zum Inhalt
zur Hauptnavigation

July 2025

Karlsruhe Institute of Technology

In June 2025 the Karlsruhe Institute of Technology (KIT), Geophysical Institute, finished its last measurement campaign in the southern Black Forest close to Freiburg, a combined effort with the State Earthquake Service of Baden-Württemberg. Additionally, to line measurements, a ring measurement was conducted to analyze the source signal of a wind turbine in more detail. The location of the measurement will also allow us to study the effects of strong topography.
Furthermore, the data analysis for the measurement campaign close to the town Havixbeck was finished. We identified several rotation rate dependent frequency peaks, determined corresponding b-values as well as the quality factor Q.
In total the KIT group did seven line measurements in the last three years and starts now with a detailed comparison between these measurement campaigns and with the results of the other DB MISS groups as well as results reported in the literature. Publications will be prepared.

Ruhr Universität Bochum

The DenoisingAutoEncoder is to be made more widely available by integrating it into the SeisBench software package (Slide 2). The performance of the new application is tested using the KWISS dataset (Limberger et al. 2021). This dataset consists of continuous waveform data recorded over up to two years along two profiles (19 + 10 stations) near a wind farm (Slide 3). In addition to evaluating the performance of the DenoisingAutoEncoder, these data are used to test up to what distances from the wind turbines, and under varying wind conditions, the noise can be effectively suppressed so that phase identification becomes possible. When training the new version of the DenoisingAutoEncoder in SeisBench, slight performance deficits compared to the original version can be observed (Slide 4). A comparison of denoised waveforms from two events with a magnitude of 0.7 originating from the same source region (distance ~93 km), recorded under different wind conditions at different stations along the profile (Slides 6–8), shows better noise suppression with the original DAE. Using up to 2,200 seismic events (source distance < 3°), their interpretability before and after applying the DAE at the stations along the profile is systematically analyzed as a function of distance, magnitude, and wind speed.

Geological Service NRW

The velocity model for North Rhine-Westphalia (NRW), derived from sonic logs, was further revised and modeled in GoCAD for selected locations. Preliminary Qp and Qs factors were determined from the Vp and Vs velocities in order to estimate the damping value Qr for the propagation of Rayleigh waves using Vr = 0.92*Vs. A decay function for the far field was defined, which describes the amplitude decrease of the Rayleigh waves as a function of Vr and Qr. This decay function was initially applied to data measured by KIT at the Tönisvorst site and checked for plausibility. In the future, the decay function is also to be applied to the other KIT sites and evaluated for its suitability in predicting the influence of wind energy plants (WEA) on seismic stations.

University Münster

This presentation highlights the latest results from newly added data processing efforts related to the 1720 quarry blast dataset. The data, provided by the University of Cologne and processed at the BNS Observatory, has been analyzed within the framework of the dbMISS project to study seismic attenuation parameters in the region.
Analysis of the quarry blast and local earthquake datasets revealed consistent S-phase travel times and similar epicentral distributions, emphasizing the reliability and compatibility of these independent sources. Moreover, the average frequency-dependent coda quality factor (Qc) calculated from both datasets shows a strong agreement, further supporting the robustness of the applied methodology.
Including of the quarry blast data has enhanced the depth resolution of the tomographic images, particularly at shallower depths, improving resolution down to approximately 1300 meters.
Preliminary processing of data from the temporary broadband network installed by the University of Münster indicates that this deployment significantly improves ray path coverage in the northern part of the study area. Initial examples of observed seismicity in this region were also presented, highlighting the value of the temporary network.
In the final stage of the project, data processing will focus on incorporating the most recent datasets and delivering an updated, refined version of the tomographic and attenuation models. These enhancements are expected to yield deeper insights into the seismic structure and attenuation characteristics of the study region.

March 2025

Karlsruhe Institute of Technology

In the last three months, the focus at the Karlsruhe Institute of Technology (KIT), Geophysical Institute, was the planning and realization of further line measurements starting at wind turbines, as well as the analysis of the data. Close to the town Havixbeck, a line measurement was conducted and currently the data is analyzed, also regarding the differences in noise level before and after wind turbine construction. For April 2025 the last line measurement campaign within the DB MISS project is planned in cooperation with the state earthquake service of Baden-Württemberg.

Geological Service NRW

As the basis for the lithological model of North Rhine-Westphalia (NRW), data from the Geothermal Portal NRW (https://www.geothermie.nrw.de/) for shallow and medium-depth geothermal energy were used. The point data were intersected with the digital terrain model of the BKG with a grid spacing of 200 and 1000 m (https://gdz.bkg.bund.de/index.php/default/digitale-geodaten/digitale-gelandemodelle.html?___store=default). Thus, virtual borehole data for all of NRW with depths of 100 and 1000 m are available at grid intervals of 200 or 1000 m, respectively. For further use, the complex lithology in the borehole data was homogenized. Approximately 175 geophysical borehole measurements (sonic logs) were used to assign interval velocities of the compressional wave (Vp) to the lithological units in the vicinity of the borehole measurements. The interval velocities were transferred to layer equivalents across all of NRW. In a simplified approach, shear wave velocities (Vs) were determined using the relationship Vs = Vp/√3. Missing velocity information for lithologies was appropriately supplemented with literature values. Preliminary Qp and Qs factors were estimated from the Vp and Vs velocities.

