Prediction of stem biomass of Pinus caribaea growing in the low country wet zone of Sri Lanka

Abstract

Forests are important ecosystems as they reduce the atmospheric CO2 amounts and thereby control the global warming. Estimation of biomass values are vital to determine the carbon contents stored in trees. However, biomass estimation is not an easy task as the trees should be felled or uprooted which are time consuming and expensive procedures. As a solution to this problem, construction of mathematical relationships to predict biomass from easily measurable variables can be used.

The present study attempted to construct a mathematical model to predict the stem biomass of Pinus caribaea using the data collected from a 26 year old plantation located in Yagirala Forest Reserve in the low country wet zone of Sri Lanka. Due to the geographical undulations of this forest, two 0.05 ha sample plots were randomly established in each of valley, slope and ridge-top areas. In order to construct the model, stem wood density values were calculated by using stem core samples extracted at the breast height point. Stem volume was estimated for each tree using Newton’s formula and the stem biomass was then estimated by converting the weight of the known volume of core samples to the weight of the stem volume. Prior to pool the data for model construction, the density variations along the stem and between geographical locations were also tested.

It was attempted to predict the biomass using both dbh and tree height. Apart from the untransformed variables, four biologically acceptable transformations were also used for model construction to obtain the best model. All possible combinations of model structures were fitted to the data. The preliminary model selection for further analysis was done based on higher R² values and compatibility with the biological reality. Out of those preliminary selected models, the final selection was done using the average model bias and modeling efficiency quantitatively and using standard residual distribution qualitatively. After the final evaluation the following model was selected as the best model to use in the field.