Incorporating the Physics of SSEs into Time-dependent Forecasting of Subduction System Main Shock Earthquakes

Abstract:

Many SSE-related phenomena occur over times-scales of seconds to weeks or months. Physical properties on or around megathrust faults are also changing on similarly short time-scales. Can these properties be used to improve prospective earthquake forecasting?

Densification of continuous GPS and seismic networks and advances in analytical methods allow us to monitor SSE-related phenomena in near real time. This monitoring provides an enticing new field of research combining observation, modelling, and statistical methods that could eventually lead to time-dependent estimates of hazard from deadly earthquakes, much like volcanic activity is now forecast in terms of 'alert levels'.

Information from SSEs has potential to improve medium-term (years to decades) and short-term (days to months) earthquake forecasting models that are used, for example, in government planning, disaster preparation, and risk mitigation. Many physical models of SSEs invoke changes in fault strength. There is communication between the megathrust and upper plate faults, suggesting megathrust events affect rates of crustal earthquakes. Models of dynamic and static triggering are beginning to explain inter-related earthquake phenomena. However, this vast new understanding has not been fully utilised in the quantification of hazard. Models of fault interaction based on our newest and best science require rigorous testing and validation prior to implementation into operational background or time-dependent forecasts.

We have implemented a method to test the effect of including time-dependent parameters associated with the subduction system in earthquake forecasting. We create multiplicative hybrid earthquake likelihood models from a baseline model and other gridded variables. The changing spatial distributions of megathrust slip during SSEs, accumulated stress, and fault strength are potential candidate variables. We are developing methods of generating useful monitoring variables and testing them. The multiplicative modelling method can independently evaluate information gain from each variable. Incorporating SSE monitoring efforts into earthquake hazard studies through multiplicative models provides a pathway toward time-dependent estimation of earthquake hazard throughout the subduction system.