Introduction
World population is expected to increase by 2.6 billion over the next 45 years. Ethiopia is one of the nine countries predicted to have the largest increase. Therefore, there is a pressing need to increase food production to feed this increasing population. Wheat is an important commodity crop that could contribute a major part in achieving the Ethiopia's agricultural objective of food grain self-sufficiency. In its strategic plan, the Ethiopian Agricultural Research Organization (EARO) considered wheat as no. 1 priority crop among cereals (EARO, 2000). Despite having high potential environments and being rich in diversity, the average national yield of wheat is low (1.8 t/ha).
The demand for wheat is continuously increasing. The estimated 2.2 Mt national wheat grain requirement is 50% greater than the total production, and therefore Ethiopia is a net importer of wheat (Eshetu, Reference Eshetu2002) suggesting the need to increase production. A dual strategy based on increasing wheat productivity through development of high yielding varieties and increase in area under wheat crops through expanding wheat cultivation to new areas was suggested to meet this demand (EARO, 2000). Although Ethiopia has potential environments for expanding wheat, meeting expected demands by continued expansion of agricultural production into marginal areas might be difficult as the economic costs of establishing new farms are high (Skovmand et al., Reference Skovmand, Reynolds and Delacy2001). Therefore, the country has to work more towards increasing productivity and total production thereof. Future gains in yield potential require exploitation of the largely untapped sources of genetic diversity housed in collections of wheat landraces and wild relatives (Skovmand et al., Reference Skovmand, Reynolds and Delacy2001). Even in the age of genomics, genetic diversity remains the cornerstone of crop improvement (Sneller et al., Reference Sneller, Nelson, Carter and Cui2005).
In Ethiopia, large amounts of wheat germplasm (about 12,000 accessions) have been maintained in the Institute of Biodiversity Conservation (IBC). Landraces constitute the lion's share of these collections. Landraces, which are locally adapted genotypes that have evolved because of natural and artificial selection forces over the millennia are one of the invaluable heritages that traditional farmers have given us (Myers, Reference Myers1994). Landraces may be used as starting populations for cultivar development (Lakew et al., Reference Lakew, Semeane, Alemayehu, Genre, Grando, van Leur and Ceccarelli1997) or as sources for the introgression of genes and quantitative trait loci conferring resistance to biotic (Huang et al., Reference Huang, Hsam and Zeller1997) and abiotic stresses (Forster et al., Reference Forster, Ellis, Thomas, Newton, Tuberosa, El-Enein, Bahri and Ben Salem2000). Lakew et al. (Reference Lakew, Semeane, Alemayehu, Genre, Grando, van Leur and Ceccarelli1997) confirmed the presence of individual genotypes within landraces that have a yield potential comparable with the best breeding lines. Various authors (Vavilov, Reference Vavilov1951; Porceddu et al., Reference Porceddu, Perrino, Olita and Mugnozza1973; Amri et al., Reference Amri, Hatchett, Cox, Bouhassini and Sears1990; Belay et al., Reference Belay, Tessema, Becker and Merker1993; Kubo et al., Reference Kubo, Jitsuyama, Iwama, Hasegawa and Watanabe2004) reported the uniqueness of the Ethiopian tetraploid wheat germplasm for different useful traits. However, it is felt that these collections have not been fully utilized in the breeding programmes. Efficient utilization of the genetic potential held in the germplasm collections requires a better knowledge of the collected material including morphological and phenological characterizations. Although several authors (Jain et al., Reference Jain, Qualset, Bhatt and Wu1975; Bekele, Reference Bekele1984; Negassa, Reference Negassa1986a, Reference Negassab; Bechere et al., Reference Bechere, Belay, Mitiku and Merker1996; Pecetti and Damania, Reference Pecetti and Damania1996; Eticha et al., Reference Eticha, Bekele, Belay and Börner2005; Hailu et al., Reference Hailu, Merker, Singh, Belay and Johansson2006) have conducted variability studies in Ethiopian tetraploid wheats, there is still a need for more information on population structure and about their potential input for breeding. Hence, in this study, phenotypic diversity was investigated using 271 accessions collected from all over the country. Moreover, the breeding opportunities prevailing in Ethiopian tetraploid wheat have been summarized and strategies to enhance yield and quality traits were suggested.
