Selection of Rice Varieties for Recommendation in Sri Lanka: A Complex-free Approach

Transcription

1 World Journal of Agricultural Sciences 6 (): , 010 ISSN IDOSI Publications, 010 Selection of Rice Varieties for Recommendation in Sri Lanka: A Complex-free Approach 1 3 S. Samita, M. Anpuas and D.S. De. Z. Abeysiriwardena 1 Faculty of Agriculture, University of Peradeniya, Sri Lanaka College of Graduate Study, University of British Colombia-Okanagan, Canada 3 Rice Research and Development Institute, Batalagoda, Sri Lanka Abstract: Performance of a good crop variety should be consistent across locations and over different seasons. This is usually measured by e adaptability parameters and is normally being set as goals in all crop breeding programmes. Numerous meods / techniques have been proposed to evaluate adaptability of crop varieties. Most of em are very complex, difficult to understand by an average crop scientist / breeder and cannot be implemented rough standard statistical packages. Therefore use of ese meodologies is limited and conclusions are error bound. The Recently proposed meod of relative superiority overcomes many of ese shortcomings. The advantage of is meod is at it uses only one parameter to select varieties for recommendation and analysis can be performed rough standard statistical software packages. A paddy data set generated on 3, 3½ and 4½ mon maturity group varieties tested over several locations and seasons were used to illustrate e value of is meod. The meod could easily detect adaptable varieties in e test so at meod was simple and straightforward. Key words:breeding programmes Environmental index Interaction Relative stability Relative superiority varieties / genotype INTRODUCTION e environmental index used in corresponding regression models is strictly not an independent variable. Bo Stability parameters based on regression models are dependent and regress variables in e model are widely used. They were first proposed by Yates and erefore random variables, us undermining e validity Cochran [1]. Finlay and Wilkinson [] later proposed e of confidence intervals and hypoesis tests of model regression of observed genotypic response values on parameters. environmental index defined as e difference between e The stability variance parameter for each genotype marginal mean of e environment and e overall mean. estimated by Shukla [4] is not statistically independent Eberhart and Russel [3] proposed a second parameter, and it can be negative as it is a difference of two sums of basically e deviation mean squares from e regression squares. As e distribution of is parameter is not clear, model of Finlay and Wilkinson, which can be used in test of homogeneity of e estimates are approximate. conjunction wi e regression coefficient. They Recently, a few new approaches had also been regarded a genotype to be stable if its regression proposed such as multivariate meods for analysis of coefficient was close to one and e deviation mean genotype-by environment data. These include e biplot square was small. The assumption here is at e [5], spatial analysis [6], e additive main effects and genotypes wi coefficients significantly greater an one multiplicative interaction (AMMI) models [7] and a few would be adapted to more favorable growing conditions oers. Eskridge [8] proposed e use of a safety first rule whereas ose wi coefficients significantly less an for selection of superior genotypes. His approach is one would be adapted to less favorable environments. subject to e choice of risk level of obtaining Some difficulties are realized in e use of regression disastrously low yields as well as choice of a stability parameters to identify stable high-yielding genotypes. measure. However, most of ese techniques consist of Heterogeneous variances result in different precisions of fairly complicated analysis, difficult to understand by an e estimates of regression coefficients, b i, us agronomist and cannot be performed by standard complicating among-genotype comparisons. Furermore, software available. Corresponding Auor: M. Anpuas,College of Graduate Study, University of British Colombia-Okanagan, Canada 189

