CONSTRUCTION OF A PRECISE GROWTH MODEL TO PREDICT THE INDIVIDUAL TREE STEM VOLUME OF Tectona grandis L.f. (TEAK) IN A 29 YEAR OLD PLANTATION IN MIHINTALE IN ANURADHAPURA DISTRICT

Abstract

Attached

Stem volume is the most important variable in commercial forestry because all the management decisions are taken on the volume production of trees. However, it is also the most difficult variable to measure and therefore it is necessary to accurate volume prediction methods. For the present study a growth model was constructed to predict the stem volume of individual Tectona grandis L.f. (TEAK) trees of the 29 year old plantation (ID No: Block 01, Sub-block 29) in Mihintale Beat of the Anuradhapura Forest Division. The age of this even-aged plantation was 29 year and the size was 34.0 ha. In order to collect the data, ten 0.02 ha circular sample pots were randomly laid. Diameter at breast height (dbh), total height and crown height of the trees in all the sample plots were measured as the first step. Tree basal area, stand basal area and top height were calculated using these data. For the second step of data collection, each tree stem was divided into 3-5 m sections without felling them using the Blume-Leiss altimeter. Then the bottom, middle and top diameters of each section were measured using Spiegal relascope. The volume of each section was calculated separately using Newton's formula and the stem volume was determined by summing the section volumes together. For this reason, the final section of the tree was considered as a cone. The sample plot data were divided into two as construction (75%) and validation (25%) and the latter was not used for building the model. A theoretical model was developed to predict the individual tree volume using the relationship of form factor with volume, basal area and total height. It was fitted to the collected data using multiple linear regression in MINIT AB. Three site factors and four transformations which are biologically accepted were used to enhance the quality of the models After fitting 13 models were selected for further analysis due to their high R2 values which were over 85% and good distribution of standard residuals. For these selected models, average model bias and modelling efficiency were tested to select the best model. The biases indicated by all the models were insignificant and the model with the highest modelling efficiency (0.982) was selected for the field use. When the final model was validated with independent data reserved at the beginning of the model construction, the results proved the ability of using the selected model in the field without producing errors. The finally selected model for the field use is .rv = 0.0567.Jba * lit + 0.00356topht.