African Crop Science Journal African Crop Science Society ISSN: 1021-9730 EISSN: 2072-6589 Vol. 6, Num. 3, 1998, pp. 323-328

Ashura Luzi-Kihupi

Department of Crop Science and Production, Sokoine University of Agriculture, P. O. Box 3005, Morogoro, Tanzania

(Received 20 September, 1995; accepted 4 January, 1997)

Thirty six rice lines/cultivars were evaluated for yield and other components at two sites within the University farm of Sokoine University of Agriculture, Morogoro, Tanzania. Data generated were used to provide information on relative direct and indirect contribution of various biological components towards yield. Correlation coefficient analysis revealed grain yield per plant to be positively correlated with all the characters except per cent unfilled grains and days to 50% flowering. Results from path analysis identified number of filled grains per panicle, number of panicles per plant and 1000 grain weight to be important characters that influence grain yield. However, number of filled grains per panicle had a significant negative indirect effects through number of panicles per plant and 1000 grain weight. Heritability estimates revealed plant height, number of filled grains per panicle and grain weight to be highly heritable characters. The study suggested that number of filled grains per panicle and grain weight could be used as selection criteria when selecting for increased yield in rice.

Key Words: Correlation coefficient, Oryza sativa, path analysis, yield components

Trente-six lignées/cultivars de riz ont été évaluées pour le rendement et autres composantes en deux sites dans une ferme universitaire, Université d'Agriculture de Sokoïne, Morogoro, Tanzanie. Les données obtenues ont été utilisées pour fournir des informations directes et indirectes relatives à la contribution de diverses composantes biologiques au rendement. L'analyse du coéfficient de corrélation a révélé le rendement en grains par plant à Étre positivement correlé avec tous les caractères, à l'exception du % des grains non remplis et les jours à la floraison. Les résultats de l'analyse de l'allée ont identifié le nombre de grains remplis par panicule, le nombre de panicules par plant et le poids de 1000 grains comme étant les caractères importants pouvant influencer le rendement en grains. Cependant, le nombre de grains remplis par panicule a eu un effet indirect significatif à travers le nombre des panicules et le pois de 1000 grains. Les estimations d'héritabilité ont revélé la hauteur des plants, le nombre de grains remplis par panicule et le pois de grains comme étant des caractères hautement héritables. L'étude a suggéré que le nombre de grains remplis par panicule et le poids de grains pouvaient Étre utilisés comme critère de sélection pendant la sélection pour accro"tre le rendement du riz.

Mots Clés: Coefficient de corrélation, Oryza sativa, analyse d'allée, composantes de rendement

Rice (Oryza sativa L.) is an important crop in Tanzania and is consumed by about 60% of the population (IRRI, 1993). Tanzania ranks second after Madagascar as a major rice producer in Eastern and Southern Africa. In 1995, it was estimated that rice was grown in about 477,900 hectares with a production of 722,900 tons (FAO, 1995). Current average yield is estimated at 1.5-2.1 tons/ha but yields as high as 6 tons/ha have been obtained in irrigated rice projects (Kanyeka et al., 1994).

Attempts have been made since 1966 to develop varieties with higher yield potential for different rice ecologies in Tanzania. Early varietal improvement concentrated primarily on improving popular local varieties by pure line selection, varietal introduction and evaluation and hybridisation between local and introduced varieties.

Yield per hectare is the most important consideration in rice breeding programme, but yield is a complex character in inheritance and may involve several related components. Rice yield is a product of number of panicles per unit area, number of spikelets per panicle, percentage of filled grains and weight of 1000 grains (Yoshida, 1981; De Datta, 1981). It is therefore important to know the factors or traits that influence grain yield directly or indirectly or both, and to determine heritability and genetic advance under selection of those traits so that response to selection can be predicted.

Knowledge of interrelationship of the phenotypic traits among each other and their influence on yield is lacking for rice in Tanzania. There has been no correlation studies of various traits of the available germplasm in the country. Information obtained from such studies will be of importance in obtaining desirable lines to be incorporated in a breeding programme and in selecting suitable lines for subsequent release as new varieties. Knowledge of the correlations that exist between important characters is also needed to facilitate the planning of a more efficient breeding programme.

