Previous PageTable Of ContentsNext Page

Efficient Evaluation of Lines Derived from Recurrent Selection in Barley

A. Taleei1 and D.E. Falk2

1. Dept. of Agronomy, Faculty of Agriculture, University of Tehran, Karaj, Iran
2. Dept. of Plant Agriculture, Ontario Agricultural College, University of Guelph, Guelph, Ontario, N1G 2W1 Canada


The RIPE (Recurrent Introgressive Population Enrichment) method has been developed to enable barley breeders to apply basic principles of recurrent selection to a normally self-pollinated crop Yield trials were conducted with F3:4 lines with 10 lines per family at each of the top level of the RIPE system. Stepwise random deletion was used to generate ten new sets of four to nine lines per family at each level; these sub-sets were compared to the original ten lines for yield means, and variances. The results indicate that five randomly selected lines were as effective as the ten original lines in identifying good families and superior lines within families. Comparisons of yield means, maximums, minimums, and variances of the Elite, High and Intermediate levels showed that there is considerable variability in all levels, including the Elite level, therefore it is possible to continue to make progress in selecting for yield within this population, Families at the Intermediate level had the highest relative yields, the greatest proportion of families above the checks in yield, and the highest average variances within families, and so may be the optimum balance of adapted and exotic germplasm for yield expression, The Intermediate level had the best resistance to powdery mildew and rust.

Key words

RIPE, barley, recurrent selection, yield, variance


The RIPE (Recurrent Introgressive Population Enrichment) method (Kannenberg and Falk, 1995) has been developed to enable barley breeders to apply basic principles of recurrent selection to a normally self-pollinated crop. This method combines the hierarchical population structure of the HOPE method developed for maize (Kannenberg 1981,2001) with the male sterile facilitated recurrent selection population (MSFRSP) methods that have been applied to barley. The RIPE method has evolved into a very efficient population breeding method which emphasizes the accumulation of new, desirable alleles in the Elite population while retaining the existing adapted background. The RIPE system is based on numerous crosses at several levels of a hierarchical population structure. Populations of F3:4 lines at each of the Elite, High (87.5% Elite germplasm), and Intermediate (75% Elite germplasm) are generated each year with limited amounts of seed. Evaluation of these lines is primarily to identify the better lines in each population relative to established check varieties, and to eliminate lines with obvious defects from further testing. Generally, the genotype by year interactions are significant, but not genotype by location interactions, so selection at a single location is considered to be valid for most traits, including yield. The main criteria for identifying and recycling desirable genotypes is an unreplicated F3:4 line yield trial. The determination of the minimum number of lines to test per cross and the number of crosses per level for maximizing the efficiency of system has not been determined. The relative performance and variance at each level of the population hierarchy needs to be determined to help guide the stringency and type of selection made within each level, and within vs among crosses at each level. Peel and Rasmusson (2000) evaluated material at several levels of introgression of two-rowed barley germplasm into six-rowed barley backgrounds. They concluded that three cycles of introgression (approximately 87.5 % six-rowed background) resulted in lines with significant yield and agronomic improvements; this would be comparable to the High level of the RIPE system which is approximately 87.5% Elite background. They evaluated 20 F2:4 lines in four sets of five lines in each population.

The objectives of these experiments were to determine the minimum number of lines necessary to identify high yielding lines and families, and to evaluate the relative performance of lines and families within each of the three top levels of the RIPE system at this time.

Materials and Methods

Development of Lines

In each cycle of the RIPE system, crosses between Elite male sterile plants (seed from the F2 of the previous round of crosses) and selected F3:4 lines (F3:5 male parents) are made to produce Elite, High and Intermediate level populations. The crosses to produce material used in this study were made in a growth room (22C/15C day/night with 16 hr days) in the fall of 1998. The F1 plants were selfed in the same growth room in the winter of 1999; the F2 populations were grown as bulk populations in the field at the Elora Research Station, near the University of Guelph, during the summer of 1999. The F3 bulk populations were grown in an off-season nursery in the winter of 1999/2000 at Brawley, California as spaced plants (30cm within rows that were 1 m apart) on irrigated, raised beds. The F4 plots were grown from seed of single F3 plants within each cross. The F3:4 generation was sown as yield plots (1.25m 3.0m) at commercial seeding rates at the Elora Research Station during the summer of 2000. Trials were harvested with a plot combine, dried to approximately 14% moisture and weighed for total yield. Locally adapted, high yielding cultivars 'Chapa is' and 'Brucefield' were grown as checks, alternating at 11 plot intervals. Thus, every trial plot was within five plots (6.25m) of a check plot. Plots were scored for agronomic characteristics and the naturally occurring diseases powdery mildew (Erysiphe graminis fs hordei) and leaf rust (Puccinia hordei).

