EVALUATION OF STEADY-STATE SOIL
CONCENTRATIONS
FOR PERSISTENT ORGANIC POLLUTANTS (POPS)
August, 1999
-------
EVALUATION OF STEADY-STATE SOIL
CONCENTRATIONS
FOR PERSISTENT ORGANIC POLLUTANTS (POPS)
United States Environmental Protection Agency
National Exposure Research Laboratory
Ecosystems Research Division
960 College Station Rd.
Athens, GA 30605-2720
Robert F. Carousel
EPA Task Leader
and
HydroGeoLogic, Inc.
1155 Herndon Parkway, Suite 900
Herndon, VA 20170
August, 1999
-------
TABLE OF CONTENTS
Page
1.0 INTRODUCTION 1-1
2.0 STUDY METHODOLOGY 2-1
3.0 BASE SIMULATIONS 3-1
3.1 PRZM-3 MODEL 3-1
3.2 SITE SELECTION PROCESS 3-1
3.3 PESTICIDE CHEMICAL DATA 3-7
3.4 DESIGN OF SIMULATIONS 3-8
3.5 RESULTS OF SIMULATIONS 3-9
4.0 SENSITIVITY ANALYSIS 4-1
5.0 DEVELOPMENT OF PREDICTIVE METHODOLOGY 5-1
5.1 INITIAL STATISTICAL ANALYSIS OF PRZM-3 DATA 5-1
5.2 ADDITIONAL PRZM-3 SIMULATIONS 5-5
5.3 FINAL STATISTICAL ANALYSIS OF PRZM-3 DATA 5-8
5.4 APPLICABILITY OF THE PREDICTIVE METHODOLOGY
FOR Css 5-8
5.5 APPLICABILITY OF THE PREDICTIVE METHODOLOGY
FOR TSS 5-14
5.6 UNCERTAINTY ANALYSIS 5-17
6.0 SUMMARY AND CONCLUSIONS 6-1
7.0 REFERENCES 7-1
APPENDIX A DOCUMENTATION OF SOURCES OF DATA USED FOR CREATION
OF PRZM-3 INPUT FILES
F:\Projeco\EPA\EPA_OOt_2I103\popsl.wpd 1 HydroGeoLogic. Inc. 09/02/99
-------
LIST OF TABLES
Page
Table 3.1 Summary Information for 32 Sites Selected for Detailed
PRZM-3 Simulations 3-6
Table 3.2 Chemical Data for Investigated POPs 3-7
Table 3.3 Computed Concentrations in mg/kg at t= 100 years: Chlordane and DDE 3-22
Table 3.4 Css and Tss: DDD and Kepone 3-23
Table 3.5 CM and T^: DDT, Dieldrin, Methoxychlor, and Toxaphene 3-24
Table 4.1 Results of Sensitivity Analyses for Dieldrin 4-2
Table 4.2 Results of Sensitivity Analyses for Kepone 4-3
Table 5.1 Summary of Candidate Three-Variable Functions to Predict C^ 5-4
Table 5.2 Summary of Candidate Two-Variable Functions to Predict Css 5-5
Table 5.3 Ks Values for "New" POPs 5-6
Table 5.4 Summary of Additional PRZM-3 Simulations Conducted to Support
Development of Screening Level Methodology 5-6
Table 5.5 Summary Statistics from Monte Carlo Analyses 5-18
LIST OF FIGURES
Page
Figure 3.1 Crop Distribution by % Land Use for Corn, from 1992 NRI data 3-2
Figure 3.2 Crop Distribution by % Land Use for Cotton, from 1992 NRI data 3-2
Figure 3.3 Crop Distribution by % Land Use for Soybeans, from 1992 NRI data .... 3-3
Figure 3.4 Crop Distribution by % Land Use for Wheat, from 1992 NRI data 3-3
Figure 3.5 Average Annual Precipitation in the United States 3-4
Figure 3.6 32 Sites Selected for Detailed PRZM-3 Simulations 3-5
Figure 3.7 Typical PRZM-3 Simulation Results - Toxaphene 3-10
Figure 3.8 Time Required to Reach Steady State - Toxaphene 3-11
Figure 3.9 Typical PRZM-3 Simulation Results - Chlordane 3-12
Figure 3.10 Typical PRZM-3 Simulation Results - DDD 3-13
Figure 3.11 Typical PRZM-3 Simulation Results - DDE 3-14
Figure 3.12 Typical PRZM-3 Simulation Results - DDT 3-15
Figure 3.13 Typical PRZM-3 Simulation Results - Dicofol 3-16
Figure 3.14 Typical PRZM-3 Simulation Results - Dieldrin 3-17
Figure 3.15 Typical PRZM-3 Simulation Results - Endosulfan 3-18
Figure 3.16 Typical PRZM-3 Simulation Results - Furan 3-19
Figure 3.17 Typical PRZM-3 Simulation Results - Kepone 3-20
Figure 3.18 Typical PRZM-3 Simulation Results - Methoxychlor 3-21
Figure 4.1 Sensitivity Analysis Results for Kh: Concentration Profiles for Dieldrin
at Site 4 4-4
F:\Prejects\EPA\EPA_OW_21103\popsl.wpd
11
HydroGeoLogic. Inc. 09/02/99
-------
LIST OF FIGURES
Page
Figure 4.2 Sensitivity Analysis Results for Ks: Concentration Profiles for Dieldrin
at Site 9 4-5
Figure 4.3 Sensitivity Analysis Results for Kd: Concentration Profiles for Kepone
at Site 22 4-6
Figure 4.4 Sensitivity Analysis Results for CN: Concentration Profiles for Kepone
at Site 28 4-7
Figure 5.1 Behavior of Numerator Function in Equation 5.1 5-9
Figure 5.2 Behavior of Denominator Function in Equation 5.2 5-9
Figure 5.3 Region of Applicability of Predictive Methodology 5-10
Figure 5.4 Computed vs. Predicted Values of Css for Selected POPs 5-11
Figure 5.5 Computed vs. Predicted Values of Cssfor Kepone; Ks = 0/day 5-12
Figure 5.6 Computed vs. Predicted Values of Cssfor Toxaphene;
Ks = 1.92E-04/day 5-12
Figure 5.7 Computed vs. Predicted Values of Cssfor POP A; Ks = 4.70E-04/day . . 5-13
Figure 5.8 Computed vs. Predicted Values of C^for Methoxychlor;
Ks = 1.89E-03/day 5-13
Figure 5.9 Computed vs. Predicted Values of TSS for Selected POPs 5-14
Figure 5.10 Computed vs. Predicted Values of T^ for Kepone; Ks = 0/day 5-15
Figure 5.11 Computed vs. Predicted Values of T^ for Toxaphene;
Ks = 1.92E-04/day 5-15
Figure 5.12 Computed vs. Predicted Values of TSS for POP A; Ks = 4.70E-04/day . . 5-16
Figure 5.13 Computed vs. Predicted Values of TSS f°r Methoxychlor;
Ks = 1.89E-03/day 5-16
Figure 5.14 Cumulative Distribution Function for Css at Point A 5-19
Figure 5.15 Cumulative Distribution Function for Css at Point B 5-19
Figure 5.16 Cumulative Distribution Function for C^ at Point C - 5-20
Figure 5.17 Cumulative Distribution Function for Css at Point D 5-20
Figure 6.1 Effect of Various Application Rates on Predicted Concentrations;
Chlordane 6-1
Figure 6.2 Effect of Various Application Rates on Predicted Concentrations;
Dieldrin 6-2
Figure 6.3 Effect of Various Application Rates on Predicted Concentrations;
Kepone 6-2
F:\Projttts\EPA\EPA_004_21103\popsl.wpd *" HydroGeoLogic. Inc. 09/02/99
-------
1.0 INTRODUCTION
The U.S. Environmental Protection Agency (EPA) is continually faced with regulatory issues
concerning the migration of organic and inorganic chemical constituents to and through multimedia
systems (water, soil, air, etc.). Each of these issues requires that the potential risk to human
health and various ecosystems be evaluated. Recently, much of this attention has been focused
on exposure to persistent organic pollutants (POPs). Historically, these POP compounds have
been associated with application hi an agricultural crop setting. These POP chemicals are of
concern to the EPA due to their classification as "endocrine disrupters," which adversely impact
the glandular system of the human body.
The purpose of this report is to present the results of a study conducted by HydroGeoLogic to
investigate the behavior of a list of specified POPs, and to develop a predictive screening level
methodology for use with other, similar chemical compounds. Detailed analysis was conducted
for the following eleven POPs:
• Chlordane
DDD
• DDE
DDT
• Dicofol
• Dieldrin
• Endosulfan
• Furan
• Kepone
• Methoxychlor
• Toxaphene
The above compounds have historically been associated with usage as pesticides hi agricultural
settings. The analysis of these compounds primarily involved the design and execution of a series
of long term simulations using Version 3.0 of the EPA's Pesticide Root Zone Model (PRZM-3).
The intent of the individual simulations was to determine whether the listed POPs demonstrated
a tendency to reach a steady state concentration hi the upper soil horizon as a result of annual
application, and, if so, to determine the time required for steady state concentrations to be
achieved. Results of these simulations were used to support the development of a predictive
screening level methodology which can be used for other POPs based on certain readily available
chemical parameters. The screening level methodology, if used as described hi this report, will
allow rapid determination of whether a specific compound is likely to demonstrate a tendency to
be persistent hi the environment.
The remainder of this report will discuss the study methodology, the results of long term PRZM
simulations for the listed POPs, the development of the screening level methodology, and a
discussion of the uncertainty associated with application of the screening level methodology.
F:\Projects\EPA\EPA_004_21103\popsl.wpd 1-1 HydroGeoLogic. Inc. 09/02/99
-------
2.0 STUDY METHODOLOGY
The approach used to develop the screening level methodology involved the application of a
detailed study methodology. Each portion of the study methodology will be described in more
detail in subsequent sections of the report. The intent of this section is to briefly summarize the
nature of the approach used for this investigation.
The basic backbone of this study involved detailed, long term simulations using PRZM-3.
Simulations were conducted for each of the 11 study POPs at each of 32 sites distributed across
the contiguous United States. The site selection process was designed to provide a group of
simulation sites which were generally representative of the wide variety of cropping practices,
climatology, and soils conditions in agricultural regions. Simulations assumed an annual unit (1
kg/ha) application of a specific POP, and were conducted for a minimum of 100 years.
PRZM-3 simulations considered the annual pesticide application as the source term, and erosion,
runoff, volatilization, decay, and leaching as the primary loss terms. For persistent compounds
(those which exhibit a tendency to increase in concentration over time), the magnitude of the loss
terms increases as the concentration increases hi the upper layer of the soil. Since the assumed
application rate is constant, many compounds eventually reach a steady-state condition where the
annual loss term matches the source term. Simulation results were analyzed to determine 1)
whether a specific compound exhibited a tendency to reach a steady state concentration, and if so,
2) what the magnitude of the steady state concentration was, and 3) how much time was required
for steady-state conditions to be achieved.
Following completion of the base simulations, a series of sensitivity analyses were conducted to
determine which of the input variables in the PRZM-3 simulations had the most effect on
computed soil concentrations. A rigorous statistical analysis was then conducted to determine the
best method to predict steady state concentrations and time required to reach steady state
conditions. The results of the sensitivity analyses were used to guide the development of the
predictive screening level methodology and the subsequent uncertainty analyses. The uncertainty
analysis was performed to help assess the limitations of the screening level methodology.
