NDVI (iNDVI) Satelite data were used in an effort to compare
VEMAP-simulated to satellite observed patterns in vegetation (Schimel et al
1997). We present a three-year average (1986-88) of
integrated annual normalized difference vegetation index (iNDVI). The data
file is in SVF format.
Data from the NASA/NOAA Pathfinder Program
were used to produce
satellite vegetation indices for use in model evaluation. The
Pahtfinder land data set is derived from measurements made by the
Advanced Very High Resolution Radiometer (AVHRR) on board the
NOAA TIROS-N series of polar orbiting satellites. The optical
data calibrated for reflectance. The data are projected onto an 8
x 8 km equal-area grid and temporally composited to scenes within
10-d windows from the original daily data. The Pathfinder
processing also includes cloud screening and partial atmospheric
correction (for ozone and Raleigh scattering). The normalized
difference vegetation index (NDVI) can be computed from the two
AVHRR optical channges (Tucker 1979, Goward et al. 1991).
There have been many studies that relate the NDVI to ecological quantities at local, regional, and global scales (Asrar et al. 1984, Fung et al. 1987). Most relevant for this study, the growing season integral of the NDVI is related to terrestrial net promary productivity (Goward et al. 1985) based on good empirical correlations and a strong theoretical basis (Goward et al. 1985, Sellers 1985, Fung et al. 1987, Schimel et al. 1991, Potter et al. 1993). Although there is uncertainty involved in calculating NPOP from the NDVI because of atmospheric and geometric bidirectional effects and vegetation type-specific relationships, it is robust as a proxy. also, by recompositing and averaging, we minimized these effects. NDVI has the important advantage of being an actual measurement made everywhere in the study domain, rather than being interpolated or extrapolated to a map, as is mapped soil organic carbon (SOC) (Schimel and Potter 1985).
We produced a data set that is the annual integral of NDVI (iNDVI) averaged over the 3 year of Pathfinder data available at the time of this study (1986-1988). We first extracted the continental US from the global grid and recomposited the data to monthly scenes by chossing the greenest pixels (an additional and conservative technique for minimizing cloud and sun/sensor angular geometry effects). The monthly values were then averaged for each pixel, and the twelve resulting maps reprojected to the VEMAP 0.5 x 0.5 degree grid. The iNVI values were then dividied by 12 to retain the familiar [-1, 1] range for NDVI.