Correlation, path coefficient analysis and heritability of grain yield components in pearl millet (Pennisetum glaucum (L.) R. Br.) parental lines

Twenty four parental lines of pearl millet A/B pairs developed jointly by Lake Chad Research Institute (LCRI), Maiduguri and International Crops Research Institute for the Semi-Arid Tropics (ICRISAT), Kano during 1997 to 1999 were evaluated along with a seed parent (ZATIB) across five locations to determine yield and yield component relationships, heritability estimates as well as genetic advance. Correlation coefficient analysis showed that stand count (r=0.249), plant height (r=0.435) and head weight (r=0.958) significantly (p<0.05) and positively correlated with grain yield while days to 50% flowering significantly but negatively correlated (r=-0.539) with grain yield. There were negative but none significant correlation between grain yield with downy mildew score (r=-0.100) and Striga count (r=-0.095) while downy mildew score and Striga count negatively correlated with stand count (r=-0.155 and r=-0.065 respectively). Head weight has high positive and significant environmental, genotypic and phenotypic correlation coefficient with grain yield (re=0.920; rg=0.900 and rp=0.980). Positive and significant genotypic and phenotypic correlation coefficient exists between plant height and grain yield (rg=0.593 and rp=0.417). Path coefficient analysis indicated that stand count had strong positive direct effect (0.970) on grain yield followed by plant height (0.953). Head weight expressed high negative direct effect (-0.846) on grain yield. The parental lines under study showed high to moderate broad-sense heritability; with panicle length expressing the highest heritability (78.95%), followed by grain yield (75.43%) and head weight (73.30%). The rest characters expressed moderate heritability values. Panicle length and head weight displayed high heritability and high genetic advance indicating that the two traits might be controlled by additive gene effects. This suggests that selection in the segregating generation may be effective. Phenotypic correlation approximates genotypic correlation coefficient in this study, indicating that the influence of environment may be probably minimal and traits with high predictive values can be selected early in the breeding program as against traits with low predictive values.


INTRODUCTION
Pearl millet is an important staple food security crop in Nigeria grown in 5.2 million hectares with a production of 4.62 million tones grain per year.It occupies about 32% of total area planted under cereals, and account for about 26% of total cereals production in Nigeria (Ndjeunga et al., 2010).Low productivity of pearl millet across all the millet growing belts in Nigeria is due to the cultivation of open pollinated varieties (OPVs) by farmers coupled with adverse biotic and abiotic stresses.It has been observed that single hybrid generally gives 20 to 30% more grain yield than OPVs (Rai et al., 2006).With the increasing population and rapid deployment of pearl millet into feed and instant value added products, significant increase in per hectare yield of the crop is required to meet the ever increasing demand, which can be made possible with the use of hybrids.Based on the availability of a commercially exploitable cytoplasmic-nuclear malesterility system LCRI, Maiduguri along with ICRISAT embarked on pioneer research of developing commercial pearl millet hybrids using indigenous germplasm and converted breeding lines (Ezeaku and Angarawai, 2005).Out of the large pool of parental lines developed, 30 male sterile lines (A-lines) and their maintainer (B-lines) were selected based on uniformity, stable sterility and other characters such as seed set, exertion and vigor.These traits were evaluated based only on visual observation.As seed parents required for the production of millet hybrids, studies on the character association and heritability of the A/B lines are the first most significant step in embarking on single cross hybrid program.Estimation of correlation, path coefficient analysis, heritability and genetic advance would be useful in developing appropriate breeding and selection strategies.Therefore, understanding the yield and yield components relationship as well as heritability estimate of hybrid parental lines is essential in determining traits that contributes significantly to yield, facilitate their selection and utilization in hybrid development.Grain yield is a complex quantitative trait and is polygenetically controlled.Therefore, selection on the basis of grain yield alone is usually not effective.However, selection based on its components and secondary characters could be more efficient and reliable (Govindaraji et al., 2011).The purpose of this study was to gain sufficient knowledge of the interrelationship, path coefficient between yield and its components, heritability and genetic advance among pearl millet parental lines to determine criteria for selection that could be effectively used to identify the desirable lines with potential for high yield in single cross hybrid development program.

