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


     

Published online 24 January 2006
Published in Crop Sci 46:264-272 (2006)
© 2006 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via ISI Web of Science (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Annicchiarico, P.
Right arrow Articles by Veronesi, F.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Annicchiarico, P.
Right arrow Articles by Veronesi, F.
Agricola
Right arrow Articles by Annicchiarico, P.
Right arrow Articles by Veronesi, F.
Related Collections
Right arrow Turfgrass
Right arrow Plant and Environment Interactions
Right arrow Plant Genetic Resources

TURFGRASS SCIENCE

Cultivar Adaptation across Italian Locations in Four Turfgrass Species

Paolo Annicchiaricoa,*, Luigi Russib, Efisio Pianoa and Fabio Veronesib

a C.R.A.–Istituto Sperimentale per le Colture Foraggere, 29 viale Piacenza, 26900 Lodi, Italy
b Dipartimento di Biologia Vegetale e Biotecnologie Agroambientali e Zootecniche, Univ. degli Studi di Perugia, 74 Borgo XX Giugno, 06100 Perugia, Italy

* Corresponding author (bred{at}iscf.it)


   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Crossover genotype x location (GL) interaction may be exploited by site-specific breeding and cultivar recommendations for distinct subregions. Turf quality and density of 110 elite cultivars belonging to four species [perennial ryegrass (Lolium perenne L.), Kentucky bluegrass (Poa pratensis L.), tall fescue (Festuca arundinacea Schreb.), and red fescue (Festuca rubra L. subsp. rubra, litoralis, and commutata)] were assessed at five locations scattered across the Italian peninsula and its main islands. The 110 cultivars were evaluated for (i) assessing the extent of GL interaction and its implications on cultivar selection and recommendation; (ii) comparing additive main effects and multiplicative interaction (AMMI) vs. joint regression modeling of GL effects; (iii) verifying the consistency between species of site similarity for GL effects; and (iv) defining test locations for the Italian national evaluation program and for local breeding programs. The trials were arranged in a Group Block Design with three replicates, grouping cultivars within species. Genotype x location interaction occurred for all traits (P < 0.05) except turf density in tall fescue and Kentucky bluegrass. The selected, one-dimensional AMMI model always outperformed the joint regression model for predictive accuracy as estimated by the mean square value of the relevant GL interaction parameter. Wide crossover GL interaction was observed for turf quality among top-ranking cultivars of red fescue and Kentucky bluegrass, justifying site-specific recommendations. Variation of top-ranking cultivars across sites also occurred for turf density in red fescue and perennial ryegrass. Site similarity for GL effects was largely inconsistent across species. However, specific adaptation either to cooler, wetter sites or warmer, drier sites emerged for cultivars of red fescue, perennial ryegrass, and tall fescue. Site-specific breeding may be valuable especially for red fescue, exploiting the trend toward specific adaptation of its subspecies.

Abbreviations: AMMI, additive main effects and multiplicative interaction • GL, genotype x location • PC, principal component • sg2, genotypic variance across environments • sgl2, genotype x location interaction variance component • sL(rg)2, lack-of-genetic-correlation variance component


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
TURFGRASS SEED TRADE is a consolidated sector of modern agriculture because of the importance of turfgrass germplasm to establish recreational and sport field areas on the one hand and grassy ground covers for orchards, vineyards, roadsides, and ski runs on the other (Beard and Green, 1994). Recommendation of turfgrass cultivars in the USA is based on the visual assessment of turf quality as a part of a national evaluation program (Ebdon and Gauch, 2002a). Turf quality ratings take into account the aesthetic and functional aspects of a turfgrass relative to uniformity, density, leaf texture, growth habit, smoothness, and color (Beard, 1973). Although subjective, the visual turf quality rating remains a fundamental characteristic used by turf researchers because it is much less time-consuming than alternative quantitative options based on individual traits and can, therefore, be easily assessed several times per year.

