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


     

Published online 19 October 2005
Published in Agron J 97:1485-1492 (2005)
DOI: 10.2134/agronj2004.0323
© 2005 American Society of Agronomy
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 HighWire
Right arrow Citing Articles via ISI Web of Science (5)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Ruffo, M. L.
Right arrow Articles by Bullock, D. G.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Ruffo, M. L.
Right arrow Articles by Bullock, D. G.
Agricola
Right arrow Articles by Ruffo, M. L.
Right arrow Articles by Bullock, D. G.
Related Collections
Right arrow Spatial Distribution
Right arrow Predictive soil mapping
Right arrow Nutrient Management
Right arrow Nitrogen

Nitrogen Management

Spatial Variability of the Illinois Soil Nitrogen Test

Implications for Soil Sampling

Matías L. Ruffo, Germán A. Bollero*, Robert G. Hoeft and Donald G. Bullock

Department of Crop Sciences, Univ. of Illinois, 1102 S. Goodwin Ave., Urbana, IL 61801

* Corresponding author (gbollero{at}uiuc.edu)



   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSION
 CONCLUSIONS
 REFERENCES
 
The Illinois Soil Nitrogen Test (ISNT) is a recently developed test that has shown promise for corn (Zea mays L.) N fertilization. However, among-field and within-field spatial variability of this test is unknown and sampling recommendations have not been defined. The objectives of this study were to estimate the spatial variability and structure of the ISNT to provide sampling recommendations and to determine the short-term effect of N fertilizer on this test. Soil samples were collected on 14 production fields and analyzed for the ISNT. Data was analyzed with traditional statistics and with geostatistical techniques. In the short term, N fertilization did not have a significant effect on the ISNT (p > 0.3). The ISNT was normally distributed in 12 fields and showed a relatively low coefficient of variation (18% on average). Field mean ISNT ranged from 98 to 255 mg kg–1 and tended to be higher in Mollisols than in Alfisols. On average across fields, at least 10 samples are required to determine the mean ISNT for a given field with a precision of 24 mg kg–1. The geostatistical analysis revealed that the ISNT has strong spatial structure, as indicated by the bounded variograms, the relatively low nugget/sill ratio (<30% in 10 fields) and the mean range of 150 m. The spatial variability indicates that the ISNT can be mapped with a relatively sparse sampling grid and with a low number of samples, critical characteristics when assessing the economic implication for variable and uniform rate N fertilization.

Abbreviations: CV, coefficient of variation • ISNT, Illinois Soil Nitrogen Test • LRT, likelihood ratio test • MSE, mean squared error • NF, nitrogen fertilizer • SE, standard error • UIN, University of Illinois Agronomy Handbook nitrogen recommendation • VRN, variable rate nitrogen


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSION
 CONCLUSIONS
 REFERENCES
 
THE ILLINOIS Soil Nitrogen Test (ISNT) is a recently developed soil N test that shows promising ability to detect soils where corn is nonresponsive to N fertilization (Khan et al., 2001). Soils where corn was nonresponsive to N fertilization had ISNT values >235 mg kg–1, whereas where corn was responsive to N the ISNT was <225 mg kg–1. The ISNT estimates the aminosugar N fraction of the soil organic N, a labile pool of N that mineralizes during the growing season and supplies mineral N to the growing crop (Hoeft et al., 2002). The results presented by Khan et al. (2001) have generated a significant interest in this test among scientists, industry, and producers (Finck, 2004). In addition, preliminary results of ongoing research show that potential soil N supply—as measured by the ISNT—affects the site-specific response of corn to N fertilizer; thus, this test could be used for variable rate N (VRN) application recommendations.

There are no published results about the sampling distribution, variability, or spatial structure of the ISNT. Thus, sampling recommendations for this test have not been defined (Hoeft and Peck, 2004; Finck, 2004) and are required for ISNT to become extensively used. Although there is no available information about the sampling distribution and spatial variability of the ISNT, insight can be gained from the information available about other N fractions. Two relatively well characterized soil N fractions are total N and nitrate-N. Soil nitrate-N is very dynamic in time and variable in space because it is affected by mineralization, immobilization, nitrification, leaching, denitrification, and plant uptake. These processes are affected by soil–water content, soil temperature, and plant growth, which also vary spatially and temporally (Yates et al., 1988; Wolf and Rogowski, 1991; Bárdossy and Lehman, 1998; Western et al., 1999). In contrast, soil total N concentration is generally accepted as being stable in time, less variable in space, and is only affected by long-term changes in soil management. Since the ISNT estimated the concentration of a labile pool of organic N, it is likely that it shares some characteristics with total N and nitrate-N. However, it probably behaves more similarly to total N than to nitrate-N because the aminosugar N fraction is affected by less dynamic processes than nitrate-N.

