VZJ Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 24 January 2007
Published in Vadose Zone J 6:53-66 (2007)
DOI: 10.2136/vzj2006.0065
© 2007 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Hong, N.
Right arrow Articles by Cassel, D. K.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Hong, N.
Right arrow Articles by Cassel, D. K.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Hong, N.
Right arrow Articles by Cassel, D. K.
Related Collections
Right arrow Vadose Zone Processes and Chemical Transport
Right arrow Ground Water Quality

ORIGINAL RESEARCH

Groundwater Nitrate Spatial and Temporal Patterns and Correlations

Influence of Natural Controls and Nitrogen Management

Nan Honga,*, Jeffrey G. Whiteb, Randy Weiszc, Marcia L. Gumpertzd, Miressa G. Dufferab and D. Keith Casselb

a Div. of Plant Science, Univ. of Missouri, Columbia, MO 65211
b Dep. of Soil Science, North Carolina State Univ., Raleigh, NC 27695-7619
c Dep. of Crop Science, North Carolina State Univ., Raleigh, NC 27695-7620
d Dep. of Statistics, North Carolina State Univ., Raleigh, NC 27695-8203
* Corresponding author (hongn{at}missouri.edu)

Received 1 May 2006.



   ABSTRACT
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
To use shallow groundwater NO3–N concentration as an indicator of groundwater quality requires understanding its patterns, correlations, and controls across space and time. Within a study comparing variable-rate and uniform N management, our objectives were to determine groundwater NO3–N patterns and correlations at various spatial and temporal scales and their association with natural controls and N management. Experiments in a random, complete block design were conducted in a 2-yr crop rotation in North Carolina that included one variable-rate and two uniform N management treatments to wheat (Triticum aestivum L.) and corn (Zea mays L.). We measured groundwater NO3–N and depth every 2 wk at 60 well nests, sampling the 0.9- to 3.7-m depth. Field-mean NO3–N varied with time from 5.5 to 15.3 mg NO3–N L–1. These variations were correlated primarily with concurrent changes in water table elevation and depth. Mean NO3–N exhibited two preferred states: high when the water table was shallow and low when the water table was deep. Temporal NO3–N fluctuations greatly exceeded treatment effects. Treatments appeared to affect NO3–N temporal covariance structure. Groundwater NO3–N spatial patterns and correlations were associated mostly with saturated hydraulic conductivity and water table fluctuations and appeared influenced by subsurface lateral flow. When treatment effects became consistently significant later in the study, they overrode natural controls, and NO3–N was spatially uncorrelated or exhibited shorter spatial correlation ranges and patterns associated predominantly with treatments.

Abbreviations: AR(1), first-order autoregressive covariance model • CS, compound symmetry covariance model • FA, remote sensing informed, in-season, uniform, field-average N management • Go, Goldsboro soil series • Ksat, saturated hydraulic conductivity • Ly, Lynchburg soil series • No, Norfolk soil series • RYE, uniform realistic yield expectation N management • SOM, soil organic matter • SS, remote sensing informed, in-season, site-specific, variable-rate N management • VR-N, variable-rate N management • WTD, water table depth • WTE, water table elevation


   INTRODUCTION
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
GROUNDWATER NO3–N is an indicator of water quality that has been used in agricultural research to assess the effectiveness of N management strategies in reducing groundwater NO3 contamination (Casey et al., 2002; Hong et al., 2006). Crop growth, N uptake, and N fertilizer requirement vary temporally among and within seasons (Baethgen and Alley, 1989), and spatially among and within fields (Ferguson et al., 2002). In-season, site-specific, variable-rate N management applies appropriate spatially and temporally variable N fertilizer rates at critical crop growth stages; thus, it attempts to match crop N needs and optimize fertilizer N-use efficiency while optimizing N inputs and minimizing losses to groundwater. Groundwater NO3–N spatial and temporal patterns and relationships with landscape, soil, and groundwater properties beneath agricultural fields have not been well characterized. This is especially true in precision agricultural experiments testing VR-N.

Landscape–soil–water systems are naturally complex due to their spatial and temporal heterogeneity (Grayson et al., 1997). Therefore, NO3 leaching through these systems is also complex and scale dependent across space and time. Consequently, groundwater NO3–N will probably exhibit distinct spatial and temporal patterns. In addition, it may be spatially and temporally correlated and have variances and covariances that change with time. Characteristics of spatial patterns in groundwater NO3–N can be defined quantitatively as: (i) no organization, that is, observations are spatially random with no or random controls; (ii) large-scale organization, where observations exhibit smooth and continuous patterns with large spatial correlation ranges that are associated with large-scale controls such as topography and the predominant surface and subsurface lateral water movement; and (iii) small-scale organization, where observations exhibit sharp and discontinuous patterns with shorter spatial correlation ranges that are associated with small-scale controls such as localized soil properties and microtopography. Characteristics of a temporal pattern can be defined quantitatively by determining the temporal correlation structure of observations.

Hergert et al. (1995) reported that soil NO3–N at the 0.3- to 1.2-m depth in a crop field was spatially correlated with spatial correlation ranges varying from 40 to 275 m. Bruckler et al. (1997) found that spatial correlation of soil NO3–N varied with time in an irrigated salad crop field under normal farming conditions. In addition, the spatial pattern of soil NO3–N has been reported to vary with time (Cahn et al., 1994). Recent research demonstrated that NO3–N exhibited spatial and temporal patterns in soil (Ghidey and Alberts, 1999; Cain et al., 1999; Eghball et al., 2003), in shallow (<2-m depth) groundwater in a riparian buffer zone (Dhondt et al., 2002), in pristine and agriculturally influenced prairie streams (Kemp and Dodds, 2001), and in deep (≤15.8-m) vadose zone soils below a small (0.8-ha) irrigated orchard in semiarid California (Onsoy et al., 2005).

We hypothesized that under agricultural fields with uniform N fertilizer management, spatial patterns and temporal dynamics of groundwater NO3–N would be associated primarily with natural controls that affect NO3 movement. These natural controls include soil physical properties, surface elevation, water table depth, and subsurface lateral flow. With the introduction of VR-N, these patterns and correlations might be disrupted and instead reflect the spatial and temporal variability in N applications. Within a study comparing uniform and VR-N management, our objectives were to determine: (i) shallow groundwater NO3–N spatial and temporal patterns and correlations at various spatial and temporal scales during a 2-yr period; and (ii) any association of these patterns and correlations with natural controls or VR-N and uniform N management.


