Developing biomass equations for Western hemlock and red alder trees in Western Oregon forests

Abstract

Biomass estimates are required for reporting carbon, assessing feedstock availability, and assessing forest fire threat. We developed diameter- and height-based biomass equations for Western hemlock (Tsuga heterophylla (Raf.) Sarg.) and red alder (Alnus rubra Bong.) trees in Western Oregon. A system of component biomass equations was fitted simultaneously with a constrained seemingly unrelated regression. Additionally, a linear model that predicts total aboveground biomass as a function of DBH and height was also fitted. The predicted total biomass was then apportioned to different components according to the predicted proportions from beta, Dirichlet, and multinomial log-linear regressions. Accuracy of these methods differed between species with higher root mean squared error (RMSE) being produced in red alder trees. Within species, the accuracy of the equation for bole biomass was better than the equations for other components. None of these methods stood out as a clear winner, but the multinomial log-linear regression produced marginally better results compared to other methods in terms of RMSE, except for Western hemlock bark biomass and red alder bole and branch biomass. The equations based on a seemingly unrelated regression provided lower RMSEs for those species-component combinations.