Crossover-type GL interactions imply a change of top-ranking genotypes (i.e., cultivars) across test sites. This gives researchers the opportunity to exploit GL effects by growing cultivars specifically adapted to a certain location (Gauch and Zobel, 1997; Annicchiarico, 2002a). Location sets with the same top-ranking cultivars may form a subregion for recommendation. Conversely, the same recommendation may apply to the whole target region when the top-ranking germplasm is the same across locations. Recommendations based on modeled cultivar responses to locations, for example, by joint linear regression (Finlay and Wilkinson, 1963) or AMMI (Gauch, 1992) models, allows for reducing the noise (i.e., the random error) that affects the observed genotype by location cell means. This improves the prediction of future responses of genotypes at each site, and usually simplifies the recommendation through a reduction in the number of subregions (Gauch, 1992, p. 134; Gauch and Zobel, 1997). Crossover GL interaction among top-ranking cultivars is frequent in grain or forage crops even within relatively small regions, such as northern Syria (Ceccarelli, 1989) and northern Italy (Annicchiarico and Piano, 2005). This has also recently been reported for turfgrass cultivars of Kentucky bluegrass and perennial ryegrass across sites of the USA in relation to differences in cultural practices (mowing height and N fertilization) and disease intensity (Ebdon and Gauch, 2002a), supporting the need for site-specific recommendations (Ebdon and Gauch, 2002b). It is unknown, however, whether such recommendations may also apply to turfgrass cultivars evaluated in smaller geographical regions such as individual European countries or in the presence of homogeneity of cultural practices among test sites.

All GL interaction effects that arise from lack of genetic correlation among sites (including those for poorly performing material) are relevant for defining crucial test sites for national testing programs (Lin and Butler, 1988) or selection locations and selection strategies for breeding purposes (Annicchiarico, 2002b). This is particularly true if results from a data set are extrapolated to produce information on GL effects that may be met in a given target region (Cooper et al., 1996).

This study reports cultivar responses for turf quality and turf density across Italian sites for four turfgrass species {perennial ryegrass, Kentucky bluegrass, tall fescue, and red fescue complex [here including the following subspecies according to the classification of Ruemmele et al., 2003: rubra Gaudin, strong creeping red fescue; litoralis (G.F.W. Meyer) Auquier, slender creeping red fescue, earlier classified as trichophylla (Gaud.) Richter; and commutata (Thuill.) Nyman, chewings fescue]}, with the aims of (i) assessing the extent of crossover GL interaction and its implications on cultivar recommendation and selection strategies; (ii) comparing AMMI vs. joint regression models for ability to describe GL effects; (iii) verifying the consistency between species of site similarity for GL effects; and (iv) defining crucial test locations for the Italian national turfgrass evaluation program and for local breeding programs.


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Plant Material and Environments
The list of tested cultivars, reported in Annicchiarico et al. (2000), included the most recent and best-performing material available on the European seed market based on the recommendations provided by 17 contributing seed companies. The evaluation included 40 cultivars of perennial ryegrass, 20 of tall fescue, 20 of Kentucky bluegrass, and 10 each of the subsp. rubra (strong creeping), litoralis (slender creeping), and commutata (chewings) of red fescue.

The evaluation was performed at five Italian locations which are described in Table 1. Test sites were scattered across the Italian peninsula and its main islands (Fig. 1 ), to adequately sample the environmental variation across the region. The trials were sown in spring 1999 at Perugia and autumn 1998 at the remaining sites. Plots were 2 by 3 m in size and were arranged in a group block design with three replicates (Gomez and Gomez, 1984, p. 75). This design is similar to a randomized complete block design, but cultivars sharing a selected characteristic, the species in this case, are grouped together within each block to facilitate their comparison. Data were analyzed separately for each species using a randomized complete block. Pooling different species into the same blocks assured that species were grown in the same environment at each site, thereby excluding within-site environmental variation as a contributing reason to possible inconsistencies of site similarity for GL effects across species.


View this table:
[in this window]
[in a new window]
  		Table 1. Name, acronym, latitude, longitude, altitude, and soil texture class of Italian turfgrass test locations.

 

Figure 1
View larger version (16K):
[in this window]
[in a new window]
  		Fig. 1. Geographic position of test locations.

