- Email: [email protected]
Evaluation of two pesticide leaching models in an irrigated field cropped with corn
Journal of Environmental Management 150 (2015) 508e515
Contents lists available at ScienceDirect
Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman
Evaluation of two pesticide leaching models in an irrigated field cropped with corn Dorothea D. Giannouli, Vassilis Z. Antonopoulos* School of Agriculture, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece
a r t i c l e i n f o
a b s t r a c t
Article history: Received 11 September 2014 Received in revised form 26 November 2014 Accepted 26 December 2014 Available online 2 January 2015
Pesticide leaching models is an easy and cost effective method used in the prediction of surface and groundwater pollution. In this paper, the ability of two pesticide leaching models, MACRO and PEARL, to describe soil water dynamics and atrazine's transport through the soil profile was examined. The data used for the comparison was obtained from an experiment in an irrigated corn field in the plain of the Ardas River, in north-eastern Greece. Both models were parameterized using pedotransfer functions, field and laboratory data. The uncalibrated simulation showed several discrepancies, therefore the retention curve and the sorption parameters were calibrated according to the trial and error method. The comparison of both models indicated that soil water flow was described similarly. The simulated results of atrazine's concentration were evaluated and compared to the measured concentrations at specific depths, using statistical criteria. Atrazine transport was simulated in a satisfactory manner as confirmed by model efficiency (EF) values, that are very close to unit. Coefficient of residual mass (CRM) values for both models are positive, indicating that both models underestimate the measured data. MACRO estimated higher accumulated actual evapotranspiration values, and less percolated water from soil profile than PEARL, and as a result, change in water content was higher in the latter. PEARL also predicted that half the amount of the applied mass was decayed two days earlier than the day estimated by MACRO. Generally, MACRO simulated the fate of atrazine in soil better than PEARL. © 2014 Elsevier Ltd. All rights reserved.
Keywords: Pesticide leaching models MACRO PEARL Field data Atrazine
1. Introduction Agriculture is one of the main sources of groundwater contamination, because of the extended use of pesticides and other agrochemicals. Several monitoring studies mention the presence of agrochemicals in surface and groundwater (Cooper, 1996; Reemtsma et al., 2013). Several studies confirm the presence of numerous pesticides in groundwater (Hallberg, 1989; Papastergiou and Papadopoulou-Mourkidou, 2001; Gonçalves et al., 2007). Apart from biotic, the extensive use of pesticides influences also the abiotic processes in soil (Zeng et al., 2012; Fenner et al., 2013). The European Union, in order to protect groundwater from pollution, promotes several directives such as the Water Framework Directive (2000/60/EC), the Groundwater Directive (2006/118/EC) and the recent Directive 2013/39/EU, regarding priority substances in the field of water policy. Besides, according to the Directive 98/83/EC, it is important that agrochemicals that reach groundwater do not
* Corresponding author. E-mail address: [email protected] (V.Z. Antonopoulos). http://dx.doi.org/10.1016/j.jenvman.2014.12.044 0301-4797/© 2014 Elsevier Ltd. All rights reserved.
exceed a concentration of 0.1 mg/L. In Greece, water pollution is caused mostly by the plant protection products used in corn, rice and cotton crops. According to Albanis et al. (1998), the maximum concentration of DEA and metolachlor detected in river water samples reached a value of 0.53 and 0.56 mg/L, respectively. Atrazine has been detected in significant concentrations (1.51 mg/L) in the groundwater of corn-growing areas of Northern Greece (Papastergiou and PapadopoulouMourkidou, 2001). High concentration levels of atrazine were also detected at the depth of 35 cm (1166 mg/L, Vryzas et al., 2012). In a recent study (Kalogridi et al., 2014), several pesticides and especially triazines, were detected in the surface waters and sediments of Lakes Kerkini, Doirani and Volvi, in Northern Greece. An efficient and low cost method for a preliminary assessment of groundwater vulnerability is the use of mathematical models. Various models like GLEAMS (Knisel and Davis, 2000), PRZM 3.21 (Carsel et al., 1998) and PEARL (Leistra et al., 2001) have been developed for the description of soil water movement and pesticide leaching. These models can simulate not only water flow and pesticides fate, but also most of the processes that compose the water
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515
and mass balance. Vanclooster et al. (2000) presented an assessment of twelve models simulating water movement and pesticide transport in soil, with the purpose of describing the problems encountered in using these models, and to introduce a Good Modelling Practice. In recent years though, there has been a strong tendency to incorporate soil water movement through macropores, commonly known as preferential flow, into leaching models. For this reason, many models that account for preferential flow, such as HYDRUS (Sim unek et al., 2005), RZWQM (Ahuja et al., 2000a,b) CRACK-NP (Armstrong et al., 2000b) and MACRO (Larsbo and Jarvis, 2003) have been developed. Despite the advantages a mathematical model can provide, when it comes to a comparison between two or more models, certain discrepancies may be observed. Armstrong et al. (2000a) and Persicani (1996) mention the variability in models' prediction. Although many studies concerning pesticide leaching in soil columns or lysimeters can be found in literature, only a few are based on data derived from field scale experiments (Flury et al., 1995; Flury, 1996; Vanclooster et al., 2000). The lack of measured values, especially at field scale, led to the use of stochastic param€hne et al., 2009). Among the limited number of compreeters (Ko hensive field-scale tests of pesticide transport models is the work of Suarez et al. (2013), who studied the transport of simazine in a vineyard using the HYDRUS 2D model. Other works with measured field data include the validation of the PEARL model using data from two sites in the Netherlands and Sweden (Bouraoui, 2007), the testing of the MACRO model in a cracked clay soil in the Netherlands by Scorza et al. (2007), and the impact of different irrigation practises on herbicide leaching by Fait et al. (2010). Recognizing the lack of site-specific data, this paper tries to contribute to the testing and the validation of two of the most commonly used in Europe models in the prediction of pesticide leaching. MACRO and PEARL models were used to simulate soil water dynamics and a solute's fate in a clayey soil, using the field data from an irrigated field cropped with corn in north-eastern Greece, under the climate and growing conditions of a Mediterranean area. 2. Method 2.1. Description of the MACRO and PEARL models MACRO 5.2 (Larsbo and Jarvis, 2003) and PEARL 4.4.4 (Leistra et al., 2001) are two representative pesticide leaching models, and were chosen in this study because of their credibility and ease of use (Jarvis, 1995; Boesten and van der Linden, 2001; Scorza and € hne et al., 2009; Fait et al., Boesten, 2005; Scorza et al., 2007; Ko 2010). FOCUS team (Boesten et al., 2000) through realistic worstcase scenarios, evaluated MACRO and PEARL models as a first-tier assessment of a pesticide leaching to groundwater. Since a full description of MACRO and PEARL is given elsewhere (Larsbo and Jarvis, 2003; Tiktak et al., 2000), only a brief summary of the models will be presented here. Both MACRO and PEARL are one-dimensional models which describe water flux, heat, and solute transport in soil matrix. Moreover, both account for a complete water balance including water flow, canopy interception and root water uptake, seepage to drains and groundwater. PEARL is linked with the SWAP model (Van Dam et al., 1997), whose soil hydrology is described by Richard's equation. MACRO is a dual permeability model which simulates preferential flow of water and solutes by dividing the soil matrix into the micropore and macropore domain, each one described by specific characteristics such as degree of saturation, conductivity and flux. Therefore, soil water flow in micropores is also described by Richards equation, and a kinematic wave
509
equation is used in the macropore domain. A convectionedispersion equation is used to describe the solute's transport, and in the macropore domain, MACRO uses only the convection equation. In both models, instantaneous equilibrium or kinetic sorption is described by either a linear or a Freundlich equation, and degradation by first order kinetics, depending on soil water content, temperature and depth. The degradation process in MACRO occurs in both domains. PEARL can also account for a pesticide's volatilization from soil or the plant's canopy, while MACRO does not account for this process. Crop growth in both models is described by a simple model where both the leaf area index (LAI) and the rooting depth are a function of the development stage of the crop. In MACRO, LAI and root depth follow a logistic curve, when in PEARL both are linear. Potential evapotranspiration is calculated using the equation of PenmaneMonteith (Allen et al., 1998) or can be pre-calculated and provided by the user, and can be partitioned into actual evaporation and transpiration through reduction functions. Finally, both models assume that runoff happens when infiltration capacity is exceeded. All these processes are also summarized in Table 1. Several studies are included in literature where PEARL and MACRO models were used to describe the water flow and the pesticide transport in soil (Scorza and Boesten, 2005; Scorza et al., 2007; Leistra and Boesten, 2010; Stenemo et al., 2007). In some cases, MACRO did not considered for preferential flow, and was used only for the one-domain approach (Antonopoulos et al., 2013). Because of the importance of the flow through the soil macropores, Tiktak et al. (2012a,b) presented two studies where PEARL, through macropore modules, could account for preferential flow (or a twodomain approach). This version of PEARL though, is unavailable to general public due to further testings.
