| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Dep. of Plant and Soil Sciences, 12F Stockbridge Hall, Univ. of Massachusetts, Amherst, MA 01003
b Soil, Crop, and Atmospheric Sciences, 1021 Bradfield Hall, Cornell Univ., Ithaca, NY 14853
* Corresponding author (sebdon{at}pssci.umass.edu)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Abbreviations: AMMI, additive main effect and multiplicative interaction • ANOVA, analysis of variance • GE, genotype by environment • IAS, interaction score • NTEP, national turfgrass evaluation program • PCA, principal component analysis • SS, sum of squares • SVD, singular value decomposition
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Ignoring the interaction is problematic when the GE interaction is larger than the genotype main effect, which is a common scenario in yield trials (Gauch and Zobel, 1996). In turf variety trials large GE interaction complicates an agronomist's or breeder's research because turfgrass performance is less predictable and cannot be interpreted on the basis of only genotype means and environment means. Furthermore, interactions complicate cultivar recommendations because genotypes must be targeted to specific locations.
Interactions associated with turfgrass performance trials have been ignored in part because of the complexities of NTEP tests. Variety trials sponsored by NTEP may evaluate 100 or more cultivars and experimental lines at 20 or more locations evaluated over a five-year period, resulting in numerous and potentially complex genotype-location-year combinations. Secondly, GE interactions are problematic because there is a lack of effective statistical tools available to adequately summarize complex data structure typical of NTEP trials.
In yield trials, standard statistical methods that have been applied include (i) ANOVA, (ii) principal component analysis (PCA), (iii) linear regression (LR), and (iv) AMMI models. These methods are often inadequate in effectively treating complex data structure typical of yield trials (Zobel et al., 1988). AMMI analysis has been shown to be effective because it captures a large portion of the GE sum of squares, it cleanly separates main and interaction effects that present agricultural researchers with different kinds of opportunities, and the model often provides agronomically meaningful interpretation of the data (Gauch, 1992). Additionally, results from AMMI are useful for performing mega-environment analysis in which a crop's growing region is subdivided into homogenous subregions that have similar interaction patterns and cultivar rankings, simplifying cultivar recommendations (Gauch and Zobel, 1997).
Gauch and Zobel (1996) provide several examples of AMMI application to understanding GE interactions in agricultural yield trials. AMMI statistics presented in biplots can be used to provide insightful interpretation of data from large, complex experiments (Gabriel, 1971; Bradu and Grabiel, 1978; Zobel et al., 1988; Gauch, 1992; Ebdon et al., 1998). Results from AMMI analysis can often reveal plant morphological and physiological relationships that cause genotypes to interact with environments (Zobel et al., 1988; Gauch and Zobel, 1996). Although AMMI analysis of yield data does not use environmental data (only yield data), environmental factors such as rainfall, average daily maximum and minimum temperature, altitude, latitude, nitrogen fertilization, irrigation, and clay content have often been found to be correlated with AMMI environmental (interaction) statistics (singular vector values) (Gauch, 1992, p. 231–236; Romagosa et al., 1996; Gauch, 1998). The results of AMMI analysis are useful in supporting breeding program decisions such as specific adaptations to target (tolerances to disease, heat and drought, cold) and selection of environments or test site locations (Gauch and Zobel, 1997).
The application of AMMI has been limited to yield trials, the results of which may not necessarily be relevant to turf trials. Turfgrass performance trials are distinctly different from yield trials. Unlike crop varieties, superior yield (productivity) in turfgrass does not equate to superior turfgrass performance (Shearman, 1985; Youngner, 1985). Turfgrass quality is based on a visual rating system (NTEP uses a 1 to 9 scale with 9 indicating the highest quality) assessing the aesthetic appeal (and function) of turf which integrates several quality components such as color, density, uniformity, and texture (Turgeon, 1980). The turfgrass rating system is subjective and evaluators may use different parts of the rating scale (Horst et al., 1984) because of evaluator bias. However, there is general agreement among evaluators in cultivar ranking (Skogley and Sawyer, 1992), therefore evaluator bias cause little interaction. It would appear likely that turfgrass agronomists and breeders could benefit from a more thorough treatment of GE interactions afforded by AMMI analysis.
