HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Annicchiarico, P. Articles by Doust, M. A. Search for Related Content PubMed Articles by Annicchiarico, P. Articles by Doust, M. A. Agricola Articles by Annicchiarico, P. Articles by Doust, M. A.

PLANT GENETIC RESOURCES

Repeatability of Large-Scale Germplasm Evaluation Results in Durum Wheat

^a Istituto Sperimentale per le Colture Foraggere, viale Piacenza 29, 26900 Lodi, Italy
^b Istituto Sperimentale per la Cerealicoltura, via Varese 43, 95123 Catania, Italy
^c Istituto di Agronomia Generale e Coltivazioni Erbacee, via Valdisavoia 5, 95123 Catania, Italy

iscfbred{at}telware.it

	ABSTRACT

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussion REFERENCES

 
Large-scale trait evaluation may enhance the utilization of germplasm collections by breeders. However, there is concern that this information may not be repeatable outside the area of testing because of large genotype x location interaction. Our objective was to assess the repeatability of large-scale evaluation results for durum wheat [Triticum turgidum (L.) Thell. ssp. turgidum conv. durum (Desf.) MacKey] across two areas of the Mediterranean region. A subset of 600 entries from a world collection was evaluated for six agronomic traits at one site in northern Syria and one in Sicily. Pearson's and Spearman's rank correlations of entry values between locations indicated that the repeatability as measured by both coefficients was high for heading time (r 0.73), moderately high for kernel weight and plant height (0.48 r 0.57), relatively low for grain yield and kernels per spike (0.18 r 0.30), and non-significant for early vigor score. Comparison between phenotypic and genetic correlations suggested that not only genotype x location interactions but also experimental errors contributed substantially to departures from complete repeatability. When each location alternated as the selection site and the other as the test site, at least 50% of the advantage shown by the top 10% entries over the remaining accessions at the selection site could be maintained at the test location for early or late heading, high kernel weight, and short or tall stature, for which 30 to 57% of selected entries were common to both sites. At least 20% of the selection advantage was maintained for high kernels per spike and grain yield. The results support the potential usefulness of large-scale germplasm evaluation trials for crop improvement.

	INTRODUCTION

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussion REFERENCES

 
LARGE EX SITU GERMPLASM COLLECTIONS have been established for the major seed crops at world-based or regional gene banks (Plucknett et al., 1987; Adham and Van Sloten, 1990). A thorough characterization of these genetic resources is an on-going activity that is hindered by high cost. Large-scale evaluations of morphophysiological traits desirable for crop improvement have also been undertaken in some instances with the aim of providing breeders with information of practical interest. The lack of such information has frequently been advocated as one of the main constraints to a more extensive and more effective utilization of germplasm collections in breeding programs (Frankel, 1970; Duvick, 1984; Williams, 1991). However, there is concern that large genotype x location interactions may substantially limit the repeatability of results from agronomic trait evaluations which, therefore, offer little useful information to breeders outside the area of testing (Marshall and Brown, 1981).

With regard to durum wheat, information on variation in large collections has been reported by Qualset and Puri (1975) for phenology; Spagnoletti-Zeuli and Qualset (1987, 1990) for quantitative spike traits and flag leaf size; Porceddu (1976) and Yang et al. (1991) for various vegetative and reproductive characters; Pecetti and Annicchiarico (1991) for grain yield potential under moderately favorable conditions and tolerance to drought and diseases; and Pecetti et al. (1992) for various vegetative and reproductive traits. The information on each accession evaluated is in general available to breeders. The only documented assessment of the repeatability of evaluation results is that reported by Spagnoletti-Zeuli and Qualset (1993) for six spike characters of a relatively small random set of accessions evaluated at a U.S. site and re-evaluated in the same location and in southern Italy. Correlations between seasons of evaluation at the same location generally were of the same magnitude as those between locations. The average repeatability between locations ranged from moderately low (r = 0.34) for spikelet number to high (r = 0.78) for kernel weight.

The objective of this study was to assess the repeatability of large-scale germplasm evaluation results on agronomic traits across different areas of the Mediterranean basin as a means of determining the reliability and potential usefulness of these results to breeders operating in this region.

