Monotonicity of Two-Band Spectral Vegetation Index in General Form Under a Two-Endmember Linear Mixture Model

Spectral vegetation indices using red and NIR bands (two-band VIs) have been used as information for vegetation monitoring. However, spatially averaged values of two-band VI have biases (known as scaling effect) that are induced by surface heterogeneity and nonlinearity of VI’s model equations. This study tries to understand the mechanism of the scaling effect for a general form of twoband VI assuming a two-endmember linear mixture model. It is proved that spatially averaged VI value changes monotonically for a certain sequence of resolution cases. The findings (monotonicity of two-band VI) lead us to identify an error bound of two-band VI caused by the scaling effect. The derived results would also provide useful information for cross-calibration among VI products from various sensors/platforms.