One generally overlooked factor that has very important implications for sampling and for developing accurate and precise recommendations is the knowledge of the distribution of the variable under analysis. Hergert et al. (1995) argued that one of the most common problems in soil testing and interpretation is the assumption of normality of soil tests. If a soil test is log-normally distributed, >50% of the samples are lower than the mean, and the mean is no longer the best estimate of central tendency. Basing fertilizer recommendations on mean values of a log-normally distributed soil test underestimates the amount of fertilizer needed. The distribution of the soil test not only affects the recommendation when a measure of central tendency is needed, but also when the goal is to develop a map of the soil test because variograms estimated from nonnormal data sets tend to be erratic (Webster and Oliver, 2001) and significantly impact kriging estimates and their variance. In general, the sampling distribution of soil nitrates is known to be log-normal (Reuss et al., 1977; Parkin et al., 1988; Hergert et al., 1995), whereas the sampling distribution of total N is, in general, normally distributed (Cambardella et al., 1994; Stenger et al., 2002).

In many cases, management decisions are based on the mean value of a soil test. In these cases, it is critical to determine the necessary number of samples to collect to estimate the mean with a given level of precision. The number of samples to collect from a field can be calculated with the following formula (McBratney and Webster, 1983; Goovaerts and Chiang, 1993):

[1]
where N is the number of samples, t{alpha} is Student's t at chosen level of probability ({alpha}), P is the precision desired, and CV is the coefficient of variation. It is evident from this formula that the CV of the soil test determines the number of samples that have to be taken from a given field to achieve the desired level of precision in the estimation of the mean. The CV of soil nitrate-N ranges between 30 and 85%, 45% being quite common (Bundy and Meisinger, 1994) whereas the CV for total N ranges between 15% for plots and 30% for fields (Meisinger, 1984).

Once the number of samples to collect has been determined, the next decision is to decide the distance between soil samples. If the objective is to determine the mean value of ISNT, the range of the variogram determines the shortest distance between samples at which a field should preferentially be sampled and provides guidelines for sampling grid spacing. When a map of ISNT is needed, the variogram can also be used to provide guidelines for optimum sampling design for interpolation by ordinary kriging (Burgess and Webster, 1980; McBratney et al., 1981). Generally, samples need to be taken at a distance shorter than the range of the variogram for developing reliable maps. The kriging variance or standard error (SE) is usually considered an appropriate criterion to optimize the sampling design (McBratney and Webster, 1981; van Groenigen, 2000), although other criteria have also been used (van Groenigen, 2000). Regular grid sampling has been shown to be superior to random sampling to minimize maximum kriging SE (Burrough, 1991; Webster and Oliver, 2001) and although triangular grids are theoretically slightly better than square grids, the latter are preferred for practical reasons (Burrough, 1991). Consequently, the main variable that needs to be determined is the optimal grid spacing for a square (or rectangular if the variable is anisotropic) grid. The kriging SE depends on the sample configuration with respect to the point to be estimated and does not depend on the actual observed values (McBratney and Webster, 1981; Webster and Oliver, 2001). Thus, when the variogram is known, the kriging SE can be estimated for any sampling configuration of interest and the optimal design determined before conducting the field sampling. Consequently, the variogram is a useful tool to provide sampling recommendations for mapping and also when the objective is to determine the mean value for a field. However, the spatial structure of the ISNT is currently unknown and there are no reports of variogram estimates for it.

The ISNT was developed with samples collected at preplanting in early spring before N fertilizer application (Khan et al., 2001), but lately Hoeft et al. (2002) recommended sampling in the fall. Under a corn–soybean [Glycine max (L.) Merr.] rotation, fall soil sampling for ISNT would be advantageous because this sampling time is usually longer and preferred for P and K soil testing and the same soil sample could be used to evaluate all three nutrients. When the samples are collected after corn in a corn–soybean rotation or in the fall in a continuous corn rotation, the field is sampled in the same growing season after N fertilization. If N fertilizer application affects the ISNT values, then fall sampling could lead to incorrect fertilization decisions. There have not been studies to evaluate the effects of N fertilization in the same year on ISNT.

