Ensemble data assimilation methods for estimating fault slip and future earthquake occurrences

In this dissertation, I explore ensemble data assimilation methods to enhance our capability to forecast earthquakes and slow slip events, focusing on the critical challenge posed by limited information on the current stress state of faults.
At the outset, the research acknowledges the inherent limitations in our current understanding of fault stress states. These limitations significantly hinder our ability to forecast seismic events accurately. The study proposes utilizing ensemble data assimilation techniques as a robust solution. Central to this dissertation, these methods enable estimating the fault’s state by integrating information from physics-based models with observational data. Importantly, this approach considers the uncertainties inherent in both the models and the data, offering a more reliable framework for forecasting. The dissertation emphasizes that probabilistic forecasts represent the highest achievable goal in earthquake forecasting. However, it also recognizes the challenges that arise from limited information on critical aspects such as stress, strength, and governing parameters of seismogenic sources. These limitations can significantly impede the accuracy of forecasts.
Throughout the dissertation, I systematically examine how ensemble data assimilation can be effectively implemented to improve earthquake forecasting. This involves exploring current fieldmeasurement techniques and the data quality they produce. The study demonstrates how ensemble data assimilation can bridge the gap between empirical observations and theoretical understanding by carefully analyzing and integrating this data with advanced theoretical models.
A core component of the research is a critical evaluation of various data assimilation techniques, mainly focusing on their ability to enhance forecasting accuracy in the context of limited information on fault stress states.
Furthermore, I explore practical applications of these techniques using 1D and 2D models. This includes investigating how data assimilation can improve the forecasting of earthquake occurrences and the inherent challenges in making such estimations.
The dissertation culminates in a forward-looking discussion on the future of earthquake forecasting. It emphasizes the role of ensemble data assimilation methods in overcoming the current limitations of stress state information and proposes ways for more informed seismic forecasts.