Safe CO₂ – II

-    Goal of the Project:
We will focus on the following problems:
- What is the value of microseismic monitoring: do event locations form a good representation of the main flow directions? Does microseismic activity allow for conclusions on rock properties or only helps to identify regions of high pore pressure?
- Is there aseismic deformation during fluid injection at the CO₂Lab? How can we quantify its amount? What can these observations contribute to the understanding of the injective/hydraulic behaviour of the rock?
- Can we make use of the continuity of seismic recordings by analysing the noise in addition to the signal? Can we employ ambient seismic noise tomography to enhance our understanding of fluid flow and changes in the subsurface qualitatively as well as quantitatively?
- Is it feasible to turn Svalbard into a CO₂ neutral community by injecting a sufficient amount of CO₂ in subsurface reservoir(s) without breaching the caprock?

-    Technical content:
The main objective of the project is to correlate CO₂ injection data at the Spitsbergen CO₂ Lab with microseismic activity, aseismic deformation, and temporal variations in the seismic velocity structure. Secondary objectives comprise the use of event locations and source mechanisms to map size and orientation of fracture systems in the rock activated by the CO₂ injection. Furthermore, we will contribute to a better understanding of the geomechanical response of the reservoir to injection. To this end, we will upgrade the existing network of borehole geophones at the Longyearbyen CO₂ Lab with temporary broadband seismometers at the surface. To process the data, we will develop algorithms to detect weak and slow slip events and utilize noise recordings to extract Greens functions to be inverted for velocity model and its temporal changes. Data interpretation will be complemented by numerical modelling of slow slip events with respect to fracture orientation and slip velocities.

-    Technical advantages:
a)    Estimating the exact position of microearthquakes provides information on pressure buildup in the formation, which constitutes vital information for realistic reservoir simulations.
b)    From a previous project (SafeCO₂), NORSAR already operates a microseismic monitoring network consisting of 10 deep and shallow borehole geophones at the Longyearbyen CO₂Lab.
c)    Recent results have shown that cross-correlation of the ambient seismic noise field between different sensors leads to the reconstruction of the impulse function between these sensors and can be used to track the temporal evolution of the crustal properties. Since the permanent microseismic monitoring network at the Longyearbyen CO₂ Lab is recording in a continuous mode, we possess records of ambient seismic noise since 2010.
d)    Tremor detected during hydrofracturing in a shale gas reservoir (Das and Zoback, 2011) seems to originate from two distinct hypothetical fracture planes matching locations of microearthquake clusters. Such slow slip events may contribute considerably to enhance fluid transport. Their signal might be contained in the recorded data.
e)    All relevant data from the Longyearbyen CO₂ Lab on Svalbard, Norway, are accessible through a MoU between NORSAR and UNIS via the SafeCO₂ project (grant no. 189994).

-    R&D challenges:
a)    The present microseismic monitoring network is tuned to observe microseismic events within the potential reservoir layers. However, both the observation of tremor or low-frequency events indicative of slow slip events as well as ambient seismic noise tomography requires the observation of much lower frequencies. Therefore, we will supplement the permanent network with a temporary surface broadband seismometer network. Due to the increased number of stations, we also hope to detect more small events, which may be hidden in the noise. However, the success of these tasks will strongly depend on data quality (difficult conditions on Svalbard) as well as the presence of tremor and events in the data.
b)    Data investigated from three CO₂ geological storage sites (Sleipner, Snøhvit and Krechba) indicate essential differences in geomechanical response to injection with respect to well-head pressures and microseismic activity (Eiken et al., 2011).

-    Results to date:
•    The permanent seismic installation at the CO₂Lab wellpark:
An additional string of 8 geophones (15 Hz, 3-components) was deployed in the well Dh4, down to 594 m in October 2013. In late November, a GPS antenna was connected to the recording system (3 digitizer, GEODEs), allowing to better synchronize the instruments. Unfortunately, the additional geophone string had to be removed on the 15th of May 2014 due to Health & Safety issues because of a gas leakage from Dh4. In addition, the cable leading to the geophones installed in DH3 was accidentally cut on the 14th of May 2014 and its repair was not successful There are substantial problems with the network and also data transfer to NORSAR since the beginning of 2015.

