3.2.1 Experimental sites
The study was conducted during two consecutive years in summer season of 2013 (2012/13) and 2014 (2013/14) at three sites in each year. In 2012/13 season, the study was carried out at Mutanda Research Station located at 12º25.959ʹ S and 26º12.620ʹ E (Environment 1), Mt. Makulu Research Station at 15º32.946ʹ S and 28º15.078ʹ E (Environment 2) and Golden valley Agricultural Research Trust (GART) at 14º58.185ʹ S and 28º06.134ʹ E (Environment 3). For 2013/14 season the experiment was evaluated at Mpongwe Seed-Co Research Farm located at 12º06.622ʹ S and 3º114.660ʹ E (Environment 4), Mt. Makulu Research Station at 13º32.831ʹ S and 28º03ʹ.626 E (Environment 5) and GART at 14º58.056ʹ S and 28º05.875ʹ E (Environment 6).
3.2.2 Experimental material, layout of the experiment and crop management One hundred and fifty wheat genotypes were used in the study. The materials comprised nine genotypes from Zambia Agricultural Research Institute (ZARI), one from Seed-Co, two from the University of Zambia (UNZA) and 138 (advanced lines and nurseries) from International Maize and Wheat Improvement Centre (CIMMYT), Mexico. The list of genotypes used for genetic diversity study is presented in Appendix 3.1.
The experimental field was laid out in a 10 × 15 alpha lattice design. Each genotype was planted in 2.5 meters long plot of two rows, 20 cm between rows with a plant to plant distance of 10 cm. Spacing of 40 cm between plots was used. Standard agronomic practices were followed for good crop management. Weeding was done by hand.
3.2.3 Measurements
Evaluation of morphological characteristics was done using descriptors recommended by the International Board for Plant Genetic Resources (IBPGR)(IBPGR, 1978). Observations were recorded on five plants per plot. Means for each trait were used for further statistical analysis. Data recorded was as given below.
1. Growth habit
a. Plant height (cm) – was recorded as height of plant at maturity, excluding awns.
b. Number of tillers per square meter – determined by counting number of tillers bearing ear spikes at the time of harvest per meter length of each row.
c. Tillers per plant – determined by counting number of tillers bearing spikes per plant based on an average of five plants.
2. Maturity
a. Days to heading (flowering) – recorded as number of days from sowing to the date when the spike completely emerged from the flag-leaf sheath on 50% of the plants in the plot.
b. Days to maturity – recorded as number of days from sowing to the date when 50% of the glumes have lost their green colour.
3. Yield and yield components
a. Spike (ear) length (cm) – was measured from the base to the tip of the spike, an average of five spikes.
b. Number of grains per spike – was determined by counting the number of grains per spike from the central portion of the spike; an average of five spikes.
c. Grain yield per plot (g/plot) – was measured by harvesting plants in a plot, threshing them and record grain weight.
d. Thousand grain weight – one thousand grains were counted from the bulk of grains of each entry and weighed on an electronic balance to determine its weight (g).
e. Peduncle length (cm) – was measured from the highest node to the base of the spike.
f. Hectolitre weight (kg h-1l) – measures the weight of hundred litres of wheat and was measured from the grain density bulk of grains of each entry using a hectolitre (hl) device.
3.2.4 Data analysis
Data obtained was subjected to analysis of variance using general linear model procedure (PROC GLM) in SAS version 9.3 (SAS Institute, 2011) to test significant differences among the genotypes. Analysis of variance (ANOVA) was performed separately on individual experiment of each environment and combined across environments. A combined ANOVA was conducted to determine the effect of genotypes, environment (location, year, and year x location) and the interaction. Genotypes and sites were considered fixed while replications and years were considered as random effects.
The following linear statistical model for combined analysis was used (Annicchiarico, 2002):
Yijkr =µ + gi + lj + (gl) ij + yk + br (ljyk) + (gy) ik + (ly) jk + (gly) ijk + eijkr)
Where Yijkr = observation of genotype i in location j in year k and block r, µ = overall mean, gi = effect of genotype i, lj = effect of location j, yk = effect of year k, br (ljyk) effect of block r within location j and year k, (gy) ik = genotype i x year k interaction, (ly) jk = location j x year k interaction, (gly) ijk = genotype i x location l x year k interaction and eijkr= residual effect The association for all the traits was estimated using simple linear correlation coefficient to determine the degree of association between the traits. Path analysis was also performed using the correlation values to assess the direct and indirect effects of different traits on grain yield following the method in Singh and Chaudhary (1995). Path coefficient values proposed by Lenka and Mishra (1973) as cited by Lule and Mengistu (2014) were used in this study. Path coefficients of < 0.09 were considered as having negligible direct effects, 0.10 to 0.19 as low, 0.20 to 0.29 as moderate and 0.30 to 0.99 as high direct effect on grain yield. Residual effects which determine how the causal factor (independent variable) accounts for variability of the dependent factor (yield) were estimated using the formula below (Singh and Chaudhary, 1995);
Residual effect (h) = √1-∑Piyriy
Where, Piy is the component of direct effect of independent ith factor and the dependent factor y (yield) as determined by path analysis, and riy is the correlation coefficient of ith factor with y (yield) as measured by correlation.
Based on the mean values for each trait, the principal component analysis was performed in GenStat version 14 (Payne et al., 2011) to detect traits that explained the most variability in the data set and also to cluster genotypes based on the similarities. In this study, the trait
with the coefficient equal to or greater than 0.3 was considered to discriminate the genotypes more than those with coefficient less than 0.3 (Badu-Apraku et al., 2006; Sanni et al., 2012). Cluster analysis based on Ward’s method (Ward, 1963) using squared Euclidean distance was used to group genotypes in to clusters using Statistical Package for Social Scientists (SPSS) 16.0 version for windows (SPSS, 2007).