Construction of a growth model to predict the individual stem volume of Tectona grandis L.f. (Teak) in Sri Lanka

Abstract

Stem volume is one of the most important variables in commercial forestry since all the managementdecisions are taken on the volume production of trees. Also volume is the most difficult variable tomeasure and therefore it has to be predicted using a reliable method. Therefore, a growth model wasnewly constructed in this study to predict the individual stem volume of teak (Tectona grandis L.f.)planted as even-aged rnonocultures in Svi Lanka.

In order to construct the model, data were collected from 11 plantations in dry zone (Anuradhapura,Harnbatota and Puttlam Forest Divsions) and intermediate zone (Kurunegala Forest Division). Theselected plantations vary in age from 27 to 44 years. Stratified random sampling was used to collectthe data with 0.02 ha circular sample plots. It was decided to use the Newton's formula for volumecalculations and for this reason, the standing tree stems were divided into 3-5 sections and enddiameters, mid diameter and length of each section were measured. The final section was assumedas a cone. The stem volume was then calculated by adding all the section volumes together. Inaddition to that, breast height diameter, total height and crown height were also measured. Usingthose data. tree volume, basal area and top height were calculated which were necessary for theanalysis.

A theoretical model was developed to predict the individual tree volume using the relationship of treevolume with form factor, basal area and total height. Three site factors and four transformationswhich are biologically accepted were used to enhance the quality of the models. When tested for thesite differences using top height/age index, five classes were identified from the 1 J plantations. At thefirst stage of analysis, the data of ench site class were fitted separately to the basic model structures.As J result, eight cornmon models v-e-e i :1c'1tified for all five site classes with R2 over 91.0 and goodstandard residua' d.stributions. However, the regression parameters were different for different siteclasses evei (or the same model structure.

In order to eliminate the difficulty of using different parameter sets for different site classes, datacollected from all plantations were fooled anc' fitted at the second stage to tile selected eight modelsand new parameter sets were estimated. Then the normal residuals were calculated separately forfiv , site classes using those models and tested using one way ANaYA. Only one model indicated thenon-significant residuals for all site types, i.e., V1/2= 0.5730 (basal area * total height')1/2+ 0.0235 [1/(top height / age)]. That model indicated negligible bias (-0.002) and very high modelling efficiency(0.91). When validated with the reserved data which were 110tused for model building, it indicated agood distribution of normal residuals. Finally it was conc!uded that the selected model proved itsability of predicting the unbiased volume of individual teak stems of all site classes in Sri Lanka andtherefore it is recommended to LIse in the field.(Financial support from the University of Sri Jayewardenepura is acknowledged.)