MATERIALS AND METHODS
A set of twelve A-lines and their maintainers (B-lines) developed by LCRI, Maiduguri and ICRISAT, Kano were evaluated in five locations namely, Minjibir, Bagauda, Zaria, Panda and Babura in 2000 during wet season in a randomized complete block design with four replications.The experimental unit was a four-row plot of 5 m long, spaced at 0.75 m apart and intra row spacing of 0.5 m.Inorganic fertilizer (NPK 15:15:15) was applied as a basal dose at the rate of 300 kg per hectare.Crops were thinned out to two plants per stand count two weeks after crop emergence.The crop was top dressed with 100 kg urea per hectare after three weeks of post crop emergence.
Data was taken from two middle rows for stand count, days to 50% flowering, downy mildew score (recorded following a 1-6 damage rating scale, where 1 = no symptom, 2 = 1-5% infected plants, 3 = 6-10% infected plants, 4 = 11-20% infected plants, 5 = 21-40% infected plants and 6 = > 40% infected plants), Striga count, plant height (cm), panicle length (cm), head weight (kg ha -1 ) and grain yield (kg ha -1 ) following the recommendation of International Board for Plant Genetic Resources (IBPGR) and ICRISAT descriptor list for pearl millet (Anonymous, 1993).Correlation coefficient was computed from variance and covariance components as suggested by Burton (1952), Wright (1960 and1968) and Narasimharao and Rachie (1964).The correlation coefficient was partitioned into direct and indirect effects according to Dewey and LU (1959) and Turner and Stevens (1959).The genotypic, phenotypic, environmental correlation between yield and its components among themselves and genetic advance were worked out as per the methods suggested by Johnson et al. (1955) while heritability in broad sense was calculated according to the procedure described by Singh and Chaudhary (1977).All the data were analyzed using GENSTAT, 2009 edition.