In early stages of a breeding programme direct estimation of yield which has low heritability is difficult to obtain. Plant breeders commonly select for yield components that indirectly increase yield (Grevois and McNew, 1993). Yield component breeding would be most effective if the components involved were highly heritable and positively correlated (Sidwell et al., 1976).

This study was therefore undertaken to determine phenotypic correlation coefficients among several agronomic traits in rice and to analyse their interrelationships through path coefficient analysis. Another objective was to estimate broad-sense heritability and genetic advance under selection for the selected traits.

Thirty six rice varieties/lines representing a wide diversity in various characters consisting of local and introduced materials from the International Rice Research Institute (IRRI) and the International Institute of Tropical Agriculture (IITA) were used for the study.

The study was carried out in 1988 main cropping season under irrigated and rainfed lowland conditions within the Sokoine University of Agriculture Farm Morogoro, Tanzania. Morogoro is located at an altitude of about 526 metres above sea level and receives mean annual rainfall of about 800 mm.

The experiments were laid out in a Triple Lattice Design using 3.6 x 2m plots with plants spaced 20cm x 20cm apart. Fertilizer was applied at a rate of 100 kg N ha^-1, split three times, at planting, maximum tillering and panicle initiation stages, while phosphorus was applied at planting time at a rate of 50 kg P₂O₅ ha^-1. At one of the sites which has a clay loam soil, the experiment was rainfed, while at the other site which has a sandy clay loam soil, the trial was supplemented with irrigation throughout the growing season.

Data collected included days to 50% flowering, plant height, tiller number, panicle length, panicle number, number of filled grains per panicle, percentage of unfilled grains, grain weight, and grain yield per plant and plot. All measurements were taken in accordance with Gomez (1972). The data collected from each experiment were subjected to analysis of variance using the procedure given by Gomez and Gomez (1976) for a Triple Lattice Design. Combined analysis was done following the method of Cochran and Cox (1957).

From the analysis of variance of individual traits, estimates of error, phenotypic and genotypic variances for different characters were computed by following the procedures described by Kaul (1973). Genetic and phenotypic coefficient of variations were estimated by the formula adopted by Kaul and Kumar (1982), namely, dividing the square root of the genotypic and phenotypic variance, respectively, by the population mean and multiplying by 100. Heritability estimates in the broad sense for the various traits were calculated using the formula proposed by Hanson et al. (1956), while expected genetic advance was estimated by the formula given by Johnson et al. (1955). From the data collected, simple correlation analysis was performed for the pooled data of two sites using Pearson's correlation coefficient.

Path coefficient analysis of data was carried out following the method outlined by Wright (1921) and adopted by Dewey and Lu (1959). Yield was assumed to be influenced by 7 components, namely, (1) number of panicles per plant, (2) panicle length, (3) number of filled grains per panicle, (4) percent unfilled grains, (5) 1000 grain weight, (6) days to 50% flowering, and (7) plant height.

The direct effects were calculated from standardised coefficients using the following formula:

PXiy = bxi + Sxi
      ----------
         Sy

where,

PXiy = direct effect of the independent variable (Xi) on the dependent variable y
bxi = Regression coefficient of Xi
Sxi = standard deviation of Xi
Sy = standard deviation of y
i = 1-7 characters listed above.

The indirect effects were calculated using a set of simultaneous equations as outlined by Dewey and Lu (1959).

The dependence of grain yield on other traits has been reported for many crops (Robinson et al., 1951; Johnson et al., 1955; Dashora et al., 1977; Chandhanamutta and Frey, 1973, etc). In this study, grain yield per plant was positively correlated with all the characters except per cent unfilled grains and days to 50% flowering which had negative but non-significant correlation with grain yield (Table 1). The negative association between yield and percent unfilled grains was expected since percent filled grains per panicle had a highly significant correlation with grain yield. Thus, in order to increase yield, it is important to reduce spikelet sterility or increase spikelet fertility. The highest correlation coefficient with grain yield was recorded with the number of panicles per plant (r = 0.72; P=0.001) indicating the importance of this component in rice. Significant positive correlation between yield and number of panicles per plant was also obtained by Sarathe et al. (1969).

