Spectral indices derived, non-parametric Decision Tree Classification approach to lithological mapping in the Lake Magadi area, Kenya

Abstract

Here, we demonstrate the application of Decision Tree Classification (DTC) method for lithological mapping from multi-spectral satellite imagery. The area of investigation is the Lake Magadi in the East African Rift Valley in Kenya. The work involves the collection of rock and soil samples in the field, their analyses using reflectance and emittance spectroscopy, and the processing and interpretation of Advanced Spaceborne Thermal Emission and Reflection Radiometer data through the DTC method. The latter method is strictly non-parametric, flexible and simple which does not require assumptions regarding the distributions of the input data. It has been successfully used in a wide range of classification problems. The DTC method successfully mapped the chert and trachyte series rocks, including clay minerals and evaporites of the area with higher overall accuracy (86%). Higher classification accuracies of the developed decision tree suggest its ability to adapt to noise and nonlinear relations often observed on the surface materials in space-borne spectral image data without making assumptions on the distribution of input data. Moreover, the present work found the DTC method useful in mapping lithological variations in the vast rugged terrain accurately, which are inherently equipped with different sources of noises even when subjected to considerable radiance and atmospheric correction.

Keywords

• decision tree classification
• aster data
• lithological mapping
• lake magadi