 
Sowing rates were 250 kg ha–1 for Kentucky bluegrass and 350 kg ha–1 for other species. Nitrogen, P, and K fertilizers were applied at the elemental rates of 50, 150, and 50 kg ha–1, respectively, before seeding. Annual rates of N applications were 150 kg ha–1, distributed as 30 kg ha–1 in February–March, 60 kg ha–1 in April, 30 kg ha–1 in May–June, and 30 kg ha–1 in September–October. Turf was mowed whenever it reached 5.5 cm. Mowing height was set at 3.5 cm and clippings were always removed. All sites were equipped with permanent irrigation systems using buried pipes and sprinklers. The irrigation amount took into account the rainfall and the evapotranspiration of the seasons, with the aim of maintaining regular turf growth throughout the growing season. Chemical weed control was applied on an as-need basis using Mecoprop [2-(4-chloro-2-methylphenoxy)propanoic acid], Dicamba (3,6-dichloro-2-methoxybenzoic acid), or fenoxaprop-P-ethyl ((2R)-2-{4-[(6-chloro-2-benzoxazolyl)oxy]phenoxy}propanoic acid). Puccinia striiformis and P. graminis in Kentucky bluegrass were the major causes of diseases. They were controlled by Propiconazole [1-[[2-(2,4-dichlorophenyl)-4-propyl-1,3-dioxolan-2-yl]methyl]-1H-1,2,4-triazole] applied only to the damaged plots when clear symptoms were present on at least 50% of the area. This fungicide treatment was allowed for Kentucky bluegrass in the evaluation because of its ordinary adoption for the species in the region.

Turf quality was assessed using a visual rating system adopted by the U.S. National Turf Evaluation Program, ranging from 1 (= very poor quality) to 9 (= outstanding, ideal turf). During the first year the evaluators met on several occasions at different sites to standardize the evaluation method, thereby minimizing the error attributable to different observers across locations in the following years. Observations were made on a monthly basis during the second and third years (from January 2000 to December 2001).

Turf density was measured in the autumn of 2001 as the number of tillers within three randomly placed 58-mm diam. turf cores per plot. This variable, recorded at the end of the evaluation period, also provided an indication of stand persistence for each genotype (besides concurring to turf quality).

Statistical Analysis
The average plot value across observations was used for statistical analysis of the response variables. The size of variance components relative to genotype (cultivar), GL interaction, and the two determinants of the latter variance, i.e., the lack of genetic correlation and heterogeneity of genotypic variance among sites, and the pooled genetic correlation among locations, were estimated as described by Cooper et al. (1996) following a combined ANOVA and ANOVAs for single experiments.

A combined ANOVA limited to red fescue data and including also subspecies as a fixed factor assessed the variation among and within red fescue subspecies. The mean squares for genotype within subspecies (random factor) and its interaction with location were used as error terms for testing subspecies main effects and subspecies x location interaction, respectively.

Cultivar mean values across sites for traits showing no GL interaction were compared by the Newman-Keuls test. Modeling of significant GL interaction effects was performed by joint regression analysis (Finlay and Wilkinson, 1963) and AMMI analysis (Gauch, 1992). The AMMI modeling of turf quality rating responses was previously used by Ebdon and Gauch (2002a; 2002b). Heterogeneity of genotype regressions in the joint regression analysis was tested on deviations from regressions. Genotype x location interaction principal component (PC) axes in the AMMI analysis were tested by the FR test as recommended by Piepho (1995). Since some degree of subjectivity in the assessment of turf quality at different sites could not be excluded a priori and may have contributed to GL interaction, Type 1 error for these tests was set to the more conservative rate of P < 0.01 for this variable while keeping P < 0.05 for turf density. Model comparison was based on the mean square value of heterogeneity of genotype regressions and GL interaction PC 1. For one-dimensional models, this value is related to the predictive ability because it takes into account the accuracy (i.e., the amount of GL interaction sum of squares) as well as the parsimony (i.e., the amount of degrees of freedom) of the model (Brancourt-Hulmel et al., 1997).

AMMI-modeled responses of cultivars were graphically expressed as nominal values of the response variable as a function of the site score on the first GL interaction PC axis (Gauch and Zobel, 1997; Ebdon and Gauch 2002b). The adoption of nominal values, which sum up the estimated genotype mean value and the product of the genotype by the site scaled scores on PC 1 (excluding the site main effect, irrelevant for genotype ranking), allows for linearizing the adaptive responses. For sake of clarity, the graphs included a subset of genotypes that were among the four top-ranking ones in at least one site.

