HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Related articles in JEQ Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via ISI Web of Science (8) Citing Articles via Google Scholar Google Scholar Articles by Miao, Z. Articles by Trevisan, M. Search for Related Content PubMed PubMed Citation Articles by Miao, Z. Articles by Trevisan, M. Agricola Articles by Miao, Z. Articles by Trevisan, M. Related Collections Organic Compounds

TECHNICAL REPORTS

Organic Compounds in the Environment

Modeling the Effects of Tillage Management Practices on Herbicide Runoff in Northern Italy

^a Istituto di Chimica Agraria ed Ambientale, Università Cattolica del Sacro Cuore, 29100 Piacenza, Italy
^b Dipartimento di Scienze e Tecnologie Agroambientali, Università di Bologna, 40126 Bologna, Italy

^* Corresponding author (ettore.capri{at}unicatt.it).

Received for publication May 26, 2003.

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
The need to quantitatively predict pesticide runoff and erosion under cropping system management has gained increasing importance. In Europe, predictive models have not yet been fully validated because of the lack of field data sets. The objective of this study was to validate the capability of PRZM (Pesticide Root Zone Model) 3.12 to predict water runoff, sediment erosion, and associated transport of atrazine (6-chloro-N²–ethyl-N⁴–isopropyl-1,3,5-triazine-2,4-diamine), terbuthylazine (N²–tert-butyl-6-chloro-N⁴–ethyl-1,3,5-triazine-2,4-diamine), and metolachlor [2-chloro-6'-ethyl-N-(2-methoxy-l-methylethyl)acet-o-toluidide] under common tillage management practices found in northern Italy. A 2-yr field data set was used to evaluate the model. Results showed that the model could qualitatively simulate significant differences of water runoff, soil erosion, and associated herbicide losses between conventional tillage (CT) and minimum tillage (MT) for a winter barley (Hordeum vulgare L.) cover crop. For MT, water runoff, soil erosion, herbicide losses in water runoff and eroded sediment, and the proportion of herbicide loss via sediment erosion were significantly lower than for CT. The model failed to correctly simulate event-based herbicide concentration, water runoff, and soil erosion. The model usually underestimated pesticide runoff events with high rainfall intensity and low daily precipitation volume, and overestimated runoff events with low intensity and high volume. The main reason was that the description of runoff and erosion processes is rather empirical in the model and not physically based. Moreover, model calculations do not adequately reflect the relationships between soil erosion intensity and chemical concentration in sediment losses, leading to discrepancies between predictions and field observations.

Abbreviations: CT, conventional tillage • IREG, location of the Natural Resources Conservation Service 24-h hyetograph • MT, minimum tillage • PRZM, Pesticide Root Zone Model • RMSE, root mean squared error • SCS, Soil Conservation Service

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
IN SLOPING AGRICULTURAL FIELDS, pesticide runoff can adversely affect the environment (Isensee et al., 1990; Vicari et al., 1999). In a review by Wauchope (1978), surface runoff was identified as a conduit for pesticide transport to surface waters with reported total losses ranging from less than 0.1 to 5% of the applied mass. In general, losses of up to 5% can be expected from fields of moderate slope (10–15%), and losses of up to 2% can be anticipated from fields of low slope (3% or less). For the majority of commercial pesticides, total pesticide runoff losses are 0.5% or less of the amounts applied, unless severe rainfall occurs within 1 to 2 wk after application. A recent compilation of runoff studies was published by the USGS, covering an extremely wide range of spatial scales (from bench top to major watersheds), varied physical locations (primarily United States and Europe), and different pesticides (Capel et al., 2001). The average runoff loss in the European study sites was 0.8% of the applied chemical. For small watersheds (0.1–100 ha), the average runoff loss was 0.7% of the applied, indicating that runoff losses are essentially independent of the size of the watershed.

Field studies have demonstrated that appropriate management practices will reduce pesticide runoff, depending on the characteristics of chemicals and soils. Around Lake Erie in the United States, for instance, conservation tillage was identified as the most cost-effective means to reduce pesticide runoff and sediment losses from agricultural land while maintaining productivity (Richards et al., 2002). In Illinois, the percentage of applied carbofuran lost in runoff and sediment ranged from 1% (contoured moldboard) to 11% (up-to-down slope moldboard), and the percentage of applied alachlor lost in runoff and sediment ranged from 1% (contoured no-till) to 2% (contoured moldboard) (Felsot et al., 1990; Villholth et al., 2000). Wauchope et al. (1990) and Hall et al. (1991) reported that the percentage of herbicide runoff loss (1984–1988) was less under no tillage than under conventional tillage corn (Zea mays L.) production. Runoff from bare plots occurred with one-third of the rainfall required to produce the same runoff from grassy plots, and the sediment loss in bare plots was double that in the grassy plots. Thus, tillage practice has a large effect on pesticide fate in farming systems.

