Published in J. Environ. Qual. 33:1487-1498 (2004).
© ASA, CSSA, SSSA
677 S. Segoe Rd., Madison, WI 53711 USA
TECHNICAL REPORTS
Vadose Zone Processes and Chemical Transport
Assessing Ground Water Vulnerability with the Type Transfer Function Model in the San Joaquin Valley, California
Iris T. Stewarta,* and
Keith Loagueb
a Scripps Institution of Oceanography, University of California San Diego, La Jolla, CA 92093-0224
b Department of Geological and Environmental Sciences, Stanford University, Stanford, CA 94035-2115
* Corresponding author (istewart{at}meteora.ucsd.edu).
Received for publication September 16, 2003.
 |
ABSTRACT
|
|---|
The recently developed type transfer function (TTF) simulation approach was applied to generate a regional-scale nonpoint-source ground water vulnerability assessment for the San Joaquin Valley, California. The computationally comparatively inexpensive TTF approach produces quantitative estimates of contaminant concentrations for large regional scales through characteristic functions based on different soil textures and their leaching properties. The TTF simulations employed an extensive soil and recharge database to estimate atrazine (1-chloro-3-ethylamino-5-isopropylamino-2,4,6-triazine) concentrations at a compliance depth of 3 m resulting from a surface application. Two different sets of TTFs with two different levels of upscaling were used for spatially uniform and distributed recharge estimates. Results show that estimated atrazine concentrations can be related to soil survey information. Areas with high potential vulnerability to atrazine leaching were found for soils with low organic carbon content and sandy loam and loam textures. Travel times for atrazine peak concentrations to the compliance depth ranged from 350 to 730 d. The extent of areas with estimated atrazine concentrations above the maximum contaminant level was less extensive when uniform annual recharge values were used. Simulated TTF concentrations were highest for eastern Fresno County, a vulnerability pattern that is also supported by field observations. The TTF modeling approach is shown to be a useful tool for quantitative pesticide leaching estimates at regional scales significantly larger than those of previous studies.
Abbreviations: MCL, maximum contaminant level • NPS, nonpoint source • TTF, type transfer function
|
INTRODUCTION
|
|---|
OVER THE PAST DECADES, nonpoint-source (NPS) ground water contamination from agrochemicals has grown into one of the most pressing environmental concerns (Corwin et al., 1999), as NPS contaminants are widely distributed, may be persistent, may be toxic below the detection limit, and are routinely applied in modern farming practices. Of the pesticides in use today, the herbicide atrazine has been the most frequently detected parent compound in national water quality sampling studies (Barbash et al., 1999) and the second most common pesticide found in drinking water wells (Howard, 1989; USEPA, 1994). Atrazine's high potential for ground water contamination is the result of a low tendency for sorption in combination with a comparatively long half-life (Wauchope et al., 1992). The total amount of atrazine applied annually has reached approximately 36 million kg in the USA alone (Barbash et al., 1999).
The intensely farmed San Joaquin Valley in California (Fig. 1) is one of the areas within the USA where NPS agrochemical contamination in well waters has been detected frequently, and where contaminant concentrations have been among the highest (Barbash et al., 1999; Burow et al., 1998). Approximately half the amount of pesticide active ingredient reported used in California, which is on the order of 100 million kg per year, is applied in the San Joaquin Valley. At the same time, the water table aquifer underlying the semiarid San Joaquin Valley represents a critical domestic and irrigation water resource. For these reasons, the assessment of ground water vulnerability to pesticide leaching in this area is of great importance for regulatory agencies.
View larger version (75K):
[in this window]
[in a new window]
|
Fig. 1. The location of the San Joaquin Valley within California. The borders of the eight counties in the San Joaquin Valley are shown.
|
|
|
CURRENT APPROACHES TO NONPOINT-SOURCE GROUND WATER VULNERABILITY ASSESSMENT
|
|---|
In the past two decades, considerable research effort has been expended to characterize the mechanisms that control vadose-zone pesticide leaching (e.g., Cheng, 1990; Hutson and Roberts, 1990; Loll and Moldrup, 2000; Nofziger et al., 1996). Data collection campaigns (e.g., Domagalski and Dubrovsky, 1992) are expensive and difficult, and therefore mostly limited in scope. The modeling approaches currently used to assess ground water vulnerability can almost exclusively be categorized as either index and overlay methods, or process-based simulation models. Index and overlay methods simplify or neglect the physical processes and characterize vulnerability as a qualitative measure of relative leaching potential. Therefore, these methods are particularly suited to and have been predominantly used for applications at the regional scale, as they usually only require the determination of statistical relationships, based on selected soil and chemical properties, and net annual recharge (e.g., Blanke, 1999; Diaz-Diaz and Loague, 2001; Khan and Liang, 1989; Loague, 1994; Loague et al., 1990; Mills, 2004). Process-based simulation models of leaching are generally based on the numerical solution of Richards' equation with the advection–dispersion equation. At the regional scale, their usefulness has been restricted by the lack of available data and the tremendous computational effort associated with solving highly nonlinear equations. Consequently, only a few (e.g., Loague et al., 1998; Petach et al., 1991) regional-scale simulation studies of NPS agrochemical contaminant leaching have been attempted, none of these at the scale of the San Joaquin Valley.
