The Influence of Observation Conditions on the Accuracy of NDVI Vegetation Index Calculation from Earth Remote Sensing Data

Abstract

The problem of calculating the normalized vegetation index (NDVI) from satellite data is considered. The index values calculated from the MODIS/Aqua radiometer data by the SeaDAS software package algorithm are compared with the values obtained at the La Crau research site (France) for 7 years. The site is located near the Mediterranean coast and is a flat field where grass grows. The measurements are carried out by the ROSAS automatic photometric station. Calculations show the proximity of satellite and field measurements: bias of 0.005 and standard deviation of 0.03. The algorithm used does not take into account the effect of aerosol on the NDVI value. However, the errors due to the lack of accounting for aerosol lie within the limits of the total calculation error. There is a slight dependence of the error on the zenith angle of the sun, which varied in the range from 20° to 70°. The bidirectional reflectance distribution function of the surface on the site is uniform, except for the directions close to the sunbeam. The measurements were far from the sunbeam zone. However, the accuracy of the calculation depends on the difference between the azimuth angle of the site survey and the azimuth to the sun. NDVI mismatches due to differences in the central wave numbers of different satellites also turned out to be significant. The comparison was made for the following radiometers: MODIS/Aqua, Landsat-8/OLI-1, Landsat-9/OLI-2, and Kanopus-V/MSS.