Comparison of lattice designs,check plots,and moving means in wheat breeding trials

Summary Efficiencies of lattice designs, check plot designs, and moving means were compared in seven environments for control of error in wheat breeding trials. Lattice analyses of lattice designs and check plot covariance analyses of check plot designs gave similar CV's and intra-site heritabilities and were superior to direct RCBD analyses of these designs. Moving mean covariance analyses were generally only slightly less efficient than lattice analyses or check plot covariance analyses. Differences between designs and methods of analysis, which were evident on an intra-site basis, were not evident from combined analyses indicating that genotype x environment interaction is much more important than intra-site error in limiting progress from selection. Moving mean covariance analyses with 2 replicates per location and as many locations as feasible is suggested as a useful strategy for early generation wheat breeding trials.     

Adjustment for Competition Between Genotypes in Single-Row-Plot Trials of Winter Wheat (Triticum aestivum)

Several statistical methods were investigated for improving yield estimation for 72 genotypes of winter wheat tested in a single-row-plot (SRP) trial. The different analytical models were compared using several criteria: the residual mean square fit; the standard error of differences of genotype means; the correlations of genotype means with means from six-row-plot (6RP) reference trial; the expected genetic gains relative to 6RP trial from selecting the 10 top yielding genotypes. There was no significant correlation between the unadjusted genotype means for SRP and 6RP. Genotype yields were positively correlated with height in SRP, but this correlation was reversed in 6RP. Using the mean height difference of a plot with its two neighbours as a covariate to adjust for competition in SRP reduced the standard error of difference of genotype means and substantially increased their correlation with the 6RP means. The predicted selection efficiency in SRP trial increased from zero, before adjustment, to 83 % of 6RP trial. Including the difference in heading date and the effect of guards at borders as covariates in the SRP analysis improved the residual mean square fit but had no effect on correlation with 6RP means or selection efficiency. A producer-competitor model in which individual competitor effects were estimated for each genotype gave a substantial improvement in fit over the covariate models but a much lower correlation with 6RP means and selection efficiency. Yield selection in early generation wheat trials based on single-row plots could be considerably strengthened by use of a plot covariate derived from height difference between neighbours.     

Inter-plot competition in yield trials of swedes (Brassica napus ssp. rapifera L.)

Summary Two experiments were done to evaluate the use of single-row plots for assessing dry-matter yield of swedes. In the first experiment the correlation between the dry-matter yields of 20 cultivars in unguarded single-row plots and those in guarded double-row plots was poor (r=0.39). It was improved (r=0.58) by adjusting the single-row plot fresh-weight yields for inter-plot competition, through an analysis of covariance in which the covariate was the mean of the adjacent plot yields, and then estimating the pure stand dry-matter yields. However, the correlation was still lower than that between unguarded double-row plots and the guarded double-row plots (r=0.73), and this latter correlation was also improved (r=0.82) by adjusting the fresh-weight yields of the unguarded plots for inter-plot competition.The second experiment provided an explanation for the partial success of the covariance adjustment. The individual neighbour effects of six cultivars were determined. Doon Major and Ruta Øtofte decreased the yields of adjacent cultivars whereas Dryden and Melfort increased them. Angela and Magres had little effect. The correlation between the dry-matter yield neighbour effects of cultivars and their fresh-weight yields was high (r=–0.81), but not complete. So although competitive ability was associated with root fresh-weight yield, other factors must also be involved.It was concluded that a two-replicate trial with unguarded double-row plots is preferable to a four-replicate trial with unguarded single-row plots for dry-matter yield assessment of early generation families in a pedigree swede breeding programme.     

Optimal regimes of selection for grain yield and harvest index in spring wheat

Summary Alternative strategies of multi-site testing of advanced lines in the northern wheat belt of New South Wales have been evaluated, using genetic parameters for large plot grain yield and hill plot harvest index estimated from dryland and irrigated trials at regional sites during 1975–1981. The average pairwise genetic correlation of large plot grain yields recorded at different sites within years was 0.45±.03, with a mean repeatability within trials of 0.56±.05. Harvest index measured in 20-grain hill plots in 1978 showe genetic correlations of 0.98±.08 with plot yield at the same site, and 0.39±0.06 with plot yield assessed at other sites in the same year.The genetic correlation between harvest index in hill plots and total biological yield in large plots at the same site was 0.84±.13, the relationship showing no evidence of curvilinearity. Selection for harvest index in hill plots is therefore expected to lead to an increase in biological yield as well as grain yield in the breeding populations studied. Quantitative genetic theory suggests that the response to selection for grain yield can be increased by approximately 40% with an initial screening using hill plot harvest index at three sites instead of one, and reallocation of resources in the first stage of large plot yield assessment to include 6–8 sites, rather than dryland and irrigated trials at a single location.     

