Document Type

Thesis - University Access Only

Award Date


Degree Name

Master of Science (MS)

Department / School



Multi-temporal Landsat TM and SIR-C images were used to classify wetlands in the Florida panhandle. TM bands 3, 4, 5, and 7, and SIR-C L-band HH, L-band HV, and C-band HH, were co-registered and clipped to create a combined data set. An unsupervised maximum likelihood classifier was used to initially classify the combined data. In the post classification phase, training samples derived from National Wetlands Inventory (NWI) data were used to train classification trees. The final classified image was compared to a classification derived without the use of radar. Despite the low overall accuracy (50%), visual comparison of the images shows that the two sensors provide unique information, which when combined should more completely describes the surface. Dropping one wetland class from the error matrix improved the overall classification accuracy by 15%. Such a large improvement suggests that the classification scheme used in this study may not be suitable for the identification of wetlands in the study area. Additional factors affecting accuracy, such as a temporal difference between the NWI data and the image data, are discussed in detail.

Library of Congress Subject Headings

Wetlands -- Florida -- Classification -- Remote sensing
Vegetation classification -- Remote sensing
Artificial satellites in remote sensing



Number of Pages



South Dakota State University



Rights Statement

In Copyright