Dynamic Rupture Modelling of the 1999 Düzce, Turkey Earthquake

The dynamic rupture process and near-source ground motion of the 1999 Mw 7.1 Düzce Earthquake are simulated. The fault rupture is governed by the slip-weakening friction model coupled to a three-dimensional viscoelastic wave equation. The problem is solved numerically by a 3-D dynamic rupture code that uses a generalized finite difference method. Initial parameterization of stress drop (\(\Delta \tau\)) and strength excess (\(S_{\text{e}}\)) for dynamic rupture calculations is obtained from the slip velocity distribution of a kinematic waveform inversion (KI) model by solving the elastodynamic equation with the kinematic slip as a boundary condition. Using the kinematic slip distribution and observed ground motion as constraints, a trial and error procedure was followed to define the stress parameterization. Preferred model describes the source in terms of stress with three asperities (located, respectively, at the deep, middle and shallow) and strong barriers between asperities. \(S_{\text{e}}\) is as high as 19 Mpa at barriers between the three asperities and \(\Delta \tau\) is maximum about 40 Mpa at the deepest asperity. This heterogeneity in stress distribution produces abrupt jumps in rupture velocity, exhibiting locally apparent rupture speed exceeding the P wave velocity at the borders between barriers and asperities, due to sharp changes of fault strength and stress drop at those areas. Overall, consistent with other studies, the rupture propagation is dominated by supershear speed toward the eastern asperities and at shallow surface. Simulated surface rupture at the eastern fault is consistent with other studies; nevertheless, the western shallower parts did not rupture during the simulation, suggesting that those regions may have already broken during the 1999 Kocaeli event, which occurred three months earlier. Ground motion simulation catches the major characteristics of the observed waveforms. Distribution of simulated peak ground velocity (PGV) in low frequency (0.1–0.5 Hz.) inside the study area reveals the propagation pattern on the field, with PGV reaching to 1.2 and 2.2 m/s in the NS and EW components, respectively.