Modeling and testing of the effect of tillage, cropping
and water management practices on nitrate
leaching in clay loam soil
H.Y.F. Ng
a,1, C.F. Drury
b,*, V.K. Serem
c, C.S. Tan
b, J.D. Gaynor
baNational Water Research Institute, 867 Lakeshore Road, Burlington, Ont., Canada L7R 4A6
bGreenhouse and Processing Crops Research Center, Agriculture and Agri-Food Canada,
Harrow, Ont., Canada N0R 1G0
cMcGill University, Ste Anne de Bellevue, Que., Canada H9X 3V9
Accepted 4 March 1999
Abstract
Intensive agricultural management practices have contributed to nitrate concentrations in surface and subsurface drainage water which exceed water quality guidelines. Numerous studies have been conducted to examine the effects of soil and crop management practices on nitrate leaching, however, there are few studies which examine new technologies to reduce nitrate leaching from agricultural land. In this study, technological advances in water table management are used in combination with conservation tillage and intercropping treatments to determine if nitrate leaching can be effectively reduced in a clay loam soil. There were two water table management treatments and four crop management treatments arranged in a 2 by 4 factorial arrangement with two replicates. The water table management treatments consisted of controlled drainage/subirrigation system (CDS) and a free drainage system (FD). The crop management treatments consisted of moldboard plow tillage with and without an annual ryegrass intercrop (MP, MPIC) and soil saver (modified chisel plow) with and without an annual ryegrass intercrop (SS, SSIC). The data was modeled using LEACHM and a mean error difference procedure was used to determine how well the model predicted the field data. The model worked well when the difference between the predicted and measured values approached zero. When the CDS system was modeled, the mean error difference values ranged fromÿ1.67 to 0.33 mg N lÿ1for the 4 crop management treatments whereas the values were considerably greater when the FD system was modeled with values ranging between 7.48 and 12.7 mg N lÿ1. Hence the LEACHM model predicted nitrate leaching better on plots under CDS system than on plots under FD. Both the Leaching Estimation and Chemistry
Agricultural Water Management 43 (2000) 111±131
1Tel.: 905-336-4905, fax: 905-336-4420, e-mail: [email protected]
* Corresponding author. Tel.: +1-519-738-2251; fax: +1-519-738-2929.
E-mail address:[email protected] (C.F. Drury)
Model (LEACHM) predicted scenarios and field sampled data showed that CDS reduced nitrate leaching substantially. Model calibration by using one full year of field data was found to be acceptable, but for predictions based on shorter calibration periods unsatisfactory results were obtained.#2000 Elsevier Science B.V. All rights reserved.
Keywords: Modelling; Nitrate; Tile drainage; Subirrigation; Controlled drainage; Water table management
1. Introduction
Nitrate nitrogen is an essential plant nutrient. Nitrogen can be released from the soil by mineralization processes, however supplemental nitrogen is often required to maximize crop yields. During rainfall events, nitrate nitrogen from soil organic matter, fertilizer, livestock manure, or legume can leach through the soil and contaminate groundwater. High NO3ÿ levels in drinking water are unsafe especially for infants, the elderly and
young animals as NO3ÿcan be converted into NO2ÿin the digestive tract and lead to
methemoglobinemia, a condition in which the NO2ÿbinds to haemoglobin and causes
suffocation (Haynes et al., 1986; Sittig, 1991).
Nitrate pollution of lakes, rivers and groundwater has been well documented (Porter, 1975; International Joint Commission, 1978; Coote et al., 1982; Great Lakes Water Quality Board, 1987; Jones and Schwab, 1992; Polglase et al., 1995), however, there is very little information available on remedial methods for reducing nitrate loss from the agricultural areas. Further, the development of these remedial schemes can be enhanced when both field data and mathematical models are used. The objectives of this study were to identify management practices (i.e. water table management, conservation tillage and intercropping) which reduce nitrate leaching using field data and the LEACHM (Hutson and Wagenet, 1989) model.
2. Materials and methods
2.1. Experimental design
The experimental field plot was located in south-western Ontario, at Eugene F. Whelan Experimental Farm, Agriculture and Agri-Food Canada, Woodslee, Ont. The field plot configuration has been reported previously (Tan et al., 1993; Drury et al., 1996). The layout of the field plot design (Fig. 1) consists of 16 plots each 15 m wide by 67 m long with an area of 0.1068 ha (including berm). Each plot, is isolated by a double layer 4 mil plastic barrier from the surface to a depth of 1.2 m to prevent leakage and subsurface interaction between the adjacent treatments. There are two 104 mm diameter tile drains in each treatment which run parallel along the length of the plot at 0.6 m depth and 7.5 m spacing with a 0.08% slope. There were two water table management treatments (controlled drainage and subsurface irrigation (CDS) versus free drainage (FD) and four management practices (conservation tillage using a soil saver (SS) with/without an annual ryegrass intercrop (IC) and conventional moldboard plow tillage (MP) with/without an
Fig.