9 1
F:\Projects\EPA\EPA_004_21103\popsl.wpd Z-~1 HydroGeoLogic, Inc. 09/02/99
-------
3.0 BASE SIMULATIONS
3.1 PRZM-3 MODEL
The Pesticide Root Zone Model (PRZM-3) is a one-dimensional, dynamic, compartmental model
that can be used to simulate chemical movement in unsaturated soil systems within and
immediately below the plant root zone (EPA, 1998). It has two major components - hydrology
(and hydraulics) and chemical transport. The hydrologic component for calculating runoff and
erosion is based on the Soil Conservation Service (SCS) Curve Number (CN) technique and
Modified Universal Soil Loss Equation (MUSLE). Evapotranspiration is divided among
evaporation from crop interception, evaporation from soil, and transpiration by the crop. Water
movement is simulated by the use of generalized soil parameters, including field capacity, wilting
point, and saturation water content. The chemical transport component can simulate pesticide
application on the soil or on the plant foliage. Dissolved, adsorbed, and vapor-phase
concentrations in the soil are estimated by simultaneously considering the processes of pesticide
uptake by plants, surface runoff, erosion, degradation or transformation, volatilization, foliar
washoff, advection, dispersion, and chemical advection due to sorption.
3.2 SITE SELECTION PROCESS
The site selection process was designed to allow identification of sites in geographically diverse
locations for which detailed PRZM-3 simulations could be conducted. Since the POPs investigated
in detail are historically associated with pesticide usage in agricultural settings, the sites selected
were chosen from areas of primary agricultural production. Simulations were conducted for four
major crops: corn, cotton, soybeans, and wheat. Figures 3.1 through 3.4 show crop distribution
maps for each of these crops in the United States. These figures were based on data published in
the National Resource Conservation Service (NRCS) 1992 National Resources Inventory (NRI).
The darker shadings refer to successively higher percentages of land in a given county which is
dedicated to production of the respective crops. Using these maps, and a map of average annual
precipitation (see Figure 3.5), eight sites were selected for each crop type, resulting in a total of
32 sites identified for detailed simulations.
While the PRZM-3 simulations were one dimensional, the sites were selected on a county basis.
PRZM-3 input files were generated using parameter values that were determined to be the most
representative of average conditions across a selected county. More information on how this was
done is presented hi Section 3.4.
Figure 3.6 shows the sites that were selected for detailed PRZM-3 simulations. Table 3.1 presents
summary information for these sites.
F:\Projeca\EPA\EPA_004_21103\popsl.wpd -3""1 HydroGeoLogic, Inc. 09/02/99
-------
Figure 3.1 Crop Distribution by % Land Use for Corn, from 1992 NRI data
Figure 3.2 Crop Distribution by % Land Use for Cotton, from 1992 NRI data
F:\Projece\EPA\EPA_004_21103V21103«.wpd
3-2
HyckoGeoLogjc, Inc 08/31/99
-------
Figure 3.3 Crop Distribution by % Land Use for Soybeans, from 1992 NRI data.
Figure 3.4 Crop Distribution by % Land Use for Wheat, from 1992 NRI data.
F:\Proj«0\EPA\EPA_004_21103U1103*. wpd
3-3
HydroGeoLogic, Inc. 08/31/99
-------
'
i
-------
-------
Table 3.1
Summary Information for 32 Sites Selected for Detailed PRZM-3 Simulations
Site?
1
i
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
County. State
Jackson, IN
Kent, MD
Hamilton, NE
Cayuga, NY
Sumter, SC
Hanson, SD
Farmer, TX
Dane, \VI
Maricopa, AZ
Kings. CA
Dooly, GA
Franklin, LA
Edgecombe, NC
Crockett, TN
Dawson, TX
Wiilacy, TX
Burke, GA
Wabash, IL
Hamilton, IA
Pornte Coupee, LA
Clay, MX
Tunica, MS
Salem, NJ
Union, OH
Sutler, CA
Hancock, IL
Mitchell, KS
Daniels, MT
Crop
Cora
Corn
Corn
1 Corn
Cora
Corn
Com
Cora
Cotton
Cotton
Cotton
Cotton
Cotton
Cotton
Cotton
Cotton
Soybeans
Soybeans
Soybeans
Soybeans
Soybeans
Soybeans
Soybeans
Soybeans
Wheat
Wheat
Wheat
Wheat
Soil Name
Peoga
Matapeake
Holder
Honeoye
Norfolk
Clarno
Olton
Piano
Momoli
Armona
Tifton
Calhoun
Goldsboro
Loring
Amarillo
Raymondville
Doman
Roby
Brownton
Commerce
Bearden
Sharkey
Mattapex
Blount
Capay
Virden
Harney
Cherry
HSG
C
B
B
B
B
B
C
B
B
C
B
D
B
C
B
D
B
C
C/D
C
C
D
C
C
D
B/D
B
C
1 Met Station
Louisville, KY
Wilmington. DE
Grand Island, NE
Syracuse, NY
Columbia, SC
Sioux Falls, SD
Amarillo, TX
Madison, WI
Phoenix. AZ
Fresno. CA
Macon, GA
Jackson. MS
Raleigh- Durham, NC
Memphis, TN
Midland, TX
Brownsville, TX
Augusta, GA
Evansville, IN
Des Moines, I A
Baton Rouge, LA
Fargo, ND
Memphis, TN
!
Wilmington, DE
Columbus, OH
Sacramento, CA
Burlington. LA
Ccncordia, KS
Williston, ND ^
3-6
-------
Table 3.1 (continued)
Summary Information for 32 Sites Selected for Detailed PRZM-3 Simulations
Site#
29
30
31
32
County, State
Fulton, OH
Lexington, SC
Travis, TX
Lincoln, WA
Crop
Wheat
Wheat
Wheat
Wheat
Soil Name
Mermill
Nason
Houston Black
Bagdad
HSG
B/D
C
D
B
Met Station
Toledo, OH
Columbia, SC
Austin, TX
Spokane, WA
3.3 PESTICIDE CHEMICAL DATA
Table 3.2 shows the chemical data used for detailed PRZM-3 simulations for each of the study
POPs. These data were obtained from a variety of sources as documented in the table.
Table 3.2
Chemical Data for Investigated POPs
POP
Chlordane
DDD
DDE
DDT
Dicofol
j Dieldrin
Endosulfan
Fur an
Kepone
Methoxychlor
Toxaohene
Da
(cmVday)
1019.52
1347.84
1244.16
1183.68
3473.281
1080
993.6
8985.6
6912
1347.84
1002.24
K,
(dim)
1.99E-03
1.63E-04
8.55E-04
3.31E-04
4.1E-071
6.19E-04
4.58E-04
0.2207
1.04E-06
6.46E-04
2.45E-04
ENPY
(kcal/mole)
20
20
20
20
20
20
20
6.562
20
20
20
K,,
(day1)
1.03E-07
1.29E-04
0.00
1.01E-03
1.68E-OP
1.73E-04
6.44E-01
0.00
0.00
1.92E-03
2.68E-04
K,
(day1)
0.00
6.85E-05
0.00
1.64E-04
1.68E-013
1.73E-04
2.05E-04
0.00
0.00
1.89E-03
1.92E-04
half-life
(soil, days)
cc
10,117
oo
4,226
4
4006
3380
CC
CO
367
3609
K«
(cm3/g)
7.76E+05
7.76E+05
4.37E+06
3.89E+06
6.06E+03
1.20E+05
3.55E+03
l.OOE+01
1.41E+04
7.94E+04
2.04E+04
'from the SPARC (Spare Performs Automated Reasoning in Chemistry) properties calculator
'from the Handbook of Chemistry- and Physics, 7921 Edition. 1995-96
3from the NRCS ARS Pesticide Properties Database for pH of 7
•from the NRCS ARS Pesticide Properties Database (selected value by ARS)
ENPY
K.
K,
vapor phase molecular diffusivity coefficient
Henry's constant
enthalpy of vaporization
liquid phase decay rate
solid phase decay rate
orsanic carbon coefficient
F:iprojec-J' £PA\E?A_«W_2 II03' Peps 1 .wp
-------
Values for Da and Kh were obtained from the EPA Region VI Delis ting Spreadsheets (except as
noted). Values for ENPY were obtained from the PRZM-3 User's Manual (except as noted).
Values for Kw and K^ were calculated based on 25 °C and rate constants from the Hazardous
Waste Identification Rule (HWIR) publication (EPA, 1995) (except as noted). Values for K^ were
taken from the HWIR publication (except as noted).
3.4 DESIGN OF SIMULATIONS
For each selected site, a generic PRZM-3 input file was generated. This generic file was then
modified as required to reflect the chemical data for a specific POP. The generic input file was
generated using the capabilities of the Pesticide Assessment Tool for Rating Investigations of
Transport (PATRIOT) (Imhoff et al, 1993). PATRIOT is a decision support system that allows
the user to interact with a series of databases (including rainfall, soils, cropping, and pesticides).
In order to generate the generic PRZM-3 input data set for each site, the PATRIOT system was
used to identify the most predominant soil type for the crop of interest, and to identify the nearest
meteorological station. The option to generate a PRZM (version 2) input file was then selected.
This file was then modified as required to reflect the changes in the input data requirements for
the PRZM-3 model. PATRIOT was used to define the following variables in the final versions
of the input files used for detailed simulations: crop emergence, maturation, and harvest dates; the
curve number (CN) data used for runoff calculations; total depth of soil core; and soil horizon data
including bulk density, initial soil water content, field capacity, wilting point, and percent organic
carbon.
Other sources of data were used to build the generic PRZM-3 input files, including the PRZM-3
User's Manual and the soils database included within the NRCS 1992 National Resources
Inventory. Once completed, the generic input files for each site were then modified for each POP,
using the chemical data in Table 3.2. Appendix A provides documentation for the major
assumptions made and for the sources of data used for the base PRZM-3 simulations.
Since study simulations were designed to be at least 100 years long, the option to use the 10 year
record in the PATRIOT rainfall database was not used. Instead, once a meteorological station had
been identified, the data for that station were downloaded from the EPA's Center for Exposure
Assessment Modeling (CEAM) site. The met stations selected are shown in Table 3.1. Most
stations had 36 years of data; the available data for each station were successively appended to
provide a meteorology input file with 100 years of data.
The PRZM-3 irrigation option was selected for corn sites 3 and 7; cotton sites 9, 10, and 16; and
wheat site 25. Irrigation was chosen for these sites based on United States Department of
Agriculture maps showing percent of cropland in irrigation by county for 1992 (see web site
address "http://www.nhq.nrcs.usda.gov/land/meta/m2289.html").
Pesticide applications were assumed to be 1.0 kg/ha, applied on the same date to the same crop
for ever}' year of the simulations. All pesticides were assumed to be soil applied, using a default
incorporation depth of 4 cm. The processes of biodegradation and plant uptake of pesticides were
not considered.
j p
-------
3.5 RESULTS OF SIMULATIONS
The results of the PRZM-3 simulations are presented in this section. Soil concentrations are
presented throughout as average concentrations over the top 15 cm of the soil column. The 15 cm
depth was judged to be reasonable to support risk assessments in which primary exposure routes
are ingestion and dermal contact. The leaching of POPs below the 15 cm depth was considered
to be a loss term, although that does not necessarily imply transport to the groundwater system.
HydroGeoLogic wrote a simple post processor which parsed the PRZM-3 output file and extracted
the total, adsorbed, dissolved, and gaseous concentrations for the top 15 cm at the end of every
month. Graphs of total concentration vs time were prepared and investigated. Simulations beyond
100 years were then conducted where required to allow concentrations to reach steady state.
