Assessment of tissue culture derived regenerants of linseed (Linum usitatissimus L.) in Ethiopia

English: I. The study was undertaken to assess the comparative performance of six linseed regenerants along with two crosses and three check cultivars across 18 linseed-growing environments of Ethiopia from 1996 to 1998. The seed yield and other agronomically desirable characters were analysed with different statistical procedures to determine the adaptation potential, G x E interactions and seed yield stability performance. The main objective of the study was to understand and describe the genotypes and their growing environments by applying different statistical methods of analyses in order to make useful recommendations for the future. Likewise, contemporary studies on the genotypes, environment and their interactions, and various analytical methods ofstability parameters were discussed. 2. Separate and combined analyses of variance across locations and years, seven types of stability parameters, correlation and canonical variate analyses were performed using MSTAT-C, AGROBASE 98 and SAS computer programmes. For the stability analyses, data of 10 varieties evaluated across six locations and three years (excluding the local checks) were analysed by following the procedures of: Francis and Kannenberg (1978) for the coefficient of variation, Finlay and Wilkenson (1963) and Eberhart and RusseIl (1966) for the joint regression, Wricke (1962) for ecovalence, Shukla (1972) for stability of variance, Lin and Binns (1978) for cultivars' superiority measure, Nassar and Huehn (1978) for variance of ranks and Gauch and Zobel (1988) for AMMI stability model. Comparisons were also made among these different stability measurements. Canonical variate analyses were undertaken on SAS CANDISC programme (SAS Institute, 1982) to classify and describe the genotypes and their test localities. 3. The separate trial analyses for the three years have shown highly significant (P<O.Ol) differences among the genotypes for seed yield and most of the measured traits. Totally four regenerants outperformed the crosses in 1996 and most of them repeated their performance during the succeeding years. Across locations and years, Chilalo ranked first (1505 kg ha¹), followed by three regenerants (RII-M20G, RIO-N27G and RII-NI266), with a yield ranging from 1414-1455 kg ha¹. The high yielding performance of the regenerants indicates their high potentials and good adaptability to the linseed growing environments of Ethiopia. In fact, R 11-M20G was already recommended for commercial production in Adet area in 1999. Among the locations, the highest yield of 2172 kg ha-I was obtained from Bekoji, followed by that of Kulumsa over the years, indicating the good potentials of these sites. The result also showed tremendous yield variations over locations and years, suggesting high G x E interactions. The average of ANOV A components over the three years showed that about 45% of the total variance was accounted for by blocks, 39% by genotypes and the remaining 16% was attributed to random errors. As higher variability for blocks was recorded at Holetta, Adet and Asasa, further analysis of environmental factors (edaphic and climatic) and close supervisions are needed. 4. The combined analysis of variance across locations showed highly significant (P<O.Ol) difference among the locations (L), genotypes (G) and their interactions for most of the measured traits, indicating high differential responses of the genotypes over the locations, due mainly to edaphic and climatic related factors. About 76-85% of the variance components was also attributed to locations, while the genotypes accounted for only 3-7% (nearly similar to that G x L component) over the three years. These indicate the confounding effects of environmental factors and thus necessity of stability analysis to select appropriate varieties for their required purposes. 5. The combined analysis of variance and the percentage of its components for the seed yield across years per location show highly significant (P<O.O 1) differences for the years, genotypes and their interactions at Bekoji, Holetta and Kulumsa. In contrast, Y x G interactions were not significant at Sinana, Adet and Asasa, indicating more yield stability over the three years at these sites than the others. The variance components of ANOV A indicate higher variabilty for years or growing seasons, ranging from 50% at Adet to 94% at Bekoji. This large seasonal variability may have been due mainly to the amount and distribution of rainfall, among other factors. Repeatability of the trials at Bekoji and Holetta was about 85% against the lowest of Asasa (48%). This also indicates the high level of environmental variations that needs further diagnosis either to adjust or cope along with them. 6. The combined analysis 'across locations, years and their interactions reveals highly significant differences (P<O.Ol) among the genotypes for all the measured traits, suggesting differential responses of the genotypes to their test environments. As significant G x E interactions tend to confound cultivar selection processes and create difficulties in identifying reliable varieties, stability analysis with appropriate statistical methods are required. The variance components of seed yield were estimated to about 55% for years, 26% for locations, 13% for Y x Linteractions, 3% for genotypes and the remaining 3% for the rest of interactions. Most of these interactions were highly significant due mainly to climatic; soil and biotic factors, and more in depth studies are needed for better understanding and further actions. As a general case, however, when G x E interaction is mainly caused by unpredictable environmental factors, such as year to year fluctuations in rainfall (like in this study), the breeder must try to develop stable varieties that can perform relatively good under a range of conditions. But if G x E interaction is due to predictable environmental factors, such as soil types and management practices, the plant breeder can develop either different varieties for different environments or broadly adapted varieties for a range of conditions. 