Forecasting the occurrence of natural hazards, such as earthquakes or landslides, remain very challenging. These hazards are often caused by stress changes in the subsurface, therefore detecting and monitoring these changes can help the prediction and mitigation. Active ultrasonic transmission experiments were performed on Red Pfaelzer sandstones to investigate the monitoring and forecasting potential of these measurements. The sandstone samples were loaded until failure at different initial confining stress conditions. The forecasting potential to failure of different analysis methods, such as coda wave interferometry or wave attenuation, is investigated and compared. Our results show we can detect the forecast the upcoming failure of the samples from 40 to 70% of its failure point. Small differences between each analysis method are visible, but the trend of the signal is leading and therefore a robust prediction of failure can be made by combining analysis methods. In this paper, we propose a traffic light forecasting system using the precursory signals from ultrasonic monitoring. This system is applicable for monitoring failure at various depths and or stress conditions, for a better prediction of small stress-induced changes in the subsurface and thus mitigation of failure (natural hazards) in the subsurface.

@en

Induced earthquakes are still highly unpredictable, and often caused by variations in pore fluid pressure. Monitoring and understanding the mechanisms of fluid-induced fault slip is essential for seismic risk mitigation and seismicity forecasting. Fluid-induced slip experiments were performed on critically stressed faulted sandstone samples, and the evolution of the actively sent ultrasonic waves throughout the experiment was measured. Two different fault types were used: smooth saw-cut fault samples at a 35° angle, and a rough fault created by in situ faulting of the samples. Variations in the seismic slip velocity and friction along the fault plane were identified by the coda of the ultrasonic waves. Additionally, ultrasonic amplitudes show precursory signals to laboratory fault reactivation. Our results show that small and local variations in stress before fault failure can be inferred using coda wave interferometry for time-lapse monitoring, as coda waves are more sensitive to small perturbations in a medium than direct waves. Hence, these signals can be used as precursors to laboratory fault slip and to give insight into reactivation mechanisms. Our results show that time-lapse monitoring of coda waves can be used to monitor local stress changes associated with fault reactivation in this laboratory setting of fluid-induced fault reactivation. This is a critical first step toward a method for continuous monitoring of natural fault zones, contributing to seismic risk mitigation of induced and natural earthquakes.

@en

Addressing climate change and transitioning to renewable energy will involve subsurface activities like carbon storage, geothermal exploitation, and underground energy storage. However, fluid injection and extraction in the subsurface can alter the pressure, temperature, stress, and rock geochemistry, potentially leading to seismicity and subsidence. Understanding the mechanisms of fault reactivation and the geomechanical response of intact reservoir rock to variations in pore fluid pressure from injection and depletion operations is thus crucial. In this thesis, we integrate our findings from multiple studies to explore the impact of parameters related to injection and depletion, such as pattern (monotonic, cyclic) and rate, on the deformation of intact reservoir rock, slip behaviour in faulted reservoir rock, and the evolution of microseismicity, with a focus on how we can mitigate induced seismicity.

Our experimental investigations employ uniaxial compressive tests on intact Red Felser sandstone samples, subjecting them to cyclic recursive (CR), cyclic progressive (CP), and monotonic stress patterns at varying stress rates. The recording of Acoustic Emission (AE) waveforms revealed that cyclic stress patterns, especially CP, are characterized by lower maximum AE amplitudes compared to the monotonic pattern. By reducing the stress rate, the maximum AE energy and final mechanical strength both decrease significantly. Moreover, high-stress rates were found to alter the AE signature of events, suggesting that cyclic stress patterns combined with low-stress rates may mitigate induced seismicity in subsurface injection operations.

For underground energy storage, we investigate the geomechanical response of Red Felser sandstone to cyclic loading, crucial for safe and efficient underground porous reservoir operations. Experimental results, complemented by constitutive modeling, revealed various deformation mechanisms, including linear elastic, viscoelastic, and inelastic responses. Our study shows that the magnitude of inelastic deformations is influenced by mean stress, amplitude, and frequency of the stress waveform, with our models closely fitting the experimental data.

As part of our investigation into mitigating induced seismicity, we examine how stress and sliding patterns affect fault slip behaviour and seismicity evolution. To achieve this we carry out displacement-driven fault reactivation experiments on saw-cut Red Felser sandstones. Our results indicated that cyclic sliding, compared to continuous sliding, reduces seismicity but can accelerate slip velocity during the reloading phase due to the healing of gouge material on the fault plane. Additionally, under-threshold cycling effectively prevents seismicity and shear slip but poses a risk of increased seismicity if shear stress exceeds critical levels.

