Effects of Selection by Yield Components on Grain Yield in Pearl Millet (Pennisetum typhoideum Rich.)
Totok Agung Dwi Haryanto, Tae-Kwon Shon and Tomohiko Yoshida (Faculty of Agriculture, Kyushu University, Fukuoka 812-81, Japan)
Abstract : Selections by yield and yield components were conducted on pearl millet (Pennisetum typhoideum Rich.) population. Sub populations selected for high grain yield, seed weight, panicle weight and productive panicles showed higher mean values than those of selected for low and original population. Realized genetic gains were obtained and heritability values were estimated as 0.74, 0.84. 0.65, and 0.50 for yield, seed weight, panicle weight and productive panicles, respectively. Genetic correlations were calculated as 1.00, 0.89 and 0.75 for between yield and seed weight, panicle weight and productive panicles, respectively. Grain yield were varied as 371~ 590 g m-2. Accordingly, selections by yield components were effective for improving grain yield.
Key words: Genetic correlation, Genetic gain, Grain yield, Heritability, Indirect selection, Pearl millet, Yield components.
Beside as the principal food cereal in semiarid regions of Africa and the Indian Subcontinent and a forage in Australia, Southern Africa, South America and the USA, pearl millet (Pennisetum typhoideum Rich.) has potential as an early-maturing summer grain crop in temperate regions (Anand Kumar and Andrews, 1993; Yoshida and Sumida, 1996).
Pearl millet has high yield potential and responds well to water and soil fertility (Poehlman, 1994). Several improved pearl millet varieties have been developed in West Africa, but the yield gains have been insignificant (Ouendeba et al., 1993, 1995; Rajat De and Gautam, 1987).
Plant breeders commonly select for yield components that indirectly increase yield. Gravois and McNew (1993) reported that panicles per unit area was the single most important component of yield in rice. Significant effect of indirect selection in soybean(Tinius et al., 1993), maize (Burgess and West, 1993), and barley (Van Oosterom and Ceccarelli, 1993) also were reported. Furthermore, effectiveness of recurrent selection procedure for improvement of quantitative traits in both self- and cross- pollinated crops (Reysack et al., 1993), kernel morphology (De Koeyer et al., 1993), stability of yield (Reysack et al., 1993) and grain yield of oat (Klein et al., 1993), seed size of soybean (Tinius et al., 1993), grain yield stability of sorghum (Chisi et al., 1996), and grain yield of maize (Burgess and West, 1993; Keeratinijakal and Lamkey, 1993) have been reported. However, study on indirect selection for yield components on pearl millet grain yield is scarce.
Our previous studies on pearl millet showed that, 1. Seed weight, panicle weight and number of productive panicle were important characters to the grain yield and, therefore, they might be used as selection criterions for high grain yield (Totok et al., 1996); 2. Using recurrent selection, genetic gain and heritability of seedling characters under low temperature were obtained; 3. Selection increased seedling characters and yield components, and shoot and root length could be used as selection criterions for low temperature germination (Totok et al., 1997).
In this study, the effect of indirect selection through yield components to improve grain yield of a short statured pearl millet population will be discussed. Heritability and genetic gain of the characters and their genetic correlations are also estimated.
High heterosis can be obtained by crossing between inbred lines with high combining ability in pearl millet (Poehlman, 1994), though adequate male-sterile and fertility-restorer lines are needed. Superior inbred lines might be derived from the open pollinated populations selected for the characters in this study. High yield populations also could be used for the practical grain production.
Materials and Methods
1. S0 population development
Seeds of pearl millet were sown in a field on August, 1995 to develop S0 population in the experimental field of Kyushu University. The material was originated from early maturing and short statured "ICVM83074" variety of ICRISAT (India) and was later seed increased by open pollinating in several seasons including May and August sowing (Yoshida and Sumida, 1996). Space planting as 15 x 60 cm within and between rows and 0.8 g N, 0.8 g P2O5, and 0.8 g K2O per plant were applied in four rows with 75 plants per row. Grain yield as seed yield per plant and yield components as seed weight per panicle, panicle weight and number of productive panicles per plant of all plants in S0 population were measured. Plants were selected for 10 % the highest and lowest yield and yield components and the plants were harvested separately.
2. Polycrossing of selected plants
Seeds from 9 groups, originated from plants selected for high and low performance and the parent, were sown in paper pots, 5x5x5 cm, filled with fertile soil. After 3 weeks, seedlings were transferred to the plastic pots in a green house on May, 1996. One pot was filled with about 4 kg sifted soil, 0.8 g N, 0.8 g P2O5, and 0.8 g K2O. Two plants per pot and 6 pots per group were arranged. Total of 6 (pots) x 2 plants were involved for each group.
