四川和云南地区场地平均剪切波速vS20和vS30经验预测模型研究

利用四川和云南地区共973个工程场地钻孔资料，分别基于常速度外推模型、对数线性模型和条件独立模型的经验外推方法建立了该区域20 m和30 m平均剪切波速v_S20和v_S30的经验预测模型。研究表明常速度外推模型的预测误差最大，当波速资料深度小于10 m时，常速度外推方法会显著低估实际场地平均波速。基于对数线性外推方法建立了四川和云南地区波速经验预测模型，对比结果表明四川和云南地区平均波速预测结果与北京和加州地区较接近，明显低于日本地区。基于三种不同外推方法的预测误差对比分析结果表明条件独立性模型的预测结果在不同深度时误差均为最小，建议优先采用该方法建立的区域波速预测模型。 

Empirical prediction models of time-averaged shear wave velocity vS20 and vS30 in Sichuan and Yunnan areas

The time-averaged shear wave velocity of overburden soil is an important parameter for site classification and reflecting site effects on ground motion, which is widely used in earthquake ground motion prediction models. Using the lithology and wave velocity profile data of 973 boreholes in Sichuan and Yunnan, we study the regional prediction model of the average shear wave velocity. Based on the bottom constant velocity （BCV） model, log-linear model and Markov independent model, the empirical prediction models of v_S20 and v_S30 in this region were established. The results show that, the BCV method has the largest prediction error. When the depth of the shear wave velocity is less than 10 m, this method will significantly underestimate the average wave velocity of the actual site. Based on the log-linear model of Boore method, we establish an empirical prediction model. By comparison, we find that the average wave speed prediction results in Sichuan and Yunnan are close to those in Beijing and California, and significantly lower than those in Japan. Through the comparative analysis of prediction error of three different extrapolation methods, we find that the prediction results based on Markov independence model have the smallest error at different depths, and it is preferred to use this method to set up regional prediction model.