An analysis of the effectiveness of selection in early generation trials for grain yield of wheat (Triticum aestivum L. em. Thell.)

Abstract

Early generation testing has been used as a method forndiscriminating objectively among large populations ofnentries for grain yield. The objective of this thesis wasnto evaluate the effectiveness of such procedures for thenidentification of entries for which further testing wasnjustified.A population of entries was evaluated in unreplicated trialsnat three sites (Toobeah, Biloela and Pirrinuan) in 1978, andnin replicated trials at five sites (Moonie, Tummaville,nBilla Billa, Muckadilla and Bongeen) in 1982/83. Fornindividual trials, coefficients of variation ranged from 7nto 17 percent in 1978 and from 7 to 15 percent in 1982/83.nGenetic coefficients of variation ranged from 14 to 22npercent in 1978 and from 9 to 21 percent in 1982/83. Thenvariance component estimate for entry x site interactionsnwas 1.3 times greater than that for entries in 1978 and 2.0ntimes greater in 1982/83. The pairwise correlations betweenneach of the sites ranged from 0.32 to 0.46 in 1978 and fromn0.02 to 0.55 in 1982/83. These results indicated thatnsubstantial variability existed for the absolute andnrelative yield performance of entries within and among sitesnin both series of trials.nHeterogeneity among regressions on an environmental indexnaccounted for 71 percent of the entry x site interaction sumnof squares in 1978 but only 27 percent in 1982/83. Greater than 50 percent of the entry x site interaction sum ofnsquares was accounted for by individual regressions for 65npercent of the entries in 1978, but only 16 percent of thenentries in 1982/83. The rank correlations between the twonseries of trials, for regression coefficients (0.08),necovalence (-0.03), deviations from regression mean squaren(0.25) and percentage entry x site interaction sum ofnsquares accounted for by regression (-0.13) indicated thatnthese statistics had no predictive value. Furthermore, theyndid not facilitate the discrimination among entries on thenform of their response across sites.nClassification retained 69 and 63 percent of the entry xnsite interaction sum of squares among groups for 1978 andnfor 1982/83, respectively. Groups were relativelynhomogeneous for mean yield and for the form of entrynresponse across sites. The membership of the groups wasnrelated to the origin of the genetic material, its level ofnproductivity and to reactions to diseases. There was only anmoderate consistency between the membership of groups fromnclassification on performance in 1978 and in 1982/83. Thisnreflected the marked differences between the two series ofntrials for the absolute and relative performance of entries.nAdjustment of data did not improve the precision of trialsnin 1978, suggesting that trials were conducted on relativelynuniform sites. The performance of entries for mean yieldnacross sites in 1978, and in 1982/83, was more closelynrelated to the performance at those sites with relatively high mean yield and/or standard deviation, and for which thencorrelation with the performance of entries at other sitesnwas relatively high. Transformation of data as ranks or asnstandard normal variates resulted in relatively highernassociations between the mean yield across sites and thosenindividual sites comprising the mean for which the mean yieldnand/or standard deviation were relatively low.nTransformation of data as relative yields or as logarithmsnwas not effective for this purpose. However, the effect ofneach transformation was small for 1982/83 where there werenrelatively large differences in the correlations between pairs of sites for the yield performance of entries. Thisnsuggested that the correlations between pairs of sites had anmore marked influence on the association between thenperformance of entries for the mean yield across sites andnthe individual sites comprising the mean, than did site meannyields and standard deviations.nDiscrimination among entries in 1978 was practised usingnrelatively high mean yield across sites, a combination ofnrelatively high mean yield and a regression coefficientnclose to zero, and relatively high mean yield of groupsndelineated by classification. These selection strategiesnretained populations of entries with many common members.nThis was probably due to the importance of the mean yieldnacross sites for discriminating among entries.nRelatively larger numbers of entries were retained withnelite performance at Biloela, than at Pirrinuan or at Toobeah. The responses to selection in 1982/83 for the populations retained using a selected fraction of 20 percentnwere similar to, or higher than, those for a selectednfraction of 10 percent in several instances. Relativelynhigh proportions of the entries with elite performance inn1982/83 were rejected using a selected fraction of 10npercent.nThese results suggested that entry x site interactions inn1978 and in 1982/83 had a marked impact on the relativenperformance of, and discrimination among, entries for yieldnperformance across environments. It was recommended that innthe presence of substantial entry x year and entry x site xnyear interactions multi-stage selection may be the mostneffective procedure for the identification of those entriesnin early generation trials for which further testing wasnjustified. Alternatively, a series of managed environmentsnwith a specific set of environmental challenges may achievenmore effective discrimination among entries.n