3. Field and computational data 3.1. Site description and data sets The data used was derived from a 0.4 experimental field site in the Ardas River plain (41370 N, 26 210 E) in north-eastern Greece and was presented elsewhere in detail (Fragkoulis, 2003; Vryzas et al., 2007; Antonopoulos et al., 2011). Ardas River plain is one of the most important areas nationwide for corn production, and is characterized by warm summers and wet and cold winters, where Table 1 Description of the MACRO and PEARL processes. MACRO Water flow
Micropores: Richard's equation Macropores: Kinematic wave Solute Micropores: Convection dispersion transport equation Macropores: convection (gravity flow) Sorption Linear or Freundlich Instantaneous equilibrium and kinetic sorption for micro- and macropores Degradation First-order kinetics, separate rate coefficients for both solid and liquid phase of micro- and macropores Water Empirical sink terms uptake Volatilization No Initial Soil water content and temperature condition ETp PenmaneMonteith or input ETa Reduction functions Crop growth LAI as a Logistic function
PEARL Richard's equation Convection dispersion equation Linear or Freundlich, Instantaneous and non equilibrium First-order kinetics
A function of transpiration Soil and canopy Groundwater level PenmaneMonteith, Makkink, or input Reduction functions LAI as a Linear function
510
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515
the average annual rainfall and temperature are 771 mm and 12 C, respectively. The soil consisted of two layers and was characterized as loam (0e40 cm) and clay loam (40e120 cm). Generally, the soils of this area are clayey with low organic matter, and are characterized by recent alluvial deposits of the rivers Ardas and Evros. The depth to groundwater is about 6 m. Selected soil properties are given in Table 2. On April 7th 2000, the field was sown with corn and was harvested on September 19th 2000. Atrazine was applied to the soil surface on April 7th, a few hours before corn was sown, using an application rate of 1 kg/ha. In this field, no crop was sown and no herbicide has been applied before. During the growing season, the rainfall was 147.8 mm, the amount of water applied by sprinkler irrigation between June and August was 328 mm, and the reference evapotranspiration was calculated at 608.8 mm. Daily evapotranspiration values were calculated using data from a nearby meteorological station, and the PenmaneMonteith method. The maximum values of LAI and root depth were 4.5 and 1.4 m respectively. Agricultural practises and the parameters of the corn's development are summarized in Table 3. Atrazine leaching was measured under filed conditions at depths of 20 and 35 cm using suction cups (Fragkoulis, 2003; Vryzas et al., 2012). 3.2. Model parameterization The soil profile was divided into four horizons of 0e30, 30e40, 40e80 and 80e120 cm each. In order to provide adequate numerical accuracy, the soil profile in MACRO was partitioned into 60 layers ranging from 0.3 to 2.5 cm, while in PEARL discretization along the z axis ranged from 2.5 to 5 cm, separating soil profile in 30 layers. According to good modelling practice, and in order to simulate the initial water content in the beginning of the simulation, a six-month warming up period from October 1st 1999 to 31 March, 2000 was used. The simulation period lasted from April 1st to September 30th, 2000. Both models were parameterized with measured data, pedotransfer functions and literature values when certain data was unavailable. Initial conditions were set as a groundwater level of 600 cm for PEARL, and by inserting the initial values of water content and temperature for MACRO, while the bottom boundary condition was set as a free drainage for both models. Soil water content at field capacity qfc was set at a pressure head of 167 cm and 400 cm for the topsoil and subsoil, respectively, and at wilting Table 2 Selected soil physicochemical and hydraulic properties of the corn field. Characteristic
Sand (%) Silt (%) Clay (%) OM (%) pH rb (gr cm3) qs (m3 m3) qr (m3 m3) avg (cm1) nvg () Ks (m d1) hb (cm) hb (cm) 1D qb (m3 m3) Kb (m d1) n* () d (mm) d (mm) 1D
Depth (cm) 0e30
30e40
40e80
80e120
48 35 17 0.8 6.3 1.550 0.36 0.01 0.0143 1.44 0.120 10 0 0.351 0.053 4 6 1
43 29 28 0.7 6.8 1.565 0.39 0.04 0.0156 1.39 0.661 10 0 0.380 0.031 4 6 1
40 30 30 0.4 7.3 1.621 0.38 0.04 0.0160 1.32 0.646 10 0 0.369 0.026 4 6 1
43 27 30 0.3 7.7 1.644 0.37 0.04 0.0180 1.29 0.042 10 0 0.310 0.029 4 6 1
Table 3 Agricultural practices and corn growth parameters. Date
Treatment
7 Apr. 2000 14 Apr. 2000 15 June 2000 30 Jun. 2000 16 Aug. 2000 19 Sept. 2000
Atrazine application þ seeding
Crop stage LAI Root Root (m2 m2) depth (m) distribution
Emergence 0.01
0.06
60%
Flowering 4.50
1.4
60%
Harvest
1.4
Start of irrigation
End of irrigation 4.50
point qpwp, was set at 15000 cm. The parameters related to van Genuchten's soil water retention curves and hydraulic conductivity were determined for each layer through soil texture and bulk density, using Rosetta Lite pedotransfer functions, as proposed by Schaap et al. (2001). For MACRO model, the parameters demarcating the two domains such as the boundary water content qb and hydraulic conductivity Kb were defined by using the incorporated pedotransfer functions (PTF) of MACRO 5.2 GUI and algorithms proposed by € sten et al. (1999) and Moeys et al. (2012). The boundary pressure Wo head hb was set at 10 cm, according to the range given by Jarvis and Larsbo (2012) and then, the boundary water content qb was estimated from the water retention curves by setting h ¼ hb. The boundary hydraulic conductivity Kb for every horizon was estimated using the MACRO 5.0/5.1 pedotransfer functions of MACRO 5.2 GUI. The same PTFs were used for the estimation of the effective diffusion pathlength d which was set to 6 mm. Tortuosity/pore size distribution coefficient n* was set to 4, the maximum default value according to Beulke et al. (2002) and Antonopoulos et al. (2013). Crop growth parameters were based on default values provided by the models or by observations. The growth stages for the corn crop were 30/50/20/50 days. In PEARL, crop factor as a function of development stage was set at 0.89 for the 30th of June, where LAI reached the maximum value of 4.5. For the estimation of potential evapotranspiration in MACRO, the crop coefficient values adjusted to local conditions were kcinit ¼ 0.4, kcmid ¼ 1.1 and kcend ¼ 0.3. Physicochemical and sorption properties for atrazine such as sorption coefficient Koc and Freundlich exponent n came from literature (Weed Science of America, 1994). Wauchope et al. (1992) ranged the Koc value of atrazine between 38 and 174 cm3 g1, and the half-life from 18 to 120 days. The half-life (DT50) value of atrazine at 20 C, and as a consequence the degradation rate coefficient mref, derived from laboratory experiments (Fragkoulis, 2003) and was set at 12.6 days and 0.055 d1, respectively. The aforementioned sorption and degradation parameters were considered only for topsoil. Depth has a significant role in transformation rates, since degradation decreases with depth. Therefore, a degradation factor is used by both models to correct each layer's transformation rate, equal to 0.5 for a subsoil layer of 30e60 cm, to 0.3 for a subsequent layer of 60e100 cm, or equal to zero for layers below 100 cm (Boesten et al., 2000). It is assumed that the applied substance is mixing with the soil water content that is stored in a 1 mm thick soil layer. No interception occurred since a preemergent herbicide was applied. Furthermore, although both models can deal with non-equilibrium sorption, in this study kinetic sorption was not assumed, according to first-tier leaching assessment (Boesten et al., 2000). A first simulation using both models was executed without any
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515
calibration, in order to test the description of water flow and atrazine's fate in soil. The initial results showed that certain parameters had to be calibrated, because of the differences between measured and calculated data. Van Genuchten's parameters a and n, affecting the soil water retention curves, and the two parameters affecting sorption, Koc and the Freundlich exponent n, were calibrated according to trial-and-error method, until the model results could give a good prediction of the measured data. It occurred that PEARL simulated better the soil water content when a lysimeter as a lower boundary condition was used. To calibrate the hydraulic parameters, the values obtained by Rosetta's PTF's were compared to those derived by using Vereecken pedotransfer functions (Vereecken et al., 1989), and the most appropriate values for the simulation are presented in Table 2. Then the sorption constant Koc and the Freundlich exponent n were adjusted to the observed concentration values at specific depths. Atrazine properties used for the simulation are given in Table 4. The Koc and n values obtained after the calibration were 70 L kg1 and 0.88, respectively. The sorption distribution coefficient Kd was calculated from the measured organic carbon content in the soil horizons using the equation:
Kd ¼ Koc OC and was set at 0.32 cm3 gr1. Weber et al. (2004) correlated Kd values with soil properties for selected pesticides, and suggested 2.65 cm3 gr1 as a mean value for atrazine. The degradation rate coefficient mref in different depths was calculated using corrective factors (Boesten et al., 2000). In MACRO model, degradation rates were assumed to be the same in both phases and domains of every horizon. The matrix dispersivity was set at 5 cm, according to FOCUS scenarios (Boesten et al., 2000), and the diffusion coefficient in water was set at the default value of 4.9 1010 m s1. The mass fraction f was set at 0.02 as proposed by Dubus et al (2003) and Jarvis and Larsbo (2012). 3.3. Model evaluation The performance of both models in simulating the measured substance distribution was evaluated using three statistical criteria in specific depths, as proposed by Loague and Green (1991) and Antonopoulos (2000), the modelling efficiency EF, the relative root mean square error RMSE and the coefficient of residual mass CRM.
EF ¼
n n 2 X X Oi O ðPi Oi Þ2 i¼1
RMSE ¼
i¼1 n X
!,
n X
Oi O
2
(1)
i¼1
! . 1=2 ðPi Oi Þ n 2
(2)
i¼1
Table 4 Selected properties of atrazine. Properties
Atrazine
Molar mass (g) Vapour pressure (Pa) Solubility in water (mg L1) Molar Enthalpy (KJ mol1) Sorption distribution coefficient kd (cm3/g) Sorption constant Koc (cm3 g1) Freundlich exponent n Degradation rate coefficient mref (d1) Half-life at 20 C DT50 (d)
215 3.9 104 (25 C) 33 27 0.32 70 0.88 0.055 12.6
CRM ¼
n X
Oi
i¼1
n X
!, Pi
i¼1
n X
511
Oi
(3)
i¼1
where Pi and Oi are the simulated and measured values of atrazine respectively, O is the mean of measured values and n is the number of measured values. The optimum values of EF, RMSE and CRM criteria are 1, 0 and 0, respectively. EF value compares the predicted values to the average value of the measurements. If EF is less than zero, the model predicted values are worse than simply using the observed mean. Positive values of CRM indicate that the model underestimates the measurements.
4. Results and discussion 4.1. Water content The simulated results of water content, using the MACRO and PEARL models, are presented in Fig. 1 at the depths of 20, 60 and 110 cm, respectively, and the components of the water balance are shown in Table 5. During the simulation period (from the 92nd to 274th day of the year), a total of 502.8 mm of rain and irrigation water was applied. Surface runoff was not observed, since the experimental field was flat. Predicted values from both models followed a similar pattern, and both can simulate interactions between soil, plant and atmosphere in a reasonable manner. As shown in Fig. 1, during the study period, the simulated soil water dynamics was similar for both models at the three different depths. At the depth of 20 cm, and during the irrigation period (June to August), MACRO estimated higher evapotranspiration values, and as a consequence soil water content was lower than the one estimated by PEARL, causing a discrepancy. The same inconsistency is also observed at the depths of 60 cm and 110 cm, for a shorter period. This discrepancy can be explained by the different approach of both models to the estimation of dynamic crop properties, especially when it comes to the separation of potential evapotranspiration into soil evaporation and transpiration from the plant. Although the same daily potential evapotranspiration values were used for the estimation of actual evapotranspiration, the predictions from both models were different. As mentioned before, LAI is modelled as a logistic curve in MACRO and a linear one in PEARL, resulting in a different estimation of evaporation and transpiration (Table 5) and, as a consequence, different water content. Using the MACRO model, the cumulative actual evaporation from the soil and the transpiration from the plants were calculated at 73.2 mm and 368.9 mm respectively while in PEARL were 116 mm and 283.5 mm, respectively. During the simulation period, the MACRO approach predicted that 61.7 mm of water was percolated from soil profile, whereas PEARL predicted 132.2 mm. Due to different values of calculated actual evapotranspiration and deep percolation, the change in water content during the whole simulation period was 3.8 mm in MACRO approach, whereas in PEARL was 19.9 mm. In a parallel simulation where MACRO considered only for flow in the micropore domain (results not shown here), higher water content values were predicted, especially after rain or irrigation events in every layer. The change in water content was decreased only by 5.6 mm, due to the change in percolated water. The same difference was also noticed by Antonopoulos et al. (2013) after comparing one and two-domain approaches using MACRO in an irrigated corn field. Garratt et al. (2002) also mentioned interesting differences in water balance when comparing seven models with different methods of estimating soil water flow and evapotranspiration. In
512
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515
100 Irrigation + Ra in
MACRO
PEARL
20 cm
0,4
80
0,3
60
0,2
40
0,1
20 0
0,0 92
122
152
182 Day of year
212
242
272
100 Irrigation + Rain
MACRO
60 cm
PEARL
0,4
80
0,3
60
0,2
40
0,1
20
0,0
irrigation + rain, mm
soil water content, m3 m-3
0,5
0 92
122
152
182 Day of year
212
242
272
100
0,5 Irrigation + Rain
MACRO
110 cm
PEARL
0,4
80
0,3
60
0,2
40
0,1
20
irrigation + rain, mm
soil water content, m3 m-3
irrigation+ rain, mm
soil water content, m3 m-3
0,5
0
0,0 92
122
152
182 Day of year
212
242
272
Fig. 1. Comparison of simulated water content at 20, 60 and 110 cm using MACRO and PEARL model.
MACRO, potential evapotranspiration is used as a direct input, whereas in this simulation PEARL uses reference evapotranspiration values and then the potential evapotranspiration is calculated internally by the model. Different methods of evapotranspiration can cause significant changes the ratio of actual and potential transpiration and eventually in water content (Jarvis et al., 2000). Although no measured data was available, the two models displayed similarly the soil water simulation. Jarvis et al. (2000) using pedotransfer functions in a MACRO model simulation, noted a value of RMSE ¼ 0.06 m3 m3 between measured and simulated water content, while Antonopoulos (2000) noted that the RMSE using direct and pedotransfer function for parameters of van Genuchten retention curves was 0.038 and 0.037 cm3 cm3, respectively, between the measured and simulated values of water content. 4.2. Atrazine behaviour The results of atrazine simulation at depths of 20 cm and 35 cm are shown in Fig. 2. Atrazine concentrations simulated with the MACRO and PEARL models showed similar patterns versus time at both depths. There is a good agreement between modelling results and measured data, even though the predicted concentrations from both models are higher in the upper layer and lower on subsoil.
