Abstract: Using single nonlinear equation with constant parameters to simulate tree biomass may result in obvious biased estimation for small diameter class saplings. Based on the aboveground biomass data of larch and oak in northeastern China, an approach with dummy variable was presented which could improve the estimation of tree biomass, and one-and two-variable tree biomass models for the two tree species were constructed. The results showed that: 1) the approach not only was effective in solving the problem of biased estimation for small saplings, but also improved the prediction results of biomass model for all sample trees to some extent; 2) the constructed models were suitable for both saplings with diameter at breast height (DBH) less than 5 cm and trees with DBH more than 5 cm for biomass estimation of larch and oak trees in all kinds of forest inventory in northeastern China, and through the two-variable biomass models, the mean predicting error was all less than 5%.