金曜日セミナー(6月26日)Ylona van Dinther 氏 (Utrecht University)
Combining physics-based models and observations to understand Tectonics and Seismicity
Abstract:
Earthquake cycle models are becoming increasingly more realistic in terms of physical ingredients and resemblance to geological observations. One such example is our seismo-thermo-mechanical (STM) modeling approach, which simulates tectonics and seismicity using a visco-elasto-plastic rheology with laboratory-derived parameters for different lithologies. Physics-based models can reveal their power through a close link to observations or by predicting new observations. Observations are our stronghold towards the truth, but they are indirect and limited in space and time. I will first discuss how these observations can be used to constrain dynamic earthquake rupture models to show that landslides are not needed to explain most of the inundation in Palu bay following the 2018 Sulawesi earthquake-tsunami [1]. Then I will highlight different examples on how we relate STM models and observations to better understand the seismic cycle.
Geological and geophysical observations can be used to tightly constrain our STM models. In an example in the Northern Apennines, a peculiar orogenic belt driven by slab pull, we show that lower crustal rheology and deep lithospheric mantle temperatures distinctly affect crustal tectonics and seismicity [2]. Through adding seismicity as a new observable we may thus be able to constrain such irretrievable states better.
One of the ultimate goals of numerical models is to predict new features or processes that can be verified with new observations. STM models predicted the existence of a secondary zone of uplift due to great megathrust earthquakes, which was subsequently identified in vertical displacements of four out of four studied M> 8.5 earthquakes [3]. Physics-based, self-consistent models can then be used to better understand the governing physical mechanisms reaching from hundreds of kilometers depth to the surface.
Ultimately, one effective way to learn uncover nature’s secrets is by combining physics-based models and observations in a quantitative and probabilistic manner, such that unrecoverable critical parameters or states can be estimated and understood. I will show how we demonstrate in a simplified proof of concept that ensemble data assimilation, as borrowed from weather forecasting, can quantitatively merge physics-based models and observations to estimate and forecast fault stresses [4].
References:
- Ulrich et al., PAGeoph, 2019: https://link.springer.com/article/10.1007/s00024-019-02290-5
- D’Acquisto et al., Tectonophysics, 2020: https://www.sciencedirect.com/science/article/pii/S0040195120301645
- van Dinther et al., PAGeoph, 2019: https://doi.org/10.1007/s00024-019-02250-z
- van Dinther et al., GJI, 2019: https://doi.org/10.1093/gji/ggz063