Atrazine remained in significant concentrations on the topsoil for more than a two-month period after application. According to the MACRO simulation, the applied atrazine appeared at 20 cm depth on the 98th day of the year and reached the underlying layer two weeks later, while at the end of the simulation, the concentration level of atrazine at 35 cm depth was 0.131 mg/m3. Meanwhile, the PEARL simulation also predicted the atrazine's appearance on the topsoil two weeks after the application day, while on the underlying layer, atrazine appeared on 122th day of year, eight days later than the day predicted by MACRO. Atrazine concentration at the end of the simulation period was 0.059 mg/m3 at the 35 cm depth. At the same depth, and during the first 55 days, the predicted by MACRO values were higher than those predicted by PEARL. These values, combined with the earlier atrazine appearance in this layer could be explained by preferential displacement, since PEARL model does not account for a solute's behaviour through macropores. According to PEARL approach, atrazine remained on the topsoil in higher concentrations (the maximum value was 177 mg/m3 on the 124th day of year) while on the subsoil, the predicted values were lower. On the contrary, the simulated by MACRO concentration levels were lower on the first layer (the maximum value was 137.9 mg/m3 on the 120th day of year), and higher on the subsoil (the maximum value was 6.5 mg/
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515 Table 5 Water balance for MACRO and PEARL models from 92 to 274 day of year 2000. MACRO (mm)
PEARL (mm)
Cumulative rain and irrigation Cumulative actual evaporation Cumulative actual transpiration Cumulative runoff Cumulative percolation Cumulative canopy evaporation Change in water content
502.9 73.2 368.9 0 61.7 2.8 3.8
502.9 116 283.5 0 123.2 e 19.9
total solute concentration, mg m-3
Components of soil water balance
20 cm 150
measured MACRO
100
PEARL 50 0
total solute concentration, mg m-3
92
122
152
182 Day of year
212
242
272
8
35 cm 6
Measured MACRO
4
PEARL 2 0 92
122
152
182
212
242
Table 6 Statistical criteria between measured and calculated values. Depth (cm)
20 35 Average
200
272
Day of year Fig. 2. Measured and simulated atrazine concentrations at 20 and 35 cm using MACRO and PEARL model.
m3 on the 128th day of year). On the180th day, when the simulated by PEARL concentration reached the maximum value of 6.3 mg/m3, the value predicted by MACRO was 4.9 mg/m3. On the topsoil, PEARL model predicted better the concentration, after the irrigation event on the 168th day of year. The irrigation event on the 192th day of the year was critical, because it led to an increase of the simulated by MACRO atrazine concentrations at 35 cm depth. According to Fait et al. (2010), the type of irrigation is important for a pesticide's leaching. Statistical criteria give a more detailed picture of both models' performance, and are summarized in Table 6. Generally, both models predict atrazine distribution at two depths acceptably well, judging from average EF values that are close to 1. Average RMSE values are lower for MACRO than for PEARL, and average CRM are close to 0. More specifically, at the depth of 20 cm, CRM values indicate that both models overestimate the measured data, while at the depth of 35 cm there is a tendency to underestimate concentration values. The combination of MACRO's RMSE and EF values indicate the latter as a better model in the description of a solute's behaviour. CRM and EF values derived from a field study using a similar Koc value for atrazine (72 mg L1) in a clay soil, ranged from 1.19 to 0.62 and from 3.07 to 0.57, respectively (Sarmah et al., 2005). The estimated by both models mass balance for atrazine as a function of time is shown in Fig. 3. In MACRO simulation, almost 99% of the applied mass was stored in the liquid and solid phase of the soil on the application day, 50% was stored on the 121th day of
513
EF
RMSE
CRM
MACRO
PEARL
MACRO
PEARL
MACRO
PEARL
0.876 0.996 0.936
0.584 0.994 0.789
10.667 1.421 6.044
19.514 1.707 10.610
0.126 0.350 0.112
0.332 0.392 0.030
the year, 25% was stored on the 140th day of the year, and 10% was stored on the 160th day of the year (62 days after application). A very small amount (0.024 mg m2) was also stored in soil macropores. Moreover, 50% of the applied mass was degradated on April 30th (121 day of the year), 23 days after application. No degradation from canopy, solute uptake by crop or leaching to groundwater were observed during the simulation period. At the end of the simulation, 0.072 and 0.001 mg m2 were stored in the liquid and solid phase of micro- and macropore domain, respectively. According to PEARL simulation, 99% of the applied mass was also stored in soil profile on the application day, 50% and 25% of the applied mass were stored two days earlier than the MACRO prediction, and 10% was stored on the 155th day of the year (57 days after application). Until the 196th day of year (98 days after application) degradation occurred faster, since 50% of the applied mass was transformed on 119th day of year, two days earlier than the day estimated by MACRO, and as a consequence, PEARL predicted lower stored mass and higher degradation levels. The latter also predicted 0.81 mg m2 of atrazine uptake by crop, 0.015 mg m2 of atrazine volatilization and 0.017 mg m2 of stored mass at the end of the simulation. Atrazine leaching to groundwater was also not observed. In Table 7, a more detailed mass balance for the 99th, 216th and 274th day of the year is presented for MACRO and PEARL models, respectively. These days represent the days after application, after the maximum irrigation event and the last day of simulation, respectively. In other field experiments with corn crops, 15% of the initially applied atrazine remained after one year of application, at 1 m depth (Tasli et al., 1996), while atrazine began to transform into DEA on topsoil 15 days after application (Martínez et al., 1996). Sarmah et al. (2009) reported that half-lives of atrazine at the topsoil and the subsoil of a silt loam soil were 18 days and 47 days, respectively. Scorza and Boesten (2005) using the PEARL model, tried to simulate the movement of water, and the transport of bentazone and imidacloprid in a cracked clay soil. Although measured and simulated values of water and imidacloprid were in a good agreement, PEARL did not predict the bentazone concentrations well because of their preferential transport to soil macropores. The PEARL model also estimated higher concentrations and persistence of bentazone in soil profile (Bouraoui, 2007). Armstrong et al. (2000a), after testing four different models in a macroporous clay soil in the UK, indicated MACRO as a better model in the estimation of isoproturon concentrations in soil. Larsson and Jarvis (2000) concluded that when using the MACRO model without macropore flow, no leaching of the applied bromide and bentazone was observed. On the other hand, MACRO was able to simulate well measured and simulated values of an early bentazone breakthrough in drainage water, indicating flow through macropores (Scorza et al., 2007). As a result of atrazine bypassing the matrix soil, Flury et al. (1995) indicated that 40 mg kg1 of atrazine were detected in soil cracks and at a depth of 80e90 cm of a loamy soil. A literature review in sensitivity analysis for both MACRO and PEARL models, indicated the sensitivity of the related to sorption and degradation parameters (Armstrong et al., 2000a; Dubus et al.,
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515
100
1,0
80
0,8
60
0,6
Tota l solute stored in profile, MACRO Tota l solute stored in profile, PEARL Degra da tion in soil, MACRO Degra da tion in soil, PEARL Solute uptake by plants, MACRO Solute uptake by plants, PEARL Solute vola tiliza tion, PEARL
40
20
0,4
0,2
0
0,0 92
122
152
182 Day of year
212
242
Solute uptake and volatilization, mg m-2
Total mass and degradation, mg m-2
514
272
Fig. 3. Mass balance for atrazine in soil profile (120 cm) using MACRO and PEARL models.
2003; Tiktak, 2012a,b). Dubus et al. (2003) after performing a sensitivity analysis in four mathematical models including PEARL and MACRO, concluded that parameters affecting sorption and degradation such as degradation rates and Freundlich exponent n as well as sorption distribution coefficient Koc were the most sensitive, while the water balance was slightly affected by variation in input parameters. Scorza et al. (2007) also indicated the mass fraction of solid material contacting water in macropores as a sensitive parameter, and Armstrong et al. (2000a) the significance of the mixing depth and the user's subjectivity. 5. Conclusion In this paper, the dual-permeability model MACRO was compared to PEARL model, to simulate the fate of atrazine in a clayey soil in Ardas River plain. According to statistical criteria, both models provided a very good simulation of the mass transport of atrazine, although both of them underestimate atrazine concentrations in the subsoil. The results of the MACRO approach were better in average values than the simulation using PEARL, indicating the former as a better model. The existence of preferential water flow may be a cause for the discrepancy in the description of atrazine levels. Discrepancies between measured and predicted values were also caused by differences in the parameterization or the different approach of both models in the description of certain processes. The initial condition was described differently, and because in PEARL model, the initial water content was not an input value, a larger warm up period was needed to simulate better a more realistic simulation. Furthermore, PEARL model predicted similar results with MACRO when instead of free drainage, a lysimeter was used as a lower boundary condition. The choice of the initial and
Table 7 Change in mass balance for atrazine using MACRO and PEARL models during the simulation period. Mass (mg m2) Day of year 99 MACRO Stored in soil Degradation Crop uptake Volatilization Total mass a
216 PEARL
MACRO
274 PEARL
MACRO
PEARL
95.410 0.517a 0.185 0.073a 0.017 97.259a 2.741 4.464 99.482 99.000 99.927 99.160 0.000 0.000 0.000 0.809 0.000 0.811 e 0.002 e 0.015 e 0.0155 100 100 100 100 100 100
Micropores and macropores.
lower boundary condition is of great importance. Moreover, differences in the estimation of the actual evapotranspiration, probably because of the crop model used in both models, led to a discrepancy in soil water dynamics, indicating a wetter soil when using MACRO. When data is not sufficient, pedotransfer functions or data derived from literature is a useful way to estimate the missing parameters. Finally, both MACRO and PEARL models were developed and evaluated in countries where rainfall, and not always irrigation, is the main inflow in water balance. In this study, both models were applied using data derived from a corn field, emphasizing the soil water dynamics and atrazine behaviour during the irrigation period. It would be really interesting to extend this study by comparing MACRO model with a version of PEARL model, where the preferential flow would be considered. In conclusion, even if the same scenario is used, the predictions of mathematical models for a pesticide's fate may vary. Nevertheless, these models can be used for many environmental scenarios for the achievement of sustainable and rational water and agrochemical management. References Ahuja, L.R., Rojas, K.W., Hanson, J.D., Shaffer, J.J., Ma, L., 2000a. The Root Zone Water Quality Model. Water Resources Publications LLC, Highlands Ranch, CO. Ahuja, L.R., Johnsen, K.E., Rojas, K.W., 2000b. Water and chemical transport in soil matrix and macropores. In: Ahuja, L.R., Rojas, K.W., Hanson, J.D., Shaffer, J.J., Ma, L. (Eds.), The Root Zone Water Quality Model. Water Resources Publications LLC, Highlands Ranch, CO. Albanis, T.A., Hela, D.G., Sakellarides, J.M., Konstantinou, I.K., 1998. Pesticide residues in surface water, groundwaters and rainfall of Imathia (Greece). In: Proc. of Environmental Protection and Restoration, Sani Chalkidiki, pp. 119e126. Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop Evapotranspirationguidelines for Computing Crop Water Requirements. FAO Irrigation and Drainage Paper 56, Rome, Italy. Antonopoulos, V.Z., 2000. Modeling of soil water dynamics in an irrigated corn field using direct and pedotransfer functions for hydraulic properties. J. Irrig. Drain. Syst. 14, 325e342. Antonopoulos, V.Z., Fragkoulis, G., Papadopoulou- Mourkidou, E., Vryzas, Z., 2011. Assessment of pesticide leaching in a corn field using PEARL 1.1.1. model. In: Proceedings of the Council of Agricultural Engineers of Greece, Athens, Greece, p. 8 (in Greek). Antonopoulos, V.Z., Georgiou, P.E., Kolotouros, C.A., 2013. Soil water dynamics in cropped and uncropped fields in northern Greece using a dual-permeability model. Hydrol. Sci. J. 58 (8), 1748e1759. Armstrong, A., Aden, K., Amraoui, N., Diekkruger, B., Jarvis, N.J., Mouvet, C., Nicholls, P., Wittwer, C., 2000a. Comparison of the performance of the pesticide-leaching models on a cracking clay soil: results using the Brimstone Farm dataset. Agric. Water Manag. 44, 85e104. Armstrong, A.C., Matthews, A.M., Portwood, P.B., Leeds-Harrison, P.B., Jarvis, N.J., 2000b. CRACK-NP: a pesticide leaching model for cracking clay soils. Agric. Water Manag. 44, 183e199. Beulke, S., Renaud, F., Brown, C.D., 2002. Development of Guidance on Parameter
