A fast operative method for NDVI uncertainty estimation and its role in vegetation analysis

Vegetation indices represent an effective and widely used tool for monitoring vegetation changes in time and space. Unfortunately, in many works index uncertainty is not reported, making interpretation unreliable. In this paper we propose an operational approach for estimating NDVI uncertainty, based on the propagation of variance of factors defining the adopted radiative transfer model. Two Landsat 8 Operational Land Imager (OLI) images were used to test the method and discuss results. An agriculture-devoted area located in NW Italy was chosen as case study. Results showed that: a) the major contribution to NDVI uncertainty comes from topographic and atmospheric factors; b) uncertainty varies in space and time and depends on sensor spectral bands; c) NDVI uncertainty estimates can be exploited to map NDVI significant differences in space and time.