Models to simulate pesticide runoff and erosion are very useful for studying the effect of cropping systems and land uses, and several comparison and validation studies have been published for PRZM in the United States (Loague and Green, 1991; Carsel et al., 1998; Warren-Hicks et al., 2002). Recently, for example, FIFRA (the U.S. Federal Insecticide, Fungicide and Rodenticide Act) Exposure Model Validation Task Force completed a validation procedure for PRZM (Version 1.0), which included comparison of simulated runoff and erosion with the results of field-scale experiments (Russell and Russell, 2001). In the literature, PRZM is regarded as an acceptable model to provide reasonable estimates of the runoff coefficient (fraction of the precipitation resulting in runoff) as well as good estimates of cumulative runoff fluxes. The predictions of PRZM for individual runoff events generally agreed with field data within one order of magnitude and cumulative values (e.g., runoff summed over the study period) agreed within a factor of approximately 3. The accuracy of runoff and erosion predictions is better for large than small rainfall events. Nevertheless, in Europe, such a model has not been fully validated as field data sets are lacking (Russell and Russell, 2001). The validation of this model under European conditions is urgently needed, since the model has been accepted for regulatory use for calculating the environmental concentration of pesticide in runoff, potentially a source of contamination of the surface water bodies (FOCUS, 2001).

The objective of the study was to validate the applicability of PRZM 3.12 to simulate water runoff, sediment erosion, and associated transport of three herbicides (atrazine, terbuthylazine, and metolachlor) under the influence of tillage management practices with a 2-yr field data set collected in northern Italy.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Field Site and Data Set
The study was conducted in a hilly area with northern aspect at Ozzano dell'Emila, 20 km east of Bologna, Italy, at 200 m above sea level (Fig. 1). The soil is classified as a fine, mixed, mesic, Udertic Ustochrept, and has a loamy texture, with 420, 340, and 240 g kg^–1 of sand, silt, and clay, respectively. The pH (1:2.5 soil to water) is 7.7 and organic carbon is 9.2 g kg^–1. The site consisted of eight plots at 15% slope. Each plot had a surface of 7 x 50 m, with the longest side perpendicular to contour lines. The plots were hydrologically isolated by means of ditches and separated by 5- x 50-m buffer plots. Runoff water and sediment from the plots were collected at the bottom of each plot by means of PVC line ditches and transported by aluminum pipes to the measuring devices located downstream in a field laboratory. Rainfall and temperature data were recorded at a weather station located about 100 m from the experimental area.

View larger version (22K): [in this window] [in a new window]   Fig. 1. Map of the study site (northern Apennine, Italy, 44°28'N, 11°28'E)

Runoff Measurement and Sampling
An automated system for runoff measurement and collection, described in detail by Rossi Pisa et al. (1994)(1999), was used. In brief, on each plot, a cylindrical stainless steel pan (150-L volume) collected and measured runoff. Each pan contained a mixer to homogenize runoff water before sampling, a level sensor, and a washing system. Runoff samples (2000 mL) from each plot were collected on a daily basis, kept in bottles, and stored by automatic sampling machines controlled by a preset central control system. The sampling systems, one for each plot, had 12 glass bottles (2 L) maintained at 4°C in the sampling machine. The samplers were set to take a 2000-mL sample of runoff from each pan. A 200-mL subsample was taken when the water level in the pan increased each 10 cm. Each 10-cm increase corresponded to a runoff of 0.055 mm. When the pan was filled (120 L, corresponding to 0.34 mm of runoff), the bottom of the pan opened and the water drained out. Then, the bottom of the pan closed and the sampler passed to take a 2000-mL sample in the subsequent glass bottle. Collected runoff samples from each plot were composited on a daily basis to give a volume of 2000 mL. The samples were kept in 2-L bottles and stored in the dark at –20°C.

