Agronomy Journal Grow Your Career With ASA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

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 ISI Web of Science (8)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Adams, M. L.
Right arrow Articles by Peverly, J. H.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Adams, M. L.
Right arrow Articles by Peverly, J. H.
Agricola
Right arrow Articles by Adams, M. L.
Right arrow Articles by Peverly, J. H.
Related Collections
Right arrow Soybean
Right arrow Crop Growth and Development
Right arrow Nutrient Management
Agronomy Journal 92:268-274 (2000)
© 2000 American Society of Agronomy

MICRONUTRIENT STATUS

Toward the Discrimination of Manganese, Zinc, Copper, and Iron Deficiency in `Bragg' Soybean Using Spectral Detection Methods

Matthew L. Adamsa, Wendell A. Norvellb, William D. Philpotc and John H. Peverlyd

a CSIRO Land and Water, Private Bag, P.O., Wembley 6014, WA, Australia
b USDA Plant, Soil & Nutrition Lab., Tower Rd., Ithaca, NY 14853 USA
c Dep. of Civil and Environmental Engineering, Hollister Hall, Cornell University, Ithaca, NY 14853 USA
d Dep. of Soil, Crop & Atmospheric Sciences, Bradfield Hall, Cornell University, Ithaca, NY 14853 USA

matthew.adams{at}per.clw.csiro.au


   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 Materials and methods
 Results
 Discussion
 Conclusions
 REFERENCES
 
Early visual symptoms of Mn, Zn, Cu, and Fe deficiency are often difficult to interpret and incorrect diagnoses are common. Reflectance and fluorescence measures may be useful for early and more reliable detection of Mn, Zn, Cu, and Fe deficiencies, but only if one or more spectral measures change uniquely with each deficiency. A discriminant analysis was performed to determine whether selected fluorescence and reflectance measures could be used to discriminate effectively marginal Mn, Zn, Cu, Fe deficiencies, and nutrient-adequate soybean [Glycine max (L.) Merr. cv. Bragg] leaves from plants grown in solution culture. Predictors were yellowness index (YI), a new measure sensitive to chlorosis; normalized difference vegetation index (NDVI); the ratio of minimal fluorescence (Fo) to variable fluorescence (Fv), Fo/Fv; and the ratio of minimal fluorescence to the fluorescence yield after 5 min of illumination (F5min), Fo/F5min. Manganese, Zn, Cu, and Fe deficiencies were correctly identified 62, 40, 92, and 30% of the time, respectively, as estimated by cross-validation. Controls were identified correctly 77% of the time. One-third to one-half of the leaves identified as nutrient deficient by tissue analysis did not exhibit visual symptoms. Lack of a spectral measure sensitive specifically to Zn and Fe deficiency contributed to the low identification rates for Zn and Fe deficiencies. While the development of spectral measures sensitive to Zn and Fe deficiencies is required for further development of this rule, discriminant analysis is a suitable method for the development of classification rules for identifying marginal stresses.

Abbreviations: F, chlorophyll fluorescence (subscripts m, o, v represent maximum, minimal, variable fluorescence • 5 min, fluorescence after 5 min illumination) • ICP, inductively coupled plasma emission spectrometry • NDVI, normalized difference vegetation index • YI, yellowness index • YFML, youngest fully mature leaf


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
Materials and methods
 Results
 Discussion
 Conclusions
 REFERENCES
 
SEVERAL STUDIES have examined the effect of one to three stresses on either leaf fluorescence induction or leaf reflectance characteristics (e.g., Carter and Young, 1993; Sun et al., 1989; Kriedemann and Anderson, 1988; Conroy et al., 1986). A few studies have examined more than three stresses (e.g., Subhash and Mohanan, 1994; Carter, 1993; Abadía et al., 1988; Al-Abbas et al., 1974; Gausman et al., 1973).

The use of several different spectral techniques simultaneously, sensitive to different aspects of stressed plant physiology, can improve the ability to discriminate among stresses. A number of studies have examined the effect of different stresses on both fluorescence and reflectance (e.g., Buschmann et al., 1994; Baret et al., 1988).