Correlations of site PC scores with climatic and soil variables of locations were assessed to reveal environmental factors related to the occurrence of GL interaction (Ebdon and Gauch, 2002a). The climatic and soil variables included in the correlations were as follows: soil content of clay, sand and silt; soil pH; mean annual rainfall; average potential evapotranspiration and total rainfall in summer (July to August), and the difference between these variables; mean rainfall in the driest month (July); and mean daily temperatures of the coldest (January) and the warmest (July) month. Climatic data were averaged across three test years. Correlations were also assessed between turf quality and turf density of cultivars for mean value across sites and adaptive response as expressed by PC 1 score values.

National variety evaluation programs may define a small set of test sites by classifying a larger set of locations on the basis of their similarity for GL effects and selecting one site from each group (Lin and Butler, 1988). Site classification by cluster analysis may follow the AMMI analysis, using as variables the site PC scores of significant PC axes (i.e., the characteristics that express the similarity of sites for GL effects) (Annicchiarico, 2002b). In this case, the aim of the analysis was producing information of general interest for a turfgrass variety evaluation program in the region. Therefore, the indications on site similarity provided by the four species were pooled together, by allowing all significant PC axes of locations relative to turf quality or density of any species to simultaneously enter the cluster analysis as variables. Ward's clustering method, and a squared Euclidean distance as the dissimilarity measure, were used for the analysis as recommended by Cooper et al. (1996).

The screening ability of each site for the target region (as represented by the sample of locations), which is of possible interest for breeding programs that aim to identify one major selection location, was assessed by the phenotypic correlation between genotype values at the site and average genotype values across all sites (Cooper et al., 1996). Correlations were computed for single species and then averaged across species to express the mean screening ability of sites, considered of special interest to European programs due to their commitment to breeding of several turf species. Averages of correlation coefficients were computed on inverse tanh (Z)-transformed values, and were preceded by the assessment of their homogeneity by the {chi}2 test described by Steel and Torrie (1960, p. 191).

The software IRRISTAT, developed and made freely available by the International Rice Research Institute (IRRI) of Manila, Philippines, was used for AMMI and joint regression modeling. The use of IRRISTAT for analysis of adaptation has been discussed and exemplified by Annicchiarico (2002a). An AMMI analysis provides a solution that is unique up to simultaneous sign change of the genotype and location PC scores (Gauch, 1992). To better verify the consistency of site ordination across species, positive PC scores were preferably assigned to locations of north and central Italy, that is, Lodi and Perugia (inverting the outputted sign of location and genotype PC 1 scores simultaneously when convenient). The Statistical Analysis System (SAS Institute, 1999) software was used to perform all other analyses.


   RESULTS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Extent of Crossover Genotype x Location Interaction
The GL interaction variance component (sgl2) for turf quality was always significant (P < 0.01). However, in perennial ryegrass the size of the GL interaction variance was modest relative to the genotypic variance across environments (sg2), while in tall fescue most of GL effects were accounted for by the heterogeneity of genotypic variance among sites [equal to the difference between sgl2 and the lack-of-genetic-correlation variance component, s\rm L(rg)2] (Table 2). As a result, the s\rm L(rg)2 exceeded the heterogeneity of genotypic variance only in red fescue and Kentucky bluegrass (Table 2). In these species, the low value of pooled genetic correlation among sites (rg ≤ 0.40) confirmed that crossover GL interaction for turf quality was extensive (Table 2).


View this table:
[in this window]
[in a new window]
  		Table 2. Estimates of variance components for genotype (sg2), genotype x location interaction (sgl2), and lack-of-genetic-correlation among locations [sL(rg)2], and pooled genetic correlation among locations (rg), for turf quality and turf density of four species.

 
Significant (P < 0.05) GL interaction variance for turf density was found only in red fescue and perennial ryegrass. However, the modest size of the sL(rg)2 relative to the sg2 and the relatively high pooled genetic correlation among locations (rg > 0.80) indicated that crossover GL effects were also fairly small in these species (Table 2).

Red fescue subspecies differed for average turf density across sites (P < 0.01) but did not differ for average turf quality (P > 0.05), while genotypes within subspecies significantly differed for mean value across sites for both variables (P < 0.01). Subspecies x location interaction was always significant (P < 0.01), whereas genotype within subspecies x location interaction was significant for turf quality (P < 0.01) but not for turf density (P > 0.05). On the whole, these results suggested that genotype main effects and GL interaction effects for turf density were mainly accounted for by the subspecies factor, whereas also the genotype within subspecies factor was important for turf quality responses.

