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


     

Published online 1 September 1999
Published in Agron J 91:801-806 (1999)
© 1999 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 (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Liu, W.
Right arrow Articles by van Santen, E.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Liu, W.
Right arrow Articles by van Santen, E.
Agricola
Right arrow Articles by Liu, W.
Right arrow Articles by van Santen, E.
Related Collections
Right arrow Crop Genetics
Right arrow Forage Management
Right arrow Other Forage Crops
Right arrow Plant Analysis
Agronomy Journal 91:801-806 (1999)
© 1999 American Society of Agronomy

FORAGE & GRAZING MANAGEMENT

Population Differentiation, Spatial Variation, and Sampling of Tall Fescue under Grazing

Weiguo Liua, Elizabeth A. Guertala and Edzard van Santena

a Dep. of Agronomy and Soils, Auburn University, 202 Funchess Hall, Auburn, AL 36849-5412 USA

evsanten{at}acesag.auburn.edu


   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 Materials and methods
 Results and discussion
 Conclusions
 REFERENCES
 
Tall fescue [Festuca arundinacea Schreb.] is the most important cool-season perennial forage grass in Alabama and the southeastern USA. Genetic variation is essential for breeding improved cultivars, and understanding factors influencing genetic variability in pastures is important if material from existing pastures is to be used in a breeding program. This study was conducted to determine the extent of differentiation for agronomic traits in pastures grazed long-term and to investigate possible spatial variation and its effect on sampling. Three populations from permanent pasture treatments of the USDA SARE cropping system trial in Virginia were sampled: (i) pure tall fescue fertilized with N, stocked continuously (Fescue + N); (ii) tall fescue–alfalfa (Medicago sativa L.) mixture used as hay and pasture (Fescue + alfalfa); and (iii) tall fescue–red clover (Trifolium pratense L.) mixture used as hay and pasture (Fescue + red clover). The tall fescue cultivar was endophyte-free Ky 31 [fescue endophyte: Neotyphodium coenophialum; syn. Acremonium coenophialum]. Plants from these paddocks were established in central Alabama in 1995. Original seed from the SARE trial were also germinated for establishing the original population. Ex situ evaluation was conducted in Alabama (1995–1997). Compared with plants derived from the original seed lot, plants derived from pastures under grazing had significantly earlier maturity, higher dry matter (DM) yield per plant, and larger plant diameter, indicating population differentiation in response to grazing. No significant differences were observed among populations with different pasture management treatments. Statistical and graphical analysis of spatial variation of agronomic traits showed no spatial relationships in any of the six sampled paddocks. Bootstrap estimates of minimum and maximum values indicated that 25 individuals per paddock captured most of the phenotypic variation within each paddock. A random walk approach covering the entire unit being sampled seems therefore to be an appropriate strategy for sampling similar pastures to obtain base material for a breeding program.

Abbreviations: DM, dry matter yield • SARE, Sustainable Agriculture Research and Extension Program


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
Materials and methods
 Results and discussion
 Conclusions
 REFERENCES
 
TALL FESCUE provides the primary ground cover on an estimated 12 to 14 million ha in the USA, making it the most widely grown cool-season perennial forage grass (Buckner et al., 1979). It forms the forage base for beef cow–calf (Bos taurus) production in the east-central and southeastern USA, supporting over 8.5 million beef cows (Hoveland, 1992). Due to its importance as a forage crop, there is a need to breed tall fescue cultivars with improved performance (e.g., high yield, late maturity, and grazing tolerance). When tall fescue pastures are used as the source materials for cultivar development, plants are generally collected in a random manner without regard to possible genetic stratification. Plant populations may undergo genetic differentiation due to environmental disturbance such as animal grazing (Allen and Marlow, 1994; Gilfedder and Kirkpatrick, 1994; Singh et al., 1995), edaphic variation (Snaydon, 1987; Snaydon and Davis, 1976), soil toxins (Ducousso et al., 1990), chemical treatment (Hassan, 1989), or soil fertility (Davies and Snaydon, 1973; Helgadottir and Snaydon, 1986; McNeilly and Roose, 1984). It is possible that genotypes with similar agronomic traits are clustered together because of the similarity of localized environmental conditions. This is particularly true in situations where there is limited or no recruitment of new individuals, as may be the case in heavily grazed pastures. If a spatial genetic structure is present in a field, it would be more appropriate to sample plants in a more systematic manner. It is, therefore, of practical importance to detect and describe possible spatial patterns in pastures to provide the necessary information for pasture sampling designs.

