Practical communication of uncertainty in quantitative landslide risk assessment

Uncertainty is inherent in all quantitative landslide risk assessment. However, rarely are the outcomes of a quantitative landslide risk assessment accompanied by an indication of the uncertainty of the assessment. Landslide risk assessment requires the prediction of natural climate, geological and geomorphological processes which cannot be done with complete certainty. Further uncertainty is introduced through error in measurement and observation, variable judgement and bias. This paper sets out practical means by which the uncertainty of landslide risk assessed quantitatively and reported as a probability can be estimated and then communicated to stakeholders or users of the quantitative landslide risk assessment. Methods for estimating epistemic uncertainty arising from error in observation and its interpretation and aleatory uncertainty arising from the unpredictability of natural systems and processes are set out. The uncertainty assessment is based on the estimation of indices using a scoring system. Indices for epistemic uncertainty are linked to the information used to inform the landslide risk assessment and the interpretation of that information. Indices for aleatory uncertainty are linked to the complexity or predictability of the natural system in which the landslide could occur. Various methods are presented for how the assessed uncertainty can then be communicated to a stakeholder or user of the landslide risk assessment to convey to that user a realistic expectation of the confidence they should have in what is reported to them. These include presenting the uncertainty as an index, describing uncertainty qualitatively, presenting the risk as a range or assuming the risk estimation follows a statistical distribution and presenting confidence limits. Examples of how to estimate and communicate uncertainty are presented for situations involving both the assessment of individual risk and societal risk.