The samples were successively analyzed for sediment content and for atrazine, terbuthylazine, and metolachlor concentrations in water and in sediment as described by Rossi Pisa et al. (1994). Briefly, the extraction of herbicides from the sediment was performed by adding 15 mL of dichloromethane and 15 mL of acetone to glass 75-mL centrifuge tubes containing 20 g of sediment. The centrifuge tubes were shaken for 2 h in a horizontal shaker and sonicated for 5 min in a sonicator bath. The slurry was centrifuged for 5 min at 4000 x g and filtered at 0.2 µm. The supernatant was then passed through a NP-SPE-NH₂ column and eluted with 3 mL of methanol. The sample was evaporated to dryness and redissolved into 0.5 mL of methanol into a 2-mL high performance liquid chromatography (HPLC) vial, and stored at –12°C until analysis. The multiresidue analysis of the three herbicides was performed by HPLC. The HPLC system was a Varian (Palo Alto, CA) with a UV-VIS detector. Detection was performed at 220 nm, injecting 100 µL of each sample into a C₈ column (15 cm long x 4.6-mm i.d.). The mobile phase, in isocratic conditions, was acetonitrile and ammonium acetate (45:55 v/v). The limit of detection of the three herbicides was 10 µg kg^–1.

The extraction of the three herbicides from the water phase was performed by concentrating 1 L of the sample on a preconditioned C₁₈ solid-phase extraction (SPE) column, eluted with 3 mL of methanol. The sample was evaporated, redissolved, and analyzed by HPLC as previously described. The limit of detection of the three herbicides was 0.1 µg L^–1.

Pesticide Root Zone 3.12 Model
A detailed description of PRZM 3.12 can be found in Carsel et al. (1998). We provide a description of the basic theories and the new developments of PRZM 3.12 used for predicting water runoff, sediment erosion, and pesticide concentration in water runoff and eroded sediment in the research.

In PRZM, the prediction of water runoff is based on the Soil Conservation Service (SCS) curve number. Precipitation and/or snowmelt are inputs into the SCS runoff equation written as (Russell, 1999):

[1]

where Q is daily water runoff volume (cm d^–1), P is the net precipitation rate (rainfall minus crop interception, cm d^–1), SM is the snowmelt (cm d^–1), and S is the watershed retention parameter estimated by:

[2]

where RCN is the SCS runoff curve number.

The RCN is a sensitive variable for estimating water runoff, and is strongly linked to soil type, soil drainage properties, soil management practices, and crop type. The SCS rainfall distributions were developed for flood control design. Nevertheless, the meteorological files of PRZM do not require values of rainfall intensity over time. Rainfall intensity is assumed to occur according to designed storm distributions developed by the Soil Conservation Service (Type I, IA, II, and III, i.e., IREG [location of the Natural Resources Conservation Service 24-h hyetograph]).

For soil erosion, PRZM 3.12 provides three methods: the Modified Universal Soil Loss Equation (MUSLE), which was contained in earlier versions of PRZM, and two recent modifications, MUST for theoretical calculations and MUSS specially designed for small watersheds. The common formula of MUSLE, MUST, and MUSS is:

[3]

where X_e is the event soil loss (Mg d^–1), V_r is the volume of daily runoff (mm), q_p is the peak storm runoff (mm h^–1), A_f is the field size (ha), K is the soil erodibility factor (dimensionless), LS is the slope-length factor (dimensionless), C is the soil cover factor (dimensionless), and P_c is the conservation practice factor (dimensionless).

In our research, the MUSS method was adopted pursuant to the small watershed characteristics of the Ozzano experimental fields, and b₁, b₂, and b₃ were assigned the values of 0.79, 0.65, and 0.009, respectively.

The peak storm runoff, q_p, was calculated using the Graphical Peak Discharge Method:

[4]

in which the variable a is a unit conversion factor, A is the cross-sectional area of the soil column (cm²), F_p is the pond and swamp adjustment factor, and q_u is the unit peak discharge:

[5]

where T_c is the concentration time (h) defined as the time it takes water to flow from the furthest point in the watershed to a point of interest within the watershed and is a function of basin shape, topography, and surface cover; and C₀, C₁, and C₂ are regional coefficients that relate storm intensity, rainfall volume, and initial abstraction. The driving variables are IREG (SCS rainfall distribution region, location of NRCS 24-h hyetograph), which affects the concentration time (T_c), and defines the SCS regional coefficients and other parameters such as SLP (slope of the hydraulic flow path) and Manning's roughness coefficient, which measures the resistance of open channels to flow.

In the model, herbicide runoff includes two parts: the dissolved phase pesticide loss via water runoff and the sorbed phase loss in eroded sediments. In the previous release (2.2) of PRZM, chemical residues in the dissolved phase were uniformly and completely available for runoff to a depth of 1 cm. Residues below 1 cm were unavailable for runoff. With the nonuniform extraction model in PRZM 3, chemical residues have decreasing availability with depth (nonlinear model) because the interaction between soil pore water and excess precipitation (runoff) diminishes with depth in the soil. The nonuniform extraction model employs an exponential curve to reduce the amount of dissolved-phase chemical that is available to mix with runoff water:

[6]

where DRI_i is the fraction of dissolved-phase chemical present in compartment i available for runoff, Midtot_i is the depth to midpoint of compartment i (cm), 0.7 is an efficiency factor, and 0.9 is the depth-reduction coefficient. Calculations were performed for all compartments from the surface to a depth of 2 cm. The efficiency factor and depth reduction coefficients were derived empirically by the model authors (Carsel et al., 1998). The daily pesticide flux due to runoff is calculated as from Eq. [7]:

[7]

where J_QR is the daily pesticide loss due to runoff (g d^–1), Q is defined as in Eq. [1], C_w is the dissolved concentration of pesticide (g cm^–3), A_w is the watershed area (cm²), and A is the cross-sectional area of the soil column (cm²).