1.
Experi
men
tal
plot
and
dra
inage
layout
syste
m.
annual ryegrass intercrop) arranged in a 2 by 4 factorial design replicated twice for a total of 16 plots (Fig. 1). The soil saver treatment was a modified chisel plow.
2.2. Soil and crop management practices
The experiment was initiated in the spring of 1991 on a Brookston clay loam soil. Corn (Zea maysL., Pioneer 3573) was seeded with a Kinze four row planter at a rate of 65,000 seed per ha in a 0.75 m wide between rows. Fertilizer (8-32-16) was banded beside the seed at a rate of 132 kg haÿ1. Annual ryegrass intercrop was seeded within corn rows at 14 kg haÿ1with a Brillion seeder. Urea was applied as a side dress application with a customized designed brush applicator at the 6th leaf stage of corn based on the average NO3ÿ soil test of soil samples collected on the day of planting. The sidedress N
application rates were 115, 140, 190 and 170 kg N haÿ1for 1991, 1992, 1993 and 1994, respectively. The N rate was reduced by 55 kg N haÿ1 in 1991 to account for the N released from the previous alfalfa crop.
Typical herbicide application methods and materials were used in this study. Atrazine (6-chloro-N-ethyl-N0-(1-methylethyl)-1,3,5-triazine-2,4-diamine) at 1.1 kg haÿ1, metola-chlor (2-metola-chloro-N-(2-ethyl-6-methyl-phenyl)-N-(2-methoxy-1-methylethyl) acetamide) at 1.68 kg haÿ1, and metribuzin (4-amino-6-(1,1-dimethylethyl)-3-(methylthio)-1,2,4-tria-zin-5(4H)-one) at 0.5 kg haÿ1were banded over the corn row immediately after planting to control weeds.
2.3. Collection and measurements of surface runoff and tile drainage water
Surface runoff and tile drainage water from the 16 individual plots flowed into sump holes by gravity (Fig. 1). A sump pump was installed in each of the 32 sump pits and when the water level rose, a float sensor activated the sump pump and the water was pumped through a water meter into an outlet drain. A multichannel datalogger was used to monitor and store the water meter signals. The data stored in the datalogger were converted into flow volumes (Tan et al., 1993).
Surface runoff and tile drainage water samples were collected automatically with 32 autosamplers (CALYPSO 2000S, Buhler Gmbh) located in an instrumentation building. The autosampler was activated by the water meter sensor based on a predetermined setting of the flow volume which ranged from 500 to 3000 l depending on the time of year. Generally, wet conditions prevailed from late fall to spring that resulted in higher volumes of surface runoff and tile drainage from the plots. The dry season usually occurred in late spring and summer and smaller volumes of surface runoff and tile drainage from the plots were expected, hence sampling frequency was increased.
2.4. Soil properties
Soil samples were collected at 0±25, 25±45, 45±80 and 80±120 cm depths from each plot and field capacity, permanent wilting point, and particle size distribution were measured. Soil moisture contents at 20, 40, 60, 80 and 100 cm depths were determined during the growing season using a subsurface neutron moisture probe (Campbell Pacific
Model 503). Soil temperatures at 5, 10, 25, 40 and 60 cm depths were monitored year round using soil temperature probe (Campbell Scientific Model 107). Hydraulic conductivity was measured using an auger hole method and averaged over the 0± 120 cm depth. Soil organic carbon was determined from soil samples collected at 0± 15 cm depth. Soil samples at 20 cm depth increments to 100 cm were taken in spring and fall and from 0 to 30 and 30 to 60 cm depths at 0, 3, 7, and 42 days after N application were analyzed for NH4ÿN and NO3ÿ using a TRAACS 800 autoanalyzer
(BranLuebbe, Buffalo Grove, IL).
2.5. Analysis of nitrate concentration in water samples
Surface runoff and tile drainage water samples were collected and stored in glass bottles at 48C. These water samples were filtered through a 0.45mm filter (Gelman GN-6, Gelman Sciences, MI) and were analyzed for nitrate using cadmium reduction method (Tel and Heseltine, 1990; Drury et al., 1996) on a TRAACS 800 autoanalyzer (BranLeubbe, Buffalo Grove, IL). Flow weighted mean nitrate concentrations were calculated from the sum of nitrate loss over the study period from 1 November 1991 through 31 October 1994 divided by the sum of the total flow volume.