The objective of this research was to provide biologically meaningful interpretation of genotype by environment interactions associated with NTEP performance trials using AMMI. In a companion paper, cultivar recommendations are discussed emphasizing AMMI predictive accuracy, statistical efficiency, and mega-environment analysis.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
For some locations, every evaluation year was represented, while for other locations not all years were available for analysis. For the Kentucky bluegrass trial, nine of the 17 NTEP locations provided data for all 125 genotypes and each year of the 5-yr evaluation period including Post Falls, ID; Ames, IA; Carbondale, IL; Lexington, KY; Beltsville, MD; North Brunswick and Adelphia, NJ; Kingston, RI; and Haymarket, VA. One location (Oregon) provided 4 yr of turfgrass quality data, six locations (Fort Collins, CO; Urbana, IL; East Lansing, MI; Martinsville, NJ; Blacksburg, VA; and Pullman, WA) provided 3 yr of yearly data, and one location (Marysville, OH) provided 2 yr of annual data. For the 1990 perennial ryegrass trial, eight of the 19 NTEP locations provided data for all 123 genotypes and each year of the 4-yr evaluation period including Ames, IA; Carbondale, IL; Lexington, KY; Beltsville, MD; North Brunswick, NJ; Oregon, Blacksburg, VA; and Pullman, WA. Seven locations provided 3 yr of turfgrass quality data (Fort Collins, CO; Rathdrum, ID; Wichita, KS; East Lansing, MI; Adelphia and Martinsville, NJ; and Kingston, RI), three locations provided 2 yr of annual data (Silver Springs, MD; Lincoln, NE; and Marysville, OH) and one location provided one year of data (Richmond Hill, ON, Canada). For a detail listing of genotypes and NTEP site information refer to NTEP Final Reports for Kentucky bluegrass and perennial ryegrass.
In the 1990 Kentucky bluegrass trial, turfgrass quality of 125 genotypes involved 69 location-year combinations. In the 1990 perennial ryegrass test, 123 genotypes were evaluated at 60 location-year combinations. The large body of data available for genotypes from NTEP reports including disease response, summer response to drought, winter color, spring density and green-up (as well as other data) were used to provide information to help explain why genotypes interact with environments. Similarly, environmental data available from NTEP reports on each location were used to provide interpretation of AMMI environmental parameters relating to interaction including nitrogen fertility level applied, mowing height, soil test results (pH, P, and K), and irrigation schedules, as well as climatic data including temperature (average, maximum, and minimum), rainfall, latitude, and elevation that were obtained on each site.
AMMI Analysis
AMMI analysis of turfgrass performance data was performed by MATMODEL (Gauch and Furnas, 1991; Gauch, 1998). The data must be organized in a two-way layout such as genotypes and environments. Accordingly, the format used for analysis of NTEP data (a three-way table involving genotype, location, and year) was analyzed by AMMI by combining location and year to form environments (location-year combinations). For other restrictions and requirements in the data, see Gauch and Furnas (1991). For a detailed description of appropriate statistical fixed-effect models for AMMI and subcases, refer to Zobel et al. (1988).
The ANOVA model is,
 |
 |
g are genotype mean deviations (mean minus the grand mean), ße are the environment mean deviations, N is the number of SVD (singular value decomposition) axes retained in the model, 
n is the singular value for SVD axis n, 
gn are the genotype singular vector values for SVD axis n, 
en are the environment singular vector values for SVD axis n, 
ge are the interaction residuals, 
ge are the AMMI residuals, and 
ger is the error term.
The eigenvalue for a given SVD axis is the sum of squares (SS) retained by that axis and it equals the square of the singular value, 2. The sum of the eigenvalues 
2 for the N axes, plus the residual SS for a reduced model, is equal to the GE interaction SS. Thus the interaction SS is partitioned by SVD into interaction axes SS and associated degrees of freedom, which allow for the use of F-tests to determine the significance of a given SVD axis (Gauch, 1992).