Removal of sorbed pesticides by eroded sediments is calculated from soil erosion amounts:

[8]

where J_ER is the pesticide loss due to erosion (g d^–1), X_e is the erosion sediment loss (Mg d^–1), P_u is a unit conversion factor (g Mg^–1), C_s is the sorbed concentration of pesticide (g g^–1), A and A_w are defined in Eq. [7], and r_om is the enrichment ratio for organic matter (g g^–1) calculated by:

[9]

Model Parameterization and Assumptions
The variables and parameters of PRZM 3.12 are summarized in Table 1. Some parameters, including crop practices (e.g., the date of crop emergence and harvest, maximum coverage area, and interception storage of the crop), water management strategies (e.g., irrigation date and methods, precipitation events and intensity), and chemical application (e.g., pesticide application date, rate, and efficiency), were measured from field experiments. Other parameters, for example, the Universal Soil Loss Equation erodibility parameter, Manning's roughness coefficient, and the pesticide decay rate in water, soil, and plant leaves, were derived from other calculations within PRZM, empirical estimation, the PRZM 3.12 manual, or other related literature (Leonard et al., 1988; Cheng, 1990; Jury et al., 1991; Rossi Pisa et al., 1994; Tomlin, 1994; Sanchez-Camazano et al., 1995; Donatelli et al., 1996; Kum et al., 1996; Carsel et al., 1998; Elliott et al., 2000; FOCUS, 2000; Eckersten et al., 2001; Miao et al., 2003). For example, the K_d values for metolachlor, atrazine, and terbuthylazine are the average values given by Tomlin (1994).

The soil was divided into two horizons and each horizon was subdivided into 0.15-cm-thick compartments. Bulk density, wilting point, and field capacity were measured to be 1.55 Mg m^–3, 0.25 m³ m^–3, and 0.35 m³ m^–3, respectively.

Solar radiation and evapotranspiration were estimated by Radest 3.0 and the Priestley–Taylor equation (Donatelli et al., 1996, 2003).

With respect to the two tillage management practices (CT and MT), all parameters were the same with the exception of the following four variables: ISCOND (surface condition of initial fallow, cropping, and residues), ICNAH (surface condition of the crop after harvest date: fallow, cropping, and residues), USLEC (universal soil loss cover management factors for fallow, cropping, and residues), and RCN (runoff curve numbers of antecedent moisture condition). The first three parameters were assigned to "fallow" for CT and "crop residues" for MT. The RCN for barley cover plots was derived based on the impact coefficient of barley residues of 1000 kg ha^–1 in the PRZM 3.0 manual.

The simulation was performed to cover the period 1991 to 1992. The predicted mass of pesticide lost in runoff water was calculated from the concentration in water (µg L^–1) and water runoff volume (m^–3). Similarly, the pesticide lost in eroded sediment was calculated from the concentration (µg kg^–1) and sediment mass (kg).

Model performance was assessed objectively by comparing the agreement of the model with field measurements (e.g., water runoff, soil erosion, pesticide mass and concentration lost in water runoff and eroded sediment). The statistical index of root mean squared error (RMSE, %) was used to express the degree of overall fit between the predicted and observed:

[10]

where P_i is the predicted value, Q_i is the observed value, is the average of the observed values, and n is the number of observations. The lower limit for RMSE is zero. So far, the standard for model evaluation using RMSE has not been established for solute transport models, but one would hope to have a perfect fit value as close to 0.0 as possible (Loague and Green, 1991).

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Water Runoff
The model could only qualitatively simulate the influences of tillage practices on water runoff. Predicted water runoff volumes for both MT and CT were higher than the measured volumes of MT, especially in 1992. For instance, in 1992, the predicted water runoff volumes of MT and CT were 19.4 and 43.6 mm, whereas the observed measurements were 2.0 and 15.3 mm, respectively (Table 2).

View larger version (28K): [in this window] [in a new window]   Fig. 2. Daily precipitation volume and maximum rainfall intensity during the simulated period (1 June–30 Sept. 1991 and 1 Jan.–31 Dec. 1992).