The above-mentioned points yield to the scientific and practical relevance for testing whether the ISNT is normally distributed; that the CV of the ISNT is relatively small (10–20%), and comparable to reported CVs for total N; that the ISNT presents spatial structure, with a low nugget/sill ratio (<30%) and a relatively long range (50 m); and finally, that the values of the ISNT are not affected by the previous growing season N fertilization.

The objectives of this study were: (i) to determine the sampling distribution of the ISNT in production fields of central and southern Illinois, (ii) to estimate the variability of ISNT values within production fields, (iii) to assess the spatial structure of ISNT, (iv) to determine if N fertilization affects the value of ISNT, and (v) to develop sampling guidelines for ISNT.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSION
 CONCLUSIONS
 REFERENCES
 
Eight production fields in 2002 and six fields in 2003 were sampled to assess the distribution and spatial variability of ISNT. Field locations are shown in Fig. 1 . Soil associations found in each field are presented in Table 1. Soils were preplant sampled in the spring (mid-March to early April) or postharvest in the fall (mid-October to mid-November). Specific sampling dates are presented in Table 2.



View larger version (31K):
[in this window]
[in a new window]
  		Fig. 1. Geographic distribution of experimental sites. Squares represent adjacent fields where experiments were conducted in 2002 and 2003, triangles represent 2002 fields, and circles represent 2003 fields.

 

View this table:
[in this window]
[in a new window]
  		Table 1. Location and soil series association of each field included in the study.

 

View this table:
[in this window]
[in a new window]
  		Table 2. Illinois soil nitrogen test (ISNT) sampling dates, field areas, and number of samples for each field involved in the study.

 
Soil sampling was on a rectangular grid (approx. 24 by 70 m) with the longest side in the row direction. In some fields (Table 2), a nested design was added to the grid sampling. The reason for this nested design was to provide information about short-range spatial variability and to estimate the variogram at short lags. The nested design consisted of eight five-sample cells. The cells were arranged as a square of 10 m in side with samples at the corners and at the geometric center. As an example, the nested sampling design used in A02 is presented in Fig. 2 . The total number of samples varied between fields, depending on the size of the field and the addition of the nested cells. Sample size by field is summarized in Table 3. Sample locations were flagged and georeferenced using a survey-grade RTK-DGPS (Leica SR 530).



View larger version (10K):
[in this window]
[in a new window]
  		Fig. 2. Soil sampling design for ISNT at A02. Detail of nested cell.

 

View this table:
[in this window]
[in a new window]
  		Table 3. Fields sample size (N), mean, median, standard deviation (SD), coefficient of variation (CV), skewness, and p value of the Shapiro-Wilk's test for normality for the ISNT.

 
Soil samples consisted of three 1.7-cm diameter cores taken 0.3 m apart and probed to a depth of 0.3 m. One sample was taken from the crop row and the other two between rows. Soil samples were kept in coolers during field collection and immediately dried on arrival to the laboratory. Samples were dried at 40°C and ground to pass a 2-mm mesh. Soil samples were analyzed following the method developed by Khan et al. (2001).

Statistical Analysis
Descriptive univariate statistics and exploratory data analysis were performed with the UNIVARIATE procedure of SAS (SAS Institute, 2003). Histograms, q-q plots, and box plots were used to examine the distribution of ISNT. Normality was tested with the Shapiro-Wilk's test ({alpha} = 0.01) (Shapiro and Wilk, 1965). Exploratory spatial analysis included mapping data points in ArcView GIS, variogram clouds, and h-scatter plots (Goovaerts, 1997; Webster and Oliver, 2001) calculated with Variowin (Pannatier, 1996). Equation [1] was used with two different approaches to determine the minimum number of samples needed to estimate mean ISNT. One approach was to define the desired level of precision as a percentage of the observed mean of each field, and the other approach was to determine the minimum number of samples for a maximum deviation of 24 (mg kg–1) of the sample mean from the true mean for all the fields, regardless of their mean ISNT. This maximum deviation was set 0.5 mg kg–1 higher than 10% of the upper value of the threshold range (235 mg kg–1), as determined by Khan et al. (2001).

Matheron's estimator of the variogram was used to assess the spatial structure of ISNT. This estimator takes the following form:

[2]
where N(h) is the number of pairs of observations, xi is the value at the start of the pair, and yi is the value at the end.

