Post-processing of mineral mixture maps for mapping surficial materials: The example of the Matmata loess, southern Tunisia

Abstract

Spectral mixture modelling algorithms can be applied to multispectral image data to provide estimates of mineral proportions within pixels. Most surficial materials are mixtures of different minerals, but their spatial distribution can be mapped by applying simple thresholding techniques to mineral proportions maps, as long as their mineralogical composition and surficial exposure are known. An example is given of the Matmata loess in southern Tunisia. Thick deposits of loess are found in valleys and basins of the Matmata plateau, and they have been shown to record Late Quaternary climatic changes as a sequence of weathered palaeosols developed between layers of unweathered loess. However, few sites have been studied in detail thus far, and more potential sites need to be identified for further investigation in the field. Remote sensing helps overcome the problem of inaccessibility in this rugged and arid terrain to facilitate such reconnaissance. Mineral proportions maps derived from Landsat Thematic Mapper data are calibrated against field samples and thresholded to show pixels that have the same mineral composition as the loess. The results indicate that loess is not only found as valley and basin fill on the Matmata Plateau; but also as loess-derived soils and stone pavement matrix in the area surrounding the plateau. To discriminate the valley and basin fill of palaeo environmental significance from these other loess-rich surficial materials, it is necessary to include digital elevation data in the analysis.