Panicles were bagged before heading, polycrossed among individuals manually within a group and harvested separately in August, 1996 for the seeds of S1 population.
3. Evaluation of S1 population
Amount of 9 populations, namely selected for high and low grain yield, seed weight, panicle weight, productive panicles and the parental populations (OP) were sown in paper pots, grown for 2 weeks and then transplanted to a field on August 18, 1996. Randomized complete block design with 2 replications was applied. One plot consisted of two rows and one row consisted of 10 plants. Space planting was 10x50 cm within and between rows. Amount of 0.8 g N, 0.8 g P2O5, and 0.8 g K2O per plant was fertilized. Average 5 plants per plot were sampled for grain yield and yield components evaluation. Seeds were harvested on November 8, 1996.
4. Statistical analysis
Data obtained in S1 population were subjected for analysis of variance continued by protected LSD test (Stell and Torrie, 1980). Heritability was estimated using formula (Falconer, 1981; Fehr, 1987) as follows:
h2 ＝ （ x1 high - x1 low ）／ （ x0 high - x0 low ）
where, h2, x0 high, x0 low, x1 high and x1 low, were heritability, mean of individuals selected of original population for high and low characters, mean of population derived from mating the selected individuals for high and low characters, respectively.
Realized genetic gain was calculated as: x1 high - x1 low.
Genetic correlations were calculated as (Falconer 1981; Fehr 1987):
rA = （ Crx ix hx ）／（ Rx iy hy ）
where, rA, Crx, Rx, ix, iy, hx and hy, were genetic correlation, amount of improvement in primary character obtained by indirect selection for secondary character (yield components) and by direct selection for primary character (yield), selection intensity for primary and secondary character (as 1.75 for 10% selection of population), square root of heritability for primary and secondary character, respectively. Percentage of improving in average for yield and yield components was measured from ratio of differences between mean value of population selected for character and OP divided by mean value of OP for character multiplied by 100 (%) as:
Percent of improving = （ x1 high - x1 OP ）／（ x1 OP ） x 100(%)
Results and Discussion
Analysis of variance for several characters after one cycle of recurrent selection was presented in Table 1. Significant effect of selection on almost all characters was observed. It revealed that upward and downward selection was effective for the characters. Populations differed in panicle weight, panicle length, seed weight and grain yield but not in plant height and number of productive panicles.
Mean values of populations selected for grain yield and yield components were presented in Table 2.
1. Population performance
(1) Plant height
Among mean values of populations selected for high or low grain yield and yield components, no significant different for plant height was observed. It revealed that the character was least influenced by selection. Furthermore, population might avoid deteriorating effect due to lodging after selection for yield and yield components.
(2) Panicle length
Mean value of panicle length of populations selected for high yield and yield components was higher than that of selected for low or OP except for number of productive panicles. It revealed the selection was effective on the changes of unselected character which have contribution to high yield.
(3) Productive panicles
Number of productive panicles was 2.5 and 1.3 in the population selected for high and low productive panicles, respectively. But no significant difference among populations selected for high, low or OP was detected for productive panicles because of the large sampling error.
(4) Panicle weight
Panicle weight of populations selected for high grain yield and yield components was higher than that of selected for low or the original population, except for productive panicles. Whereas, mean values of populations selected for low were lower than those of OP. Panicle weight increased by average 73% and decreased by average 35% following selection for high and low, respectively.
(5) Seed weight
Seed weight per panicle of populations selected for high yield and yield components was higher compared with that of selected for low or OP. No significant difference between population selected for low value and OP was observed. Seed weight increased by average 159% following selection for high.
(6) Grain yield
Grain yield per plant of populations selected for high yield and yield components was higher than that of selected for low or OP, except for productive panicles. The mean values among populations selected for yield and yield components were not significantly different to each other. No significant difference between mean value of population selected for low and OP was observed. Selection for high yield directly increased by average 186 % of grain yield.
Grain yield improvement by 26.8 % and 55.3% based in row and hill plot, respectively after five cycles of recurrent selection in oat (De Koeyer et al., 1993), a 44% yield increasing after four cycles in random-mated maize population (Burgess and West, 1993), and gain from 17.7 to 44.0 % based in hill plot from four cycles of recurrent selection for oat grain yield has been obtained (Reysack et al., 1993).
Grain yield from 371 ~ 590 g m-2 was obtained. The increase in grain yield of pearl millet revealed that selection procedure practiced in this study was effective for grain yield.
