In 2019 the KEM-04 project has successfully executed 3D seismic modelling to better understand characteristics of induced earthquakes of the Groningen gas field and the 3D propagation of seismic waves and their effects on seismic structural response in the Groningen area.
The results of KEM-04 have given suggestions to evaluate the Ground Motion Model (GMM) V5 and to possibly improve subsequent ground motion models. As ground motion model V7 is now available and the results of KEM-04 refer to ground motion model V5, the question arises whether the identified issues in V5 still apply to the latest ground motion model. The output of the project will be used to evaluate the Ground Motion Prediction Model V7 of NAM, which has been built into the public hazard and risk model for Groningen of TNO, and for other earthquake-engineering related uses, for example, to derive relevant time-histories for structural assessment or to assess the effect of larger magnitudes for which recorded data are absent or scarce, yet drive the risk figures in the Groningen area.
The output of the project will be used to validate and/or improve the existing Ground motion prediction model (GMPE) of Groningen of NAM and TNO and for deriving relevant time series for earthquake engineering related questions, that is Groningen seismic risk assessment.
The project was commissioned of Fugro, Politecnico de Milano and Seister and started in 2023 and finished in 2024.
More ...
In 2019 the KEM-04 project has successfully executed 3D seismic modelling to better understand characteristics of induced earthquakes of the Groningen gas field and the 3D propagation of seismic waves and their effects on seismic structural response in the Groningen area.
The results of KEM-04 have given suggestions to evaluate the Ground Motion Model (GMM) V5 and to possibly improve subsequent ground motion models. As ground motion model V7 is now available and the results of KEM-04 refer to ground motion model V5, the question arises whether the identified issues in V5 still apply to the latest ground motion model. The output of the project will be used to evaluate the Ground Motion Prediction Model V7 of NAM, which has been built into the public hazard and risk model for Groningen of TNO, and for other earthquake-engineering related uses, for example, to derive relevant time-histories for structural assessment or to assess the effect of larger magnitudes for which recorded data are absent or scarce, yet drive the risk figures in the Groningen area.
The output of the project will be used to validate and/or improve the existing Ground motion prediction model (GMPE) of Groningen of NAM and TNO and for deriving relevant time series for earthquake engineering related questions, that is Groningen seismic risk assessment.
The KEM 36 project provided an independent assessment of various aspect related to ground motion modelling at the state-of-the-art for Groningen, related to GMM V7, which is one of the key components of the hazard and risk assessment of the region.
The project was executed and reported in two phases.
The work carried out and reported in chapter 2 of the report finally provides a solid hypothesis to explain the underestimation of the amplitude of high-frequency (and short-distance) seismic motions in the GMMV7 model. Chapter 2 shows convincingly and didactically that the method used to compute the amplification factor (and more particularly the response spectra at base and surface levels) plays a role in the predicted level of amplification because of the unusually large high-frequency content at the reference NS_B level (very low κ0 value). This is a major result, as the influence of the choice of the computation method (and also sampling frequency) had never been discussed or evoked previously by the NAM expert group or our sub-panel.
The 3D physics-based numerical simulations are improved and reported in chapter 3. It is first shown that a simulation with rupture on an elongated fault exclusively contained within the reservoir does not reproduce the observations for a past magnitude 3.4 earthquake. This conclusion however is conditioned on assumptions imposed by a) the limited knowledge on the source parameters of such peculiar geometries (constant slip, constant local rise time consistent with usual scaling relationships) and b) the 12 Hz high-frequency limit of limitations: a much smaller local rise time would probably lead to corner frequency and stress drop values more consistent with the observations. Nevertheless, for the case of a Mw4.5 event, simulations with such elongated ruptures are compatible with GMMV7 modelling. In the second part of chapter 3, the authors show the first results of models that take plastic non-linearity into account in 3D propagation models (SPEED-NLP). Such development is a major achievement in the engineering seismology field. The results show that such models have a significant impact on maximum high-frequency vibrations (20-30% reduction) for a magnitude 4.5 earthquake. Such a decrease is not observed when conventional models (Linear Equivalent) are used in 1D modeling, which is an important result.
