Coseismic deformation and source parameters of the 24 September 2013 Awaran, Pakistan M(w)7. 7 Earthquake derived from optical Landsat 8 satellite images
Feng Guang-Cai,Xu Bing,Shan Xin-Jian,Li Zhi-Wei,Zhang Guo-Hong
DOI: https://doi.org/10.6038/cjg20150515
2015-01-01
Abstract:On September 24th 2013, an earthquake of M(w)7. 7 occurred North-Northeast (NNE) of Awaran, Balochistan Province, Pakistan, causing extensive damages to cities and counties along the Hoshab fault. At least 825 people were killed and hundreds more were injured. However, there were quite limited SAR and geodetic data available, because the ENVISAT/ASAR satellite was out of service and ALOS/PALSAR stopped working at that time. Fortunately, this event was completely mapped by the Apr. 2013 launched Landsat8 optical satellites, which is usually for the change detection of the earth. So this earthquake provides a unique chance to study how to map the coseismic deformation of great earthquake with the Landsat 8 optical images. In this study, we utilize the Landsat 8 images to map the coseismic deformation based on the cross-spectrum correlation method, which is usually used in SAR pixel offset tracking measurement. A linear model is used to remove the long wavelength phase ramp caused by the orbital error of satellite. Our results indicate there is a curved oblique strike-slip faulting along the pervious Hoshab fault and the maximum horizontal slip is up to 12 m in the center of fault trace. In order to weigh the different datasets in modeling, we propose a geostatistic approach to assess accuracy of the North-South (NS) and East-West (EW) offsets derived from the Landsat 8 optical images. The results show their accuracy is about 0. 6 m and 0. 5 m respectively. In order to reduce the effects from topography change, atmosphere and low coherence area, we mask the measurements on the water and cloud regions. We propose a modified quadtree resampling method that not only takes account of density of samples but also gradient of deformation to overcome the limitation of conventional quadtree resampling method near and far earthquake rupture area. We keep both the largest and smallest windows as 128X128 and 8X8 pixels, respectively, to ensure sufficient point density in the non-deforming area and proper samples near the surface rupture. We locate the curved fault traces and construct the fault geometry based on the discontinuity in surface offsets. We use a two-step approach, combining the simulated annealing ( SA) optimization and subsequent derivative-based Levenberg-Marquardt inversion, to solve for the fault parameters in nonlinear modeling. It is a simplification to assume uniform slip on fault planes since heterogeneity in fault displacement and rake is observed on all scales of faults. Therefore, we construct a complex curved fault plane with dip angle estimated above to match the fault trace we mapped and divide it into 34 X 5 subfaults (7 km X 7 km) to estimate variations in slip on the fault. Based on Okada half-space elastic model, the slip distribution of this earthquake is inferred from the Landsat 8 measurements. The inversion results show that the left-lateral strike-slip dominates the earthquake fault slip concentrating at the depth less than 15 km, with a slight dipslip component and the maximum slip is Up to 10 m. The inverted seismic moment is 4. 68 X 10(20) Nm, corresponding to M7. 75. We also find that some thrust faults in the Kirthar range and strike-slip faults within the Chaman fault system have been brought closer to failure by this earthquake. This research can help the earthquake disaster evaluation in this area, and also shed some lights on the earthquake study with the Landsat 8 optical images in future.