We show that near–real-time seismic monitoring of fluid injection allowed control of induced earthquakes during the stimulation of a 6.1-km-deep geothermal well near Helsinki, Finland. A total of 18,160 m3of fresh water was pumped into crystalline rocks over 49 days in June to July 2018. Seismic monitoring was performed with a 24-station borehole seismometer network. Using near–real-time information on induced-earthquake rates, locations, magnitudes, and evolution of seismic and hydraulic energy, pumping was either stopped or varied—in the latter case, between well-head pressures of 60 and 90 MPa and flow rates of 400 and 800 liters/min. This procedure avoided the nucleation of a project-stopping magnitudeMW2.0 induced earthquake, a limit set by local authorities. Our results suggest a possible physics-based approach to controlling stimulation-induced seismicity in geothermal projects.

A seismic network was installed in Helsinki, Finland to monitor the response to an ∼6-kilometer-deep geothermal stimulation experiment in 2018. We present initial results of multiple induced earthquake seismogram and ambient wavefield analyses. The used data are from parts of the borehole network deployed by the operating St1 Deep Heat Company, from surface broadband sensors and 100 geophones installed by the Institute of Seismology, University of Helsinki, and from Finnish National Seismic Network stations. Records collected in the urban environment contain many signals associated with anthropogenic activity. This results in time- and frequency-dependent variations of the signal-to-noise ratio of earthquake records from a 260-meter-deep borehole sensor compared to the combined signals of 24 collocated surface array sensors. Manual relocations of ∼500 events indicate three distinct zones of induced earthquake activity that are consistent with the three clusters of seismicity identified by the company. The fault-plane solutions of 14 selected ML 0.6–1.8 events indicate a dominant reverse-faulting style, and the associated SH radiation patterns appear to control the first-order features of the macroseismic report distribution. Beamforming of earthquake data from six arrays suggests heterogeneous medium properties, in particular between the injection site and two arrays to the west and southwest. Ambient-noise cross-correlation functions reconstruct regional surface-wave propagation and path-dependent body-wave propagation. A 1D inversion of the weakly dispersive surface waves reveals average shear-wave velocities around 3.3  km/s below 20 m depth. Consistent features observed in relative velocity change time series and in temporal variations of a proxy for wavefield partitioning likely reflect the medium response to the stimulation. The resolution properties of the obtained data can inform future monitoring strategies and network designs around natural laboratories.

No abstract

Abstract. In this study, we present a high-resolution dataset of seismicity framing the stimulation campaign of a 6.1 km deep enhanced geothermal system (EGS) in the Helsinki suburban area and discuss the complexity of fracture network development. Within the St1 Deep Heat project, 18 160 m3 of water was injected over 49 d in summer 2018. The seismicity was monitored by a seismic network of near-surface borehole sensors framing the EGS site in combination with a multi-level geophone array located at ≥ 2 km of depth. We expand the original catalog of Kwiatek et al. (2019), including detected seismic events and earthquakes that occurred 2 months after the end of injection, totaling 61 163 events. We relocated events of the catalog with moment magnitudes between Mw −0.5 and Mw 1.9 using the double-difference technique and a new velocity model derived from a post-stimulation vertical seismic profiling (VSP) campaign. The analysis of the fault network development at a reservoir depth of 4.5–7 km is one primary focus of this study. To achieve this, we investigate 191 focal mechanisms of the induced seismicity using a cross-correlation-based technique. Our results indicate that seismicity occurred in three spatially separated clusters centered around the injection well. We observe a spatiotemporal migration of the seismicity during the stimulation starting from the injection well in the northwest–southeast (NW–SE) direction and in the northeast (NE) direction towards greater depth. The spatial evolution of the cumulative seismic moment, the distribution of events with Mw≥1, and the fault plane orientations of focal mechanisms indicate an active network of at least three NW–SE- to NNW–SSE-oriented permeable zones, which is interpreted to be responsible for the migration of seismic activity away from the injection well. Fault plane solutions of the best-constrained focal mechanisms and results for the local stress field orientation indicate a reverse faulting regime and suggest that seismic slip occurred on a sub-parallel network of pre-existing weak fractures favorably oriented with the stress field, striking NNW–SEE with a dip of 45∘ ENE parallel to the injection well.