The period-to-period (p2p) correlation model in GMM-V7 are discussed and reported in chapter 4. The authors analyze the data from 11 earthquakes and evaluate p2p correlation by separating the within-event residuals into site-to-site and single -site component. They show that present physics-based simulations overestimate (as expected) p2p correlations among sites. The authors conclude that the procedure adopted in V7 could be improved and suggest to replace the BJ08 model by the Kotha et al. (2017) model (to use correlations models specifically developed for between-event and single site residual) and test the impact of these models on risk. Such suggestion makes sense given the fact that the correlation of different residual components may be controlled by different physical mechanisms.
Chapter 5 summarizes the main outcomes of the KEM36 investigations in relation to the main initial requests. The conclusions are sound with respect to all the KEM-36 findings and it is recommended they are considered for any further application and/or evolution of the GMMV7 model.
The project was evaluated by the KEM subpanel.
Similar to the KEM-04 project, KEM36 was a successful project delivering results useful to gather a deeper understanding the current capabilities and limitations of the seismic hazard and risk assessment for Groningen, and to highlight issues that need further attention. The quality of the product is high such that at least topics 2 and possibly 3 or 4 might be susceptible of international scientific journal publication.
The reviewers do have the following recommendations or comments for future work:
2..1 The value of the sampling frequency used by NAM should be verified by contacting the authors of the GMMV7 model.
2.2 Some details or explanations are still missing or could be improved at some places in the report, but do not affect the main conclusions.
3.1 The use of this new non-linear plastic model has a strong impact on the amplitude of high-frequency movements for a moderate earthquake (Mw=4.5). To our knowledge, such non-linearities have not (yet) been observed/detected for such moderate earthquakes. This also means that even greater effects are expected for strong earthquakes (M=5.5). A test of the method on recent earthquakes would be very important to validate these calculations.
3.2 The fact that the method leads to larger decreases in 3D compared to 1D calculations (difference between Figure 3.18 and 3.17) is an intriguing, potentially very important result. This shows the limits of 1D calculations (whatever the method) and that the main choice is not only in the method (or the constitutive law) chosen for the non-linear calculation but rather in the possibility of performing NL 3D calculations. It would also be useful to understand the origin of such larger NL effects in 3D simulations.
4.1 This analysis should be presented/discussed with TNO, in order to consider the possibility of developing a “Groningen-specific” p2pcorrelation model based on the available recordings.
5.1 Would any source / hazard / consequence / risk model need to be developed for similar induced seismicity purposes, it is recommended to take advantage of the KEM-36 and NAM experience to carefully think about some important initial options for these models:
- Hazard definition: response spectra or Fourier domain? The sensitivity (and in some cases inaccuracy) of the site amplification factor (AF) to be applied on response spectra, to the calculation method in case of high-frequency input motion, and their sensitivity to the earthquake scenario is a source of additional complexity in the hazard model. Deriving the hazard model in the Fourier domain could help to avoid some of these complexities (not all of course…)
- The issues related to the p2p correlation models are rather complex and uneasy to solve when local data are relatively rare. The consequences of p2p models are especially important when the final hazard is specified in a single scalar term (here SAavg). Once again, some of the complexities might be avoided by keeping a vectorial description of the hazard and specifying vulnerability / fragility as a function of the spectral content at the vibration period of the considered structure. This would also offer the advantage of better accounting for the actual site amplification at the considered period, instead of averaging it over a wide period range.
In conclusion, the project successfully addressed all research questions, except the consistency of period-to-period correlation between hazard and vulnerability modelling. Its results are expected to have an appreciable impact on SHRA in terms of possible revision of GMM V7, period-to-period correlation, and source location. The quality of some of its results is such that international scientific journal publications are expected. The quality of some of its results is such that international scientific journal publications are expected.