The goal of the analysis is to summarize the interaction SS with a few SVD axes (typically, N = 1 to 3), leaving a reduced model with residuals containing mostly noise. Only the early interaction axes retain pattern that has biological interpretation (Gauch, 1992). The units for µ, , ß, , , and are in the same units as the response Y (turfgrass quality rating). The singular vectors for genotype and environment are dimensionless. The genotype interaction scores (0.5 g) and the environment interaction scores (0.5 e) are in units which are the square root of the unit for Y. The abscissas of the AMMI biplot are the main effects (g and ße) while the genotype interaction scores (0.5 g) and environment interaction scores (0.5 e) are the ordinates. Thus the product of genotype scores and environment scores gives the estimated interaction which is in the same units as the original Y.
To provide biologically meaningful interpretation of GE interactions, simple correlations (r) between genotype interaction scores from AMMI and genotypic covariables obtained from NTEP final reports were determined for Kentucky bluegrass and perennial ryegrass. Similarly, environmental covariables for locations from the same NTEP reports were explored to identify important cultural, edaphic, and climatic factors correlated with environmental interaction scores.
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
0.5 e) and genotype scores (0.5 g) need to be explained in biologically meaningful terms. For both Kentucky bluegrass and perennial ryegrass data sets, genotype and environment (location-year) main effects and interaction were highly significant (p 
0.001). The only biologically interpretable interaction parameters were associated with the first interaction axis (IAS-1), whereas higher axes captured and discarded noise affecting predictive accuracy (Ebdon and Gauch, 2002). Although there appears to be terms beyond the first interaction axis that are not due entirely to noise (Ebdon and Gauch, 2002), IAS-1 is used here as a rough approximation of the true interaction.
Environments
Figures 1 through 4
are the AMMI plots for the Kentucky bluegrass and perennial ryegrass variety trials. To avoid clutter, genotypes and environments are shown separately (Fig. 1 through 3). The abscissa shows the main effects and the ordinate shows the IAS-1 scores that capture interaction effects. In all AMMI plots a horizontal, dashed line shows the interaction score of zero, and the grand mean is indicated as a vertical, dashed line.
|
|
|
|
The Post Falls, ID, location from the Kentucky bluegrass test (open triangle, Fig. 1) shows large variation from year-to-year in both main effect and interaction and therefore occupies a large region in the AMMI plot. Similarly, Carbondale, IL (closed triangle), is a location with a large main effect and interaction. However, the Carbondale site negative (-) interaction scores are opposite in sign and therefore opposite in the direction of its interaction with genotypes compared with Post Falls, ID. The interaction pattern with genotypes for the Carbondale, IL, location is similar for both variety trials (Fig. 1).
The Beltsville, MD, NTEP site (open circle) represents a location with small main effect and small interaction. Year-to-year turfgrass quality and environment interaction scores are highly predictable at this site because the corresponding markers occupy a small region in the interaction plot for both Kentucky bluegrass and perennial ryegrass variety trials (Fig. 1). Additionally, Ames, IA, East Lansing, MI, and Martinsville, NJ are other NTEP locations with small prediction regions (and hence high predictability). Conversely, locations such as Carbondale, IL, with a large interaction effect (and low predictability) is likely to be more variable in year-to-year cultivar rankings, making cultivar recommendations for such locations more difficult compared with NTEP locations such as Beltsville, MD. Other NTEP locations with low predictability included Lexington, KY and North Brunswick, NJ.
Similar interaction patterns were observed for many of the NTEP locations shared between the 1990 Kentucky bluegrass and perennial ryegrass trials. So, there appears to be some stability in year-to-year variation in interaction patterns and main effects across turfgrass species. It is important to recognize, however, that year-to-year variations in main effects have no effect on cultivar rankings, but interaction variations do. Also, some locations are inherently more predictable than others.
