The influence of variations in prior information and design criteria on optimal designs of seismic source location surveys

SUMMARY

Estimating accurate locations of seismic sources requires appropriately designed surveys. Optimal designs maximize location accuracy or any other desired criterion, given a budget of equipment, time, or other resources. In this study, we empirically assess how variations in prior information and design criteria affect the performance of optimized seismic networks using both real and synthetic data from the Cuolm da Vi slope instability. We benchmark network designs using 95 controlled explosions recorded by 878 sensors. We explore how prior information about source locations and velocity models (and their uncertainties), data correlations, signal-to-noise ratio reduction with distance, optimization methods, and design criteria influence the quality of optimized receiver layouts. We find that designs optimized using fully nonlinear Bayesian experimental design methods consistently outperform non-optimized designs, with median improvements of approximately 25 per cent in source localization accuracy for our real data test case. The effectiveness of optimizing designs increases with the number of receivers, up to around 6 to 10 in an area of $\sim$0.3 km$^2$, after which the marginal gain drops off. Our results highlight the importance of accurately modelling the prior distribution of source locations, and of accounting for at least one of either attenuation effects or velocity uncertainties when designing seismic networks. Optimized designs are robust to reasonable variations in design parameters, particularly the number of events used to sample the prior distribution, information gain estimation methods, and optimization approaches. This work provides practical guidance for deploying optimal seismic networks in complex geological settings, where accurate source locations are essential for hazard monitoring.