Geostatistics, a statistical method first developed by geologists (Matheron, 1963), can be used to describe phenomena showing variations that are not randomly distributed in space. Geostatistical analysis uses the geographic location of individual observations to quantify spatial correlation among treated plots from field experiments (Ball et al., 1993; Brownie et al., 1993; Stroup et al., 1994), to characterize the geographic variability of soils (Ovalles and Collins, 1988), and to improve sampling design by utilizing the spatial dependence of soil properties within sampling regions (Di et al., 1989). Recently, geostatistics has been used to identify spatial genetic structure in wild populations of perennial ryegrass (Lolium perenne L.) (Monestiez et al., 1994) and to establish a core collection of natural populations (Charmet and Balfourier, 1995; Charmet et al., 1994).

Many important cool-season forage cultivars, among them Ky 31 tall fescue, are the result of ecotype selection or selection from existing cultivars (Casler et al., 1996). Often existing pastures or hayfields serve as the source populations for cultivar development. The purpose of sampling these pastures is to select genotypes from a source population covering as much of the genetic variation as possible. Generally, the larger the sample size, the better the sample represents the source population. Large samples will capture more genetic variation, but they are more costly. It is important to achieve good genetic representation at a reasonable cost. From a practical standpoint, it is important to study the pasture sampling of genetic variability as affected by sample size.

The bootstrap method of statistical inference is a computer-based method for assigning measures of precision to statistical estimates, as first introduced in 1979 by Efron to estimate standard errors (Efron and Tibshirani, 1993). The nonparametric bootstrap can be used to estimate the sampling distribution of an estimator with an unknown probability density from the data in a single sample. Xie and Mosjidis (1996) used the bootstrap method to evaluate the performance of yield stability parameters. Bootstrap estimates and their standard errors of the parameters could be useful indicators of the relative performance of the parameters of concern. It is logical and practical to use bootstrap techniques to determine the effects of sample size on the amount of genetic variability of agronomic traits such as maturity, DM yield, and plant diameter in a sample of plants.

The objectives of this study were to (i) determine the effect of longer-term grazing and different management practices on pastures with regard to the changes in maturity, DM yield per plant, and individual plant diameter; (ii) determine the genetic variation of these agronomic traits as it relates to the distance between plants within a paddock; and (iii) use the bootstrap technique to determine the effects of sample size on the amount of genetic variability of these agronomic traits.


   Materials and methods
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and methods
Results and discussion
 Conclusions
 REFERENCES
 
The experimental material originated from the permanent pasture treatments of a USDA SARE cropping systems trial in Virginia (Allen et al., 1995). Three treatments, replicated four times in a randomized complete block design were established at the Virginia Polytechnic Institute Research Farm, Whitethorn, VA, in autumn 1988: (i) pure tall fescue fertilized with N, stocked continuously (Fescue + N); (ii) tall fescue–alfalfa mixture used as hay and pasture (Fescue + alfalfa); and (iii) tall fescue–red clover mixture used as hay and pasture (Fescue + red clover). The tall fescue cultivar in all cases was endophyte-free Ky 31 [i.e., free of the fescue endophyte Neotyphodium coenophialum (Morgan-Jones & W. Gams) Glenn, Bacon & Hanlin; syn. Acremonium coenophialum Morgan-Jones & W. Gams]. Grazing commenced in spring 1989 and continued until March 1995, when we took samples from each of three paddocks (grazing treatments) in Blocks I and II.

