Empirical and Mechanistic Modelling for Process Understanding in Digital Soil Mapping

Abstract

Empirical prediction of soil properties coupled with an understanding of soil processes, can uncover the complexity of the soil system. Digital soil mapping (DSM) has revolutionized the way soil information is delivered. While empirical DSM has greatly improved the quantitative prediction, we should be able to incorporate our physical and mechanistic understanding of the processes. Likewise, we should be able to use empirical knowledge to inform process-based models. This thesis delivers mechanistic and empirical models to improve the understanding of soil genesis and mapping of soil functional properties and finding the relationships between soil and environmental factors. Chapter 2 first critically reviews pedology models and DSM concepts, mapping soil classes, mapping soil profiles, mapping pedological features and processes, the relation between pedological knowledge and DSM, and the application of mechanistic pedological models in DSM. Chapter 3 investigates the use of a mechanistic pedogenesis model, State Space Soil Production and Assessment Model (SSSPAM) for modelling the spatiotemporal evolution of particle-size distribution (PSD). In Chapter 4, we used process-based understanding in a mechanistic model to help us make a better prediction of the 4D spatiotemporal distribution of SOC. Chapter 5 evaluates the proposition that soil properties can be evaluated at any depth by comparing the multi-layered 2.5D and 3D modelling with soil depth as a predictor variable. Chapter 6 investigates whether data provided from a rapid and non-destructive proximal sensor can be used to directly predict the provenance of soil samples. Overall, this thesis demonstrates that to comprehensively explain the complexity of the soils, their dynamics and relation to the soil-forming factors, it is beneficial to include knowledge of processes to model soil profile distribution and identify the unique pattern of soil distribution.