These were conducted by repeating the original simulation (from 0 to 100 years), but with initial
soil pesticide concentrations set equal to the concentrations predicted at tune = 100 years. At
every site, the modeled soil compartment thickness was 1 cm over the depth range from which
concentration data were investigated.
Figure 3.7 shows typical simulation results for toxaphene. The figure shows end of month average
soil concentration versus tune. The results shown are taken from Site 1, but are representative
of the behavior of toxaphene at all sites. Similarly, all other results shown hi this section are
typical for the respective POPs and are taken from Site 1.
Figure 3.7 demonstrates typical behavior for the more "well-behaved" POPS of those investigated
in detail. Initially, the sum of all of the loss terms associated with erosion, runoff, volatilization,
decay, and leaching is much lower than the annual unit source term. Average soil concentrations
increase rapidly, but the increase in concentration leads to higher loss terms, which are each a
function of soil concentration. Eventually, the annual loss term approximately balances the annual
source term, and steady state concentrations are achieved.
From Figure 3.7, it appears that steady state concentrations are achieved sometime after about 60
years, but because of the monthly variation in concentration it is not immediately clear what we
mean by steady state. It was determined that an objective means to determine "time to reach
steady state" was required, so that simulation results could be processed objectively and
consistently.
The methodology adopted to determine the time required for steady state conditions to be achieved
involved plotting the percent change in the moving average of the computed soil concentration
data. Theoretically, when the value of the percent change in moving average reaches zero, steady
state conditions have been achieved. Durations of from one to five years for the moving average
computation were considered. Figure 3.8 shows these data for the toxaphene plot presented in
Figure 3.7 for one year and five year moving averages. Successively longer moving average
periods reduce the "noise" in the data and produce smoother and more consistent curves. A
moving average period of five years was adopted in the subsequent determination of time required
to reach steady state conditions for all POPs at all sites. Based on this moving average period, the
time required for toxaphene concentrations shown in Figure 3.7 to reach steady state conditions
(hereinafter designated Tss) is 74.08 years, at which time the average concentration is 4.77 mg/kg.
-i q
F:'.P70jeca'.EPA\EPA OW_21I03'?opsl.»Td -J'J Hyd.-oGcoLogic. !r.c. C9t£99
-------
OJ
I
>—'
o
05
U)
£
X3
|Hl*
O
£L
fD
sr
S
6.0
5.0
^ c
| C 3.0
6
2.0
1.0
0.0
o
* ""
Css = fl,77 rrra/l
'408 years
10 20 30 40 50 60 70
Number of Years
80
90
100
-------
0.05
£
H
o
I
t/3
rt
m
H
o
H
te
0)
-0.03
Number of Years
-------
i j
30 T~
25
O)
*20
.£
c
O 'C
+3
s
*-»
c
O
O
O
I)
0
10 20 30 40 50 60
Number of Years
70
80
90
100
00.4.21103\pop»l.wpd
3-12
HydfoGeoLogic, Inc. 09/02^99
-------
I
UJ
n>
o
e.
re
t/i
d
d
d
14
12
D)
23 C
I CD
I °
I O
O
10
E
s .E s
c
>* o
£C •«»«•
§ *m
^ 2 6
0
0 20 40 60 80 100 120 140 160 180 200
Number of Years
-------
10
f
•
H*
^H
o'
EL
3
5
C/3
3
|
o'
%
C/5
C
ti
1
w
D)
^
D)
E
c
• •MM
C
0
1o
4>J
c
0)
o
c
o
o
1 VX
9
8
7
6
5
4
3
2
1
0
X
x
,x
0
10 20 30 40 50 60 70
Number of Years
80
90
100
3
fi
-------
o
C/3
o
a
»
(D
C/J
5T
I
a
o
H
7
6
O)
.S4
C
o
3
C
0)
O
z:
a C
O 2
0
0
10 20 30 40 50 60
Number of Years
70
80
90
100
-------
r?
O
3
o
3
"
>*1
I—<•
S o)
£
H C
O •"-«
0.20
0.18
0.16
0.14
0.12
O 0.10
O)
E
E O
£ O
§
O
O
Hi*
o
0.06
0.04
0.02
0.00
0
10
Nurriter of Years
15
20
-------
6
a
8
CD
O)
* 0)4
H E
* c
S .2 3
a c
£ CD
£.02
O f—
3 lu
» O
rt> / \
I, °
cr
o 1
0
0
10 20 30 40 50 60
Number of Years
70
80
90
100
-------
oo
0.00
0
10
Number of Years
15
20
-------
"J3
05
"2.
o'
S
ation
I
0.16
0.14
0,
^ (13
0.12
O)
E 0.10
.E
c
O 0.08
8
0
S
0
eouco
Resul
0.02
0.00
A __ /k
l\
0
10
Nunft^r of Years
15
20
-------
to
o
re
U)
O
e.
s^-<
§•
o
3
I?
t/j
a
GT
re
T3
o
3
30
25
Ol
c
.2 15
4«j
2
4_j
c
0)
g 10
LM
o
o
0
0
10
20
30 40 50 60
Number of Years
70
80
90
100
-------
0.90
to
3
n
H*
00
H
12.
o'
B.
Sa
N
£J
i
C/3
£
P
rt-
o"
a
<£.
o
O
s
o*
D)
^)
E
c
mrnrnm
C
o
"f"J
2
IE:
o
o
c
o
O
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
0
10
Number of Years
15
20
-------
The magnitude of the steady state concentration (hereinafter designated Css) is therefore assumed
to be 4.77 mg/kg. (Note that while Css and Tss data are typically reported quite precisely, this is
an artifact of the software used to generate these numbers. The accuracy of these data are
probably limited to 1 or 2 significant figures; for the above example, it would therefore be
acceptable to assume that Css = 4.8 mg/1 and Tss = 74 years.)
Inspection of Figure 3.7 and Figures 3.9 through 3.18 demonstrates that there are several types
of behavior among the investigated POPs (a discussion of each compound is presented later). The
most persistent compounds are chlordane and DDE. These compounds did not show a tendency
to reach steady state, even for simulations as long as 400 years. Both of these POPS have high
K
-------
The POPs DDD and kepone are also strongly persistent. However, these compounds show a
tendency to reach steady state concentrations after an extended period of time, often in excess of
100 years. Values of Css and Tss for these compounds are presented in Table 3.4.
Table 3.4
C^ and T^: DDD and Kepone
Site
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23 •
24
25
26
27
28
29
30
31
32
DDD
C^, mg/kg
12.65
13.13
10.69
12.17
11.09
15.97
12.51
13.43
12.98
11.29
13.04
9.56
13.35
9.96
12.77
12.04
12.18
14.69
13.84
11.93
15.39
11.91
12.69
11.43
11.73
15.56
12.81
14.49
14.6
10.75
11.94
12.61
TSJ, yrs
116.25
95.42
108.67
95.58
97.67
113.58
140.92
84.75
162.5
161.75
134
76.25
112.17
98.42
109.58
141.58
112.67
138.08
86.42
97.75
102.42
98.75
95.67
93.08
114.83
100.58
87.58
104.67
100.67
85.83
84.83
74.17
Kepone
C^, mg/kg
27.69
21.67
22.59
51.99
13.16
82.52
20.91
74.41
7.11
14.97
5.69
37.20
44.91
16.97
31.42
19.95
3.14
22.34
> 156
30.36
> 173
36.87
32.31
44.33
20.27
104.34
66.31
25.93
78.54
26.12
59.09
58.96
Ts,yrs
97.33
60.50
84.25
161.33
54.08
228.42
81.58
199.58
28.00
67.08
23.17
135.58
148.50
62.75
95.00
82.08
12.42
72.42
> 400
97.75
> 400
126.08
95.92
140.17
70.92
293.83
189.75
87.83
241.58
91.58
185.50
174.17
F:'' Projeca'.EPA1 E?A_OOi_2!! 03' pops I .*-pd
HydroGeeLog:;, I-c. WOT99
-------
The next group of POPs demonstrated a consistent tendency to reach steady state conditions in less
than 100 years. This group includes DDT, dieldrin, methoxychlor, and toxaphene. In addition,
the magnitude of the Css for a specific POP did not show as much variation from site to site as for
the more persistent POPs. Values of Css and TS3 for these compounds are presented in Table 3.5.
Table 3.5
CK and TM: DDT, Dieldrin, Methoxychlor, and Toxaphene
Site
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
3Q
31
32
DDT
Cs,mg/kg
5.98
6.19
4.87
4.96
5.51
7.80
5.85
6.60
5.94
4.95
6.13
4.47
6.53
4.64
6.52
5.48
5.46
6.98
6.99
5.83
7.58
5.90
6.30
5.73
5.71
7.63
6.38
7.42
7.10
5.28
5.98
6.38
TB, yrs
52.25
48.75
44.83
27.58
52.33
68.42
75.42
42.67
63.83
62.25
63.92
36.25
63.92
61.00
62.58
65.92
44.58
59.42
49.75
53.25
63.50
62.08
58.50
55.42
69.08
61.75
47.58
63.33
69.58
48.75
47.83
38.17
Dieldrin
Cs, mg/kg
5.00
5.40
4.55
5.30
4.31
5.76
5.02
5.92
4.63
4.65
4.58
4.11
5.16
4.38
4.32
5.06
4.23
5.27
5.82
5.08
5.78
5.04
5.24
4.85
4.68
5.93
5.28
5.08
5.32
4.71
5.11
5.43
T^yrs
74.50
58.50
63.92
57.42
83.58
75.83
75.42
63.58
92.00
81.33
83.00
95.58
74.42
62.08
80.42
76.75
62.50
85.83
49.75
61.75
65.83
62.42
59.58
68.50
69.92
77.75
64.83
81.42
69.58
50.17
48.83
72.42
Methoxychlor
CB, mg/kg
0.56
0.58
0.51
0.64
0.47
0.67
0.55
0.70
0.47
0.51
0.48
0.52
0.59
0.48
0.47
0.57
0.42
0.59
0.70
0.60
0.68
0.59
0.60
0.58
0.53
0.67
0.60
0.57
0.59
0.58
0.60
0.63
Ts, yrs
13.58
11.67
11.67
13.50
11.83
13.42
12.33
12.42
16.17
16.50
12.42
10.75
11.67
9.50
12.92
15.42
11.58
13.67
12.67
13.58
13.75
12.00
11.67
10.50
12.33
11.83
15.83
14.42
14.83
11.58
10.17
11.08
Toxaphene
C^mg/kg
4.77
5.19
4.40
4.93
3.97
5.25
4.85
5.49
3.64
4.40
3.24
4.23
4.78
4.26
4.01
4.88
2.14
4.96
5.34
4.90
5.09
4.77
5.00
4.55
4.44
5.32
4.86
4.56
4.78
4.64
4.86
4.98
Ta,yrs
74.08
54.75
65.58
57.42
67.33
73.75
69.67
63.58
33.75
61.75
27.58
56.33
79.67
62.08
66.17
74.50
15.42
63.58
60.83
71.67
76.58
70.50
57.75
69.42
68.08
77.75
66.42
64.67
66.58
59.67
64.42
80.00
EPA\EPA_OOiJ!!IC3'pcpsi.v.TK!
3-24
-------
Finally, a group of compounds exhibited no tendency to reach steady state conditions. For these
compounds, the loss terms were sufficient to reduce the average concentration in the top 15 cm
to zero (or nearly to zero in the case of endosulfan). This group includes dicofol, endosulfan, and
furan. Since these compounds exhibited no tendency towards persistence (at least as defined for
this study), computed concentration results are not presented here.