Pasture Sampling
Three considerations guided our sampling: (i) sampling points should be close enough to detect intrapaddock variation, (ii) sampling points should be far enough apart to limit the total number of plants in the evaluation trial to a manageable number (less than about 5000), and (iii), in the evaluation trial, the average distance between plants from a given paddock should be substantially less than the sampling distance within a paddock, so that the spatial relationships among plants in the original trial will not be obscured by spatial variation within plots in the evaluation trial.

Plants were collected at 6-m intervals in both X and Y directions of each paddock (average size 1.4 ha). Pieces comprising four to six tillers were removed from the nearest growing tall fescue plant (= genotype) and placed into linen soil sample bags. Each sample's position in the field was identified by X–Y coordinates. Depending on the size of the paddock, between 350 and 400 samples taken. Tillers were transplanted into Cone-tainers and grown during summer 1995. Individual genotypes were subcloned three times before transplanting. Seed was also germinated from the original seed lot that was used to establish the pasture trial. The resultant genotypes were established 6 mo prior to sampling paddocks, to achieve equal size with plants originating from pastures. These original seed lot genotypes were subcloned later along with the pasture samples.

Evaluation Trial
The ex situ evaluation experiment was conducted at the Plant Breeding Unit, Tallassee, AL (32°42' N, 85°53' W), beginning in autumn 1995, when genotypes were transplanted to the field as spaced plants (30 by 30 cm). The soil type was a Hiwassee sandy loam (clayey, kaolinitic, thermic Typic Rhodudults). Each treatment (plot) consisted of 400 (20 x 20) genotypes from a single paddock, assigned at random to position within a plot, with a 30-cm spacing between plants; extra plants of Ky 31 were used to provide even competitive conditions for all genotypes and to balance all plots in the evaluation trial to 400 plants. Plots containing plants from the original seed lot were somewhat smaller (160 plants at a 30-cm spacing) as germination of the stored seed was only 50%. There were seven treatments in all, six representing the six paddocks plus another one containing plants grown from the original seed lot. The seven treatments were evaluated in a randomized complete block design with three replicates.

Nitrogen fertilizer was applied at rates of 22.4 kg N ha-1 before transplanting, and 33.6 kg N ha-1 each in February and October. Herbicides were applied several times before and after transplanting to prevent weed competition, especially bermudagrass [Cynodon dactylon (L.) Pers.].

Individual genotypes were first harvested in late April 1996, and again in mid-April 1997, by removing all aboveground biomass at a height of 5 cm. All residual growth was removed to a height of 10 cm in late November of each year. Maturity on the Simon and Park (1983) scale, DM yield per plant, and plant diameter were determined on an individual genotype basis. Plant diameter was determined after harvest by placing a plexiglass disk which had concentric circles in 1-cm increments marked on it onto the harvested plant. The target growth stage was an average of 54 on the Simon and Park scale for the trial, equivalent to plants having tillers 50% emerged from the boot. On average, we harvested the plants slightly earlier than planned (Table 1) .


View this table:
[in this window]
[in a new window]
  		Table 1 Least squares population means and contrast P-values for maturity, dry matter (DM) yield, and plant diameter from an ex situ spaced-plant evaluation of plants from the three pasture treatments of the Virginia Cropping System Trials and plants grown from original seed used to establish the trial. Evaluation was conducted at the Plant Breeding Unit, E.V. Smith Research Ctr., Tallassee, AL

 
Data Analysis
Mixed linear models (Littell et al., 1996) were used to analyze the data from the evaluation trial. Treatment, genotype within treatment, year, and treatment x year effects were considered to be fixed; replicates and experimental error were considered to be random effects. Least square means for maturity, DM yield per plant, and individual plant diameter were calculated for each paddock and for genotypes within paddocks. Least squares genotype means for each of 18 paddock–trait combinations were analyzed using geostatistical techniques (Clark, 1979; Cressie, 1991; Isaaks and Srivastava, 1989) to estimate the variation in the original paddocks.