	Materials and methods

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussion REFERENCES

 
Germplasm Evaluation
A world collection of about 8000 durum wheat accessions from 60 countries was evaluated at the International Center for Agricultural Research in the Dry Areas (ICARDA) at Tel Hadya, northern Syria (Lat. 36°01' N, Long. 36°56' E, 284 m elevation). The accessions were divided for evaluating in 1 of 3 yr (1985-1986 through 1987-1988) and, within each year, subdivided into a number of individual, non-replicated augmented designs (Federer, 1956). The soil type was a Chromoxerertic-Rhodoxeralf, the texture class being clay according to the U.S. soil survey classification. The plot size was 3 m² (4 rows, 2.5 m long, 30 cm apart). Sowing always occurred at the end of November at a rate of ca. 150 kg ha^-1. Fertilization rates were 40 kg ha^-1 of N and 60 kg ha^-1 of P₂O₅. The recorded data were adjusted within each experiment according to the adopted design, and across experiments in the same or different year according to the mean performance of three locally adapted control cultivars used in all experiments, viz. Haurani, Sham 1, and Om Rabi 9 (Pecetti et al., 1992).

A subset of 600 accessions was further tested at Libertinia, Sicily, Italy (Lat. 37°32' N, Long. 14°35' E, 183 m elevation). These entries were randomly chosen from those that showed higher than average yield under moderately favorable conditions and drought tolerance in northern Syria (Pecetti and Annicchiarico, 1991). These thresholds were imposed to exclude materials poorly adapted to Mediterranean environments. The 600 entries, belonging to 28 countries of origin, were divided into four augmented design experiments and evaluated one year in Sicily during 1990-1991. The soil type was a Vertic Xerofluvents with a texture classified as silty-clay-loam. The plot size was 0.85 m² (2 rows, 2.5 m long, 17 cm apart). Planting occurred on 18 December at a rate of ca. 250 kg ha^-1. The fertilization included 100 kg ha^-1 of N and 92 kg ha^-1 of P₂O₅. Differences in agronomic practices between the two locations reflected ordinary crop management of the respective areas. The recorded data at Libertinia were adjusted within and across experiments as indicated for the evaluation at Tel Hadya, but data were used from three local cultivars as controls, viz. Simeto, Arcangelo, and Capeiti.

Six characters were recorded at both locations following the same procedures: (i) early vigor, visually scored at the end of tillering (beginning of February) on a scale from 1 = minimum to 9 = maximum; (ii) days to mean heading (from 1 April); (iii) plant height, measured at physiological maturity; (iv) average number of kernels per spike for five random spikes per plot; (v) grain yield per plot; and (vi) 1000-kernel weight, recorded on a bulk of 250 harvested seed.

Statistical Analysis
The repeatability of entry values across the two evaluation sites was estimated by Pearson's correlation and Spearman's rank correlation coefficients. The latter measure was considered more reliable for those traits for which the assumption of normal distribution was not fulfilled on the basis of Shapiro and Wilk's (1965) test.

Any deviation from unity of the phenotypic correlation coefficient, i.e., the lack of complete repeatability, could be due to entry x location interactions and/or experimental errors. Large errors may lead to underestimation of correlation coefficients (Dagnelie, 1975, p. 310). The lack of repeatability that is only due to genotype x location interaction can be estimated in terms of deviation from unity of the coefficient of genetic correlation (r_g) of entry values between locations (Burdon, 1977; Falconer, 1989, p. 322). The r_g value can be estimated from Pearson's phenotypic correlation coefficient (r) and broad-sense heritability on an entry-mean basis at each of the two locations (h²_x and h²_y):

and tested for significance from zero as described by Burdon (1977). The h² values for each location can be computed from the genotypic (²_g) and experimental error (²_e) components of variance:

where n is the number of replications. In the current, unreplicated experiments n is equal to 1. These formulas show that when experimental errors are neglibible, or n is high, h² values tend to unity, and the phenotypic correlation tends to coincide with the genetic one.

