Quantifying tree biomass is a core element of forest inventory and a long-standing research priority at the Hubbard Brook Experimental Forest (HBEF). Because direct measurements are impractical, aboveground live tree biomass (AGB) must be inferred from allometric equations developed from limited destructive samples. The legacy HBEF biomass estimation models (BEMs) systematically underestimated AGB in young, even-aged stands, indicating limits to their generality. We developed a revised set of BEMs, the Hubbard Brook Forest Analytics (HBFA) models, using an expanded allometry dataset that incorporated additional species, size classes, and stand ages. Nonlinear fitting methods improved the representation of variance structure and diagnostic performance. Across species, HBFA models achieved high predictive skill (pseudo R² = 0.70–0.99) and low relative errors (rRMSE ≤ 0.27 for most species). When evaluated against direct harvest measurements, predicted AGB differed from observed values by only 1.6% in mature forests and remained within one standard error of locally derived estimates for early successional stands. Monte Carlo error decomposition showed that bias accounted for less than 6% of total prediction error, with residual and coefficient variability dominating. By integrating local and regional data with reproducible analytical procedures, the HBFA framework strengthens long-term biomass monitoring and supports uncertainty-quantified forest carbon assessments.
Rutherford et al. (Tue,) studied this question.