Model Comparison, Adaptive Responses, and Top-Ranking Genotypes
The AMMI model with one GL interaction PC was selected for all of the six variables that showed significant GL interaction, based on test results for PC axes. This one-dimensional model, accounting from 48 to 79% of the GL interaction variation (Table 3), always outperformed the joint regression model in terms of accuracy as well as in terms of higher mean square of PC 1 relative to the heterogeneity of genotype regressions (Table 3). The poorer value of joint regression was confirmed, for five variables out of six, by significant (P < 0.05) deviations from regression (Table 3).


View this table:
[in this window]
[in a new window]
  		Table 3. Mean squares for genotype x location (GL) interaction, heterogeneity of genotype regressions, deviations from regressions and GL interaction PC 1, GL interaction variation accounted for by each model (R2), mean square ratio of PC 1 to genotype regressions, and number of genotypes that are top-ranking in at least one site according to observed and AMMI (additive main effects and multiplicative interaction) modelled data, for turf quality and turf density of four species.{dagger}

 
AMMI-modeled nominal turf quality ratings as a function of the GL interaction PC 1 score of sites are reported for better-performing cultivars of each species in Fig. 2 . Responses for tall fescue material confirmed the importance of heterogeneity of genotypic variance among sites, which increased with increasing PC 1 scores of the site. Crossover interaction was negligible, and one cultivar (Farandole, alias Matador) was consistently top-ranking across sites. Responses for the subset of perennial ryegrass cultivars confirmed the limited extent of GL interaction in this species. Chaparral was top-ranking across sites, but a few other cultivars (Roadrunner, Charger II, and Brightstar) responded very similarly. Large crossover interaction took place among well-performing cultivars of red fescue and Kentucky bluegrass (Fig. 2). In the former species, some cultivars of subsp. commutata (Samt, Olivia, Center and Waldorf) were specifically adapted to Lodi, while other cultivars belonging to all three subspecies were specifically adapted to Sassari. In Kentucky bluegrass, two cultivars (Cocktail and Princeton) were top-ranking at different sites, and other top-performing genotypes showed contrasting adaptation patterns.


Figure 2
View larger version (28K):
[in this window]
[in a new window]
  		Fig. 2. Nominal predicted turf quality rating (1 to 9, 9 = highest) of top-performing genotypes as a function of the site score on the first GL interaction PC axis (showed by a black triangle), for four turfgrass species.

 
Adaptive responses for turf density are reported in Fig. 3 for red fescue and perennial ryegrass, which showed significant GL effects. Crossover interaction among top-ranking cultivars was observed in both species. Responses at Foggia tended to contrast with those at other sites.


Figure 3
View larger version (14K):
[in this window]
[in a new window]
  		Fig. 3. Nominal predicted turf density of top-performing genotypes as a function of the site score on the first GL interaction PC axis (showed by a black triangle), for two turfgrass species.

 
Responses of red fescue subspecies are reported in Fig. 4 as average nominal genotype values of turf quality and density. For turf quality, subsp. commutata tended toward specific adaptation to Lodi, whereas the two subspecies with a creeping habit (rubra and litoralis) were relatively better performing in Sassari. These indications are in good agreement with those relative to the top-ranking set of cultivars reported in Fig. 2. For turf density (where the subspecies factor accounted for the entire GL interaction), subsp. litoralis, commutata, and rubra ranked in this order across locations, but their differences decreased passing from Foggia to the other locations (Fig. 4). On the whole, the adaptive responses of subspecies were little related to their creeping ability (high in subsp. rubra; modest in subsp. litoralis; null in subsp. commutata), since the subspecies with intermediate creeping (subsp. litoralis) showed a marked specific-adaptation response for turf quality and a consistently top-ranking response for turf density (rather than intermediate responses for these traits) (Fig. 4).


Figure 4
View larger version (10K):
[in this window]
[in a new window]
  		Fig. 4. Mean value across genotypes of nominal predicted turf quality rating (1 to 9, 9 = highest) and turf density for three subspecies of red fescue as a function of the site score on the first GL interaction PC axis (showed by a black triangle).