Statistical Evaluation of Field Trials

Number of lines. Mean yields of each pair of checks were used to determine variation across the field as the base for adjusting entry yields for field variation. Ten lines of each of 40 Elite level cross families were arbitrarily allocated to plots for determining minimum number of lines necessary for identifying highest yielding lines. Stepwise random deletion was used to generate ten new data subsets for each family composed of 9, 8, 7, 6, 5, and 4 lines each. These were then compared to the original set of ten lines to determine correlations of means, ranges, and variances.

Comparison of levels: In the second set of materials, comparisons of yield means and variances of 63 families of Elite (24 families, 502 lines), High (17 families, 225 lines), and Intermediate (22 families, 138 lines) levels with a minimum of five lines each were made to determine differences in yield among the levels.

Results and Discussion

Data on yield were standardized to the two running, alternating check varieties. Check variety Brucefield consistently had a higher yield than check Chapais (5.0 t/ha vs 4.2 t/ha). Brucefield was derived from a cross involving Chapais, and has generally better disease resistance and agronomic characters. In plotting the means of the check varieties across the plot area, no significant trends in yield were detected, leading us to conclude that the test plot area was reasonably uniform in fertility and other factors that might affect yield. Both checks had moderately resistant reactions to powdery mildew while Brucefield was more resistant to leaf rust than Chapais. The test entries varied from highly resistant to both diseases to highly susceptible to both, with most lines having some resistance to one or the other disease,

Number of Lines per Cross

The sets of lines generated by stepwise random deletion were compared to the original ten lines for variance among lines within the corresponding set. Figure 1 shows that correlations among the means are relatively high (> 0.75) until less than five lines are left in each subset. With only four lines in each subset, the correlations averaged less than 0.60 and were not acceptable as an estimate of the potential performance of the cross combination. Five or more lines were not greatly different from the full set of 10 lines in estimating the mean yields nor the best yielding lines. The best correlations of the reduced size subsets remained above 0.80 indicating that there generally were subsets, which were close to, the original sets in variance. The poorest correlations tended to become worse quickly as the number of lines in the subsets was reduced. Thus, the chance of having a non-representative subset increased fairly rapidly as the size of the subsets was reduced. The poorest correlation among variances was above 0.60, however, until only five lines were included in the subset. It can be seen from Figure 1 that sets with five, or more, lines are nearly as effective in representing the variance of each particular cross combination as the full set of 10 lines. By reducing the number of lines in each family, it is possible to evaluate more families, and potentially capture a greater amount of variability in the material being generated by the RIPE system.

Comparison of Levels

(a) Yield

The average yield (grid adjusted to check plot yields) of the F3:4 Elite lines tested in 2000 at the Elora Research Station was 91 % of the mean of the yield checks (Figure 2). The best Elite line (data not shown) was 158% of the checks and the poorest line 37%. The best Elite family yielded 110% of the checks and the poorest family 76% (Figure 2 and Figure 3a). The average of the best line in each family (data not shown) was 110%. The Elite level is currently a source for material performing well in replicated trials across several locations in the province.

The average yield of families at the High level was 86% of the yield checks (Figure 2) with a range of lines having 134-35% of the check means (data not shown). The best family at the High level was 106% and the worst family was 69% of the checks (Figure 2 and Figure 3b). The High lines are generally nearly as good in yield as the Elite lines, but often have poorer agronomic traits and/or seed quality, and less desirable disease resistance 'packages'. The additional dose of Elite germplasm introgressed at the next step appears to give better stability over years and locations.

