Document Type : Original Article
Authors
1 Department of Agronomy and Plant Breeding, Faculty of Agriculture, Lorestan University, Khorramabad, Iran
2 Assistant Professor. Department of Agronomy and Plant Breeding, Faculty of Agriculture, Lorestan University, Khorramabad, Iran
3 Research assistant, Seed and Plant Improvement Departement, Lorestan Agricultural and Natural Resources, Research and Education Center, AREEO, Khoramabad, Iran
Abstract
Introduction
The main objective in breeding programs is mainly based on evaluation experiments to select superior genotypes and accurate estimates of the mean of varieties of different varieties. Soil fertility, soil water-holding capacity, soil physical and mineral characteristics, wind direction and other unknown agents are often varied across an experimental site. This means that that the residuals of conventional models such as RCB are not independent of their near plots or spatial locations. So, this is an inherent issue which spatial adjustment of errors necessarily improves the precision of estimated treatment. The main objectives of this paper were to (1) explore the characterization of spatial dependence of traits using linear mixed models in large field chickpea trials in normal and drought conditions, (2) compare the efficiency of statistical analysis of spatial model to RCB model, (3) estimation of genetic parameters based on spatial models.
Materials and methods
This research was carried out in order to evaluate the yield and yield components of 64 chickpea genotypes in normal and rainfed conditions in Khorramabad Province, Iran. Experiments were conducted in an 8 × 8 Lattice square design framework in two replications for each condition during the 2018 season. Due to seasonal precipitation, there was no significant statistical difference between rainfed and normal conditions as well genotype by environment interaction in most of the traits. Therefore, rainfed data were considered as replications for the normal environment in all traits and all replications were subjected as one condition. To analyze the data for removing heterogeneity of farm condition in Lattices square design, Federer's spatial models were used to optimize the Lattice square linear model. Finally, the nearest neighbor model was selected as the best model based on the relative advantage and also the statistic -2RLL.
Results and discussion
The results of the combined analysis of variance of chickpea genotypes in two normal and dry conditions were shown in Table 2. These results showed that there was no significant difference between two normal and dry conditions in all traits except for pod weight and root dry weight. These results indicated that the two environments provided the same conditions for the genotypes on average. Also, the interaction of genotype in the environment was not statistically significant in all traits. After integrating the normal and dry data, the analysis of variance was carried out using randomized block design and lattice approach on data. The results showed that the spatial model (4), the nearest neighbor model, is the best linear model for analyzing and comparing the mean of genotypes. This model had the lowest -2RLL (-2 Res Log Likelihood). The lower the value of this statistic, the better the adequacy of the model to justify the response variable. Analysis of variance by the nearest neighbor method for different traits was shown in Table 4. The significant differences existed among genotypes in all traits except for pod weight and root dry weight at the 0.01probability level. Estimation of heritability of different traits showed that most of the studied traits had a heritability ranged from 0.39 to 0.75. For the grain yield traits with 10% intensity selection, the predicted genetic gain was 14% relative to the trait average. Therefore, this genetic gain can be used to advance the development of this trait. Also, genotypes G59, G09, G12, G30, and G41, respectively, in the nearest neighbor model, had the highest ranked estimated grain yields.
Conclusions
The results of the analysis of the nearest neighbor model showed that there was a significant difference between all genotypes for all traits, as well as high genetic variance and acceptable heritability. Therefore, the variation in the genotypes studied could be used and introduced superior genotypes for breeding programs and even introductions to farmers.
Keywords
)