- Email: [email protected]
Simulated effects of irrigation management in groundwater contamination
Agricultural Water Management, 20 (1992) 281-297
281
Elsevier Science Publishers B.V., Amsterdam
Simulated effects of irrigation management in groundwater contamination William F. McTernan a and Edward D. Mize b aSchool of Civil Engineering, Oklahoma State University, University Center at Tulsa, Tuba, Oklahoma, USA bProjectEngineer, Williams Pipe Line Company, Tulsa, Oklahoma, USA (Accepted 19 September 1991 )
ABSTRACT McTernan, W.F. and Mize, E.D., 1992. Simulated effects of irrigation management in groundwater contamination. Agric. Water Manage., 20:281-297. A computer-based evaluation of alternative irrigation management options was evaluated for one county in west-central Oklahoma. The purpose of the study was to determine the probabilities of groundwater contamination from agricultural chemicals, particularly pesticides, as the county moves from dryland farming to irrigated agriculture. A scientifically-based irrigation approach, approximating that possible with soil moisture probes, was compared to irrigation on schedule as well as to the dry-land farming base condition. Monte Carlo simulation, using the U.S. Environmental Protection Agency's (EPA) pesticide root zone model (PRZM), combined with Oak Ridge National Laboratory's AT123D code, was employed to identify the probabilities of contamination. PRZM was used to route pesticides from points of application to the top of the water table while AT 123D was employed to simulate the transport of chemicals throughout the aquifer. The results showed that the scientifically-based irrigation option very closely approximated the dry land base case, while irrigation on schedule produced sharply higher probabilities of contamination. In all cases, however, those events which resulted in pesticide reaching target locations in environmentally significant concentrations were extreme conditions where highly soluble pesticides with low decay rates were introduced during wet cycles onto highly permeable soils.
INTRODUCTION
Pesticide applications onto farm lands in the United States totaled approximately 235 000 tonnes (t) in 1984 (OTA, 1984). In Oklahoma, these figures equaled almost 2000 t of active ingredients from the 20 most commonly used chemicals (Criswell, 1982). There is growing concern that some of these chemicals could leach into shallow groundwaters and offer significant risks to the ultimate users of these resources. Over 17 pesticides have been found in the groundwaters of 23 different states (USEPA, 1986) with concentrations ranging from 0.1 to 700/zg/1. Monitoring for these chemicals may prove de0378-3774/92/$05.00 © 1992 Elsevier Science Publishers B.V. All rights reserved.
282
W.F. MCTERNAN AND E.D. MIZE
ficient, however, when attempts to evaluate previously untested management alternatives are needed. Simulation modeling has proven attractive in evaluating various, previously untried, options in a relatively short time at moderate expense. Some problems associated with modeling, however, include the uncertainties associated with the selection of input parameters for these codes. As an alternative to optimization techniques or deterministic modeling, the we used a Monte Carlo simulation approach to address some of the uncertainties associated with pesticide transport to and within an aquifer system in an area where the agricultural base is changing from dry-land farming to one based upon irrigation through which wider range of crops can be grown. The question under investigation is whether this change will significantly increase the probabilities of ground and surface water contamination from the pesticides used.
Experimental Site Selection Caddo County, Oklahoma was selected for this initial effort as a significant increase in the number of irrigation systems has been reported (USDA 1985 and Nelson, 1988 ). The county (Fig. 1 ), located in west-central Oklahoma, has an area of approximately 327 000 hectares (ha) and derives the majority of its agricultural income from the sale of peanuts, wheat, cotton, grain sorghum and hay. In 1985, 950 farms irrigated a total of 25 388 hectares (Kizer, 1985 ). Ninety-nine percent of these farms employed sprinkler systems with 90% of the land irrigated with groundwater. Irrigation used an estimated 100 million m 3 water in 1979 which comprised 88% of all municipal and irrigation water used in the county (Pettyjohn et al., 1984).
CADDOCOUNTY, OKLAHOMA
~
Fig. 1. Study area location map.
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
283
Irrigation Approaches Evaluated Simulation of the irrigation systems used in this county included a method where water was regularly added regardless of soil or climatic conditions, as well as a scientific approach that applied water based upon soil moisture. The simpler of these is often utilized and was referred to as "traditional" in this effort, while the evaluations based upon soil moisture conditions were called "scientific". This latter approach parallels a system where soil moisture probes would trigger irrigation on demand. The rainfall record randomly selected for each annual simulation was unaltered for the base or non-irrigated conditions while the traditionally managed irrigation option employed the equivalent of 770 m3/ha of additional water each month. This approximated that needed for corn growth in southwestern Oklahoma and, as such, represented an extreme value when compared to that needed for other crops (Nelson, 1988 ). This water was regularly added throughout the growing season with no regard for the existing soil moisture. It often produced conditions of increased surface runoff due to preexisting high soil water contents. The scientifically managed irrigation precipitation record was constructed by completing 25 years of daily soil moisture simulations. When the simulated soil moisture decreased to 1.5 times the wilting point within the soil surface layers (approximately the top 0.1 m), additional water was added on a daily basis to return to field capacity (Elliot, 1988; USDA, 1985 ). This data set was intended to approximate the irrigation record which would result if more sophisticated soil moisture monitoring techniques were employed.
Simulation Approach Computer codes employed Two publicly available computer codes were utilized: the U.S. Environmental Protection Agency's pesticide root zone model (PRZM) and Oak Ridge National Laboratory's AT 123D (Carsel et al., 1984; Yeh, 1981 ). PRZM is a finite difference model, developed to track agricultural chemicals from points of application through soils into runoff and percolated groundwaters. Based upon water and chemical balances coupled to partially saturated reference states for moisture content and drainage rate, PRZM was developed to allow for management-level simulations with minimal, readily available soil, chemical and meteorological data. It was designed to track agricultural chemicals from points of application to the top of shallow water table aquifers. It was also used to define the daily soil moisture profile employed in defining the scientifically managed precipitation record. AT 123D is an analytical model developed to simulate contaminant transport through saturated aquifer systems where one, two, or three dimensions
284
W.F. MCTERNAN AND E.D. MIZE
can be approximated. Different loading functions can be employed to simulate chemical introduction at the water table. Biodecay and retardance are available to account for transport in addition to the usual advective and dispersive mechanisms. Pesticide mass delivered to the top of the water table by PRZM was subsequently routed through a three dimensional aquifer volume by AT123D. The mass routed, together with the volume contaminated, produced pesticide concentrations at downgradient points. Monte Carlo methods The Monte Carlo method repeatedly and randomly selected data for individual deterministic transport models. The leaching outputs from these sequential simulations were then pooled and arranged into statistical distributions to define the probability that a given condition could occur. The distributions used to develop the necessary input values describing soil conditions were prepared from readily available sources (Gray and Roozitalab, Gray and Galloway, 1969 and USDA, 1973 ). The input file for the vadose zone model provided random selection of soil organic matter, bulk density, wilting point, field capacity, rainfall year and depth to groundwater. These data were combined with select fixed variables such as pesticide decay and partition coefficients, as well as with various cropping and tillage options. The meteorological data came from records compiled at the nearest Type 1 station maintained by the U.S. Weather Bureau (Oklahoma Climatological Survey, 1988 ) and consisted of daily rainfall and temperature as well as evaporation measured with Class A pan. A single annual rainfall period was randomly selected from a twenty-five year record for each individual, annual simulation. The Monte Carlo simulation resulted in the full range of rainfall records being accessed while the simulated depth to groundwater ranged from approximately 0.67 to almost 20 m. A normal distribution function developed from U.S. Geological Survey data describing depth to the water table aquifer in Caddo County was randomly accessed (USGS, 1986). Pesticide loads resulting from these simulations were then used as input into the 3-D transport code (AT 123D) to route the chemical delivered to the top of the water table by the vadose zone model through the receiving aquifer. A finer resolution of pesticide distribution over time, however, was needed than that provided by the previous PRZM simulations. The conditions resuiting in significant annual pesticide leaching were then repeated on a monthly basis to determine the appropriate temporal distributions of these materials at the top of the water table. Cumulative density functions describing the mass of pesticide leached below the root zone as well as that delivered to the top of the water table were determined. The probabilities associated with peak pesticide concentrations within the aquifer and the aquifer volumes affected by the contaminant plumes
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
285
were also identified. These were done on an annual basis on the last day of each simulation year but comparisons to the maximum values simulated throughout the year were also completed.
