review of the Big Spring Basin study is included due to the unique
conditions in the study area.
3.4.1 Aldicarb in Florida citrus Qroves
A three-year field study of the movement and degradation of
aldicarb in both the unsaturated zone and a shallow water table was
conducted at two locations in Florida (Hornsby et al., 1983; Jones et
al., 1987). The two locations are referred to as the Oviedo site and the
Lake Hamilton site. At both locations, aldicarb was applied to bedded
citrus trees located on coarse-textured soils. Scme of the soils at the
Oviedo site had a thick organic layer overlying the coarse textured
subsoil. The Oviedo site is in the flatwoods area and is poorly drained.
The Lake Hamilton site was located on a sand ridge. Drainage of the
upper soil layers was very rapid. The Oviedo site encrmpassed 3.6 ha of
treated area and the other site encmpassed approximately 1.7 ha.
Aldicarb was banded along each side of the trees and incorporated to
a depth of 5 cm. At the Oviedo site, soil samples were collected in 30
cm increments to a depth of 150 cm. At the lake Hamilton site, soil
samples were collected in 30 cm increments to a depth of 60 cm, and then
in 60 am increments to a depth of 300 cman. Soil samples were collected
using a bucket auger and were transferred into plastic bags for storage
at -20C until analyzed. The soil samples were collected monthly for six
months following the first application of aldicarb in 1983. Urndisturbed
soil cores and bulk soil samples were also collected from untreated areas
at each site for laboratory determination of soil and chemical
At each site, two clusters of observation wells were installed
within the treated area, and additional well clusters were located
upgradient and downgradient of the treated area. The wells were screened
at depths of 2, 3, 5, and 6 m at the Oviedo site, and at depths ranging
from 5 to 15 m, depending on location, at the Lake Hamilton site. Water
samples were usually collected with a peristaltic pump. At least 5 well
volumes were pumped prior to collecting a sample for analysis. When
flowrates into the well were too slow to permit pumping of 5 well
volumes, the wells were evacuated and the water entering the wells was
pumped for an additional 5 minutes prior to collection of the sample.
The groundwater temperature, pH, and conductivity were measured during
sampling. Water samples were collected monthly throughout the study.
The storage conditions prior to analysis were not specified.
Results from the first 6 months of the study were reported by
Hornsby et al. (1983). Results are reported in terms of total toxic
residues (TTR), which is the sum of residues of aldicarb, aldicarb
sulfoxide, and aldicarb sulfone. Fifteen days after application, TTR
were not detected below 30 ncm at the Oviedo site, but were detected in
all replicates at the Lake Hamilton site to a depth of 120 cm. After 120
days, TMR were detected at the deepest sampling depths (150 cm at Oviedo,
and 300 can at Lake Hamilton) at both sites. No TmR were observed at the
Oviedo site in any of the observation wells during the first six months.
Two wells within the treated area at the Lake Hamilton site showed
measurable levels of TMR. In one well, TmR levels as high as 1.2 mg/L
were measured approximately 130 days after application. Data relating to
rainfall or irrigation during the study were not presented.
Jones et al. (1983) used the data franom the aldicarb study above to
evaluate three pesticide transport models. One of the models tested was
PRZM (Carsel et al., 1984). Agreement between PRZM predicted movement of
aldicarb and the observed data was good. A sensitivity analysis was
performed to demonstrate the response of FPRZM to variations in several
soil and pesticide properties. Jones et al. (1987) presented results
from the Lake Hamilton site in which water samples from the shallow water
table were collected franom a network of 174 wells over a period of three
years. These data were used to evaluate predictions from a linkage of
the PRZM model to a one- or two-dimensional saturated transport model.
The linked model was developed to allow predictions of the extent of
lateral transport of TIR which have entered a shallow water table. FPRZM
was used to predict daily mass loadings of aldicarb to the top of the
water table. Results of the saturated zone modeling were in reasonably
good agreement with the observed concentrations in the wells and the rate
of lateral movement.
3.4.2 COmparison of tillagce effects in Maryland
Paired watersheds were established in 1984 on the Eastern Shore of
Maryland to observe the impacts of tillage practices on agricultural
chemical leaching and runoff. The experimental design and early results
have been reported by Brinsfield et al. (1987, 1988). One of the
objectives of the experiments was to generate a data set for testing and
validation of pesticide transport models. One watershed was used to grow
corn using conventional tillage (CT) practices, and on the other
watershed corn was grown using no-till (NT) methods. The CT watershed is
approximately 6 ha in size, and the NT watershed covers approximately 8.9
ha. The soils in both watersheds are silty, well-drained, and nearly
Soil samples were collected at 5 locations in each watershed 4 times
per year. The samples were collected using a hand auger in 15 cm
increments to a depth of 120 m.
Monitoring wells made of 6.35 cm diameter PVC were installed to
depths of 3-4 m on 30 m centers in each field. The wells were arranged
to divide each field into four quadrants. The wells were installed so
that only the top of the water table would be sampled. A bentonite-soil
mixture was used to seal the top 0.5 m of each well. Well samples were
collected monthly. The wells were pumped for 5 minutes prior to
collecting a sample for analysis.
To monitor surface runoff, H-flumes with automatic samplers and
recorders were installed in the outlets of the watersheds. All samples
were frozen for transport and storage prior to analysis. Details of the
analytical procedures used for pesticide residue analysis are presented
by Brinsfield et al. (1987).
Gravity-fed lysimeters were installed horizontally through the walls
of a reinforced pit to collect samples of percolating water. The
lysimeters were installed at a depth of 1 m below the soil surface.
Three lysimeters were installed in each pit. These lysimeters collect
samples of saturated flow percolating through the root zone. Samples
were collected from these lysimeters after each rainfall event.
Herbicide treatments were the same for both watersheds. Atrazine,
metolachlor, and simazine were applied pre-emergence at the rate of 1.68
kg AI (active ingredient)/ha. Cyanazine was applied at 2.24 kg AI/ha,
and carbofuran at 1.12 kg AI/ha, both pre-emergence. Dicamba was
applied post-emergence at a rate of 0.55 kg AI/ha.
leachate collected in the lysimeters during January, 1985, had
atrazine concentrations as high as 2 Vg/L. After planting in 1985, no
leachate samples were collected in the CT watershed until September.
Atrazine and metolachlor concentrations in those samples ranged from 1-2
gg/L. Simazine was detected in only one sample, and cyanazine was not
detected. In the NT watershed, samples collected in July had
concentrations of atrazine at 8-10 &g/L, simazine at 7-10 jg/L, cyanazine
at less than 1 jig/L, and metolachlor at less than 2 jg/L. Leachate from
the NT watershed in the fall of 1985 had similar concentrations to those
collected from the CT watershed.
Groundwater samples collected 15 days after application in 1984
showed levels of atrazine, simazine, and cyanazine exceeding 1 Jg/L. The
concentration of atrazine in one well reached 7 A/L on this date.
During late summer and fall, atrazine was the only pesticide detected in
the wells. Pesticide concentrations in the groundwater were similar for
the two watersheds, although the concentrations were somewhat higher
during the growing season in the NT watershed. In April, 1986, (prior to
application) atrazine concentrations of 4 mg/L were observed in the water
table on both watersheds. This indicates that atrazine may be persistent
in the lower soil zones and in the water table.
3.4.3 Aldicarb movement in the Dougherty Plain of Georgia
The U. S. Environental Protection Agency and the U. S. Geological
Survey began a study in 1982 to observe the movement of aldicarb and its
residues in the unsaturated and saturated zones at a field site in Lee
County, in southwestern Georgia. Aldicarb is used as an insecticide and
nematicide on peanuts grown on the site. One of the primary objectives
was to develop an extensive database of soil, land use, geologic, and
pesticide data for use in the validation and testing of the PRZM (Carsel
et al., 1984) model. Different aspects of this study have been reported
by a number of investigators (e.g. Cooper, 1986; Rao et al., 1986; Hook,
1987; Hedden, 1986; Smith and Carsel, 1986).
The study site covers an area of approximately 4.5 ha. Four soils
are mapped in the field. All have a fine-loamy texture. Below a depth
of approximately 1 m there are layers of clay separating zones of coarser
sand and gravel. The conductivity of the clay layers is very low, which
suggests that perched water tables may form on these layers and cause
lateral transport of leached pesticides. There is a shallow water table
at the site which can vary in position by as much a 6 m during a year.
Soil samples were collected from the site to characterize soil
properties and pesticide sorption and decay characteristics. Geologic
information was collected from holes bored to the top of the Ocala
Limestone. The residuum was approximately 13 m thick. The Ocala
Limestone is a part of the Floridan aquifer system. Four wells were
cased into the Floridan aquifer for measurement of water levels.
A weather station was installed at the site to collect meteorological
Statistical procedures were used to determine that 20 monitoring
sites would accurately reflect the fate and transport of aldicarb at this
site. The locations of the monitoring sites were randomly selected and
distributed between the three major soil series based upon the relative
area in each series.
Each monitoring site was equipped with an impressive array of
instruments for monitoring of conditions anid collecting samples. Five
tensiometers were installed to a depth of 150 ancm for monitoring soil
water content. The tensicmeters were monitored and serviced three times
per week. Five thermistors were installed to a depth of 115 cm for
measurement of soil temperatures. The thermistors were monitored three
times per week during the middle of the afternoon in order to estimate
maximum soil temperatures. Three soil solution samplers were installed
at depths of 1.5, 2.1, and 2.7 m. The samplers were made using stainless
steel for the body, a high-flow ceramic cup, and teflon tubing for sample
collection and applying the vacuum. A silica flour slurry was poured
around each sampler to insure good contact with the surrounding soil.