Following is a brief discussion of the PRZM-3 simulation results for each compound, presented
in order of persistence behavior (as discussed above):
Chlordane
Chlordane simulations, even those carried out as far as 400 years, produced concentration vs tune
plots that were nearly linear (similar to Figure 3.9). Chlordane had one of the highest K^ (and
consequently, Kj) values of the investigated POPs, indicating a strong tendency to remain hi the
solid or adsorbed phase. Since its solid phase decay rate (KJ is zero, soil concentrations continue
to increase year after year. Note that chlordane has a relatively high Kh value among the
investigated POPs; volatilization losses were typically the highest loss term in the PRZM-3
simulations. However, because the compound remains primarily hi the solid phase, computed
volatilization losses were relatively minor.
DDE
DDE was the only one of the investigated POPs for which the PRZM-3 parent daughter simulation
capabilities were employed. The results of these simulations were not included in the development
of the predictive screening level methodology, but are reported here for purposes of project
documentation.
We assumed that DDT degradation would produce decay products of 40% DDD and 40% DDE
(DDD was included in the parent daughter simulations to preserve mass balance). Since the DDT
K; value is 1.64E-04/day (see Table 3.2), the simulated rate for production of solid phase DDE
from degradation of DDT was therefore 0.40(1.64E-04/day) or 6.56E-05/day. Similarly, the
simulated rate for production of liquid phase DDE from degradation of DDT was therefore
0.40(1.01E-03/day) or 4.04E-04/day. The K, and K^ rates for DDE are both zero.
Simulation results for DDE indicate a high degree of persistence, similar to chlordane. In fact,
the K^ value for DDE was the highest of the POPs investigated. This compound demonstrates
a strong tendency to remain in the solid phase and not decay. The highest loss term for this
compound in PRZM-3 simulations was typically erosion.
DDD
DDD, like DDE, is a daughter product of DDT. However, DDD has historically been used as
a pesticide. Therefore, it was simulated in a fashion similar to all of the other investigated POPs.
DDD has a K^ value equal to that of chlordane, but, unlike chlordane, it has a nonzero Kj value
of 6.85E-04/day. While this value is relatively low, the decay loss term was typically the highest
for this compound in the PRZM-3 simulations. The low rates of decay account for the long Tss
values for this compound.
F:\Projec3-EPA\EPA_OW_21103.pops l.w?d J>"-~) " " H>droCecU>g;c. Ire. 09'02/99
-------
Kepone
Kepone is another compound that exhibited long Tss values for many sites. It has a zero decay rate
for both solid and liquid phases. Unlike DDD, however, kepone has a relatively low K^ value,
which means it is relatively more likely to be hi the liquid or gaseous phase. Of the true loss
terms considered, volatilization losses were typically highest in the PRZM-3 simulations.
However, because kepone is more readily transported through the soil column in the liquid phase
than some of the other POPs, some of the "loss" for this compound is associated with leaching.
As shown in Table 3.4, values of Css and Tss for kepone at sites 19 and 21 were not determined.
These two sites had the highest percent organic carbon of any of the simulated sites, and the high
Kd values for kepone produced strongly persistent behavior. Simulations were carried out to 400
years for both sites, at which point steady state conditions had not yet been achieved.
DDT
DDT was simulated as a parent with no daughter products hi order to determine its ultimate
concentration hi the upper soil profile. While DDT has one of the higher K^. values investigated,
it also has Kj, Kw, and Kh values that allow losses that are higher than the more persistent POPs
investigated. The decay losses were typically the highest losses hi the PRZM-3 simulations. As
shown hi Table 3.5, DDT was one of the most consistent POPs investigated, with a fairly narrow
range of Css and Tss values.
Dieldrin
Dieldrin simulations indicate behavior similar to that of DDT. Again, the chemical parameters
for dieldrin suggest a compound that will be persistent, but not strongly so, and subject to a wide
range of loss processes. For dieldrin, the highest loss term was decay, followed by volatilization
losses which typically were about half the decay losses. This compound, like DDT, behaved in
a very consistent fashion from site to site.
Methoxvchlor
Methoxychlor, as shown in Figure 3.18, is at the lower end of what might be considered the range
of possible persistent behaviors. Soil concentrations of methoxychlor typically increase for several
years, but quickly reach a steady state condition where the loss terms match the application rate.
Methoxychlor has a relatively low K^. value and the highest K^ value of all of the "persistent"
POPs investigated. The highest loss term for methoxychlor was typically decay.
Toxaphene
Toxaphene is similar to DDT and dieldrin in its behavior from site to site, with a tendency to
reach steady state conditions at consistent values of Css and Tss. This POP has a relatively low K^
value, and typically was subjected to losses in the solid, liquid, and gaseous phases. Its primary
loss mechanism is associated with decay.
Dicofol
Dicofol had one of the lowest K^ values investigated, and also had the highest K^ and K^, rates of
any of the POPs. Its decay losses were the highest of any POP investigated, and these losses were
sufficient to prevent any accumulation of chemical from year to year.
-> ^ /•
F:\Pro.ecu'EPA1 EPAjX>i_:i 103 popsl.wpd J-—O HytcGeoLogic. Ir.c. 09/02/99
-------
Endosulfan
Endosulfan behavior is somewhat similar to that of methoxychlor. In fact, at all sites, the
computed minimum concentrations never reached zero after the first year or two. However,
because its loss terms are relatively high, it is not judged to behave as a "persistent" pesticide.
The endosulfan K^ value is lower than methoxychlor, and the decay term for the liquid phase
significantly higher. The highest loss term for endosulfan in the PRZM-3 simulations was decay.
Furan
Furan is not a persistent POP, at least as defined for this study. Its minimum concentration
reaches zero every year at all sites, and it exhibits no tendency to accumulate in the top soil layer.
Furan has the lowest K^ value of all of the POPs investigated (by over two orders of magnitude).
Although its solid and liquid phase decay rates are zero, it has a very high Kh value. It was the
only POP investigated that had volatilization as its highest loss term.
HvdroGccLogic. I,c.
-------
4.0 SENSITIVITY ANALYSIS
Following completion of the base PRZM-3 simulations described hi Section 3, a sensitivity
analysis was conducted. The purpose of the sensitivity analysis was to determine which of the
parameters in the PRZM-3 input files had the greatest effect on computed POP concentrations.
The sensitivity analysis was performed by conducting and analyzing additional PRZM-3
simulations as described in this section.
Based on the base simulation results, it was determined to design the sensitivity analysis using
dieldrin and kepone. These two POPs were chosen for additional analysis because they both
exhibited the tendency toward steady state behavior of interest to this study. However, the
chemical nature of these two compounds (at least with respect to their Kh, Kw, K,, and K^. values)
were different enough to warrant investigating both.
Sensitivity analyses were conducted at four sites. One site was chosen for each of the four crops
simulated in this study. The intent was to pick sites that would represent a wide range hi
climatological conditions across the country. The four sites selected were site number 4 (corn;
Cayuga County, NY; cool and wet); site number 9 (cotton, Maricopa County, AZ; hot and dry);
site number 22 (soybeans, Tunica County, MS; hot and wet); and site number 28 (wheat, Daniels
County, MT, cool and dry).
A total of 21 PRZM-3 input parameters were modified (one parameter per simulation). Each
parameter selected for inclusion hi the sensitivity analysis was increased and decreased by 50%
hi separate simulations, and the resultant impact on computed soil concentrations tabulated. It
should be noted that for two parameters (runoff CN and maximum percent coverage for a crop)
it did not make physical sense to follow the +/- 50 % variation. Runoff CNs were allowed to vary
to the minimum and maximum values listed in the PRZM-3 User's Manual for a given hydrologic
soil group. The minimum and maximum percent coverages were assumed to be 80 and 100,
respectively (the base simulations assumed 95% coverage). For the purposes of this analysis,
computed concentrations at t = 50 years were compared to the base simulation to determine the
relative impact a particular parameter had on concentration.
Results of the sensitivity analyses for dieldrin are shown hi Table 4.1, and for kepone hi Table
4.2. The results indicate that the most sensitive parameters included Kh, K,, Kj, and CN. These
parameters were included hi the attempt to develop a predictive screening level methodology
described hi Section 5.
F:'.Proj«ea'.EP.VEPA 004 :il03'.R06-99.157.»,pd "*" * HydroGeoLogic. Ire. 9.7/99
-------
Table 4.1
Results of Sensitivity Analyses for Dieldrin
PRZM-3
Parameter
ANETD
USLEK
USLELS
USLEP
SLP
HL
iCINTP
AMXDR
HTMAX
USLEC
MNGN
PAIR
HENRYK
ENPY
PCDEPL
RATEAP
DWRATE
DSRATE
KD
COVMAX
CN
Site #4: Corn
Cayuga County, NY
concentration in mg/kg,
t=50yrs
50%
5.128
5.270
5.271
5.270
5.139
5.150
5.129
5.135
5.128
5.269
5.106
5.371
5.371
5.128
N/A
N/A
5.128
8.149
4.951
5.128
5.492
100%
5.128
5.128
5.128
5.128
5.128
5.128
5.128
5.128
5.128
5.128
5.228
5.128
5.128
5.128
5.128
5.128
5.128
5.128
5.128
5.128
5.128
150%
5.122
5.030
5.030
N/A
5.120
5.105
5.127
5.121
5.128
5.029
5.142
4.962
4.962
5.128
N/A
N/A
5.128
3.669
5.228
5.128
4.981
Site £9: Cotton
Maricopa County, AZ
concentration in mg/kg,
t=50yrs
50%
4.377
4.404
4.404
4.404
4.394
4.393
4.391
4.737
4.393
4.404
4.389
4.835
4.834
4.393
4.095
4.393
4.394
7.230
4.290
4.394
4.487
100%
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
4.393
150%
4.691
4.381
4.381
N/A
4.391
4.391
4.392
4.474
