2:00 PM - 2:15 PM
[SSS07-08] Slip Modeling of Large Earthquakes by Joint Inversion of W Phase and Back Projected Images
Keywords:W-phase inversion, image deconvolution back-projection, joint inversion, finite fault inversion
We present a novel joint inversion algorithm that integrates W-phase finite fault inversion with Image Deconvolution Back-Projection (IDBP) to rapidly and accurately determine seismic source parameters, including moment tensor, slip distribution, and centroid location, following large earthquakes. This innovative approach leverages the strengths of both methods: the W-phase inversion provides reliable low-frequency information for moment tensor and centroid location, while the IDBP technique offers high-resolution imaging of rupture propagation and fault extent. By combining these two techniques, our algorithm overcomes the limitations of individual methods, enabling a comprehensive understanding of the earthquake source process with minimal parameter input.
The algorithm operates through an iterative process: (1) point-source W-phase inversion to determine the initial moment tensor and centroid location; (2) back-projection to image the rupture propagation and estimate fault extent; (3) joint inversion of W-phase and back-projected data to determine the finite slip distribution; (4) finite fault W-phase inversion to update the moment tensors of subfaults; and (5) iterative until the moment tensors and slip distribution stabilize. This method is particularly advantageous for near real-time earthquake analysis, as it requires few parameters and can be automated for rapid deployment.
We applied this algorithm to two significant earthquakes: the 2015 Mw 7.8 Nepal earthquake and the 2013 Mw 7.5 Craig Alaska earthquake. For the Nepal earthquake, our results revealed a rupture length of approximately 160 km and a source duration of around 55 seconds, with slip distribution closely matching detailed post-earthquake studies. The joint inversion successfully captured the eastward rupture migration and energy release, demonstrating the algorithm's ability to resolve complex rupture processes. Similarly, for the Craig Alaska earthquake, the algorithm identified a strike-slip mechanism with a rupture length of 100–150 km and a source duration of ~30 seconds. While the results for the Craig earthquake showed some discrepancies in depth resolution compared to reference studies, the horizontal slip distribution and energy release patterns were well-constrained.
One of the key advantages of our algorithm is its timeliness. By utilizing dense seismic networks for back-projection and global networks for W-phase inversion, the joint inversion results can be obtained within approximately 30 minutes after a large earthquake. This rapid response capability makes the algorithm highly suitable for real-time earthquake analysis, disaster response, and tsunami hazard assessment.
However, challenges remain, the algorithm assumes that the high-frequency and low-frequency sources share a similar radiation pattern, which may not hold true for all earthquakes. To address this, we introduced a Frequency Dependence Index (FD Index) to quantify the degree of frequency-dependent energy radiation. Our synthetic tests showed that the FD Index effectively captures the separation between high-frequency and low-frequency sources, providing a measure of the algorithm's accuracy in cases of significant frequency-dependent rupture complexity.
In conclusion, our joint inversion algorithm represents a significant advancement in the rapid determination of earthquake slip models. Its successful application to the Nepal and Craig Alaska earthquakes highlights its potential for real-time earthquake source characterization, offering valuable insights for earthquake emergency response and research.
The algorithm operates through an iterative process: (1) point-source W-phase inversion to determine the initial moment tensor and centroid location; (2) back-projection to image the rupture propagation and estimate fault extent; (3) joint inversion of W-phase and back-projected data to determine the finite slip distribution; (4) finite fault W-phase inversion to update the moment tensors of subfaults; and (5) iterative until the moment tensors and slip distribution stabilize. This method is particularly advantageous for near real-time earthquake analysis, as it requires few parameters and can be automated for rapid deployment.
We applied this algorithm to two significant earthquakes: the 2015 Mw 7.8 Nepal earthquake and the 2013 Mw 7.5 Craig Alaska earthquake. For the Nepal earthquake, our results revealed a rupture length of approximately 160 km and a source duration of around 55 seconds, with slip distribution closely matching detailed post-earthquake studies. The joint inversion successfully captured the eastward rupture migration and energy release, demonstrating the algorithm's ability to resolve complex rupture processes. Similarly, for the Craig Alaska earthquake, the algorithm identified a strike-slip mechanism with a rupture length of 100–150 km and a source duration of ~30 seconds. While the results for the Craig earthquake showed some discrepancies in depth resolution compared to reference studies, the horizontal slip distribution and energy release patterns were well-constrained.
One of the key advantages of our algorithm is its timeliness. By utilizing dense seismic networks for back-projection and global networks for W-phase inversion, the joint inversion results can be obtained within approximately 30 minutes after a large earthquake. This rapid response capability makes the algorithm highly suitable for real-time earthquake analysis, disaster response, and tsunami hazard assessment.
However, challenges remain, the algorithm assumes that the high-frequency and low-frequency sources share a similar radiation pattern, which may not hold true for all earthquakes. To address this, we introduced a Frequency Dependence Index (FD Index) to quantify the degree of frequency-dependent energy radiation. Our synthetic tests showed that the FD Index effectively captures the separation between high-frequency and low-frequency sources, providing a measure of the algorithm's accuracy in cases of significant frequency-dependent rupture complexity.
In conclusion, our joint inversion algorithm represents a significant advancement in the rapid determination of earthquake slip models. Its successful application to the Nepal and Craig Alaska earthquakes highlights its potential for real-time earthquake source characterization, offering valuable insights for earthquake emergency response and research.