Produtividade primária bruta e biomassa em pastagem no bioma cerrado: uma análise a partir dos modelos SEBAL/CASA e MOD17 no estado de Goiás

• Ano de defesa: 2018
• Autor(a) principal: Veloso , Gabriel Alves
• Orientador(a): Ferreira , Manuel Eduardo
• Banca de defesa: Ferreira , Manuel Eduardo, Vieira , Pedro Alves, Giongo, Pedro Rogério, Luiz , Gislaine Cristina, Ferreira Júnior, Laerte Guimarães
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Goiás
• Programa de Pós-Graduação: Programa de Pós-graduação em Geografia (IESA)
• Departamento: Instituto de Estudos Socioambientais - IESA (RG)
• País: Brasil
• Palavras-chave em Português: Landsat 8; Biomassa seca; SEBAL/CASA; MOD17A2H; Suporte bovino
• Palavras-chave em Inglês: Dry biomass
• Área do conhecimento CNPq: GEOGRAFIA REGIONAL::ANALISE REGIONAL
• Link de acesso: http://repositorio.bc.ufg.br/tede/handle/tede/8625

Biophysical parameters of the soil-vegetation system, such as real evapotranspiration (ETR), radiation balance (Rn) and gross primary productivity (GPP), as well as information on dry biomass, are recognized as important in areas with agricultural activities, especially pastures (that usually do not have irrigation), and can help in the proper management of this environment. Making these measurements by satellite data, from the electromagnetic radiation reflected by the targets on the surface, makes these operations more efficient in a series of applications, such as monitoring of extensive agropastoral areas. The objective of this study was to estimate these parameters, based on the parameterization of models with specific data for pasture in the Cerrado of Goiás (mainly with regard to gross primary productivity), such as light use efficiency (LUE) and Photosynthetically Active Absorbed Radiation (FPAR). In order to improve these estimates, local meteorological data, such as Photosynthetically Active Radiation (PAR), were used, contributing to a better understanding of the spatial-temporal variability in the study area. The experiment was carried out at distinct scales, one of which was more detailed - in pasture areas in the Rio Vermelho watershed (BHRV), west portion of Goiás, using Landsat 8 OLI/TIRS imagery, and a more comprehensive one, for pasture areas in the Cerrado of the entire state of Goiás, using the NDVI images generated by the MODIS sensor - product MOD13Q1H. At BHRV, the variation of these parameters was analyzed from October 2014 to September 2015, using nine Landsat 8 images, path/roll 223/71. The estimation of these parameters, especially the GPP, was obtained through the coupling of the algorithms SEBAL (Surface Energy Balance Algorithm for Land), aimed at the estimation of the evapotranspiration, combined with the CASA (Carnegie Ames Stanford Approach) model, which calculates the Photosynthetically Active Radiation Absorbed (APAR) and, together with surface data, ends with the estimate of dry biomass. For this same area, an adaptation of the methodology of the GPP product obtained by MOD17A2H to Landsat 8 images was also carried out, in order to better understand the variation of GPP and dry biomass in medium spatial resolution images (30 m), with calibration of the model with specific pasture data and local meteorological data. Among the results, in the BHRV the values of Rn and ETR were consistent with those found in the literature, presenting significant spatial and temporal variability, with the first presenting a mean variation from 413 to 670 w/m-2, and the second from 1.6 to 4.55 mm.day-1, in which the lowest values can be related to pasture areas with some level of degradation. In relation to GPP, the SEBAL/CASA method proved to be more efficient among the methods applied in this research, following the climatic seasonality of the region and its influences on pasture areas, presenting a variation of 0.10 to 4.6 g C m-2, and with an average carbon sequestration potential of 4.8 Mg ha-1 year -1. The MOD17 method, adapted to Landsat 8 images, showed a variation of 0.5 to 4.0 g C m-2 day-1, with small variation in the monitoring of climatic seasonality of the region. The analysis of GPP, by the product MOD17A2H in the BHRV, presented a variation of 0.27 to 5.39 g C m-2 day-1, but with low spatial variation due to low image resolution and calibration data of the model (generated for the terrestrial globe). The analysis of the dry biomass followed the same pattern, with the SEBAL/CASA method being more efficient, obtaining good results with the data observed in the field, with a correlation coefficient of 0.663, mean absolute error of 0.228, root mean square error of 0.665 and Willmott's concordance index of 0.754. The dry biomass estimated with the product MOD17A2H showed good correlation (r = 0.833) with the field data, when considering the temporal variation; however, for this (dry biomass), mean absolute error and mean square error (2,133) were observed, due to the observed super-optimization. In relation to the Animal Unit by area (UA/hectare), the data obtained by the SEBAL/CASA and MOD17 method applied to the Landsat 8 images showed to be closer to the reality of the BHRV, with average values of UA/ha of 1,5 and 2,5 UA/ha, respectively. The UA/ha data obtained with MOD17A2H images appear to be high for the basin, with an average value of 3.6 UA/ha in the BHRV. In addition, data from GPP and dry biomass obtained in the pasture areas of the Cerradogoiano, from NDVI images (product MOD13Q1H), were significantly lower than those observed by the GPP product MOD17A2H, reflecting this result in the UA estimate/ha in the State of Goiás, which, with the product MOD17A2H, average values of 5.2 UA/ha were observed for the pasture areas of the State of Goiás, while the data obtained by this research presented average values of UA/ha of 2.5, which is closer to reality in such pasture areas in the Cerrado of Goiás. Therefore, the estimation of these parameters, aiming at a reading of the pasture and local climatic data, presented better results with the calibration of the models with specific data.