4.393
4.381
4.394
4.147
4.147
4.393
4.804
4.393
4.392
3.050
4.450
4.393
4.251
Site £22: Soybeans
Tunica County, MS
concentration in mg/kg,
t=50vrs
50%
4.876
4.962
4.957
4.962
4.8S5
4.889
4.883
4.880
4.879
4.961
4.868
5.147
5.146
4.879
N/A
N/A
4.879
7.846
4.729
4.881
5.155
100%
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
4.879
150%
4.879
4.819
4.825
N/A
4.874
4.874
4.875
4.879
4.879
4.818
4.889
4.723
4.723
4.879
N/A
N/A
4.879
3.462
4.977
4.879
4.813
Site #28: Wheat
Daniels County, MT
concentration in
mg/kg, t=50yrs
50%
4.825
4.848
4.849
4.849
4.839
4.838
4.839
4.835
4.837
4.849
4.835
5.233
5.232
4.837
N/A
N/A
4.838
7.814
4.635
4.838
4.872
100%
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
4.837
150%
4.846
4.827
4.827
N/A
4.836
4.837
4.835
4.839
4.837
4.827
4.839
4.654
4.654
4.837
N/A
N/A
4.836
3.416
4.998
4.836
4.81
% deviation from base simulation
Parameter
ANETD
USLEK
USLELS
USLEP
SLP
HL
CINTP
AMXDR
HTMAX
USLEC
MNGN
PAIR
HENRYK
ENPY
[PCDEPL
RATEAP
DWRATE
DSRATE
KD
COVMAX
!CN
50%
0.00
2.77
2.79
2.77
0.21
0.43
0.02
0.14
0.00
2.75
-0.43
4.74
4.74
0.00
N/A
N/A
0.00
58.91
-3.45
0.00
7.10
100%
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
150%
-0.12
-1.91
-1.91
N/A
-0.16
-0.45
-0.02
-0.14
0.00
-1.93
0.27
-3.24
-3.24
0.00
N/A
N/A
0.00
-28.45
1.95
0.00
-2.87
50%
-0.36
0.25
0.25
0.25
0.02
0.00
-0.05
7.83
0.00
0.25
-0.09
10.06
10.04
0.00
-6.78
0.00
0.02
64.58
-2.34
0.02
2.14
100%
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
150%
6.78
-0.27
-0.27
N/A
-0.05
-0.05
-0.02
1.84
0.00
-0.27
0.02
-5.60
-5.60
0.00
9.36
0.00
-0.02
-30.57
1.30
0.00
-3.23
50%
-0.06
1.70
1.60
1.70
0.12
0.20
0.08
0.02
0.00
1.68
-0.23
5.49
5.47
0.00
N/A
N/A
0.00
60.81
-3.07
0.04
5.66
100%
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
150%
0.00
-1.23
-1.11
N/A
-0.10
-0.10
-0.08
0.00
0.00
-1.25
0.20
-3.20
-3.20
0.00
N/A
N/A
0.00
-29.04
2.01
0.00
-1.35
50%
-0.25
0.23
0.25
0.25
0.04
0.02
0.04
-0.04
0.00
0.25
-0.04
8.19
8.17
0.00
N/A
N/A
0.02
61.55
-4.18
0.02
0.72
100%
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
150%
0.19
-0.21
-0.21
N/A
-0.02
0.00
-0.04
0.04
0.00
-0.21
0.04
-3.78
-3.78
0.00
N/A
N/A
-0.02
-29.38
3.33
-0.02
-0.56
F:' Pro;«3'.EPA' E?A_CC4J> 1103' R06-99.167.-*pd
4-2
Hvc.-cGecLog;c. Ire. 9.7.'99
-------
Table 4.2
Results of Sensitivity Analyses for Kepone
PRZM-3
Parameter
AXETD
USLEK
USLELS
USLEP
SLP
HL
CINTP
AMXDR
HTMAX
USLEC
MNGN
DAIR
HENRYK
ENPY
PCDEPL
RATEAP
DWRATE
DSRATE
KD
COVMAX
CN
Site #4: Corn
Cayuga County, XV'
concentration in mg/kg,
t = 50yrs
501
20.60
21.68
21.68
21.68
20.69
20.76
20.65
20.65
20.60
21.68
20.44
20.73
20.73
20.60
N/A
N/A
N/A
N/A
21.17
20.61
23.44
1001
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
20.60
1501
20.55
19.86
19.86
N/A
20.54
20.43
20.56
20.55
20.60
19.86
20.71
20.47
20.47
20.60
N/A
N/A
N/A
N/A
19.82
20.59
19.42
Site #9: Cotton
Maricopa County, AZ
concentration in
mg/kg, t = 50yrs
501
7.114
6.798
6.796
6.79S
6.763
6.757
6.792
4.939
6. '57
6.798
6.~45
6.836
6.836
6.757
13.53
6.757
N/A
N/A
3.446
6.751
6.105
1001;
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
6.757
1501
4.119
6.721
6.719
N/A
6.753
6.752
6.751
5.168
6.757
6.721
6.764
6.682
6.682
6.757
4.103
6.757
N/A
N/A
10.120
6.752
S.311
Site -22: Soybeans
Tunica County, MS
concentration in
mg kg, t = 50yrs
501
20.50
21.33
21.28
21.33
20.60
20.65
20.59
20.56
20.55
21.33
20.43
20.72
20.72
20.55
N/A
N/A
N/A
N/A
18.05
20.56
19.63
1001 I 1501
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.55
20.49
19.97
20.03
N/A
20.50
20.50
20.51
20.54
20.55
19.97
20.63
20.39
20.39
20.55
N/A
N/A
N/A
N/A
20.55 | 19.99
20.55
20.55
20.55
19.81
Site ?28: Wheat
Daniels County, MT
concentration in
mg kg. t = 50yrs
501
21.03
20.90
20.91
20.91
20.72
20.71
1001 j 1501
20.69
20.69
20.69
20.69
20.69
20.69
20.63 20.69
20.84 j 20.69
20.69 20.69
20.91
20.65
21.31
21.31
20.69
N/A
N/A
N/A
N/A
13.22
20.67
20.51
20.69
20.69
20.69
20.69
20.69
20.69
20.69
20.69
20.69
20.69
20.69
20.69
20.17
20.50
20.51
N/A
20.63
20.68
20.75
20.49
20.69
20.51
20.72
20.15
20.15
20.69
N/A
N A
N A
N/A
22.47
20.70
20.59
1 deviation from base simulation
Parameter
ANETD
USLEK
USLELS
USLEP
SLP
HL
CINTP
AMXDR
HTMAX
USLEC
MNGN
DAIR
1 HENRYK
IENPY
PCDEPL
RATEAP
DWRATE
DSRATE
KD
COVMAX
ICN
501
0.00
5.24
5.24
5.24
0.44
0.78
0.24
0.24
0.00
5.24
-0.78
1001
-
-
-
-
-
-
-
-
-
-
-
0.63 f -
0.63
0.00
N/A
N/A
N/A
N/A
2.77
0.05
13.79
-
-
-
-
-
1501
-0.24
-3.59
-3.59
N/A
-0.29
-0.83
-0.19
-0.24
0.00
-3.59
0.53
-0.63
-0.63
0.00
N/A
N/A
N/A
- | N/A
-
-
-
-3.79
-0.05
-5.73
501
5.28
0.61-
0.58
0.61
0.09
0.00
0.52
-26.91
0.00
0.61
-0.18
1.17
1.17
0.00
100.24
100%
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
0.00
N 'A
N/A
-49.00
-0.09
-9.65
-
-
-
-
-
1501
-39.04
-0.53
-0.56
N/A
-0.06
-0.07
-0.09
-23.52
0.00
-0.53
0.10
-1.11
-1.11
0.00
-39.28
O.CO
N/A
N/A
49.77
-o.o-
501
-0.24
3.30
3.55
3.80
0.24
0.49
0.19
0.05
0.00
3.80
-0.53
0.83
0.33
O.CO
N/A
N/A
N/A
NA
-12. r
0.05
23.00 | -4.43
1001
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
1501
-0.29
-2.32
-2.53
N/A
-0.24
-0.24
-0.19
501
1.64
1.01
1.06
1.06
0.14
0.10
-0.29
-0.05 ! 0."2
0.00
0.00
-2.32 1.06
0.39
-0.19
-0.~S 3.00
-0."S 3.00
1001
-
-
-
-
-
-
-
-
1501
-2.51
-0.92
-0.37
N/A
-0.05
-0.05
0.29
-0.9-7
0.00
-
-0.37
0.14
-2.6:
- ! -2.61
0.00 0.00 j
N/A
N/A
N'A N/A
O.GO
- ! N/A
N/A
N/A N/A
N/A NA
T ""•}
-36.10
0.00 -0.10
-3.60
-0.37
N/A
N/A
3.6C
0.05
-0.43
4-3
-------
ts
OK)'
rt
H
91
•t
n
5
n
3
2
>—
o'
ft
M
D,
•1
rt
4-
16.00
0.00
0
Site 9: Dieldrin - Sensitivity at T = 50 Years
Parameter: HENRYK
50% decrease
base case
50% increase
6 8 10 12 14
Soil Compartment Number
-------
I
.'„
B
t
5
r
o1
p
d
50.00
0
Site 4: Dieldrin - Sensitivity at T = 50 Years
Parameter: DSRATE
50% decrease
base case
A 50% increase
•HI—ft—ft—-ft—ft—ft-HI—ft—ft—-ft---ft—ft
6 8 10 12 14 16 18 20
Soil Compartment Number
-------
,
i
i
sr
n
e
I
a
sr
H
n
s
C/J
S'
i ;
45.00
40.00
Site 22: Kepone - Sensitivity at T = 50 Years
Parameter: KD
50% decrease
A 50% increase
o.oo
6 8 10 12 14
Soil Compartment Number
-------
;,
i
i
4-
~4~
X
n
2]
I
•3
M
BS'
i
i
S1
n
•X
n
e
n
n
e
D
5
d
•a
c
D
R
Ht
00
60.00
50.00
o> 40.00
TJ>
o 30.00
o
c 20.00
o
O
10.00
0.00
0
Site 28: Kepone - Sensitivity at T = 50 Years
Parameter: CN
decrease
base case
6 8 10 12 14
Soil Compartment Number
16
18
20
F:\PlOjeUs\KPA\KI'A 004 2UU3\KOO-99.l<>7,wpd
4-8
HydroGeoLogic, Inc 9/2/99
-------
5.0 DEVELOPMENT OF PREDICTIVE METHODOLOGY
5.1 INITIAL STATISTICAL ANALYSIS OF PRZM-3 DATA
Output from the PRZM-3 model runs were assembled into a spreadsheet file. Simulation results
for all of the POPs that exhibited steady state tendencies (DDD, DDT, dieldrin, kepone,
methoxychlor, and toxaphene) were grouped into pages in the spreadsheet, one for each
compound. Each spreadsheet page included data which described the input parameters and the
results of a particular model run. The data included the following (PRZM-3 input parameters are
shown in bold):
Kd,
Kfc,
Da,
percent organic carbon in the soil,
average annual loss due to decay,
average annual loss due to erosion,
average annual loss due to runoff,
average annual loss due to volatilization,
average annual remaining core storage,
average annual precipitation,
average annual evaporation,
average annual throughfall,
average annual snowfall,
average annual snow melt,
average annual runoff,
average annual infiltration,
average annual sediment loss,
average annual evapotranspiration,
average annual recharge,
C,s, and
T«.
All of the parameters identified in the sensitivity analysis described in Section 4 were included
with the exception of curve number CN. The CN is used to compute the amount of runoff (and
hence infiltration) associated with a given increment of rainfall. It is known to vary seasonally,
and 3 separate values for CN are required in the PRZM-3 input file. The effect of variation in CN
was considered by including average annual values for precipitation, infiltration, runoff,
evapotranspiration, and recharge. Separate consideration of each of these hydrologic water budget
parameters allowed a more rigorous investigation of the effect of these parameters on computed
values for Css and Tss. The average annual values were computed from data parsed from the
PRZM-3 output files.
g:c. I.-c. 9/L99
-------
A statistical analysis was performed, the objective of which was to determine a set of functions
that would fit the existing model runs reasonably well. Two functions were desired: one to
predict Css, and one to predict Tss. This approach required balancing the two opposing goals of
equation fit and equation simplicity. If allowed to use all the input parameters, an equation could
be constructed which would match the model output very closely. This equation would be too
complex and cumbersome to be practical, however. A simpler equation was therefore desirable.
The following discussion describes the search for an equation to fit Css data. The same approach
was used in the search for an equation to fit Tss data.
The analysis involved attempting to construct a function with the highest correlation coefficient
(R2) using the least number of terms and variables. The desired function would calculate Css in
terms of one or more independent input variables. Some of the prospective input variables were
rejected as possible candidates due to the difficulty in determining them without extensive
modeling efforts. These included total evapotranspiration, and all of the average annual the loss
terms. The list of potential predictor variables was then narrowed to these:
percent organic carbon in the soil,
average annual precipitation,
average annual evaporation,
average annual runoff,
average annual infiltration, and
total recharge.
Some of the variables on the above list were considered more likely candidates than others based
on the results of the model sensitivity analysis.
The first steps taken in the search involved looking for a suitable function of a single variable
using Microsoft Excel. In Excel, separate x-y plots were generated using each of the candidate
variables as the independent variable with Css as the dependent variable. For each of the x-y
relationships plotted, an R2 value was computed for each of the following best fit curves: linear,
logarithmic, polynomial (up to degree 6), power, and exponential. None were found to be useful.
