A technique of inter-sensor VI translations using EO-1 Hyperion data to minimize systematic differences in spectral band-pass filters

摘要

Utilization of satellite date from multiple platforms increases our chances of more frequent and accurate observations of the Earth's surface in both global and regional scale For the purpose of vegetation monitoring, this will be particularly true by combining the data from sensors of various spatial, spectral, and temporal resolutions, e.g. the combinations of data from AVHRR (broad band), MODIS (narrow band) and ETM+ (higher spatial resolution). Even though the same spectral vegetation index can be obtained from these sensors, the two main issues need to be considered, one is the systematic differences caused by the spectral response functions, and the other is the differences in spatial resolutions. This paper investigates the spectral issue and its role in the spectral calibration of NDVI among sensors. Hyperspectral data from Hyperion onboard the EO-1 platform were used to simulate outputs from various sensors by band convolution. The data were initially corrected for Rayleigh scattering and Ozone absorption to produce the top-of-the-canopy reflectance as a starting point. The technique first designs a sensor-specific vegetation index (VI) and background brightness index (BI) by accounting for the differences in band-pass filters. These VIs and BIs are then used to estimate the common parameters (sensor independent parameters) attributed to vegetation amount and background brightness. Finally, these parameters are used for the translation of VI among sensors.