3. The LEACHM model
3.1. Model description
Leaching Estimation And Chemistry Model (LEACHM) is a process-based model developed by Hutson and Wagenet (1989)) that describes the water and solute movement, transformation, plant uptake and chemical reactions in unsaturated soils to a maximum depth of 2 m. The model applies numerical solution techniques to the Richard's water flow equation (Hutson, 1983) and the convection dispersion equation (CDE) using finite difference methods. The LEACHM model contains four modules:
1. LEACH-W simulates only the water regime,
2. LEACH-N simulates nitrogen transport and transformation,
3. LEACH-P simulates the pesticide displacement and degradation, and 4. LEACH-C simulates the transient movement of inorganic ions.
The above modules are capable of simulating the absorption of water and solutes by plant roots, precipitation and evaporation. The following inputs are common to all of the four modules.
± Soil properties and initial conditions for each soil segment including water content or water potential; hydrological constants for calculating retentivity and hydraulic conductivity; particle size distribution, and the relevant soil chemical properties. ± Soil surface boundary conditions including irrigation and rainfall amounts; rates of application; mean temperatures and diurnal amplitudes (weekly means) if a temperature simulation is required; and potential evaporation (weekly totals).
± Crop details or control variable if no crops are present. The crop parameters include time of planting; root and crop maturity at harvest; root and cover growth parameters, and soil and plant water potential limits for water extraction by plants.
± Other constants were also used in determining lower boundary conditions, time steps, dispersion and diffusion coefficients, chemical reactions and output details. Some of these constants rarely require alteration. These are also listed in the data files and the user has the option of changing these constants.
3.2. Model limitations
There are limitations in LEACHM model. The model is not applicable to the following conditions:
± Profiles with unequal depth increments.
± Prediction of surface runoff water quantity and quality. ± Simulating plant responses to soil or environmental changes. ± Prediction of crop yields.
± Transport of immiscible fluids.
± Solute distributions and transport in 2 and 3 dimensional flux patterns.
± Runoff water effects on management practices and nitrate leaching are not simulated by the model.
± The model does not take into account macropore effects.
3.3. Model input and initial values
In order to apply LEACHM, the soil profile was divided into three horizontal segments of equal depth: 0±0.2, 0.2±0.4, and 0.4±0.6 m. Since the available data were measured at the soil surface and 0.6 m drain depth only, intermediate data were interpolated from the measured data. These data include water table depths, nitrate-N (NO3ÿÿN)
concentrations, soil properties, and hydrologic parameters. The data measured in December 1991 were used as initial values for January 1992 and those measured in December 1992 for January 1993. The water table depths, surface soil temperatures, and evaporation data were summarized to fit the weekly input format.
The initial NO3ÿÿN soil profile concentrations were calculated from the drain water
concentrations according to the equation:
NO3ÿNs
NO3ÿNw=
(1)
Where [NO3ÿN]w concentration (mg N lÿ1 water), [NO3ÿÿN]sNO3ÿÿN
con-centration (mg N kgÿ1 dry soil), bulk density of the soil layer (g cmÿ3), and
qvolumetric water content (cmÿ3cmÿ3).
This procedure assumes that the NO3ÿÿN concentration in the soil profile is equal to
that in the drain water. This assumption may not always be true, but it provides a logical way to estimate the soil NO3ÿÿN concentration when measured soil data are not
available.
3.4. Model calibration and prediction
Model calibration and prediction were carried out in two stages. The first stage was that the data for 1992 were used to calibrate the model and to do sensitivity analysis of model's key parameters. Available observed data between January and December 1992 were used in the calibration processes. The second stage was that the datasets of 1993± 1994 were used for validation of the calibrated parameters of the model. Donigian (1983) suggested that it is an acceptable procedure to use one year's data to calibrate a simulation model and then apply the calibrated parameters to subsequent periods.
4. Results and discussion
4.1. Tile nitrate loss Ð field data
The field tile nitrate loss data were reported previously (Drury et al., 1996), however, since the first year of this data set was used to calibrate the model and the 2nd and 3rd year of the data were used to validate the model the following is a summary of these results. Tile drainage samples (5801) were collected from November 1991 until 31 October, 1994. The CDS treatment reduced tile drainage volumes by 24%, nitrate concentration in tile drainage water by 25% (from 10.6 to 7.9 mg N lÿ1) and average nitrate loss by 43% (from 25.8 to 14.6 kg N haÿ1yearÿ1) compared to the FD treatments. There were no significant differences between crop management treatments however it was noted that the lowest nitrate loss (11.6 kg N haÿ1yearÿ1) was achieved with the combination of CDS and the conservation tillage treatment (CDS-SS).
4.2. Model calibration
The calibration process was to obtain the initial values for model parameters that would estimate NO3ÿÿN concentrations closest to the observed values in the drain flow.