Of those studies that compared differences between or among stresses, comparisons were often qualitative, comparing only a small number of treatment levels. Furthermore, none of these studies were intended to determine whether specific stresses could be detected or discriminated at an early stage of stress. They generally concluded that specific stresses could be discriminated at a specific point in time and from severely stressed plants. The ability to discriminate severely stressed plants is not useful in a practical sense.

Adams et al. (2000) reported changes in selected leaf reflectance and fluorescence induction measures to varying levels of Mn, Fe, Zn, and Cu supply. The treatments imposed in Adams et al. (2000) resulted in plants ranging from nutrient adequate to severely deficient (including marginally deficient plants) in each of the micronutrients mentioned above. This study extends the work of Adams et al. (2000) to determine whether reflectance and fluorescence induction characteristics can be effectively used to discriminate among nutrient-adequate and Mn-, Fe-, Zn-, and Cu-deficient Bragg soybean leaves. This report discusses how the method might be improved, and addresses issues associated with the application of this method to field data.


   Materials and methods
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and methods
Results
 Discussion
 Conclusions
 REFERENCES
 
A quadratic discriminant analysis was performed using Minitab V11.2 (Minitab, Inc., State College, PA1) on training data extracted from previous work (Adams, 1993, 1996; Adams et al., 2000) by the process described below. The spectral measures selected as predictors for the analysis were yellowness index (YI), normalized difference vegetation index (NDVI), Fo/Fv, and Fo/F5min. The response of these four measures in response to Mn, Zn, Fe, or Cu concentration in leaf tissue has been described previously (Adams, 1993, 1996; Adams et al., 2000). The error rate of the discriminant rule was determined by two methods: (i) by cross-validation, which omits each observation individually from the dataset, develops a new discriminant rule using the remaining observations, then classifies the omitted observation, and (ii) by dividing the training data equally into two sections, a training set and a test set.

Training Data
There are two important considerations in the development of a classification rule for discriminating early Mn, Fe, Cu, or Zn deficiency. First, the classifier must be developed from agronomically relevant observations. Being able to successfully discriminate a severely Mn-deficient plant from a severely Fe-deficient plant is of little value. Therefore, for this analysis only observations from marginally deficient and sufficient plants as determined by plant tissue analysis were used (discussed further below). Second, the spectral measures selected as predictors in the classification rule should be responsive to one or more metal deficiencies and unresponsive to sufficient levels. This helps improve the separation of stressed plants from unstressed plants by minimizing the variability within the control group (discussed further below). The YI, NDVI, Fo/Fv, and Fo/F5min from Adams et al. (2000) meet this second criterion.

The aggregate data set from which the training data was derived consisted of 540 observations from the youngest fully mature leaf (YFML), the next older leaf (YFML+1), and the next younger leaf (YFML-1) from plants supplied with varying levels of Mn, Fe, Zn, or Cu (depending on experiment), which were sampled 16, 21, 24, and 28 d after seed imbibition; see Adams et al. (2000) and Adams (1996) for additional information. One-half of these observations (i.e., approximately 270) were initially excluded, because Fo/F5min was not measured for these leaves. The following criteria were then applied to the remaining observations to generate training data consisting of observations from marginally deficient and sufficient plants.

1. Any observation from treatments that resulted in shoot yields less than 80% of the shoot yields of nutrient-adequate plants (referred to as controls hereafter) for a given harvest and experiment was excluded. Critical levels are generally determined for 90% of maximum yield.

2. After excluding observations based on shoot yield, further observations were excluded if the leaf metal concentration associated with a given observation from the specified individual element experiment was outside the range given in Table 1 . The intent with this criterion was to exclude observations from leaves severely deficient or potentially toxic in an element according to tissue analysis.


View this table:
[in this window]
[in a new window]
  		Table 1 Table of nutrient concentration criteria for including marginal and sufficient observations in the training dataset

 
The values given in Table 1 worked well for the range of nutrient concentrations found at each treatment level from the experiments from which the data were drawn. For example, Criterion 1 removed all of the observations that were severely Zn deficient by plant tissue analysis; therefore, only observations with high leaf Zn concentrations needed to be excluded from the analysis. Other values might be more appropriate for other varieties or other experiments where a greater range of leaf concentrations are encountered.