7. The ANOVA of joint regression model for seed yield showed highly significant difference between the genotypes. According to this joint regression, R 12-Nl OD was found the most stable genotype, followed by P136IIxl0314D and Chilalo (the highest yielder across the environments). All these stable varieties also had higher coefficients of determination, which were significantly correlated with the coefficient of regression and deviation from the regression. NorLin was also non-significantly different from the coefficient of regression and thus had general adaptability to diverse environments. The coefficient of variability also showed similar results. 8. According to Wricke's (1962) ecovalence, RII-M20G followed by Rl1-N1266, R12- NIOD and P13611xl0314D were the most stable genotypes. The first three genotypes were the regenerants of tissue culture, whereas the fourth was one of the crosses developed at Holetta Research Center. Chilalo, NorLin, R12-D33C and P136lIxl0314B were categorised as intermediate in stability, unlike RI0-N27G and DI2-D24C that were found unstable according to this stability measurement. 9. Shukla' s stability variance (1972) showed that R12-NI0D, PI3611xl0314D and Chilalo were the most stable genotypes, while D12-D24C, RII-NI266 and RII-M20G were classified as the least stable. R 12-N IOD, the regenerant from NorLin was the most stablegenotype as measured by both ecovalence and stability variance. Join regression was also in close agreement with these results. 10. Lin and Binns's (1988a) cultivars' superiority measure indicated Chilalo, RI0-N27G and RII-NI266 were the most stable genotypes, while 012-024C and P13611xl0314B were the least stable. In most cases, ranks of cultivar superiority measure were in harmony with the ranks of varietal mean yield rather than with other stabil ity parameters. Il. Nassar and Huehn' s (1978) non-parametric measure of stabi Iity revealed that R 12-N 100 had the smallest changes in ranks and thus was the most stable regenerant unlike D12- 024C, which was significantly unstable. The next more stable varieties were P13611xl03140 and Chilalo. This result was in agreement with most of the above stability measurements. 12. Additive main effects and multiplicative interaction's (AMMI) stability value, and scores of the interaction principal component analysis (IPCA) indicated that RI2-N 1OD, P1361Ixl0314D, RI2-D33C and Chilalo were relatively the most stable genotypes across the tested environments of Ethiopia. On the other hand, R11-N1266, RI0-N27G and Norlin were specifically adapted to low or unfavorable conditions, according to these parameters. AMMI model has been widely and successfully used during the past few years to analyse and understand the G x E interactions and stability in many crops. Since it combines the analysis of variance and principal components analysis in one model, it describes adequately both the G x E interaction and stability analysis through its response patterns. 13. Comparison of the seven stability parameters has shown that the coefficient of variability, Shukla's (1972) stability variance, Nasser and Huehn's (1978) variance of ranks and AMMl's stability value (ASY) were harmonious in detecting the most stable genotype, RI2-NIOD. Ecovalence and deviation from regression also revealed this genotype as one of the stable varieties and only cultivars superiority measure categorised it in the intermediate stability group. The same was true to with the second most stable variety (PI361Ixl0314D). In general, AMMI, Eberhart and Russell's (1966) deviation from regression, Nasser and Huehn's (1978) variance of ranks and Shukla's (1972) stability variance were found very useful in determining the comparative stability of linseed genotypes considered in this study. The coefficient of variability and ecovalence were also relatively better than the cultivar's superiority measure. All in all, the seven parameters detected R12-NIOD, PI361lxl0314D and Chilalo as the most stable varieties, and R12- D24C, RIO-N27G and P13611xl0314B as unstable ones, while the rest were intermediate between these two groups. However, repeatability study is needed to determine the best parameter. 14. The evaluation oil content and oil yield indicated that the highest location mean of 38.26% was obtained from Holetta, followed by that of Bekoji (36.6%). Of the genotypes, R 12- D33C and R12-D24C gave the highest of about 37.4% across the localities. R12-NIOD was also good in its oil percentage, like its seed yield. These varieties should, therefore, be used in the crossing programme to improve the oil contents. The analyses of variance for both oil content and oil yield across locations and years indicated highly significant difference (P<O.OI) between the genotypes. The variance components across locations and years also depicted higher variability for years, locations, and genotypes and for their interactions in this order. J 5. The assessment of agronomic characters revealed that the genotypes took 75 and 139 days to reach the flowering and maturity stages, respectively. The early flowered entries have also matured earlier than others after 134 days, unlike the late maturing ones that took up to 144 days. RII-NI266, RIO-N27G, RI2-N10D and NorLin were among the early maturing group, while R11-M20G, PI361Ixl0314D and the local checks were late maturing. The two crosses were found more susceptible to powdery mildew, while RI2-D33C, RI2-D24C, RII-M20G and Chilalo were relatively resistant to powdery mildew and pasmo diseases.16. The correlation among the measured characters showed highly significant (P<O.OI) positive correlations between oil yield and seed yield (r = 0.924), oil yield and plant height (r = 0.585), and oil yield and stand count (r = 0.656). Seed yield was, however, negatively affected by days to flowering and maturity, indicating the poor yielding ability of early maturing varieties. The same was true with seed yield and powdery mildew, and seed yield and lodging percent. Oil content was positively influenced by days to maturity, plant height and stand percentage, implying that late maturing and tall plants positively contribute to the oil content of linseed. Highly significant negative correlation was noted between the oil content, and powdery mildews, Fusarium wilt and pasmo, indicating the negative effects of these diseases oil content of linseed. 