Abstract: Discrimination of Surficial Materials in Part of Mississippi River Deltaic Plain Using Computer-Assisted Processing Techniques

Ralph N. Baker, Allan M. Thompson

Seasonal LANDSAT imagery of southeastern Louisiana (scenes 1195-16025/Feb 73, 1393-16015/Aug 73, and 1843-15502/Nov 74) were analyzed using General Electric Co.'s "IMAGE 100" multispectral image-analysis system to determine the ability to discriminate surficial materials under different conditions of vegetation vigor and illumination angle. Classifying the deltaic plain consistently to level I of the USGS Land Cover Classification System was possible using different combinations of band ratioing (primarily, bands 4/5, 5/6, and 6/7) followed by contrast stretching of the ratio images. Vegetation and sun-angle variations between the scenes made it possible to classify the surface to level II in some cases. A pseudo "true" color composite was generated by contrast stretching bands 4, 5, and 7, assigning the standard blue-green-red sequence, and adjusting the scaling factors and color balance.

The delta test site consisted primarily of salt and freshwater marsh and swamp, natural and artificial levees, lakes, and some urbanization and agriculture. Surface-sediment patterns are enhanced strikingly by band ratioing (difference/sum) as are the levees.

A simple band-ratio composite (4/5 = blue, 5/6 = green, and 6/7 = red) appeared to separate barren from vegetated or industrial areas. Best overall classification of the test site was accomplished using a normalization ratio of the winter (Feb 73) imagery.

One-dimensional training (semisupervised) was applied to a 30 × 30-km test site east of Barataria Bay, with classification accuracy to level II in many cases.

The computer-assisted enhancement and classification techniques employed can be applied to wetlands mapping, surface current/surface sediment movement, and understanding modern sedimentary environments.

AAPG Search and Discovery Article #90968©1977 AAPG-SEPM Annual Convention and Exhibition, Washington, DC