   MATERIALS AND METHODS
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
This work was part of a long-term (>5 yr) precision agriculture experiment conducted in the southeastern U.S. Coastal Plain. While the nature of the study requires that we describe the N management treatments used in the experiment (see below), our primary emphasis was on groundwater NO3–N spatial and temporal patterns, correlations, and controls during a 2-yr period, and not specifically on how the N management treatments affected groundwater NO3–N. The effects of the N management treatments on groundwater NO3–N were reported in detail elsewhere (Hong et al., 2006).

Study Site
The study site was in a relatively flat agricultural landscape at the Lower Coastal Plain Tobacco Research Station, Kinston, NC. The experiment was conducted in two adjacent fields totaling 12 ha, which we considered to be one field (Fig. 1 ). The elevation gradient in the field was 1.34 m. An Order 1 soil survey (North Carolina Agricultural Experiment Station, 1977) delineated three soil map units in this field: Norfolk (No) loamy sand with 0 to 2% slope (fine-loamy, siliceous, thermic Typic Paleudult), Goldsboro (Go) loamy sand (fine-loamy, siliceous, thermic Aquic Paleudult), and Lynchburg (Ly) sandy loam (fine-loamy, siliceous, thermic Aeric Paleaquult). A detailed characterization of particle size distribution in this field (not shown, Duffera et al., 2006), however, revealed that the Ap texture was predominantly sandy loam with a significant proportion of somewhat finer textures in the Ly soil. The No loamy sand is well drained, the Go loamy sand is moderately well drained, and the Ly sandy loam is somewhat poorly drained. These soils are representative of millions of hectares of farmland in the southeastern U.S. Coastal Plain.


Figure 1
View larger version (43K):
[in this window]
[in a new window]

  		Fig. 1. Field layout of the randomized complete block design at the Lower Coastal Plain Tobacco Research Station, Kinston, NC. Shading patterns indicate plots within the same block. The N management treatment number is in the upper left corner of each plot. Nitrogen treatments are: RYE = uniform realistic yield expectation N management; FA = remote sensing informed, in-season, uniform, field-average N management; SS = remote sensing informed, in-season, site-specific, variable-rate N management.

 
A typical stratigraphic description of this field indicates a predominantly sandy clay loam texture transitioning to sandy clay at ~2.6-m depth (North Carolina Agricultural Experiment Station, 1977). There is a clay bed at ~2.6- to 2.9-m depth that has low permeability. During periods of high rainfall, water will perch above the clay, forming a zone of saturation up to 1.8-m depth or less. The major horizontal (lateral) water flow probably occurs at the zone of a basal bed of medium coarse sand at ~2.9- to 3.7-m depth. The basal sand bed transitions in color from yellow to yellowish red through ~5.2 m, where an abrupt lower boundary indicates the base of the Wicomico morphostratigraphic unit and the start of the red to black compact loam of the Pee Dee formation. The field has a tile drainage system with lines at about 1-m depth spaced ~30 m apart oriented parallel to the northern and southern field boundaries and exiting into orthogonal tile lines across the field center and on the eastern side. Consistent with current best management practices for field crops in this region, we equipped the outlets of the tile lines with drainage control structures to maintain water tables as high as possible to foster denitrification of groundwater NO3–N, but not so high as to hinder crop production or field operations (Osmond et al., 2002).

Experimental Setup and Treatments
The experiment was established in a randomized complete block design with three treatments replicated 10 times. The treatments were applied in 30 large (0.37-ha) square plots. The three treatments were uniform realistic yield expectation N management (RYE); remote sensing informed, in-season, uniform, field-average N management (FA); and remote sensing informed, in-season, site-specific VR-N (SS).

The RYE treatment followed the current regulation-mandated N management system for the North Carolina Neuse and Tar-Pamlico river basins. The RYE rate was determined based on the published RYE and N-use factor for the predominant soil type (Go) in the field, derived from the North Carolina RYE database (North Carolina Nutrient Management Workgroup, 2003). Fertilizer N was applied uniformly to the RYE plots at planting and at the appropriate growth stages (Table 1). The FA rates were determined based on field-averaged in-season estimates of optimal N needs derived from color infrared aerial photography at critical growth stages (Flowers et al., 2003; Sripada et al., 2005; Hong et al., 2006). For wheat, this was at Zadoks' (Zadoks et al., 1974) growth stages 25 and 30 and for corn at tasseling (growth stage VT, Ritchie et al., 1993). Fertilizer N was applied uniformly to the FA treatment plots. The SS rates were determined based on site-specific estimates of N demand at critical growth stages using color infrared aerial photography. Each SS plot was divided into 78 4.6- by 9.1-m "miniplots" for spatially variable-rate N application. The SS N rates were determined by averaging N demand within each miniplot and fertilizing each miniplot accordingly (Hong et al., 2006). Mean N rates and times of application for these treatments during 3 yr of the 2-yr crop rotation are summarized in Table 1. As a result of these treatments, we applied spatially and temporally variable rates of N fertilizer at the field scale owing to the spatial arrangement of treatments, and at the plot scale in the SS treatment due to the spatially variable N inputs within the SS miniplots.


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

  		Table 1. Rate and timing of N fertilizer applications for the three N management treatments for wheat and corn in the 2-yr winter wheat–double crop soybean (Year 1)–corn (Year 2) rotation.

 
Aqueous urea–NH4NO3 (30% N) was applied to wheat using a variable-rate sprayer, and to corn by dribbling it into row middles using high-clearance, variable-rate equipment with drop nozzles. In 2002 in cases where corn row spacing prevented mechanized operations adjacent to well nests, seeding and fertilization were done manually.