 
Correlations (P < 0.10) between environmental variables and site PC 1 score for turf quality suggested that PC 1 was a negative indicator of mean temperature in the coldest month for red fescue (r = –0.91) and perennial ryegrass (r = –0.85), and a positive indicator of rainfall in the driest month for perennial ryegrass (r = 0.96) and of annual rainfall for tall fescue (r = 0.84). For all of these species, PC 1 tended to represent a contrast between cooler, wetter sites (Lodi and Perugia) on one hand, and warmer, drier sites (Agrigento, Foggia, Sassari) on the other (Fig. 2). The range of variation for these climatic variables was –1.3 to 6.0°C for mean temperature in the coldest month, 399 to 802 for annual rainfall, and 4 to 43 mm for rainfall in the driest month. No relationship between environmental variables and site PC 1 score emerged for turf quality in Kentucky bluegrass or turf density in red fescue and perennial ryegrass.

AMMI modeling allowed for reducing the number of cultivars that were top-ranking in at least one location in comparison with observed data (Table 3), thereby simplifying cultivar recommendations. The reduction was dramatic for turf quality of perennial ryegrass, where four sites out of five showed different top-ranking cultivars based on observed data, whereas Chaparral was consistently top-ranking based on modeled data.

In tall fescue, Farandole showed the highest mean value of turf density across sites (P < 0.05) concurrently with its top-ranking turf quality across locations. In Kentucky bluegrass, two cultivars (Cocktail and Alpine) outperformed the others for mean value of turf density (P < 0.05). A positive relationship between turf quality and turf density of cultivars emerged in tall fescue (r = 0.82, P < 0.001), red fescue (r = 0.56, P < 0.01), and perennial ryegrass (r = 0.75, P < 0.001) for values averaged across sites. However, relatively better or worse response to locations for turf quality was independent of the response for turf density, as indicated by the lack of correlation (P > 0.05) between PC 1 score for turf quality and PC 1 for turf density of cultivars in any species.

Site Similarity for GL Effects and Definition of Test Locations
There was a poor consistency between species for ordination of test locations on GL interaction PC 1 for turf quality, as indicated by the lack of correlation (P > 0.10) for site PC 1 score values between any pair of species. This finding highlighted the difficulty in identifying a small set of locations that contrast for adaptive response of cultivars to be used for cultivar testing of the main cool season turfgrass species in the region. Conversely, the site ordination on PC 1 for turf density showed a fairly high consistency between tall fescue and perennial ryegrass (r = 0.89, P < 0.05).

The results of cluster analysis for test locations based on site similarity for significant GL interaction effects of the four species are summarized in Fig. 5 . Foggia revealed a markedly distinct response relative to any other site. Lodi and Perugia were the most similar locations for adaptive response of turfgrass cultivars, although they were fairly dissimilar in the analysis.


Figure 5
View larger version (13K):
[in this window]
[in a new window]
  		Fig. 5. Cluster analysis of test locations performed on significant GL interaction PC scores of sites for turf quality rating and turf density of four turfgrass species (see Table 1 for acronym of location).

 
Phenotypic correlations between genotype response at the site and genotype mean response across locations are reported in Table 4 as average values across species. The marked lack of homogeneity of the correlation coefficients for turf quality of Perugia and Sassari (P < 0.01) indicated that the ability of these sites to reproduce the mean turf quality of cultivars across locations was strongly affected by the species. Either Agrigento or Lodi, tending toward higher and/or less inconsistent correlation coefficients than other sites for both variables (Table 4), could be preferred as selection location for turfgrass breeding for the region. Even these sites, however, were characterized by extreme PC 1 score values in some species (Fig. 2 and 3) and failed, therefore, to faithfully reproduce in all cases the mean response of cultivars.