Type Transfer Function Approach
The type transfer function (TTF) approach to ground water vulnerability assessments (Stewart, 2001; Stewart and Loague, 1999, 2000, 2003) combines some of the advantages of both index and overlay methods and process-based models. The TTF approach involves the upscaling of the lognormal Jury transfer function model (Jury, 1982; Jury and Roth, 1990) to soil-texture-based TTFs for regional-scale simulation of vadose-zone leaching. The TTFs developed by Stewart (2001) and Stewart and Loague (2003) were shown to (i) be capable of quantitatively estimating the spatiotemporal distributions of solute concentrations due to a specific pesticide application, (ii) rely only on available soil survey and climatic and irrigation information, and (iii) significantly minimize the computational effort compared with process-based simulations.
Type transfer functions are upscaled travel time probability density functions that describe characteristic vertical leaching behavior for soil profiles with similar soil-hydraulic properties. Once the TTF(s) are identified, a pesticide leaching simulation using the TTF approach only involves selecting the appropriate TTF(s), convolving the TTF(s) with the input concentration function, and scaling the TTF(s) by expressions for retardation and linear decay.
Stewart and Loague (2003) developed and cataloged seven sets of TTFs, representing different levels of upscaling, for six loam soil-textural classes with the aid of synthetic data. The concentrations predicted by the TTF model for synthetic test cases were compared with those estimated from process-based simulations using the finite-element code SWMS_3D (Simunek et al., 1995). The results from Stewart and Loague (2003) show that the TTF modeling approach performed well compared with physically based simulations in terms of estimating concentration peaks and their arrival times at a compliance surface. Until now, the TTF approach has not been employed for ground water vulnerability assessments at the scale of the San Joaquin Valley.
|
STUDY OBJECTIVE
|
|---|
The overall objective of this study was to demonstrate the use of the TTF approach (Stewart, 2001; Stewart and Loague, 2003) for quantitative, regional-scale pesticide leaching assessments. The vulnerability assessments for atrazine leaching were prepared for a 11610-km2 San Joaquin Valley study area. The study area consists of the eastern portions of eastern Stanislaus and Merced Counties, the western portions of Madera County, the eastern portion of Fresno County, and Kings County (see Fig. 2). The area shown in Fig. 2 is the portion of the area depicted in Fig. 1 that is in agricultural production and likely to experience pesticide applications. For this study, the TTF approach was loosely coupled with a geographic information system (GIS) and focuses on atrazine leaching as a case in point. Specific objectives for this study were to (i) demonstrate that the preparation of a pesticide leaching assessment for an area the size of the San Joaquin Valley is computationally feasible and efficient with the TTF approach, (ii) predict the spatial distribution of atrazine concentrations at a compliance surface (specified for the state of California) resulting from a single label-recommended application, (iii) relate the estimated atrazine concentrations to soil survey information, (iv) identify areas of high potential vulnerability to atrazine leaching, and (v) delineate the spatial distribution of the arrival times of peak concentrations at the compliance surface. Type transfer function model predictions are compared with field data (Domagalski and Dubrovsky, 1992) to the extent possible.
View larger version (48K):
[in this window]
[in a new window]
|
Fig. 2. The areas from the five counties in the San Joaquin Valley that were used to assess potential atrazine leaching. The location of the study area is shown in Fig. 1.
|
|
San Joaquin Valley and Ground Water Vulnerability Assessment
Setting
The composition and structure of the soils in the eastern portion of the San Joaquin Valley differ from that in the western portion of the valley. On the eastern side, the soils are derived from the Sierra Nevada, and consist mainly of highly permeable, medium to coarse-grained sands with low total organic carbon. On the western side, soils are derived from the coast ranges and exhibit a higher clay content and finer texture than those in the east (Page, 1986). In the central portion of the valley, Holocene stream channel deposits of coarse sand, gravel, and silt are found along the San Joaquin River and its major tributaries. At greater distances from the channels, flood basin deposits are often present. Soils that have developed on stream channel deposits tend to have a high clay content and low permeability (Davis et al., 1959). Holocene dune sands, which outcrop in large areas south of eastern Fresno and around the Merced River, are well-sorted (Page, 1986) and generally permeable. There are also several areas in the central portion of the valley with mostly fine-grained, generally impermeable lacustrine and marsh deposits.
The mean annual precipitation in the San Joaquin Valley is highly variable, ranging from approximately 0.15 to 0.50 m, and generally increasing to the north. Potential evapotranspiration in the study area can reach up to 1.50 m a year (Gronberg et al., 1997). Recharge in agricultural areas is therefore mainly comprised of crop irrigation. The permeability of the soils and the amount of recharge are the dominant controls on leaching.
Previous Regional-Scale Pesticide Leaching Studies for the San Joaquin Valley
Due to the intense agricultural use of the San Joaquin Valley, a number of modeling studies have sought to characterize ground water vulnerability in this area. Overlay- and index-based studies comprise the bulk of the prior ground water vulnerability assessment efforts for the San Joaquin Valley. The regression models of Teso et al. (1988)(1996) and the leaching potential approaches of Blanke (1999), Meeks and Dean (1990), and Mills (2004) are examples. None of these models can be applied outside the calibrated range or for different NPS pollutants. Blanke (1999) estimated pesticide leaching with the attenuation factor (AF) and retardation factor (RF) indices for 32 agrochemicals in the eastern San Joaquin Valley with consideration for the uncertainty in both the soil and chemical data. Blanke (1999) used average annual recharge (see Loague et al., 1998) and soil survey information from the five counties (shown in Fig. 1 and 2) to estimate vulnerability. The results from Blanke (1999) indicated that atrazine was unlikely to leach for each of the soil orders in the San Joaquin Valley. Mills (2004) improved the Blanke (1999) AF vulnerability assessments by estimating spatially distributed annual average recharge values based on monthly precipitation data, evapotranspiration estimates, and land use. Mills (2004) reported estimates of atrazine leaching greater than those of Blanke (1999) for the San Joaquin Valley. However, even with consideration for the uncertainty in the soil, chemical, and recharge data, Mills (2004) still classified atrazine as unlikely to leach in the San Joaquin Valley. Loague et al. (1998) simulated transport of DBCP (1,2-dibromo-3-chloropropane) in the unsaturated-zone with PRZM-2 in east-central Fresno County by conducting 1172 one-dimensional simulations for 1-km2 square elements for the period of 1960–1994. Although the resulting simulated concentration dataset is unique in extent and detail, the data and computational requirements for this study were tremendous. It should be noted that none of the past studies produced quantitative concentration estimates at the scale of the San Joaquin Valley.
