Improving yield projections from early ages in eucalypt plantations with the Clutter model and artificial neural networks

Casas, Gianmarco Goycochea and Fardin, Leonardo Pereira and Silva, Simone and de Oliveira Neto, Ricardo Rodrigues and Breda Binoti, Daniel Henrique and Leite, Rodrigo Vieira and Ramos Domiciano, Carlos Alberto and de Sousa Lopes, Lucas Sérgio and da Cruz, Jovane Pereira and dos Reis, Thaynara Lopes and Leite, Hélio Garcia. (2022) Improving yield projections from early ages in eucalypt plantations with the Clutter model and artificial neural networks. Pertanika Journal of Science & Technology (Malaysia), 30 (2). 1257 -1272. ISSN 2231-8526

A common issue in forest management is related to yield projection for stands at young ages. This study aimed to evaluate the Clutter model and artificial neural networks for projecting eucalypt stands production from early ages, using different data arrangements. In order to do this, the changes in the number of measurement intervals used as input in the Clutter model and artificial neural networks (ANNs) are tested. The Clutter model was fitted considering two sets of data: usual, with inventory measurements (I) paired at intervals each year (I₁–I₂,I₂–I₃,…, Iₙ–Iₙ+₁); and modified, with measurements paired at all possible age intervals (I₁–I₂, I₁–I₃,…, I₂–I₃,I₂–I₄,…, Iₙ–Iₙ+₁). The ANN was trained with the modified dataset plus soil type and geographic coordinates as input variables. The yield projections were made up to the final ages of 6 and 7 years from all possible initial ages (2, 3, 4, 5, or 6 years). The methods are evaluated using the relative error (RE%), bias, correlation coefficient (ryŷ), and relative root mean square error (RMSE%). The ANN was accurate in all cases, with RMSE% from 8.07 to 14.29%, while the Clutter model with the modified dataset had values from 7.95 to 23.61%. Furthermore, with ANN, the errors were evenly distributed over the initial projection ages. This study found that ANN had the best performance for stand volume projection surpassing the Clutter model regardless of the initial or final age of projection.

Full text available from: Official URL: http://www.pertanika.upm.edu.my/resources/files/Pe...

Abstract

A common issue in forest management is related to yield projection for stands at young ages. This study aimed to evaluate the Clutter model and artificial neural networks for projecting eucalypt stands production from early ages, using different data arrangements. In order to do this, the changes in the number of measurement intervals used as input in the Clutter model and artificial neural networks (ANNs) are tested. The Clutter model was fitted considering two sets of data: usual, with inventory measurements (I) paired at intervals each year (I₁–I₂,I₂–I₃,…, Iₙ–Iₙ+₁); and modified, with measurements paired at all possible age intervals (I₁–I₂, I₁–I₃,…, I₂–I₃,I₂–I₄,…, Iₙ–Iₙ+₁). The ANN was trained with the modified dataset plus soil type and geographic coordinates as input variables. The yield projections were made up to the final ages of 6 and 7 years from all possible initial ages (2, 3, 4, 5, or 6 years). The methods are evaluated using the relative error (RE%), bias, correlation coefficient (ryŷ), and relative root mean square error (RMSE%). The ANN was accurate in all cases, with RMSE% from 8.07 to 14.29%, while the Clutter model with the modified dataset had values from 7.95 to 23.61%. Furthermore, with ANN, the errors were evenly distributed over the initial projection ages. This study found that ANN had the best performance for stand volume projection surpassing the Clutter model regardless of the initial or final age of projection.

[error in script]

Item Type:	Article
AGROVOC Term:	Eucalyptus
AGROVOC Term:	yield forecasting
AGROVOC Term:	neural networks
AGROVOC Term:	artificial intelligence
AGROVOC Term:	forest management
AGROVOC Term:	research
AGROVOC Term:	regression analysis
AGROVOC Term:	growth models
Geographical Term:	Brazil
Uncontrolled Keywords:	Artificial intelligence, data structure, forest growth and yield, forest management, regression
Depositing User:	Ms. Azariah Hashim
Date Deposited:	12 Nov 2024 02:27
Last Modified:	12 Nov 2024 02:27
URI:	http://webagris.upm.edu.my/id/eprint/1744

Actions (login required)

View Item