The highest reported R2 value in this series was less than 0.5. Smaller data sets- one for each
compound - were generated and, where appropriate, analyzed using the same methodology. (For
instance, this would not be appropriate for Kj, since Kj is constant for a given compound.) A
slightly higher R2 value (0.7) was found in isolated cases, but in general this approach did not
provide acceptable results.
In pursuit of a possible single-variable function, more analysis was performed using another piece
of software. NCSS 2000, from Number Cruncher Statistical Systems, is a robust, elaborate
statistical analysis package with a large number of available fit models. The existing spreadsheet
data were imported into NCSS and all of the single-variable fit models were attempted. As in
F:,Prcj.:csEPA>EPA_(X4_:;iO;',R;X-99.157.wpd J~~ Hyi-cGeoLos*:. !-c. 9'2/99
-------
Excel, none of these models had an acceptable R2 value. At this point, the search for a single-
variable function was abandoned.
Next, the NCSS software was employed in the search for a suitable multi-variable function. The
number of possible/o/ras of a multi-variable function is more limited, but the number of possible
functions is staggering. Fortunately, NCSS provides a method for evaluating possible functions
without having to construct each and every one. Using a process called "multivariate ratio
search, " NCSS will evaluate whole classes of possible functions in sequence, and report the results
ranked by R2 value. The basic approach is to construct a rational function (one in which there is
a polynomial function in the numerator and another in the denominator) based on certain types of
combinations of the independent variables. The user can specify the types of combinations to be
considered, but in this case no limitations were placed on the search. In addition to the basic
independent variables, this process can also utilize some transformations of the independent
variables in the search. The allowable transformations are:
x' = -
X
x' = ln(x)
x' = V*
.x' = x2
In most cases, all of these transformations were tested. A special situation occurred, though, when
Kj was used as one of the independent variables. Since kepone has a K, of 0, some
transformations (e.g., ln(x)) would result in an overflow or underflow condition. Two approaches
to this problem were taken. In one case, K^ was allowed to go to 0 and error-producing
transformations were not considered. In the other, K, was allowed to go only to IxlO"15, and all
transformations were considered.
Using the multivariate ratio search capabilities of NCSS, all possible combinations of two and
three of the independent variables were explored. (Combinations of four variables were also
considered briefly, but most of these tests produced functions with more than 50 terms. This was
deemed too many terms for a useful application.) For each combination, all possible allowable
functions were evaluated using all the available transformations. The best R2 value for each
combination was noted, along with the number of variables involved in the equation, the
transformations applied, and the form of the equation itself. These were then ranked by R2, and
the top echelon in each category (two- and three-variable functions) were chosen for closer
inspection.
Closer inspection of each function consisted of generating several plots and exploring the overall
behavior of the function. In each case, plots were generated showing the concentration predicted
by PRZM-3 for each compound, with values predicted by the function overlaid for comparison.
F:\Pro,«ec-j,E?A'.EPA OOi_:i;C3'RC6-99.167.u.pd -'"•' Hyd-cCecLogic. Ix. 9 -'99
-------
These plots gave an indication of how well each function predicted the model output for specific
compounds. Additional plots were generated for some of the two-variable cases. These were x-y
plots which showed a curve generated by fixing one of the two variables and allowing the other
to vary over the range found in the data set. The specific model output data points were graphed
against these curves, providing further evidence of how well each function might predict the
behavior of similar compounds.
Two tools were employed in exploring the overall behavior of each function. First, simple
mathematical tests were applied to determine discontinuities, local maxima and minima, and other
behavior patterns. Second, a small visualization program was written which graphed each function
over the range of independent variables.
The three-variable functions generally had very high R2 values (>0.96) when compared to the
model data points, but had an exorbitant number of terms. Table 5.1 shows the best of the three
variable predictive functions which were considered.
Table 5.1
Summary of Candidate Three-Variable Functions to Predict C^
Variables
1
Kh
Da:r
Kh
Organic Carbon
Kh
Da,
Organic Carbon
Organic Carbon
Kh
Da,
Orsanic Carbon
2
K,
K,
K,
Kh
K,
K,
Kh
K,
K,
K,
K^
3
Infiltration
Infiltration
Precipitation
Infiltration
Runoff
Precipitation
Precipitation
Runoff
Recharge
Recharge
Infiltration
R2
0.975
0.969
0.696
0.968
0.963
0.960
0.959
0.959
0.958
0.953
0.951
number of terms
26
26
26
26
26
26
26
26
26
26
18
Ultimately all the three-variable functions were rejected due to behavior problems or
computational complexities.
The highest R2 values for the two-variable functions were all around 0.9. The two-variable
functions were generally more well-behaved than the three-variable ones, and there were several
from which to choose in the top echelon. Table 5.2 shows the best of the two variable predictive
functions which were considered.
r:' Project EPA '-E?AJX>i_:i !03 J-.06-99.167.wpJ
5-4
HydroGcoLogic. Inc. 91.99
-------
Table 5.2
Summary of Candidate Two-Variable Functions to Predict C&
Variables
1
K,
Kh
Organic Carbon
Organic Carbon
Organic Carbon
Dair
K*
Organic Carbon
2
K-
K,
Kd
Kh
K«
Kd
Koc
Da,
R2
0.901
0.900
0.898
0.893
0.893
0.893
0.888
0.852
number of terms
6
8
10
8
10
8
16
8
After considering the model sensitivity analysis, the function utilizing K^ and K^ was chosen as the
primary candidate (this function was developed with the option of allowing Ks to go to 0; this
option was adopted for all subsequent analyses). Further analysis of this function showed that it
did a good job of fitting the data already produced by the model, but had problems elsewhere.
The range of values for K, for the existing compounds was wide but highly concentrated (see Table
3.2). This caused the function to fit well on opposite ends of the range, but allowed wide latitude
in between. To overcome this problem, additional PRZM-3 simulations were designed and
conducted using existing site and meteorological data, but with compound characteristics made up
to fill the data gaps. These simulations are described hi the next section.
5.2 ADDITIONAL PRZM-3 SIMULATIONS
PRZM-3 simulations were conducted using "made up" compound characteristics that were
designed primarily to include solid phase decay rate values (K,) that fell within the range between
the value for methoxychlor (K, = 1.89E-03/day) and toxaphene (K, = 1.92E-04/day). A total of
5 different K^ values were used for these new compounds, which were designated A through E.
The K, values for these POPs are shown in Table 5.3.
F:'.Pro;scts\EP.VEPA_(XXi_:iIC3'R06-99.I67.»-?d
Hvd-oGeoLosic. I.-x. 97.99
-------
Table 5.3
Values for "New" POPs
POP
A
B
C
D
E
K^day1
4.70E-04
8.20E-04
1.12E-03
1.36E-03
1.61E-03
A total of forty additional PRZM-3 simulations were conducted using the K, data shown in Table
5.3; eight simulations were conducted for each of these values for solid phase decay rate.
Simulations were conducted by using the PRZM-3 input data files from the base simulations, and
modifying them to reflect a range of IQ and JQ values required to fill in the data gaps described
in Section 5.1. For each K, value shown in Table 5.3, a series of K^ values were assumed to
reflect the effect of variation in the percent of organic carbon. Simulations were conducted using
"default" input data files for both dieldrin and toxaphene, and using sites 5 and 27. Table 5.4
summarizes the distinguishing characteristics for the additional PRZM-3 simulations, and presents
simulation results.
Table 5.4
Summary of Additional PRZM-3 Simulations Conducted to Support Development of
Screening Level Methodology
Sim #
1
2
3
4
5
6
7
8
9
10
11
12
Site
5
5
5
5
5
5
5
5
5
5
5
5
Crop
Com
Corn
Corn
Corn
Corn
Corn
Com
Com
Corn
Com
Com
Corn
Default Input File
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
POP
A
A
A
A
B
B
B
B
C
C
C
C
Ka
7920
22090
32410
51180
9890
16100
31000
45610
2910
11110
45610
71280
C^, mg/kg
2.120
2.148
2.154
2.158
1.289
1.289
1.294
1.296
0.939
0.961
0.967
0.968
T«, yrs
25.67
25.67
25.67
25.67
17.75
17.75
17.75
17.75
15.33
15.92
15.33
14.58
5-6
-------
Table 5.4 (continued)
Summary of Additional PRZM-3 Simulations Conducted to Support Development of
Screening Level Methodology
Sim#
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
Site
5
5
5
5
5
5
5
5
27
27
27
27
27
27
27
27
27
27
27
27
27
27
27
27
27
27
27
27
Crop
Com
Corn
Corn
Corn
Com
Com
Com
Corn
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Wheat
Default Input File
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Toxaphene
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
Dieldrin
POP
D
D
D
D
E
E
E
E
A
A
A
A
B
B
B
B
C
C
C
C
D
D
D
D
E
E
E
E
K,
4210
21200
31000
49060
710
10020
14810
26300
7920
22090
32410
51180
9890
16100
31000
45610
2910
11110
45610
71280
4210
21200
31000
49060
710
10020
14810
26300
Css, mg/kg
0.791
0.803
0.804
0.805
0.632
0.681
0.682
0.684
2.44
2.54
2.56
2.58
1.50
1.52
1.54
1.54
1.06
1.13
1.15
1.16
0.91
0.95
0.95
0.96
0.66
0.80
0.80
0.81
TB,yrs
13.67
13.67
13.67
13.67
12.17
12.17
12.17
12.17
26.42
26.42
24.83
24.83
18.42
18.42
18.42
18.42
17.58
17.58
17.58
17.58
17.58
17.58
17.58
17.58
15.75
16.17
16.17
16.25
5-7
HvdrcGecLotic. !-.c. 9.1'99
-------
5.3 FINAL STATISTICAL ANALYSIS OF PRZM-3 DATA
The new model runs described in Section 5.2 were incorporated into the existing data set and a
new function of K, and Kj was generated. The revised function is shown in Equation 5.1, with
coefficients rounded to two decimal places.
-9.39+581.45,/F + 2.44,/F - 61.99,/FF
(^ - V 5 V » V s " -.- jx
•^ 1+8.31,/F - 2.76xlO"2,/F - 20.66,/FF
y .s y ^ y s d
Equation 5.1 produced an overall R2 value of 0.906 for the PRZM-3 simulations considered.
Using the same approach, an equation which estimates Tss was also developed. The Tss equation
is also a function of K, and Kj only. The best fit time prediction equation is presented in Equation
5.2.
Tss --477.04+43638.61 ^F - 763761.24^ +
- 14257.37yFln(/tp + 239032.10^1n(A'^) (5.2)
-11.161n(F)2 + 896.57,/Fln(A-)2 - 14847.59 A-In (F)2
^ dj y 3 ^ d' s ^ d'
Equation 5.2 produced an overall R2 value of 0.811 for the PRZM-3 simulations considered.
Equations 5.1 and 5.2 are interesting in that both are functions of K^ and K^ only. In other words,
the only site specific information that is required to apply these equations is the percent organic
carbon in the top 15 cm of the soil (since %OC is used in the determination of K
-------
1.E+06
1.E+05
1.E+04
1.E+03
Kd 1.E+02
1.E+01
1.E+00
1.E-01
1.E-02
1.E-10 1.E-08 1.E-06 1.E-04 1.E-02 1.E+00
Ks
Figure 5.1 Behavior of Numerator Function in Equation 5.1.
Figure 5.2 shows the behavior of the function in the denominator in Equation 5.1. This figure
shows the regions where the denominator function will predict positive, zero, and negative values.