View this table:
[in this window]
[in a new window]
  		Table 4. Average phenotypic correlation coefficients between genotype value at the site and genotype mean value across sites, and {chi}2 test for homogeneity of the averaged coefficients for test locations of turf quality and turf density.{dagger}

 

   DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
Substantial GL interaction of crossover type may occur even across a much smaller region and more homogeneous cultural practices than those reported for occurrence of crossover interaction by Ebdon and Gauch (2002a; 2002b). These authors also adopted a longer evaluation period (4 or 5 yr) than the current one (3 yr), although our averaging the turf quality scores across the second and third year rather than across the entire period (as done by Ebdon and Gauch) may partly counterbalance the shorter turf cycle by assigning a greater weight to later observations. The fairly wide climatic variation that characterizes the Italian region probably contributed to the relatively large extent of GL effects, although few relationships emerged between site similarity for these effects and the available climatic variables. The specific adaptation of cultivars either to cooler, wetter sites or to warmer, drier sites that emerged for turf quality in red fescue, perennial ryegrass, and tall fescue is similar to that reported for grain yield of winter cereals in the region (Annicchiarico, 1997). The limited consistency for site PC 1 score between turfgrass species suggests, however, that different environmental factors or sets of factors, partly undetected, contributed to GL interaction in each species. The contribution to GL effects of biotic stresses could not be ruled out, since their occurrence was limited only with regard to rusts in Kentucky bluegrass.

The superiority of AMMI over joint regression for modeling adaptive responses has already been reported for cereal and forage crops in the region (Annicchiarico, 1997; Annicchiarico and Piano, 2005) as well as for several grain crops worldwide (Brancourt-Hulmel et al., 1997). In two turfgrass species out of four, i.e., red fescue and Kentucky bluegrass, the occurrence of different top-ranking genotypes for AMMI-modeled turf quality responses leads to site-specific cultivar recommendations for each area represented by one test site. Indications on top-ranking material for turf density may contribute to, or even substitute for, those for turf quality when the persistence of the turf rather than its aesthetic characteristics is of primary importance. In this context, site-specific recommendations may be defined for red fescue and perennial ryegrass. The simplification of recommendation domains allowed for by AMMI modeling relative to observed data was already reported by Ebdon (2002) for Kentucky bluegrass and perennial ryegrass in the USA. The higher predictive ability and greater advantage for recommendations that are theoretically expected for modeled data relative to observed data (Gauch, 1992) would require an independent set of trials to be reliably verified. One such assessment recently performed on durum wheat [Triticum turgidum L. subsp. durum (Desf.) Husn.] indicates a grain yield advantage around 4% by adopting the top-ranking cultivars based on AMMI-modeled data (Annicchiarico et al., 2005).

The poor consistency of site similarity for GL effects points to the difficulty to markedly reduce the number of test sites in the Italian turfgrass evaluation program. Should such reduction be unavoidable due to budget restrictions, three sites, including Foggia (the most dissimilar), Lodi or Perugia, and Agrigento or Sassari, would provide the optimal choice. These sites may also be chosen for assessing the Value for Cultivation and Use that is required for registering new cultivars in the National List of Varieties. Budget restriction may also lead to confirm the adopted 3-yr evaluation period, which, however, allowed for the detection of better-performing material across the region.

The size of GL interaction effects, while being sufficient for exploring site-specific cultivar recommendations, may justify the breeding of distinct cultivars for different areas of the region only for red fescue. For this species, specific breeding is also favored by the trend to specific adaptation shown by the subspecies. In particular, breeding for northern Italy may concentrate on germplasm of subsp. commutata, whereas different subspecies may be of potential interest for other areas. The rather poor relationship between adaptive response and creeping ability of subspecies suggests that also other characteristics of subspecies may be important for the adaptive response. The information on the average screening ability of locations is especially valuable for the other three species, to minimize the chance to select narrowly adapted material.


   ACKNOWLEDGMENTS
 
We gratefully acknowledge the contribution of S. Bullitta, P. Martiniello, L. Stringi, and C. Tomasoni, who were responsible for the individual trials of Sassari, Foggia, Agrigento, and Lodi, respectively. We also thank M. Casler for revising the manuscript.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
The work was carried out within the Project "Inerbimenti e Tappeti erbosi per la Valorizzazione Agricola, Ricreativa e Sportiva del Territorio" funded by the Ministry of Agricultural and Forestry Policies of Italy.

Received for publication January 14, 2005.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 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 (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Annicchiarico, P.
Right arrow Articles by Veronesi, F.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Annicchiarico, P.
Right arrow Articles by Veronesi, F.
Agricola
Right arrow Articles by Annicchiarico, P.
Right arrow Articles by Veronesi, F.
Related Collections
Right arrow Turfgrass
Right arrow Plant and Environment Interactions
Right arrow Plant Genetic Resources

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