Development of biomass models for six understory seedling and sapling species in broad-leaved mixed forests of Maoershan Mountain, Northeast China utilizing seemingly unrelated regression.

Seedlings and saplings are vital elements of understory vegetation, the accurate biomass estimation of which is important for quantifying carbon storage within forest ecosystems. With data of 620 seedlings and saplings individuals from six species-<i>Acer mono</i>, <i>Populus davidiana</i>, <i>Ulmus laciniata</i>, <i>Fraxinus mandschurica</i>, <i>Quercus mongolica</i>, and <i>Syringa amurensis</i>-across 101 broadleaf mixed forest plots in Maoershan Mountain, we developed power-function biomass models utilizing basal diameter, plant height, and crown area as independent variables and identify the optimal models as the base models. We further assessed the error structure of each base model through likelihood analysis, and established a biomass equation system for the six species using seemingly unrelated regression (SUR). The results showed that the univariate model utilizing only basal diameter was the most effective for <i>F. mandschurica</i>. For <i>S. amurensis</i>, the ternary model that encompassed basal diameter, plant height, and crown area was superior. For the other species, the binary biomass models that included basal diameter and plant height yielded the best results. The adjusted coefficients of determination (<i>R</i>_a²) varied from 0.716 to 0.990, while the root mean square errors (RMSE) ranged from 0.060 to 6.403, with all model parameters showing significance. The error structure for both component and total biomass across the species was found to be multiplicative (ΔAICc>2). Consequently, linear biomass models following logarithmic transformation were employed to develop the SUR biomass models for the six species. These models had high <i>R</i>_a² values (0.713-0.987) and low RMSE values (0.062-7.408), suggesting they were appropriate for accurately estimating the biomass of seedlings and saplings in the understory.