Earthquake ground motion in Belgium : contribution to a high quality database, analysis of anelastic attenuation, and evaluation of ground motion models
(2026)
Files
Mahsa Onvani-PhD thesis.pdf
Open Access - Adobe PDF
- 9.96 MB
Details
- Authors
- Supervisors
; Kris Vanneste- Abstract
- (en) This PhD research was carried out within the framework of the BELSHAKE project, a BELSPO-funded project in the Seismology & Gravimetry department of the Royal Observatory of Belgium that aims to improve the understanding of earthquake ground motions in Belgium and adjacent regions. Belgium is characterized by a low to moderate level of seismicity, yet seismic hazard assessment remains essential for infrastructure design, risk mitigation, and nuclear safety. Reliable hazard assessment requires both high quality observational data and robust ground motion models adapted to regional conditions. The BELSHAKE project was structured around two main objectives. The first was to compile a comprehensive database of earthquake ground motions for Belgium, including systematic collection and uniform processing. The second objective was to develop the capacity to model earthquake ground motions in Belgium, through the analysis of seismic attenuation, the evaluation of existing ground-motion models (GMMs), and the development or calibration of regionally appropriate models. This thesis contributes to both objectives by: • Contribution to the compilation of a high-quality ground-motion dataset, including importing all waveforms and recording stations, and applying waveforms quality check, • implementing a windowing workflow, • applying different approaches to check data consistency, • analysis of attenuation in Belgium through the high-frequency anelastic attenuation parameter kappa, • evaluation of the goodness of fit of published ground motion prediction models with respect to the BELSHAKE database. The compilation of the database required a systematic inventory of all available digital waveform data in the archive of the Royal Observatory of Belgium (ROB), supplemented by international data repositories. Approximately 344 earthquakes events with local magnitudes )𝑀𝐿) ranging from 2.0 to 5.8, focal depths shallower than 30 𝑘𝑚, and hypocentral distances (𝑅ℎ𝑦𝑝𝑜) up to 500 𝑘𝑚, were imported, and relevant source parameters and station metadata were extracted for each. Raw waveform data were downloaded into a dedicated archive, and an automated time-windowing was developed to capture noise, P, S, and coda windows. Two windowing algorithms from the literature (Goulet et al., 2014; Perron et al., 2018) were implemented and tested on the BELSHAKE data. The comparison revealed that the Goulet method provides robust definitions of signal windows (P, S and coda), while the Perron method offers more sophisticated treatment of noise windows. This thesis therefore proposes and validated a hybrid approach, combining the strengths of both methods. This innovation increased the reliability of the windowing step and decreased the number of waveforms without a noise window. A visual inspection of each record, each component, and each window ensured the removal, correction, or flagging of problems such as timing errors, low signal-to-noise ratios, clipping, or other anomalies. This information was added to the metadata and used to categorize the waveforms into good, intermediate, and bad quality. The P- and S-wave arrivals were also checked and adjusted when necessary. This was a very important step to ensure high-quality data, and only the good and intermediate quality records identified during the visual check were considered in the next processing steps. The processing was also performed on an event-by-event basis using established tools, the reliable frequency range was evaluated for each record and the intensity measures were computed from these high-quality processed records. To evaluate data quality, a series of consistency checks were applied to the intensity measures. Instrument response problems were identified and corrected where possible, and problematic records were flagged. Residual analyses were then performed against a generic GMM. These steps allowed the detection and correction of component swaps, anomalous amplitudes, and resonance issues, while also confirming the overall compatibility of BELSHAKE with external datasets such as the French RESIF database (Traversa et al., 2020). Although some minor issues remain under investigation, the resulting dataset is of high quality and suitable for robust scientific analysis. Understanding seismic attenuation in Belgium is important for adapting ground-motion models to local conditions. The high-frequency anelastic attenuation parameter 𝜅 was computed using the classic method of Anderson and Hough (1984) for about 7,800 event–station pairs using acceleration Fourier amplitude spectra of S-waves. The resulting 𝜅 values were then analyzed as a function of distance to separate site-specific (𝜅₀) and path-dependent (𝜅𝑟) components. Three estimation methods were compared: the Free kappa gradient, the Joint kappa gradient, and Mixed effects regression methods. Station-specific 𝜅0 values range from ~14 to 59 𝑚𝑠 for the H component and from ~12 to 53 𝑚𝑠 for the Z component for all stations in Belgium. The regional 𝜅𝑟 ranges from 0.13 to 0.17 𝑚𝑠/𝑘𝑚 for the H component, and from 0.05 to 0.1 𝑚𝑠/𝑘𝑚 for the Z component. Considering an average S-wave velocity of 3.6 𝑘𝑚/𝑠, the frequency-independent quality factors (𝑄) range from ~1600 to 2100 for the H component, and from 2700 to 3900 for the Z component. These results advance our understanding of anelastic attenuation in Belgium and will support the development of improved GMMs and seismic hazard models for stable continental regions. The BELSHAKE dataset was used to test the compatibility of the 20 published GMMs with the geological context of Belgium. These GMMs were developed based on European, NGA-West (Next Generation of Ground-Motion Attenuation Models for the western United States), NGA-East (Next Generation of Ground-Motion Attenuation Models for the Central and Eastern US), Japanese, global, and regional datasets. Multiple statistical measures were employed to evaluate the goodness of fit of these GMMs. The comparison revealed that some European (Akkar et al., 2014; Bindi et al., 2014), NGA-West (Abrahamson et al., 2014; Campbell and Bozorgnia, 2014), and regional models (Edwards and Douglas, 2013) performed consistently well, while others showed systematic biases or poor fit. The final ranking highlights a subset of GMMs that are best suited for seismic hazard studies in Belgium, providing guidance for future national and regional assessments. The outcomes of this thesis support improved seismic hazard assessment in Belgium by providing a quality-checked ground-motion dataset, regional and site-specific attenuation (𝜅) estimates, and a systematic comparison of published GMMs to this dataset. Together, these contributions form a coherent and transparent foundation for hazard studies in low-to-moderate seismicity regions, strengthening the scientific basis for earthquake-resistant design and risk mitigation in Belgium and beyond.
- Affiliations
Royal Observatory of Belgium Seismology & Gravimetry
Citations
Onvani, M. (2026). Earthquake ground motion in Belgium : contribution to a high quality database, analysis of anelastic attenuation, and evaluation of ground motion models. https://hdl.handle.net/2078.5/276565