Groundwater Sampling
In March 2001, two well nests were installed in each of the 30 treatment plots, making a total of 60 well nests. One nest in each plot was placed at the plot center to minimize the influence of neighboring plots. The other was placed randomly within each plot with the constraints of being an adequate distance from the plot center, at least 4.7 m from the plot edge, and in line from plot to plot parallel to crop rows to facilitate field operations. Each well nest consisted of three polyvinyl chloride pipe groundwater-sampling wells screened to sample 0.9- to 1.8-, 1.8- to 2.7-, and 2.7- to 3.7-m depths. We sampled to 3.7-m depth to capture groundwater within the basal sand bed and to ensure that water samples were available throughout the season. Before sampling, each well was purged at least three times the volume of water in the well using a peristaltic pump. Once the well had recharged, a 25-mL sample was pumped into a plastic scintillation vial, which was capped. Samples were cooled for transport to the laboratory where they were acidified by adding a drop of 1.0 M H2SO4 and stored at –20°C until analysis. Groundwater samples were collected as available from each well from March 2001 until July 2003, and the water table depth (WTD), that is, the distance from the soil surface to the groundwater table, was measured from July 2001 until July 2003, both approximately every 2 wk or after significant rainfall. Water table elevation (WTE, i.e., water table height above mean sea level) at each well nest was calculated from the measured WTD and the wellhead soil surface elevations, which were derived from North Carolina Floodplain Mapping Program lidar data. The lidar data had an average post spacing of ~3 m and a vertical RMSE of ~16 cm (www.ncfloodmaps.com; accessed 12 June 2003, verified 19 Oct. 2006). Groundwater NO3–N was determined using an automated ion analyzer (QuikChem 8000, Lachat, Loveland, CO) using a Cd reduction method (Greenberg et al., 1992).

Soil Sampling and Precipitation Data
Before establishing fertilizer N treatments, surficial (0–20 cm) soil sampling was conducted on a 15-m equilateral triangle grid in November 2000 before wheat planting (Li et al., 2002). Soil organic matter (SOM) content was determined by loss on ignition (360°C) and was estimated at well nests by kriging. In this field, surficial SOM is well associated with soil drainage class; SOM is greatest in the somewhat poorly drained Ly soil, intermediate in the moderately well-drained Go soil, and lowest in the well-drained No soil (Li et al., 2002). In June 2003 after wheat harvest, we extracted relatively undisturbed soil cores to about 1-m depth by inserting a hydraulically driven soil tube (Giddings Machine Co., Windsor, CO). The soil cores were sectioned into depth increments that corresponded to the centers of time domain reflectometery probe profiling increments. Segments of 7.6-cm diameter by 7.6-cm length corresponding to depths of 4 to 12, 19 to 27, 34 to 42, 49 to 57, and 64 to 72 cm were cut and placed into soil cans immediately after core extraction, capped using air-tight plastic caps, and stored at 4°C until they could be processed. Saturated hydraulic conductivity (Ksat) of the intact cores was measured using a constant-head permeameter (Klute and Dirksen, 1986). For each section, we determined particle size distribution (sand, silt, and clay percentage) using the hydrometer method after pretreatment with sodium hexametaphosphate (Gee and Bauder, 1986). The results of these soil physical characterizations were reported in detail by Duffera et al. (2006). Daily precipitation data (Fig. 2 ) were collected at the experiment station rain gauge. Precipitation distribution was generally similar to the 30-yr average (1974–2003), but monthly precipitation was below average for 16 of the 26 mo in the study period, above average for 6 mo (not consecutive), and near average for 4 mo (Fig. 2).


Figure 2
View larger version (34K):
[in this window]
[in a new window]

  		Fig. 2. (a) Daily precipitation from November 2000 through July 2003 and (b) actual (November 2000–July 2003, bars) and 30-yr-average (1974–2003, line) monthly precipitation at the study site. Note that the beginning of the precipitation reporting period precedes the groundwater reporting period (Fig. 3) by 8 mo to provide the meteorological context for the latter.

 

Figure 3
View larger version (32K):
[in this window]
[in a new window]

  		Fig. 3. (a) Field-mean shallow groundwater NO3–N concentration sampled over the 0.9- to 3.7-m depth from July 2001 to July 2003; the reference line is the USEPA drinking water maximum contaminant level goal, 10 mg L–1 NO3–N; (b) field-average water table depth (0 m represents the soil surface); (c) NO3–N spatial correlation range; and (d) total (field-scale) and plot-scale NO3–N spatial variances. The field-scale spatial variance is analogous to the sill and the plot-scale variance analogous to the first lag interval variance of a semivariogram. Error bars indicate ±1 SE of all estimates.

 
Statistical Analysis
During the reporting period (July 2001–mid-July 2003), groundwater sample availability increased with sampling depth. Samples were frequently unavailable at the 0.9- to 1.8- and 1.8- to 2.7-m depths, corresponding primarily to periods of low rainfall (Fig. 2) or high evapotranspiration or both (April–August; Amatya et al., 1995). Samples were sometimes unavailable even at the 2.7- to 3.7-m depth, mostly from early September to late November in 2002, during which an average of 46% of data was unavailable. Thus, on a given sampling date, there could have been up to three samples at a well nest. We averaged groundwater NO3–N from the available samples at each well nest on individual dates. The resultant means represented groundwater NO3–N in the 0.9- to 3.7-m depth at each well nest and were used for statistical analyses, which consisted of temporal and spatial analyses conducted separately. The normality of sand, silt, and clay content, groundwater NO3–N, Ksat, SOM, WTE, WTD, and surface elevation was assessed by diagnostic plots (e.g., histograms) coupled with the Shapiro–Wilk test in SAS PROC CAPABILITY (SAS Institute, 2006). There was no strong evidence of nonnormality for the static variables sand, silt, and clay content, Ksat, SOM, and surface elevation, so the data were analyzed in the original scale. The degrees of normality of the dynamic variables groundwater NO3–N, WTD, and WTE varied with time between relatively normal and somewhat abnormal, so we assumed that the data were normally distributed. Our spatial analysis approaches (see below) are relatively robust to moderate deviations from normality (Littell et al., 1996). All tests described below were determined at the {alpha} = 0.05 significance level.