2 Kamidi [9] proposed a simple technique for varietal Specific Stability: The specific stability (hencefor selection. The uniqueness of is meod is at e refered to as stability) is defined as e correlation meodology can be used easily wi e help of standard between genotype and environmental index. The statistical software and is easy to understand. The correlation coefficient ( ) significantly different from zero objective of e present study was to illustrate e value merely signifies e presence of some association of is meod by applying it in evaluating adaptability of between two variables. This association has to be rice varieties in different maturity groups tested in sufficiently strong for a stable genotype. If = 1 en it is multi-locational yield trials over two seasons in Sri Lanka. regarded at e variety is stable. Thus in testing e stability it is necessary to determine wheer e estimate MATERIALS AND METHODS of e correlation (r ge) actually represents = 1. The test is en H o: = 1 versus H A: < 1. Depending on e The meod proposed by Kamidi [9] is based on e outcome, varieties are classified. If e is not regression model of e form, being significantly different from unity at (i.e. P > 0.05) e genotype is regarded as very stable. Similarly, if yij = i + ixij + dij (1) 0.01 < P < 0.05, e variety is considered as sufficiently stable, if < P < 0.01 e variety is considered as where: fairly stable and if P < 0.001, e variety is considered as unstable. yij-yield of e i genotype at e j environment, i -i genotype mean, Testing wi H o: = 1 versus H A: < 1 is fairly i -regression coefficient, complicated. The test used in is circumstance is e test dij-deviation from regression, proposed by Gayen [10]. However, one can use e critical xij - environmental index for e i genotype at e j values published by Kamidi [9] and make inferences environment. wiout actually performing e test. The environmental index, x ij gy j y = g 1 where: y j -marginal mean of e j g -number of genotypes ij environment Based on model (1) e ree parameters used for varital recommendation are specific stability, relative performance and relative superiority. () Relative Performance: The relative performance (hencefor refered to as performance) of e i variety (p) i is defined as b i-1, where biis e estimated regression coefficient (measure of response across environments) from model (1) i.e., by how much its response lies above or below e average (b = 1). Relative Superiority: Relative superiority (hencefor refered to as superiority) of e i variety (s i) is measured as a product of performance and stability i.e.(s i = p i r ge). This parameter is taken as e measure for selecting stable high yielding varieties. Table 1: The varieties tested in each maturity group in each season and location Season Maturity group Varieties tested Tested locations Wet (001/0) 3 mon Bg845, Ld98-3, Bg300, Ambalantota, Ampara, Arulaganwila, Bg834, At303 and Bg305 Batalagoda, Bombuwela and Vanarumollai 3½ mons Bg879, Bg780, Bw38-1, Bg835, Bg300, Bw , Ambalantota, Ampara, Arulaganwila, Bg , Bg357 and Bg359 Batalagoda,Bombuwela and Vanarumollai 4 mons Bg949-1, Bg379-, Bg937-, Ambalantota, Arulaganwila Bg403, Bg893 and Bg450 Batalagoda and Bombuwela Dry (00) 3 mon Bg845, Ld98-3, Bg300, Ambalantota, Batalagoda,Bombuwela, Bg834, At303 and Bg305 Labuduwa, Maha-Illupallama and Vanarumollai 3½ mons Bg880, Bw1059, Bw38-, Ambalantota, Ampara, Arulaganwila, Bg781, Bg358, Bg301, Batalagoda,Bombuwela, Labuduwa, Bg836, Bw1047 and Bg360 Maha-Illupallama and Vanarumollai 4 mons Bg949-, Bg379-, Bg937-, Ambalantota, Arulaganwila, Bg403, Bg893 and Bg450 Batalagoda and Labuduwa 190

