| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
a International Water Management Inst., PO Box 2075, Colombo, Sri Lanka
b USDA-CPRL-ARS, PO Drawer 10, 2300 Experiment Station Rd., Bushland, TX 79012
c Dep. Soil, Water and Climate, 1991 Upper Buford Cir., 439 Borlaug Hall, Univ. of Minnesota, St. Paul, MN 55108
d Dep. Bioproducts and Biosystems Engineering, 1390 Eckles Ave., Univ. of Minnesota, St. Paul, MN 55108
* Corresponding author (v.nangia{at}cgiar.org).
Received for publication May 4, 2007.
ABSTRACT
Nitrate losses from subsurface tile drained row cropland in the Upper Midwest U.S. contribute to hypoxia in the Gulf of Mexico. Strategies are needed to reduce nitrate losses to the Mississippi River. This paper evaluates the effect of fertilizer rate and timing on nitrate losses in two (East and West) commercial row crop fields located in south-central Minnesota. The Agricultural Drainage and Pesticide Transport (ADAPT) model was calibrated and validated for monthly subsurface tile drain flow and nitrate losses for a period of 1999–2003. Good agreement was found between observed and predicted tile drain flow and nitrate losses during the calibration period, with Nash-Sutcliffe modeling efficiencies of 0.75 and 0.56, respectively. Better agreements were observed for the validation period. The calibrated model was then used to evaluate the effects of rate and timing of fertilizer application on nitrate losses with a 50-yr climatic record (1954–2003). Significant reductions in nitrate losses were predicted by reducing fertilizer application rates and changing timing. A 13% reduction in nitrate losses was predicted when fall fertilizer application rate was reduced from 180 to 123 kg/ha. A further 9% reduction in nitrate losses can be achieved when switching from fall to spring application. Larger reductions in nitrate losses would require changes in fertilizer rate and timing, as well as other practices such as changing tile drain spacings and/or depths, fall cover cropping, or conversion of crop land to pasture.
HYPOXIA in the Gulf of Mexico affected an area of 17,500 km2 during 2006 (LUMCON, 2006). A reduction in nitrate loading by 30% has been recommended to reduce hypoxia in the Gulf of Mexico (Mitsch et al., 1999). Nitrate loadings from the Upper Mississippi River Basin (UMRB) account for roughly 35% of the nitrate entering the Gulf of Mexico (Alexander et al., 1995), yet this area covers less than 20% of the Mississippi River Basin. The UMRB is characterized by an extensive area of Mollisols managed with subsurface tile drainage systems that are used primarily for row crop production. Nitrate concentrations in the Mississippi River are generally greatest in the tributaries emanating from Illinois, Iowa, and Minnesota (Antweiler et al., 1995) where artificially drained soils planted to corn and soybean dominate the landscape (Burkart and James, 1999). Omernik (1977) reported that total N concentrations were nearly nine times greater downstream from agricultural lands than downstream from forested areas, with the highest concentrations being found in the Corn Belt states. It is important to identify and evaluate agricultural management strategies that are capable of reducing nitrate loadings from agricultural systems in the Midwest to attain improved oxygen levels in the Gulf of Mexico.
Management practices to improve water quality can be divided into agronomic management practices and nitrogen removal practices (Dinnes et al., 2002). Considerable agronomic management research has been conducted at the plot scale to evaluate the effects of drain spacing and depth, N fertilizer application rate and timing, crop rotation, or climatic variability on the quality and quantity of drainage (Randall and Mulla, 2001; Dinnes et al., 2002). Nitrogen removal practices include planting buffer strips adjacent to streams and ditches, fall planting of cover crops, restoration of wetlands and wholesale conversion of row cropped fields to perennial cover (Mitsch et al., 2001; Boody et al., 2005). Mitsch et al. (2001) estimated that reductions in nitrate loads to the Gulf of Mexico of 300,000–800,000 metric tonne/yr could be achieved by creating or restoring wetlands and riparian buffers on 0.7–1.8% of the land in the Mississippi River Basin. These reductions compare with their estimates of a 900,000 to 1400,000 metric tonne/yr reduction in nitrate loads as a result of better N fertilizer management throughout the Basin.