The main parameters were the transformation rate constants for urea hydrolysis, ammonia nitrification, and the denitrification processes. Also, included in the calibration were the molecular diffusion coefficient (Do), which accounts for the movement of solute in
response to aqueous concentration gradients, and dispersivity (), which describes the effects of the soil porosity on the overall solute transport. Since saturated hydraulic conductivity and soil bulk density are the major parameters used by LEACHM to distinguish between tillage practices, these parameters were included in the calibration process. The field measurements did not provide conductivity values for each soil layer. Hence direct comparisons of MP effects and the SS practices were not feasible. The saturated conductivity (Ks) and soil bulk density measured in 1991 for field plots are
listed in Table 1. The MP practice exerts greater disturbance to the soil than the SS treatment which results in greater porosity in the tilled layer. This results in higher hydraulic conductivity for soil under MP treatment (Table 1).
In this study, the parameters of molecular diffusion coefficient, dispersivity, hydraulic conductivity, soil bulk density and urea hydrolysis in LEACHM are considered to be key
Table 1
Saturated hydraulic conductivity (Ks) and soil bulk density for field plots (1991)
Duplicate plot Nos. 1 and 13 2 and 14 3 and 15 4 and 16 5 and 9 6 and 12 7 and 10 8 and 11
Management practices MP-IC CDS MP-IC FD SS-IC CDS SS-IC FD MP-FD SS-FD SS-CDS MP-CDS
Ks(mm/day) 58.0 30.5 36.5 24.5 52.5 26.0 65.0 93.5
Bulk density 0±25 cm 1.18 1.18 1.22 1.22 1.18 1.22 1.22 1.18
(g/cm3)
25±60 cm 1.46 1.46 1.46 1.46 1.46 1.46 1.46 1.46
118
H.Y
.F
.
Ng
et
al.
/
Agricultur
al
W
ater
Management
43
(2000)
Table 2
Sensitivity tests for selected parameters in LEACHM modela
Parameter Value 6
Observed 21.67 14.16 13.06 13.75 11.85 9.77 6.67 4.65 3.63 4.76 4.64
Diffusion coefficient (mm2/day) 60.00 23.20 17.00 13.30 12.20 8.65 5.33 4.58 3.27 2.45 1.37 0.00
120.00 23.20 17.00 13.30 12.20 8.69 5.37 4.63 3.31 2.49 1.42 0.00 Dispersivity(120) 150.00 23.20 17.10 13.30 12.20 8.71 5.39 4.65 3.34 2.51 1.44 0.00 Dispersivity (mm) 120.00 23.20 17.00 13.30 12.20 8.69 5.37 4.63 3.31 2.49 1.42 0.00 60.00 23.10 16.60 12.70 11.50 7.63 4.26 3.49 2.10 1.46 0.51 0.00 Diffusion coefficient(120) 10.00 23.00 9.32 7.00 5.25 2.24 1.19 0.58 0.27 0.18 0.13 0.09 Bulk density (g/cm3) 1.00 23.2 17.3 13.5 12.2 8.67 5.37 4.42 3.21 2.46 1.38 3.09E-05
1.18 23.20 17.00 13.30 12.20 8.69 5.37 4.63 3.31 2.49 1.42 0.00 1.30 23.1 16.9 13.3 12.2 8.75 5.41 4.91 3.51 2.64 1.58 1.87E-05 Hydraulic conductivity (mm/day) 10.00 23.4 23.7 22.6 21.5 19.8 18.2 16.8 17 18.4 16 0.00207
58.00 23.20 17.00 13.30 12.20 8.69 5.37 4.63 3.31 2.49 1.42 0.00 100.00 23.00 15.2 12.2 11.5 7.73 4.29 4.77 2.59 2.23 1.68 4.06E-05
Parameter Value 15
Observed 2.95 6.82 2.63 3.65 2.48 0.97 0.46 0.74 0.63
Diffusion coefficient (mm2/day) 60.00 0.07 0.03 0.00 0.03 0.11 0.16 0.18 0.21 0.30
120.00 0.07 0.03 0.00 0.03 0.12 0.18 0.20 0.23 0.33
Dispersivity(120) 150.00 0.07 0.03 0.00 0.03 0.12 0.19 0.21 0.24 0.35
Dispersivity (mm) 120.00 0.07 0.03 0.00 0.03 0.12 0.18 0.20 0.23 0.33
60.00 0.05 0.02 0.00 0.00 0.01 0.01 0.01 0.01 0.01
Diffusion coefficient(120) 10.00 0.05 0.05 0.05 0.04 0.02 0.01 0.01 0.01 0.02 Bulk density (g/cm3) 1.00 0.0938 1.2 3.96E-05 0.11 0.451 0.659 0.713 0.77 0.932
1.18 0.07 0.03 0.00 0.03 0.12 0.18 0.20 0.23 0.33
