Toward verification of a natural resource uncertainty model

Natural resource management models simplify reality for the purpose of planning or management. In much the same way, an uncertainty model simplifies the many uncertainties that pervade the natural resource management model. However, though a number of uncertainty models have been developed, there has been little work on verifying such models against the uncertainty they purport to represent. The central research question addressed by this work is 'can a natural resource management uncertainty model be verified in order to evaluate its utility in real-world management?' Methods to verity uncertainty models are developed in two areas: uncertainty data models, and uncertainty propagation through process models. General methods are developed, and then applied to a specific case study: slope stability uncertainty in the southern Queen Charlotte Islands. Verification of two typical uncertainty data models (of classified soils and continuous slope) demonstrates that (in this case) both expert opinion inputs and published error statistics underestimate the level of uncertainty that exists in reality. Methods are developed to recalibrate the data models, and the recalibrated data are used as input to an uncertainty propagation model. Exploratory analysis methods are then used to verify the output of this model, comparing it with a high-resolution mass wastage database—itself developed using a new set of tools incorporating uncertainty visualisation. Exploratory data analysis and statistical analysis of the verification shows that, given the nature of slope stability modelling, it is not possible to directly verify variability in the model outputs due to the existing distribution of slope variability (based on the nature of slope modelling). However, the verification work indicates that the information retained in uncertaintybased process models allows increased predictive accuracy—in this case of slope failure. It is noted that these verified models and their data increase real-world management and planning options at all levels of resource management. Operational utility is demonstrated throughout this work. Increased strategic planning utility is discussed, and a call is made for integrative studies of uncertainty model verification at this level.