Modeling and Forecasting Aftershocks Can Be Improved by Incorporating Rupture Geometry in the ETAS Model

Abstract

We implemented an extended version of the space‐time Epidemic‐Type Aftershock Sequence (ETAS) model, which simultaneously incorporates earthquake focal depths and rupture geometries of large earthquakes, and applied it to the 2016 Kumamoto earthquake sequence. Results show that the new model corrects the estimation biases of model parameters in the point source ETAS model. The reconstructed patterns of productivity density of aftershocks, along the mainshock rupture plane, form complementary patterns for coseismic slip in space and show significant spatiotemporal migrations. Large aftershocks tend to nucleate at the edges of high productivity density areas. The decay of direct aftershocks near the mainshock rupture is consistent with static stress changes caused by the mainshock. The forecasting capability of the ETAS model can be enhanced by considering the rupture geometries of mainshocks, especially for forecasting aftershocks in the first 1–2 days. In the simulation test, the incorporation of focal depths improves the forecasting resolution.

Heterogeneity of Aftershock Productivity Along the Mainshock Ruptures and Its Advantage in Improving Short‐Term Aftershock Forecast

“…It has been reported in some literatures that large aftershocks tend to be located at the edges other than the centers of areas where aftershocks are highly clustered (van der Elst & Shaw, 2015;Stallone & Marzocchi, 2018). This finding holds true for the 2016 Kumamoto earthquake sequence in Guo et al (2019), which demonstrate that most of M5.0+ aftershocks following the mainshock nucleate at the edges of aftershock asperities formed by high productivity density areas. To investigate the distribution of strong aftershocks, we mark the epicenters of M4.5+ events in the aftershock productivity density patterns (Figures 3-6).…”

“…The most significant change is in the parameter pair α and A, which control the aftershock productivity term in Equation 2. The value of α is usually underestimated in the 3D-PS ETAS model, and A is overestimated as a trade-off effect (Guo et al, 2017(Guo et al, , 2019Hainzl et al, 2008). A smaller α value means less difference between large and small earthquakes in producing aftershocks, and thus the productivity of large earthquakes is underestimated.…”

“…For the large earthquakes with ruptures extending to tens of kilometers or more, it is necessary to incorporate the rupture geometries into the model. In the 3D‐FS ETAS model (Guo et al., 2019), the spatial kernel of mainshocks is written as where τ i ( u , v , w ) is the productivity of any location ( u , v , w ) on source S i , and the integral kernel is …”

“…Furthermore, Guo et al. (2019) proposed the 3D version of the finite source (3D‐FS) ETAS model, and validated it through the Kumamoto earthquake sequence.…”

Are Large Earthquakes Preferentially Triggered by Other Large Events?

“…The productivity exponent ( a ) defined in the first term $$K{e}^{a\left({m}_{i}-{M}_{0}\right)}$$ of Equation 2 is underestimated when the anisotropy of the spatial organization of the seismicity is not considered during calibration of the parameters of the ETAS model (Guo et al., 2019; Hainzl et al., 2008; Zhuang et al., 2019). This underestimation of a then translates into a stronger apparent dominance of the smaller earthquakes in the overall triggering process.…”