Using PS-InSAR observations to detect aseismic fault slip in the seismically active Groningen gas field

Since the start of production in 1968 in the Groningen gas field (Netherlands) considerable land subsidence (>30 cm) has occurred above the field. Variability in reservoir compaction has led to earthquakes on reactivated Mesozoic age reservoir faults. Even though the impact of this seismicity (MW [?]3.6) on society has been large, due to substantial structural damage to buildings, surface deformation induced by the co-seismic slip has been too small to detect using geodetic data. It is possible that differential compaction across faults is not only accommodated by seismic slip, but also by aseismic slip (e.g., creep). Aseismically slipping reservoir faults would relax the stresses in the reservoir and, thus, reduce the severity of the seismicity. In this study we explore the potential occurrence of aseismic slip on the reservoir faults. We perform a sensitivity analysis to investigate whether aseismic slip on the different reservoir faults has the capacity to produce detectable surface signals. We use the analytical Okada (1992) model of slip on a discrete dislocation in a uniform elastic half-space to simulate the deformations originating from slip on a wide range of fault geometries, representing the variability in the field. Unsurprisingly, laterally extensive faults with strong compaction contrasts across them (large differential slip magnitudes) produce the largest surface signals. To determine which potentially aseismically slipping faults produce surface signals that could be detectable in persistent scatterer InSAR time series, we analyze the surface patterns for large differential displacements across large length scales, since InSAR observations are most sensitive to spatially extensive patterns with high spatial gradients. We use the results of the sensitivity analysis to guide our search for patterns originating from aseismically slipping reservoir faults in PS-InSAR time series data of the Groningen area. First results show that these specific patterns are rare, indicating that the amount of aseismic slip is limited. For faults lacking surface signals related aseismic slip, the results of sensitivity analysis are used to determine upper limits for the aseismic differential slip magnitudes.

the analytical Okada (1992) model of slip on a discrete dislocation in a uniform elastic half-space to simulate the deformations originating from slip on a wide range of fault geometries, representing the variability in the field. Unsurprisingly, laterally extensive faults with strong compaction contrasts across them (large differential slip magnitudes) produce the largest surface signals. To determine which potentially aseismically slipping faults produce surface signals that could be detectable in persistent scatterer InSAR time series, we analyze the surface patterns for large differential displacements across large length scales, since InSAR observations are most sensitive to spatially extensive patterns with high spatial gradients. We use the results of the sensitivity analysis to guide our search for patterns originating from aseismically slipping reservoir faults in PS-InSAR time series data of the Groningen area. First results show that these specific patterns are rare, indicating that the amount of aseismic slip is limited. For faults lacking surface signals related aseismic slip, the results of sensitivity analysis are used to determine upper limits for the aseismic differential slip magnitudes.
Using PS-InSAR observations to detect aseismic fault slip in the seismically active Groningen gas field

INTRODUCTION
Since the start of production in 1968 in the Groningen gas field (Figure 1) considerable land subsidence (>30 cm) has occurred above the field. The current subsidence rate is ~8 mm/yr (Figure 2). The gas production has also led to earthquakes on reactivated Mesozoic age faults in the reservoir.

Figure 1:
Map of the Groningen gas field area (NE Netherlands), next to the city of Groningen. Major reservoir faults are also shown (after Pijnenburg et al., 2018).
This study is part of the DeepNL/Subsidence project, which aims to identify the subsurface drivers for the subsidence above the gas field, by assimilating geodetic time series (e.g. InSAR; Figure 2) into geophysical subsurface models. In order to build an efficient model for the reservoir and overburden, we model only those features that produce a detectable surface signal. Thus we investigate: does reservoir fault slip produce surface signals that are detectable in the geodetic time series?