1.30 0.266 2.51 2.77E-05 0.0523 0.118 0.162 0.173 0.211 0.387
Hydraulic conductivity (mm/day) 10.00 0.00175 0.0202 2.64E-07 0.00296 0.0286 0.077 0.0926 0.112 0.16
58.00 0.07 0.03 0.00 0.03 0.12 0.18 0.20 0.23 0.33
100.00 1.04 4.05 0.0444 0.371 0.588 0.592 0.716 1.28 1.8
aObserved data from plots 1 and 13 (MP-IC-CDS) were used in the test. Data year1991. NO
Table 3
Testing the sensitivity of the transformation rate constants for LEACHM model using observed data from plots 2 and 14 (MP-IC-FD)
Parameter Rate
Observed 23.19 21.08 19.34 20.87 19.53 17.76 14.72 13.51 16.06 18.41 18.08
Urea hydrolysis 0.36
Nitrification 0.3 21.5 21.5 21.5 21.5 21.4 21.4 21.4 21.3 21.3 21.2 12.7
Denitrification 0.1
Urea hydrolysis 0.16
Nitrification 0.30 21.5 21.5 21.5 21.5 21.4 21.4 21.4 21.3 21.3 21.2 12.7
Denitrification 0.10
Urea hydrolysis 0.16
Nitrification 0.1 21.5 21.5 21.5 21.5 21.4 21.4 21.4 21.3 21.3 21.2 12.7
Denitrification 0.1
Urea hydrolysis 0.16
Nitrification 0.1 21.5 21.5 21.5 21.5 21.4 21.4 21.4 21.3 21.3 21.2 12.7
Denitrification 10
Urea hydrolysis 0.4
Nitrification 0.3 21.5 21.5 21.5 21.5 21.4 21.4 21.4 21.3 21.3 21.2 12.7
Parameter Rate
Observed 7.91 8.11 3.11 2.46 0.90 0.74 0.74 1.48 1.52
Urea hydrolysis 0.36
Nitrification 0.3 0.00209 0.448 1.27 1.81 2.65 4.07 4.53 5.35 8.2
Denitrification 0.1
Urea hydrolysis. 0.16
Nitrification 0.3 0.00147 0.349 0.993 1.45 2.17 3.41 3.81 4.54 7.1
Denitrification 0.1
Urea hydrolysis 0.16
Nitrification 0.1 0.000765 0.205 0.585 0.881 1.35 2.18 2.45 2.95 4.7
Denitrification 0.1
Urea hydrolysis 0.16
Nitrification 0.1 0.000765 0.205 0.585 0.881 1.35 2.18 2.45 2.95 4.7
Denitrification 10
Urea hydrolysis. 0.4
Nitrification 0.3 0.00219 0.459 1.3 1.85 4.2 12.4 13.5 15.8 19.6
in influencing model output. Such parameters in LEACHM were selected together with observed data from plots 1 and 13 under CDS and plots 2 and 14 under MP-IC-FD management practices, for sensitivity tests. The initial values contained in plots 1 and 13 and 2 and 14 are presented in Tables 2 and 3.
Initial tests indicated that the model did not respond to effects of varying calibration parameters in the winter period. The urea hydrolysis constant was varied between 0.1 dayÿ1 and 0.4 dayÿ1, nitrification constant between 0.1 dayÿ1 and 0.3 dayÿ1, and denitrification between 0.1 dayÿ1and 10 dayÿ1. After urea side dress application on the 29th June 1992, a hydrolysis value of 0.4 dayÿ1increased NO3ÿÿN concentrations in
drainage water and by December it was 10 times more than the observed. When the nitrification constant was increased to 0.3 dayÿ1, the simulated NO3ÿÿN concentrations
increased seven times between June and December. Values of denitrification above 0.1 dayÿ1 had no noticeable effects on simulated concentrations in the free drainage treatments. This is probably due to the drier soil conditions that exist under free drainage. After several trials, values of 0.36, 0.1, 0.1 dayÿ1 were chosen for urea hydrolysis, nitrification, and denitrification processes, respectively, as the predicted and measured values were in good agreement using these constants. The model predictions were consistent with no increase in urea hydrolysis for all runs following the application of urea. But a gradual increase in the predicted output between July and December suggested that the model assumes continuous hydrolysis of urea. Increasing the hydrolysis rate, however, tended to increase the deviation of the predicted NO3ÿÿN
leaching from the observed one.
