Abstract: Dynamic triaxial test in laboratory is the main approach to investigate the dynamic residual strain (seismic subsidence) of natural loess under seismic loadings. How to reasonably apply the little test data in laboratory into the quantitative evaluation of dynamic settlement in loess field is one key problem within the practice of geotechnical earthquake engineering. Based on a magnitude estimation model for dynamic residual strain of natural loess, which comprehensively considers the critical influence parameters of the dynamic subsidence, consolidation stress, structure strength, spatial volume property and seismic loadings, the authors propose a probability-based evaluation method for ground settlement of natural loess under seismic loadings by means of theoretical analysis methods of probability statistics and Monte Carlo simulation. For the application of this new probability-based evaluation method, here, we provide one practical case on a typical loess field. The relevant Chi-square test show that frequency features of the four critical parameters influencing dynamic residual strain of natural loess determined by the above-mentioned theoretical model could be described by normal distribution. In order to minimize the uncertainties (scattering and randomness) of physico-mechanical property of the soil, furthermore, Monte Carlo simulation technique is adopted to extend the laboratory data into huge numbers. Then the corresponding huge numbers of dynamic residual strain of natural loess in the case field calculated by the evaluation method could be used to analyze the probability features of dynamic settlement behaviors of the loess ground. According to occurrence probabilities of different magnitude-grades of dynamic settlement, we could acquire the detailed behaviors of dynamic residual strain of natural loess in the target field under seismic loadings. The probability distribution reveals that the quantitative evaluation method reported here can obviously decrease the influence of parameter uncertainty associated with the soil. Meanwhile, those probability characteristics of dynamic settlement in the example field accord with the existing basic-behavior knowledge of dynamic residual strain of natural loess under seismic loadings. These results mean that the probability method could provide a better/reasonable reference not only to understand dynamic settlement behaviors of natural loess ground under seismic loadings, but also to select the proper treatment method for the potentially dangerous ground.

Key words: loess field, ground motion, residual strain, laboratory test, dynamic settlement, probability-based evaluation