University Münster

We continue analyzing the available earthquake database using the attenuation tomography tool, MuRAT. Our tests include all seismic phases, incorporating both P- and S-phase arrival time data. We compared results obtained using all three components (3C) and only the horizontal components, finding consistent agreement in the calculated parameters—Qc, Q, and peak delay variation—as well as in the average frequency-dependent Q value for the study region. Additionally, we processed data from the dense local FloodRisk Seismic Network, operated by Ruhr University Bochum (Rische et al., 2022). We are currently focused on data preprocessing and selection for this network to improve the quality of our results. These efforts are part of our preparation for publishing our findings.

November 2024

Universität Münster

Karlsruhe Institute of Technology

The focus at the Karlsruhe Institute of Technology (KIT), Geophysical Institute, was the analysis of the data and the realization of further line measurements starting at wind turbines. In total 5 measurement campaigns were successfully completed. One measurement is currently deployed, and another one is planned for spring 2025. The results show similarities to previous studies and demonstrate the complexity of the observed wavefield. As a final step, we started the estimation of the quality factor Q from the line measurements.

Geological Service NRW

As part of work package 4, the geological context of Q is currently being investigated, with the main goal being the analysis, evaluation, and characterization of earthquake monitoring stations in North Rhine-Westphalia (NRW).

The focus of the evaluation and characterization is on geological and geophysical parameters, particularly lithology and seismic velocity of the subsurface. These parameters are crucial for accurately capturing the site conditions of the monitoring stations. The analysis of lithology allows for a detailed identification of rock types and their specific properties. The investigation of seismic velocity provides important information about the propagation speed of earthquake waves.

Input data used include a 3D model of the state of North Rhine-Westphalia, 3D fault models for NRW, a DEM file, and geodata from the Geothermal Portal NRW (https://www.geothermie.nrw.de) for shallow and mid-depth geothermal energy. These data were processed using GIS software and Python, and the relevant information was transferred into the Move 3D software. Based on this data, lithological models were created. In the next step, seismic velocity values from borehole geophysical measurements will be added to provide a comprehensive representation.

Ruhr Universität Bochum

As part of work package 7, the denoiser was further tested to support phase identification of noisy data in practice. The impact of amplitude attenuation due to the application of the denoiser on magnitude determination was investigated using swarm earthquake datasets. This allowed for the determination of amplitudes for up to 1900 events with magnitudes ranging from -0.7 to 4.2 ML at various stations and different epicentral distances. Amplitudes for noise just before an event, amplitudes for P- and S-phases, as well as the signal-to-noise ratio (SNR) for the different phases, were determined. All parameters were calculated for all components of the original waveforms, filtered traces (bandpass 1 - 45 Hz, 3rd order), and the denoised traces created by the denoiser.

Different parameter combinations were tested to identify correlations in the amplitude attenuation of the denoised traces. By comparing the signal-to-noise ratio between the original and denoised data, distance-dependent thresholds were determined at which amplitude attenuation due to denoising can be neglected. However, on these traces, phase arrivals could still be identified well in the original data, making processing (filtering/denoising) unnecessary. No regularity could be established for the amplitude attenuation on the denoised traces at magnitudes below the threshold, which could be used to correct amplitudes for magnitude determination on the denoised traces.

Therefore, magnitude determination should always be performed using the original data. Excessive noise in the original data can negatively impact magnitude determination. The traces produced by the denoiser are suitable for identifying phase arrivals due to their good reproduction of the waveform but are not suitable for magnitude determination.

DMT

May 2024

Universität Münster

Karlsruhe Institute of Technology

Geological Service NRW

DMT

Januar 2024 (Talks from AG Seismologie and Project meeting 2023)

Universität Münster

At Münster, we search for more data to be used with the two programs that calculate attenuation, MURAT and QSEIS. The data availablility is small, making the region that can be resolved much smaller than anticipated.