Environment scores from AMMI analysis (0.5 e) relating to interaction often have meaningful interpretation. Correlation analysis was performed between IAS-1 scores (0.5 e) and site data summarized in NTEP reports as well as climatic data that was obtained on each location. In the Kentucky bluegrass trial, environment IAS-1 scores were identified to have the highest correlation with mowing height of cut (r = -0.52, p 0.001) when compared with all other cultural and climatic factors evaluated. Mowing height was the single most important environmental factor explaining interaction. Specifically, locations in Fig. 1 for Kentucky bluegrass from the upper half of the interaction plot (locations with positive scores) were associated with a lower mowing height of cut (2.5 to 3.75 cm) compared with locations from the lower half of the plot (negative scores) which were maintained at a higher height of cut (5 to 7.5 cm). The single most important factor explaining environment interaction with perennial ryegrass was the level of annual nitrogen applied (r = -0.66, p 0.001). Locations in Fig. 1 for perennial ryegrass from the upper half of the interaction plot were associated with sites maintained under lower annual nitrogen (98 to 147 kg ha-1) compared with locations from the lower half of the plot which were maintained at higher nitrogen (245 to 294 kg ha-1). The variable year-to-year IAS scores for some locations such as Carbondale, IL, and Post Falls, ID (Fig. 1), as well as the correlation (r) reported here between cultural factors (mowing height and nitrogen) and IAS-1 scores (r2 accounting for only 27 and 44% of the total variation in interaction scores for bluegrass and ryegrass, respectively), suggest that other environmental factors may be operating to influence interaction. However, cultural intensity level (mowing height and nitrogen) seems to play a greater role than do climatic factors in GE interaction in turfgrass.
Genotypes
The AMMI plot for genotypes is shown in Fig. 2 and 3 (for Kentucky bluegrass and perennial ryegrass, respectively) where the ordinates are the IAS-1 genotype parameters (0.5 g) and the abscissas are the means for genotypes (averaged over environments). Genotypes of special interest have been identified in the interaction plots. Also, LSD (0.05) bars are included in Fig. 2 and 3 for genotype main effects.
A pattern clearly indicated in Fig. 2 and 3 is the linear relationship between interaction scores and main effects for genotypes. Specifically, the top performing and bottom performing cultivars have opposite interaction scores. For example with Kentucky bluegrass genotypes (Fig. 2), Midnight (top ranked cultivar and overall winner) and the lowest performing genotype Ginger, differ markedly in their interaction. Midnight has a large positive interaction score (0.77) while Ginger has a large negative score (-1.00). In perennial ryegrass (Fig. 3), Prelude II (the overall winner) has a large negative interaction score (-0.76) while Linn (the lowest rated) has a large positive score (1.20).
Consequently, the top performing and bottom performing cultivars are adapted to different environments. Their relative rank order varies greatly with environment, so these genotypes are narrowly adapted. Therefore, breeding genotypes having superior performance throughout the entire growing region is inconsistent with these results from turf trials as well as results from yield trials (Ceccarelli, 1989). Such hypothetical genotypes that are universal winners in all environments having an IAS-1 score of zero (broad adaptability) are so labeled in the AMMI plots (Fig. 2 and 3) as "hypothetical winners." Such "hypothetical winners" with superior performance in all environments (both favorable and unfavorable) assume that the same physiological and morphological adaptations are operating in all environments, contrary to research evidence (Burt and Christians, 1990).
Bluegrass genotypes Barzan and Preakness with interaction scores near zero (Fig. 2) do not interact much with environments, and therefore their rank orders are relatively stable. Similarly, ryegrass genotypes Lowgrow and PR 9118 have IAS-1 scores near zero (Fig. 3). These bluegrass and ryegrass genotypes are broadly adapted, but do not represent winning or superior performers. The experimental bluegrass varieties BAR VB 1169 and Ba 73-382 are examples of two genotypes that have the same means (main effect) but are distinctly different and opposite in their interactions (Fig. 2).
Using the data summarized in NTEP final reports for Kentucky bluegrass, IAS-1 scores and disease rating (leaf spot, Bipolaris spp.) were highly correlated (r = 0.70, p 0.001). Reaction to leaf spot in Kentucky bluegrass explained almost 50% of the total variation (r2) in genotype interaction scores. Bluegrass genotypes in Fig. 2 from the upper half of the interaction plot were associated with superior resistance to leaf spot compared with genotypes from the lower half of the interaction plot. Similarly for perennial ryegrass, IAS-1 scores and disease ratings (leaf spot and brown patch, Rhizoctonia solani) were correlated (r = -0.60, p 0.001 and r = -0.78, p 0.001, respectively) and explained 36 and 61% of the total variation (r2) in genotype IAS-1 scores. Regression of these two disease ratings together explained almost 72% of the total variation (R2) in perennial ryegrass IAS-1 scores. Therefore, GE interaction in NTEP variety trials is related to disease response and cultural intensity level.