Higher nitrate losses are associated with higher N application rates (Baker and Johnson, 1981), and with fall versus spring application (Baker and Melvin, 1994). Many replicated plot scale studies have been conducted in the Upper Midwestern U.S. to experimentally measure reductions in nitrate losses through tile drains in response to alternative fertilizer management strategies (Randall and Mulla, 2001; Baker and Johnson, 1981; Dinnes et al., 2002; Randall et al., 2003). Weed and Kanwar (1996) demonstrated that the amount of nitrate found in the tile drainage from a loamy soil in Iowa was highly influenced by crop rotation, but not by tillage practice. This is mainly due to application rates of N fertilizer that are greater for grain crops than for legume crops. Randall et al. (2003) concluded in a study on tile-drained Canisteo clay loam soil that nitrate N losses from a corn–soybean rotation with subsurface drainage can be reduced by 13 to 18% by either applying N in the spring or using nitrapyrin (NP) with late-fall applied ammonia.
Attempts have been made to extrapolate experimental results for nitrate leaching at the plot scale to different temporal scales using tile drain simulation models (Davis et al., 2000; Zhao et al., 2000). Davis et al. (2000) calibrated and validated the ADAPT model using tile drainage and associated nitrate losses measured on three long-term experimental plots in Minnesota under continuous corn with conventional tillage. The experimental plots were located on poorly drained Webster clay loam soil (mesic Typic Haplaquols). Davis et al. (2000) found that a decrease in the N application rate from 225 to 175 kg/ha decreased nitrate losses by 48%.
Results from these plot scale studies have been used to estimate regional impacts of alternative fertilizer management practices on nitrate losses at scales that are vastly greater than those at which the studies were conducted (Mitsch et al., 2001). There is a pressing need to evaluate the impact of alternative fertilizer management practices at the field and watershed scales. There have been few studies at the field or watershed scales in the Upper Midwest to evaluate nitrate losses in response to alternative fertilizer management practices. At these scales, replication of experimental treatments is difficult, and spatial and temporal variability make the interpretation of trends in nitrate losses difficult to evaluate. For this reason, researchers attempting to evaluate the impact of N management practices on water quality at coarse scales have often combined experimental and modeling studies. For example, Jaynes et al. (2004) conducted a paired watershed study in Walnut Creek watershed in Iowa as a function of N fertilizer application rate. One portion of the watershed was managed with fertilizer application rates typical of Midwestern corn production, another received split N fertilizer application rates based on a late spring nitrate test (LSNT). Jaynes et al. (2004) showed that use of the LSNT approach reduced nitrate concentrations in tile drainage by 29%. Baksh et al. (2004) used the Walnut Creek watershed data with the Root Zone Water Quality Model (RZWQM) to estimate nitrate losses as a function of N fertilizer application rate. They found that reducing the N application rate from 175 to 125 kg/ha resulted in a 22% decrease in nitrate losses.
Several simulation models have been developed to simulate surface and subsurface agricultural water quality. Examples of such models are AGNPS (Young et al., 1994), SWAT (Arnold et al., 1998), CREAMS (Knisel, 1980), GLEAMS (Leonard et al., 1987), ADAPT (Agricultural Drainage and Pesticide Transport; Chung et al., 1992), LEACHM (Hutson and Wagenet, 1992), RZWQM (USDA-ARS, 1992), and DRAINMOD (Skaggs and Broadhead, 1982). Some of these models do not account for all the major hydrological processes that occur in the Midwestern U.S. such as tile drainage and snow-melt. For example, the simulation models CREAMS, GLEAMS, NLEAP, and LEACHM do not have tile drainage algorithms and the NLEAP and LEACHM models do not account for frozen soil hydrology including snow-melt runoff during the spring. The ADAPT model is a daily time step, field scale water table management model that was developed by integrating GLEAMS, a root zone water quality model, with subsurface drainage algorithms from DRAINMOD. More detailed information about ADAPT can be found in Chung et al. (1992), Ward et al. (1993), Gowda (1996), and Desmond et al. (1996).
Detailed evaluation of simulation models is necessary before their use for practical purposes, and this is often achieved by calibration and validation. This helps to determine whether the model produces rational results compared to observed data. It also provides information on shortcomings of models and additional processes/factors to be considered. Long- term monitoring data are required for calibration and validation of water quality simulation models. The ADAPT model has been calibrated and validated for various hydrologic conditions in the Midwest (Desmond et al., 1995; Chung et al., 1992; Gowda et al., 1999; Sogbedji and McIsaac, 2002a; 2002b; 2006). All of the latter studies, excepting Desmond et al. (1995), evaluated the ADAPT model for situations which involved measured streamflow and/or nitrate loads at the mouth of watersheds in the absence of any experimental fertilizer or tile drainage treatments in the watershed. Limited efforts have been made to evaluate tile drainage flow models in the presence of data involving experimental agricultural management practice treatments applied at the field or watershed scales (Zhao et al., 2000).