Selection for yield in small plots of spring wheat

Summary Four selection experiments were sown: single-row plots with single row-spacing (20 cm), single-row plots with double row spacing (40 cm), three-row plots and six-row plots both with single row-spacing (20 cm). Selection entries were mimiced by 16 different varieties or advanced breeding lines, which were also sown in a yield trial. Each experiment was laid out as a 4-times replicated randomized block design. Row length was 2 m. Alley borders and border-rows of multiple-row plots were harvested separately to evaluate the effects of different harvesting procedures on the selection efficiency. Removal of alley borders was found to be disadvantageous, since the gain in precision was more than offset by the loss in sampled area. Wide spacing of single-row plots improved the selection efficiency in comparison with normal spaced single-row plots. In multiple-row plots the selection efficiency was not improved by harvesting only the central rows.For gross plot yield (= yield of net plot + yield of alley borders) the differences in selection efficiency between the various selection plot-types were explained on the basis of the genetic variance, the environmental variance and the coefficient of genetic correlation with farm yield as determined in the yield trial.     

Biometrical analysis of alternative plot types for selection in rye

Summary Maximizing the efficiency of selection for yield and baking quality of winter rye (Secale cereale L.) requires reliable knowledge on the pertinent population parameters. This study reports estimates of variances, heritabilities and genetic correlations from a) large (5 m²) drilled (LD) plots, b) micro drilled (MD) plots, and c) one-row plots of 4 spaced plants (SP). Thirty eight single crosses of rye were grown at two locations for two years in LD, MD and SP plots replicated 2, 4, and 6 times, respectively. Genotypic differences were significant in all plot types for all agronomic and quality traits, although estimates of genetic variances were smaller in LD than MD or SP plots for grain yield, 32-spike weight, and kernels per spike. Heritabilities on an entry mean basis (0.62 to 0.95) were similar among plot types, but on a single plot basis estimates were lower for SP than for MD or LD plots. Genetic correlations were high between all three plot types, with correlations between LD and SP plots being approximately 1.0 for falling number, thousand-kernel weight, and bloom date, 0.90 for grain yield, kernels per spike, test weight and height, and 0.70 for tiller number. Based on the high heritabilities achieved in small plots and the close agreement between MD- or SP- with LD-plot performance we conclude that greater use should be made of small plot types when selecting for yield and quality in winter rye.     

Genetic variance models for the evaluation of resistance to powdery scab (Spongospora subterranea f. sp. subterranea) from long-term potato breeding trials

Breeding for resistance to soil-borne powdery scab in potato is an important component of the integrated management of this disease. Different genetic variance models within a mixed model framework were applied to data from long-term potato breeding trials, for the genetic evaluation of breeding lines. The multi-environment trial (MET) data came from 12 growing seasons (“years”, synonymous with environments) of New Zealand field trials screening for resistance to powdery scab on potato tubers. Pedigree information on a total of 1,031 genotypes was available. Additive components of the genetic effects were important with narrow-sense heritability estimates (and 95 % confidence intervals) from single-year analyses ranging from 0.26 (0.20, 0.44) to 0.57 (0.53, 0.85). Spatial components estimated from the residual plot effects were not important for most years and even when they were significant, estimates were small. In MET analyses, different variance structures for the genetic effects were tested; a homogeneous correlation model (CORH) gave a better fit to the data than a factor analytic FAk model of order (k), 1 and 2. The year-to-year genetic correlation estimate from CORH was 0.81 and compared with a range of 0.59–0.95 estimated from the FA1 model. There was no strong evidence of non-additive genetic effects with zero or boundary estimates for most years. Models which included the pedigree provided a better fit to the data than models that did not include this relationship information. There was no evidence for genetic improvement in resistance for powdery scab on tubers in the breeding population studied. This suggests that selection pressure for resistance in the past has been weak and greater consideration should be given to selecting parents on empirical breeding values to genetically improve breeding populations for resistance to powdery scab.     