Many of the 125 Kentucky bluegrass genotypes evaluated in the 1990 trial have been classified into six categories or groups according to various criteria and observations made by turfgrass breeders including pedigree (parentage), morphology, disease resistance, and place of origin (Murphy, 1994). The six categories include "aggressive" types characterized by high shoot density, "Bellevue" types having moderate shoot density and good disease resistance, "BVMG" (Baron, Victa, Merit, Gnome) types characterized by high seed yields and moderate disease resistance, "compact" types having excellent leaf spot resistance and slow vertical leaf extension rates, horizontal leaf orientation, and high shoot and leaf densities, "Mid-Atlantic" types having superior heat and drought performance, and "Midwest common" types characterized by susceptibility to leaf spot, and having rapid vertical leaf extension rates, vertical leaf orientation, and low shoot and leaf densities. Many of the genotypes are "unclassified" because they do not fit clearly into any one group.
Figure 2 shows the different Kentucky bluegrass genotypes in the AMMI plot identified by category type. Genotypes from the same category share similar interaction patterns and main effects. For example, "Midwest common" types and "compact" types are located in the interaction plot at a diagonal (opposite corners) and therefore differ in both their main effect and interaction. "Bellevue" and "BVMG" types differ in their interaction only. Many of the unclassified types are hybrids derived from Bellevue types and the genotype Baron and therefore overlap with most other categories. These genotypes represent a continuum incorporating disease response and plant morphology that will provide helpful interpretation with respect to why Kentucky bluegrass interact with environments.
Interpretation of GE Interactions
Figure 4 presents the AMMI-1 biplot with genotypes and environments shown together for Kentucky bluegrass and perennial ryegrass variety trials. The bluegrass biplot captures 78.1% of the total SS due to main effects and interaction associated with 8,625 genotype-environment combinations. The ryegrass biplot captures 89.3% of the total SS associated with 7,380 genotype-environment combinations. In Fig. 4, for any genotype-environment combination, the additive part (main effects) of the AMMI model equals the genotype mean plus the environment mean minus the grand mean, and the interaction (multiplicative part) is the genotype score times the environment score (see Zobel et al., 1988). For example, the bluegrass genotype Midnight growing in Adelphia, NJ, in 1992 has an additive effect of 6.53 + 5.02 - 5.56 = 5.99, and the interaction is estimated as 0.77 x 1.41 = 1.09. Therefore the AMMI-1 model gives a turfgrass quality estimation for Midnight growing in Adelphia, NJ, (in 1992) of 5.99 + 1.09 = 7.08. This prediction is as accurate as the raw data (means summarized in NTEP Final Reports) while for perennial ryegrass the AMMI-1 predictions are as accurate as would be the means based on 3.6 times as many replications (Ebdon and Gauch, 2002).
Turfgrass quality performance is better for genotypes growing in environments with large scores of the same sign, whereas poorer quality performance is expected with large scores of opposite sign. Therefore a genotype such as Midnight Kentucky bluegrass with a large positive score (0.77, Fig. 4) performs better in environments like Adelphia, NJ, in 1992 (score = 1.41, Fig. 4) having a score of the same sign because this combination results in a large positive interaction. Midnight performs poorer in environments such as Haymarket, VA, in 1995 with a large negative score (-1.07, Fig. 4) because this combination results in a large negative interaction. The opposite is the case for the bluegrass genotype Ginger with a large negative score (-1.00, Fig. 4); Ginger performs best in environments with a large negative score (Haymarket, VA, in 1995).