RESULTS AND DISCUSSION
The correlation between pairs of variables sampled combined over five environments are presented in Table 1.Result showed that stand count (r=0.249),plant height (r=0.435) and head weight (r=0.958)significantly and positively correlated with grain yield at the 0.05 and 0.01 levels of probability while days to 50% flowering significantly but negatively correlated (r= -0.539) with grain yield.Several previous workers (Atif et al., 2012;Singh and Govila, 1989;Bidinger et al., 1993;Jindla and Gill, 1984) also found similar results in pearl millet.The negative but significant correlation of days to 50% flowering with grain yield shows that parental lines with shorter days to flowering tend to produce more grain yield and vice-versa.This negative correlation indicates that it is not possible to improve both traits simultaneously depending on the intensity of linkage or the degree of tradeoff between the two traits.Similarly, Tables 3 and 4 equally shows that the two traits exhibited negative and significant genotypic correlation coefficient (rg= -0.532) and phenotypic correlation coefficient (rp= -0.359).Although, some of the traits that exercise negative correlation with one another will be difficult to select for in characterization of desirable traits, those with negative association but none significant correlation will be disregarded in selection for crop or variety improvement (Ariyo et al., 1987;Henry and Krishna, 1990;Newall and Eberhart, 1961).There were negative but non significant *Corresponding author.E-mail: idowuezeaku@yahoo.com.Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License    (r e =0.920; r g =0.900 and r p =0.980, respectively).Similarly, positive but significant genotypic and phenotypic correlation coefficient exists between plant height and grain yield (r g =0.593 and r p =0.417, respectively).The result suggests that these two traits, are less influenced by the environment and they could be improved in diverse environments.This finding is in agreement with Ezeaku and Mohammed (2006) and Kumari et al. (2013).The positive but highly significant correlation coefficient between stand count and grain yield for genotypic (rg=0.904)and phenotypic (rp=0.609)correlation coefficient indicates that optimum plant population generally promotes higher grain yield despite other environmental variables.Downy mildew reduced plant height, head weight and grain yield since it correlated negatively with these traits for environmental, genotypic and phenotypic correlation coefficient.Ezeaku and Angarawai (2005) found downy mildew to adversely affect these traits.Negative environmental, genotypic and phenotypic correlation coefficient between Striga count with plant height, grain yield, panicle length, and head weight revealed that Striga attacks as expected would reduced plant height, grain yield, panicle length and head weight.The positive but significant genotypic correlation coefficient between head weight and plant height (rg=0.573)and their very low environmental correlation coefficient (re=0.018)indicates that selection for grain yield based upon the phenotypic performance of these characters alone may not be effective.
The significant genotypic correlation coefficient between head weight and plant height and between head weight and panicle length indicates that these two characters are independent of one another and they could be selected separately as they are components of grain yield.Both traits also influenced grain yield significantly and positively in this study.This shows that taller plants and longer panicles possess heavier head weight and greater grain yield to some extent than shorter plants, probably due to greater mobilization of assimilates to the panicle in taller plants.This result is in agreement with Gupta and Sidhy (1972) and Ezeaku and Mohammed (2006).
When large numbers of variables are included in a correlation study the association among themselves will be very complex.Thus path analysis is necessary to elucidate the true direct and indirect relationship among such characters.In this study path analysis was used to examine the relationship between grain yield and its components.Path coefficient analysis showing direct and indirect effects of yield and yield components are presented in Table 5. Stand count had strong positive direct effect (0.970) on grain yield followed by plant height (0.953).The high positive direct effect of stand count and plant height on grain yield is indicative of their important role in influencing grain yield.However, the negative indirect effects of panicle length and head weight on grain yield through stand count and plant height suggests the effect of downy mildew and Striga on these traits which is equally corroborated by the negative environmental genotypic and phenotypic correlation coefficient between downy mildew and Striga count on panicle length and head weight.Panicle length had negative direct effect (-0.214) on grain yield.Also, head weight expressed high negative direct effect (-0.846) on grain yield.This shows that increasing panicle length through selection may not necessarily lead to proportionate increase in grain yield.The disparity between Tables 1 to 4 which consistently indicated high positive correlation between head weight and grain yield and Table 5 which revealed high negative direct effect of head weight on grain yield justifies the need to clarify the nature of relationship between yield and yield components  using path coefficient analysis.This process has assisted in elucidating the true relationship between head weight and grain yield hence its direct selection will only be effective in improving grain yield in the absence of some biotic factors that affects head weight such as downy mildew and Striga infestation.Residual effect is low (0.125) indicating most of the yield component characters were considered in the present study.Combined correlation coefficient, estimates of heritability and genetic advance as percent of mean are presented in Table 6.Estimates of heritability and their roles in predicting gains in crop species have been reported by Kang et al. (1983), Kole and Saha (2013) and Suthamathi and Dorairaj (1995).The parental lines under study showed high to moderate heritability with panicle length expressing the highest heritability (78.95%), followed by grain yield (75.43%) and head weight (73.30%).The rest characters expressed moderate heritability values.High heritability with positive but highly significant correlation coefficient was observed for head weight (73.30 and 0.900 respectively).Similarly, stand count and plant height expressed moderate heritability with positive and significant correlation coefficient.This finding is in agreement with Govindaraj et al. (2011).The high heritability in broad sense recorded for panicle length, head weight and grain yield indicates that genotype plays a most prominent role than the environment in determining the phenotype suggesting the preponderance of additive gene effects in the inheritance of the traits (Panse, 1957).This showed that phenotypic selection for these traits may likely be effective in hybrid development program.Similar results were also reported by Ghorpade and Metta (1993), Lakshmana andGuggari (2001) andGovindaraj et al. (2011).Phenotypic correlation approximates genotypic correlation coefficient in this study, suggesting that the influence of environment may probably be minimized and traits with high predictive values can be selected early in the breeding program as against traits with low predictive values.In this study panicle length, head weight, grain yield, stand count, days to 50% flowering and plant height exhibited high to moderate heritability estimates suggesting that these traits may be improved upon significantly.
The high genetic advance as percent of mean (> 20%) were recorded for panicle length (20.29) and head weight (20.77).The medium genetic advance as percent of mean (10 to 20%) were recorded for traits such as days to 50% flowering (10.05), plant height (14.87), and grain yield (14.41) while the low genetic advance as percent of mean (<10%) was recorded for number of plant stand (7.05).High heritability and high genetic advance was observed respectively for panicle length (78.95 and 20.29%), and head weight (73.30 and 20.77%).The progress that can be made in advancing mean value of population through selection program will depend on the heritability of the traits under consideration, the phenotypic variation as well as the selection intensity.Therefore, result based only on heritability might not help in identifying traits that are needed to advance selection.Johnson et al. (1955) had also suggested that heritability estimates along with genetic gain is usually more helpful than the heritability alone in predicting the resultant effect from selecting the best individuals.The heritability gives information on the magnitude of the inheritance of traits, while genetic advance aid in formulating suitable selection criteria.Hence, traits that displayed high heritability and high genetic advance such as panicle length and head weight might be controlled by additive gene effects.This indicates that selection in the segregating generation may be effective.

