Abstract: [Objective]The landforms of mountains are complex in the south of China, most genetic trials of forest species were established in hilly areas where the consistency of environmental conditions was relatively poor, spatial interactions exist among adjacent individuals. Spatial analysis can help us decrease experimental errors and enhance the accuracy of estimation of genetic parameters. [Method]The Akaike Information Criterion (AIC) values and variance components for controlled-pollinated family group(G8) and open-pollinated family group(G5) from first-generation seed orchard of Pinus taeda were compared using different models of spatial analysis, and the genetic parameters for height (H), DBH, stem straightness (STR), and clear bole height (CH) were estimated using the best spatial analysis model (AR1η+Rep·Fam). [Result]The best spatial analysis was applied to G8 and G5 using individual-tree models, which improved the additive genetic variance, decreased experimental errors and increased genetic gains. In contrast to the results from base design model, the spatial analysis of field data for G8 and G5 indicated that the autocorrelations were high, and they were consistent between rows and columns. For G8, the additive variance for H was increased by 9.1%, and residual variance was reduced by 7.3%; the additive variance for DBH was increased by 0.7%, and residual variance was reduced by 1.6%; the additive variance for CH was kept the same in the two models, and residual variance was reduced by 18.2%; the additive variance for SSTR was increased by 50%, and residual variance was reduced by 1.9%. Heritability for SSTR was increased by 30.7%, and heritability for H, DBH and CH were reduced by 24.4%, 38.7% and 18.7% respectively. For G5, the additive variance for H was kept the same in the two models, and residual variance was reduced by 4.9%; the additive variance for DBH was increased by 3.2%, and residual variance was reduced by 2.9%; the additive variance for CH was increased by 100.0%, and residual variance was reduced by 20.8%; the additive variance for SSTR was increased by 33.3%, and residual variance was reduced by 2.5%. The narrow-sense heritability of H and DBH for individuals were respectively reduced by 42.1% and 34.2%, and the narrow-sense heritability of CH and SSTR for individuals were increased by 61.5% and 5.8% respectively. For the controlled-pollinated family group(G8), the six top families at a selection rate of 15% of all the tested families, were identical between the groups with and without spatial analysis. For the open-pollinated family group(G5), five of the top six families were identical between the groups with and without spatial analysis. Increased gains for DBH from selection with rate at 15% were 7.2% for G8 and 24.5% for G5.[Conclusion]Compared to the mixed linear model, spatial analysis reduces residual variance and increases genetic gains of various characters, but we cannot ignore the field experimental design, appropriate combination of field experimental design and spatial analysis can reduce the environmental variance, improve the accuracy of estimation of genetic parameters, and improve the efficiency of forest genetic trials.

Key words: Pinus taeda, spatial analysis, genetic parameter, breeding value