Estimativa de biomassa florestal integrando dados Landsat-8 e fotogrametria aérea digital em diferentes escalas : um estudo de caso em uma bacia hidrográfica no bioma Mata Atlântica

• Ano de defesa: 2021
• Autor(a) principal: Neves, Karina Milagres
• Orientador(a): Almeida, André Quintão de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Programa de Pós-Graduação: Pós-Graduação em Recursos Hídricos
• Palavras-chave em Português: Recursos hídricos; Biomassa florestal; Levantamentos florestais; Ciclo do carbono; Sensoriamento remoto; Modelos matemáticos; Bacias hidrográficas; Bacia Rio Poxim (SE); Redução de Emissões por Desmatamento e Degradação Florestal (Programa); Carbono; REDD+
• Palavras-chave em Inglês: Structure from Motion (SfM); Fourier; Carbon; Remote sensing
• Área do conhecimento CNPq: ENGENHARIAS::ENGENHARIA SANITARIA
• Link de acesso: http://ri.ufs.br/jspui/handle/riufs/17204

The estimation of aboveground biomass (AGB) is essential to guide the actions of programs to reduce deforestation, degradation and the global monitoring of the carbon cycle. In this context, it is extremely important to develop reliable and consistent AGB estimation models for monitoring. Recent advances in the combination of threedimensional and multispectral data obtained by Remote Sensing have obtained promising results to improve the ability to estimate AGB at large scales, however, studies analyzing improvements in Atlantic Brazilian Forest secondary forests have not yet been observed. The main objective of this study was to estimate the AGB of forest fragments of the Atlantic Brazilian Forest, by multispectral imaging and by 3D products obtained by digital aerial photogrammetry (DAP). The second objective was to develop a multiscale approach to estimate AGB using forest inventory, DAP and Landsat-8 (L8), for the fragments of the Poxim-SE river basin. Initially, a forest inventory was conducted in 30 plots, 0.25 ha each, to estimate AGB values. To estimate AGB from a multispectral image, multispectral orbital data from the L8 satellite OLI sensor were selected and vegetation indices and texture metrics were calculated for each plot. Spectral bands, vegetation indices and texture metrics were used as predictor variables for modeling. To obtain the 3D DAP data, a flight with a Unmanned Aerial Vehicles imagery (UAV) was performed, later a 3D point cloud and a digital terrain model (DTM) were generated for its normalization. Fourier metrics and traditional height-based metrics were extracted for each plot, and used as predictor variables. AGB estimation was performed by multiple linear regression fit. For the modeling, three data sources were considered, L8, DAP-UAV and the combination (L8 + DAP-UAV). The model obtained using three-dimensional DAP-UAV data was used as a reference AGB of the studied fragments, increasing the number of representative plots for the area. For the estimation of multiscale AGB, at the basin level, a multiple linear regression adjustment was performed between the obtained by the model from the selected DAP-UAV and the predictor variables of the spectral data of L8. Finally, the multiscale AGB model was used to estimate the AGB of forest areas present in the Poxim-SE river basin. The model based on the combination of L8 and DAP data (L8 + DAP-UAV) had better performance in the estimates, R² of 0.96 and RMSE of 7.46 Mg ha-¹ (18.1%). The error was 24% smaller than estimates made with L8 and DAP-UAV data individually. Considering the modeling for the entire forest area analyzed, a slight overestimation of the BAS values was observed in the models from L8 and L8+DAPUAV. The results indicated that the combination of multispectral and three-dimensional remote sensing information increased the accuracy of plot-level AGB features. However, considering the entire stretch of secondary forest fragments analyzed, the L8 multispectral data caused an overestimation of the AGB values. At the basin level, the multiscale model performed with R² of 0.84 and RMSE of 15.9 Mg ha-¹ (33.7%). The Atlantic BrazilianForest areas of the Poxim basin had an average AGB of 46.51 Mg ha-¹. The DAP-UAV data showed potential to be used as a reference for the adjustment of biomass estimation models from multispectral data. The performance of the AGB estimation was consistent across all sites and the multiscale scaling approach to the AGB estimation produced a biomass map for the forest fragments of the Poxim River basin.