3 Yield (grain yield) data of ree sets of rice varieties All regression models, except for 4 mon maturity from ree different maturity groups (3, 3½ and 4 mon) group, were significant (P < 0.06) wi varying coefficient 1 tested over several locations in wet (Maha) and dry of determination (R ) values. All e regression models (Yala) seasons were obtained from e coordinating rice related to 3 and 3½ mon maturity group of wet season varietal programme using randomized complete block had R above 65% while it was above 90% for 4 mon design conducted by Rice Research and Development maturity group of dry season. Among e regression Institute of Sri Lanka and used to illustrate e proposed models fitted for dry season, 3 and 4 mon maturity meodology. The varieties used under each maturity groups had e minimum R value of 85% while 3½ mon group for each season and location are presented in maturity group had 55% as its minimum value. Table 1. Note at each variety was replicated four times Among e varieties under 3 mon maturity group in respective locations. tested in e wet season, Bg305, At303 and Bg834 The statistical analysis was carried out using SAS obtained e first ree places, respectively based on Version 8.. A sample SAS program at can be used to superiority along wi a high stability level (Table 4) while carry out e analysis for ree mon maturity group for Bg845, Bg834 and Bg300 obtained e first ree places 001/0 wet season is given in e appendix. in e dry season (Table 5). However, e variety Bg834 was sufficiently stable while oer two were highly stable. RESULTS AND DISCUSSION The varieties Bg305 and Ld98-3 gave highest yield in e dry season wi a high stability but ranked at e bottom Mean yield and corresponding adaptability because of eir poor performance. parameters estimated by fitting model (1) for varieties in In e case of 3½ mon maturity group tested in e ree maturity groups in e wet and dry seasons are wet season, alough varieties had varying levels of presented in Tables 4 and 5 respectively. Fitting model (1) stability, e varieties at occupied e first ree places is in fact performing simple linear regression for each were very stable. The varieties Bw38-1, Bg879 and variety separately using yield of e i variety (y ij) as e Bg780 obtained first ree places respectively based on response variable and environmental index for e i superiority. Alough e varieties Bw , Bg357and variety (x ij) as e explanatory variable. The x ij were Bg recorded high grain yields ey were low in computed as defined in equation (). For superiority (Table 4). instance, e, y j which are required to compute x ij for The variety Bw1047 of 3½ mon maturity group e 3 mon maturity group in wet 001/0 season can be computed as shown in Table. Accordingly, for instance, environmental index for Bg845 at Ambalantota, x 11, can be computed as, x = [(6)(5.76)-(4.71)] /5 = 5.97 (Table 3). 11 cultivated in e dry season appeared unstable while only variety Bg360 was highly stable. The varieties Bg880, Bg358 and Bw1059 of 3½ mon maturity group obtained e first ree places based on superiority along wi Table : Mean yield for 6 varieties at 6 locations (ree mon maturity group, wet 001/ 0 season) Location Bg845 Bg834 At303 Bg305 Ld98-3 Bg300 mean ( y j) Ambalantota Ampara Arulaganwila Batalagoda Bombuwela Vanarumollai Table 3: Environmental index for 6 varieties at 6 locations (ree mon maturity group, wet 001/ 0 season) Location Bg845 Bg834 At303 Bg305 Ld98-3 Bg300 Ambalantota Ampara Arulaganwila Batalagoda Bombuwela Vanarumollai Rainfall in Sri Lanka is distributed bimodally. The period from September to February in e following year during which Nor-East monsoon brings e rain is called wet (Maha) season. The period March to August during which 191 Sou-West monsoon brings e rain is called dry (Yala) season.

4 Table 4: Mean yield and corresponding adaptability parameters estimated by fitting model (1) for varieties in ree maturity groups grown in 001/0 wet season Variety Yield (t/ha) Specific Stability (r ) Regression coefficient (b ) Relative performance (p = b -1) Relative superiority (s = p r) 3 mon ge i i i i i* ge Bg *** At *** Bg *** Bg *** Ld ** Bg * ½ mons Bw *** Bg *** Bg *** Bg ** Bg * Bg *** Bw ** Bg * Bg ** mons Bg *** Bg *** Bg *** Bg *** Bg *** Bg *** *, **,*** r ge not significantly different from one (P > 0.001, P > 0.01 and P > 0.05, respectively) Table 5: Mean grain yield and corresponding adaptability parameters estimated by fitting model (1) for varieties in ree maturity groups grown in dry 00 season Variety Yield (t/ha) Specific Stability (r ge) Regression coefficient (b i) Relative performance (p i = bi -1) Relative superiority s i= rge* pi 3 mon Bg *** Bg ** Bg *** At ** Ld *** Bg *** ½ mons Bg * Bg * Bw * Bg ** Bg ** Bg * Bw * Bw a Bg *** mons Bg *** Bg *** Bg *** Bg *** Bg *** Bg *** *, **,*** r ge not significantly different from one (P > 0.001, P > 0.01 and P > 0.05, respectively) a-r ge significantly different from one 19