Materials and methods
Plant materials and data collection
A total of 271 tetraploid wheat landraces collected from all geographical regions of Ethiopia were used. Because only 1, 6 and 6 accessions, respectively, were represented each region of Ilubabur, Kefa and Sidamo, accessions from these three regions were pooled together and the three provinces were designated as IKS. Figure 1 shows the different regions of Ethiopia and its neighbours. The accessions have also been classified based on the four altitudinal classes: I [2000 meters above sea level (masl)], II (2001–2500 masl), III (2501–3000 masl) and IV (>3000 masl). The number of accessions belonging to these four altitudinal classes is 47, 123, 89 and 12, respectively. In Ethiopia, wheat is mainly grown under rainfed condition at altitudes ranging from 1800 to 2800 masl. In some parts of the country, it is known to grow at above 3000 masl. Consequently, a larger number of accessions were sampled from altitude classes II and III.
Fig. 1 Map showing the different regions of Ethiopia and its neighbours.
The experiment was conducted during 2002/2003 and 2003/2004 main cropping seasons at the Alemaya University research site (rare), which is located at 1980 masl. A randomized complete block design with two replications was used. Each plot consisted of two rows, each 1 m long and 20 cm apart. The distance between the blocks and the spacing between plots were 1.5 m and 0.5 m, respectively. Fertilization of experimental plots and all other cultural management were done following the recommended cultural practices.
Thirteen quantitative phenological and morpho-agronomic traits were measured (Table 3). Data for days from emergence to anthesis (DTH) and from emergence to maturity (DTM) were determined on plot basis, and grain filling period (GFP) was determined as days between these two phenological traits. Plant height (PH) in cm and spike length (SL) in cm and spikes/plant (SP), spikelets/spike (SS), kernels/spike (KS) and kernels/plant (KP) in numbers were determined based on five randomly selected plants per plot. Grain yield per plant (GY) and biomass yield per plant (BY), both in grams, and harvest index were assessed on ten randomly selected plants per plot and 1000-kernel weight (TKW) in grams was determined from dried samples of 1000 grains.
Statistical analysis
All measured variables were subjected to analysis of variance procedures to assess differences among varieties. Mean value over the 2 years for each character was used to determine descriptive statistics such as the range, arithmetic means and standard errors of means for each of the variables. The Shannon–Weaver (Reference Shannon and Weaver1949) diversity index (
) was used as a measure of phenotypic diversity for each trait after their transformation into classes. It is estimated using 
where for a given character C, n is the number of phenotypic classes and p is the proportion of observation in the i th class. The diversity index ranges from 0 to 1, where 0 indicates complete evenness and 1 shows complete unevenness. The index was estimated for each character over all accessions and for each character within a region and altitudinal class. Due to its additive property (Kent and Coker, Reference Kent and Coker1992), Shannon–Weaver diversity indices obtained for each character were pooled for each region and altitudinal class over the respective number of accessions. To avoid the effect of the different numbers of phenotypic classes while comparing indices obtained for the different characters, a standardized index (SDIc) was calculated as 
. The diversity index computed based on the whole dataset (H t) was partitioned into within region or altitude and between region of origin and between altitudinal classes following the procedure of Paul et al. (Reference Paul, Wachira, Powell and Waugh1997). The within-region (H r) and within-altitude class (H a) diversity indices refers to the average diversity index of each character estimated based on the regions of origin and altitudinal classes, respectively. The between-region (Dstr) and between-altitude class (Dsta) diversity index were computed as H t − H r and H t − H a, respectively. (Dstr)/H t and (Dsta)/H t are the coefficients of gene differentiation based on the regions and altitudinal classes, respectively. Principal component analysis on the diversity index was conducted using the computer program NTSYS-pc (Numerical and Taxonomy and Multivariate Analysis system version 2.0; Rohlf, Reference Rohlf1998).
