A preliminary study of the duration of strong earthquake ground motion on the territory of former Yugoslavia, 1993

This report presents an empirical study of frequency dependent duration of strong earthquake ground motion on the territory of former Yugoslavia, using a uniformly processed database of 330 three-component accelerograms from 183 earthquakes, recorded at 60 sites in former Yugoslavia between 1975 and 1984. The duration is studied in 12 narrow frequency bands (channels). In each channel, the acceleration and corresponding velocity and displacement time histories are considered as one homogeneous data set. The duration of the strong motion part in a function f(t) of ground motion (acceleration, velocity or displacement) is defined in each frequency band as the sum of the time intervals during which the integral graph of the function f(τ)dτ between τ=0 and τ=t has the fastest growth and gains 90% of its final value. In each frequency band, the duration is modeled as a sum of two terms, the "basic duration" -- the duration of strong motion observed at rock sites, and the "site conditions" term -- prolongation due to specific geological and local soil conditions at the recording site. Two groups of empirical models are studied, one with the earthquake magnitude and the other one with the site intensity as a scaling parameter. An algorithm is presented for choosing a model depending on which earthquake and site parameters are available for a particular application. The results also includes probability distribution functions for the residuals of the models, which allows one to estimate the probability of a given duration at a site being exceeded, given the earthquake size and location (or site intensity) and the local site conditions.; The Yugoslav database is deficient in records from large magnitude earthquakes, and in records at sites on basement rock (geological classification) and on deep soil (local soil classification). Also, for many records the contributing earthquakes have not been identified, which reduced further the number of records that could be used. Consequently, independent development of regression models using only this database was not possible. Instead, models developed earlier by the authors for the Western U.S. using a more complete database were fit to the Yugoslav data.; The major trends in the regression models for the Western United States and for former Yugoslavia are the same, but there are some differences. In the Yugoslav models, the duration grows with magnitude at a higher rate for large magnitude earthquakes and at high frequencies. Also, the increase of duration with distance from the source is smaller. These differences and the pattern of the dependence of duration on intensity support the assumption of larger wave velocities and larger quality factor Q in the upper crust on the territory of former Yugoslavia than in the Western United States.

本报告针对前南斯拉夫（former Yugoslavia）境内强地震动持续时间的频率依赖性开展了一项实证研究，所用数据集为1975年至1984年间在前南斯拉夫60个台站记录的、来自183次地震的330组三分量加速度记录（accelerograms），且所有记录均经过统一处理。研究针对12个窄频带（通道，channels）展开，每个频带内的加速度时程及其对应的速度、位移时程被视为一个统一数据集。在各频带中，地震动函数f(t)（加速度、速度或位移）的强动持续时间被定义为：当积分∫₀ᵗf(τ)dτ的增长速率最快，且累计达到最终总幅值90%时的时间区间之和。在每个频带中，持续时间被建模为两项之和：一是“基准持续时间”——基岩台站观测到的强动持续时间；二是“场地条件项”——由记录台站特定地质与局部土层条件导致的持续时间延长。研究构建了两类实证模型，分别以地震震级和台站烈度作为缩放参数。同时提出了一种算法，可根据特定应用场景下可用的地震与场地参数选择适配的模型。研究结果还包含模型残差的概率分布函数，据此可在已知地震规模、震中位置（或台站烈度）以及局部场地条件的前提下，估算台站处给定持续时间被超越的概率。

前南斯拉夫数据集存在明显缺陷：缺乏大震级地震的记录，也缺少基岩（地质分类）与深厚土层（局部土层分类）场地的观测记录。此外，多数记录对应的触发地震尚未被明确识别，进一步缩减了可用有效记录的数量。因此，仅依托该数据集无法独立完成回归模型的开发。取而代之的是，将作者团队此前针对美国西部基于更完整数据集开发的回归模型，适配至前南斯拉夫的地震动数据中。

美国西部与前南斯拉夫的回归模型总体趋势保持一致，但仍存在部分差异。在前南斯拉夫的模型中，大震级地震场景及高频段下，持续时间随震级的增长速率更高；同时，持续时间随震源距增加的增幅更小。上述差异以及持续时间随烈度的变化规律，佐证了前南斯拉夫境内上地壳的波速与品质因数Q（quality factor Q）高于美国西部的学术假设。