The model did not predict as many runoff events as were observed (Fig. 3). For MT plots, 75 and 56% of the observed runoff events were predicted by the model in 1991 and 1992, respectively. For CT, 60% of the observed runoff events were matched by the simulation over the 2-yr period.

View larger version (35K): [in this window] [in a new window]   Fig. 3. Observed vs. predicted event-based water runoff volume in 1991 and 1992. Bars are standard deviations.

Soil Erosion
In the simulation, soil erosion events were strongly related to water runoff events. With respect to water runoff, the model was able to differentiate the two tillage practices. The predictions of erosion from MT were lower than from CT by a factor of 10 to 20 (Table 4). The results agreed with previous field studies that the bare plots required one-third less rain to produce the same amount of runoff and yielded twice as much sediment as the grassy plots (Felsot et al., 1990; Wauchope et al., 1990; Hall et al., 1991; Rossi Pisa et al., 1994; Capel et al., 2001).

The model underestimated or failed to describe the sediment erosion events with low daily precipitation volume and high intensity and overestimated soil erosion events with high daily precipitation volume and low intensity. According to the field experiments, however, soil erosion was mainly determined by rain intensity rather than daily precipitation volume. On 14 and 17 July 1991, for instance, although the total daily precipitation were just 33.3 and 15.6 mm, the peak rainfall intensities were as high as 66.6 and 32.8 mm h^–1, respectively. For these two rainfall events, 4.3 and 3.7 Mg ha^–1 of soil eroded sediment were observed in CT plots and water runoff was 0.31 and 1.09 mm, respectively. Another example was on 11 July 1992, when the daily rainfall was 11.6 mm with a water runoff of 2.47 mm, and the observed sediment loss was 2.9 Mg ha^–1, since rain intensity reached as high as 11.6 mm h^–1 (Fig. 2 and Table 4). Nevertheless, the model underestimated or failed to simulate these extreme sediment erosion events. Without considering the extreme events, for example, in 1991 for MT and CT the predicted total sediment erosion was 0.013 and 0.091 Mg ha^–1, which was close to the observed erosion of 0.013 Mg ha^–1 for MT and 0.161 Mg ha^–1 for CT.

The model overestimated sediment erosion events with high daily precipitation volume and low intensity. Similar to water runoff, for the event of 1 May 1992 with a daily precipitation of 65.2 mm and rainfall intensity less than 6.4 mm h^–1 and the event of 8 Dec. 1992 with daily precipitation of 81.8 mm and rainfall intensity less than 8.0 mm h^–1, the model greatly overestimated soil erosion (Table 4), but actually no or little sediment erosion occurred. After the removal of the two extreme soil erosion events, the simulation of erosion events and amounts was in good agreement with the observations (Table 4), and the corresponding RMSE values would decrease from 1809.5 to 245.1% for MT and 441.2 to 374.6% for CT (Table 3). The reason is that although the three theoretical soil erosion equations (MUSLE, MUST, or MUSS) consisted of the variables of storm intensity and volume of daily runoff (mm), the real driving variables were IREG (i.e., SCS rainfall distribution region, location of NRCS 24-h hyetograph) and other parameters such as SLP (slope of the hydraulic flow path) and Manning's roughness coefficient, which measures the resistance of open channels to flow (see Eq. [3], [4] and [5]) (Carsel et al., 1998). The model did not directly consider the rainfall intensity. Meanwhile, the IREG values for our study area in northern Italy were not available, so IREG was estimated from research conducted in the United States (Carsel et al., 1998; Russell, 1999).

To investigate the effects of SCS-designed storm distribution on runoff and soil erosion and determine which rain pattern was most suitable for Italy, four SCS design rain distributions (Type I, IA, II, and III, i.e., IREG = 1, 2, 3, and 4, respectively) were respectively parameterized with the same field data set. The results indicated that storm distribution parameters did not affect water runoff, and had almost no effect on soil erosion in winter and spring. In summer and autumn, however, there were apparent differences in sediment erosion between the four IREG values (Fig. 4). For IREG = 2, the prediction of erosion during a storm event was far lower than that of IREG = 1, 3, or 4. The estimated daily sediment erosion from storm events ranked as follows: IREG = 2 < IREG = 1 < IREG = 4 < IREG = 3. The predictions with IREG = 3 were closer to the observed than other IREG values.

View larger version (34K): [in this window] [in a new window]   Fig. 4. Observed vs. predicted event-based sediment erosion in 1991 and 1992 with different IREG (location of the Natural Resources Conservation Service 24-h hyetograph) values.