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515 Estimation for the Preferential Flow Model MACRO 4.2. Final Report of the DEFRA project PL0538. Cranfield Centre for EcoChemistry, UK, p. 68. University of Cranfield. Boesten, J.J.T.I., van der Linden, A.M.A., 2001. Effect of long-term sorption kinetics on leaching as calculated with the PEARL model for FOCUS scenarios. In: Pesticide Behaviour in Soils and Water. Proceedings of the British Crop Protection Council Symposium, Brighton, UK, 13-15 November 2001, pp. 27e32. Boesten, J., Businelli, M., Delmas, A., Gottesbüren, B., Hanze, K., Jarvis, T., Jones, R., Klein, M., van der Linden, T., Rekolainen, S., Resseler, H., Roquero, C., Maier, W.M., Styczen, M., Thorsen, M., Travis, K., Vanclooster, M., 2000. FOCUS Groundwater Scenarios in the EU Review of Active Substances Sanco/321/2000. Version 1. Bouraoui, F., 2007. Testing the PEARL model in the Netherlands and Sweden. Environ. Model. Softw. 22, 937e950. Carsel, R.F., Imhoff, J.C., Hummel, P.R., Cheplick, J.M., Donigian Jr., A.S., 1998. PRZM3: a Model for Predicting Pesticide and Nitrogen Fate in the Crop Root and Unsaturated Soil Zones: User's Manual for Release 3.12. National Exposure Research Laboratory. Office of Research and Development, US Environmental Protection Agency, Athens, GA. Cooper, B., 1996. Central and NorthWest RegionsWater Quality Program.1995/1996 Report on Pesticide Monitoring. TS96.048. NSW Department of Land & Water Conservation, Sidney, Australia. Dubus, I.G., Brown, C.D., Beulke, S., 2003. Sensitivity analyses for four pesticide leaching models. Pestic. Manag. Sci. 59, 962e982. Fait, G., Balderacchi, M., Ferrari, F., Ungaro, F., Capri, E., Trevisan, M., 2010. A field study of the impact of different irrigation practices on herbicide leaching. Eur. J. Agron. 32, 280e287. Fenner, K., Canonica, S., Wackett, L., Elsner, M., 2013. Evaluating pesticide degradation in the environment: blind spots and emerging opportunities. Science 341 (6147), 752e758. Flury, 1996. Experimental evidence of transport of pesticides through field soils- a review. J. Environ. Qual. 25, 25e45. Flury, M., Leuenberger, J., Studer, B., Fluhler, H., 1995. Transport of anions and herbicides in a loamy and a sandy field soil. Water Resour. Res. 31, 823e835. Fragkoulis, G., 2003. Evaluation of Leaching and Groundwater Pollution of Plant Protection Products Using PEARL Model in a Corn Plot in Evros. Master thesis. School of Agriculture, Aristotle University, Thessaloniki (in Greek). Garratt, J.A., Capri, E., Trevisan, M., Errera, G., Wilkins, R.M., 2002. Parameterisation, evaluation and comparison of pesticide leaching models to data from a Bologna field site, Italy. Pest Manag. Sci. 58, 3e20. Gonçalves, C.M., Da Silva, J.C.G.E., Alpendurada, M.F., 2007. Evaluation of the pesticide contamination of groundwater sampled over two years from a vulnerable zone in Portugal. J. Agric. Food Chem. 55 (15), 6227e6235. Hallberg, G.R., 1989. Pesticide pollution of groundwater in the humid Unites States. Agric. Ecosyst. Environ. 26, 299e367. Jarvis, N.J., 1995. Simulation of soil water dynamics and herbicide persistence in a silt loam soil using the MACRO model. Ecol. Model. 97e109. Jarvis, N., Larsbo, M., 2012. MACRO (v5.2): model use, calibration and validation. Trans. Am. Soc. Agric. Biol. Eng. 55 (4), 1413e1423. Jarvis, N.J., Brown, C.D., Granutza, E., 2000. Sources of error in model predictions of pesticide leaching: a case study using the MACRO model. Agric. Water Manag. 44, 247e262. Kalogridi, E.C., Christophoridis, C., Bizani, E., Drimaropoulou, G., Fytianos, K., 2014. Part I: temporal and spatial distribution of multiclass pesticide residues in lake waters of Northern Greece: application of an optimized SPE-UPLC-MS/MS pretreatment and analytical method. Environ. Sci. Pollut. Res. 21, 7239e7251. Knisel, W.G., Davis, F.M., 2000. GLEAMS, Groundwater Loading Effects from Agricultural Management Systems V3.0. Publ. No. SEWRL-WGK/FMD-050199. U.S.D.A., Tifton, Georgia. €hne, J.M., Ko €hne, S., Sim Ko unek, J., 2009. A review of model applications for structured soils:b) pesticide transport. J. Contam. Hydrol. 104, 36e60. Larsbo, M., Jarvis, N.J., 2003. MACRO 5.0. A Model of Water Flow and Solute Transport in Macroporous Soil. Technical Description. Department of Soil Sciences, Swedish University of Agricultural Sciences, Uppsala. Larsson, M.H., Jarvis, N.J., 2000. Quantifying interactions between compound properties and macropore flow effects on pesticide leaching. Pest Manag. Sci. 56, 133e141. Leistra, M., Boesten, J.J.T.I., 2010. Pesticide leaching from agricultural fields with ridges and furrows. Water Air Soil Pollut. 213, 341e352. Leistra, M., van der Linden, A.M.A., Boesten, J.J.T.I., Tiktak, A., van den Berg, F., 2001. PEARL Model for Pesticide Behaviour and Emissions in Soil-plant Systems. Description of Processes. Alterra report 13, RIVM report 711401009. Alterra, Wageningen, The Netherlands. Loague, K.M., Green, R.E., 1991. Statistical and graphical methods for evaluating solute transport models: overview and application. J. Contam. Hydrol. 7, 51e73. n, Y., S Martínez, L., Lecho anchez-Brunete, C., Tadeo, J.L., 1996. Persistence of metolachlor, atrazine and deethylatrazine in soil and its simulation by deterministic models. Toxicol. Environ. Chem. 54, 219e232. Moeys, J., Larsbo, M., Bergstrom, L., Brown, C.D., Coquet, Y., Jarvis, N.J., 2012. Functional test of pedotransfer functions to predict water flow and solute transport with the dual-permeability model MACRO. Hydrol. Earth Syst. Sci. 16,
515
2069e2083. Papastergiou, A., Papadopoulou-Mourkidou, E., 2001. Occurrence and spatial and temporal distribution of pesticide residues in groundwater of major corngrowing areas of Greece. Environ. Sci. Technol. 35, 63e69. Persicani, D., 1996. Pesticide leaching into field soils: sensitivity analysis of four mathematical models. Ecol. Model. 84, 265e280. Reemtsma, T., Alder, L., Banasiak, U., 2013. Emerging pesticide metabolites in groundwater and surface water as determined by the application of a multimethod for 150 pesticide metabolites. Water Res. 47 (15), 5535e5545. Sarmah, A., Close, M., Pang, L., Lee, R., Green, S.R., 2005. Field study of pesticide leaching in an Himatangi sand (Manawatu) and a Kiripaka bouldery clay loam (Norhtland). 2. Simulation using LEACHM, HYDRUS-1D, GLEAMS and SPASMO models. Aust. J. Soil Res. 43, 471e489. Sarmah, A.K., Close, M.E., Mason, N.W.H., 2009. Dissipation and sorption of six commonly used pesticides in two contrasting soils of New Zealand. J. Environ. Sci. Health Part B 44, 325e336. Schaap, M.G., Leij, F.J., van Genuchten, M.T., 2001. ROSETTA: a computer program for estimating soil hydraulic parameters with hierarchical pedotransfer functions. J. Hydrol. 251 (3e4), 163e176. Scorza, P.J., Boesten, J.J.T.I., 2005. Simulation of pesticide leaching in a cracking clay soil with the PEARL model. Pest Manag. Sci. 61432e61448. Scorza, R.P.J., Jarvis, N.J., Boesten, J.J.T.I., van der Zee, S.E.A.T.M., Roulier, S., 2007. Testing MACRO (version 5.1) for pesticide leaching in a Dutch clay soil. Pest Manag. Sci. 63, 1011e1025. Sim unek, J., van Genuchten, MTh, Sejna, M., 2005. The HYDRUS-1D Software Package for Simulating the One-dimensional Movement of Water, Heat, and Multiple Solutes in Variably-saturated Media. Version 3.0. In: HYDRUS Software Series, vol. 1. Department of Environmental Sciences, University of California Riverside, Riverside, CA, p. 270. Stenemo, F., Lindahl, A.M.L., G€ arden€ as, A., Jarvis, N., 2007. Meta-modeling of the pesticide fate model MACRO for groundwater exposure assessments using artificial neural networks. J. Contam. Hydrol. 93, 270e283. Suarez, F., Guzman, E., Munoz, J., Bachmann, J., Ortiz, C., Alister, C., Kogan, M., 2013. Simazine transport in undisturbed soils from a vineyard at the Casablanca valley, Chile. J. Environ. Manag. 117, 32e41. Tasli, S., Patty, L., Boetti, H., Ravanel, P., Vachaud, G., Scharff, C., Favre-Bonvin, J., Kaouadji, M., Tissut, M., 1996. Persistence and leaching of atrazine in corn culture in the experimental site of La Cote Saint Andre (Isere, France). Arch. Environ. Contam. Toxicol. 30, 203e212. Tiktak, A., van den Berg, F., Boesten, J.J.T.I., van Kraalingen, D., Leistra, M., van der Linden, A.M.A., 2000. Manual of FOCUS PEARL Version 1.1.1. Alterra report 28, RIVM report 711401008. Alterra, Wageningen, The Netherlands, p. 32. Tiktak, A., Hendriks, R.F.A., Boesten, J.T.I., 2012a. Simulation of movement of pesticides towards drains with a preferential flow version of PEARL. Pestic. Manag. Sci. 68, 290e302. Tiktak, A., Hendriks, R.F.A., Boesten, J.J.T.I., van der Linden, A.M.A., 2012b. A spatially distributed model of pesticide movement in Dutch macroporous soils. A spatially distributed model of pesticide movement in Dutch macroporous soils. J. Hydrol. 470e471, 316e327. Van Dam, J.C., Huygen, Wesseling, J.G., Feddes, R.A., Kabat, P., van Walsum, P.E.V., Groenendijk, P., van Diepen, C.A., 1997. Theory of SWAP Version 2.0. Simulation of Water Flow, Solute Transport and Plant Growth in the Soil-wateratmosphere-plant Environment. SC-DLO technical document 45, p. 167. , Wageningen, The Netherlands. Vanclooster, M., Boesten, J.J.T.I., Trevisan, M., Brown, C.D., Capri, E., Eklo, O.M., Gottesburen, B., Gouy, V., van der Linden, A.M.A., 2000. A European test of pesticide-leaching models: methodology and major recommendations. Agric. Water Manag. 44, 1e19. Vereecken, H., Maes, J., Feyen, J., Darius, P., 1989. Estimating the soil moisture retention characteristics from texture, bulk density and carbon content. Soil Sci. 148, 389e403. Vryzas, Z., Papadopoulou-Mourkidou, E., Soulios, G., Prodromou, K., 2007. Kinetics and adsorption of metolachlor and atrazine and the conversion products (deethylatrazine, deisopropylatrazine, hydroxyatrazine) in the soil profile of a river basin. Eur. J. Soil Sci. 58, 1186e1199. Vryzas, Z., Papadakis, E.N., Papadopoulou-Mourkidou, E., 2012. Leaching of BrL, metolachlor, alachlor, atrazine, deethylatrazine and deisopropylatrazine in clayey vadoze zone: a field scale experiment in north-east Greece. Water Res. 46, 1979e1989. Wauchope, R.D., Buttler, T.M., Hornsby, A.G., Augustijn-Beckers, P.W.M., Burt, J.P., 1992. The SCSrARSrCES pesticide database for environmental decision making. In: George, W.W. (Ed.), Review of Environmental Contamination and Toxicology, vol. 123. Springer-Verlag, p. 164. Weber, J.B., Wilkerson, G.G., Reinhardt, C.F., 2004. Calculating pesticide sorption coefficients (Kd) using selected soil properties. Chemosphere 55, 157e166. Weed Science of America, 1994. Herbicide Handbook, seventh ed. €sten, J.H.M., Lilly, A., Nemes, A., Le Bas, C., 1999. Development and use of a Wo database of hydraulic properties of European soils. Geoderma 90, 169e185. Zeng, T., Chin, Y.P., Arnolr, W.A., 2012. Potential for abiotic reduction of pesticides in prairie pothole porewaters. Environ. Sci. Technol. 46 (6), 3177e3187.
