VZJ sign up for etocs
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 8 October 2007
Published in Vadose Zone J 6:735-745 (2007)
DOI: 10.2136/vzj2006.0139
© 2007 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Alavi, G.
Right arrow Articles by Ray, C.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Alavi, G.
Right arrow Articles by Ray, C.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Alavi, G.
Right arrow Articles by Ray, C.
Related Collections
Right arrow Agricultural Pesticides
Right arrow Flow
Right arrow Fate

ORIGINAL RESEARCH

Evaluation of Dual-Permeability Models for Chemical Leaching Assessment to Assist Pesticide Regulation in Hawaii

G. Alavia, J. Dusekb, T. Vogelb, R. E. Greenc and C. Raya,*

a Dep. of Civil & Environmental Engineering and Water Resources Research Center, Univ. of Hawaii at Manoa, Honolulu, HI 96822
b Faculty of Civil Engineering, Czech Technical Univ., Prague, Czech Republic
c Natural Resources and Environmental Management, Univ. of Hawaii at Manoa, Honolulu, HI 96822
* Corresponding author (cray{at}hawaii.edu).

All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.


Received 21 September 2006.



   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 Materials and Methods
 Results and Discussion
 Conclusions
 REFERENCES
 
Groundwater is the primary source of drinking water for all the islands of Hawaii. Past agricultural practices have led to the contamination of groundwater in certain locations. As a result, the state of Hawaii emphasizes the prevention of contamination of groundwater from the leaching of pesticides. Hawaii currently uses a simple (Tier I) screening assessment model to evaluate the leaching potential of pesticides. This model is only capable of indicating if a chemical is likely to leach; it can estimate neither the concentration profile in soil nor the concentration in leachate water. The USEPA is seeking partnership with the state of Hawaii for examining the feasibility of using Tier II models in Hawaii conditions for pesticide registration. Two pesticide leaching models, MACRO 4.3 and S1D DUAL, were tested using leaching data for five pesticides from a field site on the island of Oahu. Despite deficiencies, it is one of the best data sets currently available for tropical soils. Both MACRO 4.3 and S1D DUAL models explicitly include preferential flow components but use different concepts in model formulations. The performances of the two models were generally similar. The results show that preferential flow had a minor role in transporting the chemicals compared with micropore flow because of the high saturated conductivity of micropores (matrix). We conclude that a process-based model will contribute substantially to the evaluation of chemical leaching risk and complement the Tier I model that currently is used for pesticide registration in Hawaii.

Abbreviations: CDE, convection–dispersion equations • COM, center of mass • DBCP, dibromochloropropane • PRZM, Pesticide Root Zone Model • RMSE, root mean square error


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
Materials and Methods
 Results and Discussion
 Conclusions
 REFERENCES
 
The demand for agricultural productivity has caused a dramatic increase in the use of synthetic organic pesticides. During recent decades, many concerns have been raised about their potential adverse effects on the environment and human health. The greatest potential for unintended adverse effects of pesticides is through the contamination of hydrologic systems, as water is one of the primary transport media for pesticides (Barbash and Resek, 1996). Contamination of groundwater by pesticides is a major national issue in many countries because of the common use of groundwater for drinking purposes (Ray, 2003). Small tropical islands are particularly vulnerable to environmental damage because of small land area, higher population density, and limited resources available. Protection of groundwater from contamination by pesticides, as well as other chemicals, is therefore of great importance. Groundwater resources in Hawaii are invaluable. For the island of Oahu, where more than 75% of the state's population resides, greater than 95% of all water for domestic and agricultural uses comes from groundwater, and nearly all drinking water is obtained from groundwater.

Past agricultural practices have led to contamination of groundwater with chemicals in Hawaii. The Honolulu Board of Water Supply on Oahu has already spent more than $50 million in capital expenditures building activated carbon adsorption systems to remove these contaminants from the pumped water.

More recently, the state of Hawaii investigated the occurrence of some common herbicides in groundwater on the islands of Hawaii, Kauai, Maui, and Oahu (Li et al., 2001). The number of wells sampled on each of the four islands was 5, 5, 14, and 12, respectively. Another 13 samples were collected from some of these wells 3 mo after the first sampling. In all, 49 samples were available for chemical analysis. Low concentrations of hexazinone were found in two wells on Maui, and another well showed the presence of both bromacil and hexazinone. On Oahu, bromacil was detected in two wells, and one showed the presence of hexazinone. On Hawaii Island, two wells had low concentrations of hexazinone. The wells that showed hexazinone were in current or former sugarcane growing areas. Zhu and Li (2002) also analyzed soil samples in several pineapple fields on the island of Oahu. They found bromacil in soil from all six fields in the top 0.6 m depth. Nearly 75% of the samples had residual bromacil in the top 4.0 m. At some locations, trace amounts of bromacil were found in samples collected from depths of 15 m. Hexazinone was detected in samples from three fields. Hexazinone has not been used in these fields since the closure of sugarcane plantations on Oahu in the early 1990s.

