Effects of spectral, spatial, and radiometric characteristics on remote sensing vegetation indices of forested regions

Abstract

Vegetation indices derived from satellite image data have become one o f the primary information sources for monitoring vegetation conditions and mapping land cover change. The most widely used vegetation index in this context is NDVI, the normalized difference vegetation index, which is a function of red and near-infrared spectral bands. Given that the spectral and spatial of imagery in the red and near-infrared vary from sensor to sensor, NDVI values based on data from different instruments will not be directly comparable. The present study demonstrates the impact of changes in spectral bandwidth and spatial scale on NDVI derived from Airborne VisiblelInfrared Imaging Spectrometer (AVIRIS) data acquired at 20-m resolution over a forested region in southeastern British Columbia. For this purpose, the 10-nm AVIRIS data were spectrally and spatially aggregated in the red and near-infrared to simulate bandwidths from 10 nm to 150 nm for ground resolutions varying from 20 m to 1100 m. Sensor-specific spectral bands and spatial resolutions such as those for SPOT HRV, Landsat TM, NOAA AVHRR, EOS MODIS, and Envisat MERIS were also generated. NDVI values were then calculated using atmospherically corrected surface reflectances for forestry-related targets for the entire simulated band set at the various scales. The results indicate that the NDVI is significantly affected by differences in spectral bandwidth, especially for the red band, and that changes in spatial resolution lead to less peruasive but more land cover specific effects on NDVI. Results for the forested regions also indicate that NDVI is not very sensitive to the location of the near-infrared spectral band, provided that the bandwidth is no wider than 50 nm and the atmospheric correction for water vapor absorption is adequate. If either proviso is relaxed, the wavelength placement of the near-infrared spectral band is more critical, the optimum location being in the 850–880 nm range. Finally, for the same forest targets, some results were also generated for several other vegetation indices that make straightforward use of atmospherically corrected red and near-infrared spectral bands.