Results
Descriptive statistics
Descriptive statistics such as the minimum, maximum, mean and standard error of means from the combined analysis of variance for 13 quantitative traits were presented in Table 2. The combined analysis of variance results revealed that the genotypic differences among genotypes were highly significant (P ≤ 0.05) for all the characters. Also, the range of variation, which is the difference between the maximum and minimum values of each character, with regard to the 13 quantitative traits indicated the presence of wide variation for all quantitative variables. Among traits, GY per plant ranged from 3.7 to 9.2 g/plant with a mean value of 8.3 g. DTM, DTH and GFP have shown range values of 30, 31 and 22 days, respectively. High differences between the maximum and minimum mean values were also found for the other traits.
Table 1 Geographic region, number of sampled accessions (N) and estimates of the Shannon–Weaver diversity index (H′) for the 13 metric characters across geographical regions and altitudinal classes
Table 2 Summary of descriptive statistics from combined analysis of variance
Estimates of diversity
The overall diversity index estimated based on the entire dataset is 0.72 (Table 3). Among regions, the range of mean index of diversity varied from 0.68 for the accessions from Hararghe to 0.40 for the accessions from Gamugofa. The diversity indices differed among regions for specific characters. For instance, traits such as SS, SL, seeds/spike, spikes/plant and KP showed high polymorphism in Arsi, Hararghe and Shewa. The highest diversity index (0.75) for PH was recorded in collections from Arsi. Hararghe and Shewa also exhibited high polymorphism for BY and harvest index (Table 1).
Table 3 Partitioning of the phenotypic diversity within and between geographical regions of collections and altitudinal classes
H t, diversity index for each character computed from the whole dataset; H r and H a, the average diversity indices of each character estimated based on regions of origin and altitudinal classes, respectively. Dstr and Dsta are the proportions of diversity within regions and altitudinal classes, respectively. Gstr and Gsta are the coefficients of gene differentiation based on regions and altitudinal classes, respectively.
Among the three altitudinal classes, the highest (0.72) and lowest (0.61) mean diversity indices were was noted in altitude classes II and IV, respectively. The diversity indices differed among altitudinal classes for specific characters. In altitude class I, all characters except GY and biomass showed diversity index of more than 0.70. In altitude classes II and III, most of the traits except GY and SS in the former and GY and days to maturity in the latter displayed a diversity index value of more than 0.70. In altitude class IV, however, only five traits have exhibited a diversity index of greater than 0.70 (Table 1), indicating low polymorphism for traits in this altitudinal class than others.
Partitioning of phenotypic diversity
Subdividing the variation into its components may facilitate genetic resources conservation and utilization, by determining the relative contribution of the different levels of variability to the total diversity available in any one area. To determine the significance of the different regional components, the total variation was partitioned into within- and among-region diversity. The coefficient of regional diversity differentiation is 0.29 implying that 71% of the total variation was explained by the within-region diversity (Table 3). Among the 13 traits, BY, SL, HI, GFP and SL showed high within-region variation indicating their relative significance for differentiating accessions within regions. On the other hand, the two phenological traits, DTM and DTH, and PH contributed more to the between-region variation. The partitioning of between- and within-altitude diversity revealed that 95 and 5% of the total variation was attributed to the within- and between-altitude class variation, respectively (Table 3). The contribution of individual characters to the within-altitude diversity revealed that SS, TKW, GFP and BY have slightly stronger effect than others. In Ethiopian tetraploid wheat, studies (Bekele, Reference Bekele1984; Bechere et al., Reference Bechere, Belay, Mitiku and Merker1996; Pectti and Damania, Reference Pecetti and Damania1996) showed greater contributions of the lower (within populations and among populations within regions and altitude zones) than the higher (among regions and altitude zones) level hierarchies to the total phenotypic variation.
Phenotypic diversity was also examined using multivariate analyses. To this end, principal component analysis was computed on the diversity index of regions of origin and altitudinal class for the 13 quantitative traits to examine the regional and altitudinal patterns of variation. On regional bases, the first four axes, whose eigenvalues are greater than 1, explained about 82% of the observed phenotypic diversity in the 271 tetraploid wheat accessions (data not shown). The first and second axes accounted for about 29.8 and 20.5% of the total variation, respectively. Johnson and Wichern (Reference Johnson and Wichern1988) suggested that coefficient or eigenvector greater than half divided by the standard deviation of the eigenvalue of the respective PC is more important in explaining the overall variation. Following this criteria, traits such as days to heading, KS, SL and SP accounted for much of the variation on these axes. On altitudinal bases, only the first two principal components, which produced eigenvalues greater than 1, explained 89.7% of the total variation (data not shown).
