Characterization of Hydrocarbon Microseepages in the Tucano Basin, (brazil) through Hyperspectral Classification and Neural Network Analysis of Advanced Spaceborne Thermal Emission and Reflection Radiometer (aster) Data

This study focus on the characterization of hydrocarbon microseepages in the northern Tucano Basin (Bahia State, Brazil), using geostatistical analysis of regional hydrocarbon geochemical data yielded from soil samples and digital processing of Enhanced Thematic plus (ETM+/Landsat7 satellite) and Advanced Spaceborne Thermal Emission and Reflection Radiometer imagery (ASTER/Terra satellite). A theoretical detection model was devised in which gas anomalies (seeps) indicated by hydrocarbon geochemistry should spatially match a number of surface expressions, such as bleaching of soils and rocks (i.e., reduction of Fe3+ to Fe2+ ), geobotanical markers and development of specific clays (kaolinite) and carbonates (siderite). These indirect evidences were employed for the application of remote sensing data and information extraction techniques in order to locate sites more favorable to host hydrocarbon seeps in the Tucano Basin. The ETM+ data was processed using the pseudo-ratio technique an adaptation of the classic principal components analysis. The ASTER data was processed using the Spectral Angle Mapper and the Mixture Tuned Matched Filtering techniques, which are commonly used for the processing of hyperspectral data, though adapted here for these multispectral dataset. In addition, the ASTER data were classified using three different neural network systems (Fuzzy Clustering, Radial Basis Functional Link Network and Probabilistic Neural Network). The results showed that a number of the sites predicted using the applied detection model concurred with known geochemical anomalies. Other sites with similar characteristics but for which no geochemical data were available were also revealed. These sites are taken as new potential targets for the presence of seeps and oil reservoirs. The research demonstrated the excellent potential of ASTER data and spectral-spatial methodologies for low-cost, onshore exploration of hydrocarbons in Brazil.