Omnidirectional variograms were estimated with GSLIB (Deutsch and Journel, 1998) up to a distance of half of the maximum lag, as suggested by Journel and Huijbregts (1978). At least 30 pairs of observations were used to calculate each of the semivariances, which where then in turn used to estimate the semivariogram as suggested by Journel and Huijbregts (1978). Fields with a nested sampling design provided lag distances <15 m and sample size of 32 pairs, whereas the other lag distances had larger sample sizes (>50 pairs).

Visual and statistical approaches were used for variogram modeling as suggested by Webster and Oliver (2001). In addition, the best two models were cross-validated to select the best variogram model. First, the variogram was plotted and inspected for general trends. Then spherical, Gaussian, and exponential models with and without a nugget were fitted using SAS NLIN (SAS Institute, 2003). The fit of the models was inspected by analyzing plots of residuals and the plot of the fitted model and the sample variogram. Based on the mean squared error (MSE), the best two models were selected for further analysis. These two best models were compared by cross-validation using the KT3D program of GSLIB (Deutsch and Journel, 1998). The model with the smallest MSE was selected. The parameters (nugget, sill, and range) and their standard errors were estimated with SAS NLIN.

The maximum ordinary kriging SE for each variogram at different sampling grids was calculated with the OSSFIM function of GSTAT (Pebesma, 2004). The OSSFIM function is based on the work of McBratney et al. (1981) and McBratney and Webster (1981), who developed an algorithm to estimate kriging SE. The original OSSFIM code was modified to restrict the number of observations used for ordinary kriging between 10 and 25. The search radius was defined as 10% larger than the variogram range. Several authors (Burrough, 1991; Webster and Oliver, 2001) recommend having at least 10 observations for ordinary kriging estimation and to extend the search neighborhood beyond the range of the variogram. When the number of observations within the search neighborhood was <10, the ordinary kriging SE was not calculated.

Effect of Nitrogen Fertilizer on the Illinois Soil Nitrogen Test
The effect of N fertilization on ISNT was analyzed on fields where soil samples were collected in the fall after N fertilization in the previous fall or spring (i.e., following a corn crop). The fields were subdivided in 13 to 20 sections depending on the field shape and size. Each section was composed of five N rate fertilization plots. Plot dimensions varied slightly between fields depending on field shape, size, and farming equipment, but were at least 70 m long and 24 m wide. Plot width accommodated two passes of the fertilizer applicator. Each plot randomly received one N fertilizer (NF) rate. The University of Illinois Agronomy Handbook N recommendation (UIN) (Hoeft and Nafziger, 2004) was used to determine a benchmark rate. The algorithm used was:

[3]

The yield goal (Mg ha–1) was calculated by averaging the last 5 yr of corn yield or provided by the producer. A 45 kg N ha–1 soybean credit was used because corn was planted after soybean in every field. The incidental N considered the N applied with starter fertilizer (DAP or MAP) and herbicides. The NF rates applied were UIN, UIN – 56 kg N ha–1, UIN – 28 kg N ha–1, UIN + 28 kg N ha–1, and UIN + 56 kg N ha–1. The UIN varied among fields depending on the yield goal and N management but ranged from 150 to 190 kg N ha–1.

The effect of fertilizer N on ISNT was assessed with the MIXED procedure of SAS (SAS Institute, 2003). Nitrogen fertilizer rates were considered fixed effects in the model. Spatial correlation of residuals was tested for spatial structure ({alpha} = 0.1) by a likelihood ratio test (LRT) between a model with a spatial covariance matrix and a model with independent errors (Littell et al., 1996). When the spatial structure was significant, the model of spatial covariance structure (spherical, exponential, linear, and Gaussian, with and without nugget) was selected based on the corrected Aikaike's information criteria AICC, as suggested by Littell et al. (1996).


   RESULTS AND DISCUSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSION
CONCLUSIONS
 REFERENCES
 
Nitrogen Fertilizer Effect on the Illinois Soil Nitrogen Test
Nitrogen fertilizer application did not have a significant effect (p > 0.30) on ISNT in any of the fields where samples were collected the same growing season as fertilizer N application. Other studies have shown that total N and labile fractions of soil organic N can also remain unchanged after several years of continuous N fertilizer application (McCraken et al., 1989). For example, Omay et al. (1997) found that potentially mineralizable N content was not affected by the continuous application of fertilizer N for 10 yr, both in corn–corn and corn–soybean rotations. In summary, as occurs with total N and more labile fractions of N, 1 yr of N fertilizer application did not affect ISNT for these production fields.