Suppose the distance between two samples (genotypes) within a given paddock is h. The semivariance {gamma}(h) can be used to describe the spatial relationship between these two samples. The semivariance describes the degree of dependence existing between samples as a function of the distance between them. The semivariance {gamma}(h) can be calculated as

(1)
where g equals the observed value for a given trait (e.g., yield, maturity, plant diameter), x denotes the position of one sample in a pair and x + h the position of the other, n is the total number of pairs at a given distance h, and {gamma}(h) is the experimental mean semivariance. Semivariograms are often presented in graph form, with the horizontal x-axis being increasing increments of h (the lag distance) and the vertical y-axis the semivariance at each lag distance. Positive definite models are fit to these experimental curves by a variety of mathematical procedures, providing coefficients that describe the spatial relationships within the data set. The most common model fit to geostatistical data is the spherical model, given mathematically as

(2)

where a is the range of influence; c is the sill of the semivariance, which is equal to the ordinary sample variance; and h is the lag distance, the distance between two observations.

The range is the distance at which samples become independent of one another. When samples are closer together than the range, some type of spatial relationship exists between those samples and they are not independent of each other. Samples farther apart than the range are independent, and the assumption of randomness when selecting plants would apply. The nugget variance or nugget effect is the semivariance when distance between samples equals zero (i.e., ). This represents the unexplained variance, often caused by measurement error or by variability of properties which could not be detected at the shortest sampling distance employed (Ovalles and Collins, 1988).

The Geostatistical Analysis Program (MGAP; RockWare, 1993) was used for initial data analysis and graphing, followed by the PROC MIXED procedure of SAS (Littell et al., 1996) to fit models and test significance of spatial variability.

Statistical significance of the spatial covariance parameters was tested with a likelihood ratio statistic (based on restricted maximum likelihood). The nugget effect can be tested similarly, using the restricted maximum likelihood log-likelihoods for nugget and no-nugget models (Littell et al., 1996; Self and Liang, 1987).

Seven data sets consisting of individual genotypes within the six paddocks and the original population with least squares estimates of maturity scores, DM yield per plant, and individual plant diameter were obtained from the analysis of the evaluation trial for further analysis.

The bootstrap method (Efron and Tibshirani, 1993) was used to analyze these seven data sets per dependent variable. The procedure is described thus:

  1. Select B independent bootstrap samples x*1, x*2, ..., x*B, each consisting of n data values (sample size) drawn randomly with replacement from X. Each sample is called a bootstrap replicate.
  2. For each bootstrap replicate calculate the statistic of interest (i.e., mean, maximum, minimum).
  3. Obtain an estimate of the expected value and standard error of a statistic by calculating the mean and standard deviation among the values for the B bootstrap replicates.

Based on this theory, a SAS program was developed to obtain bootstrap estimates of the expected value of the mean, maximum, and minimum and also of the corresponding standard deviations at various sample sizes for three agronomic traits. The SAS RANUNI function (SAS, 1989) was used to conduct bootstrap resampling with replacement, and a SAS macro was used to generate bootstrap replicates. The standard deviation among bootstrap replicates is a bootstrap estimate of the standard error. We used 500 replicates at each sample size (ranging from 2 to 400) for all seven data sets.

To characterize the sample size effects on bootstrap estimates of maximum, minimum, and standard error of mean maturity, DM yield, and individual plant diameter, estimates were plotted against sample size as the independent variable.


   Results and discussion
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Results and discussion
Conclusions
 REFERENCES
 
Population Comparisons
Compared with the population derived from the original seed lot, populations derived from pastures under grazing were earlier maturing, had higher DM yield per plant, and were larger in diameter (Table 1). Previous research by Vaylay and van Santen (1999) with four tall fescue cultivars indicated that the derived population after two years of grazing had significantly (P <= 0.05) earlier maturity, higher DM yield, and larger plant diameter compared with originals of the same genetic background. We confirmed those results in the present study with plant material from pastures in Virginia that had been grazed for six years. The presence of genetic variation in pastures is a prerequisite for adaptation and evolutionary change. Genetic variability enables pastures to undergo population changes to better adapt to new environments, such as animal grazing. The shift towards earlier maturity in derived populations compared with originals may have evolutionary advantages. Kemp (1988) argued that earlier-maturing genotypes in a whole range of cool-season species have better establishment, leading to better survival in the first winter. It may also be argued that plants with inherently earlier maturity could shorten the time of the life cycle required to regenerate progeny that are better adapted to the new environment. Small differences in relative maturity indicate that accumulation of change happens over time.

