An application of airborne hyperspectral and EO-1 Hyperion data for inter-sensor calibration of vegetation indices for regional-scale monitoring

In order to use satellite data from multiple sensors for monitoring of the Earth's vegetation, satellite data product continuity and compatibility need to be investigated. In this study, we assessed inter-sensor compatibility of the normalized difference vegetation index (NDVI) by simulating bandpasses of the NOAA-14 AVHRR, Terra-MODIS, and Landsat-7 ETM+ from airborne hyperspectral and EO-1 Hyperion data acquired over a savanna-forest transitional zone in Brazil. Our results showed that the simulated reflectances among the sensors examined here were near-linearly related, but with relationships that were land cover-dependent, e.g., reflectances over green vegetation targets formed separate relationships. The deviations were larger for greener targets. On the other hand, the NDVI crossplots among the sensors formed single, land-cover independent relationships. They were, however, curve-linearly related with the degree of curve-linearity increasing with atmosphere contamination. Also, land cover dependencies appeared in the NDVI cross-plots when contaminated by atmosphere. These results suggest that inter-sensor calibration and continuity of the NDVI are achievable, but require a careful examination of surface and atmospheric conditions and the development of a theoretical basis.