The sensitivity of the model to molecular diffusion coefficient was tested usingDo
values of 60, 120, and 150 mm2 dayÿ1. Using these values, no significant differences were observed in the simulated leaching. Since it is rare to encounter Do values of
150 mm2 dayÿ1 in the field (Schulin et al., 1987; Van Der Ploeg et al., 1995), higher values were not tested. The model was not sensitive to dispersivity () values between 10 and 80 mm, the range in which the model underestimated leaching, but a value of 120 mm increased the nitrate leaching by almost 10 times which improved the models predictions. Hence thevalue of 120 mm was used in the model testing phase.
Increasing the soil bulk density between 1.0 and 1.30 g cmÿ3 tended to decrease leaching after fertilizer application on 14 May, 1992 but its effects were not noticeable in the winter period. Increasing the saturated hydraulic conductivity between 10 and 100 mm dayÿ1 resulted in about 50% reductions in leaching in the winter, but in the summer a Ks value of 100 mm dayÿ1 increased leaching by up to 10 times. These
observations underscore the need to have measured data for these parameters, especially in cases where tillage practices are being evaluated.
Other parameters found to be important in LEACHM performance include precipitation rates, water table depths, and evaporation rates. Precipitation rates in the range of 100 mm dayÿ1, representing precipitation duration of 2±4 h, gave better leaching estimates. LEACHM assumes that no evaporation occurs during precipitation. It was also not possible to differentiate between snow melt and actual rainfall from the available dataset.
Under free drainage conditions, the model predictions were consistent with no increase in urea hydrolysis after urea application on 29 June, 1992. The model predictions tended
to have greater deviations from the observed data in June. These deviations were also observed towards the end of the simulation period in December.
4.3. Parameters used in the model testing phase of the study
Designated parameters for prediction of scenario from the calibration process, are listed in Table 4. These values were chosen after several runs of the model. They gave the best estimates for nitrate leaching for 1992. The simulated results and observed values, using 1992 data, are plotted in Figs. 2 and 3.
The results of simulation of the nitrate leaching by LEACHM, using the parameters listed in Table 4 together with the observed data for 1993 and 1994 are plotted in Figs. 4 and 5. The results indicated that the model performed well under CDS conditions, but under free drainage conditions (FD) the model overestimated NO3ÿÿN leaching. The
deviation becomes more pronounced with time, especially after N application and progresses in the summer months throughout the fall, winter and spring to the following summer of 1994 (Fig. 5). This overestimation may be caused by the inability of the
Table 4
Key parameters selected from LEACHM for calibration and simulation
Parameter definition Value
Crop management
Plant uptake 102/167akg N haÿ1
N fertilizer application rate 10.56 kg N haÿ1
Urea sidedress application rate 141.22/189.5/170.72kg N haÿ1
Rate constants
Denitrification 0.1 dayÿ1
Half saturation (at 50%) 10 mg lÿ1
Urea hydrolysis 0.36 dayÿ1
Nitrification 0.1 dayÿ1
Soil
Soil layer thickness 0.2 m
Molecular diffusion 120 mmbdayÿ1
Dispersivity 120 mm
Aev value in Campbell's equation ÿ0.1 kPa Value ofbin Campbell's equation 3.0/3.5/4.0c
Soil temperature response
Q10factor 3
Soil moisture response
Saturation activity 0.6
Porosity 50%
aCorn/annual ryegrass uptake, respectively. bRates for 1992/1993/1994, respectively. cSoil layers: 0-0.2/0.2-0.4/0.4-0.6 (m).
Aevair enter value (Hutson and Case, 1987).
Q10factorsoil temperature response to a 108C change of optimal temperature.
Fig. 2. Nitrate concentrations in tile drainage water with CDS treatments. Study period 1992.
124
H.Y
.F
.
Ng
et
al.
/
Agricultur
al
W
ater
Management
43
(2000)
Fig. 3. Nitrate concentrations in tile drainage water with FD treatments. Study period 1992.
H.Y
.F
.
Ng
et
al.
/
Agricultur
al
W
ater
Manageme
nt
43
(2000)
111±131
Fig. 4. Nitrate concentrations in tile drain water with CDS treatments. Study period 1993±1994.
126
H.Y
.F
.
Ng
et
al.
/
Agricultur
al
W
ater
Management
43
(2000)
Fig. 5. Nitrate concentrations in tile drainage water with FD treatments. Study period 1993±1994.
H.Y
.F
.
Ng
et
al.
/
Agricultur
al
W
ater
Manageme
nt
43
(2000)
111±131
model to distinguish between snow-melt and rain or subirrigation. In such a situation the model may predict larger flows through the soil resulting in increased NO3ÿÿN
transport to the drains. It is possible that the model simulated drier soil profile conditions which inhibit denitrification, and subsequently lead to higher N amounts in the soil available for leaching. Khakural and Robert (1993) performed tests on LEACHM±N and observed that the model simulated total leaching of NO3ÿÿN from the soil profile very
well. However, Jemison et al. (1994) found that LEACHM±N only performed well when the rate constants were calibrated for each year, a requirement that would be very time consuming.