Temporal Analysis
During the experiment, fieldwide mean WTD and WTE and mean groundwater NO3–N fluctuated considerably. We divided the entire 2-yr sampling period (considered the long-term scale) into five phases (the short-term scale) based on these temporal fluctuations (Fig. 3a and 3b). During the 2-yr period, we plotted fieldwide mean groundwater NO3–N against time to detect temporal patterns.

The foundation of our repeated measures analysis is the general linear model, which takes the form:

Formula 1[1]
where y is an n x 1 vector of responses (i.e., groundwater NO3–N), X is an n x p design matrix for fixed effects, ß is a p x 1 vector of fixed-effects parameters, and {varepsilon} is a n x 1 vector of residual random errors, where n and p are the number of responses and fixed-effect parameters, respectively. The fixed effects examined included N treatment, treatment x date, and WTE and WTD as temporal covariates, which were examined individually. Repeated measures analysis was conducted in PROC MIXED SAS Version 9 (SAS Institute, 2006). Three error covariance structures were modeled: a nontemporal covariance structure that assumed independent and identically distributed errors, and two temporal covariance structures: compound symmetry (CS) and first-order autoregressive [AR(1)]. We chose these because groundwater NO3–N could be uncorrelated (independent and identically distributed errors model), equally correlated (CS model), or autoregressively correlated [AR(1) model]. An equal temporal correlation means that observations from the same experimental unit have the same correlation regardless of temporal adjacency, suggesting that a dominant factor or process is responsible for the temporal correlation of observations. First-order autoregressive temporal correlation takes the form of

Formula 2[2]
where yi and yj are observations from the same experimental unit measured at times i and j, respectively, {rho} is the autocorrelation coefficient, and {sigma}2 is the variance (SAS Institute, 2006). An AR(1) temporal correlation means that observations from the same experimental unit measured on any two contiguous dates have the same correlation, and this correlation decreases with increasing temporal separation. This suggests significant connectivity among observations over time in that an observation from an experimental unit on a date in question is strongly dependent on an observation or observations from the previous date or dates. The two temporal covariates (WTE and WTD) with the three covariance structures resulted in six candidate covariance models. We used the Akaike Information Criterion (Akaike, 1974) to select the best-fit covariance model among these. Within each phase (the short-term scale), we also used repeated measures analysis as described above to determine the relationships of groundwater NO3–N with WTD, WTE, and N management treatments.

Spatial Analysis
Groundwater NO3–N spatial patterns and their relationships with various landscape, soil, and groundwater properties and with N management treatments were determined for each individual date. In this analysis, we considered two spatial scales: the field and the plot. At the field scale, we considered the 60 well nests as the analysis unit, while at the plot scale, the analysis unit consisted of the two well nests within each treatment plot. At the field scale, groundwater NO3–N on each sampling date was mapped using inverse distance-squared weighted interpolation to visualize the spatial pattern and how it changed with time. For groundwater NO3–N observations on a given date, isotropic and anisotropic empirical semivariograms of Formula 2 (h), which take the form

Formula 3[3]
where h is the distance between two sampling sites ua and ua + h, r(ua) is an observation value at the ath location ua, and N(h) is the number of pairs of sampling sites a lag distance h meters apart (Goovaerts, 1997), were computed in SAS PROC VARIOGRAM (SAS Institute, Cary, NC) and in GS+ Version 7 (Gamma Design Software, Plainwell, MI). The exploratory semivariograms presented here were calculated with consideration restricted to half the maximum lag distance (Journel and Huijbregts, 1978). Semivariograms provided starting parameters for the quantitative analysis of spatial correlation of groundwater NO3–N on individual dates using the model in Eq. [1] in PROC MIXED. Seven covariance structures were modeled, one nonspatial covariance model assuming spatial independence of groundwater NO3–N and six isotropic spatial covariance models: the spherical, Gaussian, and exponential, all with or without a nugget effect. We chose these because of their applicability in describing spatial covariance structures commonly encountered in agriculture and soil science, and because of the forms of our exploratory semivariograms. Of these candidate models, the one with the best fit was selected, as described by Hong et al. (2005), and used to estimate the spatial correlation range and variance of groundwater NO3–N using the restricted maximum likelihood method. Note that PROC MIXED estimates spatial correlation parameters using the full spatial extent of the data and does not calculate a semivariogram; thus spatial correlation parameters estimated by semivariography and PROC MIXED typically differ somewhat (Littell et al., 1996).

Also at the field scale, we tested N management treatment effects on groundwater NO3–N with the following parameters considered individually as spatial covariates: surface elevation; mean sand, silt, and clay contents; mean Ksat of the 0- to 72-cm depth cores; surficial SOM; and WTE and WTD. The spatial analyses were conducted in PROC MIXED for each sampling date following the same procedure for selecting a best-fit spatial model as described above, but treating N management treatments as fixed effects and block and block x treatment as random effects. The foundation of this analysis is the general linear mixed model, which takes the form:

Formula 4[4]
where y, X, ß, and {varepsilon} are the same as defined in Eq. [1], Z is an n x q design matrix for the random block and block x treatment effects, and u is a q x 1 vector of random effects, where q is the number of random-effect parameters.

At the plot scale, we quantified groundwater NO3–N spatial variances on individual dates using Eq. [3] in PROC VARIOGRAM. Each treatment plot contained two well nests that were about 25 m apart; the variance between them (i.e., the within-plot variance) was equivalent to the first lag interval (i.e., 25 m) of a semivariogram. We hypothesized that if the SS treatment, which applied spatially and temporally variable rates of N fertilizer, did not affect the spatial variance of groundwater NO3–N, the within-plot variance for each treatment should be similar and consistent throughout the study period.


   RESULTS AND DISCUSSION
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Nitrate-Nitrogen Temporal Patterns, Correlations, and Relationships
General Trends
Fieldwide mean groundwater NO3–N varied from 6 to 15 mg L–1, averaging 11 mg L–1, typical of N-fertilized agricultural fields in the North Carolina Coastal Plain (Osmond et al., 2002). Mean groundwater NO3–N seasonally exceeded the USEPA drinking water maximum contaminant level goal (MCLG) of 10 mg NO3–N L–1 (USEPA, 2002); this occurred mainly in spring (Fig. 3a). Mean groundwater NO3–N exhibited temporal fluctuations that appeared to be associated with those of field-mean WTD, which varied from 0.95 to 3.40 m below the soil surface (Fig. 3b). When analyzed across all dates, mean NO3–N was negatively correlated with field-mean WTD and positively correlated with field-mean WTE, both averaged across all well nests on individual dates (for both: r2 = 0.72; P < 0.0001; n = 54). These results were in agreement with the finding by Terry and McCants (1970) that NO3 leaching in Coastal Plain agricultural soils was associated with the amount of percolating water moving to shallow groundwater.