The main objectives of this study were to use water quality data collected in south-central Minnesota on two commercially farmed fields with experimental nitrogen fertilizer rate and timing treatments to: (i) calibrate and validate the ADAPT model for monthly subsurface tile drainage and associated nitrate losses, and (ii) determine sensitivity of nitrate losses to fertilizer application rates and timing.
Materials and Methods
Site Description
The calibration and validation of the model for subsurface tile drain flow and nitrate losses were performed using water quality measurements made on two fields of a commercial farm with a corn [Zea mays (L.)]- soybean [Glycine max (L.) Merr.] rotation. The site is located 8 km southwest of St. Peter, Minnesota (Fig. 1
). It is set up such that a 21-ha field is split roughly in half [west field = 11 ha (213 m x 540 m) and east field = 9.3 ha (174 m x 535 m)]. The site is dominated by poorly drained clay loam soils that developed under tall prairie grasses in glacial till. Soils in the study area included Cordova (Typic Argiaquolls), Cordova-Rolfe (Typic Argialbolls), Canisteo (Typic Haplaquolls), Le Sueur (Aquic Argiudolls), Harps (Typic Calciaquolls) and Okoboji (Cumulic Haplaquolls). The site is drained with concrete tile drains installed 15–30 yr ago at 30 m spacings and 1.1 m depths with an average slope of 0.3%. The diameter of tile drains was 152 mm. The average annual precipitation in the region is about 737 mm. Table 1
presents average monthly precipitation, temperature and potential evapotranspiration (PET) data for 1999–2003. Approximately 75% of the total drainage occurs in April, May and early June. The growing season typically lasts from mid/late May until early/mid October. Snow starts to melt in late March or April and high flows are observed at monitoring sites during the April-June period.
|
|
|
ADAPT uses a detailed pseudo-mechanistic approach for estimating nitrogen fate and transport, including mineralization of soil organic matter, immobilization, nitrification and denitrification, volatilization, crop uptake and N fixation (legumes only), leaching and losses in drainage and runoff. N mineralization is considered as a two-stage process in ADAPT: the first stage being a first-order ammonification process; and the second a zero-order nitrification process. The default potential mineralization rate constant value used in ADAPT is 0.003 kg/ha/day, while the potential nitrification rate constant has a value of 100 mg NO3–N/kg soil/week. Ammonification occurs from the active soil N, fresh organic N from root and surface residue, and organic N in animal waste. The two soil organic carbon pools are based on carbon:nitrogen (C:N) ratios. Mineralization rates depend on potential mineralization rates, modified by temperature and soil water factors. ADAPT considers mineralization not only from soil organic matter, but also from crop and root residues. Immobilization of nitrogen as nitrate or ammonia is estimated by ADAPT based on fresh residue mass, concentration of nitrogen in the residue and a decay rate which is a function of C to N ratio of the residue, soil water content and temperature. Denitrification is estimated by ADAPT as a function of soil nitrate concentration, a decay coefficient, and temperature and soil moisture factors. Denitrification occurs only when soil moisture content is 10% above field capacity water content.
ADAPT estimates uptake of nitrogen as either nitrate or ammonia based on concentrations of nitrate or ammonia in soil layers, daily nitrogen demand of the crop and root uptake of water from different soil layers. Daily nitrogen demand of the crop is estimated based on total dry matter nitrogen in biomass as a function of concentration of nitrogen in biomass, changes in leaf area index, potential crop yield and the ratio of total dry matter to harvestable yield. For leguminous crops, ADAPT estimates uptake using the same approach described above only when the concentrations of nitrate and ammonia in soil solution exceed 5 mg/L. If the concentrations of nitrate and ammonia are less than 5 mg/L, N fixation occurs in an amount needed to satisfy the daily nitrogen demand. Further details of the nutrient component of the ADAPT model can be found in Knisel et al. (1993). All default N process rate constants in ADAPT were used without any calibration.