Parameter Selection Vadose Zone Modeling In accordance with previous efforts, soil organic matter, bulk density, wilting point, field capacity, and rainfall year were selected as random variables (Carsel et al., 1988 and McTernan et al., 1990). Additionally, depth to groundwater was also treated as a random variable and subsequently accessed. Pesticide selection was modeled using the partition and decay coefficients, Koc and Ks, respectively, while cropping and tillage alterations were addressed by modifications to U.S. Soil Conservation Service (SCS) curve numbers (CN). Originally developed by the SCS to predict rainfall excess at ungauged watersheds, curve numbers are based on drainage and storage assumptions to partition rainfall between the various environmental compartments such as runoff, infiltration, detention storage and others. They can be approximated by soil type and agronomic practices from tables available in various texts and guidance documents (USDASCS, 1972 ). These latter three parameters were selected sequentially rather than randomly from a range of values consistent with those either practiced or possible for Caddo County, Oklahoma agriculture. The random parameters represented physical features which could occur anywhere within the study area while these fixed variables included management alternatives which would only vary in response to economic or environmental considerations. In this way, a risk-based, sequential evaluation of the effects of selected management practices upon pesticide leaching and transport was attempted for all locations within the study area. The soils simulated ranged from free draining (type A) with C N = 6 7 to fairly impermeable soils (type D) with C N = 91. The degradation rate constant per day, Ks, was chosen to be either 0.0023 (Benomyl) or 0.2961 (Parathion) thus bracketing the range of chemicals usedqn the project area. This spectrum of available decay rates allowed interpolation of almost any given Ks, and thus extended the range of chemicals which could be evaluated. Similarly, the organic carbon distribution coefficient, Koc, was chosen to be 0.001 (MSMA), 2.0 (Dicamba) or 600.0 (Phorate) to provide upper and lower limits of pesticide solubility. Table 1 summarizes the basic configuration of the simulations attempted. These data sets were based on various combinations of similar fixed input parameters consisting of curve numbers, partition and decay properties. The ranges of "Rainfall year" and "Depth to Groundwater" simulated for each data set are also shown in this table and are considered sufficient to provide
286
W.F. MCTERNAN AND E.D. MIZE
TABLE 1 Randomly chosen rainfall and depth to groundwater records Data Set
I 2 3 4 5 6 7 8 9 10 11 12
Degradation rate/day Ks 0.2961 0.0023 0.0023 0.0023 0.2961 0.2961 0.2961 0.0023 0.0023 0.0023 0.2961 0.2961
Organic carbon distribution Koc
CN
600 600 0.001 2 2 0.001 600 600 0.001 2 2 0.001
91 91 91 91 91 91 67 67 67 67 67 67
Rainfall year
Depth to groundwater (m)
Min
Max
Min
Max
1955 1954 1957 1955 1955 t956 1954 1955 1960 1954 1957 1955
1977 1969 1970 1976 1975 1976 1978 1974 1978 1977 1973 1978
0.35 2.75 0.56 5.19 3.67 0.56 1.03 0.89 2.63 2.30 1.52 0.51
17.76 19.28 17.66 18.72 17.84 19.10 19.63 17.61 19.41 18.29 19.51 19.48
plausible comparisons. Simulation was repeated until a plot of 75% probability value versus the number of simulations approached a relatively constant value. At that point the exercise had achieved an acceptable level of precision and additional simulations would add little. Admittedly, this precision level is less restrictive than that employed in other efforts but it was deemed acceptable for the proof-of-concept level of resolution required for this work. In an effort to simulate leaching extremes for crops which could be cultivated within the study area if irrigation was adopted, the pesticide delivered to 0.3 m was evaluated. This approximated the shallowest root depth expected for a crop grown within the study area. A constant root depth of 0.8 m approximating corn, a heavy water user, was used for all simulations. Pesticide application occurred on May 1 with crop emergence 10 days later while harvest was in October. These values were reasonable for a large range of crops (from wheat to corn) and allowed conservative estimates of leaching potentials for a variety of conditions. Pesticide leaching at 0.3 m illustrated the potential economic loss of chemical due to over application while data collected at the top of the water table simulated the amount of contaminant available to enter the aquifer. Cumulative density functions describing the mass of pesticide leached below the initial level as well as that delivered to the top of the water table were determined for each of the three rainfall records. The amount of pesticide delivered to the aquifer was subsequently input to a saturated zone code.
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
287
Saturated Zone Model The saturated zone model was constructed to simulate a single non-changing water table condition within the study area. Parameters for this model are shown in Table 2 and were considered typical of the area (Garner, 1988). This model was used to determine the contaminant concentration and subsequent contaminated volume at the end of each simulation year. Only those simulations where leaching from the vadose zone exceeded 1E12 kilograms per hectare (that is 1 × 10-12) could pesticide be realistically detected in the underlying saturated aquifer. Monthly chemical load data were subsequently developed only for those previously completed annual simulations where leaching exceeded this value. The monthly loads were used as inputs to the 3-D saturated zone model.
Additional Pesticide transport in surface runoff was also evaluated for each of the three irrigation practices as a comparison was necessary to address environmental partitioning of the chemicals. This was completed in conjunction with the pesticide leaching effort within the P R Z M code. P R Z M fit a trapezoidal unit hydrograph to the daily rainfall excess to determine runoff. The pesticide transported with this water was determined through mass balance. As all simulations used unit application rates of 1 k g / h a / y a significant difference in the amount of pesticide leached should be accompanied by an equivalent difference, observed in one or more of the other environmental compartments available. Improvement of the groundwater condition at the expense of surface water resources was deemed unacceptable. TABLE2 Simulated water table aquifer Parameter
Value
Porosity Hydraulic gradient Hydraulic conductivity per hour Longitudinal dispersion Transverse dispersion Vertical dispersion Thickness Width Length
15% 0.0034 0.18 m 10 m 1m 1.0 m infinite infinite infinite
288
W.F.MCTERNANAND E.D. MIZE
RESULTS
Probability determinations Pesticide leaching probabilities for all three irrigation options at 0.3 m of soil depth are shown in Fig. 2, while Fig. 3 presents the corresponding data at the top of the water table. As can be observed, the traditional irrigation simulation had a slightly higher probability of pesticide leaching beyond the initial depth than did the scientific or no irrigation management practices. The similar slope and close plotting proximity of these data indicate similar leachLEACHING AT 0.3m DEPTH f
J
,
,
,
1.00 10E-2
i
10E-4
o~ 10E-6 T R A D I T I O N A l , SCIENTIFIC IRRIGATION--J/ / J /' NO IRRIGATION.... ~'
10E-8 10E-10
lb
5'0
8o 9'o ~5
99
~99.9
PERCENT TIME LESS THAN
Fig. 2. Leachingprobabilitiesfor root zone simulations. LEACHING AT DEPTH TO GROUNDWATER A
1.00 10E 2
[
10E-4
~
'O TRADITIONALIRRIGATION • SCIENTIFICIRRIGATION • NO IRRIGATION
o~ IOE 6 10E-8 10E-10 10E-12
5 10 PERCENT TIME LESS THAN
Fig. 3. Leaching probabilities at variable water table depth.
50
SIMULATED EFFECTSOF IRRIGATION MANAGEMENTIN GROUNDWATERCONTAMINATION
289
ing characteristics for all three methods at the initial depth. The amount of pesticide leached beyond the 0.3 m depth represented an over application of pesticide for shallow rooted plants which might be feasibly grown in a the study area. Leachate beyond the simulated root zone depth, or below the plant's maximum depth for utilization of pesticide uptake, represents a potential contaminant to any underlying aquifers. A review of pesticide leaching at a depth of 0.3 m for each of the three types TABLE3 Annual leaching output at 0.3 m ( k g / h a / y ) Data Set
1 2 3 4 5 6 7 8 9 10 11 12
No irrigation (Base case) Leaching output
Traditional irrigation Leaching output
Scientific irrigation Leaching output
Min.
Max.
Min.