Temporal Patterns
Groundwater NO3–N exhibited temporal patterns of low, high, and low-to-high and high-to-low transitional periods (Fig. 3a). These corresponded, respectively, with dry conditions when the water table was relatively deep, wet conditions when the water table was shallow, and the transitions of dry to wet and wet to dry when the water table rose or fell (Fig. 3b). During relatively dry periods (i.e., fall 2001 in Phase I and summer and fall 2002 in Phase III) when the mean WTD dropped below the 3-m depth, mean NO3–N tended to decrease to less than the USEPA MCLG of 10 mg L–1 NO3–N. In fall 2002, when the mean WTD was near 3.4 m, NO3–N reached the lowest mean concentration (~6 mg L–1) observed during the study. This may have been because vertical water movement was not sufficient to recharge shallow groundwater, thus no NO3–N leaching would be expected during this period. Other factors potentially contributing to low NO3–N when the water table was deep might be dilution by groundwater with less NO3–N arriving via lateral flow, or N loss via denitrification; however, denitrification would probably not be important 3 m below the soil surface in this field (Gambrell et al., 1975).

During relatively wet periods (i.e., spring 2002 in Phase II and spring and summer 2003 in Phase V), the mean WTD was <2 m and NO3–N exceeded 10 mg L–1 (Fig. 3a). When the mean WTD was <1.5 m in spring and summer 2003, the highest mean NO3–N concentrations during the study were observed, peaking at 15 mg L–1. This was probably due to leaching of high residual NO3–N from the 2002 corn crop (Hong et al., 2006; Fig. 3) and N applied to the 2003 wheat at Growth Stage (GS) 25 (Zadoks et al., 1974; Table 1). It is also probable that the rising shallow groundwater intercepted zones of relatively high NO3 in the upper soil profile.

Groundwater NO3–N experienced two dry-to-wet transitional periods: Phase I to II and Phase III to V (i.e., Phase IV). During the Phase I to II transition (about 1.5 mo), a rapid 1.44-m rise in the mean WTD was associated with a 1.71 mg L–1 increase in groundwater NO3–N. During the Phase IV transition (about 3 mo), a 1.69-m rise in the mean WTD was associated with a 6.11 mg L–1 increase in groundwater NO3–N. The difference between the two dry-to-wet transitional periods in the magnitudes of the increases in groundwater NO3–N was probably due to N management treatment effects reported by Hong et al. (2006), that is, leaching of high residual NO3 from the 2002 corn crop and from the GS 25 N application on 2003 wheat. Given that groundwater NO3–N probably also included leached mineralized soil organic N, the longer Phase IV transition probably resulted in a greater contribution to groundwater NO3–N from this pool as well. The end of Phase II through Phase III represented a wet-to-dry transitional period, during which an ~2-m drop in the mean WTD was associated with an ~6 mg L–1 decrease in groundwater NO3–N.

Temporal Correlations
Groundwater NO3–N was temporally correlated throughout the study period, but the covariance structures differed between the long- and short-term scales. At the long-term scale (i.e., the 2-yr sampling period), the AR(1) model with WTD as the temporal covariate gave the best fit with an autocorrelation coefficient of 0.65 (Table 2). This indicated that temporal fluctuations of groundwater NO3–N were associated with those of water table depths, and groundwater NO3–N on a given date tended to be dependent on that on the previous date or dates.


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

  		Table 2. Effects of N management treatments (Trt) and water table depth (WTD) on groundwater NO3–N concentrations using temporal covariance models during the entire 2-yr study, and for shorter time periods (phases). The best-fit model; the significance of treatment, treatment x date, and any significant covariate; and the estimated variances (Var) and covariances (Cov) are also indicated.

 
At the short-term scale (i.e., within each phase), groundwater NO3–N was also temporally correlated, but the covariance structures varied by phase (Table 2). For Phases I, II, and III, a CS model with neither WTD nor WTE had the best fit, indicating that groundwater NO3–N among dates within each phase had equal correlation regardless of temporal adjacency, and that this correlation was not associated with the water table. For Phase IV, the CS model with WTD resulted in the best fit. This implies that the temporal trend of groundwater NO3–N was associated with the water table during this phase, and that the portion of this trend not accounted for by WTD followed a CS covariance structure. The AR(1) model with no temporal covariate (WTD or WTE) gave the best fit in Phase V, with an autocorrelation coefficient of 0.63, very similar to that found at the long-term scale (0.65).

Temporal Relationships with Nitrogen Treatment and Natural Controls
At the long-term scale, N management treatments significantly affected groundwater NO3–N, but there was also a significant treatment x date interaction, indicating that the impact of treatments on groundwater NO3–N differed across the 2-yr time frame. These treatment effects on groundwater NO3–N were small (maximum of 2–3 mg L–1 NO3–N; data shown in Hong et al., 2006) relative to the temporal variations (Fig. 3a) that were associated with WTD.

Because the phases were defined primarily based on WTD, it was not surprising that WTD was not a significant covariate in Phases I, II, III, or V (Table 2). Phase IV was a transitional period that had the largest range in WTD compared with the other phases (Fig. 3a), which may explain why WTD was a significant covariate during that period. The spatial analysis (below) indicated that N treatments did not consistently affect groundwater NO3–N in Phases I, II, III or IV, but did so on 70% of the sampling dates in Phase V (Table 2). The relative consistency of treatment effects late in the study period was probably due to the cumulative residual effects of the N applications on 2002 corn in combination with the GS 25 application on 2003 wheat (Hong et al., 2006). The elevated and consistent impact of N treatments in Phase V may have been responsible, at least in part, for the change in the groundwater NO3–N temporal covariance structure during Phase V (Table 2).