TABLE 1. Simple correlation coefficients (d.f. = 214) of some rice parameters (Pooled data for 2 sites)

Number of filled grains per panicle was positively correlated with panicle length (r = 0.598) indicating that the plants with large panicles tend to have a high number of fertile grains. Similarly, a positive correlation was observed between number of panicles per plant and panicle length. This implies that increasing the number of panicles per plant and panicle length would also effectively increase the number of grains per panicle and thus grain yield per plant.

Information obtained from simple correlation coefficients is useful if it can be augmented by partitioning the correlations into direct and indirect effects. Path coefficient analysis is useful in providing cause - and effect relationships such as between yield and yield components (Sethi and Singh, 1972; Singh et al., 1982).

Results from the path analysis are presented in Table 2. The number of filled grains per panicle, number of panicles per plant and 1000 grain weight influenced yield per plant directly. Simple correlation coefficient analysis also revealed that these traits had positive effect on yield.

TABLE 2. Direct (bold) and Indirect effects of various traits on grain yield

The number of filled grains per panicle was ranked as the most important component which exerted strong positive influence on yield. Grevois and Helms (1992) also observed positive direct effects for panicle number and filled grains per panicle on rice yield. However, they used panicle number and panicle weight as predictor variables while in this study grain weight and filled grain per panicle are used instead of panicle weight. Number of spikelets per panicle was also found to have the highest direct effect on yield by other workers (Saini and Gagneja, 1975).

While the number of filled grains per panicle had a strong positive direct effect on yield, it had significant indirect effects through other traits. One thousand grain weight was found to have negligible effects on yield by simple correlation coefficient analysis, however, path analysis revealed that this trait had positive direct effect. The negative correlation between 1000 grain weight and number of filled grains per panicle indicates that as the number of filled grain increases, seed size decreases.

Path analyses had indicated days to 50% flowering to be of little importance in influencing grain yield due to its low and negligible direct effects on grain yield (Table 2). Simple correlation indicated this trait to have negative, though insignificant, influence on yield. The importance of this trait in the test population is therefore negligible.

Estimates of genotypic and phenotypic coefficient of variation, heritability and genetic advance due to selection are presented in Table 3. Genotypic and phenotypic coefficient of variation values were numerically higher for plant height, number of filled grains per panicle, percent unfilled grains per panicle, and 1000 grain weight while the values for number of panicles per plant, days to 50% flowering and panicle length were relatively low.

TABLE 3. Estimates of some genetic parameters in rice

Heritability estimates were high for plant height (94.9%), number of filled grains per panicle, panicle length and 1000 grain weight. The lowest heritability estimate were obtained for number of panicles per plant (21%) and yield per plant (26.8%). High heritability estimates for plant height and 1000-grain weight were also observed by other workers (Kaul and Bhan, 1974; Kaul and Kumar, 1982).

Number of filled grains per panicle and 1000 grain weight were found to exert strong direct effect on yield as well as being highly heritable traits in this study. These traits also exhibited high expected genetic gain (Table 3). Although number of panicles per plant had strong direct effect on yield, it exhibited low heritability and expected genetic advance estimates and therefore could not be effective as a selection criterion. Effectiveness of indirect selection is enhanced when the secondary character has a higher heritability than that of the primary character (Fehr, 1987).

Since direct selection for yield in early generations of breeding programmes is difficult, estimates of number of filled grains per panicle and 1000 grain weight in early generations would be useful as indirect selection for grain yield in later generations.

The negative correlations between grain weight and number of filled grains per panicle suggests that the breeder should make selection compromises if simultaneous selection for these traits is performed in order to increase yield.

