Detecting Seasonal Changes in a Semi-arid Environment Using Hyperspectral Vegetation Indices∗

Hyperspectral image processing is a promising tool for the analysis of vegetation in remote sensing imagery, particularly in areas with low vegetation cover. The detection of photosynthetically active vegetation dynamics is essential for many environmental studies. Nonetheless, remote sensing of vegetation dynamics in the absence of land cover change is much more challenging than standard land cover analysis. A variety of multispectral vegetation indices have been adapted to hyperspectral images in order to detect these changes. However none of these indices make full use of the continuous spectra acquired from hyperspectral sensors and are limited in the detection of low vegetation cover because of varying background signals. The derivative-based green vegetation index (DGVI) proposed by Chen and Elvidge (1998), the Water Band Index (WBI) and the NDVI were used to highlight changes in green vegetation cover in bitemporal Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) Images from the Santa Monica Mountains (California) and compared with one another. The DGVI makes use of the sharp transition from the Chlorophyll absorption in the red to the high reflectance in the near-infrared. It minimizes the background effects and is insensitive to illumination changes. The WBI proposed by Peñulas et al. (1997) was used as an alternate indicator of green vegetation. The results were compared with the NDVI. Ground truth data were also obtained from low altitude aerial photos in order to test the accuracy of these methods and to calculate percent green vegetation cover. Distinct seasonal vegetation dynamics were found in the chaparral communities of the Santa Monica Mountains. The difference images show high changes in urban areas, the soft chaparral and the burned site. In multitemporal comparison the DGVI und WI performs much better than the NDVI. The highest overall accuracy was obtained by the DGVI. But unexpectedly DGVI fails in shadow-affected areas