The relationship between groundwater NO3–N and WTD suggests that, in these and similar Coastal Plain soils, shallow groundwater NO3–N may be somewhat "self regulated" by WTD. Water tables rise as a result of significant precipitation, which probably includes leaching events that carry NO3 to groundwater. Rising groundwater probably intercepts and absorbs NO3 from zones of high NO3. Coincident saturation–desaturation cycles enhance mineralization of N that may end up in groundwater. As water tables remain within the upper parts of the soil profile with high NO3–N, sufficient organic C, microbes, and favorable temperature, denitrification is enhanced and groundwater NO3–N decreases (Evans et al., 1996; Osmond et al., 2002). Given the observed temporal pattern of groundwater NO3–N, its association with WTD, and the concepts of preferred states in temperate-region spatial soil moisture patterns developed by Grayson et al. (1997), we posit that shallow groundwater NO3–N in coarse-textured temperate-region soils also exhibits preferred high and low states associated with shallow and deep shallow groundwater tables.

Nitrate-Nitrogen Spatial Patterns, Correlations, and Relationships
Spatial Patterns
Spatial patterns of fieldwide groundwater NO3–N varied within the 2-yr time frame (Fig. 4 ). Early in the experiment (i.e., Phases I and II), groundwater NO3–N exhibited smooth and continuous patterns (e.g., Fig. 4a, 22 Feb. 2002, Phase II). Groundwater NO3–N tended to be higher in the northern and western portions of the field and lower in the field center and near the southeastern edge. Toward the end of the study period (i.e., Phase V), groundwater NO3–N patterns tended to be sharp and discontinuous (e.g., Fig. 4b, 8 Apr. 2003, Phase V) in that groundwater NO3–N changed to a greater extent across a shorter distance. During this period, areas of relatively high groundwater NO3–N were distributed through a greater portion of the field. We presented these two dates because they were representative of the early and late spatial patterns of groundwater NO3–N, and because they were both on dates with relatively shallow water tables (i.e., Phases II and V; Fig. 3b).


Figure 4
View larger version (60K):
[in this window]
[in a new window]

  		Fig. 4. Spatial patterns of groundwater NO3–N concentrations sampled over the 0.9- to 3.7-m depth on (a) 22 Feb. 2002 and (b) 8 Apr. 2003. Nitrate-N concentrations were estimated using inverse distance-squared weighted interpolation. The N management treatments were: RYE = uniform realistic yield expectation N management; FA = remote sensing informed, in-season, uniform, field-average N management; SS = remote sensing informed, in-season, site-specific, variable-rate N management.

 
Spatial Correlations
At the field scale early in the study period as represented by 22 Feb. 2002 (Fig. 4a), groundwater NO3–N was spatially correlated (Fig. 5 , bottom curve) with relatively long spatial correlation ranges (mean = 265 m; Fig. 3c), which corresponded to the observed smooth and continuous spatial patterns of groundwater NO3–N. The spatial dependence of groundwater NO3–N observed early in this study was similar to that observed by Ella et al. (2001) from 23 shallow (6-m) wells sampling groundwater in glacial till in Iowa. The within-plot spatial variance (i.e., the variance between the two well nests in a treatment plot) was generally small and constituted an average of 31% of the total spatial variance (Fig. 3d). Late in the study period (i.e., Phase V) as represented by 8 Apr. 2003 (Fig. 4b), groundwater NO3–N was either spatially random or spatially correlated (Fig. 5, top curve) with a mean correlation range of 104 m. This spatial correlation range was much shorter than early in the experiment (Fig. 3c and 5) and close to the dimension of a treatment plot (61 m). The mean within-plot spatial variance constituted 70% of the total spatial variance (Fig. 3d), which corresponded to the observed sharp and discontinuous spatial patterns of groundwater NO3–N during this period. Temporal changes in the spatial correlation strength of soil solution NO3–N to 0.9 m were also observed by Bruckler et al. (1997). They reported that the spatial correlation of soil NO3–N based on 36 sampling points in a 0.24-ha plot varied from spatially structured to random in the course of a year in an irrigated salad crop field under normal farming conditions in the Mediterranean zone of France.


Figure 5
View larger version (15K):
[in this window]
[in a new window]

  		Fig. 5. Global (isotropic) semivariograms of groundwater NO3–N concentration on 22 Feb. 2002 (spherical model: nugget = 0.57, sill = 5.1, range = 196 m; R2 = 0.99, residual sum of squares = 0.03) and 8 Apr. 2003 (exponential model: nugget = 1.13, sill = 11.040, effective range = 112 m; R2 = 0.93; residual sums of squares = 1.7).

 
Anisotropy was another field-scale characteristic of groundwater NO3–N spatial correlations during the study period typified by 22 Feb. 2002. The anisotropic semivariogram of groundwater NO3–N sampled on 22 Feb. 2002 (Fig. 6a ) indicates that groundwater NO3–N exhibited substantial spatial correlation on the 40° axis (i.e., approximately southwest to northeast), and little if any on the 130° axis (i.e., approximately northwest to southeast). We posit that this anisotropy was associated primarily with subsurface lateral flow. Based on our understanding of the study field stratigraphy and drainage system, we believe that subsurface lateral flow occurs in two dominant paths: (i) to and through the drainage lines during periods of saturation within the tile zone, and (ii) within horizons of relatively high horizontal Ksat driven by the WTE gradient. As estimated by Hong et al. (2006), lateral flow rates in this field probably ranged from 0.6 to 5 cm d–1 in the upper 2.6 m, with rates several orders of magnitude lower in areas of high clay or silt, and up to an order of magnitude higher in the basal coarse sand. Nitrate-containing leachate from the upper part of the profile probably accumulated above the low-permeability firm clay layer at the 2.6- to 2.9-m depth in association with the rise of a transient perched water table. The WTE gradients in this field ranged from 1.3 to 2.9 m during the study period and maintained the general southwest-to-northeast orientation typified by the 22 Feb. 2002 WTE (Fig. 6b). Thus, any subsurface lateral flow was likely to have been from the southwest toward the northeast. The major axis of anisotropy (40°) was approximately parallel to this probable direction of lateral flow, indicating that groundwater NO3–N exhibited substantial spatial correlation on this axis, and little if any on the orthogonal axis (130° axis). These results agreed with the finding by Ella et al. (2001) of anisotropy of average and maximum groundwater NO3–N in shallow (6-m) wells such that the axis of greatest spatial continuity coincided with the direction of groundwater flow in a glacial till in Iowa.