Genetic analysis and selection for wheat yield in drought-stressed and irrigated environments

Summary Wheat (Triticum aestivum L.) cultivars grown in the southern Great Plains of the U.S.A. are exposed to a wide range of moisture conditions due to large fluctuations in the amount and frequency of rainfall. Yield stability under those conditions is therefore a desirable trait for wheat breeders. Our primary objective was to quantify various genetic parameters for grain production in drought-stressed and irrigated environments. We also attempted to predict and measure yield responses when selection is practiced in either drought-stressed or irrigated environments, or both. Seventy F₂-derived lines from the cross, TAM W-101/Sturdy, were evaluated at Goodwell, OK, under irrigated and naturally drought-stressed conditions in 1987 and 1988. Genetic variance and heritability estimates were higher in the irrigated environment than in the drought-stressed environment. The genetic correlation coefficient for yields in the two environments was 0.20±0.16, indicating that selection of widely adapted genotypes requires testing in both environments. Based on the genetic variance/covariance structure of this particular population, the linear index which maximized the combined expected gain in both environments was 0.66Y₁ + 0.34Y₂, in which Y₁ and Y₂ are yields in the irrigated and drought-stressed environments. This index is not expected to apply across all populations; rather, it further supports the hypothesis that testing in either environment alone (drought stressed or irrigated) may not be most effective for increasing either mean productivity or yield under drought stress.     

Genotypic and environmental variability of yield for cultivars from 30 different crops in German official variety trials

Genotypic, environmental and genotype by environmental variability of 30 different crops was evaluated by estimating variance components from trial series in 16 years. Yield data from official German variety trials for cultivation and use (VCU) were analysed over the period from 1991 to 2006. The natural dominant part was pure environmental variation. Genotype, genotype by environment and plot error are smaller and of the same magnitude. Interaction of genotypes with locations is larger than with years. Compared with variance component estimates of crops from a study involving 16 crops from former West German VCU trials between 1975 and 1986 overall variability of yielding performance was lower in the recent study for all components except for location. Precision of variety comparisons over locations and years is higher in recent VCU trial systems when assuming equal dimensioning of testing resources. Lower values for genotype by environmental variances, especially for the genotype by year component lead to lower standard errors for variety comparisons.     

Effect of end-border condition on small-plot yield of cucumber

Summary Twelve or 14 cultivars and breeding lines (collectively referred to as lines hereafter) of pickling and fresh-market cucumbers (Cucumis sativus L.) were evaluated in 1.5 m long plots. The plots were harvested once when the check lines reached the 10% oversized fruit stage, and total, marketable, and early (oversized) number of fruits per plot was counted. The plots were planted with or without 1.5 m long end-borders to determine whether end-borders can be eliminated in small-plot trials where seed numbers and field space are limiting. Yield in unbordered plots was inflated 5 to 21% over bordered plots, but there was generally no significant effect in the analysis of variance for line x end-border condition. Therefore, end-borders are not needed when comparing different lines.     

The effect of variation of genetic variance and correlation between mean and variance on efficiency of bulk yield testing

Summary Computer simulation techniques were used to examine the effect of variation of genetic variance and correlation between mean and genetic variance on the efficiency of bulk yield testing procedures. Assessment of the effect of variation in various parameters was made by comparing genetic gains with the maximum potential gain. The maximum potential gain was defined as that which would be achieved if the particular bulk which maximizes genetic advance were chosen for selection.Maximum potential gains increased with increasing mean genetic variance within bulks but showed only a slight response to increasing variation of the within bulk genetic variance. Negative correlations between bulk mean and genetic variance reduced maximum potential gains and positive correlations increased them but the effects were not large.A selection intensity of 5% among bulks seemed close to optimal for a wide range of parameters. Slightly more intense selection may be an advantage where the between bulk heritability is high, the within bulk variance is high, and the correlation between bulk mean and variance is high and positive.     