17. Linear discriminant analysis (canonical variate analysis) was used to classify and compare the 10 genotypes and their attributed variates. The first two canonical variates (CANl and CAN2) altogether accounted for 78.01% of the total variation among the groups of genotypes. The horizontal separation (CAN I) was accounted for about 60.63% of the total variation, while the vertical separation (CAN2) attributed for 17.38%. This vertical separation was mainly due to days to flowering, the score of powdery mildew and lodging percent. Days to flowering and lodging percentage played important roles in the horizontal separation as well. Horizontal separation that showed very highly significant contribution in the total variability was used in grouping the genotypes. R12-D33C and R12-D24C contrasted the most with the other genotypes, like P13611 xl 03140. R l2-D33C and Chilalo varieties were also dissimilar with most of other genotypes. P136llxl03l4D was very similar to Chilalo and the same was true for Rll-N1266 and P13611xl0314B. NorLin cultivar was very similar to R12-NlOD, the most stable variety that deserves a license for commercial production. In general, the 10 linseed genotypes were generally classified into two major categories, the genotypes with above mean values (i.e. positive values) and those with below mean values (i.e. negative ones). RI2-D33C, R12-D24C and Rl1-M20G were among the positive values were though the latter regenerant was largely deviated from the group and much more closer to the average. The second group of genotypes that had negative CANl values included Chilalo, P13611xl0314D, RI2-NI0D, NorLin, P13611xl0314B, RI0-N27G and RII-N1266. Nevertheless, RI2-NI0D and NorLin were slightly deviated from this group as they had relatively lower values. The percent of stand count, days to maturity and oil yield played major roles in identifying this second group.18. The same 11 variates employed to describe the genotypes were also used here to explore the similarities and differences of the six locations. The first two canonical variates (CANl and CAN2) together accounted for 96.39% of the total variations among the locations. The horizontal separation significantly (P<0.05) accounted for 91.56% of the variability, while vertical separation was responsible for 4.83%. Thus, CANl was mainly considered in classifying these locations. Bekoji contrasted the most with other locations and has verified the long-standing truth of Bekoji site. It has been very suitable site for good performance of linseed by producing highest seed yield, up to 2.5 t ha". Bekoji was dissimilar to most of the other sites based on the seed yield variable. The mean yield obtained from this site was 1752 kg ha', exceeding the remaining localities by over 40%. Hence, the environment of Bekoji needs special strategy in terms of cultivar development and crop management practices to exploit the existing potentials more effectively. The other variates attributed to distinguish this location were oil content, stand percentage, the score of pasmo and percent of lodging and the same was true with Sinana. Asasa, with its highest negative value was also different, as it has been known for its unreliable rainfall and terminal drought. Asasa was dissimilar to most of the other sites based on its oil yield and Fusarium wilt percentage though it was relatively closer to Adet. Kulumsa showed contrasting negative values with Sinana, both being equally closer to the mean value in the opposite directions. Like Asasa, Kulumsa was discriminated by the oil yield and wilt percentage. Kulumsa has a relatively warmer climate and fertile soils that are conducive for good crop growth and development, resulting in high percentage of lodging. It was also conducive for the development of wilt, powdery mildew and pasmo diseases and it can be used as one of disease screening sites. Sinana scored positively above average and differed from the other sites. This result reflects the existing environment of Sinana, as it has a very different agro-ecology, bimodal and erratic rainfall distribution. The area has got two growing seasons per annum, unlike the other research centers. Holetta scored a negative value, which was very closer to the average value and was ungrouped with any of the localities though Kulumsa was relatively closer to it. Holetta, like the other locations with negative values, was discriminated by the oil yield, wilt and other disease scores. These localities are, therefore, considered as proper sites for screening disease resistant and high oil yielding varieties. In short, the canonical discriminant analysis has confirmed the existence of adequate diversity among these six research centers, and opening some more sub-centers and testing sites are justifiable as far as the results of this study are concerned. However, additional studies are required for broader applications and to make use of the canonical discriminant analysis more effectively.