The vacuum and sample lines were buried to avoid interference during
field operations. Vacuum was applied to the samplers for a period of 24
hours prior to collection of the sample.
At 15 of the monitoring sites, permanent stainless steel monitoring
wells were drilled to a depth of approximately 4.6 m. A stainless steel
screen 0.6 m in length was placed at the bottom of each well. A gravel
pack was placed around the screen and a cement grout was placed from the
gravel to the top of the finished well. The wells were finished so that
the top of the well was approximately 0.6 m below the soil surface so as
not to interfere with cropping operations. After field operations are
complete, the wells are extended above the soil surface. Water samples
are collected using a dedicated teflon tube (one per well) and a
peristaltic pump. The 250 mL glass sample containers were connected in
line between the well and the pump to prevent cross contamination. The
wells were typically pumped for 5-7 minutes prior to collecting a sample.
Soil samples were collected using a hand auger in 15 cm incremnts
to a depth of 120 ancm. After each sample was collected, gravel and plant
material was removed and then the sample was sealed in a metal container.
The references which were obtained for this review did not report
any results of pesticide monitoring from the field site. The detailed
monitoring network and experimental design used at this site should
produce a valuable data set for model validation and testing.
3.4.4 Big Sprind Basin in Iowa.
The Big Spring Basin is a 267 km2 groundwater basin located in
northeastern Iowa. The basin is entirely agricultural with
approximately 60% of the land area in row crops. The basin has been
extensively monitored and characterized, and results from these studies
have been presented by a number of investigators (Libra et al., 1986;
Kelley et al., 1986; Hallberg, 1986; Libra et al., 1987). Over 85% of
the basin's groundwater is discharged through Big Spring. Thus,
investigators have an easily accessible point for monitoring average
groundwater quality in the basin as affected by agricultural activities.
Big Spring is a karst spring with an estimated 10% of flow contributed by
directed recharge through sinkholes. By using hydrograph separation
techniques, the investigators can determine the water quality effects of
both normal infiltration and direct recharge through the sinkholes.
Four years of monitoring have shown flow-weighted mean
concentrations of NO3-N ranging from 7-11 mg/L. Combined losses of N03-N
in both surface water and groundwater amount to 33-55% of the average
annual fertilizer nitrogen applied in the basin. Historic records
indicate that the magnitude of N03-N concentrations have increased by
200-300% over the last 20 years. This corresponds to a 200-300% increase
in the application of nitrogen fertilizers during this period.
The herbicide atrazine is the only pesticide that is always present
in the discharge from Big Spring. Flow-weighted mean concentrations of
atrazine are less than 1 g/L, but the average concentrations have
steadily increased during four years of monitoring. Total losses of
atrazine in the groundwater amount to less than 0.1% of the annual
application. Pesticide concentrations in surface waters are generally an
order of magnitude higher than observed in the discharge from Big Spring.
Peak concentrations of pesticides occur following recharge events in the
spring after crops have been planted.
There is a growing body of evidence that pesticide residues are
being leached to groundwater in some areas of the country. The U. S. EPA
recognizes that at least 19 pesticides have been found in groundwater in
24 states as a result of agricultural practices (U. S. EPA, 1987a). Many
of the pesticide detections in groundwater have been from the use of
nematicides, which in general need to be very soluble in order to be
effective. The other major class of compounds found in groundwater from
agricultural practices is herbicides. The herbicides atrazine and
alachlor are used over large areas in many regions of the country, and
have been frequently detected in groundwater. Nitrates from
agriculturally applied fertilizers have also been found in groundwater.
The health risks associated with the low concentrations (typically less
than 10 pg/L) of pesticides in drinking water are still unknown. There
are documented health risks, however, to infants drinking water with high
levels of nitrates. The agricultural production systems in this country
will continue to require large irputs of chemical fertilizers and
pesticides. Scientists and governmental regulators must identify ways to
protect groundwater supplies from contamination while allowing farmnners to
use the chemical inputs required to maintain yields. Simulation models
will play an important role in evaluating the contamination potential
from applications of pesticides and fertilizers and assess the benefits
of modifications to current agricultural management practices
(application methods, timing, chemical formulations, tillage practices,
iThere are many factors which influence the mobility and persistence
of pesticides in the environment. Soil physical properties, chemical
properties, crop characteristics, environmental factors, and the
interactions among them determine the fate of agriculturally applied
chemicals. Many of the processes involved in the transport and
degradation of pesticides in the environment are poorly understood, and
the mathematical equations used to represent these processes are
simplifications based on experimental observations. The soil properties
and environmental factors which strongly influence the fate of a chemical
can vary significantly within small areas.
Methods to predict the fate of agriculturally applied chemicals
range from simple indices to complex mathematical research models.
Models which can represent the effects of agricultural management
practices on the transport of chemicals are being used by governmental
regulators as part of the chemical registration process. Most of the
management models developed to date are designed to represent the
processes on the soil and plant surfaces, within the crop root zone and
possibly the unsaturated zone between the root zone and the water table.
New model developments will incorporate groundwater transport models to
assess how far and at what rate a chemical that leaches into groundwater
will move. There are many sources of uncertainty and error associated
with model predictions. Thus models should not be expected to provide
exact predictions of transport and fate within a field. There have been
a limited number of studies however, which provide data of sufficient
detail to assess how well the models do represent average transport
within a field. In order to gain confidence in the use of models for
pesticide fate and transport predictions, many new studies in various
regions throughout the country are needed. To be of real value in model
testing and validation, these studies should last for several years. For
models which require calibration, a year or more of data may be needed to
calibrate the models and then several additional years of data from that
site will be needed to test the predictions of the calibrated models.
4.1 Field Site Description
The study site is located near Tifton, Georgia, in the coastal plain
physiographic region of the southeastern U.S. The geology of this area
has been described by Asmussen et al. (1986). The site is part of a farm
rented by the University of Georgia Coastal Plain Experiment Station
which is referred to as Gopher Ridge in honor of the many gopher
tortoises which make their home on the edges of the fields and in the
The study site covers 0.7 ha in the northwest corner of a 2.3-ha
field. The field was developed for irrigated agricultural research.
Irrigation sprinkler risers are located on a 12 by 12 m grid. Supply
lines for irrigation water are buried approximately 1 m below the soil
surface. The field was removed from active crop research in 1982. At
that time the field was deep tilled in both north-south and east-west
directions to remove possible residual tillage effects. A permanent
bahia grass cover was established and fertilized and irrigated as needed.
No pesticides were applied after the grass cover was established. The
field was maintained in this condition until the spring of 1986 when
instrumentation for this research was installed. The site is bounded on
the north and west by trees. Just beyond the tree line on the western
edge of the field, the surface elevation drops rapidly into a seepage
area where subsurface flows reemerge as surface water that flows to the
The soil on the study site is classified as a Lakeland sand (Typic
Quartzipsamrnents, thermic, coated). The soil profile depth on this site
ranges from 1.9 to 4.4 meters, Underlying the soil is a restricting
layer consisting of tight clays. This restricting layer is the top of
the Hawthorne formation which forms the confining layer over much of the
aquifer system (Floridan) of southern Georgia and Florida. The presence
of this restricting layer causes percolating rainfall or irrigation to
saturate the soil above the layer and form a transient water table.
Previous work by Asmussen et al. (1986) had shown, through the use
of ground penetrating radar (GPR), that the restricting layer below the
study site would cause saturated flow above it to converge into channels.
The presence of these channels made the site look promising for saturated
zone monitoring because samples from wells located in these channels
would be very likely to show the presence of any chemicals which were
moving with the saturated flow. The maps drawn from the (GPR)
descriptions also indicated that the slope of the restricting layer near
the study area boundaries was generally away from the study area. This
would minimize influences of saturated flows coming into the site from
Figure 4.1 shows the elevations of both the soil surface and
restricting layer for the 0.7-ha study site. The elevations given are
relative to a local bench mark which was assigned an elevation of 30.48
m. The soil surface has a relatively uniform slope of approximately 4%
towards the west. The restricting layer shows a more complicated
3 up- ', -1, I I
/_, ,, o/ ,/ i l
07-09 07-10 07-1 0 7-1 07-1 /
ff 0 0 / 0 0, /
0 D 0. 0010 /(D ) G III
17 o / I j /
I8-07 0-09 05- 0 OB-11/ 08-11 06-13 0 /
_oi 0 0,' 0 9-
S37 2 ( p D (/
*/ <5ii/ 'i
Z 09-0o9 09 10 01, 920,3 91 / /
Q j10-8 1 10 10-111 10 12 1?-13 10414
g 0 O P 0
57 -1-07 "- \ "- /
0 a 0 G I
70 90 110 130 150 170 190
90 110 130 150 170 190
Figure 4.1. Contour maps of soil surface and restricting layer
showing locations and ID labels of monitoring wells.
topography. Slopes of the restricting layer range from 1 to 15 percent,
with the general direction of the slope also being towards the west. The
fact that the soil is a coarse sand also was considered to be
advantageous for monitoring pesticide movement. The high hydraulic
conductivities and low clay and organic matter contents should result in
increased leaching and movement of applied chemicals as compared with
tighter soils. Another advantage of this site was the availability of
irrigation which could be used to supplement the natural rainfall if
required to maintain a dynamic flow regime.
