Caracterização espaço-temporal da cobertura de pastagens no Tocantins - BR, utilizando programação em nuvem no Google Earth Engine

Abstract

Agricultural production is one of the main economic activities in Tocantins, found in practically all municipalities in the state, with a notable modernization process. In the state of Tocantins, considering the types of cattle raising, extensive livestock farming predominates over confinement, with the main form of feeding the herd being natural or planted pastures. The main function of pasture areas is to serve as a primary support system for animal production in the supply of food and derived products. In this context, several studies have been carried out in an attempt to characterize the current scenario of pasture degradation, using remote sensing techniques, with the use of three images to represent the annual scenario, but there is little record of studies considering a series of years and with the largest number of images per year. Thus, the objective of this research is to analyze pasture degradation in the State of Tocantins, evaluating pasture vegetation cover (CVP) and the variation in degradation levels in the period from 2013 to 2022, with free code formulation on the Google Earth Engine platform. The methodological process consisted of applying remote sensing techniques and cloud programming (Javascript) on the Google Earth Engine platform. For processing the images from the Landsat TM 8 sensor, clipping along the perimeter of the state of Tocantins and the pasture area obtained from the MAPBIOMAS collection 8. Then, the NDVI was calculated for all images from each year to later result in an image with the average annual NDVI used to determine the CVP Pasture Coverage Index. The results show that most of the state's pastures (95.62%) present some degree of degradation, and that, on average, 47.02% of the area is at the highest level of degradation. An average of 186 images (mean) were used, and a total of 5280 for the entire series. Pasture degradation did not show any correlation with the rainfall index. The microregions that presented the best performance were Bico do Papagaio and Araguaína, with Gurupi having the worst performance identified. As a way of contributing to research in plant health, this study discusses the use of cloud programming as a way of optimizing processing time and the possibility of using BIGDATA in environmental studies.