The tropical soils of Hawaii exhibit structural heterogeneity in addition to macroporosity caused by cracks, flora, and fauna. The highly weathered soils of central Oahu have traditionally been used for agriculture. Although the soils are rich in kaolinitic clay (60–80%), they are highly aggregated, and the water stability index of these aggregates is excellent. Thus, one can speculate that preferential flow may occur during episodic storm events. Preferential flow refers to rapid, nonequilibrium fluxes of water flow through soil macropores such as interaggregate pore space, earthworm channels, root holes, and shrinkage cracks. More than 30 yr ago, Green et al. (1972) attributed deep penetration of a herbicide front in such soils to differential pore-water velocity. Under intermittent flooding conditions in the same soil, a leaching experiment with picloram (Rao et al., 1974) showed that a significant amount of the applied herbicide moved deep into the profile (1.43 m depth, much ahead of the peak location) after an application of 240 mm of water. A study was undertaken by Gavenda et al. (1996) to evaluate the leaching of five pesticides and bromide in Hawaii (see "Materials and Methods" below). The results showed extreme variability in leaching behavior of different chemicals, depending on pesticide properties and soil characteristics.

Models of pesticide movement in the soil are potentially useful tools for predicting the likely environmental fate of field applications of pesticides. Models have been used as tools for assisting pesticide regulations at state and federal levels. For example, Hawaii has been using a simple screening assessment (Tier I) model to evaluate the leaching potential of pesticides (Li et al., 1998; Stenemo et al., 2007). In this screening model, the leachability of a compound is compared against two chemicals in the database, one a known leacher and one a known nonleacher under Hawaii conditions. If the chemical shows negligible leaching potential, it is generally registered in the state without any restrictions. Compounds with high leaching potential are either registered as "restricted use" chemicals or completely banned. Under the restricted use permit, the state monitors the locations and the amounts of the chemical applied to the soil. In addition, certain management options are expected to be prescribed to reduce leaching. Because of its past experience, Hawaii is interested in preventing contamination of groundwater rather than dealing with postcontamination treatment issues. Unfortunately, the above screening model is only capable of indicating if a chemical is likely to leach, and it cannot estimate the concentrations in the soil profile or leachate water.

In the past few decades, a number of pesticide leaching models have been developed using the convection–dispersion equations (CDE) and including the sorption and degradation properties of the pesticides (see Pennell et al., 1990). The Pesticide Root Zone Model (PRZM) is such a model and has been tested for different soils of Hawaii (Loague et al., 1995, 1996). This model was fairly successful at simulating the depth and time to peak concentrations but failed to simulate the shape of transient concentration profiles. Loague et al. (1995, 1996) concluded that the simple fluid flow algorithm that does not account for preferential flow was an important limitation of the model. Many other studies have pointed out the discrepancy between observed transport behavior of chemicals during passage through soil columns and that predicted by the classical CDE (e.g., Kay and Elrick, 1967; Davidson and McDougal, 1973; Jardine et al., 1988). Such discrepancies, to some extent, were attributed to physical and chemical nonequilibrium processes like preferential flow affecting water flow and solute transport. Consequently, two-region models were proposed by several researchers (e.g., van Genuchten and Wierenga, 1976; Cameron and Klute, 1977).

Germann and Beven (1985) and Germann (1985) used kinematic wave theory to represent the gravity-driven flow in the macropores, which are often composed of cracks, fissures, and channels. Others (notably Gerke and van Genuchten, 1993a, b; Ray et al., 1997) have proposed models that account for traditional Darcian flow in the macropore and matrix regions with significant contrast in hydraulic properties. Gwo et al. (1995) proposed a three-region model to simulate flow and transport in micropore, mesopore, and macropore sequences. Reviews of flow and transport in preferential pathways in field soils (Edwards et al., 1990; Hutson and Wagenet, 1995; Flury, 1996) and rock fractures (Wang and Narasimhan, 1985; Mills et al., 1991) have pointed out the need for alternate models to simulate flow and transport that are not adequately represented by the CDE.

Sci-Grow is the Tier-I leaching model currently used by the USEPA to predict probable groundwater concentrations of pesticides. The model, based on the results of a dozen field prospective studies at various hydrogeologic settings, uses a regression approach, thus lacking rigorous process-based estimates. Process-based models are used on a routine basis in the European Union (Linders et al., 1999), and recently, Canada has used process-based (Tier II) models to estimate groundwater concentrations for the registration of newly developed pesticides. Some of these models have the ability to account for bypass flows. The USEPA plans to use process-based models for aggregate risk of a class of compounds as well as a single compound for registration needs. Likely concentrations in leachate water in groundwater or at a reference plane are needed for risk assessment. The USEPA has made significant in-house efforts along with the USGS (Nolan et al., 2005) to identify the most suitable models from a suite of seven models using validation data from four sites. The intent is to capture most of the processes that occur under agricultural management practices. However, the data needs for Tier II models are significantly higher than those for Tier I models.

