The Oman ophiolite is well exposed, and all the major structural and lithological features are visible on suitably processed Landsat images. Visual interpretation of such images provides a synoptic view of the Oman mountains which is in many aspects as useful as inspection of published maps at scales of 1:500,000 to 1:100,000. This study is an investigation of image processing for semi-automated lithological mapping. The differences between the visible and near-infrared multispectral signatures of ophiolite rock types in the Oman mountains have been shown by in-situ measurements to be considerably less than for fresh, particulate samples of similar rocks, in the Landsat MSS spectral bands. The muting of their reflectance spectra is due to a combination of alteration and weathering, and this hinders discrimination between rock types on MSS images. The problem is further compounded by topographically induced variations in brightness (BV), and the presence of shadows containing no valid data. The first-order topographic effects can be partly overcome by the generation of intensity-normalized bands. By using training sets of less than 2 km2 from within an area of less than 2000 km2; a thematic map can be produced which shows the approximate distribution of major rock types over virtually an entire MSS scene (30,000 km2). The problem of misclassification due to topography and other effects can be largely side-stepped by the use of a suitable post-classification filter.

The Oman ophiolite is well exposed, and all the major structural and lithological features are visible on suitably processed Landsat images. Visual interpretation of such images provides a synoptic view of the Oman mountains which is in many aspects as useful as inspection of published maps at scale...