Once the training data were extracted, each observation was classified as being Mn deficient, Zn deficient, Cu deficient, Fe deficient, or nutrient sufficient based on the criteria given in Table 2 .


View this table:
[in this window]
[in a new window]
  		Table 2 Rules for classification of observations from five experiments

 
To test the results of the cross-validation of the discriminant rule, the training data were randomly split into two equal groups referred to as the training set and the test set. The training set was used to develop the discriminant rule, and the test set was used to test the rule.

Data Preprocessing for Classification
Prior to classifying the entire dataset, a screening program was used to automatically classify extreme observations, based on a set of rules determined from visual examination of all of data. For example, if an observation had a YI greater than 2.5 and Fo/Fv greater than 1.5, the screening program classified that observation as Mn deficient, because no data points from any other group had YI and Fo/Fv values exceeding 2.5 and 1.5, respectively. All observations not classified by the screening program were classified by the discriminant rule (note: 31 of 267 observations were preclassified this way). The purpose of the preprocessing was to ensure that observations being classified were within the range of the data used to develop the discriminant rule.

Fo/F5min for Manganese Data
No F5min data were collected from the Mn experiments in Adams (1993). This presented a problem for the discriminant analysis, because Fo/F5min appears to be a measure specific for Cu deficiency (Adams et al., 2000; Kriedemann and Anderson, 1988). The specificity of Fo/F5min for Cu deficiency is speculative without Fo/F5min data for Mn.

Fo/F5min was measured in an experiment subsequent to Adams (1993), in which plants were grown under similar growth conditions as described in Adams et al. (2000) and solution Mn concentration was varied. Fo/F5min of the index leaf on 23-d-old plants grown in this subsequent experiment decreased from a value of approximately 1.1 when leaf [Mn] was approximately 4 to 6 g kg-1 to a value of 0.70 to 0.85 when leaf [Mn] was greater than 10 to 12 g kg-1. These observations are consistent with fluorescence curve data presented in Hannam et al. (1985), Kriedemann et al. (1985), and Kriedemann and Anderson (1988).

A linear regression of F5min on (Fo/Fv)/total chlorophyll (r2 = 0.905) was used to estimate F5min for data from Adams (1993) based on data from the experiment just outlined above. The Fo/F5min values calculated from the estimated F5min were comparable to those measured in experiments where Mn was a treatment variable. As a check, the discriminant analysis was run both with and without the simulated data; the conclusions were not affected, so the simulated data are included in the rest of this report.

Discriminant Analysis
A qualitative estimate of the ability of the selected spectral measures to discriminate Mn-, Zn-, Cu-, and Fe-deficient leaves from each other and from nutrient-adequate leaves using the four selected spectral measures (Fo/Fv, YI, NDVI, and Fo/F5min) can be visualized by plotting data for each group in two-dimensional space as ellipses of constant probability density based on the bivariate normal distribution of the two variables (Johnson and Wichern, 1992). Figure 1 presents two such plots for YI and Fo/F5min vs. Fo/Fv. The coordinates of the center of each ellipse are the respective means for each variable. The 68.3% constant density contour (i.e., the first standard deviation contour) for the bivariate normal distributions are plotted in the figure.



View larger version (20K):
[in this window]
[in a new window]
  		Fig. 1 The 68.3% contour (i.e., 1 SD) for the bivariate normal distributions of yellowness index (top) and Fo/F5min (bottom) vs. Fo/Fv for observations from Mn, Zn, Cu, Fe, and control groups from the training data set. The coordinates of the center of each ellipse are the respective means for each variable. The number of observations used to derive the ellipses are 24 for Mn, 10 for Zn, 13 for Cu, 10 for Fe, and 110 for controls

 
The lower plot of Fo/F5min vs. Fo/Fv clearly suggests that observations from Cu-deficient leaves should be easily separated from observations from the other deficiencies and controls. Observations from Fe- and Mn-deficient leaves should be relatively easy to separate from controls, but somewhat more difficult to separate from each other. Some additional power to discriminate observations from Mn-deficient leaves from observations of Fe-deficient leaves is suggested weakly in the upper plot of YI vs. Fo/Fv.

