Abstract: Tree allometric models based on height (H) and diameter (D) are the most commonly used method to estimate forest biomass. Environments and stand characteristics are recognized to affect tree allometries. However, few studies have considered to incorporate these effects into allometric models, which restricts the use of these models in a wide domain. Adopting the power-law function Y = aG^b as a basic model where Y is either tree height or biomass and the corresponding G is tree diameter D at breast height or D²H, we developed a two-step approximation procedure to quantify the effects of environments and stand characteristics on allometric coeffcients a and b for Cunninghamia lanceolata and Pinus forest in China. Results show that most of the allometric coeffcients are dependent on stand characteristics for C. lanceolata forest, and on mean annual temperature, stand age and latitude for Pinus forest. The allometric models via the two-step approximation Y = f(α + α_jx_j) G^{f (β+βixi)} (x_j or x_i are key drivers associated with environments and stand characteristics. α, α_j,β and β_i are regression coeffcients) considerably improved the accuracy of tree height and biomass estimation. Compared to the basic model, the second approximation models signifcantly reduced the mean absolute bias between the observed and computed values by 25%–34% for C. lanceolata and by 21%–26% for Pinus forest, respectively. Our results highlight the necessity of incorporating environments and stand characteristics into the allometric models and provide a universal method to accurately estimate H-D-based tree biomass across a wider domain.

Key words: allometric coeffcient, two-step approximation, environment, stand characteristics, stepwise regression, nonlinear regression, forest