The USEPA is seeking partnership with the state of Hawaii to examine the feasibility of using Tier II models in Hawaii conditions for pesticide registration. We feel that process-based models, especially those with macropore features, may be ideal for simulating pesticide transport under high rainfall events. Two pesticide leaching models, MACRO 4.3 (Jarvis, 2001) and S1D DUAL (Vogel et al., 2000; Ray et al., 2004), were chosen on the basis of their availability and ability to simulate explicitly preferential flow. The models employ different concepts for calculation of macropore flow. MACRO has been subjected to several previous validation exercises using European data sets (see Vanclooster et al., 2000; Trevisan et al., 2003). None of the models, however, has been validated against data sets from tropical regions. We realize that the available data sets for running process-based models for pesticide leaching in Hawaii may not be fully suitable for rigorous evaluation of these models. Nevertheless, the data set used in this evaluation is one of the best currently available for tropical soils. The results of this preliminary evaluation elucidate what additional data must be collected through controlled field experiments for the calibration and testing of a suite of these models. The simulation study will further allow us to assess the possible role of macropore flow in the structured soils of Hawaii. This is, however, subjected to the validity of macropore flow models, which remains difficult to demonstrate. Our objective in this effort is to asses the suitability of process-based models for characterizing the movement of pesticides in a structured Hawaii soil, with the intent of incorporating a Tier II model for future pesticide registration for the state of Hawaii.


   Materials and Methods
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results and Discussion
 Conclusions
 REFERENCES
 
The Kunia Data Set
A description of the Kunia data set and experimental methods was given by Gavenda et al. (1996). We give here only a brief overview of the experimental set-up and the measurements. The measurements were performed at Kunia in central Oahu. The mean annual precipitation is about 600 mm, and the mean annual temperature is about 24°C. The soil is classified as very-fine, kaolinitic, isohyperthermic (Molokai series) belonging to Typic Eutrotorrox (Gavenda et al., 1996). Gavenda (1989) concluded that the A horizon developed in Asian aerosolic dust, the B horizon developed in volcanic ash, and the 2B horizon developed in basaltic residuum. Within 3 m of the surface, the soil profile was described as Ap1 horizon with a thickness of 0.2 m; Ap2, B21, B22, and B23 horizons with 0.25 m, 0.4 m, 0.35 m and 1.8 m thickness, respectively (Gavenda et al., 1996). Previous laboratory and field solute transport studies on this soil suggested that stable soil aggregates could be responsible for a bimodal flow regime when the soil was significantly wet (Green et al., 1972; Rao et al., 1974). The experiment was conducted for 131 d from 1 June 1989 to 9 October 1989. Three plots (8 m x 8 m) were established on a well-drained site with gentle slope where runoff would be minimal. The site was tilled by disking before establishing the plots. The plots were adjacent and arranged in a row, which followed the contour of the site.

Five different pesticides (atrazine, ametryn, hexazinone, fenamiphos, and chlorpyrifos) important to Hawaii agriculture were used in the study. Selected physical and chemical properties of these five compounds were extracted from the USDA-ARS Pesticide Properties Database (http://www.ars.usda.gov/Services/docs.htm?docid=14199). The pesticides were applied to each of three replicate plots in the following quantities of active ingredients: 8.96 kg ha–1 for atrazine and ametryn, 2.24 kg ha–1 for hexazinone, 3.36 kg ha–1 for fenamiphos, and 9.74 kg ha–1 for chlorpyrifos. In addition, sodium bromide was applied as a tracer at a rate of 188 g NaBr per plot. All pesticides and bromide were mixed together in a 7.6-L container and applied using a CO2– pressurized backpack sprayer. Two passes of the sprayer over the treated area were made in an attempt to achieve spatial uniformity of chemical application. A 0.02-m layer of straw was spread over each plot within hours after pesticide application to limit erosion and surface sealing and to prevent photodegradation of the pesticides.

Sprinkler irrigation systems were installed before pesticide application. The system delivered about 10 mm h–1, and the irrigation events lasted between 4 and 14 h. Immediately after pesticide application, a 5-mm irrigation was applied to help move the pesticide into the soil. Subsequently, the plots were irrigated every week or at 10-d intervals (Fig. 1 ). Daily rainfall and pan evaporation data were collected within 100 m from the experimental plots during the study. Soil samples for pesticide and bromide extraction were collected at 2, 6, 10, and 19 wk after pesticide application. Soil samples were taken in six different holes, two holes per plot on each sampling date. Samples were taken at 0.1-m intervals from soil surface to 0.2 m depth and at 0.2-m intervals from 0.2 to 2.6 m depths. The same samples were used to measure soil water content at the time of sampling. The gravimetric soil water contents for five different depth zones were also measured at the initiation of the experiment (1 June 1989) and on several others occasions.


Figure 1
View larger version (21K):
[in this window]
[in a new window]

  		FIG. 1. Daily pan evaporation, irrigation, and precipitation (i.e., rainfall) during the 1989 experiment on Molokai soil at the Kunia site, Oahu, HI.

