APPLYING SPATIAL STATISTICAL METHODS TO PREDICT GROUND VIBRATION ACCELERATIONS CAUSED BY INDUCED SEISMICITY

This paper presents the results of the research aimed at improving the accuracy of predictions regarding the maximum values of resultant components for horizontal ground vibration accelerations in areas threatened by induced seismicity. The presented solution proposes a spatial model of the ground vibration attenuation relationship based on the assumptions of the Joyner-Boore model. When performing statistical analyses to verify the models, great emphasis was placed on the correctness of applied estimation methods to meet the assumptions. The starting point for introducing spatiality into the models was the occurrence of spatial autocorrelation of the residual component when estimating the structural parameters of a model with the least-squares method. Spatial interactions were presented using weight matrices, the construction of which was based on the inverse of the distance between units. During the study, it was found that the estimated spatial model of the ground vibration attenuation relationship showed a much better match with empirical data compared to the classical Joyner-Boore attenuation model.