5 e satisfactory stability levels in dry season (Table 5). In packages. Alough e meod is simple and easy to general, stability level of 3½ mon maturity group in e implement it can effectively select varieties wi high dry season appeared low so at it has to be improved in adaptability. e future. All e varieties in e 4 mon maturity group were REFERENCE found to be highly stable regardless of e season. Bg403, Bg450 and Bg893 received e first ree places, 1. Yates, F. and W.G. Cochran, The analysis respectively in e wet season (Table 4) while Bg379-, of groups of experiments. J. Agricultural Sci., Bg403 and Bg949- received e first ree places, 8: respectively in e dry season (Table 5).. Finlay, K.W. and G.N. Wilkinson, The analysis Among e maturity groups, 3 mon maturity group of adaptation in a plant breeding programme. varieties had e highest grain yield (5.4 t/ha) and Australian J. Agricultural Res., 14: mon maturity group had e lowest yield (4.53 t/ha) 3. Eberhart, S.A. and W.A. Russel, Stability during e wet season while it was reversed in e dry parameters for comparing varieties. Crop Sci., 6: season. This is a different issue which needs furer studies. 4. Shukla, G.K., 197. Some statistical aspects of Value of e meod illustrated in is article is at it portioning genotype-environmental components of uses only one parameter to select varieties for variance. Heredity, 9: recommendation. Half of e varieties which have been 5. Kempton, R.A., The use of biplots in ranked wiin first ree places under all ree maturity interpreting variety by environment interactions. J. groups in bo seasons based on superiority are already Agriculture Sci., 103: recommended varieties owing to eir superior 6. Westcott, B., A meod of assessing e yield performances consistently over locations and stability of crop genotypes. J. Agricultural Sci., 108: seasons. This shows e validity of is meod Furermore, present example shows at e meod is 7. Zobel, R.W., M.J. Wright and J.G. Gauch Jr, easy to use, practically highly feasible, technically sound Statistical analysis of a yield trail. Agronomy J., and useful. 80: Eskridge, K.M., Selection of stable cultivars CONCLUSION using a safety first rule. J. Crop Science 30: Kamidi, R.E., 001. Relative stability, performance and The uniqueness of is meod is at only one superiority of crop genotypes across environments. parameter (relative superiority) is finally used to evaluate J. Agricultural, Biological and Environmental e varieties. The parameter is based on stability and Statistics, 6: performance and us e approach used in e meod is 10. Gayen, K.A., The frequency distribution of e very informative. The main advantage of is meod is product-moment correlation coefficient in random at an average agronomist/ plant breeder can use is samples of any size drawn from non normal meod using available standard statistical software Universes. Biometrika, 38:

6 APPENDIX DATA vrec; INPUT loc$ Bg845 Bg839 At303 Bg305 Ld983 Bg300 lmean; ei1=(6*lmean-bg845) / 5; /* loc is to define e variable location*/ ei=(6*lmean-bg839) / 5; /*lmean is e location mean*/ ei3=(6*lmean-at303) / 5; /* ei1 to ei6 are e environmental indices for 6 varieties*/ ei4=(6*lmean-bg305) / 5; ei5=(6*lmean-ld983) / 5; ei6=(6*lmean-bg300) / 5; CARDS; Ambalantota Ampara Arulaganwila Batalagoda Bombuwela Vanarumollai RUN; PROC REG; MODEL Bg845=ei1; MODEL bg839=ei; MODEL at303=ei3; MODEL bg305=ei4; MODEL Ld983=ei5; MODEL bg300=ei6; RUN; PROC CORR; VAR Bg845 ei1; RUN; PROC CORR; VAR Bg839 ei; RUN; PROC CORR; VAR At303 ei3; RUN; PROC CORR; VAR Bg305 ei4; RUN; PROC CORR; VAR Ld983 ei5; RUN; PROC CORR; VAR Bg300 ei6; RUN; 194