Lines at the Intermediate level averaged 96% of the check yields, with the highest line being 185% of the checks. The best Intermediate family was 120% of the check yields and the poorest family was 72% (Figure 2 and Figure 3c) The highest yielding line at the Intermediate level was 138% of the checks with the lowest yielding line being 43%. A higher proportion of the Intermediate families were above the mean of the checks than in the other populations.

Across the levels, 8% of the Elite families had mean yields equal or greater than the checks, 6% of the High families were equal or greater than the checks, and 27% of the Intermediate families were equal or greater than the checks (Figure 3a, b, c). Many of the lines at the Intermediate level (29%) had complete resistance to both powdery mildew and leaf rust, which gave them a decided yield advantage over the checks, and also over the higher levels of the RIPE system. These results help to emphasize the significance of disease resistance and adaptation in realizing the full yield potential of lines. It is also possible that 75% adapted germplasm, the level of the Intermediate population, is sufficient for the potential of new, exotic genes to be expressed. The experience of the second author is that it generally takes about three backcrosses (87.5% recurrent germplasm) to have a reasonable expectation of producing lines with variety potential when introducing exotic material. This is similar to the conclusions of Peel and Rasmusson (2000) in their introgression of two-rowed material into six-rowed backgrounds.

(b) Variance

The average variances for yield were relatively high in all populations tested, with the Intermediate level being the highest at 398, followed by the Elite and then the High population (Figure 4). It would be expected that the Elite should have the lowest variance as it would contain the least exotic germplasm and the highest proportion adapted background genes. However, continued recombination at the Elite level appears to be maintaining a considerable level of variability for yielding ability, and therefore, potential for further progress and response to selection. Some families in each level were highly variable, as indicated by the high maximum variances, with the Intermediate level having several families with very high variances. There were also some families which were quite uniform in yield with very low variances, with the lowest one being at the Intermediate level. This would indicate that the Elite population generally has reasonable levels of variability in all families, even though it is closest to a closed population and the furthest from the introduction of new germplasm The injection of only about 6% of new genetic material each cycle, along with the frequent recombination, seems to be quite adequate for maintaining significant variation which should lead to further progress for yield at the Elite level. By bringing some new genes for agronomic type and disease resistance into the Elite population with each cycle, it should continue to be a source of new, improved gene combinations, and higher yielding lines.

Figure 1. Correlation of subsets of lines with the original population of ten lines per family at the Elite level for yield

Figure 2. The mean relative yields of families at all three levels of the RIPE system.

Figure 3. Mean relative yields of families in each level of the RIPE system.

Figure 4. Mean variances for yield of families in each level of the RIPE system.


Although there were some differences in average yields and variances among the levels, there is no consistent pattern. The Intermediate level seems to represent the highest potential for selection for yield, at least in the presence of disease pressure, as it has highest mean yields, and the greatest variation within most families. This would suggest that there may be no need to go beyond 75% Elite background to re-capture the necessary adaptation to produce some high yielding gene combinations with the exotic material coming in from the introductions. There are some improvements in agronomic type and combinations of disease resistance likely to occur as material is advanced to the Elite level and subjected to further cycles of selection and recombination. A more in-depth comparison of the Intermediate with the higher levels to determine if there is significant benefit in further recombination warrants further investigation.


Kannenberg, L. W. 1981. Activation and deployment of genetic resources in a maize breeding program. Pages 393-399 in G. G. E. Scudder and J. L. Reveal, eds. Evolution today. Proc. Of the 2nd Inter. Congr. of Systematic and Evolutionary Biology. Hunt Institute for Botanical Documentation, Carnegie-Mellon University, Pitsburgh, PA.

Kannenberg, L.W. 2001. HOPE, a hierarchical, open-ended system for broadening the breeding base of maize. pp311-329. In: Broadening the Genetic Base of Crop Production, Ed Cooper, Spillane and Hodgkin. CABI Publishing, New York.

Kannenberg, L. W. and Falk, D. E. 1995. Models for activation of plant genetic resources of crop breeding programs. Can. J. Plant Sci. 75: 45-53.

Peel, M.D. and D.C. Rasmusson. 2000. Improvement strategy for mature plant breeding programs. Crop Sci 40:1241-1246.

Previous PageTop Of PageNext Page