Figure 6
View larger version (44K):
[in this window]
[in a new window]

  		Fig. 6. (a) Directional semivariograms of groundwater NO3–N concentrations and (b) field water table elevation gradient, both from 22 Feb. 2002. Water table elevation was estimated using inverse distance-squared weighted interpolation of measured depth to the water table at each well nest, corrected for surface elevation derived from lidar. The N management treatments were: RYE = uniform realistic yield expectation N management; FA = remote sensing informed, in-season, uniform, field-average N management; SS = remote sensing informed, in-season, site-specific, variable-rate N management.

 
The plot-scale spatial variance of groundwater NO3–N also changed during the study (Fig. 3d), with the within-plot spatial variances of each N management treatment tending to increase (Fig. 7 ). Averaged across all dates, the SS treatment had the largest mean within-plot spatial variance (4.03 mg2 L–2), which was significantly greater than those for the RYE (3.16 mg2 L–2) and FA (3.17 mg2 L–2) treatments. There were no significant differences in the mean within-plot spatial variance between the RYE and FA treatments.


Figure 7
View larger version (21K):
[in this window]
[in a new window]

  		Fig. 7. Temporal variation of the within-plot (between the two well nests in a treatment plot) variance of groundwater NO3–N concentrations by N treatment during a 2-yr study. Dates with fewer than five pairs to compute the treatment variance are not shown. The N treatments were: RYE = uniform realistic yield expectation N management; FA = remote sensing informed, in-season, uniform, field-average N management; SS = remote sensing informed, in-season, site-specific, variable-rate N management.

 
We hypothesized that the within-plot spatial variances of the RYE and FA treatments, which applied uniform N rates across individual plots, would be consistent over time if the N management treatments were the dominant controls of the within-plot spatial variance; however, the within-plot spatial variances of all treatments exhibited a moderate increasing trend with time (Fig. 7). This increase was especially apparent late in the experiment when N treatment effects were significant (Table 2, Phases IV and V). This may be another indication of lateral subsurface flow. Nitrogen fertilizer applied to one treatment plot may have reached the groundwater directly beneath it and then "bled" northeast across plot boundaries to an adjacent well nest. If these adjacent well nests were associated with different treatments, this would have resulted in within-plot differences in groundwater NO3–N even in the RYE and FA treatments.

Spatial Relationships with Nitrogen Treatment and Natural Controls
During Phases I through IV, N management treatments affected groundwater NO3–N only rarely (14% of dates) and only on isolated dates (Fig. 8 ). During Phase V, N management treatment effects were significant on seven consecutive dates out of 10. The on-and-off detection of significant N management treatment effects on groundwater NO3–N was probably due, at least in part, to subsurface lateral flow, as stated above. During both wet and dry periods, shallow groundwater probably flowed from one treatment plot to another, albeit in different directions depending on the pathway (drainage tile or soil), potentially confounding any treatment differences with time.


Figure 8
View larger version (27K):
[in this window]
[in a new window]

  		Fig. 8. Significance of N management treatments (Trt) and spatial natural controls (Elev, soil surface elevation; Sand, mean sand content; Silt, mean silt content; Clay, mean clay content; SOM, surficial soil organic matter content; Ksat, mean saturated hydraulic conductivity; WTE, water table elevation; and WTD, water table depth) in explaining within-field groundwater NO3–N concentrations on 54 individual dates across a 2-yr study. Asterisks (*) indicates dates when Trt was significant; + and – indicate dates when individual covariates were significant and whether the association between the covariate and groundwater NO3–N concentration was positive or negative, respectively.

 
Spatial relationships between groundwater NO3–N and spatial natural controls were complex (Fig. 8). First, none of these controls was consistently significant across all dates during the study period. Among these controls, Ksat was the most consistent spatial covariate; it was significant on 30% of the sampling dates, substantially more often than the other covariates examined. Second, for all covariates except Ksat and clay content, the relationships that existed with groundwater NO3–N were sometimes positive and sometimes negative; for Ksat the relationships were consistently negative and for clay content, consistently positive. Third, groundwater NO3–N was either correlated with none of these controls (e.g., 6 July 2001), with only one (e.g., 19 July 2001), or with several (e.g., 16 Aug. 2001). Such apparent complexity of the relationships between groundwater NO3–N and spatial natural controls might be associated with the various fates and complex transport pathways of NO3–N through the landscape–soil–groundwater system regulated by one or more controls simultaneously. If a static natural control were a primary regulator of one NO3-related process, the relationship between this control and groundwater NO3–N might be consistent with time. For example, Ksat is associated mainly with the NO3 leaching process. When present, the relationship between Ksat and groundwater NO3–N was consistently negative during the study. The observed discontinuity of the relationship might have been because Ksat was a functional control of NO3–N only when leaching occurred. The negative relationship between Ksat and groundwater NO3–N may have been because, during leaching events, soils with high Ksat conducted water so rapidly through macropores that little soil NO3–N was intercepted and absorbed, resulting in less NO3–N leaching to groundwater and dilution of groundwater NO3–N by macropore flow with lower NO3–N (Iqbal and Krothe, 1995). If a natural control affects more than one NO3-related process, the relationship between this control and groundwater NO3–N would probably be inconsistent, that is, sometimes positive, sometimes negative, depending on conditions. For example, when water tables were falling, areas within the field with shallower water tables might have had higher NO3–N losses via denitrification, thus lower groundwater NO3–N and a positive relationship between WTD and groundwater NO3–N. Under the opposite condition when water tables were rising, areas with shallower water tables might have had higher groundwater NO3–N due to the shallower rising groundwater intercepting and absorbing more soil NO3–N in the upper parts of the profile compared with areas where water tables were deeper, creating a negative relationship between WTD and groundwater NO3–N.