 
Soil samples (50 g moist) were extracted twice with ethyl acetate in 250-mL capped Erlenmeyer flasks and shaken on a wrist action shaker for 30 min. The supernatant was filtered through anhydrous sodium sulfate, and the extract was rotoevaporated to near dryness at 60°C and resuspended in 2 mL of ethyl acetate. The sample was then transferred to a 15-mL graduated tube in a final volume of 5 to 10 mL. Pesticide extracts were analyzed on a Hewlett-Packard 5890A gas chromatograph with 7673A auto sampler and 3392A integrator, as described in Gavenda et al. (1996). Bromide was extracted from sieved (2.0 mm) 50 g field-moist soil samples with deionized water using a 1:1 soil:water ratio, and measured in the extract by an ion selective electrode (model 94-35, Orion Research, Cambridge, MA). Minimum detection limits for soil samples were 100 µg kg–1 for bromide and 1.4 µg kg–1 for the pesticides.

Organic carbon content (foc) was measured at 0.2-m intervals from surface to 1 m depth in two replicate soil profiles. The foc from 1 to 2 m depth was estimated as the average between foc in 0.8- to 1-m and 1.8- to 2.1-m layers. The latter was obtained from Peterson et al. (1985). Soil sorption coefficients (Kd) were obtained by a batch adsorption test using soil samples from topsoil, 0 to 0.2 m, and subsoil, 0.6 to 0.8 m, taken from each experimental plot. Dividing the measured Kd by foc resulted in very close values for organic carbon partition coefficient, Koc, in topsoil and subsoil in the case of atrazine and hexazonine but rather different corresponding values in the case of ametryn, chlorpyrifos, and fenamiphos. Thus, Kd in each soil layer was estimated by multiplying organic carbon partition coefficient for atrazine and hexazonine with foc, whereas for the other pesticides, the measured Kd for topsoil and subsoil were assumed to be valid for 0 to 0.45 m and 0.45 to 2 m depths, respectively. These coefficients were then used in modeling of pesticide transport.

The same soil samples were used in laboratory degradation experiments to estimate pesticide degradation (Gavenda et al., 1996). The first-order kinetic model was used to calculate degradation rate coefficients and half-lives.

The total masses of each pesticide and bromide ion remaining in each soil depth interval at each sampling time were calculated from the concentration measured in each depth increment and the bulk density of that increment. Well-defined peaks were not generally observed for the pesticides, probably because of the complex interaction of the sorption and transport processes in the aggregated soils (Rao et al., 1974). The erratic behavior of measured concentration profiles may have been caused by field heterogeneity as well.

To compare the relative mobility of the pesticides, Gavenda et al. (1996) also estimated their center of mass (COM). The COM calculations involved summing the mass of a chemical in all measured depth increments in the soil profile, then locating the depth at which half of the mass was above the point and half below. The most mobile pesticides were hexazinone and atrazine, which moved deep into the profile. Ametryn movement was limited to the top 0.5 m. Chlorpyrifos and fenamiphos moved even less than ametryn. The focus of this modeling study was thus hexazinone, atrazine, and bromide.

Precipitation and Irrigation Data
Figure 1 presents the measured pan evaporation, rainfall, and applied irrigation water during the study period. Only three large storm events occurred, with the largest on 4 Oct. 1989 (Day 126 after chemical application). April through October is generally the dry season at this location, so it is not surprising to see little rainfall in leeward portions of the Hawaiian islands during this season of the year. Compared with rainfall, most irrigation events applied significant amounts of water. On Day 35, 140 mm of water was applied. Total irrigation significantly exceeded total rainfall. The application rate of irrigation (10 mm h–1) was substantially lower than the saturated hydraulic conductivity of the soil.

MACRO and S1D DUAL Models
The description of MACRO 4.3 was given by Jarvis (2001), and the description of S1D DUAL was given by Vogel et al. (2000) and Ray et al. (2004). Only a brief description of the models and their conceptual differences is given here.

Either of the two models can be run as a single-permeability (no macropore) model or as a dual-permeability model. The two pore regions (macropores and micropores) act as separate but interacting flow domains. While both models employ Richards equation (Richards, 1931) for calculation of water flow in micropores, S1D DUAL uses Richards equation for calculation of water flow in macropore region and MACRO treats preferential flow as a noncapillary, gravity-driven process, with a hydraulic conductivity expressed as a simple power law expression of the degree of saturation.

For the retention function and unsaturated conductivity, MACRO uses the expression of Brooks and Corey (1964) and Mualem (1976), whereas S1D DUAL uses van Genuchten (1980) and Mualem (1976) analytical models. In MACRO, the water exchange between the two regions is a function of water content gradient, effective water diffusivity (D), and an effective diffusion path length assumed to be half the aggregate width (d), whereas in S1D DUAL, the exchange of water is assumed to be a function of pressure gradient and the hydraulic conductivity at a hypothetical interface.

Both models use a similar equation for solute transport between the regions, consisting of a diffusion component and a mass flow component. The mass flow component is identical, whereas the diffusion component is different: the solute exchange coefficient is empirical in S1D DUAL but in MACRO it is a function of an effective diffusion coefficient and d similar to the function for the water exchange.

In both models, sorption is described by a Freundlich isotherm, and degradation is calculated using first-order kinetics. In this study, however, the exponent in the Freundlich isotherm was assumed to be 1, thus reducing it to the linear isotherm. In sum, it can be stated that the differences are limited to the macropore region and the coupling term between the matrix and macropore regions.