Max.
Min.
Max.
3.8E-10 2.3E-2 2.1E-1 2.7E-1 0 3.6E-4 4.8E-5 3.6E-7 6.9E-1 5.2E-1 1.1E-3 5.9E-4
2.4E-3 3.9E-1 8.9E-1 4.9E-1 5.0E-2 2.1E-2 3.1E-3 6.1E-2 9.0E-1 8.2E-1 1.4E- 1 1.3E-1
1. IE-9 3.8E-2 2.8E-1 2.3E-1 6.5E-6 3.5E-4 5.0E-5 6.0E-5 7.8E-1 7.4E-1 1.2E-3 3.6E-4
2.4E-3 3.6E-1 9.2E-1 4.7E-1 5.0E-2 2.1E-2 3.1E-3 1.1E-1 8.8E-1 8.9E-1 1.4E- 1 1.3E-1
6.0E-10 2.7E-2 3.4E-1 2.7E-1 1.5E-5 6.5 E-4 4.8E-5 6.8E-7 7.1E-1 7.0E-1 5.4E-3 6.9E-4
4.4E-2 4.2E-1 9.1E-1 4.2E-1 5.0E-2 2.1E-2 3.1E-3 6.4E-2 9.2E-1 8.4E-1 1.4E- 1 1.3E-I
TABLE 4 Leaching output at depth to groundwater ( k g / h a / y ) Data set
1 2 3 4 5 6 7 8 9 10 11 12
No irrigation (Base Case) Leaching output
Traditional irrigation Leaching output
Scientific irrigation Leaching output
Min
Max
Min
Max
Min
Max
0 0 0 0 0 0 0 0 0 0 0 0
0 0 5.3E-1 2.2E-11 0 6.6E-4 0 0 2.4E-1 1.6E-1 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 8.7E-1 2.8E-7 0 1.5E-5 0 0 8.1E-1 8.4E-1 7.6E-11 5.1E-10
0 0 0 0 0 0 0 0 0 0 0 0
0 0 7.7E-1 5.1E-11 0 6.7E-4 0 0 4.0E-I 3.6E-1 0 0
290
W.F. MCTERNAN AND E.D. MIZE
MAXIMUM CONCENTRATION SIMULATED
II
10E--4
z O I..,<
O TRADITIONAL IRRIGATION IE IIGF~TIIRNRIG ATION
10E -6
rr
t-z
10E--8
w z 0 10E-10
10E-12 99
PERCENT TIME LESS THAN
99.15
Fig. 4. Probabilities of contamination in shallow water table aquifer systems. AFFECTED VOLUME SIMULATED 10E 6
10E
,~ TRADITIONAL IRRIGAFION • SCIENTIFIC IRRIGATION • NO IRRIGATION
II
IOE
10E
10E
10E 99
PERCENT TIME LESS THAN
99.2
Fig. 5. Probabilitiesof contaminated volumes in shallow water table aquifers. of simulations is presented in Table 3. This table contains the minimum and maximum leaching values for the twelve data sets for the three water management practices simulated. Although a general increase in pesticide leaching was observed for the low runoff soils (data sets 7 through 12 ), some marked effects resulting from pesticide selection were also observed. As expected, higher decay rates together with the larger partition coefficients resulted in lowered leaching of pesticide. In particular, those conditions which allowed pesticide to be retained in the soil a sufficiently long time to bring about decay were most beneficial to reducing pesticide leaching. That is, those situations where high levels of ad-
SIMULATED EFFECTS OF IRRIGATION MANAGEMENTIN GROUNDWATERCONTAMINATION
291
TABLE 5 Pesticide runoff for various irrigation management practices ( k g / h a / y ) Data set
Maximum pesticide runoff (kg)
1
2 3 4 5 6 7 8 9 10 11 12
high runoff soils
low runoff soils
No irrigation
Sci. irrigation
Trad. irrigation
7.7E-4 4.8E-1 5.7E-1 4.9E-1 8.2E-2 3.2E-2 3.7E-4 9.3E-2 7.2E-2 9.6E-2 1.3E-2 1.5E-2
2.9E-2 4.6E-1 5.7E-1 5.1E-1 8.2E-2 3.2E-2 3.7E-4 2.2E-2 7.2E-2 9.3E-2 1.3E-2 1.5E-2
1.0E-1 5.8E-1 6.7E-1 7.1E-1 8.2E-2 3.2E-2 3.7E-4 8.4E-2 1.6E-1 1.7E-1 1.3E-2 1.5E-2
.!'1
TO) Fig. 6. Typical 3-D plot for dry-land base case.
sorption combined with proper timing of chemical applications and increased decay resulted in lowered pesticide concentrations in the shallow-root zone.
292
W.F. MCTERNAN AND E.D. MIZE
Fig. 7. Typical3-D plot for "'scientific"irrigation. The leaching at depth to groundwater, Table 4, again was more pronounced for the low runoff soils. A more significant increase in pesticide delivered to these greater depths was observed, however, for traditional irrigation practices. Of importance also is the fact that pesticides with high partition or decay coefficients were generally not delivered to the depths necessary to intercept the water table aquifers simulated in this effort regardless of the water management approach practiced. Pesticide leaching probabilities for all three irrigation options at the randomly accessed variable "depth to groundwater" presented in Fig. 3, showed that most combinations of the fixed and variable input data resulted in conditions which did not leach to groundwater. Only extreme conditions allowed pesticides to reach the water table. Figure 3 clearly showed, however, that the traditional irrigation simulations had higher probabilities of delivering contaminants to underlying aquifers than did the other two management practices. The leaching values from Fig. 3 were subsequently used as loads for the saturated zone code model (AT 12 3D ) to develop Figs. 4 and 5. These figures present the probabilities of peak pesticide concentration and affected aquifer volumes, respectively, for the three water management systems simulated. As can be observed from these figures, traditional irrigation management techniques had the highest probability of contamination over a larger portion of
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
293
Fig. 8. Typical 3-D plot for "traditional" irrigation. the receiving aquifer. Similarly, the contaminant peaks were greater for this case than for the others. Figure 4 shows that a pesticide which might result at a concentration of 10E-12 mg/1 from the No or scientific irrigation simulations would exhibit a concentration in the underlying aquifer of 10E-7 mg/1 if traditional irrigation simulation was utilized. Figure 5 also shows a dramatic increase in affected aquifer volume when traditional irrigation was compared with the scientific and No irrigation alternatives. This was further highlighted by observing the difference in slopes between the traditional and other two plots along with the close parallel plots for the No and scientific irrigation as compared to the traditional simulation. This figure indicated that a pesticide which might affect 1000 m 3 of groundwater utilizing the scientific or No irrigation simulation technique had the potential to affect 10 times that amount if the traditional irrigation simulation technique were alternatively chosen. Pesticide not leached to groundwater is potentially available for discharge with surface runoff and offers an equal or greater environmental impact. The m a x i m u m amount of pesticide carried offsite due to runoff is shown for each simulated data set in Table 5. The highest pesticide runoffobserved was 71%, occurring in data set 4, which simulated traditional irrigation practices at a high runoff site (CN = 91 ).
294
W.F. MCTERNAN AND E.D. MIZE
These data were also influenced by the rainfall year as well as the pesticide decay and partition coefficients. That is, a high amount of rainfall on a site with low permeability with a pesticide with minimal adsorptive and decay properties resulted in significant surface water contamination. A dry year combined with a low runoff soil and a readily degradable pesticide (data set 7 ) resulted in the lowest amount of pesticides in the surface runoff (0.037%). Table 5 shows that for those pesticides possessing high decay rates, the concentrations in the runoff were essentially the same regardless of the irrigation method practiced. Data set 1 was an exception to this observation and seems due to the unfortunate timing of pesticide application with a precipitation event.