Contents lists available at ScienceDirect
Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman
Evaluation of two pesticide leaching models in an irrigated field cropped with corn Dorothea D. Giannouli, Vassilis Z. Antonopoulos* School of Agriculture, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece
a r t i c l e i n f o
a b s t r a c t
Article history: Received 11 September 2014 Received in revised form 26 November 2014 Accepted 26 December 2014 Available online 2 January 2015
Pesticide leaching models is an easy and cost effective method used in the prediction of surface and groundwater pollution. In this paper, the ability of two pesticide leaching models, MACRO and PEARL, to describe soil water dynamics and atrazine's transport through the soil profile was examined. The data used for the comparison was obtained from an experiment in an irrigated corn field in the plain of the Ardas River, in north-eastern Greece. Both models were parameterized using pedotransfer functions, field and laboratory data. The uncalibrated simulation showed several discrepancies, therefore the retention curve and the sorption parameters were calibrated according to the trial and error method. The comparison of both models indicated that soil water flow was described similarly. The simulated results of atrazine's concentration were evaluated and compared to the measured concentrations at specific depths, using statistical criteria. Atrazine transport was simulated in a satisfactory manner as confirmed by model efficiency (EF) values, that are very close to unit. Coefficient of residual mass (CRM) values for both models are positive, indicating that both models underestimate the measured data. MACRO estimated higher accumulated actual evapotranspiration values, and less percolated water from soil profile than PEARL, and as a result, change in water content was higher in the latter. PEARL also predicted that half the amount of the applied mass was decayed two days earlier than the day estimated by MACRO. Generally, MACRO simulated the fate of atrazine in soil better than PEARL. © 2014 Elsevier Ltd. All rights reserved.
Keywords: Pesticide leaching models MACRO PEARL Field data Atrazine
1. Introduction Agriculture is one of the main sources of groundwater contamination, because of the extended use of pesticides and other agrochemicals. Several monitoring studies mention the presence of agrochemicals in surface and groundwater (Cooper, 1996; Reemtsma et al., 2013). Several studies confirm the presence of numerous pesticides in groundwater (Hallberg, 1989; Papastergiou and Papadopoulou-Mourkidou, 2001; Gonçalves et al., 2007). Apart from biotic, the extensive use of pesticides influences also the abiotic processes in soil (Zeng et al., 2012; Fenner et al., 2013). The European Union, in order to protect groundwater from pollution, promotes several directives such as the Water Framework Directive (2000/60/EC), the Groundwater Directive (2006/118/EC) and the recent Directive 2013/39/EU, regarding priority substances in the field of water policy. Besides, according to the Directive 98/83/EC, it is important that agrochemicals that reach groundwater do not
* Corresponding author. E-mail address: [email protected] (V.Z. Antonopoulos). http://dx.doi.org/10.1016/j.jenvman.2014.12.044 0301-4797/© 2014 Elsevier Ltd. All rights reserved.
exceed a concentration of 0.1 mg/L. In Greece, water pollution is caused mostly by the plant protection products used in corn, rice and cotton crops. According to Albanis et al. (1998), the maximum concentration of DEA and metolachlor detected in river water samples reached a value of 0.53 and 0.56 mg/L, respectively. Atrazine has been detected in significant concentrations (1.51 mg/L) in the groundwater of corn-growing areas of Northern Greece (Papastergiou and PapadopoulouMourkidou, 2001). High concentration levels of atrazine were also detected at the depth of 35 cm (1166 mg/L, Vryzas et al., 2012). In a recent study (Kalogridi et al., 2014), several pesticides and especially triazines, were detected in the surface waters and sediments of Lakes Kerkini, Doirani and Volvi, in Northern Greece. An efficient and low cost method for a preliminary assessment of groundwater vulnerability is the use of mathematical models. Various models like GLEAMS (Knisel and Davis, 2000), PRZM 3.21 (Carsel et al., 1998) and PEARL (Leistra et al., 2001) have been developed for the description of soil water movement and pesticide leaching. These models can simulate not only water flow and pesticides fate, but also most of the processes that compose the water
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515
and mass balance. Vanclooster et al. (2000) presented an assessment of twelve models simulating water movement and pesticide transport in soil, with the purpose of describing the problems encountered in using these models, and to introduce a Good Modelling Practice. In recent years though, there has been a strong tendency to incorporate soil water movement through macropores, commonly known as preferential flow, into leaching models. For this reason, many models that account for preferential flow, such as HYDRUS (Sim unek et al., 2005), RZWQM (Ahuja et al., 2000a,b) CRACK-NP (Armstrong et al., 2000b) and MACRO (Larsbo and Jarvis, 2003) have been developed. Despite the advantages a mathematical model can provide, when it comes to a comparison between two or more models, certain discrepancies may be observed. Armstrong et al. (2000a) and Persicani (1996) mention the variability in models' prediction. Although many studies concerning pesticide leaching in soil columns or lysimeters can be found in literature, only a few are based on data derived from field scale experiments (Flury et al., 1995; Flury, 1996; Vanclooster et al., 2000). The lack of measured values, especially at field scale, led to the use of stochastic param€hne et al., 2009). Among the limited number of compreeters (Ko hensive field-scale tests of pesticide transport models is the work of Suarez et al. (2013), who studied the transport of simazine in a vineyard using the HYDRUS 2D model. Other works with measured field data include the validation of the PEARL model using data from two sites in the Netherlands and Sweden (Bouraoui, 2007), the testing of the MACRO model in a cracked clay soil in the Netherlands by Scorza et al. (2007), and the impact of different irrigation practises on herbicide leaching by Fait et al. (2010). Recognizing the lack of site-specific data, this paper tries to contribute to the testing and the validation of two of the most commonly used in Europe models in the prediction of pesticide leaching. MACRO and PEARL models were used to simulate soil water dynamics and a solute's fate in a clayey soil, using the field data from an irrigated field cropped with corn in north-eastern Greece, under the climate and growing conditions of a Mediterranean area. 2. Method 2.1. Description of the MACRO and PEARL models MACRO 5.2 (Larsbo and Jarvis, 2003) and PEARL 4.4.4 (Leistra et al., 2001) are two representative pesticide leaching models, and were chosen in this study because of their credibility and ease of use (Jarvis, 1995; Boesten and van der Linden, 2001; Scorza and € hne et al., 2009; Fait et al., Boesten, 2005; Scorza et al., 2007; Ko 2010). FOCUS team (Boesten et al., 2000) through realistic worstcase scenarios, evaluated MACRO and PEARL models as a first-tier assessment of a pesticide leaching to groundwater. Since a full description of MACRO and PEARL is given elsewhere (Larsbo and Jarvis, 2003; Tiktak et al., 2000), only a brief summary of the models will be presented here. Both MACRO and PEARL are one-dimensional models which describe water flux, heat, and solute transport in soil matrix. Moreover, both account for a complete water balance including water flow, canopy interception and root water uptake, seepage to drains and groundwater. PEARL is linked with the SWAP model (Van Dam et al., 1997), whose soil hydrology is described by Richard's equation. MACRO is a dual permeability model which simulates preferential flow of water and solutes by dividing the soil matrix into the micropore and macropore domain, each one described by specific characteristics such as degree of saturation, conductivity and flux. Therefore, soil water flow in micropores is also described by Richards equation, and a kinematic wave
509
equation is used in the macropore domain. A convectionedispersion equation is used to describe the solute's transport, and in the macropore domain, MACRO uses only the convection equation. In both models, instantaneous equilibrium or kinetic sorption is described by either a linear or a Freundlich equation, and degradation by first order kinetics, depending on soil water content, temperature and depth. The degradation process in MACRO occurs in both domains. PEARL can also account for a pesticide's volatilization from soil or the plant's canopy, while MACRO does not account for this process. Crop growth in both models is described by a simple model where both the leaf area index (LAI) and the rooting depth are a function of the development stage of the crop. In MACRO, LAI and root depth follow a logistic curve, when in PEARL both are linear. Potential evapotranspiration is calculated using the equation of PenmaneMonteith (Allen et al., 1998) or can be pre-calculated and provided by the user, and can be partitioned into actual evaporation and transpiration through reduction functions. Finally, both models assume that runoff happens when infiltration capacity is exceeded. All these processes are also summarized in Table 1. Several studies are included in literature where PEARL and MACRO models were used to describe the water flow and the pesticide transport in soil (Scorza and Boesten, 2005; Scorza et al., 2007; Leistra and Boesten, 2010; Stenemo et al., 2007). In some cases, MACRO did not considered for preferential flow, and was used only for the one-domain approach (Antonopoulos et al., 2013). Because of the importance of the flow through the soil macropores, Tiktak et al. (2012a,b) presented two studies where PEARL, through macropore modules, could account for preferential flow (or a twodomain approach). This version of PEARL though, is unavailable to general public due to further testings.
