|Observatories and Research Facilities for EUropean Seismology|
|Volume 3, no 2||December 2001||Orfeus Newsletter|
The May 7, 2001 Earthquake in the Ekofisk Area, North SeaJ. Braunmiller1, L. Ottemöller2, S.L. Jensen3, A. Ojeda3 and K. Atakan3
The coincidence of a seismic event occurring near ongoing hydrocarbon extraction immediately raised the question whether the event was natural or induced. Non-induced, purely tectonic earthquakes of magnitude 5 are possible in this region (Bungum et al.,2000). To even discuss a possible relation between seismic event and oil extraction requires a high precision earthquake location and depth estimate. The event occurred far from any national network and routine locations provide only moderately accurate epicenter estimates while depth is essentially undetermined. Here, we report about our efforts to obtain an improved epicenter location and our ongoing efforts for constraining the hypocenter depth.
We collected short-period and broadband seismic data from various sources to improve the azimuthal station coverage. Totally, we obtained data from more than 150 stations with an azimuthal gap for earthquake location to 83° (Figure 1). The seismogram onsets were generally unusually emergent as compared to earlier events in the same region, and therefore we could not pick arrival times for any phases besides Pn and Sn. We picked a total of 51 P- and S-wave arrival times. Event location was performed with the HYPOCENTER program (Lienert and Havskov, 1995) using the velocity depth model of Havskov and Bungum (1987) determined for southwestern Norway. The resulting earthquake epicenter at 56.565°N (± 4 km) and 3.182°E (± 7 km) is located at the northwestern corner of the Ekofisk oil field; the event origin time was 09:43:33.8 UTC.
Figure 1. Epicenter location (circle) of the May 7, 2001 Ekofisk event, and location of stations (black triangles) that provided data.
The seismograms from the May 2001 earthquake have unusual wave shapes compared to other events recorded in the North Sea. The seismograms are dominated by long-period signals with a corresponding deficit in high frequency energy, the P- and S-onsets are emergent and difficult to pick (Figure 2). In comparison, Figure 3 shows the signals from a small event (Mc = 3.4) that occurred off-shore southern Norway on June 2, 2001; the signals are more enriched in high frequencies and the P-onset is sharp. Emergent and long-period signals may be the result of a shallow source depth within the thick sedimentary layers of the North Sea graben structures.
Figure 2. Vertical component seismograms (in counts) from the Ekofisk event recorded at the broadband stations Kongsberg (KONO) in Norway (top) and Monsted (MUD) in Denmark (bottom). Note the emergent P-wave onset and the long-period character of the waveforms. Time in [s] is relative to the file begin (date and time in upper right corner of each seismogram trace).
Figure 3. Vertical component seismograms (in counts) from the June 2, 2001 Mc = 3.4 at 56.91°N and 7.98°E recorded at MUD (top) and KONO (bottom). Note the sharp P-wave onset and the high-frequency content.
Event location and seismogram shapes cannot answer the key question if the hypocenter is located above, within or below the reservoir. We followed two additional paths to obtain a better resolved, accurate hypocenter depth estimate and present preliminary findings here.
First, we performed 2-D ray-tracing (Zelt and Smith, 1992) through a more realistic crustal structure in the North Sea, i.e. a model that contains thick sedimentary strata. The model was constructed based on the results of several studies (Havskov and Bungum, 1987; Fichler and Hospers, 1989; Landes et al., 1998; Mooney et al., 1998; Nielsen et al., 2000; Hicks and Ottemöller, in press; Jensen et al., in press). Ray-tracing appears to favor source depths around 10 km slightly over shallower or deeper depths. The differences, however, are insignificant considering that our data set consists only of direct Pn- and Sn-waves. Travel-time ray-tracing with imprecisely known hypocenter through an imprecisely known crustal structure with only direct arriving phases cannot resolve the hypocenter depth to the desired level of accuracy. Depth resolution in ray-tracing approaches usually comes from travel-time differences (between secondary and primary arrivals) that show a strong depth dependency (like the travel time differences between PmP and Pg at epicentral distances near 100 km in continental environments). We do not have such data.