2. Indirect selection effects
Results of selection through yield components for grain yield were presented in Table 2. Grain yield of populations selected for high seed weight and panicle weight was significantly higher than that of selected for low seed weight, panicle weight or productive panicles and OP. Indirect selection through yield components, as seed weight and panicle weight, significantly increased the grain yield, by average 243 % and 158 %, respectively. However, indirect selection through high productive panicles was not effective for the grain yield. Significant effect of indirect selection for seed size on seed composition in soybean (Tinius et al., 1993), for ear height on yield in maize (Burgess and West, 1993), for plant ideotype and heading date on yield of barley under stress (Van Oosterom and Ceccarelli, 1993) were reported. The high increasing levels for grain yield of population selected for yield components, as measured by their percentage increasing to the grain yield of OP, revealed that indirect selection practiced in this study was effective.
3. Estimation of heritability
Realized genetic gain by selection intensity of 10 % for yield and yield components was presented in Table 3. Genetic gain calculated from (high selection- population mean) was 12.80, 7.10, 7.86, and 1.00 for grain yield, seed weight, panicle weight and number of productive panicles, respectively.
Heritability values for yield and yield components were presented in Table 3. The values were 0.74, 0.84, 0.65, and 0.50 for grain yield, seed weight, panicle weight, and number of productive panicles, respectively. The values reflected that there were more chances to obtain higher mean value for yield and yield components by further selection for the characters.
Falconer (1981) mentioned that indirect selection was effective if the secondery characters had higher heritability and closely correlated to the primer character. Hallauer et al. (1988) reviewed many studies in corn about the effectiveness of direct and indirect selections for grain yield. Improvement of yield through yield components was also reported in rice (Gravois and McNew, 1993) and in wheat (Van Sanford and Utomo, 1995).
As seed weight and panicle weight were relatively more important characters which showed increasing in grain yield by their indirect selection with high heritability, they might be the characters having higher chance for further selection. This result suggested the usefullness of indirect selection for secondery characters to improve grain yield in pearl millet.
4. Genetic correlation
Genetic correlation between yield and yield components was presented in Table 4. The values were from 0.75 to 1.00 and significant, revealing genetically close relationships between yield and the characters. Relatively high genetic correlation between seed weight or panicle weight and grain yield implied to the more chance in selecting these characters to improve grain yield.
Concluded, selection improved pearl millet population performance as increasing in mean values of yield and yield components which varied from 73 to 187 % in average. Indirect selection through yield components as practiced in this study clearly increased and could be useful to improve grain yield. As shown in high heritability and high genetic correlation to yield, seed weight and panicle weight might be considered as indirect selection criterions for high grain yield.
Anand Kumar K. and Andrews D.J. 1993. Genetics of quantitative traits in pearl millet: a review. Crop Sci. 33:1-20.
Burgess J.C. and West D.R. 1993. Selection for grain yield following selection for ear height in
maize. Crop Sci.33:679-682.
Chisi M., Bramel-Cox P.J., Witt M.D., Claassen M.M. and Andrews D.J. 1996. Breeding for grain
stability using full-sib family recurrent selection in sorghum. Crop Sci. 36:1083-1087.
De Koeyer D.L., Stuthman D.D., Fulcher R.G. and Pomeranke G.J. 1993. Effects of recurrent
selection for grain yield on oat kernel morphology. Crop Sci. 33:924-928.
Falconer D.S. 1981. Introduction to quantitative genetics. 2nd ed. Longman, New York. 165-207.
Fehr W.R. 1987. Principles of cultivar development. Vol. 1. Theory and technique. Collier
Macmillan Publisher, London. 95-105.
Gravois K.A. and McNew R.W. 1993. Genetic relationships among and selection for rice yield
and yield components. Crop Sci.33:249-252.
Hallauer A.R., Russell W.A. and Lamkey K.R. 1988. Corn breeding. In G.F. Sprague and J.W.
Dudley eds., Corn and corn improvement third edition. ASA Inc., Wisconsin. 469-564.
Keeratinijakal V. and Lamkey K.R. 1993. Responses to reciprocal recurrent selection in BSSS
and BSCB1 maize populations. Crop Sci.33:73-77.
Klein S.J., Smith M.A. and Frey K.J. 1993. Recurrent selection for test weight and grain yield of
oat. Crop Sci.33:744-749.
Ouendeba B., Ejeta G., Nyquist W.E., Hanna W.W. and Kumar A. 1993. Heterosis and combining
ability among African Pearl Millet Landraces. Crop Sci.33:735-739.