A comparison between single seed descent and early cross selection in wheat breeding

Summary Two selection procedures in wheat breeding were compared on the basis of their ability to supply high yielding inbred lines. The first procedure consists of an early selection between crosses in the F₃ generation, based on predictions of the cross mean and the between line variance. In the second procedure selection is postponed until the F₆, which is derived by single seed descent. The two procedures are evaluated in a two year test, using pseudo-lines of spring wheat. These pseudo-lines consist of mixtures of varieties and enable an estimation of the exact genetic parameters. In this way the accuracy of the predictions can be examined.In case of early selection, it appears that the predictions of the cross mean and especially the between line variance are very inaccurate. This is caused by the effects of plot size, intergenotypic competition and, to a lesser extent, dominance and/or epistasis. It results in an erroneous ranking of the crosses and the discarding of the potentially best cross. The F₆-SSD line estimates are much more accurate and thus the better lines are indeed selected. A first comparison between the two selection procedures therefore indicates a preference to the SSD method.     

陆地棉种子性状直接效应和母体效应的遗传分析

采用二倍体种子遗传模型及其分析方法，分析了陆地棉５个品种及其配制的Ｆ１、Ｆ２和正反回交一代（ＢＣ１、ＢＣ２、ＲＢＣ１和ＲＢＣ２）世代籽指、种仁率、容重和油分４个种子性状的直接效应、母体效应和细胞质效应。分析结果表明，４个棉子性状的遗传均以母体效应为主，种子直接遗传效应作用较小。各性状的直接效应与母体效应的协方差不显著。通过母体植株的遗传表现对这些性状选择更为有效。可以根据籽指小和容量大等性状母体植     

Estimation of heritability and response to selection for cut-flower yield in gerbera

Summary Heritabilities and responses to selection for cut flower yield were estimated for a population of gerbera. Broad-sense heritabilities averaged 42 percent, but may be as high as 66 percent. Narrow sense heritabilities averaged 68 percent when based on half-sib families, and averaged 60 percent when based on parent-offspring regression. These results suggest that most genetic variance for flower yield is additive.However, estimates of realized heritability averaged only 16 percent and inbreeding depression was estimated to be at least 38 percent, suggesting a major role for non-additive genetic variance. Nonetheless, selection for cut flower yield is expected to be successful.     

A comparison of several procedures for mass selection in winter rye. II. What are the merits of adjusting phenotypic values?

Summary To get an improved impression of the genotypic value of individual plants the literature suggests to adjust their phenotypic values according to a moving mean correction or by means of an analysis of covariance, where the mean phenotypic value calculated across neighbours is used as the concomitant variable. In the present paper it is shown that the theoretical merits of these adjustments are promising: they eliminate the contribution due to the trend in soil fertility from the phenotypic value. In an experimental verification actual merits did, however, not show up.It is concluded that interplant competition is the main cause for the failure of the studied adjustments. These may be effective in virtual absence of intergenotypic competition. Thus, in case of single plant selection an extremely low plant density would be required, whereas selection among lines or families would require evaluation by means of multi-row plots.     

A low-cost procedure for multi-purpose,large-scale field evaluation of forage crop genetic resources

The proposed procedure for field evaluation of large forage germplasm collections allows for simultaneously performing germplasm characterization, stratified mass selection and the assessment of the genetic structure of populations as revealed by among-population vs. within-population genetic variances. Unreplicated genotypes are randomly assigned to different units (grids) with the constraint that each unit should have a fixed proportion of genotypes from each population. The relative homogeneity of plant material between units allows for investigating major spatial trends.Within each unit, one plant is actually a clonal control allocated randomly and used for estimating the environmental variance. The units have a square shape, with plants grown in dense planting arranged at same spacing between and within rows. A set of adjacent units forms a column, and several parallel columns may exist. Observed data can be adjusted for the effects of: i) column of units; ii) row of units (by covariance analysis, using row number as the covariate); and iii) column of plants within units. The procedure was applied to the evaluation of 4,480 lucerne genotypes subdivided into 128 units and grown at 10 cm spacing. The efficiency of data adjustment, estimated as the variance among clones for unadjusted data relative to that for adjusted data, was 33.7% for length of the main stem, 9.8% for plant dry weight, and greater than 7% for three out of four other evaluated traits.     

