S(B) distribution’s accuracy to represent the diameter distribution of Pinus taeda, through five fitting methods

• Published in: Forest ecology and management Vol. 175
• Main Authors: Scolforo, José Roberto Soares, Tabai, Fernando César Vitti, Macedo, Renato Luı́s Grisi de, Acerbi, Fausto Weimar, Assis, Adriana Leandra de
• Format: Journal Article
• Language: English
• Published: 2003
• Subjects: linear models; Pinus taeda; prediction; prognosis; spatial distribution; trees; Brazil
• ISSN: 0378-1127; 1872-7042

The objective of this study was to verify the SB distribution’s accuracy for estimating the diameter distribution of Pinus taeda by using five different fitting methods and to analyze the influence of age and spacing on these adjustments. A spacing experiment of P. taeda located in the county of Jaguariaı́va, state of Paraná, Brazil, was used. The spacing variants were 4.4 m x 2.5 m (11 m(2)); 3.6 m x 2.5 m (9m(2)); 2.8 m x 2.5 m(7 m(2)); 2.0 m x 2.5 m (5 m(2)) and 1.2 m x 2.5 m (3 m(2)). The trees were measured at ages 3.5, 4.5, 5.9, 6.9, 7.7, 8.8 and 9.8 years. The S(B) distribution was fitted to different spacing variants and measurement ages using the following methods: Maximum Likelihood, Moments, Knoebel–Burkhart, and Mode and Linear Regression. The theoretical and observed frequencies adherence was performed by the Kolmogorov–Smirnov (KS) test. A ranking procedure was adopted to establish which method gave the most accurate fitting to those plots or measurement periods that showed similarity by the KS test between theoretical and observed frequencies. All tested methods can be used to estimate the P. taeda stand’s theoretical frequency, regardless of the age and the spacing. The Moment’s and the Maximum Likelihood were the methods which provided the best fitting when the ε value was equal to 25% of the minimum diameter. Moment’s method was the best choice to estimate P. taeda current theoretical frequency, independently from stand age and spacing. All methods, regardless of the spacing, displayed at least one parameter responsible for the distribution shape, being slightly correlated with age. This fact could affect the yield prognosis. Linear Regression and Mode presented their parameters correlated with age, thus being the most suitable method for the yield prediction.