|
METHODS
|
|---|
Assessing potential atrazine leaching in the San Joaquin Valley with the TTF approach involved the selection of the particular TTF model and sets (Stewart and Loague, 2003), assembly of the database, numerical estimation of contaminant concentrations with the selected TTF sets, and the visualization and interpretation of modeling results.
Type Transfer Function Approach
The Jury transfer function model (Jury TFM) is a linear systems model that describes the distribution of travel times from the surface (Z = 0) to a depth of interest (Z = L) with an impulse response or transfer function. Here Z [L] is the spatial coordinate in the vertical direction and L [L] is the depth of interest, for which the ground water vulnerability assessment is prepared. In the context of vadose zone leaching, a lognormal transfer function (which is fully determined by the mean and standard deviation) has yielded good results (Jury, 1982; Jury and Roth, 1990; Jury et al., 1986). For reactive transport, terms for solute sorption and decay can be incorporated into the model description. For steady-state flow conditions with a flow rate s0 [L T–1], the area-averaged solute flux concentration C(L, t) [M L–3] at L and time t [T] can be expressed in terms of the recharge below the root zone I, where I = s0t [L]. The concentration estimate C(L, I) [L–3], resulting at depth L after infiltration of recharge I, is then given by the convolution integral of the solute input at the surface C(0, I') [M L–3], where I' refers to the recharge at the time of solute input, with an impulse response function at calibration depth L. This development (Jury, 1982; Jury and Roth, 1990; Jury et al., 1986) results in:
 | [1] |
where f(L, I – I') is the impulse response or travel-time probability density function (PDF) [T–1], describing the distribution of solute travel times conditional on I' for steady-state conditions and a travel distance of L.
For regional-scale pesticide leaching assessments, Stewart and Loague (2003) defined upscaled lognormal TTFs, with effective parameters µE and 
E that were tied to six agricultural soil textures (sandy loam, loam, silt loam, sandy clay loam, clay loam, and silty clay loam). The PDF for reactive transport using the TTF approach fER (L, I – I') is given (Stewart and Loague, 2003) by:
 | [2] |
where fR is the PDF for the reactive solute, R is the retardation coefficient [unitless], and k is the first-order decay rate [T–1]. The advantage of the Jury TFM lies in the fact that once the system behavior is defined by calibrating f(L, I – I'), concentrations at L may be calculated by evaluating a simple convolution integral.
The effective TTF parameters in Eq. [2] were calibrated using simulated data sets that were generated based on published soils information (Blanke, 1999; Carsel and Parrish, 1988; Mualem, 1976) and numerical simulations of unsaturated flow and atrazine transport (Stewart and Loague, 2003) using SWMS_3D (Simunek et al., 1995). Several sets of TTF corresponding to different levels of upscaling were developed by Stewart and Loague (2003). Thus, depending on the level of upscaling and prediction accuracy desired, one of several possible sets of TTFs may be selected for a ground water vulnerability study. The individual steps relevant for TTF development and application are discussed by Stewart and Loague (2003). In the study reported here, only Sets III and IV from Stewart and Loague (2003) are employed (one and two TTFs per soil texture, respectively) for assessing potential atrazine leaching in the San Joaquin Valley. Selection of the TTF sets was based on the need for minimizing computational effort at the scale of the San Joaquin Valley.
Geographic Information System Framework
The spatial resolution used in a pesticide leaching study determines the size of the databases associated with the assessment. The simplicity of the TTF model limits the soil, recharge, pesticide application, and chemical property information required for each mapped unit (polygon). However, the large number of polygons for the study area resulted in large databases with substantial computational requirements. The database compiled for this study extended the extensive soil-series-level digitized information (with 16469 polygons) that formed the basis for the soil-order database prepared by Blanke (1999) to a soil-texture database used in this study with 55410 polygons. The generation of the soil-textural database involved the determination of soil texture for each polygon at the series level, and the assignment of soil properties and recharge values (see Stewart [2001] for details)
The spatially uniform recharge value does not include an irrigation estimate, and is therefore very conservative. For eastern Fresno County, an additional separate database was generated, where the grid (with 65535 polygons) for spatially distributed recharge, estimated by Mills (2004), was overlaid onto the 18847 polygons defined by soil texture. The union of these two grids resulted in a new grid with greater resolution. Each of the 168695 polygons in the new grid for eastern Fresno County differs from its neighbor by either the soil textural and property information and/or the average annual recharge value. The spatially distributed recharge estimates of Mills (2004) were based on a water-balance approach, where recharge is the residual of monthly precipitation, irrigation (conservatively estimated as the precipitation shortfall relative to crop cover), and evapotranspiration estimates. The recharge estimates used for TTF application with spatially distributed recharge estimates are on the order of 0.1 x qcalibration, where qcalibration is the recharge rate at calibration.
