Relationships between Hyperion-derived vegetation indices, biophysical parameters, and elevation data in a Brazilian savannah environment

The relationships between 18 narrowband Hyperion vegetation indices and basal area and canopy cover values of eight Brazilian savannah physiognomies were evaluated. The best fitting regression relationships were used to estimate these parameters in the image, which were projected over a digital elevation model (DEM) and compared with an available vegetation map. Results showed that the gradual increase of biophysical parameters, from savannah grassland to semi-deciduous forest, produced large correlation coefficients with most of the indices, especially the near-infrared/red-based ones (e.g. NDVI and simple ratio). Larger canopy cover and basal area values were associated with tall woodland and semi-deciduous forest and smaller values with grasslands and shrub savannah. When projected over the DEM, the predominance of rupestrian fields at high altitudes and of tall woodland and semi-deciduous forest at low altitudes was observed.