No significant differences were observed among different pasture management treatments (Table 1). Plant populations are dynamic entities, and their genetic structure responds to various types of disturbance. Vaylay and van Santen (1999) observed significant genetic diversity among tall fescue populations derived from original seed lot, ungrazed survivors, and survivors of the same genetic background after two years of grazing for two experimental populations, but not in the old cultivar Ky 31. Because Ky 31 is an old cultivar, only two years of exposure to grazing is unlikely to cause drastic genetic shifts; however, considerable genetic diversity among Ky 31 ecotypes collected in Alabama from pastures with ages ranging from 18 to 38 years was reported by van Santen and Collins (1991). These differences among populations were largely attributed to climatic variables such as long-term normal precipitation and temperature. The material used in the present study was Ky 31, which, although subjected to six years of different pasture management treatments, showed no significance among different management treatments. It appears, Vaylay and van Santen (1999) concluded, that the mere process of establishing a cultivar through seeding changes that cultivar.

Spatial Variation
Lag distances of the constructed semivariograms covered 50% of the shorter dimensions of each paddock (40–45 m). All semivariograms were omnidirectional; no detectable anisotropy was present (data not shown). A semivariance estimate for a given lag-distance was estimated from at least 500 pairs and some estimates were based on as many as 2000 pairs (Table 2) . The spatial variation in every one of the 18 paddock–trait combinations was pure nugget effect; the estimate at a lag distance equal to the sampling distance was 95% of the population variance estimate for a given trait (Table 2). Even though some spatial statistical analyses (e.g., maturity in the fescue + alfalfa paddock of Block II) indicated significant spatial variation, the estimated range of 7.5 m was close to the actual minimum distance sampled and the graphical presentation (Fig. 1) suggested pure nugget effects. Both pasture blocks responded identically, even though the cropping history before stand establishment was different. Block I had been in crop land, whereas the other block had been in unimproved pasture (V. Allen, 1998, personal communication).


View this table:
[in this window]
[in a new window]
  		Table 2 Number of lags, average distance (h), number of pairs, and semivariance estimates of constructed omnidirectional semivariogram for dry matter (DM) yield, maturity, and plant diameter from the Fescue + N paddock in Block I of the Virginia Cropping System Trials. Evaluation was conducted at the Plant Breeding Unit, E.V. Smith Res. Ctr., Tallassee, AL

 


View larger version (18K):
[in this window]
[in a new window]
  		Fig. 1 Constructed omnidirectional semivariograms for DM yield for the three grazing treatments from Block I of the Virginia Sustainable Agriculture Research and Education (SARE) cropping systems trial

 
Previous research (Monestiez et al., 1994) identified two spatial genetic structures, with ranges of 120 and 300 km, for heading date and summer growth in wild populations of perennial ryegrass distributed from throughout France. The authors argued that the spatial structure of 120-km range was caused by gene flow, and the 300-km range structure resulted by selection pressure such as climate. Small-scale (<=7.5 m) spatial structures may exist within pastures, but could not be detected in our study. Two possible reasons may be cited: (i) our methods were not sensitive enough to detect variation because we were dealing with phenotypic differences, or (ii) genetic variation on such a small scale is truly random. If our methods were too crude, molecular markers may offer a big improvement and may allow us to detect spatial variation. It seems clear, however, that for all practical purposes in applied forage breeding we can operate on the assumption that variation within a paddock, at least for sampling on a 6-m grid, is random with respect to the distance between two samples.