Ouendeba B., Ejeta G., Hanna W.W. and Kumar A.K. 1995. Diversity among African Pearl millet
landrace populations. Crop Sci. 35:919-924.
Poehlman J.M. 1994. Breeding field crops. Third Edition. AVI Pub., Connecticut. 542-551.
Rajat De and Gautam R.C. 1987. Management practics to increase and stabilize pearl millet
production in India. In J.R. Witcombe ed., Proceedings of the International Pearl Millet
Workshop. ICRISAT, India. 247-254.
Reysack J.J., Stuthman D.D. and Stucker R.E. 1993. Recurrent selection in oat: Stability of yield
and changes in unselected traits. Crop Sci.33:919-923.
Stell G.W. and Torrie J.H. 1980. Principles and Procedures of Statistics. McGraw-Hill Inc.,
Tinius C.N., Burton J.W. and Charter,Jr T.E. 1993. Recurrent selection for seed size in
soybean:III. Indirect effects on seeds composition. Crop Sci. 33:959-962.
Totok A.D.H. and Yoshida T. 1996. Performance and interrelationship among several characters
of pearl millet (Pennisetum typhoideum Rich.) population. J. Fac. Agr., Kyushu Univ. 41:1-9.
Totok A.D.H., Shon T.K. and Yoshida T. 1997. Genetic gain and Heritability of seedling
characters under low temperature in pearl millet (Pennisetum typhoideum Rich.). Jpn. J. Crop Sci.(in press).
Van Oosterom E.J. and Ceccarelli S. 1993. Indirect selection for grain yield of barley in harsh
Mediterranean environments. Crop Sci.33:1127-1131.
Van Sanford D.A. and Utomo H. 1995. Inheritance of tillering in a winter wheat population. Crop
Yoshida T. and Sumida K. 1996 Mass selection and grain yield of improved population in pearl
millet (Pennisetum typhoideum Rich.). Jpn. J. Crop Sci. 65: 58-62.
Table 1. Mean squares by ANOVA for populations after one cycle of recurrent selection.
Source df Plant Panicle No. prod. Panicle Seed Grain
height length panicles weight weight yield
(cm) (cm) (g) (g) (g plant-1)
Repl. 1 79 26.7* 0.11 20.5* 13.7* 129*
Group 8 218 16.8** 0.39 73.0** 33.1** 176**
Residuals 8 103 1.4 0.17 1.6 1.4 27
*, ** ; significant at 5% and 1% level of probability, respectively.
Table 2. Mean values for grain yield and yield components of populations
after one cycle of recurrent selection.
Populations Plant Panicle No. prod. Panicle Seed Grain
height length panicles weight weight yield
(cm) (cm) (g) (g pcl-1) (g plant-1)
Selected for high yield 105 a 20.4 a 2.2 a 18.0 b 10.7 b 24.7 a
Sel. for high seed weight 108 a 22.9 a 1.9 a 22.4 a 15.4 a 29.5 a
Sel. for high panicle weight 108 a 21.0 a 1.7 a 20.7 ab 12.9 ab 22.2 a
Sel. for high prod. panicle 95 a 17.1 b 2.5 a 14.1 c 7.5 c 18.5 ab
Selected for low yield 89 a 16.1 bc 1.4 a 7.8 d 5.0 cd 7.0 bc
Sel. for low seed weight 86 a 14.2 c 1.4 a 6.8 d 4.5 d 6.2 c
Sel. for low panicle weight 90 a 15.3 bc 1.2 a 7.8 d 4.6 d 5.5 c
Sel. for low prod. panicle 78 a 16.7 bc 1.3 a 8.2 d 5.1 d 6.7 bc
Original population 93 a 17.4 b 1.5 a 11.9 c 5.9 cd 8.6 bc
c.v. (%) 10.7 6.7 24.5 9.8 14.7 36.1
Means followed by the same letter within a column are not significantly different at 5% level
according to ANOVA protected LSD Test.
Table 3. Realized genetic gain (G) and heritability (h2) values for grain yield and yield
components estimated after one cycle of recurrent selections.
Characters G (x1sel-x1pop) h2
Grain yield 12.8 0.74
Seed weight 7.1 0.84
Panicle weight 7.9 0.65
Prod. panicles 1.0 0.50
x1sel, x1pop: means of selected subpopulation and original population in S1.
Table 4. Genetic correlation between grain yield and yield components.
Grain yield vs. Genetic correlation
Seed weight 1.00*
Panicle weight 0.89*
Prod. panicles 0.75*
* ; significant at 5% level of probability.