Evaluating the Potential of ASTER Data for Production of Soil Information at Farm Scale in South Africa

摘要

Under the Farmer Resettlement and Agricultural Development Program, the government of South Africa is working on the implementation of an integrated program for land re-distribution and agricultural development. As many farms lack detailed soil maps, basic in the process of land capability evaluation, a method for rapid mapping of relevant soil characteristics at a scale detailed enough for sound farming planning decisions needs to be developed. Air- and satellite-borne sensors can measure energy reflected or emitted from the land surface over a wide range of spatial, temporal and radiometric resolutions that using appropriate image processing techniques can produce topographic and landcover information relevant to soil-landscape studies. Accordingly, this paper discusses the initial framework designed for the rapid production of soil information at farm scale that integrates the three 15-m resolution bands of the ASTER instrument on-board the Terra platform, with topographic and coarse scale geologic and soil data, and field sampling of selected soil properties, for spatial analysis and modelling of geomorphic units and associated soil characteristics within a geographic information system. An approach combining segmentation of the ASTER images into relatively homogenous spectral zones, and generation of terrain types from a DEM was designed to this end. The output units resulting from the overlay of these information layers were characterised by field observations of selected soil properties using a stratified random sampling scheme. Initial results suggest that combined field and remote sensing approaches as proposed here, enable more efficient location of soil samples, reducing the time and money that is spent for field survey required at this scale (for example, 1:20,000 - 1:25,000). Furthermore, the use of ASTER data as reported in this study appear as a reliable tool for supporting soil mapping at scales as detailed as 1:20,000. The availability of a good resolution Digital Elevation Model largely impacts on the accuracy of the landforms derived in the study. Evidence suggests that a reduction between 60 up to 80 percent of field samples, as compared to conventional soil survey techniques, could be achieved by integrating ASTER data into soil-landscape studies.