2:45 PM - 3:00 PM
[SCG52-17] Predictive decay functions to forecast long-term GNSS time series after the 2011 Tohoku-oki earthquake
Keywords:Predicting function model, GNSS time series, Postseismic deformation, 2011 Tohoku earthquake , Viscoelastic relaxation, Afterslip
Introduction
Logarithmic or exponential decay functions have been popular choices for predicting geodetic time series over the last few decades. In particular, combinations of logarithmic or exponential functions are demonstrated more effective in explaining the postseismic GNSS time series (Tobita 2016, EPS; Fujiwara et al. 2022, EPS). Multiple decay functions are used to proxy the Maxwell or bivious rheology (Hetland and Hager 2006, GJI) along with the aseismic slip in the postseismic period (Marone 1998, Annu Rev). The recent postseismic model illustrated an essential upgrade to power-law bi-viscous rheology which provides substantial simulation of the time-dependent postseismic deformation (Agata et al. 2019, Nat. commun.; Muto et al. 2019, Sci. Adv.; Barbot 2020, EPS). However, a closer analytical representation of such stress-dependent power-law rheology is yet to be incorporated into function models. Moreover, neither the logarithmic nor exponential decay functions can be directly correlated with the constitutive laws of rock physics, considered in most stress-dependent postseismic deformation models.
Materials and methods
We proposed alternative decay functions that incorporate power-law dependencies, learned from the laboratory experiments of rock deformations (Barbot et al. 2009, JGR; Muto et al. 2019, Sci. Adv.). We present two decay functions representing continuous afterslip on the plate boundary fault and viscoelastic relaxations in the surrounding mantle. We deploy a linear combination of these decay functions (as an analytical function model) by conducting a curve-fitting to GNSS dataset of 2 years after the mainshock (Tobita 2016, EPS). By evaluating an additional 8 years of GNSS time series, we conclude the substantial performance of our function model in the long-term prediction of most GNSS stations in NE Japan. We ensured the stability and robustness of our function model by tunning the model parameters with Bayesian optimization (Fukuda & Johnson 2021, JGR).
Results and discussion
Our decay functions mimic the contribution of the aforementioned deformation mechanisms in the surface displacements, observed after the 2011 Tohoku-oki earthquake. Therefore, the decomposition of any GNSS time series into their contributory viscoelastic flow and afterslip can be easily obtained using our function model. Our results featured two vital observations that are consistent with several recently-developed postseismic deformation model (e.g., Agata et al. 2019, Nat. commun.; Muto et al. 2019, Sci. Adv.; Fukuda & Johnson 2021, JGR). Firstly, the dominance of viscoelastic relaxations is highlighted for inland GNSS stations in the early postseismic period (Sun et al. 2014 Nature; Agata et al. 2019 Nat. commun.; Muto et al. 2019, Sci. Adv.). Secondly, the persistence of deeper afterslip (at the downdip of the main rupture area, Muto et al. 2019, Sci. Adv.) for over a decade is inferred, manifesting the rapid uplift of the Pacific coast of NE Japan. Our proposed function model can be used not only to understand the complex interplay of various deformation mechanisms but also to facilitate geodetic inversion of these processes over the coming decades.
Logarithmic or exponential decay functions have been popular choices for predicting geodetic time series over the last few decades. In particular, combinations of logarithmic or exponential functions are demonstrated more effective in explaining the postseismic GNSS time series (Tobita 2016, EPS; Fujiwara et al. 2022, EPS). Multiple decay functions are used to proxy the Maxwell or bivious rheology (Hetland and Hager 2006, GJI) along with the aseismic slip in the postseismic period (Marone 1998, Annu Rev). The recent postseismic model illustrated an essential upgrade to power-law bi-viscous rheology which provides substantial simulation of the time-dependent postseismic deformation (Agata et al. 2019, Nat. commun.; Muto et al. 2019, Sci. Adv.; Barbot 2020, EPS). However, a closer analytical representation of such stress-dependent power-law rheology is yet to be incorporated into function models. Moreover, neither the logarithmic nor exponential decay functions can be directly correlated with the constitutive laws of rock physics, considered in most stress-dependent postseismic deformation models.
Materials and methods
We proposed alternative decay functions that incorporate power-law dependencies, learned from the laboratory experiments of rock deformations (Barbot et al. 2009, JGR; Muto et al. 2019, Sci. Adv.). We present two decay functions representing continuous afterslip on the plate boundary fault and viscoelastic relaxations in the surrounding mantle. We deploy a linear combination of these decay functions (as an analytical function model) by conducting a curve-fitting to GNSS dataset of 2 years after the mainshock (Tobita 2016, EPS). By evaluating an additional 8 years of GNSS time series, we conclude the substantial performance of our function model in the long-term prediction of most GNSS stations in NE Japan. We ensured the stability and robustness of our function model by tunning the model parameters with Bayesian optimization (Fukuda & Johnson 2021, JGR).
Results and discussion
Our decay functions mimic the contribution of the aforementioned deformation mechanisms in the surface displacements, observed after the 2011 Tohoku-oki earthquake. Therefore, the decomposition of any GNSS time series into their contributory viscoelastic flow and afterslip can be easily obtained using our function model. Our results featured two vital observations that are consistent with several recently-developed postseismic deformation model (e.g., Agata et al. 2019, Nat. commun.; Muto et al. 2019, Sci. Adv.; Fukuda & Johnson 2021, JGR). Firstly, the dominance of viscoelastic relaxations is highlighted for inland GNSS stations in the early postseismic period (Sun et al. 2014 Nature; Agata et al. 2019 Nat. commun.; Muto et al. 2019, Sci. Adv.). Secondly, the persistence of deeper afterslip (at the downdip of the main rupture area, Muto et al. 2019, Sci. Adv.) for over a decade is inferred, manifesting the rapid uplift of the Pacific coast of NE Japan. Our proposed function model can be used not only to understand the complex interplay of various deformation mechanisms but also to facilitate geodetic inversion of these processes over the coming decades.