Spatial prediction of soil properties using environmental correlation

Conventional survey methods have efficiencies in medium to low intensity su rvey because they use relationships between soil properties and more readil y observable environmental features as a basis for mapping. However, the im plicit predictive models are qualitative, complex and rarely communicated i n a clear manner. The possibility of developing an explicit analogue of con ventional survey practice suited to medium to low intensity surveys is cons idered. A key feature is the use of quantitative environmental variables fr om digital terrain analysis and airborne gamma radiometric remote sensing t o predict the spatial distribution of soil properties. The use of these tec hnologies for quantitative soil survey is illustrated using an example from the Bago and Maragle State Forests in southeastern Australia. A design-bas ed, stratified, two-stage sampling scheme was adopted for the 50,000 ha are a using digital geology, landform and climate as stratifying variables. The landform and climate variables were generated using a high resolution digi tal elevation model with a grid size of 25 m. Site and soil data were obtai ned from 165 sites. Regression trees and generalised linear models were the n used to generate spatial predictions of soil properties using digital ter rain and gamma radiometric survey data as explanatory variables. The result ing environmental correlation models generate spatial predictions with a fi ne grain unmatched by comparable conventional survey methods. Example model s and spatial predictions are presented for soil profile depth, total phosp horus and total carbon. The models account for 42%, 78% and 54% of the vari ance present in the sample respectively. The role of spatial dependence, is sues of scale and landscape complexity are discussed along with the capture of expert knowledge. It is suggested that environmental correlation models may form a useful trend model for various forms of kriging if spatial depe ndence is evident in the residuals of the model. (C) 1999 Elsevier Science B.V. All rights reserved.