5:15 PM - 6:45 PM
[U15-P63] Rapid estimation of coseismic crustal deformation and fault model associated with the 2024 Noto Peninsula earthquake using REGARD
Keywords:GNSS, GEONET, REGARD, crustal deformation, fault model, 2024 Noto Peninsula Earthquake
In contrast to short-period seismometers, displacement data obtained by GNSS observations are not saturated even in the case of large earthquakes and can be calculated without underestimating the magnitude of the earthquake. Since 2012, GSI, in collaboration with Tohoku University, has been developing a REGARD system that provides finite fault models in real-time based on observation of a nationwide dense GNSS CORS network, GEONET. REGARD has been in operation since 2016 and has contributed to the initial response of government agencies by providing information in real-time in the case of several earthquakes, including the 2016 Kumamoto earthquake. In the 2024 Noto Peninsula earthquake, REGARD immediately offers a crustal deformation and fault model, which will be discussed in our presentation.
1) Crustal deformationWe succeeded in capturing crustal deformation during the earthquake over a wide area around Noto Peninsula, and information was automatically provided to related organizations within about 10 minutes after the earthquake. Horizontal crustal deformation of approximately 1.3 m was observed at "Wajima" and approximately 80 cm at "Suzu" near the epicenter, and a northwestward crustal deformation field of several to 20 cm was obtained in the coastal area from Niigata to Toyama Prefectures. The vertical displacement reached about 1 m at Wajima, on the northern coast of the peninsula, while it was about 30 cm at Suzu on the southern coast, showing steeper spatial gradient compared to the horizontal displacement field. These results are generally consistent with post-processed results based on GEONET routine analysis (F5 solution).
2) Fault model
A reverse fault with a northeast-southwest strike and south uplift was estimated in 4 minutes after the earthquake and this information was also automatically provided to the relevant agencies within about 10 minutes. The magnitude of the earthquake was almost Mw7.4, which was generally consistent with the GSI’s post-processed model (Mw7.44), which was approximated by three rectangular faults. The REGARD model yielded a fault model with a low dip angle of approximately 10 degrees. The post-processed model suggests a relatively high dip angle of 40-60 degrees for all three rectangular faults. This difference is thought to be mainly due to the inclusion of data from stations nearby epicenter, where real-time data was difficult to retrieve, and thus more accurate consideration of the spatial gradient in the vertical component.
3) MCMC-based fault model
GSI is planning to introduce MCMC-based fault estimation (Ohno et al., 2021) instead of the maximum likelihood estimation of 2) and is currently conducting a trial operation. The MCMC method is expected to reduce the risk of overfitting to noise by flexibly searching a more comprehensive range of parameter space compared to the current process. In the Noto Peninsula earthquake, the MCMC method successfully represented the northeastward extension of fault geometry while showing a consistent model with the results from maximum likelihood estimation. In addition, the final result was obtained about 3 minutes after the earthquake by speeding up the estimation algorithm, which suggests comparable timeliness with the maximum likelihood estimation.
1) Crustal deformationWe succeeded in capturing crustal deformation during the earthquake over a wide area around Noto Peninsula, and information was automatically provided to related organizations within about 10 minutes after the earthquake. Horizontal crustal deformation of approximately 1.3 m was observed at "Wajima" and approximately 80 cm at "Suzu" near the epicenter, and a northwestward crustal deformation field of several to 20 cm was obtained in the coastal area from Niigata to Toyama Prefectures. The vertical displacement reached about 1 m at Wajima, on the northern coast of the peninsula, while it was about 30 cm at Suzu on the southern coast, showing steeper spatial gradient compared to the horizontal displacement field. These results are generally consistent with post-processed results based on GEONET routine analysis (F5 solution).
2) Fault model
A reverse fault with a northeast-southwest strike and south uplift was estimated in 4 minutes after the earthquake and this information was also automatically provided to the relevant agencies within about 10 minutes. The magnitude of the earthquake was almost Mw7.4, which was generally consistent with the GSI’s post-processed model (Mw7.44), which was approximated by three rectangular faults. The REGARD model yielded a fault model with a low dip angle of approximately 10 degrees. The post-processed model suggests a relatively high dip angle of 40-60 degrees for all three rectangular faults. This difference is thought to be mainly due to the inclusion of data from stations nearby epicenter, where real-time data was difficult to retrieve, and thus more accurate consideration of the spatial gradient in the vertical component.
3) MCMC-based fault model
GSI is planning to introduce MCMC-based fault estimation (Ohno et al., 2021) instead of the maximum likelihood estimation of 2) and is currently conducting a trial operation. The MCMC method is expected to reduce the risk of overfitting to noise by flexibly searching a more comprehensive range of parameter space compared to the current process. In the Noto Peninsula earthquake, the MCMC method successfully represented the northeastward extension of fault geometry while showing a consistent model with the results from maximum likelihood estimation. In addition, the final result was obtained about 3 minutes after the earthquake by speeding up the estimation algorithm, which suggests comparable timeliness with the maximum likelihood estimation.