Comparison Of Fitting Results And Calculation Efficiency Of Stem Taper Equations

This paper compares trigonometric variable-form taper function with generalized Brink stem profile function when estimation predicting diameter under (over) bark at any height of stem using sample taper data collected from Pinus radiata plantations in New South Wales, Australia. Nonlinear least squares regression is used when estimated parameters of both functions. The mean error, mean square error and mean absolute different of prediction, as well as determination coefficient are chosen as the criteria for evaluation. The results demonstrate that:predict results of stem diameter by the two taper functions with local bias within any stem section. The accuracy of trigonometric variable-form taper function performs better than generalized Brink stem profile function in predicting diameter at any height of new stems after the parameters in the two taper functions are obtained by regression, also its bottom stem fitting effect is superior to generalized Brink stem profile function and is more stable. According to the problems existing in the process of parameters estimation and parameter estimation results, this paper also discusses the possibility of extending the adaptability of generalized Brink stem profile function and simplifying trigonometric variable-form taper function when using them to estimate artificial pure forest. Then Acacia mangium willd stem data are used to validate the results.Due to the two stem taper functions are not reversible, this paper adopts dichotomy and Newton-Raphson methods to design numerical calculation algorithms which can calculate an appropriate height corresponding to the given top end diameter respectively for the generalized Brink stem profile function and the trigonometric variable-form taper function, and implemented the algorithms by using Matlab programming. After that, the dichotomy algorithm was optimized after comparing the two convergence speed to reach the level of Newton-Raphson algorithm so that their computation speed could be guaranteed in realistic application.