Figure 1 also shows why a quadratic, rather than linear, discriminant analysis was used. The covariance matrices among the different deficiencies and controls (i.e., sufficient) are not equal, because the constant probability ellipses are not of similar shape (e.g., see control vs. Mn deficiency ellipse in the lower plot of Fig. 1). Equality of covariance matrices among groups is an assumption in the use of linear discriminant analysis (Johnson and Wichern, 1992). Differences in covariance matrices are not surprising, because the predictors (YI, NDVI, Fo/Fv, and Fo/F5min) were selected on the basis that each was responsive to one or more metal deficiencies and unresponsive to sufficient levels (Adams et al., 2000). This tends to minimize variability within the control group, resulting in a smaller ellipse of constant probability for a given probability level.


   Results
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Results
Discussion
 Conclusions
 REFERENCES
 
Discriminant Analysis
The results of the cross-validated and test-set discriminant analyses applied to the training data are presented in Table 3 . For example, for the 24 samples in the cross-validated training data that were identified as Mn deficient by tissue analysis, 15 were identified by the discriminant rule as being Mn deficient, 0 as Zn deficient, 0 as Cu deficient, 7 as Fe deficient, and 2 as nutritionally adequate. Across all five groups Mn, Zn, Cu, and Fe deficiencies were correctly identified 62, 40, 92, and 30%, respectively, of the time as estimated by cross-validation. Controls were identified correctly 77% of the time. Correct identification rates observed by splitting the training data into a training set and a test set were similar (a chi-square test could not be performed because of small cell counts).


View this table:
[in this window]
[in a new window]
  		Table 3 Summary of classification by cross-validation and by classification of a test set by a training set (described in text) for a discriminant rule using Fo/Fv, YI, NDVI, and Fo/F5min as predictors.{dagger} For each group, shading indicates the actual number of observations predicted correctly

 
Observations from leaves that were Mn deficient were most often misclassified as Fe deficient. Observations from leaves that were Zn deficient tended to be misclassified as controls. Observations from leaves that were Fe deficient were most often misclassified as Mn and Zn deficient. Controls were most often misclassified as Zn deficient. Clearly, the discriminant rule is worse at discriminating Zn and Fe deficiencies. This was suggested qualitatively in Fig. 1 by the large degree of overlap between the Zn, Fe, and control curves.

Table 4 summarizes two different classifications of the entire dataset using a classifier derived from 50% of the training data (i.e., 85 observations taken at random from the cross-validation analysis shown in Table 3). The entire dataset included all of the data removed in the process of generating the training data described above (100 observations). The classification results under the `Resub' heading for each group include those observations used to derive the classifier. The classification results under the `Test' heading for each group do not include observations used to develop the classifier. These results emphasize the consistency of the classifier in correctly classifying observations from Mn-deficient, Cu-deficient, and control leaves, because the classification rates are similar to those in Table 3. These results also emphasize that the discriminant rule is poor at identifying Zn and Fe deficiency correctly.


View this table:
[in this window]
[in a new window]
  		Table 4 Summary of classification for the entire dataset for the discriminant rule devied from 50% of the training data (85 observations) using Fo/Fv, YI, NDVI, and Fo/F5min as predictors.{dagger} For each group, shading indicated the actual number of observations predicted correctly{ddagger}

 
Utilizing Posterior Probabilities
Table 5 shows that knowing the posterior probabilities of group classification can provide additional useful information. These are two cases where the discrimination was marginal (i.e., the difference in probabilities for assigning an observation between two different groups was small). In one case, the classification was correct (control); in the other, it was not (Fe deficient was classified as Zn deficient). However, the classification rule was able to successfully reduce the possibilities down to two, from the original five. This may be useful information, especially in an area where one of several deficiencies may be observed or where other information is available to suggest the presence of a specific deficiency.


View this table:
[in this window]
[in a new window]
  		Table 5 Probabilities of group assignment for two observations. The upper observation was classified correctly, while the lower observation was classified incorrectly (refer to text)

 

   Discussion
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Results
 Discussion
Conclusions
 REFERENCES
 
The discriminant analysis method represents an improvement over earlier attempts to discriminate among nutrient deficiencies based solely on differences of spectral measures among severely deficient plants and controls (e.g., Subhash and Mohanan, 1994; Sun et al., 1989; Al-Abbas et al., 1974; Gausman et al., 1973). Observations used to develop the discriminant rule came from leaves marginally deficient in one of four metal micronutrients, or leaves that were adequate in all four metal micronutrients. Roughly one-third to one-half of the 57 leaves considered Mn, Zn, Cu, or Fe deficient by nutrient analysis in Table 3 did not exhibit visual symptoms. This is a more agronomically relevant set of data from which to determine whether spectral techniques can be used to discriminate among micronutrient deficiencies in soybean than a set of data that contains only observations from severely deficient plants and controls.