Furthermore, we explore the influence of injection pattern and rate on fault reactivation in porous Red Felser sandstone. High injection rates were linked to increased slip velocity and seismicity. Furthermore, our results from samples subjected to various injection patterns demonstrate that the cyclic recursive pattern exhibits a higher maximum slip velocity, more episodes of slow slip, and greater radiated AE energy than a monotonic pattern. A proper injection strategy must consider fault drainage, critical shear stress, injection rate, and injection pattern. Our results demonstrate that a monotonic injection pattern and low pressurization rate may mitigate seismicity on pre-existing faults in a highly permeable porous reservoir.

Finally, we investigate the fault slip nucleation within a displaced fault system. Our triaxial experiments on displaced faults reveal that differential compaction intensifies from the top of the sample towards the internal corner at the centre of the fault, indicating a variation in the stress field surrounding the fault plane. Our direct measurements near the displaced fault plane confirm the anomalies and peaks in stress observed in previous numerical and analytical studies.

This thesis offers new insights into the mechanical behaviour and seismicity evolution of intact and faulted reservoir rocks under variations in stress patterns and rates. These findings may contribute to mitigating injection-induced seismicity in intact and porous faulted rock settings. Furthermore, they enhance our understanding of the behaviour of deep geo-reservoirs subjected to diverse injection strategies, thereby expanding our knowledge of reservoir-related phenomena.@en

Seismic interferometry (SI) retrieves new seismic responses, for example reflections, between either receivers or sources. When SI is applied to a reflection survey with active sources and receivers at the surface, non-physical (ghost) reflections are retrieved as well. Ghost reflections, retrieved from the correlation of two primary reflections or multiples from two different depth levels, are only sensitive to the properties in the layer that cause them to appear in the result of SI, such as velocity, density and thickness. We aim to use these ghost reflections for monitoring subsurface changes, to address challenges associated with detecting and isolating changes within the target layer in monitoring. We focus on the feasibility of monitoring pore-pressure changes in the Groningen gas field in the Netherlands using ghost reflections. To achieve this, we utilize numerical modelling to simulate scalar reflection data, deploying sources and receivers at the surface. To build up subsurface models for monitoring purposes, we perform an ultrasonic transmission laboratory experiment to measure S-wave velocities at different pore pressures. Applying SI by autocorrelation to the modelled data sets, we retrieve zero-offset ghost reflections. Using a correlation operator, we determine time differences between a baseline survey and monitoring surveys. To enhance the ability to detect small changes, we propose subsampling the ghost reflections before the correlation operator and using only virtual sources with a complete illumination of receivers. We demonstrate that the retrieved time differences between the ghost reflections exhibit variations corresponding to velocity changes inside the reservoir. This highlights the potential of ghost reflections as valuable indicators for monitoring even small changes. We also investigate the effect of the sources and receivers’ geometry and spacing and the number of virtual sources and receivers in retrieving ghost reflections with high interpretability resolution.

@en

Robust and reliable prediction of (induced) earthquakes remains a challenging task. Seismicity predictions are made using probabilistic models, precursors such as average earthquake size distribution. Pore pressure variations cause stress perturbations along pre-existing fault planes in the subsurface, resulting in shear slip and seismicity. Monitoring these stress changes before fault reactivation and its resulting seismicity could greatly improve forecasting seismicity. Stress changes can be determined by changes in acoustic or seismic velocities. Therefore, experiments are performed to detect the preparatory phase of an earthquake using acoustic monitoring. Faulted sandstone samples are reactivated in the laboratory by imposing pore pressure changes by fluid injection under reservoir pressures, while continuously performing passive and active (transmission) acoustics measurements. Using coda wave interferometry (CWI) and decorrelation (K), changes in velocity and scattering are obtained before and during fault reactivation. We show that fault reactivation can be identified by a large velocity drop and an increase in K or by micro-seismic foreshocks. We show that CWI velocity change is most sensitive to both the preparatory phase and the fault reactivation. These results show acoustic monitoring of fault reactivation in the laboratory is feasible, which could improve the prediction of induced seismicity.@en

Considering the storage capacity and already existing infrastructures, underground porous reservoirs are highly suitable to store green energy, for example, in the form of green gases such as hydrogen and compressed air. Depending on the energy demand and supply, the energy-rich fluids are injected and produced, which induces cyclic change of state-of-the-stress in the reservoir and its surrounding. Detailed analyses of the geo-mechanical deformations under variable storage conditions i.e., storage frequency and fluid fluctuating pressures, are crucially important for safe and efficient operations. The present work presents an integrated analysis, based on experimental and constitutive modeling aspects, to investigate sandstones’ geomechanical response to cyclic loading relevant to underground energy storage (UES). To this end, sandstone rock samples were subjected to cyclic loading above and below the onset of dilatant cracking under different frequencies and loading amplitudes. Axial strains and Acoustic Emissions (AE) were measured in both regimes to quantify the total deformation (strain) of the rock and its AE characteristics. It is found that the inelastic strain and number of AE events is the highest in the first cycle and reduce subsequently cycle after cycle. Moreover, cyclic inelastic deformations are affected by the mean stress, amplitude, and frequency of the stress waveform. On the one hand, the higher the mean stress and the amplitude, the higher the total inelastic strains. On the other hand, the lower the frequency, the higher the total inelastic strain. From the modeling perspectives, five types of deformation mechanisms were identified based on the governing physics: elastic, viscoelastic, compaction-based cyclic inelastic, inelastic brittle creep, and dilatation-based inelastic deformation. To model elastic, viscoelastic, and brittle creep, the Nishihara model was used. A cyclic modified cam clay model (MCC) and hardening–softening model were applied to capture plastic deformation. The results show a very good fit of the constitutive model with the experimental results, which could help in studying the response of reservoirs to injection and production.