In general, for non-replicated experiments, an estimate of ²_e can be obtained from the response of a set of replicated control cultivars. The ²_g for the test entries can then be estimated as the difference between the phenotypic variance of the test entry values and the error variance (Bos, 1988). Were all test entries included into one augmented design, the experimental error (i.e., the interaction between control cultivars and blocks) would provide an adequate estimate of ²_e. With the full set of test entries assigned to different augmented designs planted in 1 yr at Libertinia and across 3 yr at Tel Hadya, the interaction of control cultivars with experiments in the same or different year should also be considered. The error term from a two-way analysis of variance with control cultivars and blocks, in which blocks were inclusive of all experiments in a given location, was used to estimate ²_e. Genetic and experimental error coefficients of variation were also computed for each location by dividing _g and _e by their respective mean values.

We determined the proportion of the top 10% of selected entries that were common to both locations for eight selection criteria: early and late heading; short and tall stature; and high values of early vigor, grain yield, kernel weight, and kernel number. This proportion was compared with that expected under the hypothesis of stochastic independency for selection at the two locations. Under that hypothesis, only 10% (i.e., 6) of the 60 entries selected at one location can be expected to also be selected at the other location. The expected frequency distribution of coincident and non-coincident entries was compared with the observed one by a ² test with one df (Snedecor and Cochran, 1967).

Large-scale evaluation results could be used to discard a substantial number of undesirable entries, rather than to select a relatively small group of desirable entries. Considering a discard-rate of 50% for the same selection criteria at each location, we assessed the proportion of these entries which were common to both locations. We compared by a ² test with one df the observed frequencies of coincident and non-coincident entries with those expected under the hypothesis of stochastic independency of selection. In this case, the expected frequency of coincident entries is 50%, i.e., 150 out of 300 discarded entries.

We also determined, for the same selection criteria, to what extent the advantage shown by the top 10% entries relative to the other accessions at one selection location was maintained when the materials were tested at the other location. All comparisons between the top 10% of entries and the remaining entries (in the non-selective location) were performed by a t test, using Cochran's approximation (Snedecor and Cochran, 1967, p. 115) for unequal variances of the populations as tested by Fisher's F test.

	Results and discussion

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussion REFERENCES

 
Climatic and soil variables for the two locations (Table 1) suggest that Tel Hadya was less favorable for crop production than Libertinia because of its colder winter and hotter spring. The mean yield at Tel Hadya was 23% lower than at Libertinia (Table 2) . The fact that Libertinia had only slightly higher mean kernel weight and kernel number per spike (Table 2) suggests that the yield advantage at this location was mainly attributable to more fertile tillers per unit area. The lower plant density at establishment and late frosts were probably the main determinants of lower number of fertile tillers at Tel Hadya.

The hypothesis of a normal distribution was rejected (P < 0.001) at both locations for early vigor, days to heading, and plant height. Fisher's coefficients of skewness and kurtosis indicated asymmetry towards high values for plant height, contributed by a majority of entries that were land races and old cultivars lacking dwarfing genes. For early vigor and days to heading, the departure from normality was mainly attributed to kurtosis.

Pearson's and Spearman's rank correlations indicated that repeatability was high (r 0.73) for days to heading, moderately high (0.48 r 0.57) for kernel weight and plant height, relatively low (0.18 r 0.30) for grain yield and number of kernels per spike, and non-significant for early vigor (Table 2). The genetic correlation values of days to heading and plant height were close to unity, indicating that genotype x location interaction was minimal (Table 2). This interaction was large for early vigor, and moderate for the remaining traits, based on r_g values.

The difference between r_g and Pearson's phenotypic correlation values may reflect the contribution of experimental error to the lack of repeatability between locations. In particular, the magnitude of the difference between genetic and phenotypic correlations, compared with the deviation from unity of r_g, suggests that experimental error had greater bearing than genotype x location interactions on the lack of repeatability for grain yield and plant height (Table 2). The current, high level of genotype x environment interaction for early vigor, and moderate to low interactions for grain yield, heading date, kernels per spike, and kernel weight are in agreement with the findings relative to environments of northern Syria (Annicchiarico and Pecetti, 1998).