Second, we tried to model broadband data recorded at regional distances of 400 to 2200 km from the epicenter using a moment-tensor inversion routine (Nabelek and Xia, 1995; Braunmiller et al., 2000). We modeled the data at relatively long periods (larger than 30 seconds) to avoid ringy signals caused by the thickly sedimented travel paths through the North Sea. For varying crustal models, we found that the earthquake hypocenter was shallower than about 15 km. However, the "best-fit" depth depends on the crustal model and can vary from shallow depths of 3-5 km to 12-15 km depth. The current hypocenter depth resolution from waveform modeling is insufficient to decide if the event occurred below, above or within the reservoir (depth shallower than 7 km). Use of an improved velocity-depth model and modeling the closer-by data at higher frequencies could possibly result in an accurate, tightly constrained hypocenter depth estimate.
The regional moment tensor inversions performed at relatively long periods provided stable source parameter results independent of the exact choice of crustal structure. For each model and each frequency range tested, we obtained a normal faulting solution with north-south trending nodal planes. Figure 4 shows the waveform fits in the 40-60 s pass band for our preferred crustal model. The azimuthal distribution of the three component data constrains the source parameters tightly. The vertical component amplitude variations, for example, constrain the trend of the nodal planes tightly. The dips of the nodal planes and the seismic moment are probably less well constrained, because at shallow source depth, two moment tensor components (Mxz and Myz) are ill-determined causing large vertical dip-slip contributions. The misfit-vs-depth plot (Figure 5) shows that a source depth of 5 km fits the data best for that crustal model. Our preferred solution has one nodal plane with strike, dip, and rake of 356°, 85°, and -92°, respectively. The inversion results contained negligible amounts of compensated vector dipole or isotropic source components and a shear dislocation (faulting) is sufficient to explain the data.
Figure 4. Observed (red solid) and synthetic (blue dashed) seismograms for azimuthally selected stations in the 40-60 s pass-band. Z, R, T are the vertical, radial, and transverse components. Stations are listed in azimuthal order; numbers beneath station codes are event-station azimuth and distance. Triangles on the fault plane solutions (lower hemisphere projection) depict the station coverage (the red stations are shown above, black additional stations used for inversion not shown). Seismogram amplitudes are normalized to 100 km epicentral distance assuming cylindrical spreading.
Figure 5. Misfit versus hypocenter depth plot. Bottom: misfit when inverting in the 40-60 s pass-band using stations up to 2200 km from the epicenter. Top: misfit when inverting in the 30-50 s pass-band with data recorded at distances of up to 1200 km. The best-fit depth is marked by the red fault plane solution. The minima at shallow depth is pronounced. The fault plane solutions are stable over a wide depth range. The number beneath the fault plane solution is the moment magnitude. The steepening of one of the nodal planes and the increase in magnitude for shallow sources is caused by the in-determined nature of the Mxz and Myz moment tensor components. The increase in magnitude may thus not be real and a magnitude of 4.9-5.2 is probably a good estimate of the earthquake strength.
An accurate hypocenter depth estimate with small uncertainties is very difficult to obtain with the available data. Various methods (location, ray-tracing, broadband waveform modeling) result in hypocenter depth estimates in the upper 15 km of the crust. This accuracy is insufficient to ascertain if the event occurred above or below the reservoir. A source depth within the reservoir, however, seems unlikely since no disturbances were observed on the reservoir level (~3 km).
Discussion about a possible causal relationship between the hydrocarbon recovery and the May 7, 2001 event, is beyond the scope of this paper. This issue can only be addressed if additional data apart from the seismological observations presented in this paper, are available. Some general mechanisms that relate hydrocarbon extraction and seismicity (e.g. Segall, 1989; Grasso, 1992; Gomberg and Wolf, 1999) could be investigated in the future. Previously observed microseismicity within the reservoir (Maxwell et al., 1998), on the other hand, represents probably a gradual response of the rocks to stress accumulation, and thus appears to be different from the May 7, 2001 earthquake.
The main problem hampering the analysis of this event severely is the lack of data recorded at near distances. Data from seismograph in the vicinity of the reservoirs are needed to adequately record possible future earthquakes. Such data would allow an accurate determination of the hypocenter parameters and provide the associated ground motions, which would be of potential importance for the continued oil extraction in the North Sea.