Effectiveness of spatial analysis in Cryptomeria japonica D. Don (sugi) forward selection revealed by validation using progeny and clonal tests

Abstract

Accurate evaluation of genetic performances of trees is crucial in order to improve the efficiency of forest tree breeding. We revealed that spatial analysis is effective for predicting individual tree breeding values at the forward selection stage of Cryptomeria japonica D. Don (sugi) breeding program by using a novel validation approach. In the process of selecting genetically superior trees for breeding, appropriate handling of environmental effects is important in order to precisely evaluate candidate trees. Spatial analysis has been an effective statistical approach for genetic evaluation at sites with heterogeneous microenvironments. However, the efficiency of spatial analysis on forward selection has not been validated on a practical scale to date. This study aimed to reveal the effectiveness of spatial analysis, which incorporates spatially autocorrelated residuals into mixed models, for the prediction of breeding values at the forward selection stage by validation using progeny or clonal tests of forward-selected individuals. Tree height was analyzed by ordinary randomized complete block design models and spatial models incorporating spatially autocorrelated residuals in a linear mixed model framework, and model selection was conducted at thirty Cryptomeria japonica D. Don breeding population sites having various topographical ruggedness. For validation, three clonal tests and one progeny test of individuals selected from three and four breeding populations, respectively, were used. The effectiveness of forward selection using the two models was evaluated based on the correlation between individual breeding values at the stage of forward selection and genotypic and breeding values that were estimated by clonal and progeny tests. Spatial models were more predictive than ordinary models in all cases. Spatial correlation parameters tend to increase with the topographical ruggedness index of each site. The correlation coefficients between breeding values at the time of forward selection and genotypic or breeding values evaluated in succeeding clonal and progeny tests were significantly higher in spatial models than in ordinary models in six out of nine cases. Validation using progeny and clonal tests of forward-selected individual trees revealed that spatial analysis is more effective for the evaluation of genetic performance of individuals at the stage of forward selection in Cryptomeria japonica.