Bootstrap Sampling
The questions remains, however, how many plants should be sampled per pasture. Given a fairly high broad-sense heritability (repeatability), quantitative traits can be used to address this question. Maturity in Ky 31–derived populations generally fits these constraints. Eight out of 10 Ky 31–derived populations in a study by van Santen and Collins (1991) had broad-sense heritabilities greater than or equal to 0.74.

To avoid bias in favor of high seed yield, sampling of vegetative material is generally preferred over collection of seed (Tyler et al., 1987). The authors furthermore state that "vegetative sampling provides a sample of what is actually growing in a given environment and thus is more likely to reflect the adaptation." They argue that a minimum of 25 to 30 plants should be collected from each population. Similarly, Burton and Davies (1984) suggest taking 30 "vegetative units" from each grazed pasture.

Bootstrap sampling from the population of 160 individuals from the original population or 350 to 400 individuals from each paddock of the grazing trial indicated that 25 to 30 samples indeed captured most of the phenotypic variation for maturity (Fig. 2) . The response to sampling of individuals from paddocks in Block II was very similar (data not shown). Early-maturing entries (high score) were obtained at rather small sample sizes. This is probably the result of our approach to harvesting the trial, in that we were aiming to harvest at an average maturity score of 54. It was difficult to select the latest-maturing (low score) individuals without going to larger sample sizes (Table 3) . This may not be all that important in practical breeding, because of the relative ease of shifting maturity in populations of cross-pollinated forage grasses. For purposes of germplasm conservation, however, it may be a different matter.



View larger version (29K):
[in this window]
[in a new window]
  		Fig. 2 Bootstrap estimates of the mean, minimum, and maximum maturity value as a function of sample size based on sampling from the original population and the three grazing treatments from Block I of the Virginia Sustainable Agriculture Research and Education (SARE) cropping systems trial. The top line in each panel represents the maximum, the middle line the mean, and the bottom line the minimum value

 

View this table:
[in this window]
[in a new window]
  		Table 3 Estimated sample size needed for obtaining genotypes with latest maturity in tall fescue paddocks

 

   Conclusions
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Results and discussion
 Conclusions
REFERENCES
 
Results obtained in this study confirmed previous findings by Vaylay and van Santen (1999), that the population after seeding differs from the population used to establish a stand. The absence of any spatial dependency within paddocks, at least for sampling on a 6-m grid, indicates that a `random walk across a pasture' approach is entirely appropriate for obtaining base material for a breeding program. Bootstrap estimates for maximum and minimum maturity scores indicated that rather small sample sizes (25 individuals) capture most of the variation. None of the foregoing means that there may never be any spatial variation nor that random sampling is always appropriate. Any collection efforts need to be tempered by a knowledge of the particular location being sampled.SAS Institute 1989; Thompson 1992


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 Materials and methods
 Results and discussion
 Conclusions
 REFERENCES
 
Contribution is research conducted in partial fulfillment of the first author's M.S. degree requirements at Auburn University. Research supported in part by Hatch funds allocated to Alabama Agric. Exp. Stn. Project ALA03-005.

Received for publication March 25, 1998.
   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Results and discussion
 Conclusions
 REFERENCES
 




This article has been cited by other articles:


Home page
 Agron. J.Home page
K. R. Brye, J. M. Norman, E. V. Nordheim, S. T. Gower, and L. G. Bundy
Refinements to an In-Situ Soil Core Technique for Measuring Net Nitrogen Mineralization in Moist, Fertilized Agricultural Soil
Agron. J., July 1, 2002; 94(4): 864 - 869.
[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 (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Liu, W.
Right arrow Articles by van Santen, E.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Liu, W.
Right arrow Articles by van Santen, E.
Agricola
Right arrow Articles by Liu, W.
Right arrow Articles by van Santen, E.
Related Collections
Right arrow Crop Genetics
Right arrow Forage Management
Right arrow Other Forage Crops
Right arrow Plant Analysis

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