Model Inputs
Model simulations were made using climatic data from 1994–2003. Precipitation was measured on site using a tipping bucket rain gauge during 1999–2003 (Table 1). Precipitation data for the remaining years and other climatic data such as daily values of average temperature, solar radiation, wind speed, and average relative humidity were taken from the nearby St. Peter weather station. Subsurface tile drain flows were measured from 1999 onward at a 1-min frequency using an ISCO area-velocity meter (ISCO, Inc., Lincoln, NE) and outputs were 15-min average discharges. Water quality samples were taken from 1999 onward using automated sampling equipment during storm events and grab samples were collected during base flow conditions. Samples were measured for nitrate, total phosphorus, ortho-phosphorus, fecal coliform and E-coli bacteria, turbidity, and total suspended solids. The fields were planted with a corn–soybean rotation under conventional tillage making it very typical of the upper Midwest region cropping system. Since nutrient management data were available at the site from 1994 to 2003, model simulations were conducted starting from 1994.
For corn, di-ammonium phosphate (DAP) and urea were broadcast and anhydrous ammonia was injected. A variable rate N application study (Montgomery et al., 2000) was performed at the site from 1997 to 1999. For this purpose, the field was divided into multiple strips receiving different rates of fertilizer N. For corn, strips received N fertilizer at rates of 61, 101, 146, and 179 kg N/ha. Details of planting and harvesting dates, fertilizer application rates and tillage operations implemented during 1994–2003 are presented in Table 2 .
|
|
In the THRU formation process, spatial data layers of variable N application rates for 1997 and 1999 and soil types were overlaid using ArcView 3.0 GIS software (ESRI, Redlands, CA) to capture the variability in N fertilizer application rate against soil type. Differences in application rates across the field during the variable application rate study were handled in the model setup by treating areas with unique N application rates as separate HRUs. N fertilizer application rates for other years were uniform throughout the fields and were not spatially overlaid for THRU formation. The result was a GIS layer consisting of 11 THRUs for the west (calibration) field and 15 THRUs for the east (validation) field containing unique combinations of soil type and N fertilizer application rates.
Model Calibration and Validation
Since rigorous sampling methodology for measuring water quality monitoring data were instituted after 1999, the first 3 yr (1999–2001) of high quality monitoring data were used for calibration and the remainder of the data (2002–2003) were used for validation of the ADAPT model using monthly subsurface tile drainage and nitrate losses. The model was validated again using independent flow and water quality monitoring data from the east field. We modified parameters one at a time to check sensitivity of output to their change. We searched for optimum values of parameters in increments of 5% between specific lower and upper bounds, based on literature and default values available. The model was calibrated by varying hydrologically sensitive parameters such as saturated vertical hydraulic conductivity (Table 3), rooting depth, leaf area index, drainage coefficient, and soil moisture retention curves to achieve the closest agreement between predicted and observed subsurface tile drainage and nitrate losses.
Other parameters modified during the calibration of ADAPT, included soil freeze/thaw, soil storage, runoff, and crop growth parameters. These parameters affected the prediction of both ET and surface runoff. It is important to note that there were no observed ET data at the study site. Crop ET was indirectly adjusted by increasing the final leaf area index (LAI) coefficient by 30% (final LAI = initial LAI*1.3). The LAI database built into the ADAPT model is for older cultivars with lower biomass and crop yields that grew in locations different from Minnesota. Table 4 lists the parameters that were adjusted during model calibration.
|
Although the model was continuously run for the entire simulation period (1999–2003), observed data were not available for comparison during winter and fall months of some years (measured data were missing for: Jan., Sept.-Dec., 1999; Jan.-April and Aug.-Dec., 2000; Jan.-Mar. and Aug.-Dec., 2001; Jan.-Mar., Aug.-Sept. and Nov.-Dec., 2002; Jan.-Mar. and Aug.-Dec. 2003). Flow and water quality data are not collected in these months because tile flows are generally low to nonexistent, as a result of either frozen soils or limited rainfall. As a result, measures of model performance are a comparison of the months in which observed data were available. Although the ADAPT model is capable of predicting runoff and subsurface tile drainage during winter and early spring conditions, evaluation of model performance for these events was not possible due to a lack of measured data.
Performance Criteria
Four statistical procedures were used to assess the level of agreement between the predicted and observed data for calibration years:
 |
 |
 |
 |
 |
Index of agreement is a measure of the degree to which the predicted variation precisely estimates the observed variation. The value of d is unity when there is a perfect agreement. Nash-Sutcliffe efficiencies can range from-
to 1. An efficiency of 1 (E = 1) corresponds to a perfect match between modeled values and observed data. An efficiency of 0 (E = 0) indicates that the model predictions are as accurate as the mean of the observed data, whereas an efficiency less than zero (- < E < 0) occurs when the observed mean is a better predictor than the model. Essentially, the closer the model efficiency is to 1, the more accurate the model is.