Estimation of Concentrations
The soil textural information that was added to the soil database was mapped for each of the five counties in the San Joaquin Valley (see Fig. 3a). Soil information (e.g., soil bulk density, soil water content at field capacity, and fraction of soil organic carbon) exported from the geographic information system database was used in conjunction with TTF parameter values, recharge rates, and chemical information for TTF concentration estimates. All concentration estimates were calculated for a compliance depth d [L] of Z = 3 m, which is equal to the TTF calibration depth L [L] used by Stewart (2001) and Stewart and Loague (2003). The depth d was selected for purposes of prediction accuracy, but differs from the d = 1 m used by Blanke (1999) and Mills (2004). All concentration estimates assumed a single, label-recommended surface application of 2.24 kg ha–1 for atrazine, a value of 0.1 m3 kg–1 for the organic carbon–soil partition coefficient Koc [L3 m–1], and 60 d for the pesticide half-life t1/2 [T] (Hornsby et al., 1996). The TTF parameters µE and E for TTF Sets III and IV for all simulations were obtained from Stewart (2001) and are listed in Table 1. Type transfer function Set III contains one TTF per soil texture and therefore one associated set of TTF parameters. Set IV contains two TTFs per soil texture, and therefore two associated sets of TTF parameters, one for a lower estimate, and one for a higher estimate. Following Stewart and Loague (2003), each set applies to 50% of the area with a given soil texture. For this study, however, the lower and the higher TTF concentration estimate for Set IV are each displayed separately as Set IVa and IVb due to the large number of polygons already considered.
View larger version (51K):
[in this window]
[in a new window]
|
Fig. 3. (a) Soil texture in the San Joaquin Valley study area. (b) Type transfer function (TTF)-simulated peak atrazine concentrations for a compliance depth d = 3 m in the San Joaquin Valley study area. (c) TTF-simulated arrival time of the peak concentration at depth d = 3 m in the San Joaquin Valley study area.
|
|
Two groups of TTF simulations for the San Joaquin Valley were performed in this study. Group I used a spatially uniform daily recharge rate of s0 = 8.7 x 10–4 m d–1 and assessed atrazine leaching for all five counties in the San Joaquin Valley study area. The recharge rate was estimated by dividing the annual recharge of Iyr = 3.175 x 10–1 m used by Blanke (1999) into equal daily values. Group II employed the spatially distributed recharge information reported by Mills (2004) for the San Joaquin Valley. This part of the study focused on eastern Fresno County only, as the union of the soil textural and recharge information led to a tremendous expansion in the number of polygons and in data storage and manipulation requirements.
The soil, chemical, and recharge information was used with Eq. [1] and [2] to estimate atrazine concentrations for both groups of TTF simulations at d over a period of two years. There were three TTF concentration estimates for each group, one for TTF Set III, and two for Set IV for all 730 d of the simulation period. For Set III the peak concentration reaching d during the simulation period and the simulation time at which this peak concentration occurs were calculated. It should be pointed out that Stewart and Loague (2003) only determined TTF parameters for the six loamy soil textures that would most likely be found in agricultural settings. The TTF model could therefore not be applied in this study to the comparatively limited area where one of the remaining five soil textures (sand, loamy sand, sandy clay, silty clay, and clay) is present. The information contained in the San Joaquin Valley database for this study is summarized in Table 2.
|
RESULTS AND DISCUSSION
|
|---|
The results from the TTF simulations of atrazine leaching in the San Joaquin Valley are visualized in maps of expected concentrations at the compliance depth L. In the following discussion the term "susceptible areas" will be used for areas where TTF model estimates of atrazine concentrations lie above the maximum contaminant level (MCL) of 3 x 10–3 mg L–1. It should be emphasized again that the simulated concentrations at a depth of 3 m result from a single, surface application of atrazine. Agricultural practices, however, often entail annual or more frequent applications.
Atrazine Concentrations for the San Joaquin Valley, Spatially Uniform Recharge
For all five counties the loamy textures cover a large portion of the total area, ranging from approximately 50% for Kings County to approximately 90% for Madera County (Fig. 3a). Clayey and sandy soils are mainly restricted to isolated pockets of marsh and fluvial deposits, respectively. Notable exceptions are the Tulare lakebed in Kings County and more extensive sandy and clayey soils in Merced County. It should be noted that eastern Fresno County has the highest percentage of sandy loams, which are more susceptible to leaching than the other soil textures considered in this study.
Figures 3b and 3c depict the TTF model simulation results for spatially uniform recharge for the five counties in the San Joaquin Valley study area. Figure 3b shows the maximum concentration estimated with TTF Set III to arrive at d at any time during the simulation period, with concentrations above the MCL highlighted in dark red. The predicted arrival times of the maximum concentrations shown in Fig. 3b are illustrated in Fig. 3c.
Comparing Fig. 3a and 3b illustrates that in 100% of the area with loamy soil textures in Kings County, and in approximately 50% of the areas with loamy soil textures in all other counties, the estimated peak atrazine concentrations do not surpass 1 x 10–4 mg L–1. Most of the remaining peak concentrations lie below 1 x 10–3 mg L–1. Only eastern Stanislaus County and eastern Fresno County contain some areas (<5%) where peak concentrations reach values above 1 x 10–3 mg L–1, and only predictions for eastern Fresno County exceed the MCL. For eastern Fresno County, the maximum simulated concentration is 7.3 x 10–3 mg L–1. As expected, all high peak concentrations are found in areas with sandy loam and loam soils. However, not all sandy loam and loam soils showed high simulated concentrations. It is important to point out that the extensive sand and loamy sand soils in Merced County could be expected to be highly susceptible, but as this study was restricted to the loamy textures, for which the TTFs were developed, no estimates can be made.
