Use of Vegetation Index “Fingerprints” From Hyperion Data to Characterize Vegetation States Within Land Cover/Land Use Types in an Australian Tropical Savanna

Abstract

Suites of spectral indices may be derived from hyperspectral sensors such as Hyperion on EO-1. Spectral indices linked to vegetation and landscape function that are scalable to multi-spectral global sensors, could provide “fingerprints” for vegetation states in tropical savannas. In this study, Hyperion images were acquired on three occasions throughout the dry season over each of two consecutive years in the tropical savanna near Darwin, Northern Territory, Australia (12°25 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">'</sup> N,130°50 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">'</sup> E) during 2005 and 2006. This paper examines the changes in fractional cover of photosynthetic and non-photosynthetic vegetation and bare soil and key diagnostic narrow band vegetation indices for major land cover/land use (LCLU) types over two contrasting post-monsoon seasons. The fractional cover proportions and vegetation indices responded strongly to the additional month of full monsoon rains in 2006 versus 2005. There were differences in vegetation indices sensitive to pigments, canopy water and cellulose between LU and LC classes, but within class variation was very high for large sized sample areas. When fine scale variation in vegetation indices and fractional cover were examined as “fingerprints” for small, more uniform areas of specific LC, distinct differences were evident. Vegetation indices and derived vegetation properties can be used to characterize vegetation states at the scale of natural and management-induced variation. The vegetation indices and fractional cover methods used here can be translated and scaled-up to current and new global sensors to improve description of vegetation structure and function in savannas.