KEM-09 Sensitivity analysis of the public seismic hazard and risk assessment model train

It is the objective to verify and quantify the cumulative effect on the predicted hazards and risks and inherent uncertainties under different production scenarios caused by model train component (1) input data, (2) model choices and assumptions, uncertainty propagation in the model train etc. and their relative contribution to overall uncertainty. It is the objective to especially review the scientific and probabilistic soundness of the working hypotheses with respect to the state-of-the-art knowledge about performance-based seismic risk assessment.
To rank model parameters, assumptions, model choice uncertainties in terms of contributions on resulting impact uncertainty in calculated hazard or risks. To show whether the statistical approach followed in the Groningen hazard and risk model of NAM and the public model of TNO over- or underestimating the predicted impacts? To formulate recommendations to validate the probabilistic framework and to possibly reduce the uncertainties with respect to the working assumptions of the Groningen seismicity hazard and risk model train.

FULL RESEARCH QUESTION

The KEM-09 project aimed at evaluating how much the modelling assumptions and the uncertainty characterization, necessarily taken in the seismic risk assessment for Groningen, reflect on the risk results. This is useful to understand the robustness of the risk results and evaluate further proposals/needs of future model developments.

KEM PROJECT FINAL REPORT

The project has developed and applied an extensive Seismic Hazard and Risk Analysis (sub) models testing framework, which allows for extensive sensitivity analysis and (component) model comparisons.

One of the main conclusions is that risk is sensitive to all model components and all model components (SSM, GMM, FCM) have a significant impact on the results, at least according to the criteria chosen by the team. The study has highlighted the difficulty in reproducing the risk assessment from the Groningen field operator, which should be solved in further developments of the seismic risk assessment for Groningen. The study has highlighted some possible inconsistencies in some model components, which may call for adjustment and fix.

The way the logic tree is shaped, and the weights given to the branches have a relevant effect on the risk such that it could be helpful to establish a codified procedure to determine these weights in future developments of the risk assessment. The study has shown how some parts of the logic tree produce extremely right-skewed risk results, which have a significant relatively large effect on the risk assessment, this may call for further analysis and deepening as well as considering other risk metrics on which to evaluate the impact.

The study has shown the importance of the assessment of the impact of model components based on the effect on risk metric and has paved the way to include this kind of work in future developments, also for other mining.

The project addressed two main questions: (i) how some aggregate risk metrics for Groningen are sensitive to some models in the risk analysis pertaining to all the three main components (SSM, GMM, FCM); (ii) how the modelling of the epistemic uncertainty propagates through the logic tree and, therefore, which elements of it most affect the results.

KEM PROJECT EVALUATION

The study has shown the importance of the assessment of the impact of model components based on the effect on risk metric and has paved the way to include this kind of work in future developments, also for other mining. These are valuable results and have provided insights that were not available so far. The team also alerted about possible inconsistencies in the risk analysis; for example, between GMM and FCM models.

KEM-09 can set the base for a standard for the assessment of the need and relevance of risk model updates/development for Groningen and possibly for other mining activities in the Netherlands. In more general terms it could be an important part of the risk assessment itself and of its quality assurance.