The distribution of peak arrival times shown in Fig. 3c is limited to the 350- to 730-d range for all areas with TTF estimates. For the susceptible areas, the arrival time of the concentration peak at d ranges between 304 and 361 d with a median of 351 d. Therefore, peak atrazine concentrations for the loamy soil textures would be expected at the end of the first year after a single pesticide application.
Figure 4 compares TTF estimates for TTF Sets III and IV for eastern Fresno county. Results for the other four counties in the San Joaquin Valley are reported by Stewart (2001). While daily TTF concentrations were simulated, only the results for Days 200 and 700 are shown Fig. 4. Figure 4 shows the results for the simulations using TTF Sets III, IVa, and IVb. It is clear from Fig. 4 that there are some consistent characteristics for the three TTF predictions. For example, the estimates for Set IVa are generally lower than, and those for Set IVb are generally higher than the estimates for Set III. Arrivals of concentration peaks above the MCL in areas of eastern Fresno County can be observed 200 d after pesticide application. These early arrivals correspond to the fast travel times through the sandy loam and loam soils. It is also clear that none of the estimates made for Day 700 exceed the MCL for any area, as the concentration peaks have moved beyond d. The concentration map for Set IVb in eastern Fresno County shows concentrations that exceed 1 x 10–3 mg L–1, which is on the same order of magnitude as the MCL. Figure 4 illustrates that the estimates are influenced by the number of TTFs used per texture.
View larger version (61K):
[in this window]
[in a new window]
|
Fig. 4. Type transfer function (TTF) concentration estimates at d for Day 200 (upper panels) and Day 700 (lower panels) of the simulation period in response to a single, label-recommended application of atrazine for eastern Fresno County. All estimates were made under consideration of spatially uniform annual recharge and for a compliance depth of d = 3 m. (a) TTF Set III (one TTF for each texture). (b) TTF Set IVa (two TTFs for each texture, lower estimate). (c) TTF Set IVb (two TTFs for each texture, higher estimate). (I) Simulation Day 200. (II) Simulation Day 700.
|
|
The relationship between the occurrence of peak concentrations and the soil organic carbon content and soil texture present for a given polygon was also examined. Figure 5 illustrates the distribution of foc in the surface soil layer throughout the San Joaquin Valley study area. The values for foc range between 1 and 4%. For most of the study area foc lies below 1.5%, and foc is particularly low in most of Fresno County (<0.75%). Figure 6 correlates all 133 polygons in the study area with predicted peak concentrations above 1 x 10–3 mg L–1 with surface organic carbon content in the soil profile and soil texture. It is clear from Fig. 6 that peak concentrations occur where low foc in conjunction with sandy loam or loam textures exist.
View larger version (19K):
[in this window]
[in a new window]
|
Fig. 6. Correlation of (a) peak concentrations above a threshold of 1 x 10–3 mg L–1 with (b) surface organic carbon and (c) soil texture. In (c), 1 = sand, 2 = loamy sand, 3 = sandy loam, 4 = loam, 5 = silt, 6 = silt loam, 7 = sandy clay loam, 8 = clay loam, 9 = silty clay loam, 10 = sandy clay, 11 = silty clay, and 12 = clay.
|
|
It is possible that the absence of more extensive susceptible areas and higher peak concentrations throughout the San Joaquin Valley study area may result from using an upscaled set of TTFs. Loague and Corwin (1996) discuss uncertainty (resulting from model, input, and/or parameter errors) related to regional-scale assessments of nonpoint-source pollutants. Obviously, any assessment of regional-scale ground water vulnerability will have some uncertainty. Stewart and Loague (2000), investigating the model error associated with upscaled TTFs, found that the uncertainty in estimated concentrations was significantly reduced when sets containing two (as opposed to one) TTFs were employed (note, adding more TTFs did not further reduce the uncertainty). For future regional-scale applications of the TTF model (i.e., beyond the first demonstration reported here) it will be important to quantify the uncertainty associated with both input and parameter errors, especially if those assessments are to be used in a regulatory or decision-making context. Further characterization of uncertainty inherent to regional-scale applications of the TTF model will require significantly more information than was available for this study.
Atrazine Concentrations for Eastern Fresno County, Spatially Distributed Recharge
Figures 7 and 8 illustrate TTF estimates for eastern Fresno County with spatially variable recharge. Figure 7a shows the average annual recharge values for each polygon. The recharge map in Fig. 7a illustrates that high annual average recharge values can be expected for large areas of the western, eastern, and southern portions of eastern Fresno County, and that great differences in estimated annual recharge values can exist between adjacent polygons.
