GROWTH AND YIELD PROGNOSIS MODELING OF EUCALYPTUS AT DIFFERENT APPROACH LEVELS.
Name: CLAYTON VIEIRA FRAGA FILHO
Publication date: 04/02/2016
Advisor:
Name |
Role |
|---|---|
| ADRIANO RIBEIRO DE MENDONÇA | Co-advisor * |
| GILSON FERNANDES DA SILVA | Advisor * |
Examining board:
| Name |
Role |
|---|---|
| ADRIANO RIBEIRO DE MENDONÇA | Co advisor * |
| GILSON FERNANDES DA SILVA | Advisor * |
Summary: FRAGA FILHO, Clayton Vieira. GROWTH AND YIELD PROGNOSIS MODELING
OF EUCALYPTUS AT DIFFERENT APPROACH LEVELS. 2016. Thesis (Doctor in
Forest Sciences) Federal University of Espírito Santo, Jerônimo Monteiro, ES.
Advisor: Prof. Dr. Gilson Fernandes da Silva. Co-adivisor: Prof. Dr. Adriano Ribeiro de Mendonça.
Growth and yield models are efficient and reliable to describe forests dynamics and perform the prediction of long-term production. There are different approach levels for these models, such as whole-stand models, diameter distribution, and individualtree. This work aimed to compare different levels for the modeling of the growth and production in terms of the accuracy and bias level of the production estimates in eucalyptus stands, with fixed and mixed models. Two criteria were used for dividing the data. In criterion 1, the data were splited into two groups: the first for fitting plots with future age between 24 and 60 months, and the second for projection, having plots with future age between 72 and 96 months. In the second criterion, 70% of the
plots were used to models fitting and 30% for the validation of these models.
Alternatives at the whole-stand level were proposed, being two of them fixed model alternatives to criterion 1, and two for criterion 2. In addition, three mixed modeling alternatives were proposed for criterion 1 and seven alternatives for criterion 2.According to analyzes for fixed models, the Logistics model was more accurate and less biased. Mixed modeling projection presented superior results compared to fixed effects modeling. Furthermore, an evaluation was performed with fixed and mixed modeling alternatives for diameter distribution level. A fixed model was evaluated for both criteria: an alternative using mixed modeling was proposed for criterion 1 and three alternatives for criterion 2. The conclusion was that alternatives adjusted and
validated with criterion 2 data proved better, and the diameters distribution with mixed model was more accurate and less biased compared with approaches with fixed effects models. From these results, fixed and mixed alternatives were evaluated at the individual-tree level. Two fixed models were analyzed for criterion 1 and one fixed model for criterion 2. Five mixed models were proposed for criterion 1 and six models for criterion 2. Among the fixed models, the linear model (criterion 1) was better; however, it was utperformed by mixed modeling in both fit and validation (criteria 1 and 2). Finally, the most accurate and less biased models previously obtained were compared. For performing the projection (criterion 1) using fixed modeling, the Logistic model is more accurate and less biased. The linear mixed model of individual-tree proposed in this work was better than Logistic model. For criterion 2, the Clutters adapted mixed model was better. Thus, it was concluded in this work that modeling at individual-tree level produced similar results as the holestand
level, being both more accurate than the results found for the diameter distribution models. It was also concluded that the mixed modeling at these three levels provided gains in accuracy and lack of bias in prognosis of growth and yield modeling for eucalyptus stands.
Keywords: Forest management, Growth and yield models, Mixed models.