Discussion
Genetic diversity: a raw material and an opportunity for breeding
Efficient utilization of the genetic potential held in the germplasm collections requires, among other things, a better knowledge of the genetic diversity present in the collected material. Knowledge of the extent of variability for plant traits and association of specific traits with geographic origin not only facilitates breeding programmes but also helps to define needs and locations for future collections of germplasm. In this study, results of analysis of variance and descriptive statistics indicated that the genotypes differed significantly for all the traits revealing the presence of a considerable diversity, which could be utilized in developing high yielding cultivars through selection breeding. The overall Shannon–Weaver diversity index for all traits was 0.74. Because Shannon–Weaver diversity index is sensitive to both the type of phenotypic descriptor and the number of descriptor classes used (Grenier et al., Reference Grenier, Bramel, Dahlberg, El-Ahmadi, Mahmoud, Peterso, Rosenow and Ejeta2004), direct comparison of Shannon–Weaver diversity indices from different studies involving different descriptors and descriptor classes need caution. In Ethiopian tetraploid wheats, Jain et al. (Reference Jain, Qualset, Bhatt and Wu1975), Negassa (Reference Negassa1986a), Bechere et al. (Reference Bechere, Belay, Mitiku and Merker1996) and Eticha et al. (Reference Eticha, Bekele, Belay and Börner2005) employed the same method to investigate phenotypic diversity and reported a Shannon–Weaver diversity index of 0.70, 0.81, 0.87 and 0.71, respectively. As with the findings of phenotypic diversity studies, the presence of appreciable genetic diversity in Ethiopian tetraploid wheats has been reported from variability studies using microsatellites (Messele, Reference Messele2001; Alamerew et al., Reference Alamerew, Chebotar, Huang, Roder and Borner2004; Teklu et al., Reference Teklu, Hammer, Huang and Röder2006a, Reference Teklu, Hammer, Huang and Röderb), cytological markers (Belay and Merker, Reference Belay and Merker1999), isozymes (Tsegaye et al., Reference Tsegaye, Becker and Tessema1994, Reference Tsegaye, Tessema and Belay1996), glutenine and gliadine storage protein and amplified fragment length polymorphisms (Messele, Reference Messele2001). Many microsatellite loci might be significantly linked to agronomically important traits (Teklu et al., Reference Teklu, Hammer, Huang and Röder2006a, Reference Teklu, Hammer, Huang and Röderb). Various authors also confirmed the uniqueness of the Ethiopian tetraploid wheat germplasm for different useful traits. For example, they have valuable features such as early ripening, short culm, long coleoptiles and low tillering (Porceddu et al., Reference Porceddu, Perrino, Olita and Mugnozza1973); high degree of allelic variation for quality traits such as the seed storage proteins (glutenins and gliadins; Messele, Reference Messele2001); resistance to powdery mildew and glume blotch (Negassa, Reference Negassa1986a); Hessian fly (Amri et al., Reference Amri, Hatchett, Cox, Bouhassini and Sears1990) and stripe rust and moderate resistance to pH and drought (Porceddu et al., Reference Porceddu, Perrino, Olita and Mugnozza1973). Vavilov (Reference Vavilov1951) found Ethiopian tetraploid wheat that had 20% protein. Ethiopian tall type (rht) landraces of durum wheat (Triticum durum Desf.) showed higher root penetration ability than semi-dwarf (rht) varieties bred in North America (Kubo et al., Reference Kubo, Jitsuyama, Iwama, Hasegawa and Watanabe2004). High variation was also reported for seed colour, kernel texture, flour colour, seed size and protein content (Negassa, Reference Negassa1986b).
Strategies to utilize genetic diversity
The best way to exploit the diversity contained in landraces and wild relatives is to introduce the valuable variation for qualitative and quantitative traits into adapted breeding materials using wide-crossing (Skovmand et al., Reference Skovmand, Reynolds and Delacy2001) and complex crosses (Vetelainen, Reference Vetelainen1994). In barley, the study of Vetelainen (Reference Vetelainen1994) indicated that hybrids from the complex crossing programme exceeded parents in earliness and TKW. Also, it is useful to consider the paradigm shift proposed by Tanksley and McCouch (Reference Tanksley and McCouch1997) in which the most divergent accession(s) relative to elite cultivars are used to increase genetic variability for the improvement of quantitative traits. This strategy was used successfully to improve quality traits in tomatoes (Fulton et al., Reference Fulton, Grandillo, Beck-Bunn, Fridman, Frampton, Petiard, Uhlig, Zamir and Tanksley2000).
