S3.5: Evaluation and quantification of errors and uncertainty in models and data to support volcanic hazard and risk assessment

Convener(s)

Valentin Gueugneau

School of Geosciences, University of South Florida, Tampa, United States of America
v.gueugneau@opgc.univ-bpclermont.fr

Andrea Bevilacqua

National Institute of Geophysics and Volcanology (INGV) Pisa, Italy
andrea.bevilacqua@ingv.it

Sylvain Charbonnier

School of Geosciences, University of South Florida, Tampa, United States of America
sylvain@usf.edu

Errors and uncertainties are inherent components in any attempt to observe,measure and predict (‘model’) our volcanic environments and their impacts on society. Epistemic uncertainty arises from a lack of perfect knowledge of the physical system, the possibility of alternative models, and limitations in our ability to pragmatically describe the system. Aleatoric uncertainty is associated with the difficulty of measurements of the natural phenomenon, the scarcity of data,the limited repeatability of observations, and irreducible randomness of volcano behavior. While we can work to reduce both epistemic and aleatoric uncertainty,they can never be eliminated; thus, it is important to quantify them when conducting any volcanic hazard assessment. The presence of errors or uncertainties do not necessarily make any model or data invalid; rather, effective assessments of accuracy and uncertainty can (1) identify limitations, (2) support model calibration,validation and benchmarking, and (3) give confidence in measurements and predictions. Furthermore, well constrained measurements of differences between reality and modelled/measured systems can improve our understanding of volcanic processes, support critical assessment of risk and ensure decisions are made using the best available data and models.This session aims to bring together scientists and practitioners to improve our understanding of volcanoes, their hazards and risk through the measurement,analysis and quantification of errors or uncertainty in both models and data. We welcome submissions from all aspects of volcanic environments, hazards and risk that quantify, use, and account for uncertainty in numerical, statistical and experimental models, as well as field, laboratory and remote sensing data. This session is supported by the IAVCEI Commissions on Statistics in Volcanology and Volcanic Hazards and Risks.

Core connection between the proposed session and societal risk mitigation: Errors and uncertainties are present in all models and data of volcanic processes as well a sin assessments of hazards, risks and benefits to society. Despite their pervasiveness, their evaluation and quantification is sometimes limited in volcanic hazard and risk assessment. This session seeks to open discussions on error and uncertainty, highlighting the benefits that quantification of data/model errors and uncertainties can bring and demonstrate ways it can enhance decision making for risk mitigation. Examples of submissions we expect to this session include model validation and benchmarking studies, model averaging approaches to improve hazard estimates and techniques to measure error/uncertainty in field and laboratory data. This session will be most relevant to decision makers such as volcano observatories, government officials and civil protection authorities who need to make effective decisions despite the presence of errors and uncertainty.

Scroll to Top