Stability of performance of mixtures and multilines

Summary The effect of intra-populational genetic diversity on the stability of performance of equiproportional mixtures, as measured by their variance in yield over a range of environments, was investigated with a simple theoretical model.In the absence of intergenotypic interactions it was found that a mixture will be more stable than the most stable line (V_min) if:{fx405-1} where n is the number of lines in the mixture and {ie405-1} and {ie405-2} are the mean variance and covariance in yield of the component lines. In other words, the yield of multilines will vary less than that of their most invariant component only when the component lines respond quite differently to environmental changes.In the presence of intergenotypic interactions it was found that a mixture will be more stable than its best or better component (V_min) provided:{fx405-2} where n, {ie405-1} and {ie405-2} take the same meaning as before and {ie405-3} and {ie405-4} represent the mean variance and covariance of the interaction effects among the n lines. That is, when the components of a mixture significantly interact, then the dissimilarity in response of each component to environmental fluctuations must be even greater for there to be a stability benefit to mixtures.These findings are discussed in relation to the problem of developing commercially acceptable multiline varieties.     

The effect of interplot competition on yield assessment in triticale trials

Summary A series of experiments was carried out to investigate the effects of interplot competition on grain yield among hexaploid triticale cultivars of widely differing heights grown in 1.5 m wide plots. Three experiments in 1982 tested four triticale cultivars, ranging in final height from 63 to 104 cm, with four wheat standards, 62 to 82 cm in height, using a design balanced for nearest neighbours. Two further experiments in 1983 with the same design included eight triticale cultivars ranging in final height from 92 to 183 cm. In four of the five experiments, plot yields of both wheat and triticale were decreased by between one and two grammes per square metre for every centimetre by which the plot was exceeded by the mean height of its two neighbours. Plots which were taller than their neighbours showed a similar increase in yield. Such systematic biases in cultivar means are not reduced by randomisation and increased replication. Hence, in breeding trials, every effort should be made to group genotypes according to their final height so that competition effects between plots are reduced. For the final evaluation of varieties differing widely in height, where an estimate of error is required for variety comparisons and varietal differences are likely to be small, guard plots should be grown on either side of each experimental plot or harvesting should be restricted to the centre of larger plots.     

Selection in Early Generations of Dried Peas, Pisum sativum L. III. Plot Size and Plot Type

Investigations were carried out to study the effect of different plot sizes and plot types on the performance of peas and to determine plot size and plot type combinations suitable for the reliable evaluation of varietal yield performance realized in commercial stands of pea crops. Four pea varieties were tested in 1981 and 1982 in combinations of plot sizes ranging from 0.25 m2 to 19.50 in2 with plot types either unbordered, bordered with faba beans, surrounded with perforated plastic foil or only central sections of plots harvested (net plots). With decreasing plot size the performance mean as well as the error variance of grain yield and thousand grain weight increased disproportionately. Thus, changes in grain yield were much higher than those for thousand grain weight. The four varieties tested reacted differently to changing plot sizes and as a result, highly significant variety × plot size interactions were found. Although overestimation of grain yield in small plots could be strongly reduced by bordering the plots with faba beans or surrounding them with plastic foil, their suitibility for yield testing was limited by highly significant variety × plot type interactions. Compared to the largest net plot (15 m²) reliable results were still obtained with plot sizes larger than 6.5 m², no matter whether edge effects were eliminated or not. If only a small quantity of seed is available, plots may be reduced even to 1.00 m², if surrounded by plastic foil. The results presented in peas were compared to those in other crops and discussed concerning their relevance for variety testing as well as for selection of grain yield in early generations.     

Spatial Dependence of Check Plot Yield and Local Control

For maximizing genetic gains, with given resources, plant breeders generally conduct large but un- or inadequately-replicated field trials especially during early generation evaluations and population improvement programmes. Checks arc often used to improve the precision of test plot comparisons. Adjustments by moving or weighted means etc. arc, however, empirical or based on assumptions too difficult to verify. An alternative stochastically best linear unbiased optimal locally weighted method of spatial prediction of micro-environmental heterogeneity indices of treatment plots was examined. Two checks used in an unreplicated evaluation of 1560 test lines of winter wheat expressed the micro-environmental variation in similar ways. About 46, 38 and 16 percent of the total variation in yield was a simple row and column effects, spatially dependent, and random errors, respectively. Optimal weights derived to simulate the micro-environmental heterogeneity indices were sensitive to the position of the test plot and thus provided a better local control. The model was validated, and prediction standard deviations of simulations computed.