View larger version (22K): [in this window] [in a new window]   Fig. 5. Observed vs. predicted pesticide loss percentage via water runoff relative to their application rates in 1992. Bars are standard deviations.

View larger version (25K): [in this window] [in a new window]   Fig. 6. Predicted metolachlor concentrations in eroded sediment in 1991 and 1992 with different Kd values.

The model failed to calculate the influences of tillage management on pesticide concentration in water runoff. For instance, depending on the experiments, the pesticide concentrations in water runoff were much higher for CT than for MT, particularly in 1992 (Table 6). Nevertheless, in the simulation, the predicted herbicide concentrations in water runoff were not greatly affected by the tillage systems (Table 6). The reason for the simulated concentrations being similar in the two tillage regimes can be deduced from the following theoretical extrapolation of chemical concentration in water runoff:

[11]

where J_QR (g d^–1), Q (cm³ d^–1), C_w (g cm^–3), A_w (cm²), A (cm²), and DRI_i are defined as in Eq. [7]. In other words, the predicted chemical concentration in water runoff had a positive relationship with two parameters such as C_w and DRI_i, which were usually affected by soil sorption for a given A_w and A (see Eq. [6]). Because the herbicide sorption parameters including K_d, organic matter, and decay rate were set to the same values for both MT and CT in the simulation, no differences in herbicide concentrations of water runoff between CT and MT occurred from the model. As a consequence, the sorption, degradation, and organic matter parameters under different tillage practices need to be more exact in future research.

View larger version (25K): [in this window] [in a new window]   Fig. 7. Observed vs. predicted percentage of applied pesticides lost in eroded sediments in 1992. Bars are standard deviations.

The predicted chemical concentrations in eroded sediment of CT were usually lower than their counterparts in MT, though sometimes the differences between CT and MT were not significant. The cause was due to the theoretical extrapolation linking the sorbed phase concentration (g g^–1) to the dissolved-phase concentration by:

[12]

where K_d is the partition coefficient between the dissolved and solid phases (cm³ g^–1) and C_w (g cm^–3) as defined before. For this reason, the chemical concentration in eroded sediment (g Mg^–1) could be written as (see Eq. [8] and [9]):

[13]

in which J_ER (g d^–1), X_e (Mg d^–1), r_om (g g^–1), P_u (g Mg^–1), C_w (g cm^–3), A_w (cm²), and A (cm²) were defined by Eq. [8]. As all other parameters are constant for a given herbicide field experiment during a given rainfall event, chemical concentration in the eroded sediment will be determined by r_om (the enrichment rate of chemical). Thus, the chemical concentration had a negative exponential relationship with soil erosion intensity (X_e/A_w) (Mg cm^–2 d^–1) (see Eq. [9] and [13]), which is in contrast with the field measurements reported in this paper (Tables 4 and 7). As a result, the empirical equations employed in the model (Eq. [8] and [9]) should be improved for European environments before being applied to predict herbicide concentration in soil erosion.

The rainfall pattern had more significant effects on chemical loss via soil erosion than via runoff. The lowest loss via soil erosion was with IREG = 2 and the best fit to the observed mass loss was with IREG = 3 (Table 8).

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
There is a current need to simulate pesticide runoff and erosion under the condition of tillage management practices. In Europe, however, such a model has not yet been fully validated as field data sets are lacking. In this study, with a 2-yr field experimental data set, the model of PRZM 3.12 was validated to predict water runoff, sediment erosion, and associated transport of atrazine, terbuthylazine, and metolachlor under the effects of different tillage practices and cover crops. The results demonstrated that the model could qualitatively simulate significant differences of water runoff, soil erosion, and associated herbicide losses between conventional tillage (CT) and minimum tillage (MT) covering a barley cover crop. For MT, water runoff, soil erosion, herbicide losses in water runoff and eroded sediment, and proportion of herbicide loss via sediment erosion were significantly lower than for CT. The main medium of herbicide loss under both tillage practices was water runoff. The predictions with IREG = 3 were closer to the observed than other IREG values.

However, the distribution of losses between aqueous and sediment phases in each runoff event was not correctly simulated. The model usually underpredicted or failed to simulate water runoff events when rainfall intensity was high and daily precipitation volume was low, and overestimated water runoff in events with high daily rainfall volume and low rainfall intensity. Similarly, the model often overestimated sediment erosion events with high daily precipitation volume and low intensity, and underpredicted or failed to simulate soil erosion events where precipitation intensity was low and daily volume was high. This was due to the fact that the model considered runoff to be mainly caused by daily rainfall volume and rainfall distribution type, and soil erosion was simulated through the empirical Universal Soil Loss Equation rather than a physically based approach. In fact, in northern Italy, the rainfall pattern was characterized by occasional intense rainfall as well as by long and low-intensity rainfalls. Hence, an alternative approach is needed in the future to replace the SCS method because the SCS rainfall distributions were sometimes biased to reflect rainfall and storm intensity.

