A systems based modification of the NDVI to minimize soil and atmospheric noise

Abstract

The Normalized Difference Vegetation Index (NDVI) equation has a simple, open loop structure. This renders the NDVI susceptible to large sources of error and uncertainty over variable atmospheric and soil background conditions, which is less than satisfactory in meeting the need for accurate, long term vegetation measurements for the Earth Observing System (EOS) program. In this study, a systems analyses approach is used to examine noise sources in existing VIs and to develop a stable, modified NDVI (MNDVI) equation. The MNDVI, a closed-loop version of the NDVI, was constructed by adding: (I) a soil and atmospheric noise feedback loop, and (2) an atmospheric noise compensation forward loop. The coefficients developed for MNDVI can be used with data uncorrected for atmosphere, as well as with Rayleigh corrected and atmospherically corrected data. In field observational and simulated data, as well as satellite imagery, the MNDVI was found to reduce combined soil and atmospheric noise to less than 4% for any complex soil and atmospheric situation. The resulting uncertainty, expressed as vegetation equivalent noise (VEN), was /spl plusmn/0.11 LAI units, which was seven times less than encountered with the NDVI (/spl plusmn/0.8 LAI), and three times less (/spl plusmn/0.36 LAI), than with the Soil Adjusted and Atmospherically Resistant Vegetation Index (SARVI). >