Conclusion
Twenty four parental lines of pearl millet A/B pairs were evaluated along with ZATIB, a seed parent across five locations in northern Nigeria to determine yield and yield component relationships, heritability and genetic advance.The study is useful in developing appropriate hybrid breeding and selection strategies aimed at enhancing the performance of the resulting hybrids as traits that contributes significantly to yield will be found, selected and utilized.
The results revealed that stand count, plant height and head weight expressed positive and significant correlation with grain yield.Head weight had high positive and significant environmental, genotypic and phenotypic correlation coefficient with grain yield.Stand count also had strong positive direct effects with grain yield.The lines showed high to moderate broad-sense heritability; with panicle length expressing the highest heritability (78.95%), followed by grain yield (75.43%) and head weight (73.30%).With panicle length and head weight displaying both high heritability and high genetic advance, selection in the segregating generation may be effective.

Table 1 .
Combined correlation coefficients between yield and yield components in pearl millet parental lines grown across five environments.

components Grain yield Stand count Days to 50 % flowering Downy mildew score Striga count Plant height Panicle length Head weight
*, ** Significant at 5% and 1% probability level respectively.

Table 2 .
Environmental correlation coefficients of grain yield on other parameters.

Table 3 .
Genotypic correlation coefficients of grain yield on other parameters.

Days to 50% flowering Downy mildew score Striga count Plant height Panicle length Head weight Grain yield
correlation between grain yield with downy mildew score (r= -0.100) and Striga count (r= -0.095) while downy mildew score and Striga count negatively correlated with stand count (r= -0.155 and r= -0.065, respectively).This result indicates that these biotic stresses reduced grain yield and this could be due probably to the reduction in plant population.The environmental, genotypic and phenotypic correlation coefficients of grain yield on other parameters are presented in Tables 2, 3 and 4. The correlation coefficient for most of the pairs of characters revealed the presence of strong positive and negative genotypic association between grain yield and other parameters assessed.The result further showed that genotypic correlation coefficients were higher than both the environmental and phenotypic correlation coefficients for

Table 4 .
Phenotypic correlation coefficients of grain yield on other parameters.

Table 5 .
Path coefficient analysis of the direct and indirect effects of the yield components and their genotypic correlation coefficients with grain yield.

Table 6 .
Combined correlation coefficient (r), heritability (broad sense, h 2 bs) and genetic advance (GA) (as per cent of mean) of the yield component characters in pearl millet parental lines ** Significant at 1% levels of significant.