Application of AVIRIS data in detection of oil-induced vegetation stress and cover change at Jornada, New Mexico

The polynomial fitting and Lagrangian interpolation, and spectral mixture analysis (SMA) are applied to the AVIRIS data sets. The first two methods are applied for the detection of the “red-edge” shift in vegetation reflectance spectra, and the last for the detection of change in vegetation fraction. The results indicate that the polynomial fitting and Lagrangian interpolation both are able to detect a red-shift of the vegetation “red-edge”, but the latter's performance depends on the band combination used and is sensitive to data noise. The polynomial fitting results are inconsistent in detection of “the red-edge” shift, while Lagrangian interpolation is not. Within the oil spill area, the fraction estimates of vegetation resulting from SMA demonstrate a decrease (10-30%) of the vegetation fraction after the accident, indicating stressed vegetation and cover change. The result also indicates that areas of extremely large decrease (>40%) in plant cover outside of the oil spill area is due to the response of grasses due to the water stress in 2000, and that the integration of some auxiliary data on ecological and climatological data with the analysis of remotely sensed data is thus very important to the interpretation of the detection results. A sensitivity analysis indicates that the detected vegetation cover change is insensitive to the noise introduced by the radiometric normalization.