@en

Seismic interferometry (SI) retrieves new seismic responses between receivers or sources using, e.g., cross-correlation. Applying SI to a reflection survey with active sources and receivers at the surface, one retrieves ghost reflections besides the physical reflections. Ghost reflections are retrieved from the correlation of two primary reflections or multiples from two different depth levels. They are only sensitive to the changes in the layer that cause them to appear in the result of SI. Using ghost reflections from SI, we investigate the possibility of monitoring pore-pressure depletion due to gas extraction in the Groningen gas field, Netherlands. We performed an active-source transmission laboratory experiment to measure S-wave velocities at pore pressures of 50, 80, 100, 200, and 300 bar. Using these values; we numerically model scalar reflection data with sources and receivers at the surface for the Groningen subsurface model. Applying SI by auto-correlation to these datasets, we retrieve zero-offset ghost reflections. We show that using only the reflections from the top and the bottom of the reservoir is essential for retrieving a specific ghost reflection from inside the reservoir. The retrieved ghost reflections showed clear time differences, indicating they can be utilized to monitor reservoir pore-pressure depletion changes.@en

Recent laboratory and field studies suggest that temporal variations in injection patterns (e.g., cyclic injection) might trigger less seismicity than constant monotonic injection. This study presents results from uniaxial compressive experiments performed on Red Felser sandstone samples providing new information on the effect of stress pattern and rate on seismicity evolution. Red Felser sandstone samples were subjected to three stress patterns: cyclic recursive, cyclic progressive (CP), and monotonic stress. Three different stress rates (displacement controlled) were also applied: low, medium, and high rates of 10⁻⁴ mm/s, 5 × 10⁻⁴ mm/s, and 5 × 10⁻³ mm/s, respectively. Acoustic emission (AE) waveforms were recorded throughout the experiments using 11 AE transducers placed around the sample. Microseismicity analysis shows that (i) Cyclic stress patterns and especially cyclic progressive ones are characterized by a high number of AE events and lower maximum AE amplitude, (ii) among the three different stress patterns, the largest b-value (slope of the log frequency-magnitude distribution) resulted from the cyclic progressive (CP) stress pattern, (iii) by reducing the stress rate, the maximum AE energy and final mechanical strength both decrease significantly. In addition, stress rate remarkably affects the detailed AE signature of the events classified by the distribution of events in the average frequency (AF)—rise angle (RA) space. High stress rates increase the number of events with low AF and high RA signatures. Considering all elements of the AE analysis, it can be concluded that applying cyclic stress patterns in combination with low-stress rates may potentially lead to a more favourable induced seismicity effect in subsurface-related injection operations.

@en

Over the last few years, several experimental and numerical studies have investigated the mitigation and managing of fluid injection-induced seismicity. A cyclic fluid injection has been suggested to have a different seismic response than a monotonic injection, and a cyclic injection may cause less seismicity. However, most studies have been allocated for intact rock medium (not faulted). In this study, faulted (saw-cut) Red Pfaelzer sandstones were subjected to fault reactivation experiments to investigate the effect of stress cycling on seismicity evolution. During the stress-driven fault reactivation experiments, three different reactivation scenarios were carried out: continuous sliding, cyclic sliding, and under-threshold cycling sliding. The results showed that in comparison to continuous sliding, cyclic sliding triggers less seismicity in terms of b-value and significant AE events due to the uniform reduction in roughness and asperities on the fault plane. In addition, increasing the number of cycles decreases the number of AE events. The under-threshold cycling strategy prevents seismicity and pure shear slip; however, if the stress exceeds the previous maximum stress (critical), seismicity risk increases drastically in terms of b-value, maximum AE energy, and magnitude.

