Spectral band difference effects on vegetation indices derived from multiple satellite sensor data

Vegetation indices based on satellite image data are widely used for change monitoring but, when derived from different satellite sensors, differ as a function of the uncorrectable differences between the analogous spectral bands used to generate them. This is an important issue because multiple satellite sensors in the Landsat class or in the AVHRR and MODIS classes are being used increasingly to monitor vegetation dynamics. This paper reports on an investigation of the impact of spectral band difference effects (SBDEs) on cross-comparisons between vegetation indices (VIs) derived from multiple satellite sensors in the solar-reflective spectral domain. Results from the simulation study, which encompassed three vegetation target types and eight VIs, indicate how large SBDEs can be and for which VI cross-comparisons they are significant. They also indicate that the spectral dependence of atmospheric gas transmittance is the key factor that gives rise to such significant spectral band difference effects. Among the vegetation indices considered, the GEMI proved to be the least sensitive to spectral dissimilarities between sensors, and hence GEMI is worth considering for quantitative monitoring of vegetation using images from multiple sensors. In the context of potential candidates to fill the forthcoming gap in Landsat data continuity, either one of the IRS-P6 sensors or the SPOT-5 HRG is preferable to the CBERS-2 HRCC as a replacement sensor from the standpoint of agreement with Landsat-based vegetation indices.