View larger version (55K):
[in this window]
[in a new window]
|
Fig. 7. Results from the type transfer function (TTF) pesticide leaching assessment for eastern Fresno County at a compliance depth of d = 3 m using spatially distributed recharge. (a) The spatially distributed annual average recharge for the Fresno study area. (b) Peak atrazine concentrations. (c) Arrival time of the peak concentrations at d.
|
|
View larger version (66K):
[in this window]
[in a new window]
|
Fig. 8. Type transfer function (TTF) concentration estimates for Day 200 (a–c) and Day 700 (d–f) of the simulation period in response to a single, label-recommended application of atrazine for eastern Fresno County. All estimates were made under consideration of spatially distributed annual recharge and for a compliance depth of d = 3 m. (a) TTF Set III (one TTF for each texture). (b) TTF Set IVa (two TTFs for each texture, lower estimate). (c) TTF Set IVb (two TTFs for each texture, higher estimate). (d) TTF Set III (one TTF for each texture). (e) TTF Set IVa (two TTFs for each texture, lower estimate). (f) TTF Set IVb (two TTFs for each texture, higher estimate).
|
|
Figure 7b and 7c illustrate the peak atrazine concentrations and the arrival time of these concentrations at a depth of 3 m. It is clear from Fig. 7 that atrazine concentrations above the MCL can be expected for areas in the western, eastern, and southern portions of the eastern Fresno County study area. In addition, extensive areas in the eastern portion of the eastern Fresno County study area show estimated atrazine concentrations above 1 x 10–3 mg L–1. The results show that all peak arrival times are greater than 200 d, with shorter arrival times for susceptible areas.
Figure 8 compares the predicted concentrations simulated with the TTF sets used in this study for Days 200 and 700. By Day 200 of the simulation, concentrations above the MCL are predicted in eastern Fresno County for all TTF estimates. In addition, extensive susceptible areas are identified for western and southern portions of eastern Fresno County for TTF Set IVa. By Day 700 of the simulation, estimated atrazine concentrations for all TTF sets fall below 1 x 10–3 mg L–1 in the entire study area, although approximately 60% of the study area still exhibits estimated concentrations above 1 x 10–4 mg L–1 for TTF Set IVb. A direct comparison of susceptible areas for spatially uniform and spatially distributed recharge is given in Fig. 9. Clearly, the extent of susceptible areas is less extensive when only the uniform annual recharge values are used (Fig. 9a). For spatially variable recharge TTF simulation results (Fig. 9b), extensive areas with estimated concentrations exceeding the MCL are identified for western, southern, and eastern portions of Fresno County.
View larger version (45K):
[in this window]
[in a new window]
|
Fig. 9. Comparison of areas with type transfer function (TTF) predictions exceeding the maximum contaminant level (MCL) for TTF Set III in eastern Fresno County. (a) Spatially uniform recharge values. (b) Spatially distributed recharge values.
|
|
Comparison with Field Measurements
Domagalski and Dubrovsky (1992) assessed the distribution of NPS pesticide residues in the San Joaquin Valley from measurements in wells. The historical and regional studies showed that atrazine was detected in some of the same general locations as simazine, mainly eastern Fresno and eastern Tulare counties. Concentrations ranged from 0.1 to 0.4 x 10–3 mg L–1, with the water table depths in the sampled wells at 16.8 m and greater. At a site in eastern Fresno County, where dune sand is present, atrazine was detected at concentrations of 0.05 to 0.2 x 10–3 mg L–1, 13 to 61 m below the surface. Atrazine was detected in the western and southern parts of the valley at only a few sites. Measured concentrations ranged from 1 to 80 x 10–3 mg L–1, with detection limits of 0.05 to 0.1 x 10–3 mg L–1.
The TTF model predicts peak atrazine concentrations at a detectable level at d = 3 m for large portions of the study area (Fig. 3). For this study, it was assumed that atrazine, as a case in point, was applied throughout the area at an agricultural use rate. As a result, the highest TTF estimated concentrations of 1 to 3 x 10–3 mg L–1 (and larger) were predicted in western San Joaquin, and western, eastern, and southern Fresno County, where sandy loams are present. These findings are not in agreement with the observations reported by Domagalski and Dubrovsky (1992) for two reasons: (i) atrazine was not used on crops during the time of the field measurements, and (ii) the samples were from wells with depths much greater than 3 m, rendering a direct comparison impossible.
The effect of soil texture and organic carbon content on leaching susceptibility, discussed by Domagalski and Dubrovsky (1992), can clearly be observed in the results of this study (e.g., Fig. 3). In addition, when the spatially distributed recharge values were taken into consideration in the TTF calculations, a peak concentration pattern more similar to that seen in the observed data emerges (Fig. 7). For both studies, the areas of highest atrazine concentrations were found in eastern Fresno County. This suggests that using spatially distributed recharge is an important component to obtaining a realistic ground water vulnerability assessment.
Model Limitations
The regional-scale vulnerability assessments for the San Joaquin Valley represent a unique tool, at a scale significantly larger than that of previous studies, which can play an important role in the identification of susceptible areas through quantitative leaching estimates. The limitations of the TTF model are summarized below.
The TTF model is an upscaled approach estimating average leaching behavior and not concentration peaks that might exist locally. It is reasonable to suspect that the most likely source of agrochemical contamination in shallow water table aquifers is preferential flow of pesticides through local macropores. These processes are not captured by the TTF model as employed in this study. The soil-texture-based TTFs used here were calibrated for a given depth and constant recharge. The recharge estimates used to assess ground water vulnerability with TTFs were lower than those used for calibration. The estimated (and idealized) soil property, recharge, and water table depth values for the San Joaquin study area can be quite different than the conditions at TTF calibration. Neither sand nor sandy loam soils, which would be expected to comprise the most susceptible areas for chemical leaching, were included in this study. The focus of this study was on the loamy textures, for which TTFs had been developed, and which cover 50 to 90% of the study area. Finally, the performance of the TTF approach must be evaluated against measured concentration data sets from the unsaturated zone. Despite the limitations of the TTF model as it is demonstrated in this study, the authors are unaware of another approach that is currently capable of estimating agrochemical concentrations for areas the size of the San Joaquin Valley with comparable speed and efficiency.
|
CONCLUSIONS
|
|---|
The effort reported here is the first application of the TTF modeling approach to quantitative, regional-scale (San Joaquin Valley) pesticide leaching assessments. Type transfer functions, as developed by Stewart (2001) and Stewart and Loague (2003), are characteristic functions that (i) are capable of quantitatively estimating the spatiotemporal distributions of solute concentrations due to a specific pesticide application, (ii) rely only on available soil survey and climatic and irrigation information, and (iii) significantly minimize the computational effort compared with process-based simulations. The TTF simulation approach was used to estimate (and map through time) the spatial distribution of atrazine concentrations at a compliance surface resulting from a label-recommended surface application.