Descriptive Statistics
A total of 1490 samples for ISNT were taken from 14 fields and are summarized in Table 3. Mean ISNT ranged from 98 to 255 mg kg–1 with lowest values occurring at RB02, RS02, and RS03 on Alfisols (<150 mg kg–1) and highest values occurring on the rest of the sites on Mollisols (218–255 mg kg–1). The ISNT values observed in this study are within the range reported by Khan et al. (2001) for similar soils. Five sites in this study were within 220 and 240 mg kg–1, indicating that it would be fairly common to find fields where the ISNT would not give a clear recommendation for N fertilization. Mean and median ISNT were generally very similar.

The SD for ISNT ranged from 13.5 mg kg–1 (RB02) to 53 mg kg–1 (RS02), but for most of the fields the SD ranged between 30 and 45 mg kg–1 (Table 3). Similarly, Boast et al. (2003) found a SD of 15 mg kg–1 for an Alfisol under continuous corn (mean ISNT 157 mg kg–1), 24 mg kg–1 for a Mollisol under a corn–soybean rotation (mean ISNT 304 mg kg–1), and 43 mg kg–1 for a Mollisol under a corn–soybean rotation that received manure annually (mean ISNT 493 mg kg–1). In addition, Alfisols had CVs ranging from 9.5 to 44% and Mollisols ranged from 10 to 20%, with a mean of 15% (Table 3). The CVs for ISNT are smaller than the reported values for nitrate-N or total N. Reports of nitrate-N CVs range from 20 to 80% (Bundy and Meisinger, 1994; Cambardella et al.,1994). Coefficient of variation for total N range from 22 to 45% (Lengnick, 1997; Cambardella and Karlen, 1999).

In general, the value of skewness of ISNT was close to 0 (i.e., no skewness) in all the fields except for RS03, which was positively skewed. Except for RS03 and RS02, ISNT was normally distributed according to the Shapiro-Wilk's test. The main reason for the lack of normality in RS02 was a bimodal distribution associated with a cluster of large values. In RS03, these large values did not produce a bimodal distribution but a positively skewed one. Since normality was the general observation, we chose to not transform values before additional analyses. No extreme outliers were detected in any of the fields under study. The low degree of skewness and the general normal distribution of ISNT contrasts with nitrate-N, which is in general highly positively skewed (Tabor et al., 1985; Cahn et al., 1994; Röver and Kaiser, 1999; Stenger et al., 2002) and log-normally distributed (Reuss et al., 1977; Tabor et al., 1985; Robertson et al., 1997).

Sample Size for Field Mean Estimation
The number of samples required to estimate the mean of a site are presented in Table 4. As expected from Eq. [1], sample number increased as the level of precision increased. Similarly, a high CV indicated a high number of samples (e.g., RS02 and RS03). If RS02 and RS03 are excluded, about 35 (5% precision) and 9 (10% precision) samples were needed to estimate the mean ISNT. On average across all fields, only 10 samples were required to estimate the mean ISNT with a maximum error of 24 mg kg–1. The number of samples required for the two fields with the largest CV decreases abruptly from >50 samples to ≤20 samples. The reason for this decrease is that 24 mg kg–1 is 16 and 24% of the mean field ISNT for RS02 and RS03, a much lower precision than the 10 and 5% used in the first approach. A similar effect occurred with the RB02 field, which had a low mean ISNT (143 mg kg–1) and only two samples were required to estimate the mean with an error of 24 mg kg–1. Relatively low variability of ISNT determines that the mean value for a given field can be determined with few samples, an attractive characteristic for a soil test.


View this table:
[in this window]
[in a new window]
  		Table 4. Number of samples necessary to estimate the mean with 5 and 10% precision with respect to the mean, or with a maximum allowed deviation of 24 mg kg–1.

 
Spatial Structure of the Illinois Soil Nitrogen Test
The ISNT presented spatial structure in most fields involved in the study (Table 5). The exception was S03, which had a pure nugget variogram, probably because the first lag (24 m) was not short enough to capture short range variability. In contrast, the adjacent field (S02) with a nested design and a much shorter first lag (7 m) demonstrated spatial structure.