Parameterization Procedure
General
The models were parameterized using measured data and run without calibration for the whole experiment period, 1 June 1989 to 9 Oct. 1989. We deliberately avoided calibration because in a registration process, calibration of the models is almost impossible (Trevisan et al., 2003). Parameters used for soil hydraulic properties in the models were either measured or estimated from the measured data. Bulk density, {gamma}, and total porosity, {theta}s, were taken from Green et al. (1982). They measured bulk density and calculated {theta}s from bulk density and particle density. Saturated conductivity, Ks, was given by Yabusaki (1993) for the topsoil (0–0.2 m) and by Green et al. (1982) for the deeper layers. Initial water content, {theta}ini, was measured by Gavenda et al. (1996). The boundary-water content ({theta}b), water pressure head ({psi}b) and hydraulic conductivity (Kb) used in MACRO are assumed to be the same as the micropore porosity, air-entry pressure, and saturated conductivity of micropores, respectively.

Residual water content, {theta}r, pore-size distribution, n, flow path tortuosity, l, and {theta}b were estimated by use of the RETC program (van Genuchten et al., 1991) and by the measured retention data and unsaturated hydraulic conductivity reported by Yabusaki (1993) and Green et al. (1982) (Table 1).


View this table:
[in this window]
[in a new window]

  		TABLE 1. Measured (bulk density, {theta}s, Ks, foc, {theta}ini) and estimated (Kb, d, {theta}b, {theta}r, {alpha}, n, l) soil parameters for Molokai soil.{dagger} {theta}r, {alpha}, n, l were estimated by Brooks and Corey–Mualem (B&C) model for MACRO and by van Genuchten–Mualem (VG) model for S1D DUAL. The measured data were from Green et al. (1982), Yabusaki (1993), and Gavenda et al. (1996). The RETC program of van Genuchten et al. (1991) was used for the B&C and VG Parameters.

 
Parameter {alpha} in Brooks and Corey–Mualem expressions was taken as the inverse of {psi}b reported by Yabusaki (1993) for Ap and B21 horizons but estimated by RETC for the deeper horizons. For van Genuchten–Mualem expressions, {alpha} in all horizons was estimated by RETC. Based on the value of {psi}b hydraulic conductivities, Kb, were estimated from the measured unsaturated conductivity in different layers, K({psi}b). Sorption and degradation data were taken from Gavenda et al. (1996) (see Table 2). The soil characteristics of the macropore domain were chosen to resemble a coarse-textured porous medium with a relatively high saturated hydraulic conductivity (Ray et al., 2004).


View this table:
[in this window]
[in a new window]

  		TABLE 2. Soil sorption coefficients (Kd) and half-lives (t1/2) for Molokai soil obtained from laboratory experiments (data from Gavenda et al., 1996).

 
MACRO
The depth of the soil profile in the model was set to 3 m, comprising 20 numerical layers. The uppermost layer was set to 0.02 m, and the bottom numerical soil compartment had a thickness of 0.2 m. A unit hydraulic gradient was assumed for the bottom boundary condition since the water table is far beyond the 3-m depth profile in these soils. Evaporation from the straw-covered soil was simulated assuming a soil covered by dead crop, that is, switching off the transpiration but assuming that interception and soil evaporation were going on. Transpiration was prevented by setting the water content at wilting point, a value close to the porosity, {theta}b. Assuming a canopy interception capacity of 0.2 mm, the ratio of the total simulated evaporation to the pan evaporation became 0.17. This value is fairly close to 0.2 reported by Adams et al. (1976) for the ratio of actual evaporation to the pan evaporation for straw mulch. The aggregate half-width, d, was estimated using data about the size and shape of aggregates and degree of structural development given by Green et al. (1982) and the MACRO_DB expert system (Jarvis et al., 1997) (see Table 2). Other solute transport parameters were set to the default values in the model.

S1D DUAL
Similar to MACRO, S1D DUAL considered a profile depth of 3.0 m. Five different soil zones comprised the 3.0-m depth profile. The depth of each zone ranged from a low of 0.2 m to a high of 0.8 m. Altogether, 305 nodes were used to numerically discretize the 3.0-m domain. Near the soil surface, the domain was discretized to have a nodal spacing of 0.001 m. The maximum nodal spacing was 0.01 m. In essence, the numerical discretization in S1D DUAL was almost 14 times finer than that in MACRO. This was done to ensure that the Richards' equation converges to continuous solution during simulation. An atmospheric boundary condition was assumed for the top boundary for flow in which the time-dependent rainfall and irrigation values were used. A pan coefficient of 0.2 was used as the soil evaporation and was limited to the top 0.1 m. The bottom boundary was a unit hydraulic gradient (McCord, 1991). For solute transport, we assumed the pesticides were applied within a period of 30 min in 7.6 L of water. The bottom boundary was a zero concentration gradient condition, which would allow the pesticides to pass through it.