This study was not intended to represent typical agricultural
chemical usage and practices in the coastal plain of Georgia. The
primary objective was to observe chemical movement within the unsaturated
soil and in a shallow water table aquifer (shallow groundwater). The
properties of the soil and restricting layer described above were suited
for this objective.
4.2 Site Instrumentation
The movement of the applied chemicals was monitored using a
combination of soil samples, soil-water (soil solution) samples from the
unsaturated zone, and saturated zone samples. These samples were
collected using augers, soil solution samplers, and monitoring wells,
respectively. Soil sample collection required no permanent
instrumentation, and the collection procedure will be described in a
Samples of the water from unsaturated soil were collected using
soil solution samplers. An excellent review of soil solution samplers is
presented by Litaor (1988). Smith and Carsel (1988) presented a design
for a stainless steel solution sampler that do not react with pesticides
during sampling. Samplers constructed of stainless steel were not
economically feasible in this research.
The samplers used in this study were constructed of 4.2 cman OD,
schedule 40 PVC and a 1 bar, high flow ceramic cup, 3.99 can diameter by
19.05 cm long attached to the PVC pipe with epoxy. A 0.64 cm
polypropylene tube was extended from the inside bottanom of the ceramic cup
to a bulkhead fitting on the top of the sampler. A second fitting was
placed on the side of sampler, near the top, for attaching a vacuum line.
The samplers were constructed in two lengths, 1.08 and 2.09 m. The
polypropylene tubing used in the samplers was not tested for adsorption
of atrazine and alachlor. This was not done since the solution samplers
were to be used for collection of samples to be analyzed for inorganic
chemicals such as bromide and nitrate. The volumes of samples collected
from the solution samples were in general too small (less than 50 mL) to
permit reliable quantification of pesticide residues.
The samplers were placed such that the center of the ceramic cup
was located at depths of 61, 122, and 183 cm. The samplers were
installed by augering a 5 cm hole to a depth that was approximately 10
ancm less than the desired sampling depth. A piece of thin wall aluminum
tubing with ID slightly smaller than the ceramic cup was then driven 20
cm beyond the augered hole, removing a core of soil. The sampler was
pushed into the hole until seated on the bottan. A slurry of water and
soil taken from a borrow pit adjacent to the site was then poured into
the amnnulus between the PVC and soil. The top 12 cm of the hole was
widened to approximately 10 an in diameter and was filled with a
bentonite slurry to form a seal around the sampler and prevent direct
flow from the surface down the side of the PVC pipe to the ceramic cup.
Figure 4.1 shows the location of the application area, which is the
strip to which the herbicides and bromide tracer ware applied. The
samplers and three monitoring wells (08-09, 09-09, and 10-09) are within
this area. The samplers were installed in groups of three at a distance
of 3.05 m on either side of each well along a north-south axis. The
samplers were assigned identification labels that indicated their
position with respect to the wells and the depth to which they were
installed, eg. the group of samplers located 3.05 m south of well 09-09
were assigned labels of 09S-2, 09S-4, and 09S-6 for depths of 61, 122,
and 183 an (2, 4, and 6 ft), respectively. A cross-section of the
application area showing the locations of solution samplers and
monitoring wells is shown in Figure 4.2.
A small instrument trailer was located approximately 4 m east of the
application area. A vacuum pump located in the trailer was used to
apply a vacuum to all samplers simultaneously through a manifold. The
pump was capable of creating a vacuum of approximately 28 cm of Hg. A
0.64 an diameter polypropylene sample tube was connected to the sample
fitting on each sampler and routed back to the trailer where it was
connected to a 50 mL polycarbonate sample container. All sample tubing
was enclosed in a protective PVC conduit which was located 1.5 m from the
samplers to avoid interference with chemical application in the vicinity
of the samplers. Between the samplers and the PVC conduit, the stiff
polypropylene vacuum and sample lines were fastened together and held
<- 36.6 M
-> 3 r, 1<2- M ., >< 12.2 --- > 3
0. n Inn
0.6 - SOLUTION
0.6 M T SAMPLERS %3
_,k_ .CR -. ,2 M
'-'^ ^ CELL ^^^ '^^\
__ RESTRICTING LAYER
Figure 4.2. Cross-section of soil profile through application area
showing locations of monitoring wells and soil solution
about 0.6 m above the soil surface to minimize interference around the
samplers. When it was time to collect the samples from the samplers,
the vacuum was switched to the sample collection side and the vacuum side
was opened to atmospheric pressure. Any water that had accumulated in
the samplers would then be pulled into the sample containers.
Monitoring wells for collection of samples frcm the saturated zone
and measurement of the water table elevation were installed during May
of 1986. The wells were constructed of 6.3 cm diameter, schedule 40 PVC
and were slotted using a veneer blade on a radial arm saw. Three rows of
slots were cut along the circumference of the well. Each slot was
approximately one-sixth as long as the pipe circumference. The slots
were spaced approximately 2.5 cm apart over the bottComn 1.22 m of the
well. Although several manufacturers offer slotted PVC well screens,
these were beyond the budgetary constraints of this project. Scmne
existing wells, installed by the USDA-ARS Southeast Watershed Research
Laboratory in conjunction with the GFPR mapping by Asmussen et al. (1986)
were also used. Conversations with the technicians involved with the
installation revealed that the wells were constructed of the same
materials as described above but were perforated with tiny drill holes
instead of being slotted. They were unable to recall the diameter of the
holes or the extent of the perforation above the bottom of the well.
Appendix A lists statistics associated with each well and indicates which
wells were installed by USDA.
The wells were installed by hand augering a 10 an diameter hole
down to the top of the restricting layer. The restricting layer was
identified by a sudden change from yellcowish-white sand to red clay
mixed with small rocks. The change was very abnrupt and the effort
required to auger through the clay was significantly greater than to
auger through the coarse sand. A PVC end cap was slipped onto the end
of the well and held in place by friction during installation of the
well. Fine gravel was placed around the wells in the zone of the slots
in an attempt to minimize sand migration into the wells through the
relatively coarse slots. Soil from a nearby borrow pit was used to fill
the hole around the well to within 12 cm of the surface. Ten centimeters
of a water and bentonite slurry was then added to form a seal around the
well. The final 2 cm was filled with soil. The USDA augered down to
the restricting layer with a hollow stemn auger on a small drill rig.
The well was slipped into the center of the auger and the auger was
withdrawn. No details on backfill procedures used by the USDA are
The wells were continuous from the restricting layer to a height of
approximately 50 an above the soil surface. No solvent weld joints or
connections were used. Each well was fitted with a 0.64 ncm diameter
polypropylene tube which extended from approximately 2.5 an above the
bottom of the well through a #2 rubber stopper placed in a vented PVC
cap. The polypropylene tubes were not tested for adsorption of atrazine
and alachlor. The caps were vented by milling small channels into
opposite sides of the inside of the cap. This was done to prevent a
vacuum from forming in the well during sampling.
The wells were installed on a 12 by 12 m grid corresponding to the
locations of existing sprinkler risers. The wells were offset by
approximately 50 cm to the west of the sprinkler risers to avoid hitting
supply lines beneath the soil surface. In the application area the
wells and solution samplers were offset from the sprinkler risers by
about 3.05 m so that the risers and supply lines would not interfere
with the downward movement of chemicals within the vicinity of the wells
and samplers. The monitoring wells were assigned ID labels to indicate
their position relative to the sprinkler riser in the SE corner of the
2.7 ha field on which the experimental site is located. Figure 4.1 shows
the locations of the wells and the associated ID labels. Well 08-15 is
located at the edge of the tree line bordering this side of the study
site. One well is located in the woods approximately 30 m west northwest
of well 08-15. This area is a seepage zone where subsurface flows
reemerge as surface water and flow to the Little River. This well is
referred to as "I1'".
Additional instrumentation installed at the site consists of a
weighing rain gage which was located near well 09-12, a water table
recorder which was located near well 08-12, tensicneters for measuring
the water content of the unsaturated soil, and a "deep" well located
between wells 08-13 and 09-13.
The rain gage is a US Weather Bureau standard weighing bucket rain
gage with a seven day chart and 30.5 ancm capacity. The rain gage was
calibrated using standard calibration weights after installation.
The water table recorder was installed to provide a continuous
record of the water table elevation to supplement the weekly
measurements taken on all wells. The recorder was equipped with a seven
day clock and a drum type chart recorder.
Eight tensicmeters were installed around each set of solution
samplers. The tensicmeters were water filled and connected with very
small tubing to a mercury manometer board. The tensicmeters were
installed at depths of 30, 60, 90, 110, 122, 140, and 183 cmn. There were
two tensiameters located at the 60 ancm depth. The tubing connecting the
tensiumeters to the manometer boards was routed above the soil surface
with the vacuum and sample lines as noted above.
The "deep" well was installed by drilling a 10 ancm diameter hole
inside of a 11.4 ancm diameter PVC casing. The casing was pushed down as
the auger advanced. The casing was installed in this manner using 1.5 m
sections to a depth of approximately 12.2 m below soil surface. The
last 3 m of the well was located in a saturated formation that liquified
during drilling. The observation well was made of a 1.5 m section of
6.3 cm diameter commercial PVC well screening solvent-welded to two 6.1 m
sections of schedule 40 PVC. This was jetted down inside of the casing
until it bottomed cut at the bottom of the previously augered hole.