View this table:
[in this window]
[in a new window]
  		Table 5. The Illinois Soil Nitrogen Test semivariogram model results, including nugget (C0), partial sill (C), range, nugget/sill ratio [C0/(C0 + C)] and coefficient of determination (r2) of the model for ISNT.

 
All the variograms were bounded (i.e., had a sill). Unbounded variograms are, in general, an indication of a trend in the spatial structure or lack of second-order stationarity (Webster and Oliver, 2001). For six of the fields that showed spatial structure, the best theoretical variogram model did not include a nugget (nugget/sill ratio of 0%). For the rest of the fields that had spatial structure, the nugget/sill ratio ranged from 14% (S02) to 51% (RB02). Ten of 14 fields presented a nugget/sill ratio smaller than 30%, indicating a strong spatial structure for ISNT. Excluding S03, the shortest range of the theoretical variogram models was 45 m for A03 and the longest range was 400 m (SY03), with a mean range of 150 m. Generally, the theoretical variograms were spherical, with only three fields having an exponential variogram. The addition of the nested samples did not affect the variogram nor the significance of the nugget.

It is possible to compare the spatial structure of ISNT with that of soil nitrate-N and total N concentration, which have been proposed and used as diagnostics tools for N fertilization. Cahn et al. (1994) reported that the range of nitrate-N was only 5 m for an intensively sampled area of 0.25 ha, but had no spatial structure when the sampled area was 3.3 ha. Similarly, Röver and Kaiser (1999) found that nitrate-N showed pure nugget effect even though the soil was sampled at a 7 by 7 m grid. Although other authors reported soil nitrate-N to have spatial structure, in general the nugget/sill ratios have been quite high in those cases. For example, Robertson et al. (1997) reported a range of 90 m with a nugget/sill ratio of 63%, Cambardella et al. (1994) found a range of 200 m with a nugget/sill ratio of 79 and 41% for two different fields, Lengnick (1997) reported a range of 300 m with 78% of nugget/sill ratio, and the range found by Tsegaye and Hill (1998) was 21 m with a nugget/sill ratio of 67%. Cambardella et al. (1994) studied the spatial structure of soil total N in two fields of Central Iowa. These authors reported a nugget/sill ratio of 11% for a pothole field and 12% for a no-till field, with ranges of 89 and 115 m, respectively. In addition, Stenger et al. (2002) reported a nugget/sill ratio of 34% and a range of 42 m.

In summary, the variograms of ISNT were always bounded and in general showed a strong spatial structure, with nugget/sill ratio smaller than 30% and a mean range of about 150 m. The spatial structure of ISNT seems to agree more with the spatial structure reported for total N than for nitrate-N.

Grid Spacing for Ordinary Kriging
The maximum ordinary kriging SEs as a function of grid spacing for all the fields included in this study are presented in Fig. 3 and 4 . These figures can be used to determine the grid spacing based on the maximum allowed ordinary kriging SE. For example, if the maximum allowed ordinary kriging SE is set at 25 mg kg–1 and we assume a variogram as the one estimated for J03, then the optimal grid spacing would be approximately 90 m. Alternatively, when resources are limiting and a fixed grid spacing is used, the maximum ordinary kriging SE for that design can be obtained using Fig. 3 and 4 before sampling.



View larger version (22K):
[in this window]
[in a new window]
  		Fig. 3. Maximum ordinary kriging standard error (SE) (mg kg–1) as a function of grid spacing (m) for the 2002 fields.

 


View larger version (18K):
[in this window]
[in a new window]
  		Fig. 4. Maximum ordinary kriging standard error (SE) (mg kg–1) as a function of grid spacing (m) for the 2003 fields.

 
In all the cases where the variogram showed spatial structure, the ordinary kriging SE increased with grid spacing and the differences between fields depended on the parameters of the variogram (nugget, total sill, their difference, and the range). The nugget set a minimum to the ordinary kriging SE, whereas the range determined the absolute maximum grid spacing and affected the rate of increase of the ordinary kriging SE as a function of grid spacing.

In some instances, particularly when the nugget was large, the ordinary kriging SE might be higher than the maximum allowed, or the optimal grid spacing might be too short, and in consequence prohibitively expensive. For example, if the maximum ordinary kriging SE tolerated was set to 20 mg kg–1, it would not be possible to achieve that level of precision with the variogram of M02. Whereas using the W02 variogram, the optimal sampling grid would be 20 m, which could be regarded as too short for economic reasons.


