Improving biomass estimation for Eucalyptus globulus Labill at stand level in Portugal.

Two systems of equations for predicting stand level aboveground biomass and biomass per tree component (wood, bark, branches and leaves) were developed taking into consideration the specific characteristics of the stands. The two systems differ by considering or not stand age as one of the regressors. Data from permanent plot, trials and continuous forestry inventory was used. It included data from first rotation and coppiced stands that represent the existing range of ages, stand densities, sites and management options in Portugal. The models have a good predictive capacity (adjusted R 2 > 0.97) using only stand variables easily accessible in forest inventories; stand age, stand density, dominant height, basal area and cultural regime. The proposed equations where used on National Forestry Inventory plot data and the results obtained where compared to the inventory results and estimates obtained using tree level allometric equations and Biomass Expansion Factors (BEF's). The comparison of the four approaches showed that the use of stand level methods gives as accurate results as the use of tree level methods. The use of BEF's resulted in the least accurate method for estimating biomass, with an average mean error of 17%. In contrast, the proposed equations estimate biomass values with a mean error of 9% and are the best option for biomass estimates in short rotation forestry, where the objective is to obtain biomass estimates for short rotation stands (2–5 years). They also allow tree component biomass estimates, something that is not possible using the existing BEF's for Eucalyptus globulus . [ABSTRACT FROM AUTHOR]