The Pearson's correlation coefficients reported by Spagnoletti-Zeuli and Qualset (1993) for kernel weight and number of kernels per spike are distinctly greater than those found here for the same characters. Although they adopted similar measurement procedures (e.g., number of sampled spikes per plot), their experiments included three replications in each of two environments and a non-replicated experiment in a third environment. The presence of replications is expected to increase broad-sense heritability on an entry-mean basis, thereby increasing repeatability as Pearson's phenotypic correlation towards a limit corresponding to the genetic correlation value (see previous formulas). Not surprisingly, the highest level of repeatability for all traits reported by Spagnoletti-Zeuli and Qualset (1993)(Table 1) occurred for the two environments with replicated trials. In this case the repeatability values were r = 0.66 for kernel number and r = 0.84 for kernel weight. Setting the number of replicates to three, we recomputed broad-sense heritabilities and obtained repeatabilities of r = 0.39 for kernels per spike and r = 0.66 for kernel weight. These values are distinctly higher than those computed for the actual non-replicated experiments (Table 2), though still lower than those reported by Spagnoletti-Zeuli and Qualset (1993). A reliable comparison of repeatability values must consider the number of replicates from which evaluation data originate. Large-scale evaluations are difficult to plant in a replicated design, mainly because of cost limitations, and are bound, therefore, to produce results whose repeatability can be markedly reduced by experimental errors. The use of experimental designs that allow for the greatest control of these errors can do much to alleviate this constraint.

Repeatability of wheat trait values has extensively been investigated only for grain yield. Despite the use of yield data from replicated trials, the reported repeatability of yield tends to be low, especially between agro-climatically contrasting locations, within countries such as Mexico (Braun et al., 1992) or Australia (Cooper et al., 1997). When genetic correlations were reported, their values between locations of Australia (0.21 r_g 0.67; Pederson and Rathjen, 1981) or contrasting environments of Oklahoma (r_g = 0.20; Ud-Din et al., 1992) were lower than the genetic correlation for yield estimated in this study.

The frequency of selected entries that were coincident in the two locations was always greater (P < 0.05) than expected under the assumption of stochastic independency of selections (Table 3) . The frequency was relatively high (at least 25 entries out of 60) when selecting for short or tall plant height, high kernel weight, and early heading. The coincidence between locations for discarded entries was greater (P < 0.05) than that expected under the hypothesis of independent selections for all traits except early vigor (Table 3). However, the difference between observed and expected frequency was very wide only for days to heading.

The level of repeatability found for all traits except early vigor supports the use of information from evaluation of large germplasm collections to enlarge the genetic base available to breeding programs in the region. This conclusion also applies to an important character such as grain yield, which has been considered too location-specific for inclusion in evaluation data made available in germplasm catalogs and other databases (IBPGR, 1981; Marshall and Brown, 1981). Greater values of repeatabilities between locations would probably have resulted in our study with the use of a completely random sample of accessions from the world collection. The preliminary exclusion from the sample of material poorly adapted to Mediterranean environments undoubtedly reduced entry variation for various adaptive traits.

More generally, our results point to the potential usefulness of large-scale evaluation for crop improvement and the pivotal role that genetic resources institutions, especially international research centers or large national centers possibly organized in networks (Adham and Van Sloten, 1990), may play in assessing and making available relevant information to users. In this respect, the encouraging results from this study require confirmation from investigations on this or other crops including a larger set of evaluation locations.

	NOTES

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussion REFERENCES

 
The work has partly been carried out at the International Center for Agricultural Research in the Dry Areas (ICARDA), P.O. Box 5466 Aleppo, Syria, within the project "Evaluation and documentation of durum wheat germplasm" funded by the government of Italy.

Received for publication April 19, 1999.

	REFERENCES

TOP NOTES ABSTRACT INTRODUCTION Materials and methods Results and discussion REFERENCES

 

