The contribution of probability theory in assessing the efficiency of two frequently used vegetation indices

Abstract

Frequency band ratios are often used as vegetation indices in environmental studies as a measure of the amount of vegetation in a digital image. The Simple Vegetation Index (SVI) and the Normalized Difference Vegetation Index (NDVI) are two frequently used vegetation indices and their mathematical form is the basis for the development of other modified vegetation indices, such as TVI, SAVI or SAVI₂. Vegetation indices are mainly defined and evaluated empirically. In the present paper, a different approach based on probability theory is developed in order to evaluate the efficiency of SVI and NDVI and to suggest two modified vegetation idices, MSVI and MNDVI. According to the mathematical analysis and experimentation with a Landsat 7 Enhanced Thematic Mapper (ETM) image of an island in western Greece, it is concluded that NDVI provides better results than SVI, since the image of the former has a much broader brightness histogram and the targets of interest are more clearly expressed in the satellite image. The image of MSVI has a broader histogram from those of NDVI and SVI and a more diverse tonality. MNDVI may provide better results than NDVI if the standard deviations of the images of the near infrared (NIR) and red bands vary considerably. It is also concluded that the signal-to-noise ratio of the MSVI image is better than that of the SVI image. The signal-to-noise ratio of the MNDVI image may be better than that of the NDVI image if a proper value for a characteristic parameter in the expression for the MNDVI is chosen.