In this study, we focus on the identification of various volcanic products of Mt. Etna (eastern Sicily, Italy). Mt. Etna is one of the most active basaltic composite stratovolcanoes with the ability to change its land field rapidly, vigorously and continuously. For this purpose, NASA EO-1/Hyperion hyperspectral data (HSI) are used for lava flow differentiation and mapping within the historic “1536” to “1669” era of the Torre del Filosofo formation. These volcanic products are selected, due to a) their distinct spatial distribution, b) spectral similarity and c) field segregation from surrounding younger lavas. Due to their high compositional variability, the corresponding HSI pixels are mixed thus the problem is tackled with Spectral Unmixing (SU) techniques. First, endmember extraction, for each lava type, is performed by averaging over a Gaussian distribution of pixel reflectances of its associated region of interest (ROI). Then, abundances are estimated by two unmixing models: Constraint Linear Least Squares Unmixing (LLSU) and Bilinear Unmixing (BLU), applied on both spectral signatures and transformed versions of them, i.e. in the frequency domain through a Fourier Transform (FT). Historic lava flow delineation results are presented creating abundance maps per method. Also, a qualitative evaluation is performed using the Etna geological map, while a quantitative assessment is performed through derivation of the Structural Similarity Index (SSIM) of each implemented method. Ultimately, we address the degree of lava flow separability with the intercomparison of all methods and give an estimation of the computationally most efficient method. The specific research shows that HSIs provide useful information for individual lava flow differentiation and mapping in a complex environment such as Mt. Etna, despite limitations such as the relatively coarse pixel size, noisy bands and the sparse number of bibliographic references on lava spectral measurements.
Application of Spectral Unmixing on Hyperspectral data of the Historic volcanic products of Mt. Etna (Italy).
Published: 22 March 2018 by MDPI in 2nd International Electronic Conference on Remote Sensing session New Image Analysis Approaches
Keywords: hyperspectral data, Etna, lava flow characterization, linear spectral unmixing, bilinear spectral unmixing, fast fourier transform, structural similarity index