|
Leonardo Disperati, University of Siena (Italy)
Riccardo Salvini, University of Siena (Italy)
Mauro Coltorti, University of Siena (Italy)
Alberto Pizzi, University of Chieti (Italy)
Bekele Abebe, University of Addis Ababa (Ethiopia)
Giuseppe Pomposo, University of Chieti (Italy)
Giorgio Sacchi, University of Siena (Italy)
Laura Pontarelli, University of Chieti (Italy)
Dario Firuzabadi, University of Siena (Italy)
|
|
The electromagnetic spectral interval 0.4-2.5 m holds relevant information concerning rock-forming minerals. Diagnostic capability is particularly effective in the 1.5-2.5 m wavelength domain, where processing of hyper- multi-spectral data permits improved rock discrimination for geologic mapping. In order to support processing of remote sensing imagery, data on spectral properties of ground features can be collected by means of spectroradiometers, which allow to get both field and lab spectra. In fact, measured vs. literature spectra allow to account for the effects of weathering processes, local lithologic features and mixing.
The Dire Dawa area provides insights in the southern margin of the East Africa Rift system. Late Triassic-Jurassic sandstones (Adigrat formation) and carbonate (Antalo limestone) sequences lie in nonconformity over the pre-Cambrian metamorphic basement and are disconformably covered by Early Cretaceous sandstones and conglomerates (Amba Aradam formation). A further disconformity was modelled before the emplacement (Oligocene) of flood basalts. Characters of outcropping rocks, lack of widespread weathering and scarce vegetation cover make the Dire Dawa region a good area in order to extract lithology information by means of ASTER satellite imagery. ASTER sensor is made up of three independent sub-systems: the VNIR system (3 bands, spectral range 0.52-0.86 m, GIFOV 15 m), the SWIR system (6 bands, spectral range 1.60-2.43 m, GIFOV 30 m) and the TIR system (5 bands, spectral range 8.125-11.65 m, GIFOV 90 m). In this study we processed VNIR and SWIR data related to two ASTER scenes L1B of 2001 and 2005.
We acquired during a fieldwork mission in 2007, reflectance spectra (range 0.35-2.5 m) for the most widespread geologic formations, some superficial formations and grass-bush vegetation. More spectra were also collected in laboratory from rock samples.
The images were registered by rectification to the official topographic maps at the scale of 1:50,000. The above spectra were used in order to support calibration of images to reflectance. To account for the variation of atmospheric effects caused by large elevation variation within the study area, we implemented a procedure considering discrete (200 m thick) atmosphere layers.
Lithologic discrimination was achieved by two approaches. Taking into account the characters of spectra from images and lab/fieldwork, we searched for the optimum thresholds of multidimensional compositions of ratios and normalised ratios of bands 4-9. Moreover, we classified principal components of bands 1-9. By integrating the outputs we obtained a map of outcropping formations. The Adigrat formation was not mapped because of its thickness, generally smaller than the ASTER GIFOV. The Antalo limestone and basalts were accurately recognised and mapped. Finally, the metamorphic basement and the Amba Aradam formation, both containing large quartz quantities, were discriminated with higher uncertainty.
|
