The assessment of vegetation seasonal dynamics using multitemporal NDVI and EVI images derived from MODIS

Abstract

The objectives of this work were to characterize seasonal dynamics of cerrado, deciduous and semideciduous forests in the north of Minas Gerais, Brazil. Time series of NDVI (Normalized Difference Vegetation Index) and EVI (Enhanced Vegetation Index) derived from MODIS sensor, were compared by analyzing temporal profiles and image classification results. The results showed that: (1) there is an agreement between vegetation indexes and the monthly precipitation pattern; (2) deciduous forest showed the lowest values and the highest variation; (3) cerrado and the semideciduous forest presented higher values and lower variation; (4) based on the classification accuracies the best vegetation index for mapping the vegetation classes in the study area was the NDVI, however both indexes might be used to assess the vegetation seasonal dynamic; and (5) further research need to be carried out exploring the use of feature extractions algorithms to improve classification accuracy of cerrado, semideciduous and deciduos forests in Minas Gerais, Brazil. Key words: Remote sensi