Abstract. In this study, we present a high-resolution dataset of seismicity framing the stimulation campaign of a 6.1 km deep Enhanced Geothermal System (EGS) in Helsinki suburban area and discuss the complexity of fracture network development. Within St1 Deep Heat project, 18 160 m3 of water was injected over 49 days in summer 2018. The seismicity was monitored by a seismic network of near-surface borehole sensors framing the EGS site in combination with a multi-level geophone array located at ≥ 2 km depth. We expand the original catalog of Kwiatek et al. (2019) and provide the community with the dataset including detected seismic events and earthquakes that occurred two month after the end of injection, totalling to 61 163 events. We relocated events of the catalog with sufficient number of available phase onsets and moment magnitudes between Mw −0.7 and Mw 1.9 using the double-difference technique and a new velocity model derived from a post-stimulation vertical seismic profiling campaign. The analysis of the fault network development at reservoir depth of 4.5–7 km is one primary focus of this study. To achieve this, we investigate 191 focal mechanisms of the induced seismicity using cross-correlation based technique. Our results indicate that seismicity occurred in three spatially separated clusters centered around the injection well. We observe a spatio-temporal migration of the seismicity during the stimulation starting from the injection well in northwest (NW) – southeast (SE) direction and in northeast (NE) direction towards greater depth. The spatial evolution of the cumulative seismic moment, the distribution of events with Mw ≥ 1 and the fault plane orientations of focal mechanisms indicate an active network of at least three NW–SE to NNW-SSE orientated permeable zones which is interpreted to be responsible for migration of seismic activity away from the injection well. Fault plane solutions of the best-constrained focal mechanisms as well as results for the local stress field orientation indicate a reverse faulting regime and suggest that seismic slip occurred on a sub-parallel network of pre-existing weak fractures favorably oriented with the stress field, striking NNW-SEE with a dip of 45° ENE, parallel to the injection well.

Abstract:We use Matlab 3D finite element fluid flow/transport modelling to simulate localized wellbore temperature events of order 0.05-0.1 • C logged in Fennoscandia basement rock at~1.5 km depths. The temperature events are approximated as steady-state heat transport due to fluid draining from the crust into the wellbore via naturally occurring fracture-connectivity structures. Flow simulation is based on the empirics of spatially-correlated fracture-connectivity fluid flow widely attested by well-log, well-core, and well-production data. Matching model wellbore-centric radial temperature profiles to a 2D analytic expression for steady-state radial heat transport with Peclet number P e ≡ r 0 φv 0 /D (r 0 = wellbore radius, v 0 = Darcy velocity at r 0 , φ = ambient porosity, D = rock-water thermal diffusivity), gives P e~1 0-15 for fracture-connectivity flow intersecting the well, and P e~0 for ambient crust. Darcy flow for model P e~1 0 at radius~10 m from the wellbore gives permeability estimate κ~0.02 Darcy for flow driven by differential fluid pressure between least principal crustal stress pore pressure and hydrostatic wellbore pressure. Model temperature event flow permeability κ m~0 .02 Darcy is related to well-core ambient permeability κ~1 µDarcy by empirical poroperm relation κ m~κ exp(α m φ) for φ~0.01 and α m~1 000. Our modelling of OTN1 wellbore temperature events helps assess the prospect of reactivating fossilized fracture-connectivity flow for EGS permeability stimulation of basement rock.

Earthquakes are frequently accompanied by public reports of audible low-frequency noises. In 2018, public reports of booms or thunder-like noises were linked to induced earthquakes during an Engineered Geothermal System project in the Helsinki Metropolitan area. In response, two microphone arrays were deployed to record and study these acoustic signals while stimulation at the drill site continued. During the 11 day deployment, we find 39 earthquakes accompanied by possible atmospheric acoustic signals. Moment magnitudes of these events ranged from $$-0.07$$ - 0.07 to 1.87 with located depths of 4.8–6.5 km. Analysis of the largest event revealed a broadband frequency content, including in the audible range, and high apparent velocities across the arrays. We conclude that the audible noises were generated by local ground reverberation during the arrival of seismic body waves. The inclusion of acoustic monitoring at future geothermal development projects will be beneficial for studying seismic-to-acoustic coupling during sequences of induced earthquakes.

Abstract:We use Matlab 3D finite element fluid flow/transport modelling to simulate localized 10 wellbore temperature events of order 0.05-0.1 o C logged in Fennoscandia basement rock at ~ 1.5km 11 depths. The temperature events are approximated as steady-state heat transport due to fluid 12 draining from the crust into the wellbore via naturally occurring fracture-connectivity structures. 13Flow simulation is based on the empirics of spatially-correlated fracture-connectivity fluid flow 14 widely attested by well-log, well-core, and well-production data. Matching model wellbore-