Interactions between Nitrogen Treatments and Spatial Natural Controls
During Phases I through IV, N management treatment effects on groundwater NO3–N were rarely and inconsistently significant (Fig. 8), and groundwater NO3–N was related primarily to spatial natural controls, including both relatively static controls (e.g., surface elevation and Ksat) and dynamic controls (e.g., WTD and WTE). The static controls were spatially correlated (Duffera et al., 2006), and the dynamic controls were spatially correlated during the study period (data not shown); consequently, we expected the smooth and continuous spatial patterns with longer correlation ranges during these phases. During Phase V, however, when N management treatments more consistently and significantly affected groundwater NO3–N, individual spatial natural controls were significant on only one or two of the 10 dates (sand, silt, Ksat, elevation, WTD; and WTE, respectively; Fig. 8). This suggests that the N management treatments overrode the natural controls during this period. Because the spatial arrangement of the N management treatments was randomized (Fig. 1), and groundwater NO3–N during this period was driven mainly by N management treatments, groundwater NO3–N tended to be spatially random or exhibited sharp and discontinuous spatial patterns (e.g., Fig. 4b) with short correlation ranges (Fig. 3c, Phase V).

Additional Considerations on the Spatial and Temporal Analyses
The temporal and spatial analyses were generally consistent in indicating that N management treatments had little or no effect on groundwater NO3–N during Phases I through IV, but did during Phase V. There was some disagreement, however, between the separately conducted spatial and temporal analyses as to the significance of the N treatments on groundwater NO3–N. For example, the spatial analysis of groundwater NO3–N on individual dates indicated that N treatments were significant on 3 of 14, 2 of 7, and 2 of 15 dates during Phases I, II, and III, respectively (Fig. 8). These results disagreed with those from the temporal analysis, which showed that N treatments were not significant during each of these phases (Table 2). The strength of the temporal analysis was that it considered the temporal correlation of groundwater NO3–N across multiple dates. A potential weakness of this analysis was that it failed to account for the spatial correlation in the model residuals. The spatial analysis had just the opposite characteristics; it took spatially correlated residuals into account, but did not account for the temporal correlation of the data across multiple dates.

There was a similar disagreement between the spatial and temporal analyses regarding the relationships between WTD and groundwater NO3–N. The temporal analysis indicated that groundwater NO3–N tended to be negatively associated with the WTD. The spatial analysis, however, showed that relationships between NO3–N and WTD were sometimes positive (Phases I, II, and IV, Fig. 8) and sometimes negative (Phases III and IV). Such disagreement was not surprising, because the temporal analysis determined the general relationship of field-mean groundwater NO3–N with WTD across all dates, while the spatial analysis used only the data from a specific date. The former considered an extensive (2-yr) temporal scale, while the latter addressed a fieldwide spatial scale on a single date.

Finally, it should be noted that the sample for our spatial analysis consisted of 60 points, which might be an insufficient number to reliably estimate the spatial range (Fig. 3c) and spatial variance (Fig. 3d) of groundwater NO3–N (Webster and Oliver, 1992). In particular, it would be desirable to have more samples to investigate more reliably the anisotropy of groundwater NO3–N. It remains difficult, however, to determine the sample size needed for estimating such parameters reliably for vadose-zone-related variables in a particular experimental setup (Skøien and Blöschl, 2006).


   CONCLUSIONS
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
Temporal patterns of field-mean groundwater NO3–N sampled over the 0.9- to 3.7-m depth exhibited two preferred states associated with water table depth: a high state when the water table was shallow and a low state when the water table was deep. The switch between the two states occurred when the water table rose or fell. Groundwater NO3–N was temporally correlated throughout the study period, but the covariance structures differed between the long- and short-term scales. The magnitudes of groundwater NO3–N temporal fluctuations greatly exceeded N management treatment effects. Further research is needed to quantify water table spatial and temporal dynamics and their controls in the context of landscape–soil–water systems. This will facilitate developing period-, season-, or site-specific strategies for managing water tables to reduce groundwater NO3 contamination in Coastal Plain soils. Managing shallow water tables via controlled drainage is already a recommended best-management practice to reduce the quantities of excess agricultural N that escape from fields to surface waters (Osmond et al., 2002). Such reductions are obtained primarily by maintaining water tables shallow enough to foster enhanced denitrification. While N management treatment effects were small compared with seasonal fluctuations in groundwater NO3–N, they appeared to contribute to a change in the temporal covariance structure of groundwater NO3–N during Phase V.

Spatial patterns and correlations of groundwater NO3–N exhibited distinct characteristics varying from random to highly organized. Early in the experiment, N treatment effects were infrequent and minor, and groundwater NO3–N exhibited smooth and continuous spatial patterns with longer correlation ranges that were primarily associated with natural controls. Groundwater NO3–N was related more with Ksat, WTD, surface elevation, and probably subsurface lateral flow, and less with WTE, surficial SOM, and particle size distribution. Later in the experiment, N treatment effects became consistently significant, and groundwater NO3–N was either spatially random corresponding to the spatially randomized assignment of the N treatments, or exhibited sharp and discontinuous spatial patterns with much shorter correlation ranges that were predominantly controlled by N treatments.

The complex spatial and temporal patterns, correlations, and relationships of groundwater NO3–N demonstrated in this study reflect the complexity of agricultural landscape–soil–water systems. They also suggest that the traditional sampling of NO3–N only at or after harvest in agricultural experiments is likely to be insufficient to capture the dominant features of groundwater NO3–N behavior throughout the growing season and beyond. Such measurements are also likely to be inadequate to assess the effectiveness of N management strategies. This is especially true in the southeastern U.S. Coastal Plain and other coarse-textured soils where NO3–N leaching during the growing season may be pronounced. Our data suggest that frequent and periodic monitoring of groundwater NO3–N is essential to capture treatment effects in the growing season and beyond. Simultaneous measurement of water table depth may also be critical to enhanced understanding of shallow groundwater NO3–N spatial and temporal behavior and its use as an environmental indicator of the consequences of N management.


   ACKNOWLEDGMENTS
 
This research was sponsored in part by the Initiative for Future Agricultural and Food Systems Grant no. 00-52103-9644 from the USDA Cooperative State Research, Education, and Extension Service.


   REFERENCES
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only