Earthquakes and seismic hazard in Sweden

Transcription

1 Earthquakes and seismic hazard in Sweden Björn Lund, Roland Roberts & Reynir Bödvarsson Uppsala University

2 Outline Nordic and Swedish seismicity Comparison to plate boundary seismicity in Japan. Extrapolation to large earthquakes and endglacial fault seismicity Basics in seismic hazard evaluation. Some comments on current seismic hazard evaluation of nuclear power plants.

3 The joint Nordic bulletin FENCAT Instrumental from 1904 but sparse networks. Seismicity in Scandinavia and Sweden and Finland are low seismicity regions far from an active plate boundary, i.e. an intraplate region. Finland

4 The modern Swedish National The growth of the network from 6 stations in 2000 to 67 stations in 2012 means that the data set is temporally and spatially inhomogeneous during these 12 years of operation. Seismic Network (SNSN) Seismic epoisodicity in space and time is therefore difficult to assess.

5 Events in Sweden Thousands of events! Mining blasts and mining induced earthquakes, construction blasts, quarries and earthquakes.

6 Earthquake epicenters All events classified as earthquakes by SNSN since the start of automatic operations in August 2000, until December events. Note similar distribution to the FENCAT data.

7 Earthquake epicenters A more homogeneous data set, from 1 Jan events

8 Earthquake epicenters An even more homogeneous data set, from 1 Jan 2008 with M > events

9 Earthquake depths Preliminary! These are the most reliable routine processing depths.

10 SNSN earthquake frequency vs. magnitude The graph shows the (log) number of events exceeding the magnitude on the x-axis. Note the stable straight-line behaviour (normally observed for all areas). The apparent lack of events at small magnitudes is just an artifact the data set used for this graph is not complete for small events Very few large events randomness clearly seen for these line not so straight at the right, large M, small N

11 Swedish b-values Stable b-values require data over three magnitudes, i.e. some 1,000 events. People often make do with 100 or fewer events. 11 years of SNSN data, Mc 0.1 => b = years of FENCAT data, Mc 2.0 => b = 0.72 Spatial variations in b-value may be possible with SNSN data.

12 Comparison to plate boundary seismic activity: Japan Japan has complicated plate tectonics and a LOT of earthquakes!

13 Comparison to plate boundary seismic activity: Japan Yellow dots show earthquakes larger than M 8

14 Comparison to plate boundary seismic activity: 2011 Tohoku-Oki Magnitude 9.0 earthquake followed by tsunami

15 Comparison to plate boundary seismic activity: 2011 Tohoku-Oki Peak Ground Acceleration Early slip inversion showed slip over 600 x 250 km (30% of Sweden) for more than 3 min. We now know that the maximum slip was shallow, and more than 50 m. Maximum measured acceleration was 2933 gal (cm/s 2 ) 2.99g

16 Comparison to plate boundary seismic activity: 2011 Tohoku-Oki Fukushima Measured accelerations at Fukushima nuclear power plant are up to 0.56g, higher than the design tolerance of 0.46g. The fission process is stopped, external power is lost and the emergency diesel generators start as they should. The tsunami hit the power plant at 15:41, almost one hour after the earthquake. The protective sea wall was 5.7 m high, the tsunami more than 15 m. Explosion in reactor 3, 14/3 Damage to reactors 3 and 4 16/3

17 Comparison to plate boundary seismic activity: 2011 Tohoku-Oki Locating nuclear power plants in Japan is difficult.

18 Comparison to plate boundary seismic activity: Japan Frequency-magnitude graph for Japan , blue. Earthquake rate in Norden a factor ~1000 less than in Japan It is plausible that the red line can be extrapolated i.e. with a longer observation interval more and larger events will be observed

19 Swedish b-values A very rough extrapolation gives one magnitude 6 event per 1,000 years and one magnitude 7 every 10,000 years. Can we have a Tohoku earthquake in Sweden or Finland?

20 Swedish b-values A very rough extrapolation gives one magnitude 6 event per 1,000 years and one magnitude 7 every 10,000 years. Can we have a Tohoku earthquake in Sweden or Finland? No, we do not have such a plate boundary in our vicinity. But can we have M 8 earthquakes in Sweden?

