NDVI Sensitivity to Atmospheric Water Vapor as a Function of Spectral Bandwidth'

The Normalized Difference Vegetation Index (NDVI) has. been an important data product of the AVHRR series of instruments that have flown on the NOAA polar orbiting satellites. This index has been extensively used to monitor global vegetation and to study deforestation. As planning begins for follow-on versions, a proposal has been made to improve the robustness of the index by narrowing the spectral bands that are used in the NDVI calculation to reduce sensitivity to atmospheric water vapor. Without dramatic improvements in instrument design, this narrowing will also degrade the instrument's signal-to-noise ratio. In this paper, results are presented of an investigation into the sensitivity of the NDVI to water vapor as a function of the width of the spectral bands. Consideration was also given to the increase in instrument noise. This study concluded that spectral bandwidths of around 50 nanometers centered on two water vapor transmittance windows provide sufficient insensitivity to water vapor without significantly degrading the instrument's noise performance. However, it should be noted that changing the bandwidths from the current AVHRR will complicate the use of the index for long term environmental studies by changing the values obtained for similar scenes. BACKGROUND AND INTRODUCTION The Normalized Difference Vegetation Index (NDVI) has been an important data product of the Advanced Very High Resolution Radiometer (AVHRR) series of instruments that have been canied by the NOAA polar orbiting satellites. The NDVI has been used to monitor global vegetation (Gatlin, et al., 1984), crop growth (Schneider and McGinnis, 1982) and to conduct deforestation studies (Malingreau and Tucker, 1987). The historical database acquired by the AVHRR has proven the utility of long term satellite data collection through these and other scientific applications. The next generation of NOAA polar orbiting satellites (NOAA KLM) are slated for launch beginning in 1996 and will continue the collection of NDVI data with an improved version in this series known as the AVHRW3 (Owens, et al., 1989). While the main improvement in this instrument will be the addition of a band at 1.6 pm for vegetation moisture stress and snow/cloud discrimination, improved signal-to-noise ratio's are also planned. While this next generation has yet to be launched, planning has already commenced for the satellite series to follow. The NOAA OPQ series (Needham, 1992) has been proposed to meet the needs of operational meteorological and environmental monitoring in the next century. The imaging radiometer which will continue the collection of data for NDVI is to be known as the Visible, Infrared, Scanning Radiometer (VIRSR) and incorporates several improvements and changes from the AVHRRL.3 instrument. The specification for the VIRSR (GSFC, 1991) calls for improvements in the radiometric accuracy as well as changes in the spectral bandpasses of several of the bands. In particular, the two bands which are used in the NDVI calculation are proposed to be substantially narrower but with higher signal-to-noise ratio as compared to the AVHRR/3. These conflicting requirements will be demanding to satisfy. A goal of reducing the spectral bandwidth of these two bands is to reduce the sensitivity of the measurements to atmospheric water vapor. Since the primary application of these channels is to measure surface characteristics, unknown and highly variable atmospheric water vapor results in "noise" and complicates interpretation of the data. This change in the bands should result in increased consistency in the NDVI measurements. However, it also will dramatically change the characteristics of the measurement and complicate the use and interpretation of the index for long term studies. 'This work was sponsored by the National Oceanic and Atmospheric Administration under Air Force. Contract F19628-90-C-0002. 0-7803-1497-2194 $4.00