Abstract: Large-scale estimation of forest biomass has received much attention. Constructing a stand-level biomass model is a method for estimating tree layer biomass. In this study, we constructed stand biomass models of Korean pine plantations based on aggregation method 1, aggregation method 2, adjustment method, and disaggregation method. The prediction precision of four additive methods was compared and analyzed to provide theoretical basis for biomass prediction of Korean pine plantations in Heilongjiang Province. Weighted functions were used to eliminate the heteroscedasticity of each model, with the leave-one-out cross validation (LOOCV) as the validation method. The results showed that the overall prediction ability of the adjustment method was slightly better than other methods. The specific prediction precision was ranked as adjustment method > aggregation method 1 > aggregation method 2 > disaggregation method. The prediction precision of four additive methods was not consistent when considering their prediction ability of different stand basal areas. When the stand basal area of Korean pine plantations was distributed in the interval of 0-10 or 50-60 m²·hm^-2, the parameter estimation values of disaggregation method performed better. When the stand basal area was distributed in other intervals, the parameter estimation values of adjustment method was better.

Key words: Pinus koraiensis plantation, additivity method, stand biomass model, leave-one-out cross validation, prediction precision