Comparison and Statistical Analysis
Graphical comparisons of simulated and measured water content and concentration and also comparison of simulated and measured COM were made to observe the general agreement between the models and the field measurements. Model performances in predicting quantitatively the concentration profiles were tested for bromide and the two most mobile pesticides: atrazine and hexazinone. Measured concentrations were compared with corresponding simulated values, and root mean square errors (RMSE) were calculated (Willmott, 1981). This parameter is defined by

Formula 1[1]
where Si and Mi are the simulated and measured concentrations, respectively, for the ith data point of n observations. The RMSE indicates the accuracy of the model. If the simulated and the observed values are the same, the RMSE is zero (the lower limit). Values of RMSE were calculated for each depth and sampling date.


   Results and Discussion
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
 Results and Discussion
Conclusions
 REFERENCES
 
Observed and Simulated Water Flow
During the 130-d experiment, a total of 1140 mm of water infiltrated into the soil profile (sum of natural rainfall and artificial irrigation). The simulated total evaporation from soil was about 270 mm during this period. On two occasions (Days 50 and 125 in Fig. 1), the models indicated that the soil surface reached fully saturated conditions, but only over short durations; thus, no surface ponding or surface runoff were simulated. Both models simulated saturation conditions after short-duration but high-intensity natural rainfalls. In total, 12 irrigation events with moderate intensity over the season were applied. The irrigation intensity (~10 mm h–1) was limited by the design capacity of the irrigation system.

The S1D DUAL code was run for a single domain initially to examine if there is sufficient surface saturation or saturation at layer interfaces to initiate macropore flow. From the single-domain simulations, the rainfall and irrigation over the simulation period induced a close-to-saturation situation at the soil surface after almost all events. Because of the unsaturated condition at the beginning of the simulated period, the first outflow at a depth of 3 m started on Day 40 (Fig. 2 ), only after prolonged irrigation on Days 34 and 35. The response to this irrigation input was relatively slow—the first effluent water left the 2-m-deep profile 34 h after the beginning of irrigation. The highest water percolation was obtained after the irrigation event on Day 70, when 120 mm of water infiltrated in 12 h. As expected, irrigation caused a regular pattern of deep percolation over the simulated time. Recorded natural rainfall did not cause any large water outflow from the soil profile because the total water input was small. The MACRO model showed a low macropore flow only after high intensity rainfalls (Days 2, 50, and 125). It was limited to the uppermost 0.12-m layer.


Figure 2
View larger version (12K):
[in this window]
[in a new window]

  		FIG. 2. Simulated (S1D DUAL) water percolation from 3-m soil profile.

 
Neither of the models showed any substantial preferential flow. The reason is probably the low intensity of water input (rainfall or irrigation) relative to the high hydraulic conductivity of matrix domain. Ray et al. (2003) measured in situ the unsaturated hydraulic conductivity using a tension disc infiltrometer. They found generally high values for the whole soil profile. The unsaturated conductivity at –100 mm pressure was about 20 mm h–1 at 1 m depth. In the present study, the simulated pressure head in the macropore domain at the soil surface remained close to –100 mm after almost all rainfall or irrigation events.

To test the impact of rainfall intensity on the occurrence of preferential flow, the models were run using soil and site properties from the experiment site but with the rainfall data from another site on March 2006 when the area experienced an unprecedented amount of rain. During a 31-d period, 1600 mm of rainfall occurred at an experiment station of the University of Hawaii approximately 12 km from the Gavenda et al. (1996) site used for the present study. Except for four days, all other days were rainy. The models showed macropore outflow from 1 m depth only in one event, namely after 10 rainy days and following 120 mm of rain in 5 h (Fig. 3 ). The proportion of macropore flow to the total outflow was only 15%. This test indicates that the high saturated conductivity of micropores (matrix) is the major reason for minimal flow of water in the preferential flow domain. The Oxisols of Hawaii can conduct large quantities of water because of their water-stable aggregates. The test also illustrates the usefulness of these models, which are capable of simulation of preferential flow, for pesticide registration.


Figure 3
View larger version (15K):
[in this window]
[in a new window]

  		FIG. 3. MACRO-simulated outflow from the soil profile at 1 m depth during a large episodic storm that occurred in March 2006.

 
The (composite) soil water distribution simulated by MACRO reveals a closer agreement with measured water content than the S1D DUAL model using the dual-permeability approach (Fig. 4 ). The simulated water contents from S1D DUAL are generally higher than measured data. The largest deviation occurred in the top layer. Note that the measurements of soil water contents over time were few and probably not sufficient to capture the dynamic nature of soil moisture in a well-structured soil. The differences between the model outputs, however, may be due to different expressions used for the water retention curve. As mentioned earlier, MACRO uses the soil-water retention equation proposed by Brooks and Corey (1964), whereas S1D DUAL uses the equation proposed by van Genuchten (1980). These two equations behave particularly differently in the near-saturation zone. Volcanic soils can be distinguished from other soils by having higher water contents at saturation and at wilting point compared with most other soils that are less aggregated and have less "bound" water (Maeda et al., 1977). This gives the water retention curve a typical shape that looks like a sandy soil at near saturation but a clay soil for the other parts (Ray et al., 2003).


