CONSTRUCTION OF A GROWTH MODEL TO PR~DICT THE INDIVIDUAL STEM VOLUME OF Eucalyptus grandis HILL EX MAIDEN (TOOLUR) IN SRI LANKA

Abstract

Toolur (Eucalyptus gram/is Hill ex Maiden) belongs to family Myrtaceae has been plantedwidely in the hill country of Sri Lanka for high quality timber and fuel wood. For thepresent study a growth model was constructed to predict the total stem volume ofindividual toolur trees. Data were obtained using temporally sampled plots from the treeplantations in Badulla forest division and 6 plalltations.~:. Nuwara-Eliya forest division.

Data were first partitioned according to the quality of the site. An index developed using atop height / plantation age was used for site classification and three significantly differentsites could be identified. In order to construct the model, 75% of sample plots from eachplantation was used and the rest was reserved for the purpose of validating the constructedmodel with independent data.

The basic model structure was developed using the relationship between volume and formfactor, basal area and total height. Form factor is difficult to measure because it varies withage, site quality, depsHyetc. ~ven!or the same species. Therefore keeping the other twovariables, it was tried to replacctii.cfortnfactQLH~ing. other explanatory variables, whichcan be measured easily.

All possible combinations of the explanatory variables were tested ill order to obtain thebest models. The performance of the models was observed using statistical tests andstandard residual distributions. For the selected models, the intercept was not significantlydifferent from zero and number of trees per hectare and total basal area tested to representthe stand density was also not significant.

After selecting the four suitable models at the first stage, the possibility of using one set ofparameters instead of separate parameters sets for different site types was tested. First theparameters were estimated for the pooled 'data. Using the resulted models, normal residualswere calculated separately for three site types. Then the significance of the residuals wastested using analysis of variance for select the final model with common set of parameters.A selected model contained basal area and total height Q~ individual trees and a site qualityfactor as explanatory variables. It was then compared -directly with the particular observedvalues. After all the tests, the results indicated that tile model coustructed in this studyproved its strength on predicting the total volume of individual stems precisely because itindicated a very low bias and a high modeling efficiency of over 0.09.

The resultant model which can be used to predict the individual toolur stem volume for allsite types is:(Volume/12=0.5946*(basal area * total height )1/2 + O.0356*(total basal area / lop height).