Specific simulation comparisons Evaluation of selected simulations which showed contamination of the underlying aquifer were performed using three dimensional plots (Figs. 6-8 ) to provide increased interpretation. This particular simulation set was chosen as it approximated the average depth of pesticide penetration for the low Ks and Koc trials which leached to groundwater. Figure 6 represents the No irrigation simulation which utilized only natural rainfall in the meteorological file while Figs. 7 and 8 represent the scientific and traditional irrigation practices respectively. Particular attention should be paid to comparisons of maximum concentration, maximum affected volume, and the shape of the 3-D plots. The affected volume of the traditional irrigation figure required a scale change and was approximately 10 times the volume of the No irrigation simulation, while the scientific simulation was only twice that of the No irrigation simulation. The maximum concentration in the aquifer was the peak concentration of the pesticide observed. It should be noted that the maximum simulated concentration of the traditional irrigation plot was approximately 5E-7 mg/1 as compared to the No irrigation concentration of 8E-12. Again, scientific irrigation approximated that of the No irrigation with a value of approximately 3E- 11. These results were very similar to the overall values observed earlier in Figs. 4 and 5 and are due to a' rapid flushing of the chemical by the increased water volumes associated with traditional irrigation. DISCUSSION
The traditionally managed irrigation system exhibited a greater pesticide leaching probability than did the other systems. The pesticides which leached beyond the initial root depth but did not reach the water table decayed, were
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
295
adsorbed, or were stored within the soil column. Of this group of chemicals, those pesticides with the higher decay rates (Ks) generally exhibited in excess of 90% loss within the soil column, while those with lower decay rates generally remained in the soil regardless of the Koc or the water management technique employed. This high percentage of storage by pesticides with low decay rates suggests a potential for future migration and contamination of the water table by these chemicals in subsequent years. The maximum depth of groundwater contaminated with detectable pesticide was 7.9 m. This depth resulted from random selection and, as such, did not represent the maximum depth to which the simulated pesticides could leach. The deepest penetration of pesticide not reaching groundwater was almost 15 meters and was attributable to a low Koc and Ks, highly impermeable soils, and a significant amount of rainfall and irrigant. The similarities between the No and scientific irrigation data suggest that the use of scientific irrigation practices have the potential to increase farm revenues while causing minimal additional groundwater contamination. Furthermore, scientific, as compared to traditional irrigation generally resulted in less pesticide runoff in surface waters. Results of these simulations indicated that, in general, excess irrigation water rather than pesticide selection increased the percentage of pesticide runoff. Approximately 20% (0.2 k g / h a / y) of the additional pesticide load in the runoff resulted when traditional irrigation was compared to the other options. This additional contamination occurred simultaneously with approximately a three-fold increase in water volume applied during traditional irrigation. Reduction of pesticides in the runoff has the potential for increasing profits as less money would be spent on "unused" pesticide in addition to reducing the potential for surface water contamination. The large volume of water associated with traditional irrigation provided a vehicle for higher percentages of pesticides to be flushed through the vadose zone into the underlying water table aquifer. Increased peak concentrations were observed even when the annual pesticide mass discharged to the aquifer was constant between irrigation methods as less dilutant was available in the shortened delivery times relative to the other approaches. A review of the plots presenting the simulations which leached to groundwater showed those pesticides with large decay constants generally exhibited peak concentrations immediately upon entering the aquifer, decreasing rapidly from that point. Similarly, those pesticides which had low decay rates generally showed a gradual increase in aquifer concentration before peaking and subsequently decreasing. This was true for the entire data set except for those simulations describing very shallow water tables. These conditions allowed for rapid movement of the chemical into the aquifer while providing a very limited storage and consequently, limited reaction time within the vadose zone.
296
W.F. M C T E R N A N A N D E.D. MIZE
It should be stated that the vast majority of the possible locations addressed by this effort exhibited poor probabilities of groundwater contamination. However, it should also be noted that the amount of groundwater contamination indicated by this effort could be severe for selected conditions. Site specific evaluations should be conducted prior to the adoption o f irrigation systems in high risk areas to insure proper compliance with ground and surface water quality goals. SUMMARY A risk based evaluation of select management alternatives potentially available to control agricultural groundwater contamination from pesticide leaching was completed for a single county in Oklahoma. The methodology employed in this report utilized existing software and data to determine pesticide leaching probabilities. Further, the interlinking of the Monte Carlo techniques, the pesticide root zone model, and the AT123D saturated zone code provided a detailed evaluation o f pesticide leaching in the vadose zone and subsequent m o v e m e n t through an affected water table aquifer. This analysis indicated that pesticide selection as well as imprudent irrigation practices were more critical in allowing pesticides to leach to and be transported in water table aquifers than were other alternatives available to the agricultural community. N o t surprisingly, pesticide leaching concentration was most severe in areas of extremely shallow water tables. Additionally, regardless o f the water management approach practiced, pesticides with high partition or decay coefficients were generally not delivered to the depths necessary to intercept the aquifers simulated. However, at the extreme conditions of shallow water tables, low partition or decay coefficients, and high rainfall years, the water management technique employed exhibited dramatic effect in terms of the contaminated volume and the peak concentration simulated within the water table aquifer.
REFERENCES Carsel, R.F., Smith, C.N., Mulkey, L.A., Dean, J.D. and Jowise, P., 1984. User's Manual for the Pesticide Root Zone Model (PRZM): Release 1. EPA-600/3-84-109. Athens, GA: U.S. EPA. Carsel. R.F., Parrish, R.S., Jones, R.L., Hansen, J.L. and Lamb, R.L., 1988. Characterizing the uncertainty of"pesticide" leaching in "agricultural" soils. J. Contaminant Hydrol., V2:N2. Criswell, J.T., 1982. Use of pesticides on major crops in Oklahoma. 1981. Research Report P833. Division of Agriculture. Oklahoma State University. Stillwater, Oklahoma. Elliott, R., 1988. Personal communication. School of Agricultural Engineering. Oklahoma State University. Stillwater, Oklahoma. Garner, T.V., 1988. Application of Monte Carlo techniques to the determination of groundwater contamination risk in saturated two-dimensional aquifer systems. Unpublished Masters thesis. Oklahoma State University. Stillwater, Oklahoma, 123 pp.
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
297
Gray, F. and Galloway, H.M., 1969. Soils of Oklahoma. Miscellaneous publication. Agricultural Experiment Station. Oklahoma State University. Stillwater, Oklahoma. Gray, F. and Roozitalab, M.H., Undated. Benchmark and Key Soils of Oklahoma Agricultural Experiment Station. Oklahoma State University. Stillwater, Oklahoma. Kizer, M.A., 1985. Extension Irrigation Specialist. Irrigation Survey, Oklahoma. Oklahoma State University. Stillwater, Oklahoma. McTernan, W.F., Daniels, B.T. and Garner, T.V., 1990. A preliminary approach to determining probability of non-point source contamination fo shallow aquifers by agricultural chemicals. In: W.F. McTernan and E. Kaplan (Editors), Risk Assessment for Groundwater Pollution Control. American Society of Civil Engineers, New York. Nelson, J.R., 1988. Personal Communication. Department of Agricultural Economics. Oklahoma State University. Stillwater, Oklahoma. 1988. Oklahoma Climatological Survey, 1988. Computerized Meteorological Data Sets. Norman, Oklahoma. Office of Technology Assessment, 1984. Protecting the Nation's Groundwater from Contamination. OTA-O-233. U.S. Congress. Washington D.C. Pettyjohn, W.A., White, H. and Dunn, S., 1983. Water Atlas of Oklahoma. University Center for Water Research. Oklahoma State University. Stillwater, Oklahoma. March. United States Department of Agriculture, 1972. Soil Conservation Service. National Engineering Handbook. Section 4. Hydrology. Washington, D.C. United States Department of Agriculture. Soil Conservation Service, 1973. In cooperation with Oklahoma Agricultural Experiment Station. Soil Survey of Caddo County, Oklahoma. April. United States Department of Agriculture, 1985. Soil Conservation Service. Oklahoma Irrigation Guide. Stillwater, Oklahoma. United States Environmental Protection Agency. Pesticides in Groundwater. Office of Ground Water Protection. Washington, D.C., 1986. United States Geological Survey, 1986. Ground-Water levels in observation wells in Oklahoma. Period of Record to March, 1985. U.S.G.S. Open-File Report 86-314. Oklahoma City, Oklahoma. Yeh, G.T. AT 123D: analytical transient one- two- and three-dimensional simulation of waste transport in the aquifer system. Oak Ridge National Laboratory. Publication 1439. Oak Ridge, Tenn. 1981.