Under CDS conditions, there were no differences in model performance between MP and SS tillage systems with IC, but the model overestimated leaching between February and August 1994 under MP without an annual ryegrass intercrop. Overall the model is not sensitive to immediate leaching following fertilizer applications as shown by model underestimation of leaching after 17 May, 1993 the date when fertilizer was applied. The model output indicated that the SS tillage system, would reduce nitrate leaching. This is expected because MP tillage has a greater water conductivity rate which enhances leaching especially under saturated flow conditions.
Under FD conditions, the model predicted highest nitrate leaching in MP without an intercrop, and no notable differences between MP-IC, SS-IC and SS treatments. In all treatments, the model overestimated NO3ÿÿN leaching, with greater deviations
between predicted and actual data after September, 1993. A further calibration, using consecutive multi-year data may be required to improve the prediction for the FD treatments.
4.4. Comparison of the model predictions with the field observations
To determine how well a model performs, the model outputs are compared to the measured independent dataset that was not used in the calibration process. It follows that the model input data and rate constants are adjusted within a range of measured values until a minimum difference between measured and predicted values is obtained. Methods for achieving this include tests of means and variances, analysis of variances, mean error difference and goodness of fit testing (Harrison, 1990; Loague and Green, 1991; Power, 1993). In this study, method of mean error difference, Eq. (2) was used.
M Pin
i1 OiÿPi
n (2)
whereMis the mean error difference in mg N lÿ1,nis the number of observations,Oiand Piare individual observed and predicted values (mg N lÿ1), respectively. A mean error
difference (MED) which approaches zero is an indication of a good predictive model. Overall there was no significant difference between the average MED of 1.26 mg N lÿ1 for the 1992 dataset and the average MED of 1.28 mg lÿ1for the 1993 dataset (Table 5). The model overestimated the predictive values by about 1.2%. This suggests that the model accurately predicts NO3ÿÿN leaching with independent dataset using a
single-year calibrated rate constant (1992 data) for a factorial design of management treatments. It also suggests that the model should be calibrated using yearly data.
Table 5
Error mean difference of nitrate concentration (mg lÿ1) between the model (predicted) and observed data for tile drainage events from 1992±1994
Treatments MP-IC-CDS MP-IC-FD SS-IC-CDS SS-IC-FD MP-FD SS-FD SS-CDS MP-CDS Treatment Averages
January±December 1992 ÿ0.072 0.776 ÿ3.68 1.69 9.03 5.84 ÿ2.65 ÿ0.85 1.26 January±December 1993 ÿ2.21 3.09 ÿ1.8 5.95 2.46 6.26 ÿ2.03 ÿ1.46 1.28
January±August 1994 0.426 18.6 0.461 20 26.5 20 0.61 3.31 11.2
Mean across years ÿ0.618 7.475 ÿ1.67 9.22 12.7 10.7 ÿ1.36 0.332
H.Y
.F
.
Ng
et
al.
/
Agricultur
al
W
ater
Manageme
nt
43
(2000)
111±131
The average mean error difference of 11.24 mg lÿ1of 1994, using partial data series from January to August 1994 showed significant difference compared to the average mean error difference of 1.26 mg lÿ1 of 1992 dataset. It was about nine times overestimated by the model. Thus, modelling of NO3ÿÿN leaching for at least 1 year is
required. In temperate regions, most of the cropping cycle occurs between May and October. There is a 6-month non-cropping period from November to April. Drury et al. (1996) found that most NO3ÿÿN loss occurs during the noncropped period, both in
surface runoff and in subsurface tile drainage. Nitrogen lost in surface runoff is considerably less than from tile drainage.
The LEACHM model performed a better prediction of nitrate leaching on plots under CDS treatments (Table 5). The average values of mean error difference, regardless of the tillage and cropping treatments, for MP-IC-CDS, SS-IC-CDS, SS-CDS, and MP-CDS, respectively, were 0.43, 0.46, 0.61 and 0.33 mg N lÿ1(Table 5). Both the predicted values using the LEACHM model and the field measured data showed that CDS decreases NO3ÿÿN leaching. Thus, both the field observations and the model predictions indicate
that CDS is a new technology which can be used to reduce nitrate leaching from tile drained agricultural soils.
5. Conclusions
The simulated results indicated that free drainage (FD) management systems would result in higher nitrate leaching whereas controlled drainage and subsurface irrigation (CDS) showed reduced nitrate leaching. The LEACHM model performed better predictions for nitrate leaching on plots under CDS than on plots under FD. For all the four management treatments plots with CDS showed smaller values of mean error difference compared with those plots under FD. The calibration process using one year of data appeared to be acceptable. However, application of partial data series for model prediction resulted in significantly higher values of mean error difference.