Estimated atrazine concentrations could be related to soil survey information, with early arrivals corresponding to sandy loam and loam soils. Areas with high potential vulnerability to atrazine leaching were found where sandy loam or loam soil textures with low foc exist. Type transfer function simulation results were also used to estimate and map the spatial distribution of the arrival times of peak concentrations at the compliance surface. Fast travel times through sandy loam and loam soils result in early arrivals of concentration peaks at the compliance depth about 200 d after pesticide application. By Day 700 of the simulation, elevated atrazine concentrations have generally moved beyond the compliance depth. Comparison of the results corresponding to different sets of TTFs illustrate that TTF prediction estimates will depend on the number of TTFs used per texture. It was found that the characterization of recharge greatly influences the pesticide leaching estimates, and that the extent of the susceptible areas was underestimated when only uniform annual recharge values were used. For spatially uniform recharge values, all peak concentrations (with the exception of <5% of the areas in eastern Stanislaus and eastern Fresno Counties) lie below the MCL. For the spatially distributed recharge estimates used in leaching assessments for eastern Fresno County, the estimated peak concentrations above the MCL cover a larger total area (particularly in the eastern portion). Although field measurements on comparable temporal and spatial scales do not exist, comparison with available field observations show that the pattern of peak concentrations predicted by the TTFs is similar to that seen in the observed data for spatially distributed recharge rates.
The work presented here illustrates that the TTF approach is promising and that a comprehensive catalog of TTFs that includes soil textures, recharge rates, and chemical characteristics not included in this study would be a worthwhile endeavor. With the ability to consider a broader range of conditions, the TTF approach can become an efficient and useful tool for those in the decision-management arena faced with assessing ground water vulnerability and designing field-sampling campaigns at the regional scale.
|
ACKNOWLEDGMENTS
|
|---|
The first author gratefully acknowledges financial support for this work from a USEPA STAR fellowship and a Theresa Heinz Fellowship for Environmental Research. We are also grateful to Jim Blanke and Melissa Mills for making their data available. The computing facilities used in this study were funded through an NSF equipment grant. The work reported here is a Center for Earth Science Information Research (CESIR) contribution.
|
REFERENCES
|
|---|
- Barbash, J.E., G.P. Thelin, D.W. Kolpin, and R.J. Gilliom. 1999. Distribution of major herbicides in ground water of the United States. Water-Resour. Investigations Rep. 98-4245. USGS, Reston, VA.
- Blanke, J.S. 1999. Vulnerability to groundwater contamination due to agrochemical use in the San Joaquin Valley, California. M.S. thesis. Stanford Univ., Palo Alto, CA.
- Burow, K.R., S.V. Stork, and N.M. Dubrovsky. 1998. Nitrate and pesticides in ground water in the eastern San Joaquin Valley, California: Occurrence and trends. Water-Resour. Investigations Rep. 98-4040. USGS, Reston, VA.
- Carsel, R.F., and R.S. Parrish. 1988. Developing joint probability distributions of soil-water retention characteristics. Water Resour. Res. 24:755–768.
- Cheng, H.H. (ed.) 1990. Pesticides in the soil environment: Processes, impact, and modeling. SSSA, Madison, WI.
- Corwin, D.L., K. Loague, and T.R. Ellsworth. 1999. Introduction: Assessing non-point source pollution in the vadose zone with advanced information technologies. p. 1–20. In D.L. Corwin, K. Loague, and T.R. Ellsworth (ed.) Assessment of non-point source pollution in the vadose zone. Geophys. Monogr. 108. Am. Geophys. Union, Washington, DC.
- Davis, G.H., J.H. Green, F.H. Olmstead, and D.W. Brown. 1959. Ground-water conditions and storage capacity in the San Joaquin Valley, California. Water-Supply Paper 1469. USGS, Reston, VA.
- Diaz-Diaz, R., and K. Loague. 2001. Assessing potential for pesticide leaching for pine forest areas of Tenerife. Environ. Toxicol. Chem. 20:1958–1967.[ISI][Medline]
- Domagalski, J.L., and N. Dubrovsky. 1992. Pesticide residues in ground water of the San Joaquin Valley, California. J. Hydrol. (Amsterdam) 130:299–338.
- Gronberg, J.M., N.M. Dubrovsky, C.R. Kratzer, J.L. Domagalski, L.R. Brown, and K.R. Burow. 1997. Environmental setting of the San Joaquin-Tulare Basins, California. Water-Resour. Investigations Rep. 97-4205. USGS, Reston, VA.
- Hornsby, A.G., R.D. Wauchope, and A.E. Herner. 1996. Pesticide properties in the environment. Springer, New York.
- Howard, P.H. (ed.) 1989. Handbook of environmental fate and exposure data for organic chemicals, pesticides. Lewis Publ., Chelsea, MI.
