Análise de séries temporais de focos de calor em biomas brasileiros utilizando o método Gráfico de Recorrência

Vegetation fires are complex processes that can have natural or man-made causes, and their effects on an ecosystem vary according to its sensitivity. The recurrence of fire outbreaks can affect the environmental balance and human health. Thus, it is necessary to monitor the occurrence of fire and understand its dynamics. In Brazil, monitoring is carried out via satellite, with which hot pixels are detected. This process is accomplished by the National Institute for Space Research (INPE). In order to investigate the temporal variability of fires in the Amazon, Cerrado, Caatinga and Atlantic Forest biomes, the Recurrence Plot method and the Recurrence Quantification Analysis, developed to study the nonlinear dynamics of time series, were used. So, the daily series of hot pixels in these biomes, generated from data made available by INPE, recorded between July/2002 and December/2019, were analyzed. The recurrence plots of hot pixels for the Amazon and Cerrado biomes showed that their dynamic systems of vegetation fires have parameters that vary slowly and are non-stationary. It was also observed a semi-periodic structure for the Caatinga biome and discontinuous for the Atlantic Forest biome, indicating, in the latter, sudden changes in the dynamics. Besides that, according to the Recurrence Quantification Analysis, the time series of vegetation fires in the Atlantic Forest biome presented the least predictability and the highest degree of disorder, while the Caatinga biome revealed a more predictable hot pixels time series with more stable dynamics.