1.E+07
1.1
1.1
1.1
1.1
1.1
1.E+01
Kd
— Denominator = 0
1.E-10 1.E-08 1.E-06 1.E-04 1.E-02 1.E+00
Ks
Figure 5.2 Behavior of Denominator Function in Equation 5.2.
F:\ Pro; :c s\ E? A' EPA_OW_211031RC6-99.! 67. -*-pd
5-9
Hyi-cGceLog:c, I-c.
-------
Figure 5.3 shows the region where Equation 5.1 is applicable This figure was generated by
overlaying Figures 5.1 and 5.2, and investigating the behavior of the predictive equation for Css at
each point. The figure also includes 4 discrete points labeled A, B, C, and D; these points will be
discussed in Section 5.5.
'» Mum erstor = 0
— —Denominator = 0
1.E-01 -
1.E-02 -
1.E-10
1.E-08
1.E-06
1.E-04
1 .E-02
1.E+00
Figure 5.3 Region of Applicability of Predictive Methodology.
Figures 5.4 through 5.8 demonstrate how well the predictive equation for Css matches the simulation
results from PRZM-3 for specific compounds. Figure 5.4 shows the computed (based on PRZM-3
simulations) vs. predicted (based on Equation 5.1) values for Css for a range of different POPs. The
x axis on Figure 5.4 is simply an arithmetic scale utilized to allow plotting of Css data from individual
(i.e. site specific) PRZM-3 simulations.
F:\Pngeds\EPAVEFA_004_21103VR06-99.167.wpd
5-10
HydroGeoLogic* Inc 9/2/99
-------
120.00
100.00
80.00
= 60.00
:
;
40.00
20 00
0 00
o Model
$ Equation
New" POPs Methoxychlor
Figure 5.4 Computed vs. Predicted Values of C^ for Selected POPs.
Figures 5.5 through 5.8 show the computed vs predicted values for Css for kepone, toxaphene, POP
A, and methoxychlor, respectively. These figures each contain discrete points from the site specific
PRZM-3 simulations and a curve generated by applying the respective K^ values for each compound
to Equation 5.1.
F:\Projaas\EPA\EPA_004_21103\R06-99.167 wpd
5-11
HydroGeoLogrc, Inc. 9/T/99
-------
O)
c
O
125
100
75
t5 50
g 25
c
O
O 0
0
Equation
Model
100
200
300
400
500
K
Figure 5.5 Computed vs. Predicted Values of Css for Kepone; Ks = 0/day.
0
0
750
Figure 5.6 Computed vs. Predicted Values of C^ for Toxaphene; Ks = 1.92E-04/day.
F:\Projeots\EPA\EPA_004_21103\R06-99 167. wpd
5-12
HydroGeoLogic,Iiic. 9/V99
-------
O) q
E
_- ? -J
c *•
.0
s
(0
i: 1
c
0)
u
S n
•
•
- Equation
• Model
.
.
J
0 10000 20000 30000 40000 50000 60000
Kd
Figure 5.7 Computed vs. Predicted Values of Css for POP A; Ks = 4.70E-04/day.
2.0
O)
I) 1.5
c 0.5
o
U
0.0
:
1000
2000
3000
Figure 5.8 Computed vs. Predicted Values of C« for Methoxychlor; K, = 1.89E-03/day.
F:\ProJ8c6\EPA\EPA_004_2II03\R06-99.I67.wpd
5-13
Hy*oGeoLoglc, Inc. 9/TJ99
-------
Since Equation 5.1 was developed by applying statistical methods to empirical data, one would be
ill-advised to apply it beyond the bounds of the data used to construct the equation. In this regard,
the KJ values do not appear to be a limiting factor. However, the bounds of K, do provide some
guidance. In the current study, K, varied from 0 to 1.89xlO~3 for the POPs exhibiting steady state
behavior. The function should therefore not be applied when values of K, are greater than about
1.9xlO~3. Additionally, care should be taken when applying the methodology to very small values of
K,. In particular, small values of K, (lower than approximately 1x10"6) should only be combined with
small values of Kd as shown in Figure 5.3. (Compounds with low values of Ks and high values of Kd,
such as chlordane and DDE, did not reach steady state conditions in the PRZM-3 simulations
described in this report)
5.5 APPLICABILITY OF THE PREDICTIVE METHODOLOGY FOR Tss
The equation to predict Tss (Equation 5.2) is a simple polynomial (and not a rational function), it has
no discontinuity. It is therefore possible to apply this equation to any set of parameters for which the
concentration equation can be applied. The only limitation to the time equation is that Kd can never
be 0, since ln(0) is not defined. Since this condition is not likely to be satisfied, particularly for
agriculturally managed soils, this limitation is not judged to be significant.
Figures 5.9 through 513 demonstrate the applicability of Equation 5.2, and are a repeat of Figures
5.4 through 5.8 for Tss instead of C,,.
350.00
300.00
250.00
u> 200.00
af
E
150.00
100j
0.00
jo o
Figure 5.9 Computed vs. Predicted Values of TM for Selected POPs.
FAProj«clsVEPA\EPAJ
5-14
HydroGeoLogK, Inc. 9/2/99
-------
[0
0
300
200
100
:
500
Figure 5.10 Computed vs. Predicted Values of T^ for Kepone; K, = 0/day.
150
300
450
600
750
K
Figure 5.11 Computed vs. Predicted Values of T,, for Toxaphene: K, = 1.92E-04/day.
F:\Proje05\EPA\EPA_004_21103VR06-99.167. wpd
5-15
HydroGeoLogic, Inc 9/2/99
-------
e
>
&
50
40
30
20
10
C
Equation
Model
0 10000 20000 30000 40000 50000 60000
Kd
Figure 5.12 Computed vs. Predicted Values of Tss for POP A; K, = 4.70E-04/day.
25
20
15
10
1000
2000
3000
K,
Figure 5.13 Computed vs. Predicted Values of TM for Methoxychlor; K, = 1.89E-03/day.
F:\Projeds\EP A\EPA_004_Zll03
-------
Figures 5.9 through 5.13 show that Equation 5.2 can be used with reasonable confidence to predict
Tss. While the R2 value for this equation is not as high as the R2 value for the equation to predict
Css, it is judged to be within acceptable tolerance for its intended application. Our approach was
based on the assumption that the critical feature of the predictive methodology described herein
was to provide the ability to estimate Css. Section 5.5 describes the analyses conducted to
investigate the uncertainty associated with the Css predictive methodology.
5.6 UNCERTAINTY ANALYSIS
In order to quantify the uncertainty in the predictive analysis and provide confidence levels for the
results, the concentration equation (Equation 5.1) was encoded in an Excel spreadsheet and the
software package @Risk was applied. @Risk is a spreadsheet add-in that performs Monte Carlo
or Latin Hypercube uncertainty analysis, and provides summary statistics and graphs of the result.
Four representative points were chosen from different regions of the domain in which the Equation
5.1 is applicable (see Figure 5.3). The K^ and Kd values from these points were entered into a
spreadsheet that was linked to @Risk. Input variables in @Risk can be defined by specifying an
underlying distribution and some descriptive information about the distribution.
Previous research (Baes and Sharp, 1983) in the development of a model to predict leaching
constants for solutes in agricultural soils was conducted assuming a lognormal distribution for K-!/ HyiroGecLog-c. Inc. 9.T99
-------
as described herein. The model could easily be modified to provide this capability, but the effort
required to do so was beyond the scope of the current study.) Additional research is required
before CV estimates for K^ and Kj can be refined further. However, a CV of 100% was judged
to be suitable for the purposes of assessing the uncertainty associated with usage of the predictive
methodology described in this report. Since the coefficient of variation is equal to the standard
deviation divided by the mean, a value of standard deviation was easily calculated for each of the
four test points. (The standard deviation in each case was set to the same value as the mean.)
The Monte Carlo analysis spreadsheet consisted of four pairs of test values of Kj and K
-------
.n
C3
.Q
O
20
40 60
Css, mg/kg
80
100
Figure 5.14 Cumulative Distribution Function for C^ at Point A.
.a
re
.a
O
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0 ^
100
200
Css, mg/kg
300
400
Figure 5.15 Cumulative Distribution Function for
at Point B.
F:\Projtc3\EP A\EPA_OW_:!103\RC6-99.157. wpd
5-19
Hyi-cGeoLog:c. I". 91'9
-------
.a
f3
o
CL
20
Css, mg/kg
Figure 5.16
Cumulative Distribution Function for CL at Point C.
_>,
!5
o
^
Q.
12
Css, mg/kg
Figure 5.17 Cumulative Distribution Function for C^ at Point D.
jera' £PA'HPA_OC-4_21103'.RC<>99.1
5-20
-------
Figures 5.14 through 5.17 indicate that the uncertainty associated with application of Equation 5.1
can be quite large. For instance, at point A in Figure 5.3 (where Kj = l.OOE-06/day and Kj =
100, Equation 5.1 predicts a value for Css of 15.9 mg/kg. The cumulative distribution function
shown in Figure 5.14 indicates that the expected value of Css would be less than or equal to about
35 mg/kg with 95% confidence. Similar results are evident at the other 3 points investigated. The
wide range in predicted values for Css is associated with the relatively large coefficient of variation
adopted for this analysis.
F:\Pro:ecs\EPA'.EPA «W :il03'.R06-99.167.»7
-------
6.0 SUMMARY AND CONCLUSIONS
The predictive methodology described in Section 5 can be used to quickly determine whether a
particular chemical compound is likely to be persistent, that is, to degrade so slowly that repeated
applications of that compound will cause increasing concentrations in the upper soil layer. All that
is needed is knowledge of the solid phase decay rate, K,, and the soil partition coefficient IQ at
the site for which the analysis is required.
The methodology was developed assuming annual unit applications of 1 kg/ha. However, the
methodology can be applied for other constant annual application rates. The Css value predicted
by Equation 5.1 should then be multiplied by the ratio of (actual application rate/unit rate).
Figures 6.1 through 6.3 show the effect of doubling and tripling the unit application rate at Site
1 for chlordane, dieldrin, and kepone, respectively. In each case, application of two and three
times the unit rate produces concentrations two and three times as high as those for the unit rate,
respectively. (Even though chlordane is not one of the steady state compounds, the concentration
curves are included here to lend credence to the assumption of linearity.)
--
~
3
H
I
I
0
-
•:
3X unit application
2X unit application
unit application
10 20 30 40 50 60
Number of Years
90
100
Figure 6.1 Effect of Various Application Rates on Predicted Concentrations;
Chlordane.
F:\Proiects\EPA\EPA.004_21103\R06-991S7wpd
6-1
HydroGeoLogK. toe 131/99
-------
18
16
oi 14
JC
o
E 12
2
M
2
M
B
•
•J
0
•J
5
2
10
— 3X unit application
— 2X unit application
unit application
^,j^N^]WW^^^^
mm
0 10 20 30 40 50 60 70 80 90 100
Number of Years
Figure 6.2 Effect of Various Application Rates on Predicted Concentrations; Dieldrin.
90
80
3x unit application
— 2x unit application
with volatilization
10 20 30 40 50 60 70
Number of Years
90
100
Figure 6.3 Effect of Various Application Rates on Predicted Concentrations; Kepone.
6-2 HydroGeoLogic, Inc 8/31/99
F:\Proj«ffl\EPA\EPA_a04__21103\RQ6-99.1S7wp
-------
The methodology is appropriate for compounds falling within the shaded region of Figure 5.3.
Application of the methodology outside that region is not recommended, and may lead to
unreliable estimates of steady state concentration and the time required to reach steady state.
