Soil surveyor knowledge in digital soil mapping and assessment in Australia

Abstract In the last three decades, land resource assessment and soil survey has shifted from an expert-driven qualitative soil-landscape paradigm to a quantitative, correlative environmental modelling approach now commonly referred to as ‘digital soil mapping and assessment’ (DSMA). The role of soil surveyors in producing spatial soil and land suitability assessments under an operational paradigm of digital soil mapping is discussed using the recently completed Northern Australia Water Resources Assessment (NAWRA) project as a case study. Real world DSMA problems are presented, and pragmatic solutions suggested. A stratified random sampling plan for data collection on ~500 new sites in the field was produced using conditioned Latin hypercube sampling using 30-m resolution environmental data layers that each represented a factor of soil formation but had low correlations between them. Some free survey sites were added in soil-landscapes that were under-represented in the stratified random sampling plan in the judgment of the soil surveyors. Predictive random forest models and gridded maps for soil classes and continuous soil attributes (and their prediction uncertainty) were produced at 90-m resolution in an iterative process that enabled the identification of issues, particularly in the legacy data included for model training. As a consequence errors in data entry in the state soil databases have been corrected. Another stratified random sampling plan using the prediction uncertainties was produced to select sites for field validation. Soil surveyors played a key role in all aspects of the DSMA workflow including identifying soil-landscapes that need to be mapped accurately and applying expert knowledge to the selection of final maps of soil classes and attributes necessary to implement land suitability rules.