Evaluation of four methods of fitting Johnson's SBB for height and volume predictions

Abstract

Johnson's SBB is the most commonly used bivariate distribution model in forestry. There are different methods of fitting Johnson's distribution, and their accuracies differ. In this article, the method of conditional maximum likelihood (CML), moments, mode and Knoebel and Burkart (KB) were used to fit Johnson's SBB distribution. A total of 4,237 diameter and height data obtained from 90 sample plots of Eucalyptus camaldulensis Dehnhardt were used. Evaluation was based on tree height and volume predictions. The predicted and observed tree heights and volumes were compared using the paired sample t-test. The average relative (%) bias and root mean square error of heights and volumes were computed for the four methods. The results showed that CML- and moments-based methods were more suitable than KB and mode methods for predicting tree height and volume. The level of significance and percentage bias were much lower in CML and moments. The mode-based method had the worst performance. The ranking order was: CML ≍ moments > KB > mode.

Keywords

• conditional maximum likelihood
• moments
• mode
• knoebel and burkhart
• eucalyptus camaldulensis