Modelagem de efeito misto para calibração de equações na predição de altura e volume em povoamentos de Pinus spp.
Visualizar/ Abrir
Data
2024Autor
Woycikievicz, Antonio Pedro Fragoso
Metadata
Mostrar registro completoResumo
Resumo: As empresas florestais conduzem inventários para quantificar e qualificar variáveis, que servem de base para monitorar o crescimento e a produção florestal. A coleta de dados nos inventários florestais é realizada de diversas formas e intensidades nos povoamentos florestais, o que existe em comum é a geração de um extenso conjunto de dados e informações dos povoamentos. No que diz respeito à redução do tamanho da amostra para modelagem de altura e volume, os modelos de efeito misto têm sido cada vez mais aplicados na Ciência Florestal, pois permitem o ajuste de novas equações com um número reduzido de árvores sem prejuízo nas estatísticas de ajuste, por meio da Calibração. Dentro da Calibração surgem três principais preocupações: i) Definição dos níveis hierárquicos; ii) Tamanho da amostra e critério para seleção de Árvores; iii) Manutenção das estatísticas de ajuste das equações. Diante disso, a pesquisa visa responder às seguintes perguntas: 1) Qual é o tamanho da amostra e quais árvores devem ser selecionadas para calibrar equações de altura e volume para espécies do gênero Pinus? 2) Qual é a qualidade das estimativas de equações de altura e volume calibradas em relação à modelagem tradicional? O objetivo geral da pesquisa foi testar modelos hipsométricos e volumétricos por meio da calibração de equações lineares mistas, com o teste de seleção de até dez árvores com 15 diferentes critérios de seleção. Os povoamentos são de três diferentes espécies do gênero Pinus (Pinus taeda – altura e volume; Pinus caribaea var. hondurensis – volume; Pinus maximinoii - altura), localizados no Norte de Santa Catarina e no Centro-Oeste do Paraná. A base para a modelagem hipsométrica foi composta por mais de 200.000 observações e para a modelagem volumétrica foi constituída de mais 2.000 observações. Foi selecionado um modelo, tanto para o volume como para a altura, que foi aplicado à modelagem mista ao nível de povoamento, e selecionando o melhor método e quantidade de árvores para calibrar a nova equação. Para a diferença entre as estatísticas das equações calibradas e a equação tradicional foi estabelecido um limite de 5%. Na calibração das equações de volume, para a espécie Pinus caribaea var. hondurensis é recomendada a seleção de uma árvore com diâmetro à altura do peito (DAP) superior à média mais um desvio padrão, e para a espécie Pinus taeda, a seleção de sete árvores (critério J7). Na calibração das equações de altura, para o Pinus maximinoii e o Pinus taeda a seleção indicada é de nove árvores, sendo que três devem possuir diâmetro inferior à média menos um desvio padrão, outras três devem ter diâmetro entre a média mais ou menos um desvio padrão, e as três restantes devem apresentar diâmetro superior à média mais um desvio padrão. Um plano de calibração foi elaborado para aplicação da metodologia. Um aplicativo (CAPLIN) foi desenvolvido para a calibração de novas populações de Pinus em nível de povoamento, para as três espécies avaliadas na pesquisa. Abstract: Forest companies carry out inventories to quantify and qualify variables that serve as a basis for monitoring forest growth and production. The collection of data in forest inventories is carried out in different ways and at different intensities in forest stands, but what they all have in common is the generation of an extensive data set and information on the stands. To reduce the sample size for modelling height and volume, mixed effect models have been increasingly applied in Forest Science, as they allow new equations to be adjusted with a reduced number of trees without affecting the adjustment statistics, by means of Calibration. In Calibration, three main concerns arise: i) Defining the hierarchical levels; ii) Sample size and criteria for selecting trees; iii) Maintaining the fit statistics in the equations. In view of this, the research aims to answer the following questions: 1) What is the sample size and which trees should be selected to calibrate height and volume equations for species of the genus Pinus? 2) What is the quality of the estimates of calibrated height and volume equations compared to traditional modeling? The general objective of the research was to test height–diameter and volumetric models by calibrating mixed linear equations, with the test of selecting up to ten trees with 15 different selection criteria. The stands are of three different species of the genus Pinus (Pinus taeda - height and volume; Pinus caribaea var. hondurensis - volume; Pinus maximinoii - height), located in northern part of Santa Catarina and central-western part of Paraná. The basis for the height– diameter modeling was made up of more than 200,000 observations and for volumetric modeling added with 2,000 observations. A model was selected for both volume and height, which was applied to mixed modeling at the stand level, and the best method and number of trees were selected to calibrate the new equation. A limit of 5% was set for the difference between the statistics of the calibrated equations and the traditional equation. In calibrating the volume equations for the Pinus caribaea var. hondurensis species, it is recommended to select one tree with a diameter at breast height (DBH) greater than the mean plus one standard deviation, and for the Pinus taeda species, to select seven trees (criterion J7). In the calibration of the Height equations, for Pinus maximinoii and Pinus taeda the selection indicated is nine trees, three of which must have a diameter below the mean minus one standard deviation, another three must have a diameter between the mean plus or minus one standard deviation, and the remaining three must have a diameter above the mean plus one standard deviation. A calibration plan was structured to apply the methodology. An application (CAPLIN) was developed for the calibration of new Pinus populations at stand level, for the three species evaluated in the research.
Collections
- Teses [387]