   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSION
 CONCLUSIONS
REFERENCES
 
Spatial variability in ISNT was evaluated on representative crop production fields in central and southern Illinois. Generally, ISNT was found to be normally distributed, with a relatively low CV and skewness. Spatial structure of ISNT was evident at most locations with a mean range of 150 m. These characteristics make ISNT an attractive soil test for N management. When the strategy is to estimate the mean ISNT, the relatively low SD and CV determine that only a small number of samples (10, on average across these fields) were necessary to estimate the mean with a maximum error of 24 mg kg–1. Samples should preferably be taken at spacing longer than the expected range, which fluctuated between fields but on average was 150 m across all the fields, or 130 m when the shortest and longest ranges were excluded.

Spatial structure is a desirable property when the objective is to map ISNT. In all but one field, the ISNT had spatial structure, and in most cases the nugget/sill ratio was smaller than 30%. Variograms were also used to determine the optimal grid spacing by minimizing the maximum ordinary kriging SE. Alternatively, Fig. 3 and 4 can be used to determine the maximum ordinary kriging SE before sampling.

Finally, finding that N fertilization within the same year as sampling had no effect on ISNT was desirable. The result is an expanded time frame between soil sampling and N fertilization and allows testing for ISNT, P, and K with the same samples.

All the characteristics associated with the spatial variability mentioned above strongly suggest that ISNT has the potential to become a routine test for uniform and variable rate N fertilization. However, it will be necessary to correlate and calibrate the ISNT to different cropping systems, soil classes, and crops.


   ACKNOWLEDGMENTS
 
The authors thank Trish Lehman for running the ISNT and Marisol Benitez, Bryan D. Young, and Cecilia Azpiroz for their help with the field work. Many thanks to Dr. Steve Ebelhar for coordinating the work in southern Illinois. We also appreciate the patience and help of the cooperators in this study, Darwin Builta, John and Susan Adams, Dean Sasse, John Reifsteck, Kelly Robertson, the Roser brothers, and Ken Dalenberg.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSION
 CONCLUSIONS
 REFERENCES
 




This article has been cited by other articles:


Home page
 Soil Sci.Home page
H.-J. Kim, J. W. Hummel, K. A. Sudduth, and P. P. Motavalli
Simultaneous Analysis of Soil Macronutrients Using Ion-Selective Electrodes
Soil Sci. Soc. Am. J., October 29, 2007; 71(6): 1867 - 1877.
[Abstract] [Full Text] [PDF]

Home page
 Crop Sci.Home page
D. S. Bullock, N. Kitchen, and D. G. Bullock
Multidisciplinary Teams: A Necessity for Research in Precision Agriculture Systems
Crop Sci., September 1, 2007; 47(5): 1765 - 1769.
[Abstract] [Full Text] [PDF]

Home page
 Soil Sci.Home page
J. D. Williams, C. R. Crozier, J. G. White, R. W. Heiniger, R. P. Sripada, and D. A. Crouse
Illinois Soil Nitrogen Test Predicts Southeastern U.S. Corn Economic Optimum Nitrogen Rates
Soil Sci. Soc. Am. J., April 5, 2007; 71(3): 735 - 744.
[Abstract] [Full Text] [PDF]

Home page
 J. Environ. Qual.Home page
N. Hong, P. C. Scharf, J. G. Davis, N. R. Kitchen, and K. A. Sudduth
Economically Optimal Nitrogen Rate Reduces Soil Residual Nitrate
J. Environ. Qual., January 25, 2007; 36(2): 354 - 362.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
A. N. Kravchenko, G. P. Robertson, X. Hao, and D. G. Bullock
Management Practice Effects on Surface Total Carbon: Differences in Spatial Variability Patterns
Agron. J., October 3, 2006; 98(6): 1559 - 1568.
[Abstract] [Full Text] [PDF]

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 HighWire
Right arrow Citing Articles via ISI Web of Science (5)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Ruffo, M. L.
Right arrow Articles by Bullock, D. G.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Ruffo, M. L.
Right arrow Articles by Bullock, D. G.
Agricola
Right arrow Articles by Ruffo, M. L.
Right arrow Articles by Bullock, D. G.
Related Collections
Right arrow Spatial Distribution
Right arrow Predictive soil mapping
Right arrow Nutrient Management
Right arrow Nitrogen

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Crop Science Vadose Zone Journal
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome