A multivariate mixed model system for wood specific gravity and moisture content of planted loblolly pine stands in the southern United States

Specific gravity (SG) and moisture content (MC) both have a strong influence on the quantity and quality of wood fiber. We proposed a multivariate mixed model system to model the two properties simultaneously. Disk SG and MC at different height levels were measured from 3 trees in 135 stands across the natural range of loblolly pine and the stand level values were used for the modeling SG-MC system. Regional variation in mean trend of the properties was incorporated in the model. Contemporaneous correlation between the SG and MC was accounted by defining within stand error structure appropriately. Compared to univariate models, predictions based on the multivariate model were improved by 29 and 26 % in root mean square prediction error for disk SG and MC after taking account of the contemporaneous correlation.