3. Field and computational data 3.1. Site description and data sets The data used was derived from a 0.4 experimental field site in the Ardas River plain (41370 N, 26 210 E) in north-eastern Greece and was presented elsewhere in detail (Fragkoulis, 2003; Vryzas et al., 2007; Antonopoulos et al., 2011). Ardas River plain is one of the most important areas nationwide for corn production, and is characterized by warm summers and wet and cold winters, where Table 1 Description of the MACRO and PEARL processes. MACRO Water flow
Micropores: Richard's equation Macropores: Kinematic wave Solute Micropores: Convection dispersion transport equation Macropores: convection (gravity flow) Sorption Linear or Freundlich Instantaneous equilibrium and kinetic sorption for micro- and macropores Degradation First-order kinetics, separate rate coefficients for both solid and liquid phase of micro- and macropores Water Empirical sink terms uptake Volatilization No Initial Soil water content and temperature condition ETp PenmaneMonteith or input ETa Reduction functions Crop growth LAI as a Logistic function
PEARL Richard's equation Convection dispersion equation Linear or Freundlich, Instantaneous and non equilibrium First-order kinetics
A function of transpiration Soil and canopy Groundwater level PenmaneMonteith, Makkink, or input Reduction functions LAI as a Linear function
510
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515
the average annual rainfall and temperature are 771 mm and 12 C, respectively. The soil consisted of two layers and was characterized as loam (0e40 cm) and clay loam (40e120 cm). Generally, the soils of this area are clayey with low organic matter, and are characterized by recent alluvial deposits of the rivers Ardas and Evros. The depth to groundwater is about 6 m. Selected soil properties are given in Table 2. On April 7th 2000, the field was sown with corn and was harvested on September 19th 2000. Atrazine was applied to the soil surface on April 7th, a few hours before corn was sown, using an application rate of 1 kg/ha. In this field, no crop was sown and no herbicide has been applied before. During the growing season, the rainfall was 147.8 mm, the amount of water applied by sprinkler irrigation between June and August was 328 mm, and the reference evapotranspiration was calculated at 608.8 mm. Daily evapotranspiration values were calculated using data from a nearby meteorological station, and the PenmaneMonteith method. The maximum values of LAI and root depth were 4.5 and 1.4 m respectively. Agricultural practises and the parameters of the corn's development are summarized in Table 3. Atrazine leaching was measured under filed conditions at depths of 20 and 35 cm using suction cups (Fragkoulis, 2003; Vryzas et al., 2012). 3.2. Model parameterization The soil profile was divided into four horizons of 0e30, 30e40, 40e80 and 80e120 cm each. In order to provide adequate numerical accuracy, the soil profile in MACRO was partitioned into 60 layers ranging from 0.3 to 2.5 cm, while in PEARL discretization along the z axis ranged from 2.5 to 5 cm, separating soil profile in 30 layers. According to good modelling practice, and in order to simulate the initial water content in the beginning of the simulation, a six-month warming up period from October 1st 1999 to 31 March, 2000 was used. The simulation period lasted from April 1st to September 30th, 2000. Both models were parameterized with measured data, pedotransfer functions and literature values when certain data was unavailable. Initial conditions were set as a groundwater level of 600 cm for PEARL, and by inserting the initial values of water content and temperature for MACRO, while the bottom boundary condition was set as a free drainage for both models. Soil water content at field capacity qfc was set at a pressure head of 167 cm and 400 cm for the topsoil and subsoil, respectively, and at wilting Table 2 Selected soil physicochemical and hydraulic properties of the corn field. Characteristic
Sand (%) Silt (%) Clay (%) OM (%) pH rb (gr cm3) qs (m3 m3) qr (m3 m3) avg (cm1) nvg () Ks (m d1) hb (cm) hb (cm) 1D qb (m3 m3) Kb (m d1) n* () d (mm) d (mm) 1D
Depth (cm) 0e30
30e40
40e80
80e120
48 35 17 0.8 6.3 1.550 0.36 0.01 0.0143 1.44 0.120 10 0 0.351 0.053 4 6 1
43 29 28 0.7 6.8 1.565 0.39 0.04 0.0156 1.39 0.661 10 0 0.380 0.031 4 6 1
40 30 30 0.4 7.3 1.621 0.38 0.04 0.0160 1.32 0.646 10 0 0.369 0.026 4 6 1
43 27 30 0.3 7.7 1.644 0.37 0.04 0.0180 1.29 0.042 10 0 0.310 0.029 4 6 1
Table 3 Agricultural practices and corn growth parameters. Date
Treatment
7 Apr. 2000 14 Apr. 2000 15 June 2000 30 Jun. 2000 16 Aug. 2000 19 Sept. 2000
Atrazine application þ seeding
Crop stage LAI Root Root (m2 m2) depth (m) distribution
Emergence 0.01
0.06
60%
Flowering 4.50
1.4
60%
Harvest
1.4
Start of irrigation
End of irrigation 4.50
point qpwp, was set at 15000 cm. The parameters related to van Genuchten's soil water retention curves and hydraulic conductivity were determined for each layer through soil texture and bulk density, using Rosetta Lite pedotransfer functions, as proposed by Schaap et al. (2001). For MACRO model, the parameters demarcating the two domains such as the boundary water content qb and hydraulic conductivity Kb were defined by using the incorporated pedotransfer functions (PTF) of MACRO 5.2 GUI and algorithms proposed by € sten et al. (1999) and Moeys et al. (2012). The boundary pressure Wo head hb was set at 10 cm, according to the range given by Jarvis and Larsbo (2012) and then, the boundary water content qb was estimated from the water retention curves by setting h ¼ hb. The boundary hydraulic conductivity Kb for every horizon was estimated using the MACRO 5.0/5.1 pedotransfer functions of MACRO 5.2 GUI. The same PTFs were used for the estimation of the effective diffusion pathlength d which was set to 6 mm. Tortuosity/pore size distribution coefficient n* was set to 4, the maximum default value according to Beulke et al. (2002) and Antonopoulos et al. (2013). Crop growth parameters were based on default values provided by the models or by observations. The growth stages for the corn crop were 30/50/20/50 days. In PEARL, crop factor as a function of development stage was set at 0.89 for the 30th of June, where LAI reached the maximum value of 4.5. For the estimation of potential evapotranspiration in MACRO, the crop coefficient values adjusted to local conditions were kcinit ¼ 0.4, kcmid ¼ 1.1 and kcend ¼ 0.3. Physicochemical and sorption properties for atrazine such as sorption coefficient Koc and Freundlich exponent n came from literature (Weed Science of America, 1994). Wauchope et al. (1992) ranged the Koc value of atrazine between 38 and 174 cm3 g1, and the half-life from 18 to 120 days. The half-life (DT50) value of atrazine at 20 C, and as a consequence the degradation rate coefficient mref, derived from laboratory experiments (Fragkoulis, 2003) and was set at 12.6 days and 0.055 d1, respectively. The aforementioned sorption and degradation parameters were considered only for topsoil. Depth has a significant role in transformation rates, since degradation decreases with depth. Therefore, a degradation factor is used by both models to correct each layer's transformation rate, equal to 0.5 for a subsoil layer of 30e60 cm, to 0.3 for a subsequent layer of 60e100 cm, or equal to zero for layers below 100 cm (Boesten et al., 2000). It is assumed that the applied substance is mixing with the soil water content that is stored in a 1 mm thick soil layer. No interception occurred since a preemergent herbicide was applied. Furthermore, although both models can deal with non-equilibrium sorption, in this study kinetic sorption was not assumed, according to first-tier leaching assessment (Boesten et al., 2000). A first simulation using both models was executed without any
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515
calibration, in order to test the description of water flow and atrazine's fate in soil. The initial results showed that certain parameters had to be calibrated, because of the differences between measured and calculated data. Van Genuchten's parameters a and n, affecting the soil water retention curves, and the two parameters affecting sorption, Koc and the Freundlich exponent n, were calibrated according to trial-and-error method, until the model results could give a good prediction of the measured data. It occurred that PEARL simulated better the soil water content when a lysimeter as a lower boundary condition was used. To calibrate the hydraulic parameters, the values obtained by Rosetta's PTF's were compared to those derived by using Vereecken pedotransfer functions (Vereecken et al., 1989), and the most appropriate values for the simulation are presented in Table 2. Then the sorption constant Koc and the Freundlich exponent n were adjusted to the observed concentration values at specific depths. Atrazine properties used for the simulation are given in Table 4. The Koc and n values obtained after the calibration were 70 L kg1 and 0.88, respectively. The sorption distribution coefficient Kd was calculated from the measured organic carbon content in the soil horizons using the equation:
Kd ¼ Koc OC and was set at 0.32 cm3 gr1. Weber et al. (2004) correlated Kd values with soil properties for selected pesticides, and suggested 2.65 cm3 gr1 as a mean value for atrazine. The degradation rate coefficient mref in different depths was calculated using corrective factors (Boesten et al., 2000). In MACRO model, degradation rates were assumed to be the same in both phases and domains of every horizon. The matrix dispersivity was set at 5 cm, according to FOCUS scenarios (Boesten et al., 2000), and the diffusion coefficient in water was set at the default value of 4.9 1010 m s1. The mass fraction f was set at 0.02 as proposed by Dubus et al (2003) and Jarvis and Larsbo (2012). 3.3. Model evaluation The performance of both models in simulating the measured substance distribution was evaluated using three statistical criteria in specific depths, as proposed by Loague and Green (1991) and Antonopoulos (2000), the modelling efficiency EF, the relative root mean square error RMSE and the coefficient of residual mass CRM.
EF ¼
n n 2 X X Oi O ðPi Oi Þ2 i¼1
RMSE ¼
i¼1 n X
!,
n X
Oi O
2
(1)
i¼1
! . 1=2 ðPi Oi Þ n 2
(2)
i¼1
Table 4 Selected properties of atrazine. Properties
Atrazine
Molar mass (g) Vapour pressure (Pa) Solubility in water (mg L1) Molar Enthalpy (KJ mol1) Sorption distribution coefficient kd (cm3/g) Sorption constant Koc (cm3 g1) Freundlich exponent n Degradation rate coefficient mref (d1) Half-life at 20 C DT50 (d)
215 3.9 104 (25 C) 33 27 0.32 70 0.88 0.055 12.6
CRM ¼
n X
Oi
i¼1
n X
!, Pi
i¼1
n X
511
Oi
(3)
i¼1
where Pi and Oi are the simulated and measured values of atrazine respectively, O is the mean of measured values and n is the number of measured values. The optimum values of EF, RMSE and CRM criteria are 1, 0 and 0, respectively. EF value compares the predicted values to the average value of the measurements. If EF is less than zero, the model predicted values are worse than simply using the observed mean. Positive values of CRM indicate that the model underestimates the measurements.
4. Results and discussion 4.1. Water content The simulated results of water content, using the MACRO and PEARL models, are presented in Fig. 1 at the depths of 20, 60 and 110 cm, respectively, and the components of the water balance are shown in Table 5. During the simulation period (from the 92nd to 274th day of the year), a total of 502.8 mm of rain and irrigation water was applied. Surface runoff was not observed, since the experimental field was flat. Predicted values from both models followed a similar pattern, and both can simulate interactions between soil, plant and atmosphere in a reasonable manner. As shown in Fig. 1, during the study period, the simulated soil water dynamics was similar for both models at the three different depths. At the depth of 20 cm, and during the irrigation period (June to August), MACRO estimated higher evapotranspiration values, and as a consequence soil water content was lower than the one estimated by PEARL, causing a discrepancy. The same inconsistency is also observed at the depths of 60 cm and 110 cm, for a shorter period. This discrepancy can be explained by the different approach of both models to the estimation of dynamic crop properties, especially when it comes to the separation of potential evapotranspiration into soil evaporation and transpiration from the plant. Although the same daily potential evapotranspiration values were used for the estimation of actual evapotranspiration, the predictions from both models were different. As mentioned before, LAI is modelled as a logistic curve in MACRO and a linear one in PEARL, resulting in a different estimation of evaporation and transpiration (Table 5) and, as a consequence, different water content. Using the MACRO model, the cumulative actual evaporation from the soil and the transpiration from the plants were calculated at 73.2 mm and 368.9 mm respectively while in PEARL were 116 mm and 283.5 mm, respectively. During the simulation period, the MACRO approach predicted that 61.7 mm of water was percolated from soil profile, whereas PEARL predicted 132.2 mm. Due to different values of calculated actual evapotranspiration and deep percolation, the change in water content during the whole simulation period was 3.8 mm in MACRO approach, whereas in PEARL was 19.9 mm. In a parallel simulation where MACRO considered only for flow in the micropore domain (results not shown here), higher water content values were predicted, especially after rain or irrigation events in every layer. The change in water content was decreased only by 5.6 mm, due to the change in percolated water. The same difference was also noticed by Antonopoulos et al. (2013) after comparing one and two-domain approaches using MACRO in an irrigated corn field. Garratt et al. (2002) also mentioned interesting differences in water balance when comparing seven models with different methods of estimating soil water flow and evapotranspiration. In
512
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515
100 Irrigation + Ra in
MACRO
PEARL
20 cm
0,4
80
0,3
60
0,2
40
0,1
20 0
0,0 92
122
152
182 Day of year
212
242
272
100 Irrigation + Rain
MACRO
60 cm
PEARL
0,4
80
0,3
60
0,2
40
0,1
20
0,0
irrigation + rain, mm
soil water content, m3 m-3
0,5
0 92
122
152
182 Day of year
212
242
272
100
0,5 Irrigation + Rain
MACRO
110 cm
PEARL
0,4
80
0,3
60
0,2
40
0,1
20
irrigation + rain, mm
soil water content, m3 m-3
irrigation+ rain, mm
soil water content, m3 m-3
0,5
0
0,0 92
122
152
182 Day of year
212
242
272
Fig. 1. Comparison of simulated water content at 20, 60 and 110 cm using MACRO and PEARL model.
MACRO, potential evapotranspiration is used as a direct input, whereas in this simulation PEARL uses reference evapotranspiration values and then the potential evapotranspiration is calculated internally by the model. Different methods of evapotranspiration can cause significant changes the ratio of actual and potential transpiration and eventually in water content (Jarvis et al., 2000). Although no measured data was available, the two models displayed similarly the soil water simulation. Jarvis et al. (2000) using pedotransfer functions in a MACRO model simulation, noted a value of RMSE ¼ 0.06 m3 m3 between measured and simulated water content, while Antonopoulos (2000) noted that the RMSE using direct and pedotransfer function for parameters of van Genuchten retention curves was 0.038 and 0.037 cm3 cm3, respectively, between the measured and simulated values of water content. 4.2. Atrazine behaviour The results of atrazine simulation at depths of 20 cm and 35 cm are shown in Fig. 2. Atrazine concentrations simulated with the MACRO and PEARL models showed similar patterns versus time at both depths. There is a good agreement between modelling results and measured data, even though the predicted concentrations from both models are higher in the upper layer and lower on subsoil.
