Uso de imagens do Sentinel 2 na estimativa de parâmetros biofísicos da vegetação em áreas de Mata Atlântica

• Autor(a) principal: Perez, Gabriel Guariglia
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFSCAR
• Texto Completo: https://repositorio.ufscar.br/handle/ufscar/10609
• topic: LiDAR; Sensoriamento remoto; Estrutura da vegetação; Remote sensing; Vegetation structure; CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL::CONSERVACAO DA NATUREZA::CONSERVACAO DE AREAS SILVESTRES; CIENCIAS EXATAS E DA TERRA::GEOCIENCIAS::GEOFISICA::SENSORIAMENTO REMOTO; CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL::CONSERVACAO DA NATUREZA

Information about vegetation biophysical parameters can be used in many applications, and remote sensing is showing itself to be a good tool to obtain it. In this study we investigated the possibility of using Sentinel 2 images to estimate vegetation biophysical parameters measured in field and with LiDAR (Light Detection and Ranging). The study was conducted in areas covered by Atlantic Forest in the state of São Paulo (Brazil) using three Sentinel 2 images. Additionally, we used a Landsat-8/OLI image for temporal proximity with the LiDAR data, tested the effect of topographic correction on the images and made an analysis of leaf reflectance in lab. The analyzed field variables were height, DBH (Diameter at breast height), percentage of canopy cover, and number of individuals. The LiDAR variables were height of the first returns, height of the last returns and number of returns per pulse. A total of 26 variables were extracted for comparisons with the images through OLS (Ordinary Least Squares) and RF (Random Forest) regression models. These comparisons were made with single bands, the vegetation indices RVI (Ratio Vegetation Index), NDVI (Normalized Difference Vegetation Index), SAVI (Soil Adjusted Vegetation Index), EVI (Enhanced Vegetation Index), NDWI (Normalized Difference Water Index), NDI45 Normalized Difference Index B4 and B5), IRECI (Inverted Red-Edge Chlorophyll Index) and S2REP (Sentinel 2 Red Edge Position), and with all the possible rations between two bands. The results show that many biophysical parameters are related to the images (r² up to 0.62), and the topographic correction seems to have a positive effect in the estimates, especially for LiDAR derived variables. The best models generated with field data were regressions between multiple Sentinel 2 bands and tree height, canopy cover and biomass. For the LiDAR data, the results were in general better than with field data and also involved regressions with multiple Sentinel 2 bands in comparison with canopy height and cover. Among the images, the one that presented better relations with the biophysical parameters were Sentinel 2 image of December 26, 2016. The validation results of the LiDAR models show that they work in different areas of the same image in which they were trained but can only be applied in different images after an appropriate atmospheric correction. In general, vegetation indices did not show better results than individual bands. Among the bands, we highlight the role of B5 (705 nm, red-edge) for the success of many estimates, which is confirmed by the results of the analysis of leaf reflectance made in lab. For future studies, we recommend better investigation of wavelengths close to 705 nm and the potential of Sentinel 2 band 5. For the creation of models that can be used in different images, we recommend the use of level 2A Sentinel 2 imagens (surface reflectance), that will be available globally by the end of 2018. Finally, we also recommend to consider other variables such as classes of successional stages and the comparison of Sentinel 2 imagery with other phytophysiognomies.