MATERIALS AND METHODS

Forty-nine Tef lines (Table 1) including one local and two standard checks were brought from Debreziet Agricultural Research Center and tested in 2008 main season. The field experiment was conducted at locations, Sinana and Adaba. Sinana station is located at 2400 m.a.s.l and 463 km south east of Addis Ababa in Bale zone and 33 km east of Robe town. The total annual rain fall for the year 2008 was 1475.90 mm. The mean minimum and maximum annual temperature of the same year was 9.56 and 20.19°C, respectively. The soil type of Sinana is dark-brown (pellic Vertisol) with slightly acidic reaction (Geremew et al., 1998). Adaba station is located 347 km from Addis Ababa on the way to Bale Robe. Total rain fall of the test location during growing season was 1172.2 mm with the average temperature maximum of 20.36°C and minimum of 10.66°C, respectively (source: Herero State Farm Meteorology Station). The soil type of Adaba station is clay soil (source: SARC).

Simple Lattice Design of 7x7 with two replications and spacing of 1 m between plots, 1.5 m between blocks were used. The treatments were sown on 2x2 m plot area in accordance with the recommended seeding rate of 30 kg ha^-1 and fertilizer rate of 60 kg ha^-1 N and P₂O₅, respectively.

Table 1:	Lists of lines used in an experiment and their origin

Data were collected on plant and plot basis for 16 agronomic traits. Data per plant were collected after thinning to 5 cm between plants on plot size of 0.60 mx2 m from 2x2 m experimental unit so that to record data per plant precisely and all data per plot were taken from a plot size of 1 m² using a quadrant.

Statistical analysis: Pooled Analysis of variance for locations were done using the mean values of ten sampled plants for all data taken on plant basis as well as for those taken on plot base. The data of the two locations were tested for normal distribution and homogeneous variance and analyzed by using simple lattice design (Gomez and Gomeze, 1998). Least Significant Difference (LSD) were used to separate the means both at 1 and 5% probability levels using SAS statistical software. Genotypic (σ²g) and phenotypic (σ²P) components of variances were estimated as suggested by Burton and De Vane (1953), Heritability (h²) for all characters was computed as suggested by Falconer and Mackay (1996).

and Expected Genetic Advance (GA) for each character at 5% selection intensity were computed using the methodology described by Johnson et al. (1955).

GA = Kxσ²xh²

RESULTS AND DISCUSSION

Combined analysis for traits showed that highly significant (p<0.01) differences for days to panicle emergence, lodging percentage, thousands kernel weight, grain yield per plant and grain yield per hectare. Similar result was also reported by Asfaw and Danno (2011) in case of days to panicle emergence (days to head). Significant differences (p<0.05) were observed for panicle length, shoot biomass and number of branch per main panicle (Table 2). Non significant (p<0.05) variations in genotypes were observed for days to maturity, culm length, plant height, number of node, productive tiller number, grain yield per panicle and grain filling period. Non significant variation (p<0.05) in location was observed only for traits harvest index, lodging percentage, grain filling period and thousand kernel weight; but the rest of the traits were significant (p<0.05). Coefficient of variations ranges from 2.34-23.71%. Highest coefficient of variation (23.71%) was observed for productive tiller number and the lowest coefficient of variation (2.34%) was observed for days to maturity.

Genotype by location interaction were non significant for all traits which indicated performance of the genotypes were consistent for these traits across locations. Balcha et al. (2003) reported that the presence of substantial genetic variability for grain yield and component traits was observed among tef genotypes comprising land races and improved cultivars.

Genotypic coefficients of variation, phenotypic coefficients of variation, Heritability and Genetic advance: High phenotypic coefficient of variation was observed for thousand kernel weight (78.82), productive tiller (46.45) and grain yield per plant (34.15). Greater genotypic coefficient of variation were also observed for traits, thousand kernel weight (40.34), productive tiller (32.46) and grain yield per plant (21.18) (Table 3). All phenotypic coefficient of variation result is greater than genotypic coefficient of variation in this study which is congruent with the present study of Kotal et al. (2010), Mohammed et al. (2011), Yadav et al. (2011) and Jalata et al. (2011).

Table 2:	Combined analysis of variance for 16 traits of 49 Tef genotypes planted at Sinana and Adaba in 2008
*: Significant (p<0.05), **: Highly significant (p<0.01), ns: Not significant, BMS: Block mean square, TrtMS: Treatment mean square, CV: Coefficient of variation, DPE: Days to panicle emergence, DM: Days to maturity, PL: Panicle length (cm), CL: Culm length (cm), PH: Plant height (cm), NN: No. of node, PTLN: Productive tiller number, HI: Harvest index, GPa: Grain yield per main panicle (g), LOD: Lodging percentage, SBM: Shoot biomass (kg), GP: Grain filling period, NPB: No. of panicle main branch, TKW: Thousand kernel weight (g), GYpl: Grain yield per plant (g), GYha: Grain yield per hectare (kg)

Table 3:	Components of variances, coefficients of variation, heritability, genetic advance for sixteen characters in Tef genotypes grown at Sinana and Adaba in 2008
DPE: Days to panicle emergence, DM: Days to maturity, PL: Panicle length (cm), CL: Culm length (cm), PH: Plant height (cm), NN: Number of node, PTLN: Productive tiller number, HI: Harvest index, GPa: Grain yield per main panicle (g), LOD: Lodging percentage, SBM: Shoot biomass (kg), GP: Grain filling period, NPB: No. of panicle main branch, TKW: Thousand kernel weight (g), GYpl: Grain yield per plant (g), GYha: Grain yield per hectare (kg)

Previous study reported that high phenotypic coefficient of variation for productive tillers (21%) and grain yield per panicle (22%) in tef (Kebebew et al., 2001). Wide difference of Phenotypic coefficient of variability (78.82%) and genotypic coefficient of variability in 40.34% in thousand kernel weight, contradict with the reports of Riaz and Chowdhry (2003). High genotypic variability facilitates selection for improvement and widens the chance of heritability of traits from parent to offspring.

Maximum heritability estimate were observed for days to panicle emergence (75.50%) and grain filling period (66.79) (Table 3). Heritability estimates along with expected genetic gain is more useful than the heritability value alone in predicting the resultant effect for selecting the best genotypes (Johnson et al., 1955). High genetic gain was observed for lodging (10.90%) and days to panicle emergence (8.05%).

CONCLUSION

Most of traits under study were showed significant variations, from low to high magnitude of heritability and genetic advance. These can facilitate selecting and utilizing the most preferred traits of interest and also hint the potential of tef for further improvement. High phenotypic and genotypic coefficient of variation especially for traits like thousand kernel weight, productive tiller and grain yield per plant are detrimental which make possible tef improvement breeding program. Broad sense heritability estimates together with genotypic coefficient of variation is important to exploit selection in germplasm improvement, since it bases all genetic effects.

ACKNOWLEDGMENT

The authors would like to acknowledge Tesfaye Letta, Tilahun Bayisa, Deme Nuguse, Tesfaye Tadess, Habtamu Legese, Gezahegn Tadess for their technical support and Oromia Agricultural Research Institute for financing this study.