The observations used to develop the rule came from leaves of different ages (YFML; the next older leaf, YFML+1; and the next younger leaf, YFML-1) sampled on different days that were either marginally deficient in a specific micronutrient or adequate. Therefore, this discriminate rule is appropriate for classification of observations from YFML±1 leaf trifoliolate leaves measured during the early part of the growth cycle. A more sensitive discriminant rule could potentially be developed from observations from the YFML only, which is the leaf usually selected for plant tissue analysis for these micronutrients. However, the application of such a rule may be similarly limited to the YFML. Sufficient observations for the development of such a rule were not available from these studies.

The discriminant rule developed in this study is probably only useful in growth chamber or greenhouse situations, as the data were obtained under highly controlled conditions. While observations from Cu-deficient leaves were successfully discriminated from observations from Fe-, Zn-, and Mn-deficient leaves and observations from Mn-deficient leaves were discriminated from observations from Zn-deficient leaves, there was some confusion between observations from Mn-deficient leaves and from Fe-deficient leaves. The ability to discriminate observations from Fe- and Zn-deficient leaves appears limited, at least with the predictors used in this analysis; however, measuring absorbance at 0.830 µm at the same time as F5min, a predictor suggested by Adams et al. (2000), may be useful in improving the discrimination of Fe deficiency from other deficiencies.

Considerations for Improvement and a Classifier for the Field
The discrimination analysis method described here shows promise for the development of a similar discriminant rule for field detection of micronutrient deficiencies or other stresses in soybean, however there are a number of additional considerations regarding the method.

This analysis assumed that the prior probability of any observation being classified as a control or Mn, Fe, Zn, or Cu deficient was equal. This is not an unreasonable assumption, since the rule is intended for situations where leaves are not showing obvious symptoms. This analysis also assumed, however, that the costs of misclassification are equal (e.g., the cost of fertilizing vs. the cost of yield loss), which is not necessarily the case. For example, if a discriminant rule determines that an observation comes from a Mn-deficient plant when it actually comes from a Mn-adequate plant, then there is potential for a grower to spend money on a treatment that is not necessary. Conversely, if a discriminant rule determines that an observation comes from a Mn-adequate plant when it actually comes from a Mn-deficient plant, then there is potential for the grower to lose income from reduced yields. Differences in these types of costs can be incorporated into the construction of a discriminant rule and soaffect how observations are classified.

The number of observations per group in this discriminant analysis with four predictors was small, except for the control group. Therefore, the error rates found for the various deficiencies may be somewhat inaccurate. A discriminant rule developed for field use should contain at least 60 observations per group to get a 0.95 reliable estimate of the error rate (Van Genderen et al., 1978).

Another potential source of error may stem from the assignment of observations to groups. Assignment of observations to groups was based on the concentration of the selected nutrients in tissue as determined by ICP analysis. Small errors in the ICP analysis could result in observations being classified to the wrong groups. Incorrect classification into groups could significantly alter group means and inflate covariance. This would reduce the sensitivity and accuracy of a discriminant rule, particularly if the covariance is greatly increased.

One way to address this is to incorporate uncertainty into the rules; e.g., attach probabilities to an observation being Mn deficient when it has a given leaf [Mn]. For example, the probability that an observation is Mn deficient increases as leaf [Mn] decreases below 12 g kg-1. One may wish to assign a probability of 0.50 to a leaf [Mn] of 10 g kg-1. By use of appropriate statistical methods, this probability can be used to modify the probabilities of group assignment determined by the classification rule and so improve accuracy.