The GE interaction involves genotypes and environments; genotype and environment IAS-1 scores when considered alone are dimensionless (hence meaningless), so both are required to explain interaction. Interpretation of the GE interaction must make biological sense when genotype IAS-1 scores which reflect disease and environment IAS-1 scores that are associated with cultural intensity level are considered together. Results from AMMI analysis of the interaction suggest that the leaf spot susceptible "Midwest common" bluegrass types such as Ginger prefer the Haymarket VA NTEP location (in 1995), indicated by their large positive interaction for this site (Fig. 4). This most likely is due to the upright growth habit of common types (poor tolerance to close mowing) and susceptibility to leaf spot are better adapted to the higher mowing height of cut of Haymarket, VA (5 to 7.5 cm). The high mowing regime of Haymarket, VA would be associated with less leaf spot pressure (Beard, 1973). Conversely, under the low mowing typical of Adelphia, NJ (2.5 to 3.75 cm), leaf spot pressure would limit turfgrass quality performance of common types (indicated by their corresponding large negative interaction for this site, Fig. 4).
The opposite is true for "Compact" types such as the bluegrass Midnight. The low growth habit of compact types and superior tolerance to close mowing (Sheffer et al., 1978) along with excellent leaf spot resistance affords superior turfgrass quality performance at NTEP sites such as Adelphia, NJ (Fig. 4). The potential for leaf spot is greatest at this site because of close mowing (Beard, 1973).
Leaf spot and brown patch susceptible perennial ryegrass such as Linn prefer the East Lansing NTEP location in 1993 (Fig. 4). This most likely is due to the low nitrogen regime of East Lansing (98 to 147 kg ha-1), which would be associated with less leaf spot and brown patch pressure (Beard, 1973). Conversely, under the high nitrogen typical of North Brunswick, NJ (245 to 294 kg ha-1), leaf spot and brown patch pressures would limit turfgrass quality performance of Linn as well as other genotypes with large positive scores (Fig. 4). The opposite is true for the ryegrass genotype Prelude II. Excellent leaf spot and brown patch resistance affords superior turfgrass quality performance at NTEP sites such as North Brunswick (Fig. 4). The potential for nitrogen-induced disease is greatest at this site because of high nitrogen (Britton, 1969; Beard, 1973).
The two species are similar in their interaction with environments as it relates to cultural intensity and disease. However, they are distinctly different because environment scores in Kentucky bluegrass were correlated with mowing height of cut while nitrogen fertilization levels were important in explaining GE interaction in perennial ryegrass. This difference can be explained in part because the two species respond differently to mowing and nitrogen fertilization as it relates to disease. Kentucky bluegrass as a species is very diverse in its plant morphology and growth habit (Beard, 1973; Nittler and Kenny, 1976; Ebdon and Petrovic, 1998) and morphology is important in cutting height and disease tolerance (Beard, 1973; Sheffer et al., 1978). Compared with perennial ryegrass, there are greater intraspecific differences in Kentucky bluegrass morphology and tolerance to mowing. Accordingly, mowing height is relevant in explaining genotype by environment interaction in this species. Kentucky bluegrass and perennial ryegrass are both responsive to fertilizer nitrogen. However perennial ryegrass is more responsive to fertilizer nitrogen, and Kentucky bluegrass has a greater nitrogen requirement than ryegrass for a given rate of shoot growth (Adams et al., 1974). Leaf spot and brown patch diseases are generally associated with excessive nitrogen and rapid shoot growth (Bloom and Couch, 1960; Cheesman and Roberts, 1964; Beard, 1973). Consequently, nitrogen level is relevant in explaining GE interaction in perennial ryegrass.
Although there are unexplained variation unaccounted for in the interaction, there exist clear interpretations in biologically meaningful terms as to why Kentucky bluegrass and perennial ryegrass genotypes interact with environments. Specifically, there was general agreement in the interpretation of GE interactions between Kentucky bluegrass and perennial ryegrass trials indicating the interaction between genotype and environment was associated with disease reaction and cultural intensity factors, respectively. Alternative models to AMMI for studying and interpreting interaction include partial least squares regression (Aastveit and Martens, 1986) and factorial regression (Denis, 1988). In this study, external environmental and cultivar covariables available from NTEP final reports were limited in number and inconsistent (spotty) among NTEP locations for a thorough examination of the interaction. However, comparative studies have found that AMMI, partial least squares regression, and factorial regression models are all useful and may identify similar cultivar and environmental variables in explaining the interaction (Vargas et al., 1999).
