The calculation of earthquake hypocenters requires careful treatment, particularly when prior knowledge of the study area is limited. The prior knowledge, such as wave velocity and data noise, is often assumed to be known in earthquake location algorithms. Such assumptions can greatly simplify the inverse problem but are less general than nonlinear approaches. A nonlinear treatment is of particular importance when the uncertainty quantification of locations is of interest. We present a nonlinear multiple‐earthquake location method that is applicable when little prior knowledge of the area exists. Efficient Markov chain Monte Carlo (MCMC) sampling is employed in conjunction with a hierarchical Bayesian model that treats earthquake hypocenter parameters, as well as P‐wave velocity, ratio in P‐/S‐wave velocities, and P‐ and S‐data noise standard deviations as unknown. Hypocenters for multiple earthquakes are located concurrently to provide sufficient constraints for the parameter’s P‐wave velocity, ratio in P‐/S‐wave velocity, and P‐ and S‐data noise standard deviations, which are shared among events. The algorithm is applied to simulated and field data. With field data, 47 event hypocenters are located in 1 yr of data from 10 sensors in the Canadian Rocky Mountain trench. To analyze the probabilistic solutions, we compare single‐earthquake and multiple‐earthquake locations for the 47 events and find that the multiple‐earthquake location produces better‐constrained solutions when compared with the single‐event case. In particular, depth uncertainties are significantly reduced for the multiple‐earthquake location. The algorithm is inexpensive, considering that it is based on an MCMC approach and highly objective, requiring little practitioner choice for tuning.

中文翻译：

---

加拿大落基山海沟的非线性多重地震定位和速度估算

地震震源的计算需要仔细处理，尤其是当研究区域的先验知识有限时。在地震定位算法中通常假定已知先验知识，例如波速和数据噪声。这样的假设可以大大简化反问题，但不如非线性方法普遍。当需要对位置进行不确定性量化时，非线性处理尤为重要。我们提出了一种非线性多地震定位方法，该方法在该区域的先验知识很少的情况下适用。有效的马尔可夫链蒙特卡洛（MCMC）采样与分层贝叶斯模型结合使用，该模型处理地震震源参数以及P波速度，P / S波速度比，以及P和S数据噪声标准偏差未知。同时定位多个地震的震源，以为参数的P波速度，P / S波速度比以及P和S数据噪声标准差（在事件之间共享）提供足够的约束。该算法适用于模拟和现场数据。借助现场数据，在1年的数据中，有47个事件震源位于加拿大落基山海沟的10个传感器中。为了分析概率解，我们比较了47个事件的单地震和多地震位置，发现与单事件相比，多地震位置产生了更好的约束解决方案。特别是，多震位置的深度不确定性大大降低了。该算法价格便宜，