The impact of soil properties on soil NDVI and the quantification of the green vegetation fraction.

摘要

Many applications use the Normalized Difference Vegetation Index (NDVI) to study vegetation and other ground components. Because soils cover most of the Earth's terrestrial surface, they constitute an important portion of the NDVI signal. In most cases, this contribution is ignored as the NDVI of soils (NDVIo) is considered to be very low (0.1) and invariant. Soil reflectance measurements gathered from various sources show, however, that most NDVIo are larger (mean NDVIo=0.17) and can vary significantly (standard deviation=0.07). We show that underestimating NDVIo yields overestimation of Fg that is the largest when 0.2 NDVIo yields overestimation of Fg that is the largest when 0.2NDVI0.4 over the studied area. Using a widely-used Fg model, we show that the Fg overestimation can be larger than 0.2 depending on the land cover type and the parameters used in the Fg model.