Geologischer Dienst NRW

Ruhr Universität Bochum

Presentation (AG Seismologie)

DMT

Karlsruhe Institute of Technology

Presentation (AG Seismologie)

März 2023

Westfälische Wilhelms-Universität Münster

The work done by the WWU in the period 01-10-2022 - 31-12-2022 focused on collecting the available data from the parties involved, e.g. the Geologischer Dienst (GD), DMT, and open-access data retrievable trough available FDSN clients. Data received form the GD is a good addition to the open-access data and can be processed in the same way. Furthermore, some events occur in both the open-access as well as the GD set, increasing the coverage. The data presented by DMT stands out with respect to the other data, in the sense that it contains seismic linesurveys rather than waveforms from seismic stations. As such, it contains many receivers, but at relatively small distances to the sources, which in turn tend to be rather small as they are all induced explosions. Consequently, the DMT data will have to be processed separately, and the initial focus will be placed on retrieving Q-values (attenuation) on a larger scale using the GD and open-access waveforms. The total attenuation is the sum of scattering attenuation on the one hand, and intrinsic attenuation on the other. The method we propose to determine the attenuation, separating scattering and intrinsic, is Qopen (Eulenfeld and Wegler, 2016). Qopen is based on envelop inversion of the S-coda. This method has proven usefull for weak events (Eulenfeld et al., 2022), and has been applied in frequency ranges of 1 Hz and higher (Neuffer et al., 2021), comparable to the second to forth eigenfrequencies of windturbines (Eulenfeld and Wegler, 2016).

In addition to the preparation of the data, a rough start has been made on the numerical implementation of the viscoelastic wave equation in general, using the code of Martin and Komatitsch (2009), and adding the free surface formulation as proposed by (Zeng et al., 2012). For now however, determination of the attenuation has been prioritized over the numerical implementation at this point. Thus, further numerical implementation will be carried out when the attenuation determination is further along.

Following these updates, the project is on schedule concerning the initial work and time planning, and no changes with regards to the targets of the DBMISS project, nor the expected planning in meeting these targets have been made.

Geologischer Dienst NRW

Beim Geologischen Dienst (GD) soll der im Vorgängerprojekt entwickelte Filter zur Trennung von Erdbebensignal und Störsignal in der Praxis erprobt und in Zusammenarbeit mit der Ruhr-Universität Bochum weiterentwickelt werden. Dazu ist es erforderlich, die Filter in den Routineauswerteprozess einzubinden. Zu dem Zweck wird zurzeit das IT-System im GD umgestellt.
Der Filter wurde nach einer ersten Erprobung mit dem Projektpartner RUB auf einer neuen Linux-Maschine dauerhaft durch den GD installiert und eingerichtet. Erste Tests mit Archiv-Daten unter Verwendung des allgemeinen Noise-Modells wurden durchgeführt. Ein bekanntes Event wurde mit den gefilterten Daten ausgewertet. Die Lokalisierung lieferte gute Ergebnisse, allerdings ist die Magnitudenbestimmung bisher noch sehr ungenau. Hier muss der Filter in Zukunft noch optimiert werden.
Der Zugriff auf Echtzeit-Daten ist derzeit noch nicht möglich. Zu diesem Zweck wird die vorhandene IT-Infrastruktur optimiert. Hier sind Fragen der IT-Sicherheit zu klären und die Vorgaben für das Landesnetz NRW zu berücksichtigen.
Für das Training des Filters speziell auf die Stationen des GD werden derzeit Noise-Fenster geschnitten. Kurzfristig erfolgt das Training des Filters durch die RUB. Mittelfristig wird auch das Training durch den GD selbst erfolgen.

Karlsruher Institut für Technologie (KIT)

The main focus in the first project phase at the Karlsruher Institute of Technology (KIT) is the identification of suitable measurement sites to observe the ground motion emissions of wind turbines. Furthermore, we focused on agreements with wind turbine operators regarding data exchange. The first measurement campaign took place in North Rhine-Westphalia near Tönisvorst. Along a profile of 5.5 km length, beginning at two 4 MW wind turbines, 11 seismic recording stations were installed for 4 weeks of continuous recording on loose Tertiary sediments. A preliminary analysis of the data was started. To analyze the data we recoded existing MATLAB scripts to Python. Those software programs ensure automatic reuse for further measurement campaigns.

Ruhr Universität Bochum

Der an der Ruhr-Universität Bochum entwickelte Filter wurde sowohl beim Geologischen Dienst NRW als auch bei der Firma DMT in Betrieb genommen. Durch die Praxiserprobung durch weitere Personen konnten schon etliche kleine Fehler behoben werden und auch die Anwendung ist intuitiver geworden. Noch treten Probleme im Bereich der Coda-Wellen auf, da diese z.T. im Trainingsprozess des Filters vernachlässigt worden sind. Hierzu wird das Training zeitnah überarbeitet. Des Weiteren wird derzeit ein Code entwickelt, der mit Hilfe maschinellen Lernens automatisch Erdbeben erkennt und diese auswertet. In Kombination mit dem Filter kann eine Datenbank erstellt werden, um herauszufinden wie viele neue Erdbeben nach Anwendung des Filters detektiert werden können und welchen genauen Einfluss dies auf die Magnitudenbestimmung hat. Dabei werden die vorhandenen Algorithmen zur automatischen Erdbeben Detektion Phasenassoziation in der Praxis erprobt und ebenfalls verbessert.

Der Quellcode ist unter https://github.com/JanisHe frei zugänglich.