Estimativa de variáveis dendrométricas individuais de Araucaria angustifolia por meio de dados laser scanner aéreo e terrestre
Abstract
Resumo: A tecnologia LiDAR (light detection and ranging) está sendo muito utilizada no setor florestal, porém, os estudos aplicados em florestas nativas do Brasil ainda são incipientes, principalmente em relação ao laser scanner terrestre (TLS). O objetivo principal deste estudo foi estimar variáveis dendrométricas: altura, diâmetro a 1,30 m do solo (dap), volume e biomassa acima do solo de Araucaria angustifolia (Bertol.) Kuntze utilizando dados LiDAR e modelos de estrutura quantitativa. Além disso, estabeleceu-se um protocolo de coleta de dados LiDAR em florestas naturais do Brasil, testou-se a redução da densidade da nuvem de pontos e, ainda, realizou-se a integração de dados de laser scanner aéreo (ALS) e TLS no intuito de obter resultados mais acurados em relação aos obtidos com eles individualmente. O estudo foi desenvolvido em um fragmento de 15,24 ha de Floresta Ombrófila Mista urbana, localizado no Campus III da UFPR. Inicialmente, foram processados os dados ALS, gerou-se o modelo digital do terreno (MDT), modelo digital de superfície (MDS) e canopy height model (CHM) e, a partir disso, foram delimitadas as parcelas a serem inventariadas com o TLS e inventário florestal convencional (IFC). Realizou-se o levantamento TLS, registro da nuvem de pontos, filtragem, segmentação dos indivíduos de araucária e a aplicação do algoritmo TreeQSM para modelagem tridimensional das árvores. Foram estimados valores de diâmetro a 1,30 m do solo (dap) e altura total utilizando o TLS e mensurados em campo (IFC), ao aplicar análise estatística multivariada, por meio do Teste T² de Hotelling, constatou-se que não houve diferença estatística entre os métodos a nível 99% de significância. Obteve-se a raiz do erro quadrático médio (RMSE) de 5,31% entre os dados observados, um viés de 3,32% e uma precisão de 7,94%. Quanto à reamostragem da nuvem de pontos, foram testados os métodos octree, espacial e aleatório, sendo o método octree o único que apresentou perda estrutural das árvores, impossibilitando a modelagem tridimensional com o algoritmo TreeQSM. A integração entre os dados ALS e TLS possibilitou estimar a biomassa acima do solo em função da área de copa mais acurados em relação aos estimados com TLS e ALS individualmente. Os resultados evidenciaram que, tanto o TLS quanto o ALS são ferramentas excelentes para estimativas acuradas de variáveis dendrométricas, assim como a tecnologia LiDAR possibilita a geração de estimativas acuradas da floresta, tanto em micro quanto em macroescalas. Os modelos de estruturas quantitativas são capazes de gerar uma modelagem tridimensional da árvore com grande acurácia, possibilitando que se tenha informações a qualquer altura, bem como de volume e afilamento. Palavras-chave: Mata Atlântica, LiDAR, Sensoriamento Remoto, Floresta nativa, Biomassa acima do solo. Abstract: LiDAR (light detection and ranging) technology have been used a lot in the forestry sector, however, studies applied in native forests in Brazil are still incipient, especially about to the terrestrial laser scanner (TLS). The main objective of this study was to estimate dendrometric variables and above-ground biomass of Araucaria angustifolia using LiDAR data and quantitative structure models. In addition, a LiDAR data collection protocol was established in natural forests in Brazil, the reduction of the point cloud density was tested, and the integration of ALS and TLS data was carried out in order to obtain more accurate results in relation to those obtained with ALS and TLS individually. The study was conducted in a 15.24 ha fragment of mixed urban ombrophilous forest, deominated Capão do Tigre, located on Campus III of UFPR. Initially, ALS data were processed, DTM, DSM and CHM were generated, and from that, the parcels to be inventoried with TLS and IFC were delimited. The TLS survey, point cloud registration, filtering, Araucaria segmentation and application of the TreeQSM algorithm were performed. Dbh and height values were estimated using the TLS and measured in the field (IFC), so that, there was no statistical difference between the methods, an RMSE of 5.31% was obtained between the observed data, a bias of 3, 32% and an accuracy of 7.94%. As for resampling of the point cloud, the octree, spatial and random methods were tested, with the octree method being the only one that presented structural loss of the trees, making three-dimensional modeling with the TreeQSM algorithm impossible. The integration between ALS and TLS data enabled the generation of above-ground biomass estimates as a function of the more accurate canopy area in relation to those estimated with TLS and ALS individually. The results of all chapters showed that both TLS and ALS are excellent tools for estimating dendrometric variables accurately, so that it is concluded that LiDAR technology allows the generation of accurate forest estimates, both in a micro and macro-scale. The models of quantitative structures are capable of generating a three-dimensional modeling of the tree with great accuracy, making it possible to have information at any time, and information on volume and tapering. Keywords: Atlantic Forest, LiDAR, Remote Sensing, Natural forest, aboveground biomass.
Collections
- Teses [347]