Local shear wave (S-wave) velocity structure and sedimentary cover thickness are essential parameters that control local amplification of ground motion and associated seismic hazard during earthquakes. Recently, analysis of environmental background noise from individual stations has been used to estimate horizontal-to-vertical spectral ratio (H/V) curves. In the present study, we inverted H/V curves using environmental noise immediately prior to earthquake P-wave arrival to retrieve S-wave velocity (or shear velocity, V _S) profiles at four temporary seismic monitoring stations (YDF, YDS, YDU, and YDD) near the Yedang Reservoir Dam. In the first step, we used a random search algorithm to constrain the subvolume of the parameter space (S-wave velocity structure) where the minimum of the misfit was located. In the second step, we independently applied two non-linear processes (a Monte Carlo sampling algorithm and a simulated annealing algorithm) to force the inversion towards an optimal solution, using the minimum misfit model determined in the first step as an initial estimate; we then compared the results. The feasibility and effectiveness of this two-step approach were verified by inversion of H/V curves for seismic noise recorded at four seismic monitoring stations near the Yedang Reservoir Dam. Borehole and topographical data from the four stations provided a well-constrained estimation of the local shear wave profile. Comparisons of synthetic and observed H/V curves showed that combining the two inversion algorithms efficiently overcame the extreme non-linearity of the inversion problem and provided a good resolution of S-wave structures at the Yedang Reservoir Dam. The S-wave velocity profile at the YDF station, which is situated on fresh, uniform bedrock, ranged from ∼2300 to 2700 m/s, which was consistent with the borehole data. Both the YDU station (which exhibited fundamental and first-order resonance frequency harmonics) and the YDD station (which exhibited fundamental, first-, and second-order resonance harmonics) showed significant stepwise velocity profiles for the entire depth. Thus, the entire depth can be regarded as three layers, including bedrock. At the YDS station, located near the dam spillway, the uppermost layer showed the lowest V _S values among all four stations (∼726–728 m/s), suggesting that the concrete dam body is well weathered or partially infiltrated by water. Because of these low V _S values of the concrete dam body, the Yedang Dam may require a full safety diagnosis in the near future. Analysis of the velocity profiles of all four stations near the Yedang Dam indicated that the bedrock, composed of fresh rock (∼V _S > 2700 m/s), varies greatly in depth among stations.

局部剪切波（S波）速度结构和沉积物覆盖层厚度是控制地震期间地面运动和相关地震危险的局部放大的基本参数。近来，来自各个站点的环境背景噪声分析已用于估计水平与垂直光谱之比（H / V）曲线。在本研究中，我们在地震P波到达之前立即使用环境噪声反转了H / V曲线，以获取S波速度（或剪切速度V _S）在野马水库大坝附近的四个临时地震监测站（YDF，YDS，YDU和YDD）的剖面。第一步，我们使用随机搜索算法来约束参数空间（S波速度结构）的子体积，其中失配的最小值位于其中。在第二步中，我们使用第一步中确定的最小失配模型作为初始估计，独立地应用了两个非线性过程（蒙特卡洛采样算法和模拟退火算法），将反演力推向最优解。然后，我们比较了结果。通过H / V反演验证了这种两步法的可行性和有效性。野党水库大坝附近四个地震监测站记录的地震噪声曲线。来自四个站点的钻孔和地形数据提供了对局部剪切波剖面的良好约束估计。合成H / V与观察到的H / V的比较曲线表明，两种反演算法的有效结合克服了反演问题的极端非线性问题，并为野马水库大坝的S波结构提供了良好的分辨率。位于新鲜均匀的基岩上的YDF站的S波速度剖面范围为〜2300至2700 m / s，与井眼数据一致。YDU站（表现出基波和一阶共振频率谐波）和YDD站（表现出基波，一阶和二阶共振谐波）在整个深度上都表现出明显的逐步速度分布。因此，整个深度可以视为包括基岩在内的三层。在大坝溢洪道附近的YDS站，最上层显示出最低的V _S这四个站之间的数值（约726–728 m / s）表明混凝土坝体风化良好或部分被水渗透。由于混凝土坝体的V _S值较低，因此在不久的将来，Yedang大坝可能需要进行全面的安全诊断。对叶当水坝附近所有四个站的速度分布图的分析表明，由新鲜岩石组成的基岩（〜V _S > 2700 m / s）在站之间的深度上变化很大。