Acknowledgements
The authors wish to thank the Great Lakes Water Quality Program for the financial support of this project. We also express appreciation for expert technical assistance to M. Soultani, Dr. T. Oloya, D. MacTavish, K. Rinas, J. St. Denis, V. Bernyk, G. Stasko and J. Stowe, and to the farm crew S. Mannel, A. Szabo, and M. Bissonnette at the Eugene F. Whelan Experimental Farm.
References
Coote, D.R., MacDonald, E.M., Dickson, W.T., Ostry, R.C., Frank, R., 1982. Agriculture and water quality in the Canadian Great Lakes Basin: I. Representative agricultural watersheds. J. Environ. Qual. 11, 473±481. Donigian, A.S. Jr., 1983. Model predictions vs. field observations: The model validation/testing process. In:
R.L., Swan, Eschenroeder, A. (Eds.), Fate of Chemicals in the Environment, Compartmental and Multimedia Models for Predictions. Am. Chem. Soc., Washington, D.C., pp. 151±171.
Drury, C.F., Tan, C.S., Gaynor, J.D., Oloya, T.O., Welacky, T.W., 1996. Influence of controlled drainage-subirrigation on surface and tile drainage nitrate loss. J. Environ. Qual. 25, 317±324.
Great Lakes Water Quality Board, 1987. Reports on Great Lakes Water Quality to the International Joint Commission, Appendix A, Windsor, Canada and Detroit, Michigan, p. 208.
Harrison, S.R., 1990. Regression of a model on real-system output: An invalid test of model validity. Agric. Syst. 15, 183±190.
Haynes, R.J., Cameron, K.C., Goh, K.M., Sherlock, R.R., 1986. Mineral nitrogen in the plant-soil system. Academic Press, p. 483.
Hutson, J. L., 1983. Estimating hydrological properties of South African soils. Ph.D. Thesis. University of Natal, Pietermaritzburg, South Africa.
Hutson, S.R., Case, A., 1987. A retentivity function for use in soil-water simulation models. J. Soil Sci. 38, 105± 113.
Hutson, J.L., Wagenet, R.J., 1989. LEACHM: Leaching Estimation and Chemistry Model, Department of Crop and Atmospheric Sciences, Cornell University, Ithaca, N.Y. (updated 1992).
International Joint Commission, 1978. Contribution of phosphorus from agricultural land to stream by surface runoff. Task Group C (Canadian Section). Activity 1.
Jemison, J.M., Jabro, J.D., Richard, H.F., 1994. Evaluation of LEACHM: II. Simulation of nitrate leaching from nitrogen-fertilized and manure corn. Agron. J. 86, 852±859.
Jones, R.D., Schwab, A.P., 1992. Nitrate leaching and nitrite occurrence in a fine-textured soil. Soil Sci., 155, 4. Khakural, B.R., Robert, P.C., 1993. Soil nitrate leaching potential indices: Using a simulation model as a
screening system. J. Environ. Qual. 22, 839±845.
Loague, K., Green, R.E., 1991. Statistical and graphical methods for evaluating solute transport models: Overview and application. J. Contam. Hydrol. 7, 51±73.
Polglase, P.J., Tompkins, D., Stewart, L.G., Falkiner, R.A., 1995. Mineralization and leaching of nitrogen in an effluent-irrigated pine plantation. J. Environ. Qual. 24, 911±920.
Porter, K.S., 1975. Nitrogen and phosphorus - food production, waste and the environment. Ann Arbor Science, Ann Arbor, MI.
Power, M., 1993. The predictive validation of ecological and environmental models. Ecological Modelling, vol. 68. Elsevier, Amsterdam, pp. 33-50.
Schulin, R., van Genuchten, M.Th., Flhhler, H., Ferlin, P., 1987. An experimental study of solute transport in a stony field soil. Water Resour. Res. 23, 1785±1794.
Sittig, M., 1991. Handbook of Toxic and Hazardous Chemicals and Carcinogens. Noyes Publications, Park Ridge, NJ, USA.
Tan, C.S., Drury, C.F., Gaynor, J.D., Wellacky, T.W., 1993. Integrated soil, crop and water management system to abate herbicide and nitrate contamination of the Great Lakes. Water Sci. Technol. 28, 497±507. Tel, D.A., Heseltine, C., 1990. The analyses of KCL soil extracts for nitrate, nitrite and ammonium using a
TRAACS 800 analyzer. Commun. Soil Sci. Plant Anal. 21, 1681±1688.
Van Der Ploeg, R.R., Ringe, H., Machulla, R., 1995. Late fall site-specific soil nitrate upper limits for ground water protection purposes. J. Environ. Qual. 24, 725±733.
