https://doi.org/10.1140/epjp/s13360-023-04662-2
Regular Article
A novel seismic topographic effect prediction method based on neural network models
1
Shandong Earthquake Disaster Prevention Center, Shandong Earthquake Agency, 250014, Jinan, China
2
Shandong Institute of Earthquake Engineering, 250021, Jinan, China
3
Publicity and Education Center, Shandong Earthquake Agency, 250014, Jinan, China
4
Shandong Earthquake Station, Shandong Earthquake Agency, 250014, Jinan, China
Received:
31
July
2023
Accepted:
3
November
2023
Published online:
21
November
2023
Topography has a great effect on ground motions, and may aggravate seismic hazards. The quantification of topographic effects is vital for hazard prevention in practice. A series of studies were conducted to achieve this quantification. First, three bell-shaped hills’ numerical models with three input motions for each were established and calculated using finite element method. Second, topographic variables that were closely related to the amplification factors were analysed and selected. Third, back propagation (BP) neural network models using different input variables were tested to quantify the topographic effect. Core methods used in this study were validated using recorded seismic data. The results showed that altitude, slope, aspect, and frequency are important variables that can affect topographic amplification factors. Thin hills or high frequencies of incident waves generally cause great amplification factors at crests. High-frequency waves tend to be captured by small-scale topography and induce amplifications of ground motions at these positions. Altitudes can be positively correlated with amplification factors when wavelengths of incident waves are apparently longer than the dimensions of local topography. Slopes generally do not have a good correlation with amplification factors for low hills until the upper part and lower part of a hill are separated for analysis, owing to slopes at the tops and feet of low hills being very similar. The BP neural network method is adequate for the quantification of the topographic effect. There are at least two combinations of input variables of BP models that can achieve good performance. One combination includes altitude, slope, aspect, and frequency, while the other includes altitude, x-axis gradient, y-axis gradient, and frequency. The former combination is less precise but has better expandability of the predictive region, while the latter combination is more precise but lacks expandability of the predictive region.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to Società Italiana di Fisica and Springer-Verlag GmbH Germany, part of Springer Nature 2023. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
