2010/11/01 Workshop on "Earthquake Forecast Systems Based on Seismicity of Japan: Toward Constructing Base-line Models of Earthquake Forecasting" Seismicity Model Modified by the Information of Seismic Velocity Structure #Fuyuki Hirose and Kenji Maeda (Meteorological Research Institute) Acknowledgements We thank J. Nakajima and A. Omuralieva for the use of the data of D90 and seismic velocity structure.
Relation between velocity structure and earthquakes Okada et al. (2010, EPS) Depth=24 km The distinct low-velocity regions are continuously distributed from the mantle wedge to the lower crust just below active seismic zones in Tohoku, Japan. This low-velocity zone can be interpreted to be a region of partial melting associated with upwelling from deeper portion.
Data Earthquakes catalog: Earthquakes in Japan with M≥6.0 and depth≤30 km are taken from Usami (1995), Utsu (1982), and JMA catalog for the period of 679-1884, 1885-1923/7, and 1923/8-2009/12, respectively.
Data Seismic velocity and D90 [Omuralieva et al. (2010)]: Estimation of 3D seismic velocity structure by using a seismic tomography method of Zhao et al. (1992). Relocation of all the shallow inland earthquakes in the period of 2000/1-2010/1 in JMA catalog by using the 3D seismic velocity structure. Estimation of spatial distribution of D90 that is defined as the depth shallower than which 90% of earthquakes in 0.3˚×0.3˚-sized tiles occur by using relocated hypocenters. Omuralieva et al. (2010)
Method We set grids with a spacing of 0.1° in latitude and longitude over Japan and count the number of earthquakes (N) with M ≥ 6.0 and depth ≤ 30 km occurring within a region with 0.3° × 0.3°-sized tiles at an each grid. We compare N obtained in step 1 with the seismic velocity (Vp and Vs) at the depths of D90, D90+5 km, +10 km, +15 km, +20 km, as well as at the fixed depth of 10, 15, 20, 25, 30, 35, 40 km. #: M≥6
Results D90 Depth=30 km Grids Incline toward low velocity Earthquakes D90+5 km D90+10 km D90+15 km D90+20 km Seismic activity is relatively high in the regions where the seismic velocity at the lower crust near the Moho is lower than the average of that depth.
Log-normal distribution Results Grids Log-normal distribution Earthquakes Earthquakes/Grids Normalization Vs=3.42 Weight=1.0W W is a normalizing factor which makes the total occurrence rate the same as that of the MGR model [Hirose and Maeda, EPS accepted]. Vs=4.10 Weight=0.2W
MGR-V model based on velocity structures Occurrence rate [/year] 5 6 7 8 9 M ×1.0W Occurrence rate [/year] 5 6 7 8 9 M MGR model Occurrence rate [/year] 5 6 7 8 9 M ×0.2W Vs=3.42 Weight=1.0W Vs=4.10 Weight=0.2W
Results (MGR model vs. MGR-V model) Forecast: 2004 MGR model MGR-V model ∆(MGR-V – MGR) (a) (b) (c) (d) (e) (f)
Results (MGR model vs. MGR-V model) Retrospective forecasts by the MGR model and MGR-V model. E[n]: expected numbers of target earthquakes in the forecast periods. N: actual numbers. : Better model The total performance of the MGR-V model is slightly better than the MGR model.
Summary We found that seismic activity is relatively high in the regions where the seismic velocity at the lower crust near the Moho is lower than the average of that depth. We introduced the velocity information into the MGR model based on the G-R law and the modified G-R law that imposes a maximum magnitude, and thus produced new earthquake forecast model (MGR-V model). The total performance of the MGR-V model is slightly better than the MGR model.