Correlation coefficient analysis revealed a positive relationship between yield and all the characters tested except percent unfilled grains and days to 50% flowering. Path analysis, however, revealed that number of filled grains per panicle, number of panicles per plant and 1000 grain weight had positive direct influence on grain yield. Number of filled grains per panicle and 1000 grain weight had high heritability estimates and expected genetic advance and could be used as selection criteria in early generation of the test population. The breeder must, however, pay attention to the negative correlation that exists between the two characters.

Chandhanamutta, R. and Frey, K.J. 1973. Indirect mass selection for grain yield in oat population. Crop Science 13:470-473.

Chochran, W.G. and Cox, G.M. 1957. Experimental Designs. John Wiley and Sons, Inc. New York 611pp.

Dashora, S.L., Rathora, A.K., Tikka, S.B.S. and Sharma, R.K. 1977. Correlation and path coefficient analysis for morphological characters in barley. Indian Journal of Agricultural Science 49:381-385.

De Datta, S.K. 1981. Principles and Practices of Rice Production. John Wiley and Sons, New York. 618pp.

Dewey, D.R. and Lu, K.H. 1959. A correlation and path coefficient analysis of components of crested wheat grass seed production. Agronomy Journal 51:515-519.

Food and Agricultural Organisation (FAO), 1995. Production year book. FAO, Rome. 49:70.

Fehr, W.R. 1987. Principles of Cultivar Development. Macmillan Publishing Company. New York. 525pp.

Gomez, K.A. 1972. Techniques For Field Experiments With Rice. IRRI, Los Banos. 46pp.

Gomez, K.A. and Gomez, A.A. 1976. Statistical Procedure for Agricultural Research, with Special Emphasis on Rice. IRRI, Los Banos, Philippines. 294pp.

Grevois, K.A. and McNew, R.W. 1993. Genetic relationship among and selection for rice yield and yield components. Crop Science 33:249-252.

Hanson, H.L., Robinson, H.F. and Comstock, R.E. 1956. Biometrical studies of yield in segregating populations of Korean Lespodeza, Agronomy Journal 48:268-272.

IRRI, 1993. IRRI Rice Almanac. International Rice Research Institute Manila Philippines. p97-99.

Johnson, H.H., Robinson, H.F. and Comstock, R.E. 1955. Genotypic and phenotypic correlations in soybean and their implications in selection. Agronomy Journal 47:477-483.

Kanyeka, Z.L., Msomba, S.W., Kihupi, A.N. and Penza, M.S.F. 1994. Rice ecosystems in Tanzania: Characterisation and classification. Research and Training Newsletter 9:13-15.

Kaul, M.L.H. 1973. Perfomance, interrelationship, and heritability estimates of certain morphological traits of Oryza sativa L. Journal of Indian Botanical Society 51:286-290.

Kaul, M.L.H. and Bhan, A.K. 1974. Studies on some genetic parameters of rice. Theoritical and Applied Genetics 44:178-183.

Kaul, M.L.H. and Kumar, V. 1982. Variability in rice. Genetica Agraria 36:257-268.

Robinson, H.F., Comstock, R.E. and Harvey, P.H. 1951. Genotypic and Phenotypic correlations in corn and their implication in selection. Agronomy Journal 43:282-287.

Saini, S.S. and Gagneja, M.R. 1975. Interrelationship between yield and some agronomic characters in short statured rice cultivars. Indian Journal of Genetics and Plant Breeding 35:441-445.

Sethi, G.S. and Singh, H.B. 1972. Interrelationship of quantitative traits with grain yield in triticale. Indian Journal of Agriculture Science 42: 281-285.

Sidwell, R.J., Smith, E.L. and McNew, R.W. 1976. Inheritance and interrelationships of grain yield and selected yield related traits in a hard red winter cross. Crop Science 16:650-654.

Singh, M., Singh, D.R. and Singh, R.V. 1982. Character association and path coefficient analysis of deep water rices. International Rice Research Newsletter. International Rice Research Institute, Los Banos, Philippines. 7:12.

Yoshida, S. 1991. Fundamentals of Rice Crop Science. International Rice Research Institute, Los Banos Manila, Philippines. 269pp.

Copyright 1998, African Crop Science Society