Figure 4
View larger version (21K):
[in this window]
[in a new window]

  		FIG. 4. Measured (mean values and standard deviations, squares and bars) and simulated (MACRO, dashed lines; S1D DUAL, solid lines) soil water contents at different depths during the experiment period (summer 1989). All values are a composite of water content in the matrix and macropore domains.

 
Observed and Simulated Chemical Transport
Observed Movement of the COM
The progressive downward movement of the COM was measured based on the amount of recharge water entering the soil profile (Gavenda et al., 1996). With 50 mm of recharge, the COM values for bromide, atrazine, and hexazinone were 0.24, 0.22, and 0.10 m, respectively (Table 3). Toward the midpart of the experiment, when the total recharge was 600 mm, the COM values for these three chemicals were 1.45, 0.25, and 0.72 m. Toward the end of the season, with 1000 mm of recharge, the COM moved to 2.33, 0.35, and 1.71 m, respectively, for the three chemicals. The depths of COM for ametryn, chlorpyrifos, and fenamiphos remained unchanged at 0.1 m during the whole study period (Table 3).


View this table:
[in this window]
[in a new window]

  		TABLE 3. Chemical leaching in Molokai soil: measured (from Gavenda et al., 1996) and simulated depth of COM on four sampling dates.{dagger} Measured values represent means and one standard deviation range.

 
Bromide
A comparison of simulated and measured bromide profiles in soil provides a means of evaluating how well a model simulates the hydrodynamic process in transporting solute without the complications of sorption and transformation. The field-measured and simulated bromide concentration profiles at 2, 6, 10, and 19 wk after bromide application are shown in Fig. 5 . The pulse of bromide moved deeper with time in response to water input and was beyond 2.8 m at the end of the sampling campaign. MACRO overpredicted peak depth at 6 and 10 wk; however, at 19 wk it underpredicted peak depth. S1D DUAL predicted peak depth reasonably well at 6 and 10 wk, but it was less successful at 19 wk compared to MACRO. Peak concentration predictions were generally below (2 wk) or above (6 and 10 wk) field-measured values for both models, although MACRO was close to the measured value at 10 wk. MACRO provided a closer match to the measured value at 19 wk than S1D DUAL. The bromide RMSE values were generally smaller for MACRO than S1D DUAL, with the largest difference at 19 wk (Table 4). The bromide COM estimate by MACRO (Table 3) was better than that of S1D DUAL at all four sampling dates. The overall better estimation of COM depth and smaller RMSE for MACRO and a closer match to the measured profile for MACRO at 19 wk suggest that MACRO is likely the better of the two models for simulating nonadsorbed solute transport in the Molokai soil.


Figure 5
View larger version (23K):
[in this window]
[in a new window]

  		FIG. 5. Measured (mean values and standard deviations) and simulated bromide profiles in the Molokai soil at four times after bromide application.

 

View this table:
[in this window]
[in a new window]

  		TABLE 4. Root mean square error (RMSE) for bromide, atrazine, and hexazinone concentrations at 19 wk after chemical application. These statistics compare the MACRO and S1D DUAL outputs at different depths.

 
Pesticides
Only hexazinone and atrazine moved sufficiently beyond the soil surface 0.1 m to be useful for evaluating model performance at greater depths (Table 3). Simulated concentration profiles for atrazine and hexazinone are compared with measured field data in Fig. 6 and 7 . Both models matched field concentrations quite well at 2 and 6 wk but failed to represent profile shape at 10 and 19 wk. Field-measured profiles show retention of atrazine and hexazinone in the top 0.3 m at concentrations exceeding 100 µg kg–1 for atrazine and 10 µg kg–1 for hexazinone. Both models predict diminishing concentrations over time at these shallow depths. Deeper in the profile, measured concentrations of both chemicals stay fairly constant with depth at about 1 to 10 µg kg–1 for atrazine and 10 to 100 µg kg–1 for hexazinone. These profiles resemble those measured by Rao et al. (1974) for picloram on the same soil. Similar shapes for atrazine concentration profile have been found for other soils (Butters et al., 2000; Loague and Green 1991). Preferential flow or enhanced movement through sorption to mobile colloids has been suggested as the reason for the deep movement of atrazine (Butters et al., 2000). Both models predict deep penetration of hexazinone at concentrations similar to those measured in the field, but neither model accurately represents atrazine movement at 10 and 19 wk. Roulier et al. (2006) and Shirmohammadi et al. (2001) used MACRO to simulate the movement of atrazine in a calisol and a sandy loam soil, respectively. MACRO simulated the atrazine concentration better in the upper soil layers than at lower depths. The most sensitive parameters were the degradation rate coefficient and sorption coefficient (Roulier et al., 2006). The RMSE values for atrazine and hexazinone in Table 4 suggest similar prediction error for both models at 19 wk. Considering that none of the models were calibrated, these discrepancies are not surprising. Probable explanations for the inabilities of the models to accurately represent field movement of mobile pesticides include the inadequate representation of (i) chemical adsorption–desorption throughout the profile—Swanson and Dutt (1973) and Johnson et al. (1995) demonstrated adsorption–desorption hysteresis (i.e., adsorption rates much faster than desorption rates) for atrazine in a variety of soils; (ii) chemical degradation or transformation throughout the profile; and (iii) the dynamics of transport into and out of aggregates and the interaction of this process with solution flow through larger pores. Given the deficiencies of both models, we are unable to suggest whether one model is better than the other for use in pesticide registration on the structured soils in Hawaii. However, MACRO is in a more fortunate situation as it better predicted the water content and bromide profiles and because it is used for pesticide registration in Sweden, Denmark, Norway, and the EU (Harmonized Registration of Active Ingredients).


