Inferring the Spatial Distribution of Regolith Properties Using Surface Measurable Features
Abstract
The aim of this research is to determine to what extent properties of the regolith may be inferred using only features easily measured from the surface. To address this research question, a set of regolith properties from Weipa, Queensland, Australia, are analysed. The set contains five variables, oxides of Aluminium, Iron, Silica and Titanium, as well as Depth to Ironstone. This last represents the depth of the layer from which the oxides are sampled.¶ The research question is addressed in two ways. First, locations where the properties are related to modern surface hydrology are assessed using spatially explicit analyses. This is done by comparing the results of spatial association statistics using geometric and watershed-based spatial samples. Second, correlations are sought for between the regolith properties and geomorphometric indices of land surface morphology and Landsat Thematic Mapper spectral response. This is done using spatially implicit Artificial Neural Networks (ANN) and spatially explicit Geographically Weighted Regression (GWR). The results indicate that the degree to which regolith properties are related to surface measurable features is limited and spatially variable.¶ ... ¶ The implications of these results are significant for anyone intending to generate spatial datasets of regolith properties. If there is a low spatial density of sample data, then the effects of landscape evolution can reduce the utility of any analysis results. Instead, spatially dense, direct measurements of subsurface regolith properties are needed. While these may not be a direct measurement of the property of interest, they may provide useful additional information by which these may be inferred.