• Adham Y.J., Van Sloten D.H. The case for a wheat genetic resources network. In: Srivastava J.P., Damania A.B., eds. Wheat genetic resources: Meeting diverse needs. Chichester, UK: John Wiley & Sons, 1990:139-144.
• Annicchiarico P., Pecetti L. Yield vs. morphophysiological trait-based criteria for selection of durum wheat in a semi-arid Mediterranean region (northern Syria). Field Crops Res. 1998;59:163-173.
• Bos I. Theoretical aspects of plant breeding. Wageningen, the Netherlands: Part II. International Agricultural Centre, 1988.
• Braun H.-J., Pfeiffer W.H., Pollmer W.G. Environments for selecting widely adapted spring wheat. Crop Sci. 1992;32:1420-1427.[Abstract/Free Full Text]
• Burdon R.D. Genetic correlation as a concept for studying genotype-environment interaction in forest tree breeding. Silvae Genet. 1977;26:168-175.
• Cooper M., Stucker R.E., DeLacy I.H., Harch B.D. Wheat breeding nurseries, target environments, and indirect selection for grain yield. Crop Sci. 1997;37:1168-1176.[Abstract/Free Full Text]
• Dagnelie P. Théorie et méthodes statistiques. Gembloux, Belgium: Les Presses Agronomiques, 1975 .Vol. 2.
• Duvick D.N. Genetic diversity in major farm crops on the farm and in reserve. Econ. Bot. 1984;38:161-178.
• Falconer D.S. Introduction to quantitative genetics, 3rd ed New York: Longman, 1989.
• Federer W.T. Augmented (or hoonuiaku) designs. Hawaiian Plant. Rec. 1956;55:191-208.
• Frankel O.H. Evaluation and utilization – Introductory remarks. In: Frankel O.H., Bennett E., eds. Genetic resources in plants – their exploration and conservation. Oxford and Edinburgh, UK: Blackwell, 1970:395-402.
• IBPGR. Revised descriptors for wheat. Rome: IBPGR Secretariat, 1981.
• Marshall D.R., Brown A.H.D. Wheat genetic resources. In: Evans L.T., Peacock W.J., eds. Wheat science – Today and tomorrow. Cambridge, UK: Cambridge Univ. Press, 1981:21-40.
• Pecetti L., Annicchiarico P. Yield potential under moderately favourable conditions, drought tolerance, resistance to yellow rust and to common bunt in a world collection of durum wheat. J. Genet. Breed. 1991;45:207-214.
• Pecetti L., Annicchiarico P., Damania A.B. Biodiversity in a germplasm collection of durum wheat. Euphytica 1992;60:229-238.
• Pederson D.G., Rathjen A.J. Choosing trial sites to maximize selection response for grain yield in spring wheat. Aust. J. Agric. Res. 1981;32:411-424.
• Plucknett D.L., Smith N.J.H., Williams J.T., Murthi-Anishetty N. Gene banks and the world's food. Princeton, NJ: Princeton Univ. Press, 1987.
• Porceddu E. Variation for agronomical traits in a world collection of durum wheat. Z. Pflanzenzüchtg. 1976;77:314-329.
• Qualset, C.O., and Y.P. Puri. 1975. Heading time in a world collection of durum wheat. Photo- and thermal-sensitivity related to latitudinal responses and geographic origins. p. 165–178. In G.T. Scarascia Mugnozza (ed.) Genetics and breeding of durum wheat. Univ. of Bari, Bari, Italy.
• Shapiro R.S., Wilk M.B. An analysis of variance test for normality (complete samples). Biometrika 1965;52:591-611.[Free Full Text]
• Snedecor G.W., Cochran W.G. Statistical methods, 6th ed Ames, IA: Iowa State Univ. Press, 1967.
• Spagnoletti-Zeuli P.L., Qualset C.O. Geographical diversity for quantitative spike characters in a world collection of durum wheat. Crop Sci. 1987;27:235-241.[Abstract/Free Full Text]
• Spagnoletti-Zeuli P.L., Qualset C.O. Flag leaf variation and the analysis of diversity in durum wheat. Plant Breed. 1990;105:189-202.
• Spagnoletti-Zeuli P.L., Qualset C.O. Evaluation of five strategies for obtaining a core subset from a large genetic resource collection of durum wheat. Theor. Appl. Genet. 1993;87:295-304.[ISI]
• Ud-Din N., Carver B.F., Clutter A.C. Genetic analysis and selection for wheat yield in drought-stressed and irrigated environments. Euphytica 1992;62:89-96.
• Williams J.T. Plant genetic resources: Some new directions. Adv. Agron. 1991;45:61-91.
• Yang R.C., Jana S., Clarke J.M. Phenotypic diversity and associations of some potentially drought-responsive characters in durum wheat. Crop Sci. 1991;31:1484-1491.[Abstract/Free Full Text]

This article has been cited by other articles:

				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]

This Article Abstract Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Annicchiarico, P. Articles by Doust, M. A. Search for Related Content PubMed Articles by Annicchiarico, P. Articles by Doust, M. A. Agricola Articles by Annicchiarico, P. Articles by Doust, M. A.