Name
Fault frictional parameter optimization using Bayesian data assimilation in a subduction zone earthquake and slow slip model
Date & Time
Tuesday, May 9, 2023, 11:00 AM - 11:15 AM
Description
Fault frictional properties play key roles in controlling the slip history in the rate-state-friction framework. The prevalent approach to constrain frictional parameters is through physical-based modelling. A first-order comparison between simulation results and observations can be useful for constraining the in-situ rate-state friction parameters applied on a natural fault zone. However, such studies are essentially conceptual, as a quantitative prediction of fault displacement requires stronger constraints on the rate-state fault parameters. Here we use GPS observations along the Northern Cascadia subduction zone (NCSZ) during the past two decades to establish a Bayesian data assimilation framework based on rate-state friction to better constrain the subduction zone fault frictional parameters.To demonstrate the feasibility of the inversion framework, we conduct synthetic experiments with surface displacement time series (i.e., synthetic GPS records) generated by earthquake cycle modeling. The synthetic time series encompassing SSEs are split into sliding time windows and used to optimize the friction parameters. Synthetic tests show that displacement time series of only 0.6 SSE (9 months for NCSZ) is sufficient to optimize the friction parameters (characteristic slip distance Dc, effective normal stress ?eff) with a reasonable initial guess. We then apply the inversion framework to the NCSZ and conduct a temporal piecewise inversion using cut GPS time series with a sliding 9-month time window. The inversion results show clearly cyclic variations of Dc and ?eff in an SSE cycle, which could be related to the proposed pore pressure build-up and release processes (fault-valve model) at the SSE depth range.
Location Name
Aspen
Full Address
Banff Park Lodge Resort Hotel & Conference Centre
201 Lynx St
Banff AB T1L 1K5
Canada
Abstract
Fault frictional properties play key roles in controlling the slip history in the rate-state-friction framework. The prevalent approach to constrain frictional parameters is through physical-based modelling. A first-order comparison between simulation results and observations can be useful for constraining the in-situ rate-state friction parameters applied on a natural fault zone. However, such studies are essentially conceptual, as a quantitative prediction of fault displacement requires stronger constraints on the rate-state fault parameters. Here we use GPS observations along the Northern Cascadia subduction zone (NCSZ) during the past two decades to establish a Bayesian data assimilation framework based on rate-state friction to better constrain the subduction zone fault frictional parameters.To demonstrate the feasibility of the inversion framework, we conduct synthetic experiments with surface displacement time series (i.e., synthetic GPS records) generated by earthquake cycle modeling. The synthetic time series encompassing SSEs are split into sliding time windows and used to optimize the friction parameters. Synthetic tests show that displacement time series of only 0.6 SSE (9 months for NCSZ) is sufficient to optimize the friction parameters (characteristic slip distance Dc, effective normal stress ?eff) with a reasonable initial guess. We then apply the inversion framework to the NCSZ and conduct a temporal piecewise inversion using cut GPS time series with a sliding 9-month time window. The inversion results show clearly cyclic variations of Dc and ?eff in an SSE cycle, which could be related to the proposed pore pressure build-up and release processes (fault-valve model) at the SSE depth range.
Session Type
Breakout Session