After placement of the well, a mixture of bentonite and water was poured
into the area between the casing and well up to the top of the casing.
The purpose of this well was to observe the piezcmetric head difference
across the restricting layer.
4.3 Chemical Applications
The first application of chemicals to the experimental site
occurred on 11/12/86. On the day preceding this first application, the
entire field site was rotary mowed. The grass within the application
area was cut to a height of approximately 5 cm. All dead grass and
clippings were raked up and removed from the area prior to application.
The application area consists of a strip 36.6 m long by 9.14 m wide as
shown in Figure 4.1.
Atrazine and alachlor were applied simultaneously on November 12.
Atrazine was applied in the form of AAtrex (Ciba-Geigy Corp.) which is
an emulsifiable concentrate containing 0.479 kg AI (active ingredient)/L
(4 lb/gal). Alachlor was applied in the form of Lasso (Monsanto
Agricultural Products Co.) which is also an emulsifiable concentrate
containing 0.479 kg AI/L. The herbicides were applied using a self-
propelled born sprayer with provisions for injecting chemicals directly
into the water stream (Sumner et al., 1987). The boom length on the
sprayer is 9.14 m.
The intended application rate of the herbicides was 4.5 kg AI/ha (4
lb/ac). One liter of herbicide solution was prepared by mixing 313 mL of
each of the chemicals with 374 mL of deionized water. The sprayer was
operated at a velocity of 3.05 m/min moving from south to north along the
edge of the application area. The chemical solution was injected into
the water stream of the sprayer using an injection pump calibrated to
deliver 83 ml/min. The water pressure at the inlet of the sprayer was
207 kPa. The pressure at the spray nozzles was maintained at 138 kPa.
Despite the calibration of the injection pump and measurement of sprayer
ground speed, the chemical solution ran out as the sprayer reached well
10-09 which is 6.1 m short of the intended end of the application area.
The application of the herbicides was considered complete at this point,
and the remaining portion of the application area was left untreated.
Fourteen 0.35 1 plastic cups were randomly placed within the
application area. The cups were supported upright in holders at a
height of approximately 35 cm above the soil surface. The cups were
used to collect samples of the application solution for analysis of
chemical concentrations, determination of the depth of water applied to
the area during application, and uniformity of chemical and water
application. The use of these cups and holders has been tested and
found to accurately reflect water application amounts (Stansell et al.,
Immediately after completing the application of the herbicides, the
sprinkler irrigation system was started and 5.1 cm of water was applied
over the entire study site. This irrigation was intended to wash the
herbicides off of the grass foliage and move them into the soil profile
to minimize volatilization losses from the soil and plant surfaces.
Bromide in the form of potassium brmanide (KBr) solution was applied
to the application area using the same sprayer on 11/17/86. There was a
light drizzle of rain during application. Five hundred grams of KBr was
dissolved in 1050 mL of deionized water and injected into the sprayer as
described above. This was equivalent to an application rate of 10 kg/ha
of bromide. The sprayer again was operated at a nozzle pressure of 138
kPa, and a ground speed of 3.05 m/min. Fourteen collectors in the same
locations as for the herbicide application were used to collect
No further chemical applications were made prior to fertilization
of the grass on 4/16/87. The fertilizer was applied at a rate of 560
kg/ha of 5-10-15 using a tractor-nounted broadcast spreader. Fifteen
percent of the nitrogen in the fertilizer was in the form of nitrate
nitrogen (NO3-N). This is equivalent to applying 18.6 kg/ha of nitrate
In an effort to further characterize bromide movement within the
soil profile and determine flow velocities within the grmrdwater, an
additional bromide application and a separate chloride application were
made. On 4/27/87, KBr was applied to the application area using the
sprayer utilized during the previous application. The application
solution was made by dissolving 1.2 kg of KBr in 2.3 1 of deionized
water. The sprayer made two passes across the application area during
which it applied approximately 1.6 1 of the KBr solution. This was
equivalent to an application rate of 17 kg/ha of bromide. Fifteen
plastic cups were used to catch application solution samples in the same
manner as previously described. Immediately following application the
field was irrigated. Irrigations continued daily for a week in order to
raise the water table and provide plenty of infiltrating water to
transport the bromide through the soil.
Prior to the beginning of irrigations on April 27, most of the wells
on the site were dry. The water table began to rise on May 1 at which
time 10 1 of solution containing 500 g of potassium chloride (KCI) was
poured directly into well 07-09 which is near the top of the site. It
was anticipated that the high concentration of chloride in the solution
could be followed downslope with the saturated flow and allow for
determination of water table flow velocities.
4.4 Sample Collection and Storage
Samples were collected weekly (Monday) throughout the study. There
were occasional periods of more frequent sampling immediatelyy after
initial applications and during the last week of April through the
middle of May).
The flint glass containers used for well sample collection and
storage throughout this study did not have teflon lined caps. Therefore,
a small piece of parafilm (a common laboratory film used for sealing
beakers and containers to prevent evaporation and contamination) was
stretched over the mouth of the bottle prior to screwing on the caps.
Samples of the application solutions of the herbicides and bromide
were transferred to glass containers and placed in a cooler with ice
packs immediately after the sprayer had passed completely by the sample
Soil samples were collected in one of two ways: taking 2.5 ncm
diameter by 5 cm long cores using a soil sampling probe, or using a 5 cmn
diameter stainless steel bucket auger. The soil sampling probe was
generally used for collection of samples from the top 5 cm of the soil
profile. When this method was used, 5 to 8 samples would be collected
franom a small area and composited into a plastic sample bag. This would
represent one final sample for the soil surface at that location. For
deeper samples the auger was used. The auger was used to remove soil to
within approximately 5 cm of the desired sample depth. The auger was
then cleaned to remove traces of soil, and a sample was collected which
represented soil from a depth of 5 cm above to 5 cm below the target
depth. The sample was carefully poured from the top of bucket into a
plastic sample bag and sealed. The number of locations within the
application area sampled depended on the number of depths to be sampled
and ranged from 5 to 14. After collection, the soil samples were placed
into coolers with ice packs for transport to Gainesville. Soil samples
franom outside of the application area were collected several times during
the study to use as blanks and spikes during soil residue analysis.
Vacuum was applied to the solution samplers immediately upon
arrival at the study site (7:00 8:00 am). The vacuum would typically
be left on for 8 to 10 hours prior to collecting the soil solution
samples. The vacuum applied to the samplers was approximately 28 cm of
Hg. Near the end of the day the vacuum would be released and the samples
would be drawn directly into individual containers. The containers are
made of polycarbonate with a capacity of 50 mL. The containers had
polycarbonate screw caps for sealing. Immediately after all soil
solution samples were collected, the sample containers were placed in a
cooler for transport.
The elevation of the water table was measured using a well depth
indicator with a stainless steel probe. The depth indicator would sound
an audible alarm when the probe contacted water in the wells. The probe
was attached to a flat cable with markings in feet and inches, with the
smallest division being 1/4 in. Water depths were recorded early in the
morning after arrival.
After the water table elevations in all wells were measured, the
wells were pumped out to insure that the water to be sampled was
representative of the surrounding water in the saturated zone at that
time. When possible, the well was pumped dry. When the depth of water
in the well exceeded approximately 0.4 to 0.5 m, the pumping rate was
less than the rate of flow into the well. When this occurred, the well
was pumped long enough to insure that 2-3 well volumes had been pulled
from the well. A small, 12v battery powered, peristaltic pump was used
to pump the wells and collect samples.
Samples were collected after all wells had been pumped as described
above. A rubber stopper was attached to the intake side of the pump
through a 1 m length of 1.25 cm OD Tygon tubing. The stopper also had a
piece of Tygon tubing approximately 10 cm in length connected to it for
attachment to the sampling tube in each well. The stopper was placed in
a sample bottle and the short piece of tygon tubing was attached to the
well sample tube. Since the pump was downstream from the sample con-
trainer, the only components common to each sample collection were the
rubber stopper and the short piece of Tygon tubing. When a sample was
collected, the initial water from the well was swirled around the bottle,
and the bottle was then inverted to remove this rinse water. This served
to rinse the Tygon tubing, stopper, and bottle. Thus, cross
contamination between wells was minimized. The sample bottles used for
the collection and storage of well samples had a capacity of 250 mL.
Well samples were placed in coolers with ice packs for storage until they
could be refrigerated.
All soil samples were frozen as soon as possible after collection
and maintained at -18C until preparation for analysis. All water
samples were refrigerated after collection and maintained at 4C until
prepared for analysis.
4.5 Sample Analysis
4.5.1 Inorganic tracer analysis
The concentrations of the three inorganic chemicals of interest
which were applied during this study were quantified using ion
chromatography. A DIONEX QIC ion chrcmatograph (IC) was used for all
A DICNEX AS4A anion separation column was used with an AG4A guard
column in place to remove organic contaminants prior to reaching the
separation column. The IC was equipped with a conductivity detector
connected to an external integrator/recorder. The IC was also equipped
with an anion micromembrane suppressor to reduce the background
conductivity of the eluant and thus allow for greater sensitivity to the
conductivity changes caused by the presence of anions in the sample.