Atrazine remained in significant concentrations on the topsoil for more than a two-month period after application. According to the MACRO simulation, the applied atrazine appeared at 20 cm depth on the 98th day of the year and reached the underlying layer two weeks later, while at the end of the simulation, the concentration level of atrazine at 35 cm depth was 0.131 mg/m3. Meanwhile, the PEARL simulation also predicted the atrazine's appearance on the topsoil two weeks after the application day, while on the underlying layer, atrazine appeared on 122th day of year, eight days later than the day predicted by MACRO. Atrazine concentration at the end of the simulation period was 0.059 mg/m3 at the 35 cm depth. At the same depth, and during the first 55 days, the predicted by MACRO values were higher than those predicted by PEARL. These values, combined with the earlier atrazine appearance in this layer could be explained by preferential displacement, since PEARL model does not account for a solute's behaviour through macropores. According to PEARL approach, atrazine remained on the topsoil in higher concentrations (the maximum value was 177 mg/m3 on the 124th day of year) while on the subsoil, the predicted values were lower. On the contrary, the simulated by MACRO concentration levels were lower on the first layer (the maximum value was 137.9 mg/m3 on the 120th day of year), and higher on the subsoil (the maximum value was 6.5 mg/
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515 Table 5 Water balance for MACRO and PEARL models from 92 to 274 day of year 2000. MACRO (mm)
PEARL (mm)
Cumulative rain and irrigation Cumulative actual evaporation Cumulative actual transpiration Cumulative runoff Cumulative percolation Cumulative canopy evaporation Change in water content
502.9 73.2 368.9 0 61.7 2.8 3.8
502.9 116 283.5 0 123.2 e 19.9
total solute concentration, mg m-3
Components of soil water balance
20 cm 150
measured MACRO
100
PEARL 50 0
total solute concentration, mg m-3
92
122
152
182 Day of year
212
242
272
8
35 cm 6
Measured MACRO
4
PEARL 2 0 92
122
152
182
212
242
Table 6 Statistical criteria between measured and calculated values. Depth (cm)
20 35 Average
200
272
Day of year Fig. 2. Measured and simulated atrazine concentrations at 20 and 35 cm using MACRO and PEARL model.
m3 on the 128th day of year). On the180th day, when the simulated by PEARL concentration reached the maximum value of 6.3 mg/m3, the value predicted by MACRO was 4.9 mg/m3. On the topsoil, PEARL model predicted better the concentration, after the irrigation event on the 168th day of year. The irrigation event on the 192th day of the year was critical, because it led to an increase of the simulated by MACRO atrazine concentrations at 35 cm depth. According to Fait et al. (2010), the type of irrigation is important for a pesticide's leaching. Statistical criteria give a more detailed picture of both models' performance, and are summarized in Table 6. Generally, both models predict atrazine distribution at two depths acceptably well, judging from average EF values that are close to 1. Average RMSE values are lower for MACRO than for PEARL, and average CRM are close to 0. More specifically, at the depth of 20 cm, CRM values indicate that both models overestimate the measured data, while at the depth of 35 cm there is a tendency to underestimate concentration values. The combination of MACRO's RMSE and EF values indicate the latter as a better model in the description of a solute's behaviour. CRM and EF values derived from a field study using a similar Koc value for atrazine (72 mg L1) in a clay soil, ranged from 1.19 to 0.62 and from 3.07 to 0.57, respectively (Sarmah et al., 2005). The estimated by both models mass balance for atrazine as a function of time is shown in Fig. 3. In MACRO simulation, almost 99% of the applied mass was stored in the liquid and solid phase of the soil on the application day, 50% was stored on the 121th day of
513
EF
RMSE
CRM
MACRO
PEARL
MACRO
PEARL
MACRO
PEARL
0.876 0.996 0.936
0.584 0.994 0.789
10.667 1.421 6.044
19.514 1.707 10.610
0.126 0.350 0.112
0.332 0.392 0.030
the year, 25% was stored on the 140th day of the year, and 10% was stored on the 160th day of the year (62 days after application). A very small amount (0.024 mg m2) was also stored in soil macropores. Moreover, 50% of the applied mass was degradated on April 30th (121 day of the year), 23 days after application. No degradation from canopy, solute uptake by crop or leaching to groundwater were observed during the simulation period. At the end of the simulation, 0.072 and 0.001 mg m2 were stored in the liquid and solid phase of micro- and macropore domain, respectively. According to PEARL simulation, 99% of the applied mass was also stored in soil profile on the application day, 50% and 25% of the applied mass were stored two days earlier than the MACRO prediction, and 10% was stored on the 155th day of the year (57 days after application). Until the 196th day of year (98 days after application) degradation occurred faster, since 50% of the applied mass was transformed on 119th day of year, two days earlier than the day estimated by MACRO, and as a consequence, PEARL predicted lower stored mass and higher degradation levels. The latter also predicted 0.81 mg m2 of atrazine uptake by crop, 0.015 mg m2 of atrazine volatilization and 0.017 mg m2 of stored mass at the end of the simulation. Atrazine leaching to groundwater was also not observed. In Table 7, a more detailed mass balance for the 99th, 216th and 274th day of the year is presented for MACRO and PEARL models, respectively. These days represent the days after application, after the maximum irrigation event and the last day of simulation, respectively. In other field experiments with corn crops, 15% of the initially applied atrazine remained after one year of application, at 1 m depth (Tasli et al., 1996), while atrazine began to transform into DEA on topsoil 15 days after application (Martínez et al., 1996). Sarmah et al. (2009) reported that half-lives of atrazine at the topsoil and the subsoil of a silt loam soil were 18 days and 47 days, respectively. Scorza and Boesten (2005) using the PEARL model, tried to simulate the movement of water, and the transport of bentazone and imidacloprid in a cracked clay soil. Although measured and simulated values of water and imidacloprid were in a good agreement, PEARL did not predict the bentazone concentrations well because of their preferential transport to soil macropores. The PEARL model also estimated higher concentrations and persistence of bentazone in soil profile (Bouraoui, 2007). Armstrong et al. (2000a), after testing four different models in a macroporous clay soil in the UK, indicated MACRO as a better model in the estimation of isoproturon concentrations in soil. Larsson and Jarvis (2000) concluded that when using the MACRO model without macropore flow, no leaching of the applied bromide and bentazone was observed. On the other hand, MACRO was able to simulate well measured and simulated values of an early bentazone breakthrough in drainage water, indicating flow through macropores (Scorza et al., 2007). As a result of atrazine bypassing the matrix soil, Flury et al. (1995) indicated that 40 mg kg1 of atrazine were detected in soil cracks and at a depth of 80e90 cm of a loamy soil. A literature review in sensitivity analysis for both MACRO and PEARL models, indicated the sensitivity of the related to sorption and degradation parameters (Armstrong et al., 2000a; Dubus et al.,
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515
100
1,0
80
0,8
60
0,6
Tota l solute stored in profile, MACRO Tota l solute stored in profile, PEARL Degra da tion in soil, MACRO Degra da tion in soil, PEARL Solute uptake by plants, MACRO Solute uptake by plants, PEARL Solute vola tiliza tion, PEARL
40
20
0,4
0,2
0
0,0 92
122
152
182 Day of year
212
242
Solute uptake and volatilization, mg m-2
Total mass and degradation, mg m-2
514
272
Fig. 3. Mass balance for atrazine in soil profile (120 cm) using MACRO and PEARL models.
2003; Tiktak, 2012a,b). Dubus et al. (2003) after performing a sensitivity analysis in four mathematical models including PEARL and MACRO, concluded that parameters affecting sorption and degradation such as degradation rates and Freundlich exponent n as well as sorption distribution coefficient Koc were the most sensitive, while the water balance was slightly affected by variation in input parameters. Scorza et al. (2007) also indicated the mass fraction of solid material contacting water in macropores as a sensitive parameter, and Armstrong et al. (2000a) the significance of the mixing depth and the user's subjectivity. 5. Conclusion In this paper, the dual-permeability model MACRO was compared to PEARL model, to simulate the fate of atrazine in a clayey soil in Ardas River plain. According to statistical criteria, both models provided a very good simulation of the mass transport of atrazine, although both of them underestimate atrazine concentrations in the subsoil. The results of the MACRO approach were better in average values than the simulation using PEARL, indicating the former as a better model. The existence of preferential water flow may be a cause for the discrepancy in the description of atrazine levels. Discrepancies between measured and predicted values were also caused by differences in the parameterization or the different approach of both models in the description of certain processes. The initial condition was described differently, and because in PEARL model, the initial water content was not an input value, a larger warm up period was needed to simulate better a more realistic simulation. Furthermore, PEARL model predicted similar results with MACRO when instead of free drainage, a lysimeter was used as a lower boundary condition. The choice of the initial and
Table 7 Change in mass balance for atrazine using MACRO and PEARL models during the simulation period. Mass (mg m2) Day of year 99 MACRO Stored in soil Degradation Crop uptake Volatilization Total mass a
216 PEARL
MACRO
274 PEARL
MACRO
PEARL
95.410 0.517a 0.185 0.073a 0.017 97.259a 2.741 4.464 99.482 99.000 99.927 99.160 0.000 0.000 0.000 0.809 0.000 0.811 e 0.002 e 0.015 e 0.0155 100 100 100 100 100 100
Micropores and macropores.
lower boundary condition is of great importance. Moreover, differences in the estimation of the actual evapotranspiration, probably because of the crop model used in both models, led to a discrepancy in soil water dynamics, indicating a wetter soil when using MACRO. When data is not sufficient, pedotransfer functions or data derived from literature is a useful way to estimate the missing parameters. Finally, both MACRO and PEARL models were developed and evaluated in countries where rainfall, and not always irrigation, is the main inflow in water balance. In this study, both models were applied using data derived from a corn field, emphasizing the soil water dynamics and atrazine behaviour during the irrigation period. It would be really interesting to extend this study by comparing MACRO model with a version of PEARL model, where the preferential flow would be considered. In conclusion, even if the same scenario is used, the predictions of mathematical models for a pesticide's fate may vary. Nevertheless, these models can be used for many environmental scenarios for the achievement of sustainable and rational water and agrochemical management. References Ahuja, L.R., Rojas, K.W., Hanson, J.D., Shaffer, J.J., Ma, L., 2000a. The Root Zone Water Quality Model. Water Resources Publications LLC, Highlands Ranch, CO. Ahuja, L.R., Johnsen, K.E., Rojas, K.W., 2000b. Water and chemical transport in soil matrix and macropores. In: Ahuja, L.R., Rojas, K.W., Hanson, J.D., Shaffer, J.J., Ma, L. (Eds.), The Root Zone Water Quality Model. Water Resources Publications LLC, Highlands Ranch, CO. Albanis, T.A., Hela, D.G., Sakellarides, J.M., Konstantinou, I.K., 1998. Pesticide residues in surface water, groundwaters and rainfall of Imathia (Greece). In: Proc. of Environmental Protection and Restoration, Sani Chalkidiki, pp. 119e126. Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop Evapotranspirationguidelines for Computing Crop Water Requirements. FAO Irrigation and Drainage Paper 56, Rome, Italy. Antonopoulos, V.Z., 2000. Modeling of soil water dynamics in an irrigated corn field using direct and pedotransfer functions for hydraulic properties. J. Irrig. Drain. Syst. 14, 325e342. Antonopoulos, V.Z., Fragkoulis, G., Papadopoulou- Mourkidou, E., Vryzas, Z., 2011. Assessment of pesticide leaching in a corn field using PEARL 1.1.1. model. In: Proceedings of the Council of Agricultural Engineers of Greece, Athens, Greece, p. 8 (in Greek). Antonopoulos, V.Z., Georgiou, P.E., Kolotouros, C.A., 2013. Soil water dynamics in cropped and uncropped fields in northern Greece using a dual-permeability model. Hydrol. Sci. J. 58 (8), 1748e1759. Armstrong, A., Aden, K., Amraoui, N., Diekkruger, B., Jarvis, N.J., Mouvet, C., Nicholls, P., Wittwer, C., 2000a. Comparison of the performance of the pesticide-leaching models on a cracking clay soil: results using the Brimstone Farm dataset. Agric. Water Manag. 44, 85e104. Armstrong, A.C., Matthews, A.M., Portwood, P.B., Leeds-Harrison, P.B., Jarvis, N.J., 2000b. CRACK-NP: a pesticide leaching model for cracking clay soils. Agric. Water Manag. 44, 183e199. Beulke, S., Renaud, F., Brown, C.D., 2002. Development of Guidance on Parameter