The model failed to differentiate the observed differences in pesticide concentrations in water runoff between MT and CT. This is thought to occur because the herbicide sorption, degradation parameters, and soil organic matter have been set to the same values for both CT and MT. The pesticide concentration in the eroded soil was not well simulated either, because the description of runoff and erosion processes is rather empirical in the model and not physically based. Moreover, model calculations do not adequately reflect the relationships between soil erosion intensity and chemical concentration in sediment losses, leading to discrepancies between predictions and field observations.

The results suggest that the model could be used for general management and registration purposes where estimates of chemical compound losses are not quantitatively required. In other words, the model has serious limitations for real-word surface water risk assessments where detailed event-specific runoff concentrations and losses are needed. If the model is to be used for such purposes, it requires modification to take into account the effects of rainfall intensity and the measured positive relationship between soil erosion and chemical loss concentration known for European environments.

	ACKNOWLEDGMENTS

 
The modeling work has been conducted as a part of "X-SAR SRTM Mission Italian Activities" (ARS-B.0/99-08) and "Protection of the surface water bodies by pesticide" (MIUR 2003). A. Vicari and F. Ventura, the owners of the data set, wish to thank SYNGENTA Ltd., Basel, Switzerland, for the financial support of the field experiment. The authors thank three anonymous referees for their helpful comments. This publication was financed by Università Cattolica del Sacro Cuore for the quality of the results obtained (esercizio 2004).