Nitrogen Fertilizer Application Rate and Timing
Long-term simulations (1954–2003) were made to determine sensitivity of nitrate losses to changes in N application rates and timings. Input parameters used in the simulations for evaluating sensitivity of nitrate losses were the same as those used in the model calibration and validation. Alternative management practices included three different N application rates (0, 123, and 180 kg N/ha) and three different timings- fall, spring pre-plant, and 50% in spring pre-plant and 50% in fall.
Results and Discussion
Model Calibration
Table 5
shows good agreement between predicted and measured subsurface tile drainage and nitrate losses for the calibration and validation periods. In the calibration phase, attempts were made to minimize the RMSE and obtain E and d values closest to unity. Comparison of measured and calibrated values for monthly subsurface tile drainage (Fig. 3
) shows that the model underpredicted drainage in early spring. This is primarily due to the difficulty in predicting the onset of subsurface tile drain flow during spring snowmelt runoff. Statistical evaluation of the monthly predicted and observed subsurface tile drain flow gave an E value of 0.75. The index of agreement was about 0.92 and the RMSE was 48% of the observed mean monthly subsurface drainage (Table 5). Table 6
compares monthly observed and predicted subsurface tile drainage and Nash-Sutcliffe efficiencies for the calibration period. For a majority of the months, E values are close to 1, showing good agreement between observed and predicted values. In June, 1999 and May and June, 2001, the E values are negative, suggesting that observed mean drainage is a better predictor than the model. We can conclude from these results that the model performed reasonably well in predicting subsurface tile drainage during the non-snowmelt period.
|
|
|
|
Model Validation
Validation on the West Field
Comparison of measured and predicted values of monthly subsurface tile drainage for validation years (2002–2003) on the west field shows (Fig. 5
) that the magnitude and trend in the predicted monthly subsurface tile drainage closely followed that of measured data in most months. There was fair agreement between predicted mean monthly subsurface tile drain flows of 1.1 mm/day and measured subsurface tile drain flows of 0.8 mm/day (Table 5). The model overpredicted subsurface tile drain flows partly due to errors in the prediction of timing and magnitude of snowmelt events in early spring. A comparison of predicted and measured monthly subsurface tile drain flows values gave an E value of 0.67 and the index of agreement was about 0.91. Table 6 indicates that the E values were close to unity for a majority of the period, with some exceptions in May and July, 2002.
|
|
|
|
The crop nitrogen uptake was 149.5% of the N applied in fertilizer because results presented here are for a corn–soybean rotation with no N fertilizer application in soybean cropping years. N fixed by the soybean crop is not considered applied fertilizer, but can be taken up by the corn crop. Average nitrogen fixation was 109.8 kg/ha which compares well with rates of 80–100 kg/ha determined by Johnson et al. (1975) and more recent estimates of 100 kg/ha for Illinois conditions (Hoeft and Peck, 2002). Soils high in organic matter can mineralize substantial amount of nitrate, which is susceptible to loss in subsurface tile drainage (Randall and Mulla, 2001). The predicted mineralization was 64.4 kg/ha which compares well with 69.8 kg/ha predicted by Davis et al. (2000) on a nearby Minnesota soil with 60 g/kg organic matter. The predicted annual average nitrate loss through subsurface tile drains (59.6 kg/ha) was about 46% of the applied N and about 1.4% higher than the measured nitrate losses (57.8 kg/ha). The predicted nitrate loss by denitrification was about 10.3% of the total N applied, which is comparable to estimated values (10–25%) reported by Meisinger and Randall (1991).
