An examination of a pixel replacement algorithm for monitoring postfire chaparral recovery using indices derived from aviris data

The objective of this thesis is to evaluate the effectiveness of a pixel replacement algorithm for monitoring postfire vegetation recovery in chaparral vegetation. The algorithm creates a new scene which replaces burned pixel values with theoretical unburned values. This process was performed on several index scenes derived from hyperspectral data. Vegetation and water absorption indices, as well as spectral mixture analysis were used to evaluate the pixel replacement algorithm, which predicted values within an area of the Santa Monica Mountains that had been burned by the Calabasas Fire in 1996. The study used 13 scenes acquired by the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) that spanned 8 years (six scenes acquired before the fire and seven after). Each predicted index was subjected to three evaluations. The first evaluation was a direct comparison to the difference Normalized Burn Ratio (dNBR) to examine recovery trends. After each index was processed through the pixel replacement algorithm the result was a more stable trend toward recovery without the acute seasonal variability which is inherent in differenced indices such as dNBR. A trendline analysis was also performed on RMSE values from each index. The RMSE was found to decrease as the number of unburned pixels used to model the burned pixels increased. A statistical test, the Kolmogorov-Smirnov (K-S) test, was the third evaluation tool used. The K-S test was used to determine the similarity between the reference pixels and the theoretical pixels after they had been processed through the pixel replacement algorithm. All three of the analyses provided support for using a pixel replacement algorithm to monitor postfire vegetation recovery as feasible. Since the pixel replacement method did not exhibit seasonal variability, may be useful in vegetation types that exhibit seasonal changes in cover, such as deciduous vegetation.