There was no automatic sampling device with this instrument, so
each sample was injected by hand. Approximately 1 mIL of sample was
loaded onto a sample loop prior to injection onto the column. The
sample loop retained approximately 0.25 mL of the sample with the rest
being discarded to a waste line. Between injections, the sample loop
was flushed with 1 mL of deionized (DI) water. Prior to the guard
column, the sample was passed through two filters rated at 30 and 5
microns, respectively, to remove solids from the sample. These filters
were an integral part of the IC and the filter elements were changed
whenever the system pressure was observed to be increasing above nominal
The eluant used in these analyses was 2.2 millimolar (mM) sodium
carbonate (0.933 g Na2CO3 in 4 1 DI H20) and 0.75 mM sodium bicarbonate
(0.25 g NaHCD3 in the same 4 1 of DI water). This solution was made
fresh as needed. All references herein to deionized (DI) water refer to
water that had been passed through carbon filters and ion exchange resins
to produce DI water with a resistivity of at least 1 megohm. The
regenerant solution was 0.025 N sulfuric acid (2.8 mL of concentrated
H2SO4 in 4 1 of DI water). The regenerant solution was passed through
the micromembrane suppressor to reduce background conductivity.
In normal operation the eluant flowrate was approximately 2.0-2.5
mL/min, and regenerant flowrate was approximately 4 mL/min. The
pressure at the discharge of the pump at the stated eluant flowrate was
approximately 8,274 kPa (1200 psi).
A single injection could reveal the presence and concentrations (if
standards were prepared) of the following anions (in order of increasing
retention times): fluoride (F-), chloride (Cl-), bromide (Br), nitrate
(NO3-), phoshate (P043"), and sulfate (SO42-).
In ion chrImatograpy, the retention times of the various anions are
dependent upon total ionic strength of the sample, ionic strength of the
individual ions, eluant flowrate, and temperature of the sample and
eluant. Thus, the retention times could, and did, vary. When the
integrator was set to calculate the areas under the peaks, it was found
that the automatic selection of the beginning and ending points of peaks
was not consistent. A small change in retention time could
significantly alter the reported area for a given ion even when using
duplicate injections. Better results were achieved by setting the
integrator to report maximum peak height. This value was not sensitive
to variations in the starting and stopping times of a peak as long the
baseline was relatively stable.
Initial injections of water table samples from the experimental
site revealed the presence Cl- and S042- in all samples, and relatively
high concentrations of N03- in samples from well IOW. Standards were
prepared using DI H20 and oven-dried quantities of certified or primary
standard grades of potassium chloride, potassium bromide, potassium
nitrate, and potassium sulfate. The standards initially included Cl-
and S042- because of their presence in the samples. They were not used
to quantify the concentrations of either species during initial sample
analysis. When C- was introduced into the water table as a tracer late
in the experiment, sample concentrations of Cl- were quantified. It was
observed that the response of the chromatogram to Cl- was approximately
twice the response to the other anions of interest. Thus, standards were
prepared with Cl- at half of the concentrations of the other species so
that the responses of all species would be similar and on scale
simultaneously. Standards were prepared with concentrations of the main
species of interest ranging frcum 0.01 to 100.0 mg/L. A typical standard
would contain 0.5 mg/L Cl- and 1.0 mg/L of Br, N03-, and SO42-. It
should be noted here that all references to nitrate (N03-) indicate
concentrations of N03- and not N03-N (nitrate-nitrogen).
Samples were taken from the refrigerator and allowed to cme to
room temperature prior to injection. No sample preparation was
performed. The analysis protocol typically followed the pattern of 3
standards, 10 samples, 3 standards, etc. The standards used were
selected based upon the concentrations of the analytes being observed in
the samples. Thus, standards were selected to bracket the observed
values as closely as possible. Occasionally a sample would be reinjected
to observe the repeatability of the analysis procedures. Samples from
wells located away from the experimental site, including the deep well
which supplies irrigation water to the farm, were regularly analyzed as
field blanks (no Br- would be expected; however, C-, N03-, and S042-
could be present).
A piecewise linear fit to each sequence of injected standards was
used to generate a function of the form: concentration = f(peak height).
The piecewise fit was selected instead of linear regression because for
some species (particularly high concentrations of Cl-) a linear
regression resulted in a negative y-axis intercept. This implies that
the species can generate a noticeable positive response and yet be
calculated to have a negative concentration. It is expected that a
positive intercept would be a normal result, i.e. there is sane threshold
concentration of the species required before the detector indicates a
response. The piecewise fit insured that any positive response would
result in a positive calculated concentration. Conversations with an
organic chemist at the Pesticide Residue Laboratory of the USDA-ARS
Southeast Watershed Research Laboratory indicated that the piecewise fit
would be an acceptable method of generating a standard curve (Marti,
The standard curves generated on either end of the analysis of 10
samples were used to compute the concentrations of the species in the
samples. The resulting concentrations were then linearly weighted based
on sample position with respect to either set of standards, i.e. the
first sample following the standards would be weighted towards the curve
generated by the preceding standards and only slightly effected by the
standard curve generated 9 samples later. The sample midway between
standards would be evenly weighted between the two standard curves. In
general, there was little drift in the instrument and the standard
curves did not vary much over a period of hours.
4.5.2 Herbicide residue extraction
Analysis of samples for residues of atrazine and alachlor required
different extraction procedures for soil and water samples. Once the
extractions were completed, all samples were analyzed on a ccnmmon gas
chrcnatograph with identical operating conditions.
There are many published methods for preparing samples for
determination of atrazine and alachlor residues (e.g., Rdhde et al.,
1981; Voznakova and Tatar, 1983). In general, the procedures are rather
detailed and costly in terms of time and supplies. Dr. Willis Wheeler,
Director of the Pesticide Residue Laboratory in the Food Science and
Human Nutrition Department at the University of Florida, Gainesville,
indicated that a simpler method of sample preparation might be developed
which would be satisfactory in terms of the recovery of the pesticide of
interest while being more economical in both the time required for sample
preparation and expense (Wheeler, 1987). The more steps involved in the
extraction and preparation of a sample, the more likely it is that
something can go wrong and the residues lost. With these factors in
mind, methods for extraction of the two herbicides from soil and water
samples were developed which were in keeping with the budgetary
constraints of the project and the time constraints on the analyst.
Extraction of the herbicide residues from water samples was done
using a liquid-liquid extraction procedure. One hundred fifty
milliliters of the sample water were placed in a quart mason jar along
with 150 mL of pesticide-grade hexane. The jar was sealed with a
standard mason jar lid which had been lined with a self-adhesive teflon
film. The jar was agitated on a shaker table for thirty minutes after
which the contents of the jar were poured into a labeled 500 mL
separatory funnel. The sample water was drained through the separatory
funnel back into the jar, leaving the hexane in the funnel. Another 150
mL of hexane was added to the jar and it was again agitated for 30
minutes. The contents of the jar were again added to the funnel. The
water was then drained and discarded. The hexane was filtered into a 500
mL boiling flask through sodium sulfate to remove any remaining water.
Next, the boiling flask was placed on a rotary evaporator and the
reduced in volume to near dryness. A few mL of fresh hexane was then
added to the flask and swirled to wash the sides of the flask. This
hexane was placed in a 10 mL volumetric flask. A few mL of hexane was
again added to the boiling flask and the swirling repeated. After two
rinses of the flask, the hexane in the volumetric flask was brought to a
final volume of 10 mL. The finished sample was stored under
refrigeration in test tubes with teflon lined screw closures. The
initial sample volume of 150 mL and the final extracted volume of 10 mL
indicate that this procedure concentrated the residues by a factor of 15.
Blanks taken from remote wells as well as DI water were extracted as a
check on the procedure. Spiked samples of well water and DI water were
extracted to determine the recovery levels of the herbicides using the
procedure described above.
Soil samples were removed from the freezer and allowed to air dry
for one day prior to extraction. This was done to allow excess moisture
to evaporate since the acetone used in the extraction process can absorb
water which will not be removed by filtration through sodium sulfate as
was done for the hexane extractions. Acetone was chosen as the solvent
for soil extractions because it was observed that the water in the soil
samples, or sane other factor, tended to repel hexane. It did not
appear that the hexane was adequately dispersing the soil particles.
A fifty gram portion of the air-dried soil sample was weighed and
placed into the mason jar. One hundred mL of pesticide grade acetone
was then added to the soil sample. The sample was agitated for 30
minutes and the acetone was carefully decanted into a separate labeled
glass container. Another 100 mL portion of acetone was added to the soil
and the agitation was repeated for another 30 minutes. The acetone was
poured into the container with the previous portion. Several rinses of
acetone were poured through the extracted soil to remove any acetone
left from the extractions. The acetone rinses were added to the acetone
frcm the extractions and the entire volume was vacuum filtered through a
47 nmm diameter filter with a 0.45 micron pore size. The filtrate was
then reduced to near dryness on the rotary evaporator and brought to a
final volume of 10 mL in acetone. Soil samples taken from outside the
study area were extracted as field blanks and were spiked with known
quantities of the herbicides and extracted as a check on the extraction
and analysis procedures.
A subsample of approximately 10 g of each soil sample was taken to
determine the soil-water content so that the concentrations of the
herbicides in the soil could be reported on dry weight basis. Soil
samples for soil-water content determinations were dried for 24 hours in
an oven at 100C.
4.5.3. Herbicide residue analysis
All herbicide residue samples were analyzed on a Hewlett Packard
model 5840A gas chromatograph (GC). The GC was equipped with a 35
position autosampler and a model 5840 GC terminal (integrator). The GC
had both electron capture (EC) and nitrogen-phosphorus (NP) detectors.