@en

Increased seismicity due to subsurface activities has led to increased interest in monitoring and seismic risk mitigation. In this study we combined passive and active acoustic monitoring methods to monitor fault sliding and reactivation in the laboratory. Acoustic emission (AE) and ultrasonic transmission measurements were performed during stress-cycling to monitor stress-driven fault reactivation. We show the use of the transmissivity and coda wave interferometry of the active acoustic measurements and the number of generated AE events for fault reactivation monitoring. Combining these two methods, we are able to detect the different phases of fault reactivation process under stress cycling including, early aseismic creep (pre-slip), fault slip, and continuous sliding. Combining both active and passive monitoring increases accuracy of monitoring and can lead to better seismic risk mitigation@en

The risk of induced seismicity is increasing, due to the increasing human activities in the subsurface, such as gas extraction, geothermal energy production, or CO2 storage. To mitigate the seismic hazard and limit the risk, monitoring and forecasting are essential. We performed stress-driven and fluid-injection-driven fault reactivation on saw-cut faulted sandstone samples. Simultaneously to the fault reactivation, active acoustic transmission measurements were performed. The transmission data was analyzed using transmissivity, Coda Wave Interferometry and Decorrelation. They show clear changes in their signal before fault reactivation. This is attributed to the pre-slip, therefore these acoustic parameters are precursory signals to the imminent fault slip and can be used to forecast the upcoming seismicity. They provide a potential method for monitoring and forecasting to minimize the risk of natural and induced earthquakes.

@en

Underground energy storage (UES) in porous and cavity reservoirs can be used to balance the mismatch between the production and demand of renewable energy. Understanding the geomechanical behaviour of these reservoirs under different storage conditions, i.e., storage frequency and fluid pressure, is key in defining their capacity and effective lifetime. This work presents an analysis performed on sandstones to unravel their geomechanical response under cyclic loading. It includes, importantly, both experimental and numerical investigations under deviatoric stress conditions below the rock dilatant cracking threshold. From the experimental point of view, axial strains and acoustic emissions indicated that inelastic strains accumulated cycle after cycle, following a decreasing rate per cycle. Four types of deformations were interpreted: elastic, viscoelastic, plastic, and cyclic-plastic. Based on these experimental results and observations, the Modified Cam-clay model was extended to account for cyclic plastic deformations and the Kelvin-Voigt model was used to model visco-elasticity. This approach can be used to study cyclic sandstone deformation’s implications on subsidence, fault reactivation, and cap rock flexure, among other physical phenomena impacting a reservoir’s storage capacity.@en

With human activities in the subsurface increasing, so does the risk of induced seismicity. For mitigation of the seismic hazard and limiting the risk, monitoring and forecasting are essential. A laboratory study was performed to find precursors to fault failure. In this study, Red Pfaelzer sandstones samples were used, which are analog to the Groningen gas reservoir sandstones. A saw-cut fault was cut at 35 degrees, and the samples were saturated. Fault slip was induced by loading the sample at a constant strain rate, and simultaneously active acoustic transmission measurements were performed. Every 3 seconds 512 S-waves were sent, recorded, stacked to reduce the signal-to-noise ratio, and analyzed. The direct seismic wave velocity, coda wave velocity, and transmissivity were monitored before and during the reactivation of the faulted samples. Different loading patterns and confining pressures were investigated in combination with active acoustic monitoring. Velocity and amplitude variations were observed before the induced fault slip and can be used as precursory signals. Two methods to determine changing velocities were used. Direct S-wave velocities are compared to velocity change obtained by coda wave interferometry. Both analyses gave similar precursory signals, showing a clear change in slope, from increase to decreasing velocities and amplitudes prior to fault reactivation. Fault reactivation is preceded by fault creep and the destroying of some of the asperities on the fault plane, causing the seismic wave amplitude and velocity to decrease. Combining all precursors, the onset of fault slip can be determined and therefore upcoming slip can be forecasted in a laboratory setting. Our results show precursory changes in seismic properties under different loading situations and show a clear variation to the onset of fault reactivation. These results show the potential of continuous acoustic monitoring for detection and forecasting seismicity and help the mitigation of earthquakes.@en

Gas extraction has caused pressure differences along the field, triggering earthquakes, which are causing a lot of damage and social unrest in the Groningen area. Predicting the degree of these stress changes, and as a result, the potential onset and exact location of failure and seismicity, is very challenging. Therefore, developing good techniques that can monitor these changes is crucial for a better prediction and thus mitigation of failure and seismicity in the subsurface. Laboratory active acoustic-monitoring techniques are used to determine parameters that can forecast upcoming failure and seismicity. We show the use of coda wave decorrelation as a monitoring tool using sandstones analogues for the Groningen reservoir. Failure of the rock sample is preceded by the formation of micro-fractures. These fractures change the scattering properties of acoustic waves. The decorrelation coefficient K, as the indicator of the amount of scattering and thus be used as precursor to failure. We show that by monitoring K we can forecast the upcoming failure of the rock samples in the laboratory.@en