21 Yes! Endglacial earthquakes of magnitude 7-8 in northern Sweden The Pärvie fault at Kaitumjaure (B. Lund) The Pärvie fault at Kamasjaure (R. Lagerbäck) Large earthquakes occurred at the end of the latest glaciation, approximately 9,500 years ago. These are high stress reverse faults with up to 30 m throw. The largest, the Pärvie fault, is 155 km long and runs just west of Kiruna. Estimated magnitude 8.2.

22 The endglacial faults are still active 80% of the earthquakes in Sweden north of 66N occur within 45 km of an endglacial fault! Most events are located to the southeast of the faults. Red circles: SNSN, Blue circles: FENCAT,

23 What drives the earthquakes? Focal mechanisms and stress inversions suggest that the tectonic loading is the main source of stress accumulation. The large activity along the northeast coast could be influenced/triggered by glacial rebound, but that trigger is small nowadays. In northern Sweden stress is released preferentially along the zones of weakness related to endglacial faulting. Correlation with older zones need be explored. What is the rate of stress/strain accumulation in Sweden/Finland?

24 Fennoscandian strain rates Scherneck et al. (2010): Maximum 4 nano/yr in the center of the Shield. Other estimates for shield areas vary from less than 2 nano/yr (GPS) to /yr (geology). -> We don t really know! How often can we have a Pärvie earthquake, M 8+? Stress release ~10 MPa. Strain rate of /yr, Young s modulus of 100 GPa => 1 million years return time. Plus an Ice Age??

25 Seismic hazard calculation: source zonation The region is divided into zones based on relevant criteria such as seismic activity and geology. Various zonations are included in the final logic tree. Example from Wahlström & Grünthal (2001)

26 Seismic hazard calculation: b-values for the zones Determine annual mean rates and b-values for the different zones defined above. Necessary with zones that contain enough earthquakes for the calculation. Propagate uncertainties to a logic tree.

27 Seismic hazard calculation: maximum magnitude The most difficult task in a low seismicity area, where we have not even been close to observing a full seismic cycle. We need to estimate a maximum earthquake magnitude in each zone. This is often done rather ad hoc by adding one or one half magnitude to the maximum observed magnitude. Propagate uncertainties to a logic tree.

28 Landslides and collapse in Kaliningrad. Felt widely in the Baltic countries, Polen, Germany, Denmark and Sweden, up to 800 km distance. Seismic hazard calculation: maximum magnitude Out of the blue: The Kaliningrad M ~5 earthquakes of 21 Sept Grünthal et al. (2008)

29 Seismic hazard calculation: attenuation relations or Ground Motion Prediction Equations (GMPE) These relate the observed Peak Ground Acceleration (PGA) to the earthquake magnitude, source to receiver distance, source characteristics, local site conditions etc. A large number of GMPEs exist, none derived in Fennoscandia. The Baltic Shield propagates high frequencies very well! Work in progress in Finland for the hazard assessment of the proposed new NPP in northern Finland. Propagate uncertainties to a logic tree.

30 Among others: Example result: Fennoscandian seismic hazard assessment Kijko et al. (1993) GSHAP (1999) Wahlström & Grunthal (2001) SHARE (EU, upcoming 2012/13) A mirror of recorded events. Map of 90% probability of non-exceedence of horizontal PGA (m/s2) in 50 years, corresponding to a mean return period of 475 years. Wahlström & Grünthal (2001)

31 Some points regarding seismic hazard for Swedish nuclear power plants Response spectra used for seismic hazard estimation at Swedish NPPs are based on near-field data collated in Japan, from mostly Japanese earthquakes. The Japanese spectra need be adjusted for the difference in bedrock conditions between Sweden and Japan, as the near-surface layers in Japan are considerably softer than those in Sweden, leading to less amplification here. Earthquake source depth is assumed to be 30 km in the calculations, SNSN data shows it to be mostly in the km range. Seismic hazard is dominated by events at km hypocentral distance. CAV is a good companion to PGA in a hard rock, low attenuation environment.

32 Summary Extrapolating from the freq-magnitude pattern for small events to assess the probablity of much larger ones may overestimate risk in slowly-deforming areas Risk is certainly very low, but not negligable. More research is necessary to assess it better. There is now much more data on Swedish earthquakes available so we should be able to improve on most aspects of the seismic hazard assessment. Some of these issues will be addressed in the Fennovoima project.