IDuiring the early phases of the study (methods development), the electron
capture detector was used. This detector had a high sensitivity to
alachlor and a lower sensitivity to atrazine. However, when field
samples were analyzed, as opposed to DI water spiked with pesticides, it
was observed that the EC detector was sensitive to many naturally-
occurring cmpourds (the EC detector is sensitive to chlorine molecules)
which had been extracted along with the pesticides. Thus, the output was
so crowded with peaks that the quantities of atrazine and alachlor
present could no longer be reliably determined. The NP detector is very
sensitive to molecules of nitrogen and phosphorus and insensitive to
almost anything else. When the samples were analyzed on this detector,
the chromatograms were much cleaner and atrazine and alachlor were often
the only visible peaks. The NP detector is somwhere between 5 and 10
times more sensitive to atrazine than it is to alachlor due to the
greater number of nitrogen atoms in the structure of atrazine. The NP
detector was operated with detector gas flowrates of air at 50 mL/min and
hydrogen at 3 mL/min. The carrier gas was helium with a flowrate of 30
The column used in the GC was 1.83 m long with a 4 mm ID. The
column was packed with 3% OV-17 (50% phenyl, methyl silicone) on 100/120
mesh Gas Chrom Q. The operating conditions were: injector port
temperature set at 2400C, column temperature at 200C, and detector
temperature at 300C.
The sample protocol used with the auto sampler was usually of the
form: 4 standards, 1 solvent blank, 10 samples, 4 standards, 1 solvent
blank, 10 samples, 4 standards, 1 solvent blank. The solvent blank was
used to help prevent carryover from the last (and highest) standard into
the following sample. Carryover was normally not a problem, but the
solvent blank would often show some trace of the herbicides.
Working standards of the two herbicides were prepared using
reference standards provided by the EPA in Research Triangle Park, North
Carolina. The initial standard was made by dissolving 10 mg of the
reference into 10 mL of solvent yielding a concentration of 1 mg/mL
(1000 mg/L). This standard was then used as the basis for subsequent
working standards. The standards were prepared with the concentration
of alachlor being 5 times greater than the concentration of atrazine so
that the detector response to the chemicals would be nearly equal.
Working standards of atrazine ranged from 0.01-500.0 mg/L with the
concentrations of aladhlor generally being 5 times higher as noted above
(very high atrazine standards did not include alachlor). The 0.01 mg/L
standard of atrazine represented the smallest concentration which was
detectable without pushing the limits of machine and detector. The
autosampler injected a volume of 4.8 AL. So an injection of 4.8 AL of
0.01 mg/L atrazine represented a lowest detectable mass of 10 nanograms
Whenever new standards were prepared, they were compared with the
previous standards to assure similar response and that no mistakes had
been made during the dilutions. Two sets of standards were made. One
set was made in acetone for use in analyzing the soil sample extracts.
The other set was made in hexane for use in analyzing the water sample
extracts. Comparisons between the two sets of standards revealed no
differences in response, but recommended procedures (Wheeler, 1987)
dictate that the standards be made up in the same solvent as the samples.
The integrator available with the GC did not work properly. In
order to quantify the amount of herbicides present in the samples, all
peak heights were hand measured (a peak height mode, as used with the IC
was not available on this integrator).
A linear regression was performed from the standards to give a
relationship of the form: concentration = f(peak height). There were no
problems with negative intercepts in these regressions as discussed in
the section on ion chromatography. Sample concentrations were computed
with the regression equations calculated from standards on either side of
the samples. In order to account for drift in the detector response, the
final concentrations were calculated by linearly weighting the
concentrations computed from each set of standards as discussed
previously. The NP detector had a tendency to drift and this correction
for drift was usually needed.
MOELIG THE EXPERDfIMETAL SITE
Two models were selected for the purpose of simulating the movement
of the chemicals applied to the experimental site. The models selected
for use were the Pesticide Root Zone Model, PRZM (Carsel et al., 1984),
version 2, and Grourxdwater Loading Effects of Agricultural Management
Systems, GLEAMS (leonard et al., 1987), version 1.8.53. These models
were selected because they were designed specifically to simulate
pesticide transport through the root zone of a crop and reflect
influences of agricultural management practices such as tillage.
Reference to the manual for GLEAMS includes the documentation for the
CREAMS model (Knisel, 1980) from which GLEAMS was derived as well as the
supplementary GLEAMS user manual which is provided with the model code
and describes the differences in input data sets between CREAMS and
This chapter will discuss the selection of input values used in the
models. Only the parameters relevant to the results will be discussed.
There are parameters in both models for which values were selected, but
which had no bearing on the results to be presented. For example the
models utilize the Universal Soil Loss Equation, USLE, (Wischmeier and
Smith, 1978) for prediction of soil erosion, however, in the simulations
surface runoff was never predicted and therefore the parameters chosen
for use with the USLE did not affect the reported results. There are
many parameters such as rainfall, pesticide properties, and soil
characteristics which are common to both models. Thus, the selection of
common parameter values will be discussed first, followed by selection of
values for parameters which are unique to each model.
5.1 Selection of Common Input Parameter Values
5.1.1 Rainfall and irrigation
Both models use daily rainfall data. The formats of the rainfall
files for the models are different and PRZM reads in values in
centimeters while GLEAMS reads in values in hundredths of an inch.
The data used to create the precipitation files came from a
combination of two rain gages. As stated previously, one weighing rain
gage is centrally located within the experimental site. This rain gage
was installed in June of 1986. A second rain gage is located
approximately 300 m north-northeast of the experimental site. This gage
is a tipping bucket type rain gage that is connected to a data logger.
All rainfall data used in the simulations came from this off-site gage
since the data fram it was already on a computer and available. The
rainfall amounts recorded in both gages agreed well. The gage located on
the site was used to supplement the rainfall record with amounts of
irrigation applied to the site. This rainfall plus irrigation file was
then manipulated into the units and formats required by each model, and
checked to insure that the two formats contained identical information.
5.1.2 Soil properties
Both models use the Soil Conservation Service (SCS) curve number
method (USDA-SCS, 1972) to partition rainfall between runoff and
infiltration. The runoff curve number was selected from Table 9 in the
PRZM manual and was chosen based on the assumption that the bahia grass
cover could be assumed to be similar to a pasture in good condition
without contouring (Lakeland sand is classified as being in hydrologic
soil group A).
Physical descriptions of the soil profile were obtained from two
sources. Hook (1985) measured properties of the soil in another section
of the farm on which the experimental site is located. The data
collected by Hook (1985) were measured within the top 120 cm of the soil
profile. The soils on the farm are uniformly classified as either a
Lakeland sand or Bonifay sand. The primary differentiation between the
soils is the depth to the restricting layer with the Lakeland soil being
deeper than the Bonifay. Soil physical properties such as water
retention and hydraulic conductivities are considered to be the same
between these soils on this farm. The second source of data used to
determine soil properties is a soil characterization report for Florida
soils prepared by the Soil Science Department at the University of
Florida (Carlisle et al., 1978). This report contains characterization
data for a Lakeland sand.
The properties that were selected to represent the soil profile of
the experimental site are summarized in Table 5.1. The average depth to
the restricting layer for the three wells within the application area is
2.62 m and this was chosen to represent the profile depth.
Table 5.1 Soil properties used in simulations.1
Organic Bulk Hydraulic Water Content (%)
Depth Carbon Density Conductivity Effective Field Wiltng
(cm) (%) (g/cm3) (cm/hr) Saturation Capacity2 Point
0- 13 0.55 1.45 13.3 33.0 10.0 2.1
13- 20 0.28 1.60 13.3 33.0 10.0 2.1
20- 51 0.08 1.58 22.9 33.3 10.0 2.1
51-102 0.04 1.59 36.5 33.8 9.0 2.1
102-262 0.03 1.59 45.0 34.0 9.0 2.1
Data frcm Hook (1985) and Carlisle, et al., (1978).
2Field gravimetric water content following 48 hours of drainage under
plastic following 16 hours of flooding.
3Water content at 15 bars tension.
5.1.3 Pesticide chemical properties and applications
The pesticide chemical properties required by the models are
solubility, partition coefficient, degradation rates on foliage and in
the soil, and a coefficient of plant uptake. The models may require
these parameters in slightly different forms, but the basic information
required is the same. In addition to the model manuals, the manual for
LEACH (Dean et al., 1984) was a good source of estimates of the required
parameters. No exhaustive search of possible values of parameters for
each chemical was undertaken. At this point in the study, the models
were run using values that a user could obtain frcm the manuals. This
may result in less than optimum performance of the models, but probably
reflects the way in which a typical user would apply them.
Reported properties of pesticides are often widely variable. For
properties such as partition coefficient and degradation rates, the
variability is understandable and expected. However, even water
solubility was found to vary by a factor of two. The solubility of
atrazine is not available from the PRZM manual. The CFEAMS manual lists
a value of 33 mg/L. However, the Agrichemical Handbook (1983) gives a
value of 70 mg/L at 20C. The solubility of alachlor is given in the
PRZM manual as 220 mg/L at 20-25C. The CEEAMS manual reports a value of
242 mg/L at an unspecified temperature. The Farm Chemical Harxndbook
(1984) does not report a value for alachlor and the Agrichemical Handbook
(1983) gives the solubility of alachlor in water as 148 mg/L at room
temperature. The final values selected for all chemical properties are
shown in Table 5.2.
