Radiometric quality and performance of TIMESAT for smoothing moderate resolution imaging spectroradiometer enhanced vegetation index time series from western Bahia State, Brazil

Abstract

Empresa Brasileira de Pesquisa Agropecuaria, BR-020 km 18, CEP: 73301-970,Planaltina/DF, BrazilAbstract. The launch of the Moderate Resolution Imaging Spectroradiometer (MODIS) sensoraboard the Terra and Aqua platforms in 1999 and 2002, respectively, with temporal resolutionsof1to2daysopenedthepossibilityofusingalongtimeseriesofsatelliteimagestomaplanduseand land cover classes from different regions of the Earth, to study vegetation phenology, and tomonitor regional and global climate change, among other applications. The main objectives ofthis study were twofold: to analyze the radiometric quality of the time series of enhanced veg-etation index (EVI) products derived from the Terra MODIS sensor in western Bahia State,Brazil, and to identify the most appropriate filter to smooth MODIS EVI time series of thestudy area among those available in the public domain, the TIMESAT algorithm. The 2000to2011timeperiodwasconsidered(atotalof276scenes).Theradiometricqualitywasanalyzedbased on the pixel reliability data set available in the MOD13Q1 product. The performances ofthe three smoothing filters available within TIMESAT (double logistic, Savitzky-Golay, andasymmetric Gaussian) were analyzed using the Graybill’s F test and Willmott statistics.Five percent of the MODIS pixels from the study area were cloud-affected, almost all ofwhich were from the rainy season. The double logistic filter presented the best performance.