ASTER AND LANDSAT LINEAMENTS IN CENTRAL NYS: IMAGE PROCESSING AND GROUNDTRUTHING FOR FRACTURES

Lineaments can be surface representations of deeper structures such as faults and fractures. Knowledge of fault locations is important for several reasons including the safe disposal of hazardous waste products, risk assessment for earthquakes, as well as potential locations of oil and gas reservoirs. Satellite images were used to select lineaments in the Finger Lakes region of New York State.

Satellite images were enhanced using various algorithms to accentuate the visibility of lineaments. Image enhancements included: 1) a principal component analysis of ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) data, 2) a combination of ASTER and Landsat bands that have the least correlation, and 3) Landsat TM image (7,4,2) from an older study. ASTER is a sensor that has high band sensitivity and spectral resolution, which makes it particularly sensitive to small wavelength changes. ASTER has one data point per 0.01 µm, which makes it an order of magnitude better spectral resolution than Landsat TM, which has one data point per 0.27 µm. ASTER also has a higher spatial resolution in the Visible and Near Infrared: 15m compared to 30m for Landsat. These factors make the ASTER sensor potentially more powerful than Landsat, for lineament identification.

Lineaments were identified on the satellite images in regions of existing detailed structure field data. In order to groundtruth the lineaments from the three different image processing methodologies, we utilized the weights of evidence method. In this technique we determined whether fractures (including Fracture Intensification Domains) and lineaments of the same trend are coincident.

Results from this study show that along Seneca Lake there is good spatial correlation between fractures and lineaments that trend NNE to E; correlation was better if the lineaments were constrained to greater than 500m in length. The Cayuga Lake region also showed a good correlation in the NE-striking to E-striking orientations; however, the correlation was worse for long lineaments. Decreasing the buffer size of the lineaments from 500m to 250m in each of the lineament orientations did not improve the values.

ArcGIS was used to complete a quantitative comparison between structural field data and lineaments drawn from satellite imagery in ENVI.