D.D. Giannouli, V.Z. Antonopoulos / Journal of Environmental Management 150 (2015) 508e515 Estimation for the Preferential Flow Model MACRO 4.2. Final Report of the DEFRA project PL0538. Cranfield Centre for EcoChemistry, UK, p. 68. University of Cranfield. Boesten, J.J.T.I., van der Linden, A.M.A., 2001. Effect of long-term sorption kinetics on leaching as calculated with the PEARL model for FOCUS scenarios. In: Pesticide Behaviour in Soils and Water. Proceedings of the British Crop Protection Council Symposium, Brighton, UK, 13-15 November 2001, pp. 27e32. Boesten, J., Businelli, M., Delmas, A., Gottesbüren, B., Hanze, K., Jarvis, T., Jones, R., Klein, M., van der Linden, T., Rekolainen, S., Resseler, H., Roquero, C., Maier, W.M., Styczen, M., Thorsen, M., Travis, K., Vanclooster, M., 2000. FOCUS Groundwater Scenarios in the EU Review of Active Substances Sanco/321/2000. Version 1. Bouraoui, F., 2007. Testing the PEARL model in the Netherlands and Sweden. Environ. Model. Softw. 22, 937e950. Carsel, R.F., Imhoff, J.C., Hummel, P.R., Cheplick, J.M., Donigian Jr., A.S., 1998. PRZM3: a Model for Predicting Pesticide and Nitrogen Fate in the Crop Root and Unsaturated Soil Zones: User's Manual for Release 3.12. National Exposure Research Laboratory. Office of Research and Development, US Environmental Protection Agency, Athens, GA. Cooper, B., 1996. Central and NorthWest RegionsWater Quality Program.1995/1996 Report on Pesticide Monitoring. TS96.048. NSW Department of Land & Water Conservation, Sidney, Australia. Dubus, I.G., Brown, C.D., Beulke, S., 2003. Sensitivity analyses for four pesticide leaching models. Pestic. Manag. Sci. 59, 962e982. Fait, G., Balderacchi, M., Ferrari, F., Ungaro, F., Capri, E., Trevisan, M., 2010. A field study of the impact of different irrigation practices on herbicide leaching. Eur. J. Agron. 32, 280e287. Fenner, K., Canonica, S., Wackett, L., Elsner, M., 2013. Evaluating pesticide degradation in the environment: blind spots and emerging opportunities. Science 341 (6147), 752e758. Flury, 1996. Experimental evidence of transport of pesticides through field soils- a review. J. Environ. Qual. 25, 25e45. Flury, M., Leuenberger, J., Studer, B., Fluhler, H., 1995. Transport of anions and herbicides in a loamy and a sandy field soil. Water Resour. Res. 31, 823e835. Fragkoulis, G., 2003. Evaluation of Leaching and Groundwater Pollution of Plant Protection Products Using PEARL Model in a Corn Plot in Evros. Master thesis. School of Agriculture, Aristotle University, Thessaloniki (in Greek). Garratt, J.A., Capri, E., Trevisan, M., Errera, G., Wilkins, R.M., 2002. Parameterisation, evaluation and comparison of pesticide leaching models to data from a Bologna field site, Italy. Pest Manag. Sci. 58, 3e20. Gonçalves, C.M., Da Silva, J.C.G.E., Alpendurada, M.F., 2007. Evaluation of the pesticide contamination of groundwater sampled over two years from a vulnerable zone in Portugal. J. Agric. Food Chem. 55 (15), 6227e6235. Hallberg, G.R., 1989. Pesticide pollution of groundwater in the humid Unites States. Agric. Ecosyst. Environ. 26, 299e367. Jarvis, N.J., 1995. Simulation of soil water dynamics and herbicide persistence in a silt loam soil using the MACRO model. Ecol. Model. 97e109. Jarvis, N., Larsbo, M., 2012. MACRO (v5.2): model use, calibration and validation. Trans. Am. Soc. Agric. Biol. Eng. 55 (4), 1413e1423. Jarvis, N.J., Brown, C.D., Granutza, E., 2000. Sources of error in model predictions of pesticide leaching: a case study using the MACRO model. Agric. Water Manag. 44, 247e262. Kalogridi, E.C., Christophoridis, C., Bizani, E., Drimaropoulou, G., Fytianos, K., 2014. Part I: temporal and spatial distribution of multiclass pesticide residues in lake waters of Northern Greece: application of an optimized SPE-UPLC-MS/MS pretreatment and analytical method. Environ. Sci. Pollut. Res. 21, 7239e7251. Knisel, W.G., Davis, F.M., 2000. GLEAMS, Groundwater Loading Effects from Agricultural Management Systems V3.0. Publ. No. SEWRL-WGK/FMD-050199. U.S.D.A., Tifton, Georgia. €hne, J.M., Ko €hne, S., Sim Ko unek, J., 2009. A review of model applications for structured soils:b) pesticide transport. J. Contam. Hydrol. 104, 36e60. Larsbo, M., Jarvis, N.J., 2003. MACRO 5.0. A Model of Water Flow and Solute Transport in Macroporous Soil. Technical Description. Department of Soil Sciences, Swedish University of Agricultural Sciences, Uppsala. Larsson, M.H., Jarvis, N.J., 2000. Quantifying interactions between compound properties and macropore flow effects on pesticide leaching. Pest Manag. Sci. 56, 133e141. Leistra, M., Boesten, J.J.T.I., 2010. Pesticide leaching from agricultural fields with ridges and furrows. Water Air Soil Pollut. 213, 341e352. Leistra, M., van der Linden, A.M.A., Boesten, J.J.T.I., Tiktak, A., van den Berg, F., 2001. PEARL Model for Pesticide Behaviour and Emissions in Soil-plant Systems. Description of Processes. Alterra report 13, RIVM report 711401009. Alterra, Wageningen, The Netherlands. Loague, K.M., Green, R.E., 1991. Statistical and graphical methods for evaluating solute transport models: overview and application. J. Contam. Hydrol. 7, 51e73. n, Y., S Martínez, L., Lecho anchez-Brunete, C., Tadeo, J.L., 1996. Persistence of metolachlor, atrazine and deethylatrazine in soil and its simulation by deterministic models. Toxicol. Environ. Chem. 54, 219e232. Moeys, J., Larsbo, M., Bergstrom, L., Brown, C.D., Coquet, Y., Jarvis, N.J., 2012. Functional test of pedotransfer functions to predict water flow and solute transport with the dual-permeability model MACRO. Hydrol. Earth Syst. Sci. 16,
515
2069e2083. Papastergiou, A., Papadopoulou-Mourkidou, E., 2001. Occurrence and spatial and temporal distribution of pesticide residues in groundwater of major corngrowing areas of Greece. Environ. Sci. Technol. 35, 63e69. Persicani, D., 1996. Pesticide leaching into field soils: sensitivity analysis of four mathematical models. Ecol. Model. 84, 265e280. Reemtsma, T., Alder, L., Banasiak, U., 2013. Emerging pesticide metabolites in groundwater and surface water as determined by the application of a multimethod for 150 pesticide metabolites. Water Res. 47 (15), 5535e5545. Sarmah, A., Close, M., Pang, L., Lee, R., Green, S.R., 2005. Field study of pesticide leaching in an Himatangi sand (Manawatu) and a Kiripaka bouldery clay loam (Norhtland). 2. Simulation using LEACHM, HYDRUS-1D, GLEAMS and SPASMO models. Aust. J. Soil Res. 43, 471e489. Sarmah, A.K., Close, M.E., Mason, N.W.H., 2009. Dissipation and sorption of six commonly used pesticides in two contrasting soils of New Zealand. J. Environ. Sci. Health Part B 44, 325e336. Schaap, M.G., Leij, F.J., van Genuchten, M.T., 2001. ROSETTA: a computer program for estimating soil hydraulic parameters with hierarchical pedotransfer functions. J. Hydrol. 251 (3e4), 163e176. Scorza, P.J., Boesten, J.J.T.I., 2005. Simulation of pesticide leaching in a cracking clay soil with the PEARL model. Pest Manag. Sci. 61432e61448. Scorza, R.P.J., Jarvis, N.J., Boesten, J.J.T.I., van der Zee, S.E.A.T.M., Roulier, S., 2007. Testing MACRO (version 5.1) for pesticide leaching in a Dutch clay soil. Pest Manag. Sci. 63, 1011e1025. Sim unek, J., van Genuchten, MTh, Sejna, M., 2005. The HYDRUS-1D Software Package for Simulating the One-dimensional Movement of Water, Heat, and Multiple Solutes in Variably-saturated Media. Version 3.0. In: HYDRUS Software Series, vol. 1. Department of Environmental Sciences, University of California Riverside, Riverside, CA, p. 270. Stenemo, F., Lindahl, A.M.L., G€ arden€ as, A., Jarvis, N., 2007. Meta-modeling of the pesticide fate model MACRO for groundwater exposure assessments using artificial neural networks. J. Contam. Hydrol. 93, 270e283. Suarez, F., Guzman, E., Munoz, J., Bachmann, J., Ortiz, C., Alister, C., Kogan, M., 2013. Simazine transport in undisturbed soils from a vineyard at the Casablanca valley, Chile. J. Environ. Manag. 117, 32e41. Tasli, S., Patty, L., Boetti, H., Ravanel, P., Vachaud, G., Scharff, C., Favre-Bonvin, J., Kaouadji, M., Tissut, M., 1996. Persistence and leaching of atrazine in corn culture in the experimental site of La Cote Saint Andre (Isere, France). Arch. Environ. Contam. Toxicol. 30, 203e212. Tiktak, A., van den Berg, F., Boesten, J.J.T.I., van Kraalingen, D., Leistra, M., van der Linden, A.M.A., 2000. Manual of FOCUS PEARL Version 1.1.1. Alterra report 28, RIVM report 711401008. Alterra, Wageningen, The Netherlands, p. 32. Tiktak, A., Hendriks, R.F.A., Boesten, J.T.I., 2012a. Simulation of movement of pesticides towards drains with a preferential flow version of PEARL. Pestic. Manag. Sci. 68, 290e302. Tiktak, A., Hendriks, R.F.A., Boesten, J.J.T.I., van der Linden, A.M.A., 2012b. A spatially distributed model of pesticide movement in Dutch macroporous soils. A spatially distributed model of pesticide movement in Dutch macroporous soils. J. Hydrol. 470e471, 316e327. Van Dam, J.C., Huygen, Wesseling, J.G., Feddes, R.A., Kabat, P., van Walsum, P.E.V., Groenendijk, P., van Diepen, C.A., 1997. Theory of SWAP Version 2.0. Simulation of Water Flow, Solute Transport and Plant Growth in the Soil-wateratmosphere-plant Environment. SC-DLO technical document 45, p. 167. , Wageningen, The Netherlands. Vanclooster, M., Boesten, J.J.T.I., Trevisan, M., Brown, C.D., Capri, E., Eklo, O.M., Gottesburen, B., Gouy, V., van der Linden, A.M.A., 2000. A European test of pesticide-leaching models: methodology and major recommendations. Agric. Water Manag. 44, 1e19. Vereecken, H., Maes, J., Feyen, J., Darius, P., 1989. Estimating the soil moisture retention characteristics from texture, bulk density and carbon content. Soil Sci. 148, 389e403. Vryzas, Z., Papadopoulou-Mourkidou, E., Soulios, G., Prodromou, K., 2007. Kinetics and adsorption of metolachlor and atrazine and the conversion products (deethylatrazine, deisopropylatrazine, hydroxyatrazine) in the soil profile of a river basin. Eur. J. Soil Sci. 58, 1186e1199. Vryzas, Z., Papadakis, E.N., Papadopoulou-Mourkidou, E., 2012. Leaching of BrL, metolachlor, alachlor, atrazine, deethylatrazine and deisopropylatrazine in clayey vadoze zone: a field scale experiment in north-east Greece. Water Res. 46, 1979e1989. Wauchope, R.D., Buttler, T.M., Hornsby, A.G., Augustijn-Beckers, P.W.M., Burt, J.P., 1992. The SCSrARSrCES pesticide database for environmental decision making. In: George, W.W. (Ed.), Review of Environmental Contamination and Toxicology, vol. 123. Springer-Verlag, p. 164. Weber, J.B., Wilkerson, G.G., Reinhardt, C.F., 2004. Calculating pesticide sorption coefficients (Kd) using selected soil properties. Chemosphere 55, 157e166. Weed Science of America, 1994. Herbicide Handbook, seventh ed. €sten, J.H.M., Lilly, A., Nemes, A., Le Bas, C., 1999. Development and use of a Wo database of hydraulic properties of European soils. Geoderma 90, 169e185. Zeng, T., Chin, Y.P., Arnolr, W.A., 2012. Potential for abiotic reduction of pesticides in prairie pothole porewaters. Environ. Sci. Technol. 46 (6), 3177e3187.