The solubilities of brmnide, nitrate, and chloride listed in Table
5.2 are considerably less than the actual solubilities. However,
solubilities of ionic species are seldom listed for reference. The
values in Table 5.2, while less than the actual solubility values, are
sufficiently high such that solubility was not a limiting factor in the
mobility of the chemicals.
Table 5.2 Chemical properties used in simulations.
Property Atrazine Alachlor Bromide Nitrate Chloride
Solubility 33 148 800 800 800
Partition coef., 163 268 0 0 0
Half-life 78 18 999 999 999
Plant Uptake Coef. 1.0 1.0 1.0 1.0 1.0
0.0 0.0 0.0 0.0 0.0
The user manuals for the models contain tables drawn from many
sources which provide values of the partition coefficient for many
pesticides. The IEAL H manual also contains a number of tables. The
tables in CREAMS and IEACH are more extensive than those in FPRZM.
Different tables within the same manual my give differing values. Table
values often list coefficients of variation (CV) on the order of 50-130
percent. PRZM presents equations by which the organic carbon partition
coefficient, Koc, can be calculated if solubility or the octanol-water
partition coefficient, Kow, is known.
The partition coefficients of alachlor and atrazine were not
measured on this soil and thus values were chosen from the sources
described above. A Koc value for atrazine of 163 cm3/g with a CV of 49%
was found in the IEACH manual. No listing of a Koc was found for
alachlor. PRZM did list a value for log(Kow) which was 2.78. PRZM
presents a relationship between log(KFw) and Koc which is
log Koc = 1.00 (log Kow) 0.21 5.1
Using this equation a Koc of 371 was calculated for alachlor. IEACH
lists a Kw value for alachlor as 434, using this and equation 5.1, a Koc
value of 268 was calculated. Since alachlor has been reported as one of
the pesticides ocamnmly found in groundwater, it was decided to choose
the lower Koc value for use in the model as this would tend to cause
higher predicted leaching losses.
The inorganic tracers bromide, chloride, and nitrate were assumed to
act as non-adsorbed pesticides. Thus they were assigned a Koc value of
Mhe degradation rate constant of atrazine in soil was given in the
LEACH manual as ranging fromn 0.0149 to 0.0063 days-1, corresponding to a
half-life of between 46 and 110 days respectively. The midpoint value
of 78 days was chosen for use in these simulations. A degradation rate
constant of 0.0384 days-1, which corresponds to a half-life of 18 days,
was given for alachlor in both the CREAMS and IEACH manuals.
The coefficient of plant uptake of pesticides with transpiration was
determined using the relationship given in the PRZM manual in which the
uptake factor is a function of KOw and is given as
UPlKF = 0.784 exp [(log Kow 1.78)2/2.44] 5.2
where UPMKF = plant uptake efficiency factor.
Using the Kow for alachlor of 434 as discussed above, the uptake factor
was calculated to be 0.52. Back calculating a Kow for atrazine, based
upon the chosen Koc of 163 using equation 5.1, yields a KOw of 264.
Using this value in equation 5.2 gives a plant uptake factor of 0.65.
The models were also run using uptake factors for both atrazine and
alachlor of 0.0 and 1.0 in order to assess the sensitivity to these
factors. GLEAMS currently recmmends that in the absence of well defined
uptake coefficients, a value of 1.0 should be used. Uptake factors for
the inorganic tracers were assigned values of both 0.0 and 1.0 to observe
the sensitivity of these non-adsorbed chemicals to plant uptake.
The dates and amounts of the chemicals applied to the soil surface
of the experimental site were entered into the models in a direct manner,
ie. there were no calculations or transformations required to convert the
actual values into model parameters. The application rates used in the
simulations correspond to the intended application rates as opposed to
the measured rates. The dates and application rates used as model inputs
are shown in Table 5.3. The applications were assumed to reach the soil
surface with no chemical residues remaining on plant surfaces. The
amount applied was further assumed to be uniformly incorporated into the
top 1 cm of the soil.
5.1.4 Length of simulations
Both models simulated the two year time period from 1/1/86 through
12/31/87. This was done to allow the models to overcome any effects of
initial conditions prior to beginning simulation of chemical
applications. Continuing simulations for six months beyond the end of
field data collection would show if any significant leaching would likely
have occurred after sampling was terminated.
Table 5.3 Chemical applications summary.
Date Chemical Applied Application Rate
11/12/86 Atrazine 4.9
11/17/86 Brcnide 10.0
4/16/87 Nitrate 18.6
4/27/87 Braomide 17.0
5.2 Parameters Unique to PRZM
5.2.1 Evapotranspiration prediction.
PRZM uses either measured daily pan evaporation or daily mean
temperature to determine evaporation from the soil and plant surfaces,
and transpiration by the crop. In these simulations, daily pan
evaporation values recorded at a weather station located on the
University of Georgia Coastal Plain Experiment Station campus were used.
This weather station is located approximately 8 km from the research
site. PRZM requires a pan coefficient which is a multiplier used to
adjust pan readings to represent total daily potential evapotranspiration
(PET). A pan factor of 0.75 was selected front a figure provided in the
manual showing pan factors as related to geographic location.
Two additional parameters required for evapotranspiration (ET)
prediction are the depth in the soil profile fram which evaporation can
occur, and the depth of rainfall and irrigation that can be stored on the
plant canopy (interception storage). A figure in the manual indicates
that for southern Georgia, the depth of soil contributing to evaporation
is approximately 25 cm. Data pertaining to the interception storage
capacity of bahia grass was not included in the manual. Values presented
in the manual for various crops ranged fram 0.0-0.3 an. Assuming that
the storage capacity of bahia grass is relatively small, a value of 0.05
cm was assumed.
5.2.2 Crop related parameters
Characteristics of the crop(s) grown during the simulation period
are defined by parameters related to rooting depth, maximum canopy
coverage, and dates for emergence, maturation, and harvest. A rooting
depth for bahia grass was difficult to obtain. Most references to bahia
grass simply stated that it is a deep or very deep rooted plant. Based
on such generalizations and without more specific information, an active
rooting depth (depth containing 90% of root mass) of 91 cm (3 ft) was
selected. Based on observations at the research site, a maximum areal
coverage of the bahia grass foliage was estimated as 98 percent.
Dates of crop emergence, maturation, and harvest are used in the
model to define the development and growth of roots. Roots are assumed
to begin growing at crop emergence and reach the maximum rooting depth at
plant maturation. The roots stay at maximum depth until harvest when
they are set back to zero. These parameters are somewhat difficult to
define in relation to a perennial crop such as bahia grass. In view of
the use made of the selected dates, some artificial dates for emergence
and harvest could be selected. Since the emergence date really defines
the beginning of plant transpiration as a component of the water balance,
an emergence date of March 1 was selected to correspond to the time when
the grass begins to "green up" in the spring. Since the majority of the
roots already exist when the grass begins active growth in the spring, a
maturation date of April 1 was selected to reflect water extraction from
the full maximum rooting depth shortly after growth for the new year
begins. The harvest date was chosen to be December 1 which reflects the
approxiate date when the grass turns brown in the winter and active
5.2.3 Soil related parameters
PRZM will allow simulation of a larger section of the profile than
just the root zone. A soil core depth of 2.62 m was chosen to represent
the average profile depth beneath the application area. The model will
allow the user to select the number of opartments used to represent the
profile. Too many compartments will increase simulation times and too
few will lead to increased numerical errors in the solution of the
equations used within the model. The manual suggests that no fewer than
30 compartments be used. Thirty-five ccpartments were selected because
this gives a compartment depth that is one-half the crpartment depth in
GLEAMS for a root zone depth of 91 cman (see section 5.3.3).
PRZM4 and GLEAMS both normnnally operate on the assumption that any
soil water in excess of field capacity will drain into the next lower
compartment. This drainage will continue until any excess water drains
from the bottom of the simulated profile. This drainage is assumed to
occur within one day. In the PRZM model, this type of drainage is called
'free drainage'. PRZM has a feature to allow selection of a drainage
rate parameter that serves to slow down the percolation process. This is
an empirical constant and requires calibration. A figure is provided in
the manual whereby the user can make a first cut selection of the
drainage parameter based upon the soil type (sand, clay loam, etc.) and
the number of simulation coapartments selected as described above.
Observations at the experimental site clearly show that there is a delay
of 2-4 days from the time excess water is applied to the time that the
water table responds. Therefore a few simulations were made using the
restricted drainage option to compare with the experimental data. The
drainage rate parameter for this case was selected from the graph in the
manual for a sand soil with 35 ccmpartments to be 2.63 day-1.
5.3 Parameters Unique to ES
5.3.1 Evapotranspiration prediction
Potential evapotranspiration (PET) is computed in GLEAMS based on
mean monthly maximum and minimum temperatures and mean monthly solar
radiation. The mean monthly temperatures and solar radiation data were
obtained from the same weather station that was used for the pan
evaporation data input to PRZM (section 5.2.1). GLEAMS also
differentiates between soil evaporation and plant transpiration. A soil
evaporation parameter is required as a model input and the manual gives a
suggested value of 3.3 (no units) for this parameter for sands.