It is difficult to quantify the uncertainty associated with application of the predictive methodology
because too little is known about the descriptive statistics for K; and Kj . However, the cumulative
distribution functions presented in Section 5.5 can be used as a guide to understanding the valid
ranges of concentrations possible for a particular chemical at a specific site.
F:'.Projrea\EPA'.EPAO&i :i!03.SC«-99.157.»7d ^'^ Hyi-oGeoLogic. Inc. 9Z'99
-------
7.0 REFERENCES
Baes, C.F. Ht, and R.D. Sharp. A Proposal for Estimation of Soil Leaching and Leaching
Constants for Use in Assessment Models. J. Environ. Qual. Vol. 12, no. 1, 1983.
Carsel, R.F., R.S. Parrish, R.L. Jones, J.L. Hansen, and R.L. Lamb. Characterizing the
Uncertainty of Pesticide Leaching in Agricultural Soils. Journal of Contain. Hydrol.,
25:111-124. 1988.
Imhoff, J.C., P.R. Hummel, J.L. Kittle, Jr., and R.F. Carsel. PATRIOT - A Methodology and
Decision Support System for Evaluating the Leaching Potential of Pesticides. U.S. EPA,
Athens, GA, 30605. 1993.
U.S. Environmental Protection Agency, National Exposure Research Laboratory. PRZM-3, A
Model for Predicting Pesticide and Nitrogen Fate in the Crop Root and Unsaturated Soil
Zones: User's Manual for Release 3.0. Athens, GA. 1998.
U.S. Environmental Protection Agency, Office of Solid Waste. Technical Support Document for
the HWIR: Risk Assessment for Human and Ecological Receptors. Washington, D.C.
August, 1995.
U.S. EPA Headquarters Library
Mai: code 320'
1200 Pennsylvania Avenue NW
Washington DC 20460
7 1
F:\Prtvec3\EP.V.EPA 0« :i:C3'.R06-99.167.»p! '~l HydroGccLcsic. Inc. S.1'99
-------
APPENDIX A
DOCUMENTATION OF SOURCES OF DATA USED FOR
CREATION OF PRZM-3 INPUT FILES
-------
RECORDS
PFAC:
SFAC:
ANETD:
pan factor used to estimate daily evapotranspirtation; obtained from Figure
5.1 of the PRZM-3 User's Manual.
snowmelt factor; taken from Table 5-1 of the PRZM-3 User's Manual.
Range for open areas reported to be 0.20 - 0.80 (cm per degree C per day);
constant value of 0.50 used for all sites.
minimum depth from which evaporation is extracted hi the dormant season.
Values obtained from Figure 5.2 of the PRZM-3 User's Manual.
When the simulation start date occurs before the crop emergence date, we assumed an initial crop
(INICRP =1) with a surface condition of "fallow" (ISCOND = 1).
RECORD 6
ERFLAG:
RECORD?
USLEK:
USLELS:
USLEP:
AFIELD:
IREG:
SLP:
HL:
erosion flag; set to 2 for all simulations to invoke Modified Universal Soil
Loss Equation.
USLE soil erodability factor; obtained from the SOILS5 database. The
mode of all of the values for individual samples within a county.
USLE topographic factor LS; computed in SOILS5 spreadsheet as a
function of hydraulic length and slope.
USLE practice factor P; obtained from the SOILS5 database. The mode of
all of the values for individual samples within a county.
area of the field in hectares; assumed to be the square of the hydraulic
length.
SCS rainfall distribution region; obtained from Figure 5.12 of the PRZM-3
User's Manual.
land slope in percent; obtained from the SOILS5 database. The average of
all of the values for individual samples within a county.
hydraulic length in meters; obtained from the SOILS5 database. The
average of all of the values for individual samples within a county.
RECORDS
All PRZM-3 simulations were performed for a single crop (NDC = 1).
F:',Pro;ec3\EPA'.EPA CCJ ;i;0:-'.R06-99.I67.»p
-------
RECORD 9
CINTP:
AMXDR:
COVMAX:
ICNAH:
CN:
HTM AX:
RECORD 9A
CROPNO;
NUSLEC:
RECORD 9B
GDUSLEC:
GMUSLEC:
RECORD 9C
USLEC:
max crop interception storage (cm); values from Table 5-4 of the PRZM-3
User's Manual. Corn = 0.28, cotton = 0.23, soybeans = 0.23, wheat =
0.08 (midpoint of respective ranges).
max crop rooting depth (cm). Based on Table 5-9 of the PRZM-3 User's
Manual and conversations with Dr. Paul Denton of the UT Agriculture
Extension Service, the following values were adopted: corn (100 cm);
cotton (90 cm); soybeans (60 cm); wheat (50 cm). The only value outside
of the ranges published hi Table 5-9 is wheat (range is 15-30 cm). Dr.
Denton felt strongly that this was too low for wheat.
max areal crop coverage in percent; assumed to be 95 for all crops. The
PRZM-3 User's Manual says this number should be between 80 and 100;
Dr. Denton told me the number "should approach 100%".
crop surface condition after harvest, we assumed all crops were in residue
condition after harvest (ICNAH = 3).
SCS curve numbers for fallow, cropping, and residue conditions. Values
from PATRIOT database.
max crop canopy height. Table 5-16 of the PRZM-3 User's Manual reports
a range of 80 - 300 cm for corn; values for our other crops not reported.
Based on conversations with Dr. Denton, the following values were
adopted: corn (210 cm); all others (90 cm).
crop number; 1 for all simulations.
number of USLE cover management factors; 3 for all simulations.
starting days for 3 cover management factors, from PATRIOT
starting months for 3 cover management factors, from PATRIOT
USLE cover management factors for fallow, cropping, and residue
conditions. This factor reflects the tillage and erosion control practices
most widely utilized for a given site. We assumed the cropping period
begins with plant emergence and ends with plant harvesting, that fallow
conditions (corresponding to tilling under the old crop residue to prepare
the ground for planting of the new crop) begin one month prior to crop
emergence, and that residue conditions begin at harvest and end with tilling.
Values of C for cropping conditions were based on the SOILS5 database
(the mode of all of the values for individual samples within a county), with
most weight given to the 1992 values (1982, 1987, and 1992 values of C
are included in the database). Values of C for fallow conditions varied
from 0.4 to 0.6, and were based on the USDA map "Tillage Practices by
NRCS Region, 1995". Where the predominant tillage practice in a region
is conventional, a C value of 0.6 for fallow conditions was assigned.
F:',Pro:ccts:EP.-VEPA_OOi_: 1 !C3 R06-99.!67.»-pd
A-2
H\
-------
RECORD 9D
MNGN:
RECORD 10
Where the predominant tillage practice is NOT conventional (i.e. the
combination of conservation tillage and reduced tillage practice), a C value
of 0.4 for fallow conditions was assigned. If the area was about 50%
conventional and 50% other, a value of 0.5 was assigned. C values for
residue conditions were chosen to be about half way between cropping and
fallow conditions.
Manning n roughness coefficient. PRZM-3 User's Manual recommended
default value of 0.17 used for all seasons at all sites.
One cropping period assumed per year: NCPDS = 1.
RECORD 11
Crop emergence, maturation, and harvest dates per PATRIOT.
RECORD 13
A typical simulation is for 100 years, with 1 pesticide application per year; NAPS = 100.
For most simulations, the number of pesticides (NCHEM) is 1; for DDT/DDD/DDE
simulations, NCHEM = 3. The (FRMFLG = 1) flag was used to test for ideal soil
moisture conditions, and bi-phase half lives were not used (DKFLG2 = 0).
RECORD 15
APD:
APM:
LAPYR:
WINDAY:
CAM:
DEPI:
TAPP:
APPEFF:
DRFT:
RECORD 17
FILTRA:
UPTKF:
target application day; assumed to be 5 days after crop emergence.
target application month.
target application year.
number of days to check for ideal soil moisture; 10 days at most sites.
chemical application method, set equal to 1 (soil applied, default
incorporation depth, linearly decreasing with depth).
depth of pesticide application (cm); set to 4 cm for all simulations.
target application rate; set to 1 kg/ha for all simulations.
application efficiency; set to 1 for all simulations to preserve unit load.
not used.
not used, set to 0.
plant uptake factor, set to 0 to indicate no plant uptake.
F:\Prci«cs'.EPA\EPA OW 21103'R06-99.iS7.wp
-------
RECORD 19
Soil type identified from PATRIOT; most common agricultural soil for the specified crop
in a given county used.
RECORD 20
CORED:
BDFLAG:
THFLAG:
KDFLAG:
HSWZT:
MOC:
IRFLAG:
ITFLAG:
IDFLAG:
BIOFLG:
RECORD 26
soil core depth in cm; from PATRIOT.
bulk density flag; set to 0 to allow input by user.
field capacity and wilting point flag; set to 0 to allow input by user.
adsorption coefficient flag; set to 0 to allow input by user.
drainage flag; set to 0 to allow free drainage.
method of characteristics flag; set to 0 (not used).
irrigation flag, set to 1 if irrigation simulated, 0 otherwise.
soil temperature simulation flag; set to 0 (not used).
thermal conductivity and heat capacity flag; set to 0.
biodegradation flag; set to 0 (not used).
DAIR: diffusion coefficient, see Table 3.2 of this report.
HENRYK: Henry's law constant, see Table 3.2 of this report.
ENPY: enthalpy of vaporization, see Table 3.2 of this report.
RECORD 27 (for irrigated sites only)
IRTYP:
PLEACH:
PCDEPL:
RATEAP:
RECORD 34
HORIZN:
THKNS:
BD:
THETO:
AD:
DISP:
ADL:
type of irrigation; set to 4 (under canopy sprinkler);
leaching factor as a fraction of irrigation water depth;.set to 0.0.
fraction of water capacity at which irrigation is applied; set to 0.5.
maximum rate at which irrigation is applied; set to 1.3 cm/hr.
horizon number, obtained from PATRIOT.
horizon thickness, obtained from PATRIOT.
bulk density, obtained from PATRIOT.
initial soil water content in the horizon, obtained from PATRIOT.
soil drainage parameter, set to 0.
pesticide hydrodynamic solute dispersion coefficient, set to 0.
lateral soil drainage parameter, not used.
F: Projeca'.EPA'EPA OC-J 2!1
A-4
-------
RECORD 36
DWRATE:
DSRATE:
DGRATE:
RECORD 37
DPN:
THEFC:
THEWP:
OC:
KD:
RECORD 39 (used only for DDT/DDE/DDD parent-daughter simulations)
DKRW12: dissolved phase transformation rate for DDT to DDE, set to 4.04E-04.
DKRW13: dissolved phase transformation rate for DDT to DDD, set to 4.04E-04.
DKRW23: dissolved phase transformation rate for DDE to DDD, set to 0.0.
DKRS12: adsorbed phase transformation rate for DDT to DDE, set to 6.56E-05.
DKRS13: adsorbed phase transformation rate for DDT to DDD, set to 6.56E-05.
DKRS23: adsorbed phase transformation rate for DDE to DDD, set to 0.0.
dissolved phase decay rate, see Table 3.2 of this report.
adsorbed phase decay rate, see Table 3.2 of this report.
vapor phase decay rate, set to 0.0.
thickness of compartments in the horizon, obtained from PATRIOT.
field capacity in the horizon, obtained from PATRIOT.
wilting point in the horizon, obtained from PATRIOT.
percent organic carbon in the horizon, obtained from PATRIOT.
pesticide partition coefficient, computed as Kd = (^OC)*!!^); values of
from Table 3.2 of this report.
r:'P:c;«-j'EPA\EPA OOi_:i!03'.R06-99.I67.wpd
A-5
HydroGeolcgic. Inc. 9199
------- |