Fitting Volume Models for Parana Pine With a Nonlinear Regression, Genetic Algorithm and Simulated Annealing

Emanuel Arnoni Costa,Cristine Tagliapietra Schons,César Augusto Guimarães Finger,André Felipe Hess

doi:10.5539/jas.v14n2p36

Emanuel Arnoni Costa, Cristine Tagliapietra Schons + Show 2 more

Open Access

https://doi.org/10.5539/jas.v14n2p36

Copy DOI

Abstract

Improving volumetric quantification of Parana pine (Araucaria angustifolia) in Mixed Ombrophilous Forest is a constant need in order to provide accurate and timely information on current and future growing stock to ensure forest management. Thus, the present study aimed to evaluate and compare the volume estimates obtained through Nonlinear Regression (NR), Genetic Algorithm (GA) and Simulated Annealing (SA) in order to generate accurate volume estimates. Volumetric equations were developed including the independent variables diameter at breast height (dbh), total height (h) and crown rate (cr) and from the fit through the NR, GA and SA approaches. The GA and SA approaches evaluated proved to be a reliable optimization strategy for parameter estimation in Parana pine volumetric modelling, however, no significant differences were found in comparison with the NR approach. This study therefore contributes through the generation of robust equations that could be used for accurate estimates of the volume of the Parana pine in southern Brazil, thus supporting the planning and establishment of management and conservation actions.

Highlights

Timber valuation is a prime requirement to ensure forest management
In the evaluation of the models generated by the different approaches, we considered as goodness-of-fit criteria the root mean square error (RMSE) (Expression vi), the Bias (B) (Expression vii), the Akaike’s information criterion (AIC) (Expression viii) and the graphical analysis of the residuals
The coefficients fitted through Nonlinear Regression (NR), Genetic Algorithm (GA) and Simulated Annealing (SA), followed by their respective values RMSE, B and AIC, are presented in the Table 3, considering each based volumetric model

Summary

Introduction

Timber valuation is a prime requirement to ensure forest management. In this regard, forest managers constantly resort to estimating forest attributes in order to provide accurate and timely information on current and future growing stock and to assess the economic benefits of their forests, at large scale industrial forests and at small scale forests (Tiryana et al, 2021). The development of accurate prediction tools is crucial to support forest management decisions, which have to be adapted to suit the particular circumstances (Özçelik et al, 2010). In the early 1900s, it is estimated that this species covered approximately 200,000 km in the south of the country (Hueck 1972)

Methods

Results

Discussion

Conclusion