In summary, interaction patterns revealed by AMMI plots indicate that Kentucky bluegrass and perennial ryegrass genotypes are narrowly adapted. No genotype has superior performance in all environments (broad adaptability). In Kentucky bluegrass, compact morphology and superior resistance to leaf spot are important adaptations when targeting genotypes to closely mown sites (2.5 to 3.75 cm). In perennial ryegrass, superior resistance to leaf spot and brown patch disease are important when targeting genotypes to high nitrogen sites (245 to 294 hg ha-1).
Some NTEP locations are highly predictable in year-to-year interaction with genotypes (making cultivar recommendations more predictable and reliable), whereas other locations are less predictable. However, NTEP locations were consistent in their interaction patterns across species. Environments interacted with genotypes according to a cultural intensity gradient (mowing height and fertilizer nitrogen). This suggests the potential for subdividing bluegrass and ryegrass growing regions into homogenous subregions having similar interaction patterns and cultivar recommendations (simplifying selection, see Ebdon and Gauch, 2002) that can be altered by cultural practices.
To validate the results reported here and to better understand GE interaction to achieve specific agricultural goals (Gauch, 1992; Gauch and Zobel, 1996 and 1997), future research targeting AMMI analysis of NTEP data for warm-season grasses and other major cool-season grass species is suggested. Furthermore, if the goal is to understand interaction in turfgrass systems of a complex nature which conform to a two-way classification (or three-way structure that can be reorganized to conform to a two-way classification), then priority should be given to AMMI analysis over ordinary ANOVA.
|
ACKNOWLEDGMENTS |
|---|
Received for publication April 2, 2001.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  H. G. Gauch Jr., H.-P. Piepho, and P. Annicchiarico Statistical Analysis of Yield Trials by AMMI and GGE: Further Considerations Crop Sci., May 1, 2008; 48(3): 866 - 889. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Worku, M. Banziger, G. S. a. Erley, D. Friesen, A. O. Diallo, and W. J. Horst Nitrogen Uptake and Utilization in Contrasting Nitrogen Efficient Tropical Maize Hybrids Crop Sci., March 1, 2007; 47(2): 519 - 528. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. Yan, M. S. Kang, B. Ma, S. Woods, and P. L. Cornelius GGE Biplot vs. AMMI Analysis of Genotype-by-Environment Data Crop Sci., March 1, 2007; 47(2): 643 - 653. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 X.-M. Fan, M. S. Kang, H. Chen, Y. Zhang, J. Tan, and C. Xu Yield Stability of Maize Hybrids Evaluated in Multi-Environment Trials in Yunnan, China Agron. J., January 1, 2007; 99(1): 220 - 228. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. G. Gauch Jr. Statistical Analysis of Yield Trials by AMMI and GGE Crop Sci., May 18, 2006; 46(4): 1488 - 1500. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Annicchiarico, L. Russi, E. Piano, and F. Veronesi Cultivar Adaptation across Italian Locations in Four Turfgrass Species Crop Sci., January 24, 2006; 46(1): 264 - 272. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. O. PB. Samonte, L. T. Wilson, A. M. McClung, and J. C. Medley Targeting Cultivars onto Rice Growing Environments Using AMMI and SREG GGE Biplot Analyses Crop Sci., October 27, 2005; 45(6): 2414 - 2424. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. S. Ebdon and H. G. Gauch Jr. Additive Main Effect and Multiplicative Interaction Analysis of National Turfgrass Performance Trials: II. Cultivar Recommendations Crop Sci., March 1, 2002; 42(2): 497 - 506. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | 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 | |||
IAS-1 = interaction scores for the first interaction axis, IAS-1 scores close to zero indicate little or no interaction (IAS-1 scale in units equivalent to the square root of turfgrass quality). Displacement along the vertical axis indicates interaction differences between environment. Displacement along the horizontal axis indicates differences in environment main effects. Dash lines show year-to-year variation within individual locations.
