Characterising truncated tree diameter data from tropical forest stands in Nigeria

Abstract

In forest inventory, diameter threshold is often set during data collection and as such lead to left truncation. Such data are preferably fitted with a truncated distribution. In this study, newly introduced truncated Log-Logistic and well-known truncated Weibull were used for characterising tree diameter from natural forest. The truncated distributions were compared with Burr, Logit-Logistic, generalized Weibull, Johnson’s S_B, 3-parameter Log-Logistic, 3-parameter Weibull, 2-parameter Log-Logistic and 2-parameter Weibull distributions. Data were obtained from 11, 10, 10 and 7 sample plots of size 0.25 ha in Akure, Cross River South, Ikrigon and Oluwa forest reserves, respectively. Distributions were fitted with maximum likelihood. Model assessment was based on Kolmogorov-Smirnov, Reynold’s error index, mean square error, mean absolute error and Akaike information criterion. The result showed that the performance of the truncated distributions was comparable to other distributions considered in this study. The truncated distributions were more suitable than the commonly used Weibull and Johnson’s SB distributions. Thus, whenever a diameter threshold is fixed, truncated distribution should be used for modelling the data set.

Keywords: truncated distributions; Akure forest reserve; Cross River South forest reserve; Ikrigon forest reserve; Oluwa forest reserve