•    Microseismic data analysis of CO₂ Lab network data:
First, a continuous data stream had to be constructed due to clock drift among data channels recorded by different GEODE recorders or even among different ADC boards within one GEODE. In addition, much of the data recorded at the CO₂ Lab site is affected substantially by electronic noise. We have demonstrated that the addition of dedicated noise sampling channels allows us to use Widrow's adaptive noise cancellation algorithm to eliminate components of this electronic noise. The data recorded during the 2010 injection has been reprocessed using waveform cross-correlation and lead to the detection of 8 events in addition to the master event. The processing of cleaned data recorded at permanent network during and after the injection in 2014 did not reveal any microseismic events, but 3 types of repeating characteristic signals were observed.

•    Ambient seismic noise analysis:
Noise characteristics have been analysed using recordings on the SPITS array. A bifurcation of power spectral density functions is observed around 0.1-0.25 Hz and 2-5 Hz. The f-k analysis shows abrupt seasonal changes: in winter, ambient seismic noise is dominated by body waves from SW direction, whereas in summer, surface waves from SSE direction are dominant. We computed daily noise cross-correlation functions (CCFs) from data recorded at the CO₂ Lab for a period of 7 months using the Whisper code (developed by ISTerre). One surprising finding was that the presence of gas travelling upwards within well DH4 was not only detected on the waveform recordings, but also changed the behaviour of the CCFs. In addition, cross-correlation functions have been computed for stations of the SPITS array situated close-by indicating a variation in the direction of sources of ambient seismic noise at different times of the year.

•    The temporary field installation in summer 2014 (SEISVAL):
From 6th to 16th May 2014, we installed 12 broadband seismic stations rented from SisMob (University of Grenoble, France) within the Adventdalen region. The instrument responses have been obtained from ISTerre and have been removed from the data. The data from the whole time period has been checked manually for local events and a set of semi-synthetic data has been constructed to model the detection of slow slip events. The NORSAR 3D Adventdalen velocity model has been updated and extended in order to comprise the locations of SEISVAL sensors. It has been used to perform FD wave propagation modelling in order to test a model-based matched field processing approach as alternative to the classical event location method. With help of simulations on the Oslo’s Universities’ supercomputer Abel (additional NOTUR grant), a Green’s functions library has been computed to examine the theoretical resolution of the SEISVAL network.

•    Paralana data:
Source parameters have been computed and a set of master events has been defined in order to test different methods for detection of weak events.

•    Integration of supplemental information:
Since performing new VSP measurements was not feasible, the 2012 shear wave downhole experiment at the observation well Dh3 has been reanalysed instead. It reveals a low velocity zone in the shallow subsurface above 100 m depth. The derived velocities at the near-surface indicate predominantly soft soil characteristics rather than the “permafrost” characteristics previously assumed for this zone. Ice-like characteristics were, however, derived at a depth of nearly 100 m. NORSAR's proprietary software SeisRoX has been used to model the effect of CO₂ presence in potential reservoir layers of the CO₂ Lab on PSDM (pre-stack depth migrated) seismic profiles employing the Gassmann model equations. Results indicate that in the seismic profiles, the interface between CO₂ and brine is visible and that the seismic amplitude will be higher the higher the layer porosity is, but that concluding on a CO₂ concentration will be difficult. In addition, the geometry of a potential an active seismic survey will have to be constructed carefully in order to reveal the shape of the CO₂ plume. An analytical geomechanical model has been set up to evaluate potential slip of preexisting fractures at: a) 428 m in Dh6 and b) 920 m in Dh4. The result indicates that the majority of fractures in the reservoir (which are close to vertical) seem to slip easily with small injection pressures. In addition, a velocity stepping direct shear test on a shale sample (386 m in Dh6) indicates that slip on bedding-parallel fractures may be non-seismic.