Assessing the value of information in site-response analysis

In seismic hazard assessments the importance of knowing different input parameters accurately depends on their weights within the hazard model. Many aspects of such assessments require inputs based on knowledge and data from experts. When it comes to decisions about data collection, facility owners and seismic hazard analysts need to estimate the possible added value brought by acquiring new data along with the budget and time available for its collection. In other words, they need to answer the question "Is it worth paying to obtain this information?". Assessing the value of information (VoI) before data collection should lead to optimising the time and money that one is willing to invest. In this article, we propose a method that combines available data and expert elicitation to facilitate the decision-making process within the site-response component of seismic hazard assessment. The approach combines Bayesian networks and decision trees to translate the causal-relationships between the input parameters in site-response analysis and Bayesian inference to update the model when new evidence is considered. Here, we assess VoI for a hypothetical site. Our analysis shows that VoI is highly sensitive to prior probabilities and the accuracy of the test to be performed. This highlights the importance of defining those from available information as well as considering only tests that are suitable for our needs and budget.