Time series investigation for correlations between rainfall and phenology in Cerrado (Brazilian savanna)

• Ano de defesa: 2020
• Autor(a) principal: Oliveira, Brener de Almeida
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa
• Palavras-chave em Português: Fenologia; Sensoriamento remoto; Cerrados; Vegetação - Mapeamento - Índices; Fotointerpretação Florestal
• Link de acesso: https://locus.ufv.br//handle/123456789/28744

The Brazilian savanna (Cerrado) is the second-largest biome in Brazil, following the Amazon. The influence of climate in this biome is pronounced, dry winters and rainy summers determines the characteristics of the Cerrado typologies. The effect of climate on vegetation can be assessed through the variation in biophysical responses of each Cerrado typology provided by time-series of spectral indices as Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI). This study aims to test for the correlation between rainfall and Cerrado typologies spectral responses from 2001 to 2018, in central Brazil. As specific objectives, we intend to spectrally separate such physiognomies and characterize them according to their phenological metrics. Thus, we performed a Principal Component Analysis (PCA) for the MOD13Q1 NDVI and EVI time-series, to test for the spectral separation of Cerrado typologies. We extracted the phenological metrics (e.g. growth rate, base value) in TIMESAT software for both indices time-series, in order to investigate the seasonal behavior of the different vegetation types. Further, we performed wavelet analyses for EVI and NDVI and rainfall time-series in order to detect whether there are anomalies in the series and whether there are coherence and causal relationships between vegetation spectral response and rainfall. PCA showed a clear spectral-temporal separation among forest and savanna physiognomies. Our findings indicated that phenological metrics are efficient in characterizing the seasonal behavior of the Cerrado typologies, and forest-like ones have greater total productivity of green biomass compared to the savanna types. Moreover, we observed coherence between vegetation response and rainfall, in the short-run (32 to 64 and 64 to 128 days). The middle-run wavelet analyses showed a cyclical correlation and coherence between phenology of the physiognomies and rainfall in every 8.5 months. The wavelet NDVI data present an in-phase situation where vegetation leads to rainfall. Conversely, in the long-run, another in-phase is observed and rainfall leads vegetation response. For EVI data, we did not observe any leading phase in the long-run. We concluded that it is possible to separate the Cerrado typologies by their spectral- temporal biophysical responses, enabling the distinction between savanna and forest physiognomies, which show the highest seasonal productivity. It is also possible to conclude that there is a strong seasonal relationship between rainfall and vegetation biophysical response throughout the time, and that extreme events and long-term trends of rainfall also show coherence with vegetation. These findings suggest that, any rainfall regime change impacts the green biomass productivity of various Cerrado biome physiognomies, from the less productive savannas to the most productive forest patches. Keywords: Remote sensing. Phenology. Cerrado. TIMESAT. NDVI and EVI.