- Hutson, D.H., and T.R. Roberts. 1990. Environmental fate of pesticides. John Wiley & Sons, New York.
- Jury, W.A. 1982. Simulation of solute transport using a transfer function model. Water Resour. Res. 18:363–368.
- Jury, W.A., and K. Roth. 1990. Transfer functions and solute movement through soil. Birkhaeuser Verlag, Basel, Switzerland.
- Jury, W.A., G. Sposito, and R.E. White. 1986. A transfer function model of solute transport through soil, 1. Fundamental concepts. Water Resour. Res. 22:243–247.
- Khan, M.A., and T. Liang. 1989. Mapping pesticide contamination potential. Environ. Manage. 13:233–242.
- Loague, K. 1994. Regional scale ground water vulnerability estimates: Impact of reducing data uncertainties for assessments in Hawaii. Ground Water 32:605–616.
- Loague, K., and D.L. Corwin. 1996. Uncertainty in regional-scale assessments of non-point source pollutants. p. 131–152. In D.L. Corwin and K. Loague (ed.) Applications of GIS to the modeling of non-point source pollutants in the vadose zone. Spec. Publ. 48. SSSA, Madison, WI.
- Loague, K., R.E. Green, T.W. Giambelluca, T.C. Liang, and R.S. Yost. 1990. Impact of uncertainty in soil, climatic, and chemical information in a pesticide leaching assessment. J. Contam. Hydrol. 5:171–194.
- Loague, K., D. Lloyd, A. Nguyen, S. Davis, and R. Abrams. 1998. A case study simulation of DBCP groundwater contamination in Fresno County, California. 1. Leaching through the unsaturated subsurface. J. Contam. Hydrol. 29:109–136.
- Loll, P., and P. Moldrup. 2000. Stochastic analysis of field-scale pesticide leaching risk as influenced by spatial variability in physical and chemical parameters. Water Resour. Res. 36:959–970.
- Meeks, Y.J., and J.D. Dean. 1990. Evaluating ground-water vulnerability to pesticides. J. Water Resour. Planning Manage. 116:693–707.
- Mills, M.B. 2004. Groundwater vulnerability assessments for the San Joaquin Valley, California. M.S. thesis. Stanford Univ., Palo Alto, CA.
- Mualem, Y. 1976. A catalogue of the properties of unsaturated soils. Res. Rep. 442. Technion, Israel Inst. of Technol., Haifa, Israel.
- Nofziger, D.L., J.S. Chen, and A.G. Hornsby. 1996. Uncertainty in pesticide leaching risk due to soil variability. Soil Sci. Soc. Am. J. 47:99–129.
- Page, R.W. 1986. Geology of the fresh ground-water basin of the Central Valley, California, with textural maps and sections. Professional Paper 1401-C. USGS, Reston, VA.
- Petach, M.C., R.J. Wagenet, and S.D. DeGloria. 1991. Regional water flow and pesticide leaching using simulations with spatially distributed data. Geoderma 48:245–269.
- Simunek, K., K. Huang, and M.Th. van Genuchten. 1995. The SWMS_3D code for simulating water flow and solute transport in three-dimensional variably-saturated media. Rep. 139. USDA Salinity Lab., Riverside, CA.
- Stewart, I.T. 2001. Development of a type transfer function approach for modeling non-point source vadose-zone pesticide leaching at the regional scale. Ph.D. diss. Stanford Univ., Palo Alto, CA.
- Stewart, I.T., and K. Loague. 1999. Identification of "type" transfer functions for regional-scale assessment of non-point source groundwater contamination in Fresno County, California. J. Environ. Qual. 28:378–387.[Abstract/Free Full Text]
- Stewart, I.T., and K. Loague. 2000. Uncertainty introduced by upscaling type transfer functions. p. 617–624. In G.B.M. Heuvelink and M.J.P.M. Lemmens (ed.) Proc. of Accuracy 2000. Delft Univ. Press, the Netherlands.
- Stewart, I.T., and K. Loague. 2003. Development of type transfer functions for regional-scale non-point source groundwater vulnerability assessments. Water Resour. Res. 39(12):1359.
- Teso, R.R., M.P. Poe, T. Younglove, and P.M. McCool. 1996. Use of logistic regression and GIS modeling to predict groundwater vulnerability to pesticides. J. Environ. Qual. 25:425–432.[Abstract/Free Full Text]
- Teso, R.R., T. Younglove, M.R. Peterson, D.L. Sheeks, and R.E. Gallavan. 1988. Soil taxonomy and surveys: Classification of areal sensitivity to pesticide contamination of groundwater. J. Soil Water Conserv. 43:348–352.
- USEPA. 1994. National water quality inventory. Tech. Rep. EPA 841-R-94-001. USEPA, Washington, DC.
- Wauchope, R.D., T.M. Buttler, A.G. Hornsby, P.W.M. Augustijn-Beckers, and J.P. Burt. 1992. SCS/ARS/CES Pesticide properties database for environmental decision making. Rev. Environ. Contam. Toxicol. 123:1–157.[ISI][Medline]
Related articles in JEQ:
- This Issue in Journal of Environmental Quality
JEQ 2004 33: 1177-1182.
[Full Text]
This article has been cited by other articles:
|
|
|
|
 
A. Tiktak, J. J. T. I. Boesten, A. M. A. van der Linden, and M. Vanclooster
Mapping Ground Water Vulnerability to Pesticide Leaching with a Process-Based Metamodel of EuroPEARL.
J. Environ. Qual.,
July 1, 2006;
35(4):
1213 - 1226.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