Concerns for the Field
There are several concerns that should be addressed before attempting to apply the results of this work to the field. Chief among these are potential interferences from other stresses, especially macronutrient deficiency, water stress, and light stress. Fluorescence and reflectance are affected by these stresses (see references in Table 6) . Severe deficiencies in N and S were classified by the rule developed in this study most often as Fe deficient (Adams, 1996). Should the spectral measure for Fe proposed briefly above and in Adams et al. (2000) be developed and shown to be specifically sensitive to Fe deficiency, this problem should be eliminated.


View this table:
[in this window]
[in a new window]
  		Table 6 References reporting on the effects of certain stresses on fluorescence or reflectance

 
Spectral measures sensitive specifically to water stress have been developed and are capable of separating water-stressed leaves from N-stressed and control leaves (Peñuelas et al., 1993, 1994). Other spectral measures—either fluorescence, reflectance, or other nondestructive sampling techniques—will need to be developed, preferably spectral measures that are specific to one or two stresses. While this may not be possible for all stresses, we emphasize a point made earlier. A discriminant rule can, if properly trained, reduce the range of possible stresses from several to just one or two. With additional information, such as a priori knowledge of what stresses are likely to occur at a give site, a reduction in the number of likely stresses or an indication of the likelihood that a specific stress is present may be all that is necessary to make a management decision.

Morning measurements are advisable, to eliminate any potential interferences from midday water stress. Furthermore, the device used to measure fluorescence should be light-proof to allow for dark adaptation of the leaf tissue and to control all incident light levels during fluorescence measurement. Varying light intensities during the day from sun or cloud movement affect fluorescence (Demmig-Adams et al., 1995).

In terms of field scouting individual plants within a field over a growing season, the most unstable periods for measuring fluorescence and reflectance are the beginning and end of a growing season (Kharuk et al., 1994; Gates, 1970). Since monitoring plants during the early growing season is desirable for detecting early deficiencies, this is an important concern.


   Conclusions
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Results
 Discussion
 Conclusions
REFERENCES
 
The four spectral measures YI, NDVI, Fo/Fv, and Fo/F5min were used to discriminate effectively Mn and Cu deficiencies from each other and from Fe and Zn deficiencies, as well as controls at the leaf level from plants grown in solution culture. Fe and Zn deficiencies were not discriminated very well, although the number of observations from Fe- and Zn-deficient leaves used to develop the discriminant rule were relatively small. Spectral measures sensitive specifically to Fe and Zn deficiency are needed. A spectral measure sensitive specifically to Fe deficiency was proposed in Adams et al. (2000).

In addition to developing spectral measures sensitive specifically to Fe and Zn deficiency, the accuracy and utility of the discriminant rule may further be improved by considering posterior probabilities of group membership, introducing uncertainty into the rules used to determine group membership of the training data, and by limiting the training and use of the discriminant rule to YFML, as is common practice in plant tissue analysis for these nutrients.

The discriminant analysis method is appropriate for developing a tool capable for stress discrimination in the field, although additional spectral measures beyond those used in this analysis will be required. While it seems unlikely that a spectral measure will be identified that is specific to every stress (including, for example, insect damage), the ability to reduce the list of likely stresses from many to one or two may be sufficient for management purposes.


   ACKNOWLEDGMENTS
 
We wish to acknowledge the support of Tom Owens, Section of Plant Biology, Cornell University for the use of his PAM fluorimeter, and for input into the interpretation of fluorescence measures. We also wish to thank Danielle Brady and Simon Cook for helpful comments during the preparation of the manuscript.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 Materials and methods
 Results
 Discussion
 Conclusions
 REFERENCES
 
Research conducted while the senior author was with the USDA Plant, Soil & Nutrition Lab.

1 Mention of a trademark, vendor, or proprietary product does not constitute a guarantee or warranty of the product by USDA-ARS and does not imply its approval to the exclusion of other products or vendors that may also be suitable. Back

Received for publication September 4, 1998.
   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Results
 Discussion
 Conclusions
 REFERENCES
 





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 ISI Web of Science (8)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Adams, M. L.
Right arrow Articles by Peverly, J. H.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Adams, M. L.
Right arrow Articles by Peverly, J. H.
Agricola
Right arrow Articles by Adams, M. L.
Right arrow Articles by Peverly, J. H.
Related Collections
Right arrow Soybean
Right arrow Crop Growth and Development
Right arrow Nutrient Management

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