CO-SEISMIC AND ASEISMIC FAULT SLIP
Seismicity in the Groningen reservoir started in 1991 (28 years after the start of production), triggered by differential compaction (Van Wees et al., 2014). The largest earthquake that has occured was the the 2012 M 3.6 Huizinge earthquake.
In a preliminary experiments we have applied estimates of the co-seismic slip magnitude and slip patch size (Dost & Kraaijpoel, 2013) to the analytical model by Okada (1992) of a rectangular dislocation in a uniform elastic half-space. This results in a surface displacement signal with an amplitude of ~1.5 mm and a spatial wavelength of 5-10 km. The associated spatial gradients of the signal are very small. This, combined with the signal formation being instantaneous and uncertianties in the InSAR time series (e.g. atmospheric noise and shallow soil deformation), makes it very unlikely that the signal is detectable. The co-seismic signal of lower magnitude earthquakes will be even smaller.
However, fault slip can also occur without producing earthquakes: aseismic slip (creeping faults). Here we investigate the surface displacement footprint of potential aseismic fault slip, for three different driving mechanisms (A-C) and a combined model (D). Reservoir compaction contrasts across a fault can induce fault slip. For example, area A in Figure 4 shows a large fault with a significant compaction constrast.
To simulate to potential aseismic fault slip for such differential compaction, we use a 2D plane strain finite element model (GTECTON; Govers & Meijer, 2001) of a vertical fault seperating two reservoir sections, of which only one is compacting (Figure 5). Compaction is driven by a 0.36 MPa/yr pressure change (ΔP), representing the approximate yearly field-average pore pressure drop. We run two version of the model: one with the fault locked (no aseismic slip) and one with an unlocked frictionless fault (continuous aseismic slip). With an applied yearly pressure drop, the surface modelled surface deformation represent the surface displacement rate (Figure 6a). In order to visualise the difference between the results of the two model versions, Figure 6b shows the residual. This residual represents the impact of aseismic fault slip. Differential compaction can also occur across faults with pre-existing reservoir offsets, without the necessity for a compaction contrast. Seismic studies have shown that some faults in the Groningen field offset the reservoir by up to 200 m (Figure 7). The finite element model is adjusted to simulate this effect of reservoir offset (Figure 8). The residual signal amplitude increases as the reservoir offset gets larger (Figure 10). However, the signal is still over an order of magntitude smaller than the overall signal of Figure 2. Reservoir fault slip can also be induced without differential compaction. Gas production causes a pore pressure reduction and vertical contraction (compaction) of the reservoir. The reservoir also wants to contract horizontally, leading to horizontal tensional (effective) stresses (e.g. Zoback & Zinke, 2002;Van Wees et al., 2014).
We isolate the effect of these horizontal stresses, using a model without any differential compaction (no offsets or compaction contrest), but with a non-vertical fault (Figure 11). An example area for this driving mechanims (C) is indicated on Figures 4 and 7. In this area the compaction is relatively strong, and compaction contrasts and reservoir offsets are small (as the geometry of Figure 11). Figure 11. A fault dip of 65° is applied.

Figure 12: Difference between surface displacement rate results of the locked and unlocked versions of the model of
The magnitude of the surface signal caused by aseismic fault slip due to the horizontal stresses (Figure 12) is smaller than that caused by reservoir offset of driving mechanism B (Figure 9b). The surface signal amplitude increases as the fault dips shallower (Figure 13). We combine the models B and C to simulate the effect of differential compaction from reservoir offset and the horizontal tensional stresses together (Figure 14). Such faults could be present in area D in Figures 4 and 7, where reservoir compaction and fault offsets are relatively large.
The resulting surface displacement residuals are shown in Figure 15. The combined residual amplitude is larger than that of driving mechanism B and C individually: 0.23 versus 0.13 and 0.06 mm/yr, for the vertical component.

Conclusions
Aseismic fault slip due to differential compaction caused reservoir offset (B) induces the strongest surface displacement signal.
Even with the combined effect of reservoir offset and fault dip (D), the footprint of the aseismic slip (0.23 mm/yr) is small compared to the ~8 mm/yr overall subsidence signal over the Groningen field.
Because of the low signal amplitude, long spatial wavelength (leading to small spatial gradients), and uncertianties in the InSAR time series (e.g. atmospheric noise and shallow soil deformation), we see detecting the tiny potential aseismic slip signal within the overall signal as impossible.
For the design of the geophysical models of the reservoir and overburden, reservoir fault slip can be ignored. ABSTRACT Since the start of production in 1968 in the Groningen gas field (Netherlands) considerable land subsidence (>30 cm) has occurred above the field. Variability in reservoir compaction has led to earthquakes on reactivated Mesozoic age reservoir faults. Even though the impact of this seismicity (M ≤3.6) on society has been large, due to substantial structural damage to buildings, surface deformation induced by the co-seismic slip has been too small to detect using geodetic data. It is possible that differential compaction across faults is not only accommodated by seismic slip, but also by aseismic slip (e.g., creep). Aseismically slipping reservoir faults would relax the stresses in the reservoir and, thus, reduce the severity of the seismicity. In this study we explore the potential occurrence of aseismic slip on the reservoir faults.

DISCLOSURES
We perform a sensitivity analysis to investigate whether aseismic slip on the different reservoir faults has the capacity to produce detectable surface signals. We use the analytical Okada (1992) model of slip on a discrete dislocation in a uniform elastic half-space to simulate the deformations originating from slip on a wide range of fault geometries, representing the variability in the field. Unsurprisingly, laterally extensive faults with strong compaction contrasts across them (large differential slip magnitudes) produce the largest surface signals. To determine which potentially aseismically slipping faults produce surface signals that could be detectable in persistent scatterer InSAR time series, we analyze the surface patterns for large differential displacements across large length scales, since InSAR observations are most sensitive to spatially extensive patterns with high spatial gradients.
We use the results of the sensitivity analysis to guide our search for patterns originating from aseismically slipping reservoir faults in PS-InSAR time series data of the Groningen area. First results show that these specific patterns are rare, indicating that the amount of aseismic slip is limited. For faults lacking surface signals related aseismic slip, the results of sensitivity analysis are used to determine upper limits for the aseismic differential slip magnitudes. W