Challenge

Often, hazard forecast models are implemented within a discipline specific context which means they integrate discipline specific concepts as the expense of reusing some of the existing methodological approaches being developed more broadly within the statistical community. At the same time, some of these more generic frameworks lack specific functionality required to model specific processes observed in the earth’s system. Here we exploit a close collaboration between the statistical experts developing the inlabru framework and geoscientists with hazard specific expertise to build a new generation of hazard forecasts.

inlabru is a Bayesian spatio-temporal modelling code that holds great potential for modelling, simulating and forecasting in many aspects of earth and environmental science, particularly for natural hazards. It is built on R which allows us to re-use many of the existing packages for handling spatial datasets; for example, we can load shape files for fault networks and directly use them as a covariate in our earthquake forecasts.

Currently, inlabru does not implement the Hawkes process – this means it currently lacks some of the functionality to incorporate self-exciting clustering seen during aftershock sequences. This is the focus of Francesco’s PhD which is funded by the H2020 project.

Blog Articles

• PRE-PRINT PAPER: Ranking earthquake forecasts using proper scoring rules: Binary events in a low probability environment (2021)
• PRESENTATION: Ranking earthquake forecasts: On the use of proper scoring rules to discriminate forecasts (2021)
• PRESENTATION: Developing Data-Driven Earthquake Forecasts Using Inlabru (2021)
• Hot off the press: Data-driven optimisation of seismicity models using diverse datasets (2020)
• Diary of a PhD Student: What a year! (2020)

Project Team

Dr Kirsty Bayliss (H2020 PDRA) Forecasting time independent seismicity using inlabru

Francesco Serafini (H2020 PhD) Forecasting the spatio-temporal evolution of seismicity using inlabru

Gina Geffers (NERC E4 PhD) Identifying links between induced and natural seismicity

Lead: Dr Mark Naylor (Edinburgh): Hazard modelling with specific expertise in seismic hazard

Prof Finn Lindgren (Edinburgh): Statistician and core inlabru developer

Prof Janine Illian (Glasgow): Applied statistician

Prof Ian Main (Edinburgh): Seismologist and RISE work package lead

Publications

• Serafini, Naylor, Lindgren, Werner, and Main, Ranking earthquake forecasts using proper scoring rules: Binary events in a low probability environment [PrePrint Server]
• Bayliss, Naylor, Illian and Main (2020) Data-driven optimisation of seismicity in models using diverse datasets: generation, evaluation and ranking using inlabru, Journal of Geophysical Research: Solid Earth, 125, e2020JB020226, https://doi.org/10.1029/2020JB020226
• Bayliss, Naylor and Main (2019) Probabilistic identification of earthquake clusters using rescaled nearest neighbour distance networks, Geophysical Journal International 217 (1), 487-503

Funding

• CoI 2019-22 £450k to Edinburgh (€3M) H2020Real-time earthquake risk reduction for a resilience Europe (RISE) Within an innovation work-package, we are adapting and trialling the use of inlabru to model time dependent seismic hazard.
  
• CoI 2017-20  £323k to Edinburgh (£700k) NSFGEO-NERC: The central Apennines earthquake cascade under a new microscope This collaborative project between the BGS, UoE, Stanford and INGV will work up the data collected from the urgency grant to deliver a world leading earthquake catalogue for exploring earthquake process and aftershock forecasting.
• EPSRC Studentship for Kirsty Bayliss
• NERC E4 Studentship for Gina Geffers

Share