Advanced seismic monitoring tools for monitoring Dutch geothermal systems

Abstract

One of the major challenges of modern society is the goal of reducing the output of anthropogenic greenhouse gases to the atmosphere. To contribute to this goal, the Netherlands is scaling up the use of geothermal energy (GE), a low-carbon technology that provides heating for infrastructure. Geothermal energy requires the extraction of geothermal fluids from a geologic reservoir to provide heat at the surface, followed by re-injection of the fluid into the subsurface. The re-injection, circulation, and extraction of the fluids affect the local stress conditions and can lead to the generation of low-magnitude seismic events. The detection, characterisation, and interpretation of these events improve the efficiency and safety of geothermal operations. This thesis aims to contribute to the Dutch government’s efforts to upscale the use of geothermal energy by enhancing and extending the existing passive seismic tools for monitoring the safety and efficiency of geothermal energy.

In Chapter 2, we introduce a seismic monitoring workflow to detect and characterise low-magnitude seismic events in noisy environments. We incorporate uncertainties from the open-access regional seismic velocity model into the hypocentre estimations. We apply the workflow to the recorded data from a temporary passive network deployed around the Kwintsheul geothermal operation in South Holland, where one low-magnitude seismic event had previously been reported. The network is located in a high-noise environment, characteristic of most geothermal operations in the Netherlands. Despite the high noise levels, we identify five additional low-magnitude seismic events that occurred close to a local fault and the injection well. These are the first events ever recorded in the region. However, large hypocentre uncertainties—due to limitations in the seismic velocity model and sparse azimuthal coverage—prevent a clear interpretation of the underlying processes. From these two limiting factors, only the velocity model can be refined after an event has been recorded. However, refining the available seismic velocity model implies significant costs, as active seismic sources are usually used.

In Chapter 3, we introduce a workflow for validating the seismic velocity model based on body-wave seismic interferometry as a cost-effective alternative. Our workflow is motivated by the possibility of retrieving virtual-offset reflection responses when seismic energy arrives with near-vertical incidence to the receivers. We apply our workflow using the low-magnitude seismic events that we detected. We find that the P-wave velocity model effectively explains the observed retrieved reflections at shallow depths. In contrast, the available S-wave models do not match the data. We conclude that the P-wave model is reliable for hypocentre studies, but that the S-wave model requires refinement.

In Chapter 4, we address how the network geometry influences the detectability and hypocentre resolution of seismic events and implement a workflow for designing seismic networks. In our workflow, we integrate open-access subsurface information to generate a synthetic earthquake catalogue using knowledge of faults and areas of expected higher seismicity risk. We then apply a non-linear design strategy and a global search algorithm to ensure approximately optimal configurations. Finally, we validate the network designs through synthetic hypocentre inversions. We identify the Dutch North Sea as the area in most need of seismic receivers, due to (i) upcoming carbon capture and storage (CCS) initiatives, (ii) the lowest existing network coverage, and (iii) the potential future use of existing oil-and-gas infrastructure for offshore geothermal-energy developments. We apply our workflow to the K-14 offshore field, where carbon capture and storage is planned. The results show that the optimised networks provide sufficient azimuthal coverage and location accuracy, even under simplified assumptions. This workflow can guide the design of cost effective networks in both onshore and offshore environments.

In Chapter 5, we focus on accurate time synchronization of seismic networks. We introduce a data-driven method to detect and correct clock errors using the time-symmetry of ambient-noise correlations. We apply our method to the IMAGE network in Reykjanes, Iceland, deployed to monitor offshore geothermal activity. Offshore geothermal-energy operations introduce additional challenges due to the need for ocean-bottom seismometers (OBS), which lack direct access to GNSS signals, leading to clock-drift errors that affect event timing and localisation. We show that most OBS in the network experienced clock drift, and some had large initial time offsets. We provide an open-source Python package (OCloC) that implements this method, enabling better timing accuracy and improved hypocentre estimation in future offshore monitoring, which can be applied in future offshore geothermal energy and the upcoming carbon capture and storage operations in the Dutch North Sea.

Together, in this thesis we introduce new or adapted workflows to tackle specific limitations in current low-magnitude seismic monitoring practices. By addressing these challenges, this thesis advances the capabilities of seismic monitoring in both onshore and offshore settings. By improving detection, location, velocity model validation, network design, and timing correction, this thesis contributes to the development of robust and cost effective seismic monitoring systems. These tools support operators and regulators in making informed decisions for the safe and sustainable scaling of geothermal energy and carbon storage in the Netherlands and beyond.