Data-driven and machine learning identification of seismic reference stations in Europe

Self Cite

124

233

To complement the new European Strong-Motion dataset and the ongoing efforts to update the seismic hazard and risk assessment of Europe and Mediterranean regions, we propose a new regionally adaptable ground-motion model (GMM). We present here the GMM capable of predicting the 5% damped RotD50 of PGA, PGV, and SA(T = 0.01 − 8 s) from shallow crustal earthquakes of 3 ≤ M W ≤ 7.4 occurring 0 < R JB ≤ 545 km away from sites with 90 ≤ V s30 ≤ 3000 m s −1 or 0.001 ≤ slope ≤ 1 m m −1. The extended applicability derived from thousands of new recordings, however, comes with an apparent increase in the aleatory variability (σ). Firstly, anticipating contaminations and peculiarities in the dataset, we employed robust mixed-effect regressions to down weigh only, and not eliminate entirely, the influence of outliers on the GMM median and σ. Secondly, we regionalised the attenuating path and localised the earthquake sources using the most recent models, to quantify region-specific anelastic attenuation and locality-specific earthquake characteristics as random-effects, respectively. Thirdly, using the mixed-effect variancecovariance structure, the GMM can be adapted to new regions, localities, and sites with specific datasets. Consequently, the σ is curtailed to a 7% increase at T < 0.3 s, and a substantial 15% decrease at T ≥ 0.3 s, compared to the RESORCE based partially non-ergodic GMM. We provide the 46 attenuating region-, 56 earthquake localities-, and 1829 site-specific adjustments, demonstrate their usage, and present their robustness through a 10-fold cross-validation exercise.

Section: Towards Non-ergodic Ground-motion Predictionsmentioning

confidence: 99%

A regionally-adaptable ground-motion model for shallow crustal earthquakes in Europe

Kotha

Weatherill

Bindi

et al. 2020

Self Cite

124

233

“…Recent analysis suggests that rock sites located in the shallow crustal part of Europe show similar values (0.01-0.02s) (Pilz et al 2019), lower than the estimations considered by Van Houtte et al (2011). Furthermore, using a database of seismic stations predominantly from central and southern Europe, Pilz et al (2020) find that more than 40 % of sites with measured V S30 larger than 800 m/s show peaked horizontal-to-vertical spectral H/V ratios at frequencies greater than 1 Hz, implying strong near surface impedance contrasts that may arise due to weathering or zones of variable fracture density. Though their dataset did not include stations from Scandinavia and northeastern Europe, the presence of short period amplification at rock sites elsewhere in Europe suggests a stronger influence of high-impedance in rock ground motion than previously assumed.…”

Section: Discussionmentioning

confidence: 99%

A ground motion logic tree for seismic hazard analysis in the stable cratonic region of Europe: regionalisation, model selection and development of a scaled backbone approach

Weatherill

Cotton

2020

Self Cite

Regions of low seismicity present a particular challenge for probabilistic seismic hazard analysis when identifying suitable ground motion models (GMMs) and quantifying their epistemic uncertainty. The 2020 European Seismic Hazard Model adopts a scaled backbone approach to characterise this uncertainty for shallow seismicity in Europe, incorporating region-to-region source and attenuation variability based on European strong motion data. This approach, however, may not be suited to stable cratonic region of northeastern Europe (encompassing Finland, Sweden and the Baltic countries), where exploration of various global geophysical datasets reveals that its crustal properties are distinctly different from the rest of Europe, and are instead more closely represented by those of the Central and Eastern United States. Building upon the suite of models developed by the recent NGA East project, we construct a new scaled backbone ground motion model and calibrate its corresponding epistemic uncertainties. The resulting logic tree is shown to provide comparable hazard outcomes to the epistemic uncertainty modelling strategy adopted for the Eastern United States, despite the different approaches taken. Comparison with previous GMM selections for northeastern Europe, however, highlights key differences in short period accelerations resulting from new assumptions regarding the characteristics of the reference rock and its influence on site amplification.

“…Site-responses can become extremely complex, especially at the numerous Résif stations located in basins and valley. Results from this study can help in selecting sites with large 𝜙 0,𝑠 for site characterisation missions (e.g., Hollender et al 2021), to identify reference sites with small 𝜙 0,𝑠 (e.g., Pilz, Cotton and Kotha 2020, Lanzano et al 2020, Thompson et al 2012, and to improve regional site-response maps (e.g., Weatherill et al 2023, Parker andBaltay 2022)…”

Section: Random-effectsmentioning

confidence: 96%

A Bayesian update of Kotha et al. (2020) ground-motion model using Résif dataset

Kotha,

Traversa

2024

Self Cite

Recent updates of pan-European seismic hazard and risk maps adopted the partially non-ergodic Kotha et al. ( 2020) ground-motion model. This model was regressed from the Engineering Strong Motion dataset, containing ground-motion data of 𝑀 𝑊 ≥ 3 events mostly from Italy, Turkey, Greece, and in smaller fractions from rest of the active shallow crustal tectonic regions of Europe. Through mixed-effects regressions, the non-ergodic model partially resolved the spatial variability of attenuation characteristics across most of seismically active Europe, but not in France due to the then lack of a regional dataset. With the availability of a manually processed dataset from Résif network, and a computationally viable Bayesian inferencing algorithm, this study extends the non-ergodic applicability of the model to 𝑀 𝑊 < 3 earthquakes, attenuating regions, tectonic localities, and sites located in France. In process, a few important decisions had to be made concerning the updating methodology, and the interpretation of spatial variability of attenuationspecifically, that of the tectonic localities producing earthquakes. The methodology and results are discussed, emphasising the need to revise the current ground-motion regionalisation approach, and to tailor the updating procedure to be application specific. This study anticipates and supports a shift from frequentist to Bayesian approach of ground-motion modelling, in order to maintain continuity of knowledge regressed from various ground-motion datasets.