Classificação de estágios sucessionais da vegetação secundária na floresta ombrófila mista : um enfoque de métricas UAV-LiDAR E LiDAR MLS
Visualizar/ Abrir
Data
2025Autor
Santos, Lucas Henderson de Oliveira
Metadata
Mostrar registro completoResumo
Resumo: A presente tese investigou o potencial de diferentes tecnologias LiDAR para classificação e caracterização de estágios sucessionais em remanescentes da Floresta Ombrófila Mista (FOM) no sul do Brasil. O estudo integrou abordagens de sensoriamento remoto de alta resolução e análises ecológicas para testar metodologias robustas de classificação sucessional, essenciais para o monitoramento e gestão de ecossistemas florestais em regeneração. A pesquisa foi estruturada em cinco capítulos complementares. Inicialmente, uma revisão bibliográfica estabeleceu os fundamentos teóricos sobre sucessão florestal, estrutura da vegetação e sensoriamento remoto aplicado a florestas nativas, contextualizando a problemática e justificativa do estudo. O segundo capítulo explorou a aplicação de dados UAV- LiDAR em 30 parcelas permanentes distribuídas em diferentes estágios sucessionais (inicial, médio e avançado), onde foram derivadas métricas estruturais, incluindo variáveis discretas e simuladas full waveform (FWF). A análise exploratória multivariada revelou padrões estruturais distintos entre os estágios, com perfis verticais evoluindo de estruturas simples no estágio inicial para configurações multimodais e estratificadas no estágio avançado. O terceiro capítulo apresentou uma revisão sistemática sobre o uso da tecnologia Mobile Laser Scanning (MLS) em inventários florestais, analisando 56 artigos científicos publicados entre 2013 e 2023. Esta revisão documentou a evolução do MLS como ferramenta de alta precisão para coleta de dados florestais, sua crescente integração com algoritmo Random Forest e as limitações técnicas em ambientes naturais complexos. No quarto capítulo, foi utilizada uma abordagem de classificação sucessional utilizando sistema portátil de varredura a laser móvel (LiBackpack) em 12 parcelas. Por meio dos resultados foram inferidos padrões, identificando vantagens e limitações específicas da plataforma para caracterização estrutural dos estágios sucessionais da FOM. Para classificação dos estágios sucessionais, foi utilizado o algoritmo Random Forest e abordagens descritivas das variáveis. As variáveis mais importantes para discriminação sucessional incluíram métricas de altura, complexidade estrutural e penetração do dossel, evidenciando a complementaridade entre diferentes tipos de dados LiDAR. A análise ecológica dos resultados confirmou que as métricas LiDAR capturam efetivamente os processos de desenvolvimento estrutural ao longo da sucessão florestal. As considerações finais integraram os resultados dos estudos com UAV-LiDAR e MLS, discutindo suas implicações práticas para monitoramento florestal e planejamento ambiental. A tese demonstrou que dados LiDAR, quando analisados por meio de metodologias que integram conhecimento ecológico e técnicas avançadas de aprendizado de máquina, permitem classificação acurada e objetiva de estágios sucessionais em FOM, oferecendo potencial para aplicação em programas de monitoramento, restauração e conservação Abstract: This thesis investigated the potential of different LiDAR technologies for the classification and characterization of successional stages in remnants of the Mixed Ombrophilous Forest (MOF) in southern Brazil. The study integrated high-resolution remote sensing approaches and ecological analyses to test robust successional classification methodologies, which are essential for the monitoring and management of regenerating forest ecosystems. The research was structured into five complementary chapters: Chapter 1: A literature review established the theoretical foundations of forest succession, vegetation structure, and remote sensing applied to native forests, providing context for the study's problem and justification. Chapter 2: Explored the application of UAV-LiDAR data across 30 permanent plots distributed among different successional stages (initial, intermediate, and advanced). Structural metrics were derived, including discrete variables and simulated Full Waveform (FWF) data. Multivariate exploratory analysis revealed distinct structural patterns, with vertical profiles evolving from simple structures in the initial stage to multimodal and stratified configurations in the advanced stage. Chapter 3: Presented a systematic review of Mobile Laser Scanning (MLS) technology in forest inventories, analyzing 56 scientific articles published between 2013 and 2023. This review documented the evolution of MLS as a high-precision tool for forest data collection, its increasing integration with the Random Forest algorithm, and the technical limitations encountered in complex natural environments. Chapter 4: Utilized a successional classification approach using a handheld mobile laser scanning system (LiBackpack) across 12 plots. The results allowed for the inference of patterns, identifying specific advantages and limitations of the platform for the structural characterization of MOF successional stages. The Random Forest algorithm and descriptive variable approaches were used for classification. The most significant variables for successional discrimination included height metrics, structural complexity, and canopy penetration, highlighting the complementarity between different types of LiDAR data. The ecological analysis confirmed that LiDAR metrics effectively capture structural development processes throughout forest succession. The final considerations integrated the results from the UAV-LiDAR and MLS studies, discussing their practical implications for forest monitoring and environmental planning. This thesis demonstrated that LiDAR data— when analyzed through methodologies integrating ecological knowledge and advanced machine learning techniques—allows for the accurate and objective classification of successional stages in the Mixed Ombrophilous Forest, offering significant potential for application in monitoring, restoration, and conservation programs
Collections
- Teses [300]