Automação de inventários florestais com terrestrial laser scanning : um enfoque em integração lavoura-pecuária-floresta
Resumo
Resumo: O uso do Light Detection and Ranging (LiDAR) em inventários florestais tem se mostrado cada vez mais eficiente. Essa tecnologia permite a aquisição de informações detalhadas sobre a estrutura das árvores sem a necessidade de cortálas, preservando a integridade do ecossistema florestal, o que é particularmente útil em sistemas agroflorestais, onde as árvores frequentemente possuem grande porte e valor econômico. O uso do Terrestrial Laser Scanner (TLS) para obter diâmetros ao longo do fuste das árvores enfrenta desafios devido à oclusão causada pela presença de galhos e folhas na porção da copa. Quando a copa das árvores é densa, os pulsos podem ser bloqueados pelos galhos e folhagem, resultando em distorções nas medidas dos diâmetros, o que compromete as estimativas dos volumes das árvores. Nesse sentido, o objetivo desta tese consistiu em superar essa limitação, tornando as variáveis extraídas das varreduras de TLS aplicáveis na estimativa do volume individual total e de sortimentos de madeira. Para atingir o objetivo proposto, foi incorporado neste estudo uma abordagem multifacetada. Em um primeiro momento, foram desenvolvidas e propostas rotinas de processamento específicas para os dados de TLS obtidos em plantios de Eucalyptus benthamii em um sistema de integração Lavoura-Pecuária-Floresta (iLPF). Na sequência, foi avaliado o desempenho das estimativas do volume individual total das árvores com base em um menor número de medições do fuste, até uma boa representação sem influência de galhos. E por último, foram propostas metodologias para estimar os diâmetros nas diferentes alturas do fuste, acima de cinco metros, tendo como base os diâmetros e alturas obtidos pelo TLS na porção do fuste sem a presença de galhos, visando quantificar os sortimentos de madeira. A hipótese de que uma representação parcial do fuste das árvores, obtida por meio do TLS, pode estimar com acurácia o volume total individual das árvores foi confirmada pelas estatísticas de avaliação das estimativas, medidas de concordância e gráficos de concordância de Bland-Altman. Os resultados indicam que quanto maior o volume parcial utilizado na estimativa do volume total individual, mais exata será a estimativa. No entanto, estimativas de volume individual baseadas em volumes parciais de secções menores do fuste também são estatisticamente satisfatórias e podem ser usadas com confiança nas estimativas dos volumes individuais. A utilização de varreduras de TLS para estimar os diâmetros ao longo do fuste das árvores em inventários florestais é uma abordagem promissora. No contexto da área de estudo, os resultados obtidos pelo Método 03 proposto não diferiram dos resultados do Método 01, tradicionalmente utilizado na área florestal. Em essência, o processo gerador de dados pelo Método 03, seguido pelo ajuste do Polinômio de 5º Grau e estimativa dos sortimentos, é equivalente ao método tradicional que envolve a derrubada e cubagem das árvores. Utilizando a cubagem via dados de TLS, foi possível superar as limitações decorrentes da oclusão, que afetam a obtenção precisa dos diâmetros ao longo do fuste das árvores. Sugere-se que o método proposto (Método 03) seja aplicado em outras condições de plantio para o aprimoramento da metodologia. Abstract: The use of Light Detection and Ranging (LiDAR) in forest inventories has proven to be increasingly effective. This technology allows for the acquisition of detailed information about the structure of trees without the need to cut them down, preserving the integrity of the forest ecosystem. This is particularly useful in agroforestry systems, where trees often have large size and economic value. The use of Terrestrial Laser Scanner (TLS) to obtain diameters along the trunk of trees faces challenges due to occlusion caused by the presence of branches and leaves in the crown portion. When the tree canopy is dense, the pulses can be blocked by the branches and foliage, resulting in distortions in the diameter measurements, which compromises the estimates of tree volumes. In this regard, the goal of the current work was to overcome this limitation, making the variables extracted from TLS scans applicable in estimating the total individual volume and assortments of wood. To achieve the proposed objective, in this study it was adopted a multifaceted approach. Initially, specific processing routines were developed and proposed for TLS data obtained from Eucalyptus benthamii plantations in a Crop-Livestock-Forestry Integration System (iCLF). Subsequently, the performance of the estimates of the total individual volume of the trees was evaluated based on a smaller number of trunk measurements, until a good representation without the influence of branches was achieved. Finally, methodologies were proposed to estimate the diameters at different heights of the trunk, above five meters, based on the diameters and heights obtained by the TLS in the portion of the trunk without branches, aiming to quantify the assortments of wood. The hypothesis that a partial representation of the tree trunk, obtained through TLS, can accurately estimate the total individual volume of the trees was confirmed by the evaluation statistics of the estimates, measures of agreement, and Bland-Altman concordance graphs. The results suggest that the larger the partial volume used in estimating the total individual volume, the more accurate the estimate will be. However, estimates of individual volume based on partial volumes from smaller sections of the stem are also statistically satisfactory and can be confidently used in estimating individual volumes. The use of TLS scans to estimate diameters along the trunk of trees in forest inventories is a promising approach. In the context of the study area, the results obtained by the proposed Method 03 did not differ from the results of Method 01, which is traditionally used in the forestry area. Essentially, the data generation process by Method 03, followed by the adjustment of the 5th Degree Polynomial and the estimation of assortments, is equivalent to the traditional method that involves felling and scaling the trees. Using scaling through TLS data, it was possible to overcome the limitations resulting from occlusion, which affect the precise measurement of diameters along the tree trunk. It is suggested that the proposed method (Method 03) be applied under different planting conditions to refine the methodology.
Collections
- Teses [387]