281
Elsevier Science Publishers B.V., Amsterdam
Simulated effects of irrigation management in groundwater contamination William F. McTernan a and Edward D. Mize b aSchool of Civil Engineering, Oklahoma State University, University Center at Tulsa, Tuba, Oklahoma, USA bProjectEngineer, Williams Pipe Line Company, Tulsa, Oklahoma, USA (Accepted 19 September 1991 )
ABSTRACT McTernan, W.F. and Mize, E.D., 1992. Simulated effects of irrigation management in groundwater contamination. Agric. Water Manage., 20:281-297. A computer-based evaluation of alternative irrigation management options was evaluated for one county in west-central Oklahoma. The purpose of the study was to determine the probabilities of groundwater contamination from agricultural chemicals, particularly pesticides, as the county moves from dryland farming to irrigated agriculture. A scientifically-based irrigation approach, approximating that possible with soil moisture probes, was compared to irrigation on schedule as well as to the dry-land farming base condition. Monte Carlo simulation, using the U.S. Environmental Protection Agency's (EPA) pesticide root zone model (PRZM), combined with Oak Ridge National Laboratory's AT123D code, was employed to identify the probabilities of contamination. PRZM was used to route pesticides from points of application to the top of the water table while AT 123D was employed to simulate the transport of chemicals throughout the aquifer. The results showed that the scientifically-based irrigation option very closely approximated the dry land base case, while irrigation on schedule produced sharply higher probabilities of contamination. In all cases, however, those events which resulted in pesticide reaching target locations in environmentally significant concentrations were extreme conditions where highly soluble pesticides with low decay rates were introduced during wet cycles onto highly permeable soils.
INTRODUCTION
Pesticide applications onto farm lands in the United States totaled approximately 235 000 tonnes (t) in 1984 (OTA, 1984). In Oklahoma, these figures equaled almost 2000 t of active ingredients from the 20 most commonly used chemicals (Criswell, 1982). There is growing concern that some of these chemicals could leach into shallow groundwaters and offer significant risks to the ultimate users of these resources. Over 17 pesticides have been found in the groundwaters of 23 different states (USEPA, 1986) with concentrations ranging from 0.1 to 700/zg/1. Monitoring for these chemicals may prove de0378-3774/92/$05.00 © 1992 Elsevier Science Publishers B.V. All rights reserved.
282
W.F. MCTERNAN AND E.D. MIZE
ficient, however, when attempts to evaluate previously untested management alternatives are needed. Simulation modeling has proven attractive in evaluating various, previously untried, options in a relatively short time at moderate expense. Some problems associated with modeling, however, include the uncertainties associated with the selection of input parameters for these codes. As an alternative to optimization techniques or deterministic modeling, the we used a Monte Carlo simulation approach to address some of the uncertainties associated with pesticide transport to and within an aquifer system in an area where the agricultural base is changing from dry-land farming to one based upon irrigation through which wider range of crops can be grown. The question under investigation is whether this change will significantly increase the probabilities of ground and surface water contamination from the pesticides used.
Experimental Site Selection Caddo County, Oklahoma was selected for this initial effort as a significant increase in the number of irrigation systems has been reported (USDA 1985 and Nelson, 1988 ). The county (Fig. 1 ), located in west-central Oklahoma, has an area of approximately 327 000 hectares (ha) and derives the majority of its agricultural income from the sale of peanuts, wheat, cotton, grain sorghum and hay. In 1985, 950 farms irrigated a total of 25 388 hectares (Kizer, 1985 ). Ninety-nine percent of these farms employed sprinkler systems with 90% of the land irrigated with groundwater. Irrigation used an estimated 100 million m 3 water in 1979 which comprised 88% of all municipal and irrigation water used in the county (Pettyjohn et al., 1984).
CADDOCOUNTY, OKLAHOMA
~
Fig. 1. Study area location map.
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
283
Irrigation Approaches Evaluated Simulation of the irrigation systems used in this county included a method where water was regularly added regardless of soil or climatic conditions, as well as a scientific approach that applied water based upon soil moisture. The simpler of these is often utilized and was referred to as "traditional" in this effort, while the evaluations based upon soil moisture conditions were called "scientific". This latter approach parallels a system where soil moisture probes would trigger irrigation on demand. The rainfall record randomly selected for each annual simulation was unaltered for the base or non-irrigated conditions while the traditionally managed irrigation option employed the equivalent of 770 m3/ha of additional water each month. This approximated that needed for corn growth in southwestern Oklahoma and, as such, represented an extreme value when compared to that needed for other crops (Nelson, 1988 ). This water was regularly added throughout the growing season with no regard for the existing soil moisture. It often produced conditions of increased surface runoff due to preexisting high soil water contents. The scientifically managed irrigation precipitation record was constructed by completing 25 years of daily soil moisture simulations. When the simulated soil moisture decreased to 1.5 times the wilting point within the soil surface layers (approximately the top 0.1 m), additional water was added on a daily basis to return to field capacity (Elliot, 1988; USDA, 1985 ). This data set was intended to approximate the irrigation record which would result if more sophisticated soil moisture monitoring techniques were employed.
Simulation Approach Computer codes employed Two publicly available computer codes were utilized: the U.S. Environmental Protection Agency's pesticide root zone model (PRZM) and Oak Ridge National Laboratory's AT 123D (Carsel et al., 1984; Yeh, 1981 ). PRZM is a finite difference model, developed to track agricultural chemicals from points of application through soils into runoff and percolated groundwaters. Based upon water and chemical balances coupled to partially saturated reference states for moisture content and drainage rate, PRZM was developed to allow for management-level simulations with minimal, readily available soil, chemical and meteorological data. It was designed to track agricultural chemicals from points of application to the top of shallow water table aquifers. It was also used to define the daily soil moisture profile employed in defining the scientifically managed precipitation record. AT 123D is an analytical model developed to simulate contaminant transport through saturated aquifer systems where one, two, or three dimensions
284
W.F. MCTERNAN AND E.D. MIZE
can be approximated. Different loading functions can be employed to simulate chemical introduction at the water table. Biodecay and retardance are available to account for transport in addition to the usual advective and dispersive mechanisms. Pesticide mass delivered to the top of the water table by PRZM was subsequently routed through a three dimensional aquifer volume by AT123D. The mass routed, together with the volume contaminated, produced pesticide concentrations at downgradient points. Monte Carlo methods The Monte Carlo method repeatedly and randomly selected data for individual deterministic transport models. The leaching outputs from these sequential simulations were then pooled and arranged into statistical distributions to define the probability that a given condition could occur. The distributions used to develop the necessary input values describing soil conditions were prepared from readily available sources (Gray and Roozitalab, Gray and Galloway, 1969 and USDA, 1973 ). The input file for the vadose zone model provided random selection of soil organic matter, bulk density, wilting point, field capacity, rainfall year and depth to groundwater. These data were combined with select fixed variables such as pesticide decay and partition coefficients, as well as with various cropping and tillage options. The meteorological data came from records compiled at the nearest Type 1 station maintained by the U.S. Weather Bureau (Oklahoma Climatological Survey, 1988 ) and consisted of daily rainfall and temperature as well as evaporation measured with Class A pan. A single annual rainfall period was randomly selected from a twenty-five year record for each individual, annual simulation. The Monte Carlo simulation resulted in the full range of rainfall records being accessed while the simulated depth to groundwater ranged from approximately 0.67 to almost 20 m. A normal distribution function developed from U.S. Geological Survey data describing depth to the water table aquifer in Caddo County was randomly accessed (USGS, 1986). Pesticide loads resulting from these simulations were then used as input into the 3-D transport code (AT 123D) to route the chemical delivered to the top of the water table by the vadose zone model through the receiving aquifer. A finer resolution of pesticide distribution over time, however, was needed than that provided by the previous PRZM simulations. The conditions resuiting in significant annual pesticide leaching were then repeated on a monthly basis to determine the appropriate temporal distributions of these materials at the top of the water table. Cumulative density functions describing the mass of pesticide leached below the root zone as well as that delivered to the top of the water table were determined. The probabilities associated with peak pesticide concentrations within the aquifer and the aquifer volumes affected by the contaminant plumes
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
285
were also identified. These were done on an annual basis on the last day of each simulation year but comparisons to the maximum values simulated throughout the year were also completed.
Parameter Selection Vadose Zone Modeling In accordance with previous efforts, soil organic matter, bulk density, wilting point, field capacity, and rainfall year were selected as random variables (Carsel et al., 1988 and McTernan et al., 1990). Additionally, depth to groundwater was also treated as a random variable and subsequently accessed. Pesticide selection was modeled using the partition and decay coefficients, Koc and Ks, respectively, while cropping and tillage alterations were addressed by modifications to U.S. Soil Conservation Service (SCS) curve numbers (CN). Originally developed by the SCS to predict rainfall excess at ungauged watersheds, curve numbers are based on drainage and storage assumptions to partition rainfall between the various environmental compartments such as runoff, infiltration, detention storage and others. They can be approximated by soil type and agronomic practices from tables available in various texts and guidance documents (USDASCS, 1972 ). These latter three parameters were selected sequentially rather than randomly from a range of values consistent with those either practiced or possible for Caddo County, Oklahoma agriculture. The random parameters represented physical features which could occur anywhere within the study area while these fixed variables included management alternatives which would only vary in response to economic or environmental considerations. In this way, a risk-based, sequential evaluation of the effects of selected management practices upon pesticide leaching and transport was attempted for all locations within the study area. The soils simulated ranged from free draining (type A) with C N = 6 7 to fairly impermeable soils (type D) with C N = 91. The degradation rate constant per day, Ks, was chosen to be either 0.0023 (Benomyl) or 0.2961 (Parathion) thus bracketing the range of chemicals usedqn the project area. This spectrum of available decay rates allowed interpolation of almost any given Ks, and thus extended the range of chemicals which could be evaluated. Similarly, the organic carbon distribution coefficient, Koc, was chosen to be 0.001 (MSMA), 2.0 (Dicamba) or 600.0 (Phorate) to provide upper and lower limits of pesticide solubility. Table 1 summarizes the basic configuration of the simulations attempted. These data sets were based on various combinations of similar fixed input parameters consisting of curve numbers, partition and decay properties. The ranges of "Rainfall year" and "Depth to Groundwater" simulated for each data set are also shown in this table and are considered sufficient to provide
286
W.F. MCTERNAN AND E.D. MIZE
TABLE 1 Randomly chosen rainfall and depth to groundwater records Data Set
I 2 3 4 5 6 7 8 9 10 11 12
Degradation rate/day Ks 0.2961 0.0023 0.0023 0.0023 0.2961 0.2961 0.2961 0.0023 0.0023 0.0023 0.2961 0.2961
Organic carbon distribution Koc
CN
600 600 0.001 2 2 0.001 600 600 0.001 2 2 0.001
91 91 91 91 91 91 67 67 67 67 67 67
Rainfall year
Depth to groundwater (m)
Min
Max
Min
Max
1955 1954 1957 1955 1955 t956 1954 1955 1960 1954 1957 1955
1977 1969 1970 1976 1975 1976 1978 1974 1978 1977 1973 1978
0.35 2.75 0.56 5.19 3.67 0.56 1.03 0.89 2.63 2.30 1.52 0.51
17.76 19.28 17.66 18.72 17.84 19.10 19.63 17.61 19.41 18.29 19.51 19.48
plausible comparisons. Simulation was repeated until a plot of 75% probability value versus the number of simulations approached a relatively constant value. At that point the exercise had achieved an acceptable level of precision and additional simulations would add little. Admittedly, this precision level is less restrictive than that employed in other efforts but it was deemed acceptable for the proof-of-concept level of resolution required for this work. In an effort to simulate leaching extremes for crops which could be cultivated within the study area if irrigation was adopted, the pesticide delivered to 0.3 m was evaluated. This approximated the shallowest root depth expected for a crop grown within the study area. A constant root depth of 0.8 m approximating corn, a heavy water user, was used for all simulations. Pesticide application occurred on May 1 with crop emergence 10 days later while harvest was in October. These values were reasonable for a large range of crops (from wheat to corn) and allowed conservative estimates of leaching potentials for a variety of conditions. Pesticide leaching at 0.3 m illustrated the potential economic loss of chemical due to over application while data collected at the top of the water table simulated the amount of contaminant available to enter the aquifer. Cumulative density functions describing the mass of pesticide leached below the initial level as well as that delivered to the top of the water table were determined for each of the three rainfall records. The amount of pesticide delivered to the aquifer was subsequently input to a saturated zone code.
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
287
Saturated Zone Model The saturated zone model was constructed to simulate a single non-changing water table condition within the study area. Parameters for this model are shown in Table 2 and were considered typical of the area (Garner, 1988). This model was used to determine the contaminant concentration and subsequent contaminated volume at the end of each simulation year. Only those simulations where leaching from the vadose zone exceeded 1E12 kilograms per hectare (that is 1 × 10-12) could pesticide be realistically detected in the underlying saturated aquifer. Monthly chemical load data were subsequently developed only for those previously completed annual simulations where leaching exceeded this value. The monthly loads were used as inputs to the 3-D saturated zone model.
Additional Pesticide transport in surface runoff was also evaluated for each of the three irrigation practices as a comparison was necessary to address environmental partitioning of the chemicals. This was completed in conjunction with the pesticide leaching effort within the P R Z M code. P R Z M fit a trapezoidal unit hydrograph to the daily rainfall excess to determine runoff. The pesticide transported with this water was determined through mass balance. As all simulations used unit application rates of 1 k g / h a / y a significant difference in the amount of pesticide leached should be accompanied by an equivalent difference, observed in one or more of the other environmental compartments available. Improvement of the groundwater condition at the expense of surface water resources was deemed unacceptable. TABLE2 Simulated water table aquifer Parameter
Value
Porosity Hydraulic gradient Hydraulic conductivity per hour Longitudinal dispersion Transverse dispersion Vertical dispersion Thickness Width Length
15% 0.0034 0.18 m 10 m 1m 1.0 m infinite infinite infinite
288
W.F.MCTERNANAND E.D. MIZE
RESULTS
Probability determinations Pesticide leaching probabilities for all three irrigation options at 0.3 m of soil depth are shown in Fig. 2, while Fig. 3 presents the corresponding data at the top of the water table. As can be observed, the traditional irrigation simulation had a slightly higher probability of pesticide leaching beyond the initial depth than did the scientific or no irrigation management practices. The similar slope and close plotting proximity of these data indicate similar leachLEACHING AT 0.3m DEPTH f
J
,
,
,
1.00 10E-2
i
10E-4
o~ 10E-6 T R A D I T I O N A l , SCIENTIFIC IRRIGATION--J/ / J /' NO IRRIGATION.... ~'
10E-8 10E-10
lb
5'0
8o 9'o ~5
99
~99.9
PERCENT TIME LESS THAN
Fig. 2. Leachingprobabilitiesfor root zone simulations. LEACHING AT DEPTH TO GROUNDWATER A
1.00 10E 2
[
10E-4
~
'O TRADITIONALIRRIGATION • SCIENTIFICIRRIGATION • NO IRRIGATION
o~ IOE 6 10E-8 10E-10 10E-12
5 10 PERCENT TIME LESS THAN
Fig. 3. Leaching probabilities at variable water table depth.
50
SIMULATED EFFECTSOF IRRIGATION MANAGEMENTIN GROUNDWATERCONTAMINATION
289
ing characteristics for all three methods at the initial depth. The amount of pesticide leached beyond the 0.3 m depth represented an over application of pesticide for shallow rooted plants which might be feasibly grown in a the study area. Leachate beyond the simulated root zone depth, or below the plant's maximum depth for utilization of pesticide uptake, represents a potential contaminant to any underlying aquifers. A review of pesticide leaching at a depth of 0.3 m for each of the three types TABLE3 Annual leaching output at 0.3 m ( k g / h a / y ) Data Set
1 2 3 4 5 6 7 8 9 10 11 12
No irrigation (Base case) Leaching output
Traditional irrigation Leaching output
Scientific irrigation Leaching output
Min.
Max.
Min.
Max.
Min.
Max.
3.8E-10 2.3E-2 2.1E-1 2.7E-1 0 3.6E-4 4.8E-5 3.6E-7 6.9E-1 5.2E-1 1.1E-3 5.9E-4
2.4E-3 3.9E-1 8.9E-1 4.9E-1 5.0E-2 2.1E-2 3.1E-3 6.1E-2 9.0E-1 8.2E-1 1.4E- 1 1.3E-1
1. IE-9 3.8E-2 2.8E-1 2.3E-1 6.5E-6 3.5E-4 5.0E-5 6.0E-5 7.8E-1 7.4E-1 1.2E-3 3.6E-4
2.4E-3 3.6E-1 9.2E-1 4.7E-1 5.0E-2 2.1E-2 3.1E-3 1.1E-1 8.8E-1 8.9E-1 1.4E- 1 1.3E-1
6.0E-10 2.7E-2 3.4E-1 2.7E-1 1.5E-5 6.5 E-4 4.8E-5 6.8E-7 7.1E-1 7.0E-1 5.4E-3 6.9E-4
4.4E-2 4.2E-1 9.1E-1 4.2E-1 5.0E-2 2.1E-2 3.1E-3 6.4E-2 9.2E-1 8.4E-1 1.4E- 1 1.3E-I
TABLE 4 Leaching output at depth to groundwater ( k g / h a / y ) Data set
1 2 3 4 5 6 7 8 9 10 11 12
No irrigation (Base Case) Leaching output
Traditional irrigation Leaching output
Scientific irrigation Leaching output
Min
Max
Min
Max
Min
Max
0 0 0 0 0 0 0 0 0 0 0 0
0 0 5.3E-1 2.2E-11 0 6.6E-4 0 0 2.4E-1 1.6E-1 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 8.7E-1 2.8E-7 0 1.5E-5 0 0 8.1E-1 8.4E-1 7.6E-11 5.1E-10
0 0 0 0 0 0 0 0 0 0 0 0
0 0 7.7E-1 5.1E-11 0 6.7E-4 0 0 4.0E-I 3.6E-1 0 0
290
W.F. MCTERNAN AND E.D. MIZE
MAXIMUM CONCENTRATION SIMULATED
II
10E--4
z O I..,<
O TRADITIONAL IRRIGATION IE IIGF~TIIRNRIG ATION
10E -6
rr
t-z
10E--8
w z 0 10E-10
10E-12 99
PERCENT TIME LESS THAN
99.15
Fig. 4. Probabilities of contamination in shallow water table aquifer systems. AFFECTED VOLUME SIMULATED 10E 6
10E
,~ TRADITIONAL IRRIGAFION • SCIENTIFIC IRRIGATION • NO IRRIGATION
II
IOE
10E
10E
10E 99
PERCENT TIME LESS THAN
99.2
Fig. 5. Probabilitiesof contaminated volumes in shallow water table aquifers. of simulations is presented in Table 3. This table contains the minimum and maximum leaching values for the twelve data sets for the three water management practices simulated. Although a general increase in pesticide leaching was observed for the low runoff soils (data sets 7 through 12 ), some marked effects resulting from pesticide selection were also observed. As expected, higher decay rates together with the larger partition coefficients resulted in lowered leaching of pesticide. In particular, those conditions which allowed pesticide to be retained in the soil a sufficiently long time to bring about decay were most beneficial to reducing pesticide leaching. That is, those situations where high levels of ad-
SIMULATED EFFECTS OF IRRIGATION MANAGEMENTIN GROUNDWATERCONTAMINATION
291
TABLE 5 Pesticide runoff for various irrigation management practices ( k g / h a / y ) Data set
Maximum pesticide runoff (kg)
1
2 3 4 5 6 7 8 9 10 11 12
high runoff soils
low runoff soils
No irrigation
Sci. irrigation
Trad. irrigation
7.7E-4 4.8E-1 5.7E-1 4.9E-1 8.2E-2 3.2E-2 3.7E-4 9.3E-2 7.2E-2 9.6E-2 1.3E-2 1.5E-2
2.9E-2 4.6E-1 5.7E-1 5.1E-1 8.2E-2 3.2E-2 3.7E-4 2.2E-2 7.2E-2 9.3E-2 1.3E-2 1.5E-2
1.0E-1 5.8E-1 6.7E-1 7.1E-1 8.2E-2 3.2E-2 3.7E-4 8.4E-2 1.6E-1 1.7E-1 1.3E-2 1.5E-2
.!'1
TO) Fig. 6. Typical 3-D plot for dry-land base case.
sorption combined with proper timing of chemical applications and increased decay resulted in lowered pesticide concentrations in the shallow-root zone.
292
W.F. MCTERNAN AND E.D. MIZE
Fig. 7. Typical3-D plot for "'scientific"irrigation. The leaching at depth to groundwater, Table 4, again was more pronounced for the low runoff soils. A more significant increase in pesticide delivered to these greater depths was observed, however, for traditional irrigation practices. Of importance also is the fact that pesticides with high partition or decay coefficients were generally not delivered to the depths necessary to intercept the water table aquifers simulated in this effort regardless of the water management approach practiced. Pesticide leaching probabilities for all three irrigation options at the randomly accessed variable "depth to groundwater" presented in Fig. 3, showed that most combinations of the fixed and variable input data resulted in conditions which did not leach to groundwater. Only extreme conditions allowed pesticides to reach the water table. Figure 3 clearly showed, however, that the traditional irrigation simulations had higher probabilities of delivering contaminants to underlying aquifers than did the other two management practices. The leaching values from Fig. 3 were subsequently used as loads for the saturated zone code model (AT 12 3D ) to develop Figs. 4 and 5. These figures present the probabilities of peak pesticide concentration and affected aquifer volumes, respectively, for the three water management systems simulated. As can be observed from these figures, traditional irrigation management techniques had the highest probability of contamination over a larger portion of
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
293
Fig. 8. Typical 3-D plot for "traditional" irrigation. the receiving aquifer. Similarly, the contaminant peaks were greater for this case than for the others. Figure 4 shows that a pesticide which might result at a concentration of 10E-12 mg/1 from the No or scientific irrigation simulations would exhibit a concentration in the underlying aquifer of 10E-7 mg/1 if traditional irrigation simulation was utilized. Figure 5 also shows a dramatic increase in affected aquifer volume when traditional irrigation was compared with the scientific and No irrigation alternatives. This was further highlighted by observing the difference in slopes between the traditional and other two plots along with the close parallel plots for the No and scientific irrigation as compared to the traditional simulation. This figure indicated that a pesticide which might affect 1000 m 3 of groundwater utilizing the scientific or No irrigation simulation technique had the potential to affect 10 times that amount if the traditional irrigation simulation technique were alternatively chosen. Pesticide not leached to groundwater is potentially available for discharge with surface runoff and offers an equal or greater environmental impact. The m a x i m u m amount of pesticide carried offsite due to runoff is shown for each simulated data set in Table 5. The highest pesticide runoffobserved was 71%, occurring in data set 4, which simulated traditional irrigation practices at a high runoff site (CN = 91 ).
294
W.F. MCTERNAN AND E.D. MIZE
These data were also influenced by the rainfall year as well as the pesticide decay and partition coefficients. That is, a high amount of rainfall on a site with low permeability with a pesticide with minimal adsorptive and decay properties resulted in significant surface water contamination. A dry year combined with a low runoff soil and a readily degradable pesticide (data set 7 ) resulted in the lowest amount of pesticides in the surface runoff (0.037%). Table 5 shows that for those pesticides possessing high decay rates, the concentrations in the runoff were essentially the same regardless of the irrigation method practiced. Data set 1 was an exception to this observation and seems due to the unfortunate timing of pesticide application with a precipitation event.
Specific simulation comparisons Evaluation of selected simulations which showed contamination of the underlying aquifer were performed using three dimensional plots (Figs. 6-8 ) to provide increased interpretation. This particular simulation set was chosen as it approximated the average depth of pesticide penetration for the low Ks and Koc trials which leached to groundwater. Figure 6 represents the No irrigation simulation which utilized only natural rainfall in the meteorological file while Figs. 7 and 8 represent the scientific and traditional irrigation practices respectively. Particular attention should be paid to comparisons of maximum concentration, maximum affected volume, and the shape of the 3-D plots. The affected volume of the traditional irrigation figure required a scale change and was approximately 10 times the volume of the No irrigation simulation, while the scientific simulation was only twice that of the No irrigation simulation. The maximum concentration in the aquifer was the peak concentration of the pesticide observed. It should be noted that the maximum simulated concentration of the traditional irrigation plot was approximately 5E-7 mg/1 as compared to the No irrigation concentration of 8E-12. Again, scientific irrigation approximated that of the No irrigation with a value of approximately 3E- 11. These results were very similar to the overall values observed earlier in Figs. 4 and 5 and are due to a' rapid flushing of the chemical by the increased water volumes associated with traditional irrigation. DISCUSSION
The traditionally managed irrigation system exhibited a greater pesticide leaching probability than did the other systems. The pesticides which leached beyond the initial root depth but did not reach the water table decayed, were
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
295
adsorbed, or were stored within the soil column. Of this group of chemicals, those pesticides with the higher decay rates (Ks) generally exhibited in excess of 90% loss within the soil column, while those with lower decay rates generally remained in the soil regardless of the Koc or the water management technique employed. This high percentage of storage by pesticides with low decay rates suggests a potential for future migration and contamination of the water table by these chemicals in subsequent years. The maximum depth of groundwater contaminated with detectable pesticide was 7.9 m. This depth resulted from random selection and, as such, did not represent the maximum depth to which the simulated pesticides could leach. The deepest penetration of pesticide not reaching groundwater was almost 15 meters and was attributable to a low Koc and Ks, highly impermeable soils, and a significant amount of rainfall and irrigant. The similarities between the No and scientific irrigation data suggest that the use of scientific irrigation practices have the potential to increase farm revenues while causing minimal additional groundwater contamination. Furthermore, scientific, as compared to traditional irrigation generally resulted in less pesticide runoff in surface waters. Results of these simulations indicated that, in general, excess irrigation water rather than pesticide selection increased the percentage of pesticide runoff. Approximately 20% (0.2 k g / h a / y) of the additional pesticide load in the runoff resulted when traditional irrigation was compared to the other options. This additional contamination occurred simultaneously with approximately a three-fold increase in water volume applied during traditional irrigation. Reduction of pesticides in the runoff has the potential for increasing profits as less money would be spent on "unused" pesticide in addition to reducing the potential for surface water contamination. The large volume of water associated with traditional irrigation provided a vehicle for higher percentages of pesticides to be flushed through the vadose zone into the underlying water table aquifer. Increased peak concentrations were observed even when the annual pesticide mass discharged to the aquifer was constant between irrigation methods as less dilutant was available in the shortened delivery times relative to the other approaches. A review of the plots presenting the simulations which leached to groundwater showed those pesticides with large decay constants generally exhibited peak concentrations immediately upon entering the aquifer, decreasing rapidly from that point. Similarly, those pesticides which had low decay rates generally showed a gradual increase in aquifer concentration before peaking and subsequently decreasing. This was true for the entire data set except for those simulations describing very shallow water tables. These conditions allowed for rapid movement of the chemical into the aquifer while providing a very limited storage and consequently, limited reaction time within the vadose zone.
296
W.F. M C T E R N A N A N D E.D. MIZE
It should be stated that the vast majority of the possible locations addressed by this effort exhibited poor probabilities of groundwater contamination. However, it should also be noted that the amount of groundwater contamination indicated by this effort could be severe for selected conditions. Site specific evaluations should be conducted prior to the adoption o f irrigation systems in high risk areas to insure proper compliance with ground and surface water quality goals. SUMMARY A risk based evaluation of select management alternatives potentially available to control agricultural groundwater contamination from pesticide leaching was completed for a single county in Oklahoma. The methodology employed in this report utilized existing software and data to determine pesticide leaching probabilities. Further, the interlinking of the Monte Carlo techniques, the pesticide root zone model, and the AT123D saturated zone code provided a detailed evaluation o f pesticide leaching in the vadose zone and subsequent m o v e m e n t through an affected water table aquifer. This analysis indicated that pesticide selection as well as imprudent irrigation practices were more critical in allowing pesticides to leach to and be transported in water table aquifers than were other alternatives available to the agricultural community. N o t surprisingly, pesticide leaching concentration was most severe in areas of extremely shallow water tables. Additionally, regardless o f the water management approach practiced, pesticides with high partition or decay coefficients were generally not delivered to the depths necessary to intercept the aquifers simulated. However, at the extreme conditions of shallow water tables, low partition or decay coefficients, and high rainfall years, the water management technique employed exhibited dramatic effect in terms of the contaminated volume and the peak concentration simulated within the water table aquifer.
REFERENCES Carsel, R.F., Smith, C.N., Mulkey, L.A., Dean, J.D. and Jowise, P., 1984. User's Manual for the Pesticide Root Zone Model (PRZM): Release 1. EPA-600/3-84-109. Athens, GA: U.S. EPA. Carsel. R.F., Parrish, R.S., Jones, R.L., Hansen, J.L. and Lamb, R.L., 1988. Characterizing the uncertainty of"pesticide" leaching in "agricultural" soils. J. Contaminant Hydrol., V2:N2. Criswell, J.T., 1982. Use of pesticides on major crops in Oklahoma. 1981. Research Report P833. Division of Agriculture. Oklahoma State University. Stillwater, Oklahoma. Elliott, R., 1988. Personal communication. School of Agricultural Engineering. Oklahoma State University. Stillwater, Oklahoma. Garner, T.V., 1988. Application of Monte Carlo techniques to the determination of groundwater contamination risk in saturated two-dimensional aquifer systems. Unpublished Masters thesis. Oklahoma State University. Stillwater, Oklahoma, 123 pp.
SIMULATED EFFECTS OF IRRIGATION MANAGEMENT IN GROUNDWATER CONTAMINATION
297
Gray, F. and Galloway, H.M., 1969. Soils of Oklahoma. Miscellaneous publication. Agricultural Experiment Station. Oklahoma State University. Stillwater, Oklahoma. Gray, F. and Roozitalab, M.H., Undated. Benchmark and Key Soils of Oklahoma Agricultural Experiment Station. Oklahoma State University. Stillwater, Oklahoma. Kizer, M.A., 1985. Extension Irrigation Specialist. Irrigation Survey, Oklahoma. Oklahoma State University. Stillwater, Oklahoma. McTernan, W.F., Daniels, B.T. and Garner, T.V., 1990. A preliminary approach to determining probability of non-point source contamination fo shallow aquifers by agricultural chemicals. In: W.F. McTernan and E. Kaplan (Editors), Risk Assessment for Groundwater Pollution Control. American Society of Civil Engineers, New York. Nelson, J.R., 1988. Personal Communication. Department of Agricultural Economics. Oklahoma State University. Stillwater, Oklahoma. 1988. Oklahoma Climatological Survey, 1988. Computerized Meteorological Data Sets. Norman, Oklahoma. Office of Technology Assessment, 1984. Protecting the Nation's Groundwater from Contamination. OTA-O-233. U.S. Congress. Washington D.C. Pettyjohn, W.A., White, H. and Dunn, S., 1983. Water Atlas of Oklahoma. University Center for Water Research. Oklahoma State University. Stillwater, Oklahoma. March. United States Department of Agriculture, 1972. Soil Conservation Service. National Engineering Handbook. Section 4. Hydrology. Washington, D.C. United States Department of Agriculture. Soil Conservation Service, 1973. In cooperation with Oklahoma Agricultural Experiment Station. Soil Survey of Caddo County, Oklahoma. April. United States Department of Agriculture, 1985. Soil Conservation Service. Oklahoma Irrigation Guide. Stillwater, Oklahoma. United States Environmental Protection Agency. Pesticides in Groundwater. Office of Ground Water Protection. Washington, D.C., 1986. United States Geological Survey, 1986. Ground-Water levels in observation wells in Oklahoma. Period of Record to March, 1985. U.S.G.S. Open-File Report 86-314. Oklahoma City, Oklahoma. Yeh, G.T. AT 123D: analytical transient one- two- and three-dimensional simulation of waste transport in the aquifer system. Oak Ridge National Laboratory. Publication 1439. Oak Ridge, Tenn. 1981.