5.3.2 Crop related parameters
Crop related parameters which are different from PRZM are leaf area
index (IAI), and winter cover factor. The IAI data are used to reflect
changes in plant transpiration with stage of growth. In keeping with the
observed behavior of bahia grass discussed in section 5.2.2, the IAI of
bahia grass was assumed to be zero until March 1 of each year and to
return to zero on December 1st. The maximum IAI for bahia grass was
assumed to be similar to the IAI for pasture as presented in the CRFAMS
manual which had a maximum IAI of 3.0. The IAI for the bahia grass was
assumed to reach 2.0 on April 1 and to reach 3.0 May 1 where it remained
until it began to decline again to a value of 2.0 on Nov. 1.
The winter cover factor is used to moderate evaporation from the
soil surface during the winter due to presence of a bare surface or cover
crops. Since the soil surface at the site was not bare during winter,
the recommended value of the winter cover factor of 0.5 was used.
5.3.3 Soil related parameters
GLEAMS was designed to simulate only the root zone of a crop. The
vadose zone below the crop is ignored. Thus, the soil profile
description input to GLEAMS covers only the 91 cm root zone as opposed to
the 2.62 m profile input to PRZM. There are no soil parameters which are
significant to the results presented in this work which differ from those
discussed in section 5.1.2. In order to highlight one of the major
differences between the two models, a short discussion of the
computational layering of the soil profile in GLEAMS will be presented
here. GLEAMS divides the crop root zone into seven camputational layers.
The first layer is defined to be 1 cm in depth and the other 6 layers are
equal in thickness (layer 2 is 1 an less than layers 3-7) and together
account for the rest of the root zone. The 1 cm upper layer is
considered an important feature of CREAMS and GLEAMS because it increases
the sensitivity of erosion and runoff to chemical concentrations near the
soil surface. PRZM divides the entire profile into a specified number of
ccompartments of equal depth. For the case presented here, GLEAMS layers
would be 15 cm thick after the surface layer (91 can / 6 layers). For
FPRZM with a 2.62 m profile and 35 ccmpartments, each compartment will be
7.5 cm thick. Thus, PRZM may be better able to describe effects of
actual soil layering since properties of adjoining layers would not be
averaged over as large a layer as in GLEAMS. However the surface layer
in PRZM will also be 7.5 cm thick and pesticide storage will be averaged
over this whole layer effectively lowering the mass at the surface which
is available for transport with eroded sediment and runoff water.
RESULTS AND DISCUSSION
There are many areas in this study for which results can be
reported. The approach used in this chapter is to report significant
results relative to each area such as sample collection, sample analysis,
simulation results, etc., and then relate the individual results where
appropriate. A summary of activities at the field site is presented in
Table 6.1 for reference.
6.1 Data Collection
The site was visited on at least a weekly basis throughout the
study. At each visit, samples were collected from the monitoring wells
and soil solution samplers. The water table elevation was also recorded
for every monitoring well on the site and for several wells located
adjacent to the study area. Soil samples were collected on an irregular
basis during the study.
The procedure used for collecting samples from the groundwater
worked well in general. Sand intrusion into a few of the wells, however,
was a consistent problem. It was possible in some instances to fill the
250 mL sampling bottle with sand. This problem was overcome by slowly
raising the bottom of the sampling tube in order to stay above the sand.
These wells were pumped out using a 1.6 cm diameter suction line to
remove accumulated sand as needed. The problem could have been avoided
Table 6.1 Chronological summary of field site activities.
11/12/86 Apply 4.9 kg/ha of atrazine and alachlor, apply 5.1 an
of irrigation after herbicide application.
11/17/86 Apply 500 g of KBr (10 kg/ha Br-), collect soil solution
11/18/86 Collect soil solution samples. First soil samples taken
frcm top 5 an in application area.
Collect soil solution and soil samples.
apply 5.1 cm of
solution and well samples.
solution, well, and soil samples.
solution and well samples.
solution and well samples.
solution and well samples.
solution and well samples.
solution and well samples.
solution and well samples.
solution, well, and soil samples.
solution and well samples.
solution and well samples.
solution and well samples.
solution and well samples.
Collect soil solution, well, and soil samples.
Table 6.1 Contir
Collect soil solution and well samples.
Collect soil solution and well samples.
Collect soil solution and well samples.
Collect soil solution and well samples.
Apply 560 kg/ha of 5-10-15 to entire study site.
Collect soil solution and well samples, Very few wells
Collect soil solution and well samples, Apply 2.5 an o,
irrigation, apply approximately 17 kg/ha Br-, apply 2.
ancm of irrigation.
Collect soil solution and
Collect soil solution and
Collect soil solution and
Collect soil solution and
solution containing 500 g
an of irrigation.
Collect soil solution and
Collect soil solution and
well samples, apply 4.2 can of
well samples, apply 3 can of
well samples, apply 3 man of
KCl into well
pour 10 1 of
07-09, apply 2.7
well samples, apply 3.2 ancm of
well samples, apply 1.9 an of
samples, apply 0.9 an of
solution and well samples, apply 1.9 an of
solution and well samples, apply 5.2 an of
solution and well samples.
solution and well samples.
solution, well, and soil samples.
solution and well samples.
cxmercially available PVC well screening in the monitoring
No problems were observed with the operation of the soil solution
samplers. The volumes of water collected during sampling varied from 0
to greater than 50 mL. Even though the samples were pulled back to the
instrument trailer through as much as 35 m of 0.64 ancm ID polypropelene
tubing, the vacuum applied to the sample lines was sufficient to pull
very small droplets of water along the wall of the tube and into the
containers. No carryover of the tracers between sampling periods was
The tensicneters were difficult to maintain properly on the weekly
sampling schedule. Even when the tensicmeters were working, their
usefulness was limited by the short length of time during which they
could be observed. If an irrigation was initiated early in the morning,
a few of the 30 cm tensiameters might have responded by the end of the
day. Weekly observations of soil-water tension were insufficient for the
purpose of calculating the flux of water through the unsaturated zone.
6.2 Sample Analysis
Water samples collected from the monitoring wells were analyzed for
residues of atrazine and alachlor and concentrations of the inorganic
tracers utilized in this research (Br-, NO3-, and Cl-). The volume of
samples from the solution samplers ranged from 0 to more than 50 mL, with
the average sample volume being approximately 10 mL. Water samples from
the soil solution samplers were generally analyzed for the inorganic
tracers only. Most of the solution samples were not extracted for
residues of the herbicides due to the small sample volumes.
Approximately 50 of the soil solution samples were extracted and analyzed
for residues of the herbicides. Soil samples were extracted and analyzed
for residues of atrazine and alachlor. The presence of alachlor and
atrazine residues in the soil samples was confirmed in a composite of
soil sample extracts that was analyzed using gas chrcamiatography/mass
During extraction of water samples with hexane, an emulsion of
hexane and water sometimes formed. This would usually separate if
allowed to sit undisturbed for a few minutes. If the enulsion did not
separate, herbicide residues contained within the emulsion could have
been discarded. There was an apparent contamination of the glassware
used for water sample collection or extraction. Samples of deionized
water were extracted and they showed concentrations of atrazine of
approximately 1 gg/L. All glassware was routinely washed with a
cumnercial glassware detergent, triple rinsed with deionized water, and
rinsed with acetone. The contamination levels appeared to be consistent
in that water samples, in which atrazine was not expected, showed a
concentration of approximately 1 J/L. The source of the contamination
was not determined. Several attempts were made to isolate and identify
the source of the contamination, however these attempts were
unsuccessful. To account for the contamination, the measured
concentrations were reduced by 1 pg/L. Thus no concentrations of less
than 1 Mg/L are reported even though the sensitivity of the GC would
allow detection down to the range of tenths of a pq/L.
The gas crciduatograph performed well except that the N-P detector
response had a tendency to drift. This was accounted for by running
standards frequently and adjusting calculated concentrations as described
in the methods section.
The recoveries of atrazine and alachlor from spiked water samples
were nearly identical and ranged frcm 68 to 104% with a mean of 79% and a
standard deviation of 12 percent. The standard deviation would probably
be reduced if these analyses had been performed by an experienced
analyst. Experience contributes significantly to the ability to produce
accurate and consistent results.
The recoveries of atrazine and alachlor from spiked soil samples
could not be determined from the data collected. Moisture contents of
the soils used for spiking were not recorded. The spike levels were
based on the moist weight of the soil. After spiking, the soil was air-
dried for 1 day in order to allow the solvent from the herbicide
standards to evaporate prior to taking subsamples for analysis. The loss
of water from the samples during this time resulted in an increased mass
of pesticide per unit weight of wet soil. Since the initial moisture
contents were unknown, it was not possible to calculate the expected
concentrations in the air dried samples. Calculations based on the
spiking levels of the moist soil resulted in apparent recoveries that
exceeded 100 percent. It can be concluded from the data that recoveries
from the soil were high. When the missions in the soil recovery study
were discovered, the facilities used to perform extractions and analyses
were no longer available.
All concentrations (water and soil) reported in the following
sections are the analytically determined concentrations and have not been
adjusted for expected recovery rates.
6.3 Chemical Applications
Applications of the herbicides and bromide tracer were monitored to
determine the concentration of the chemicals in the application water and
the volume of water applied. Concentrations of the chemicals were
multiplied by the volume of water collected and divided by the opening
area of the collectors in order to determine the application rate on a
mass per unit area basis. Table 6.2 summarizes the measured chemical
application data. The intended application rates were presented in Table
The application of atrazine and alachlor to the designated
application area was not uniform as can be seen in Figures 6.1 and 6.2.