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 

• Capel, P.D., T.A. Winterstein, and S.J. Larson. 2001. Selected data from field studies of pesticide runoff to surface waters. National Water Quality Assessment Program, USGS Water Resour. Investigations Rep. 00-4284. USGS, Sacramento, CA.
• Carsel, R.F., J.C. Imhoff, P.R. Hummel, J.M. Cheplick, and A.S. Donigian, Jr. 1998. PRZM-3, A model for predicting pesticide and nitrogen fate in the crop root and unsaturated soil zones: Users manual for Release 3.0. Natl. Exposure Res. Lab., Office of Res. and Development, USEPA, Athens, GA.
• Cheng, H.H. (ed.) 1990. Pesticides in the soil environment: Processes, impacts, and modeling. SSSA Book Ser. 2. SSSA, Madison, WI.
• Donatelli, M., M. Acutis, and S. Kosovan. 1996. Tools for agrometeorology and agricultural modelling [Online]. Available at www.sipeaa.it/tools (verified 24 May 2004). ISA Modena-DASGT, Modena, Italy.
• Donatelli, M., G. Bellocchi, and F. Fontana. 2003. RadEst 3.00: Software to estimate daily radiation data from commonly available influence on the predicted fate of the pesticide. Nodal meteorological variables. Eur. J. Agron. 18:363–367.
• Eckersten, H., K. Blombäck, T. Kätterer, and P. Nyman. 2001. Modelling C, N, water and heat dynamics in winter wheat under climate change in southern Sweden. Agric. Ecosyst. Environ. 86:221–235.
• Elliott, J.A., A.J. Cessna, W. Nicholaichuk, and L.C. Tollefson. 2000. Leaching rates and preferential flow of selected herbicides through tilled and untilled soil. J. Environ. Qual. 29:1650–1656.[Abstract/Free Full Text]
• Felsot, A.S., J.K. Mitchell, and A.L. Kenimer. 1990. Assessment of management practices for reducing pesticide runoff from sloping cropland in Illinois. J. Environ. Qual. 19:539–545.[Abstract/Free Full Text]
• FOCUS. 2000. FOCUS groundwater scenarios in the EU review of active substances. Report of the FOCUS Groundwater Scenarios Workgroup. EC Document Reference Sanco/321/2000 Rev. 2. European Commission, Brussels.
• FOCUS. 2001. FOCUS surface water scenarios in the EU evaluation process under 91/414/EEC. Report of the FOCUS Working Group on Surface Water Scenarios, EC Document Reference SANCO/4802/2001 Rev. 2. European Commission, Brussels.
• Hall, J.K., R.O. Mumma, and D.W. Watts. 1991. Leaching and runoff losses of herbicides in a tilled and untilled field. Agric. Ecosyst. Environ. 37:303–314.
• Isensee, A.R., R.G. Nash, and C.S. Helling. 1990. Effect of conventional vs. no-tillage on pesticides leaching to shallow groundwater. J. Environ. Qual. 19:434–440.[Abstract/Free Full Text]
• Jury, W.A., W.R. Gardner, and W.H. Gardner. 1991. Soil physics. 5th ed. John Wiley & Sons, New York.
• Kum, S.W., K. Schaeffer, and T. Tapley. 1996. A dynamic programming approach to storm water management systems design. In Watershed 96, Proc. of the Water Environment Federation, Baltimore, MD [Online]. Available at www.epa.gov/OWOW/watershed/Proceed/wong.html (verified 26 Apr. 2004). USEPA, Washington, DC.
• Leonard, R.A., A. Shirmohammadi, A.W. Johnson, and L.R. Marti. 1988. Pesticide transport in shallow groundwater. Trans. ASAE 31:776–788.
• Loague, K., and R.E. Green. 1991. Statistical and graphical methods for evaluating solute transport models: Overview and application. J. Contam. Hydrol. 7:51–73.
• Miao, Z., M. Trevisan, E. Capri, L. Padovani, and A.A.M. Del Re. 2003. An uncertainty assessment of the RICEWQ model. p. 545–554. In A.A.M. Del Re, E. Capri, M. Trevisan, and L. Padovani (ed.) Pesticide in air, water & soil system. Proc. of the XII Symp. Pesticide Chem. La Goliardica Pavese, Pavia, Italy.
• Richards, R.P., F.G. Calhoun, and G. Matisoff. 2002. The Lake Erie agricultural systems for environmental quality project: An introduction. J. Environ. Qual. 31:6–16.[Abstract/Free Full Text]
• Rossi Pisa, P., F. Preti, M. Rossi, F. Ventura, and B. Mazzanti. 1999. Water, soil and chemical losses: Field experiments and model analysis. Water Sci. Technol. 39:93–102.
• Rossi Pisa, P., A. Vicari, and P. Catizone. 1994. Effetto della copertura vegetale di orzo sull'erosione e sulla qualità delle acque di ruscellamento. Riv. Agron. 28:384–391.
• Russell, L.J., and M.H. Russell. (ed.) 2001. FIFRA Environmental Model Validation Task Force: Final report. USEPA, Washington, DC.
• Russell, M.H. 1999. Predicting pesticide residues in surface water. p. 29–43. In A.A.M. Del Re et al. (ed.) Human and environmental exposure to xenobiotics. Proc. XI Symp. Pesticide Chem., Cremona, Italy. 11–15 Sept. 1999. La Goliardica Pavese, Pavia, Italy.
• Sanchez-Camazano, M., M.J. Sanchez-Martin, and T. Crisanto. 1995. Occurrence of atrazine in surface and ground waters in the province of Salamanca (Spain). Toxicol. Environ. Chem. 47:203–211.
• Tomlin, C. 1994. The pesticide manual incorporating the agro-chemicals handbook. 10th ed. British Crop Protection Council, The Royal Soc. of Chem., Farnham, UK.
• Vicari, A., P.R. Pisa, and P. Catizone. 1999. Tillage effects on runoff losses of atrazine, metolachlor, prosulfuron and triasulfuron. p. 140. In D. Gut (ed.) Proc. of the 11th EWRS Symp. 28 June–1 July. European Weed Res. Soc., Basel, Switzerland.
• Villholth, K.G., N.J. Jarvis, O.H. Jacobsen, and H. de Jonge. 2000. Field investigation and modeling of particle-facilitated pesticide transport. J. Environ. Qual. 29:1298–1309.[Abstract/Free Full Text]
• Warren-Hicks, W., J.P. Carbone, and P.L. Havens. 2002. Using Monte Carlo techniques to judge model prediction accuracy: Validation of the Pesticide Root Zone Model 3.12. Environ. Toxicol. Chem. 21:1570–1577.[Medline]
• Wauchope, R.D. 1978. The pesticide content of surface water draining from agricultural fields—A review. J. Environ. Qual. 7:459–472.[Abstract/Free Full Text]
• Wauchope, R.D., G.W. Randall, and L.R. Marti. 1990. Runoff of sulfometuron-methyl and cyanazine from small plots: Effects of formulation and grass cover. J. Environ. Qual. 19:119–125.

Related articles in JEQ:

This Issue in Journal of Environmental Quality

JEQ 2004 33: 1589-1599. [Full Text]  

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Related articles in JEQ Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via ISI Web of Science (8) Citing Articles via Google Scholar Google Scholar Articles by Miao, Z. Articles by Trevisan, M. Search for Related Content PubMed PubMed Citation Articles by Miao, Z. Articles by Trevisan, M. Agricola Articles by Miao, Z. Articles by Trevisan, M. Related Collections Organic Compounds