Effects of Alternative Fertilizer Management Scenarios Based on a 50 Year Climate Record
Nitrogen Fertilizer Application Rate and Timing
Decreases in N fertilizer application rate resulted in reductions in nitrate losses (Fig. 9
). For example, annual predicted nitrate losses decreased from 50.4 kg/ha to 43.7 kg/ha when fall applied N was decreased from 180 kg/ha to 123 kg/ha. This is a 13% decrease in nitrate losses for a 32% reduction in N fertilizer rate. Further reductions in nitrate losses were predicted when fall N applications were switched to spring or split N application timings. For a N application rate of 180 kg/ha, the model predicted a reduction of 9% (from 50.4 to 45.9 kg/ha) when application timing was changed from fall to spring. Reductions in nitrate losses were also predicted at N application rates other than 180 kg/ha. Averaged across twenty-five rotation cycles, the lowest nitrate losses were found with reduced rates of N fertilizer applied during spring. Overall, reductions in N application rate had a bigger impact on nitrate losses than switching N application from fall to spring.
|
It is worthwhile to note that reducing N application rates to zero did not eliminate nitrate losses in subsurface tile drainage. Even when no N fertilizer was applied, nitrate losses of about 27 kg/ha were predicted. It appears that losses less than this are not possible for this site and cropping system because the source of this nitrate should be mineralized soil N that was fixed in the soybean portion of the crop rotation.
The results of this modeling study are consistent in trend with those obtained by Davis et al. (2000) and Baksh et al. (2004), who also showed significant reductions in nitrate losses from tile drainage after reducing N fertilizer application rate. In the present study, we obtained a 13% reduction in nitrate losses by decreasing spring N fertilizer application rates from 180 kg N/ha to 123 kg N/ha. In the Davis et al. (2000) Minnesota study, a 93% reduction in nitrate losses was obtained by reducing spring N fertilizer application rates from 175 kg N/ha to 125 kg N/ha. In the Baksh et al. (2004) Iowa study a 22% reduction in nitrate losses was obtained by reducing spring N fertilizer application rates from 175 kg N/ha to 125 kg N/ha.
These different magnitudes of reduction can partly be explained by differences in cropping system, scale and method for collecting experimental data in the two Minnesota studies. The Davis et al. (2000) study involved a continuous corn rotation where N fertilizer is applied every single year, and there is no soybean crop or N fixation. In contrast, the present study is for a corn–soybean rotation in which there is carryover of some N fixed by the soybean crop and N fertilizer is applied every other year. The Davis et al. (2000) study involved experiments at the plot scale (13 m x 15 m), whereas the present study is for measurements collected at the field scale (9–11 ha). The measured values of subsurface tile drain flow and nitrate losses on which the model results in the two studies are based may differ as a result of these scale issues. Finally, the subsurface tile drain flow and nitrate concentration measurements in the two studies were collected using two different methods. In the Davis et al. (2000) study, measurements of nitrate were analyzed from grab samples collected weekly, while flow was measured daily. In the present study, measurements of nitrate were collected using both grab samples and automated samplers during storm events, while flow was measured at 15 min intervals. Thus, the experimental data in the present study includes water quality information during peak flows, whereas the Davis et al. (2000) study does not. For all the reasons mentioned above, results from the present modeling study are an improvement on the results from Davis et al. (2000).
Conclusions
The ADAPT model was calibrated and validated for monthly subsurface tile drainage and associated nitrate losses on two commercial fields with a corn–soybean rotation under conservation tillage for the period 1999–2003. The predicted monthly subsurface tile drain flows and nitrate losses agreed reasonably well with the measured trends for both calibration and validation periods. Validation results on the east field gave better statistics than validation results on the west field. Comparison of water and nitrogen budgets against measured data and the literature showed that the model accurately partitions water and nitrogen.
The calibrated model was also used to evaluate the effects of changes in rate and timing of fertilizer application on nitrate losses. Simulation results indicated that reductions in nitrate losses are possible by reducing N fertilizer application rate. A 13% reduction in losses was found when fall N application rate was reduced from 180 kg/ha to 123 kg/ha. Further reductions in nitrate losses were obtained by changes in timing of N application. Changing the N application timing from fall to spring at an application rate of 180 kg/ha resulted in a 9% reduction in nitrate losses. Changes in timing and amount of N fertilizer applications may help reduce nitrate loads to the Gulf of Mexico. However, attaining a 30% or greater reduction in nitrate losses to the Gulf may require other alternative management practices such as changes in tile drain spacing and/or depth, planting cover crops in fall, restoration of wetlands, or conversion of cropland to pasture.
ACKNOWLEDGMENTS
The field operations were carried out by anonymous farmers and personnel from the Minnesota Department of Agriculture. The assistance of Brian Williams from the Minnesota Department of Agriculture in supplying farm management data and water quality monitoring data is appreciated. Financial support for computer modeling was provided by a grant from the National Science Foundation Biocomplexity program.
NOTES
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
