Lithologic mapping of a forested montane terrain from Landsat 5 TM image

Abstract

Thick forest cover and poor infrastructures are the major hindrances for detailed lithologic mapping in an inaccessible montane landscape. To overcome these limitations, we utilize a Landsat 5 TM image to map lithology using vegetation and drainage pattern as an indicator of underlying rock types in a heavily forested region of the Chittagong Hill Tracts area located in southeastern Bangladesh. We use supervised and unsupervised classifiers for a vegetation-based approach while on-screen digitization is used for drainage patterns-based mapping. Field observations were used for mapping lithology and evaluating accuracy. Overall, our results agree well with the current geologic map and improve it by providing a more spatially detailed distribution of the sandstone and shale. The performances of all approaches are good at the inner and outer flanks of anticlines located in the study area while the drainage pattern mapping performs best at the mid-flank area.

Keywords:

• Trellis
• dendritic
• sandstone
• shale
• supervised classification
• unsupervised classification

Acknowledgements

We would like to thank Dr. J. W. Taco Bottema, team leader of CHT development project and Delta Study Center, the University of Dhaka and NSF funded ND EPSCoR (NSF grant #IIA-135466) for providing financial support for this study in the Chittagong Hill Tracts area. Special thanks to Meghna Estuary Research Center and CEGIS, for providing satellite imagery and other related dataset. Finally, we would like to thank Dr. Nels Forsman of the University of North Dakota for his valuable suggestions and edits which have improved our manuscript markedly.