Figure 6
View larger version (23K):
[in this window]
[in a new window]

  		FIG. 6. Measured (mean values and standard deviations) and simulated atrazine profiles in the Molokai soil at four times after atrazine application.

 

Figure 7
View larger version (24K):
[in this window]
[in a new window]

  		FIG. 7. Measured (mean values and standard deviations) and simulated hexazinone profiles in the Molokai soil at four times after hexazinone application.

 
Comparison of Results from Tier I (AFR) with a Dynamic Model
While the MACRO and S1D DUAL models may not yet be suitable for use in detailed prediction of pesticide movement for the site studied, it is instructive to compare the results from a dynamic model (MACRO) with the presently used Tier I Revised Attenuation Factor (AFR) index. This comparison is accomplished by way of the COM calculations given in Table 3. The depth of COM from field measurements and MACRO are shown for the five pesticides in Fig. 8 . The AFR values for the same chemicals are shown in the bottom graph in Fig. 8. For AFR calculation, we used Molokai soil at 0 to 3% slope, depth of seasonal high water table of 1.0 m (standard deviation, 0.2 m), and a daily recharge rate of 0.001 m day–1 (standard deviation of 0.0002 m day–1). The two reference chemicals for comparison are diuron and dibromochloropropane (DBCP). Diuron has never been detected in the groundwaters of Hawaii, while DBCP has been detected in many wells. Diuron is considered "unlikely to leach" and DBCP "likely to leach" in Fig. 8. The AFR for atrazine lies close to DBCP, suggesting that it is a probable leacher. The uncertainty band for ametryn (not shown here) extends to the boundaries that define it to be either a leacher or a nonleacher. This uncertainty arises from the variabilities in soil properties (Loague et al., 1996). Figure 8 illustrates that the relative leachabilities of the five pesticides in the present study are represented similarly by both the AFR and the MACRO-simulated COM at 19 wk. As new protocols for registration require concentrations of a given chemical in groundwater, dynamic models such as MACRO and S1D DUAL are needed. Consistency in characterizing pesticide leachability is necessary for a pesticide registration protocol that will use both Tier I and Tier II models.


Figure 8
View larger version (16K):
[in this window]
[in a new window]

  		FIG. 8. Relative mobilities of five pesticides in Molokai soil as indicated by the Tier I AFR index, dynamic MACRO model, and field data. DBCP, dibromochloropropane.

 

   Conclusions
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
 Results and Discussion
 Conclusions
REFERENCES
 
The two process-based models included in this evaluation identified the most leachable compounds; these results were qualitatively consistent with the results from the Tier I AFR model currently in use for pesticide registration in the state of Hawaii. However, the two Tier II models have the capability to predict concentration profiles in the soil and the concentrations of the chemicals leaving the soil profiles, which the AFR model cannot do. The concentration profiles obtained from MACRO and S1D DUAL were somewhat different from the measured profiles. For some compounds, MACRO performed well in the early part of the study period. For bromide, MACRO-simulated concentration matched field data better than that for S1D DUAL. However, no large differences were found between the predictive performances of the two models. It should again be pointed out that both models were run uncalibrated in this study.

The apparent absence of substantial preferential flow in the experiment was probably the result of the relatively high matrix flow of this aggregated soil, so that most of the water infiltrated during rainstorms or irrigation was conducted in the matrix region of the soil. Simulation of water movement for a selected test storm of very high intensity and long duration illustrated that a dual-porosity model would be an asset for areas of Hawaii for which high-intensity storms are a common occurrence.

Chemical transport in the structured soils of Hawaii has not been easily modeled using the standard convection–dispersion reaction equations. Dynamics of sorption and desorption during flow are thought to be the cause of such discrepancies. More efforts are needed to parameterize and validate these process-based models.


   ACKNOWLEDGMENTS
 
We thank the Organization for Economic Cooperation and Development (OECD) for providing the fellowship to the first author to come to Hawaii to conduct this research. Additional support was provided by the research fund of the Ministry of Education of the Czech Republic (MSM 6840770002) to support the contribution of J. Dusek to this work. This is Water Resources Research Center, University of Hawaii contributed paper CP-2007-08.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
 Results and Discussion
 Conclusions
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Alavi, G.
Right arrow Articles by Ray, C.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Alavi, G.
Right arrow Articles by Ray, C.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Alavi, G.
Right arrow Articles by Ray, C.
Related Collections
Right arrow Agricultural Pesticides
Right arrow Flow
Right arrow Fate

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome