Uso de técnicas de sensoriamento remoto para estimar variáveis biofísicas em floresta tropical seca, no município de Floresta - PE

• Autor(a) principal: OLIVEIRA, Géssyca Fernanda de Sena
• Data de Publicação: 2020
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFRPE
• Texto Completo: http://www.tede2.ufrpe.br:8080/tede2/handle/tede2/8936
• topic: Sensoriamento remoto; Floresta tropical seca; Caatinga; Volume; Biomassa; Carbono; Manejo florestal; CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL

The use of remote sensing techniques to assist the conventional forest inventory has been promoted to moderate the need for fieldwork, reducing time and costs. Therefore, the present work has as main objective to evaluate the potential of using remote sensing tools to estimate the biophysical variables volume, biomass, and carbon in dry forest in the Northeast of Brazil. The study area is located at the Itapemirim Farm, municipality of Floresta – PE. In this area we have two fragments, with different usage background, were analyzed at different times of the year, being called Area I (Transposition) the one considered preserved and Area II (Chaining) the one considered degraded since its vegetation was removed with the help of chains for approximately 34 years for forest management purposes. 40 plots in each area with dimensions 20 m x 20 m (400 m²) were evaluated, totaling 3.2 ha of the sampled area. Scenes from the months of April/2018 and August/2018 of the satellite Landsat 8 / OLI were used, in orbit/point 216-66, referring to the wet and dry period, respectively. These scenes were converted to surface reflectance from the radiometric calibration and, subsequently, the GNDVI, NDVI, SR, SAVI L=0,5, DVI, MVI, ARVI, LAI, GVI, GARI, EVI and GEMI vegetation indices were generated. The multispectral bands, as well as the vegetation indices (IV), were related to the volume, biomass, and forest carbon estimated from dendrometric data measured in the same passage period of Landsat 8 / OLI. The data were fitted to the multiple linear regression model, leading to the selection of variables using the Stepwise method. The adjusted Criteria of Determination Coefficient (R²aj), Standard Error of the Estimate (Sxy%), and the Graph of Waste Distribution (Res (%)) were used to select the best equations. Statistical analysis was performed using software R® version 3.6.1. The dry period was the most suitable to estimate biophysical variables using orbital images and remote sensing techniques in Tropical Dry Forest (TDF). The best equations for estimating volume, biomass, and carbon obtained a R²aj of 0.634, 0.650 and 0.649, and a Sxy of 44.894%, 6.030%, and 6.842%, respectively. Biomass and carbon showed better adjustments after logarithmizing the IVs with positive values, while volume showed an opposite behavior. The vegetation indices EVI and SAVIL = 0.5 did not prove to be appropriate to estimate the biophysical variables, regardless of seasonality, while NDVI was efficient only in the wet season. Therefore, observing the due restrictions and the equations with the best statistical adjustment, as well as the residual graphs, it appears that it is possible to use Landsat 8/OLI images to make estimates of forest parameters, demonstrating the importance and applicability of this method for the estimation of biophysical variables in TDF, as well as management and conservation actions in the Caatinga Domain.