Natural populations harbour rich genetic diversity, which is eco-geographically structured and largely adaptive (Nevo, Reference Nevo, Miller and Koebner1988). As a result, landraces provide a valuable resource for plant breeding as well as for the preservation of genetic diversity. Under Ethiopian condition, where wheat landraces cultivation is predominant (Tessema et al., Reference Tessema, Becker, Belay, Mitiku, Bechere and Tsegaye1993; Bechere et al., Reference Bechere, Belay, Mitiku and Merker1996; Tessema and Bechere, Reference Tessema and Bechere1998), the first step in breeding should be the utilization of indigenous materials (Tessema, Reference Tessema, Engels, Hawkes and Worede1991). Enhancing the yield of landraces, while maintaining an appreciable level of genetic diversity, is crucial to improve their competitiveness with modern varieties and maximize their utilization (Tessema and Bechere, Reference Tessema and Bechere1998). A modification of phenotype mass selection by selecting pure lines from genetically mixed landrace populations through yield testing and then bulking two or more superior pure lines has been suggested as one of the best strategies to improve the productivity of the landrace cultivars grown by the farmers (Tessema, Reference Tessema, Engels, Hawkes and Worede1991).
The huge diversity contained in the large amounts of wheat germplasm that have been collected and maintained mainly in the Ethiopian IBC has not been exploited effectively in cultivar development. Hoisington et al. (Reference Hoisington, Khairallah, Reeves, Ribaut, Skovmand, Taba and Warburton1999) described this phenomenon as valuable genetic resources are essentially ‘sitting on the shelf’ in what have been dismissively termed ‘gene morgues’. The most attributed reason could be the large numbers of accessions, which make it difficult and time consuming to evaluate for all useful yield traits in the field trials and choose the most promising ones with which to work. Development of wheat core collections based on species could facilitate utilization of the huge diversity stored in genebanks in Ethiopia. Core collections could provide plant breeders a manageable number of accessions to use in the search of new characters or character combinations and a structured way to evaluate whole collections (Rao and Hodgkin, Reference Rao and Hodgkin2002). In addition, germplasm enhancement may be one of the keys for maximizing utilization of germplasm. It has become an important tool for the genetic improvement of breeding populations by gene introgression or incorporation of wild and landrace genetic resources into respective crop breeding pools.
Conservation
In crop improvement, it is not only working with the existing genetic variation that is central, but also parallel and periodic assessment of the threat of loss of diversity is necessary.
Using the calculation scheme, gene erosion = 100% − gene integrity, i.e. the still extant landraces, a genetic erosion up to 100% was detected in T. durum, Triticum dicoccon and Triticum turgidum in some districts of eastern Ethiopia (Teklu and Hammer, Reference Teklu and Hammer2006). Other reports (Worede, Reference Worede, Holmes and Tahir1983; Hailu, Reference Hailu, Gebremariam, Tanner and Huluka1991; FAO, 1996; Tsegaye and Berg, Reference Tsegaye and Berg2007) also reported the problem of genetic erosion in Ethiopian tetraploid wheats. Therefore, priority should also be placed on collection and conservation of landraces, which are irreplaceable materials, if lost. The best method of conservation is the use of complementary approach of the different ex situ and in situ conservation techniques. Apart from conservation, creation of sustainable agricultural systems that actively use as much biodiversity as possible should remain the major goal. The guiding principle of ‘conservation through use’ should be respected because only in use can diversity be appreciated enough to be saved, only in use it can continue to evolve and thus retain its value.
Acknowledgements
This research was conducted at the Alemaya University research site (Rare), Ethiopia. The accessions were provided by the IBC and the Debre Zeit Agricultural Research Centre, Ethiopia. The principal author is also grateful to DAAD (German Academic Exchange Service) for providing him the scholarship.
