Regional Spectral Analysis of Moderate Earthquakes in Northeastern North America: Resolving Attenuation!

Transcription

1 Regional Spectral Analysis of Moderate Earthquakes in Northeastern North America: Resolving Attenuation. John Boatwright and Linda Seekins 1997 M4.4 Cap-Rouge, Quebec 2002 M5.0 Ausable Forks, New York 2005 M4.7 Rivière-du-Loup, Quebec 1988 M5.8 Saguenay, Quebec : The attenuation results for the Appalachian Province were first obtained by analyzing regional data from the four earthquakes listed above The sttenuation technique assumes r -1 - r -1/2 geometric spreading but uses SRI (impedance) to model hard and stiff rock sites In 2011, we extended the dataset to include earthquakes and stations in Ontario, and obtained a different attenuation model 1 Regional Spectral Analysis of Small and Moderate Earthquakes in Northeastern North America John Boatwright and Tim MacDonald 2000 M4.6 Kipawa, Quebec 200 M.7 Maniwaki, Quebec 200 M.5 Cap-Saint-Fidele, QC 2006 M.7 Thurso, Ontario 2006 M.5 L Ile-aux-Coudres, QC 2010 M.1 Merrimack, NH 2006 M.4 Eagle Lake, Maine 200 M4.1 La Malbaie, QC 2012 M2.6 La Malbaie, QC 2012 M2.8 St Anne-des-Monts 2012 M.8 Happy Valley, QC 2006 M4.0 Bar Harbor, Maine 2008 M.5 Rivière-du-Loup, QC 2010 M5.0 Val-des-Bois, Quebec 2010 M.4 Donnacona, Quebec 2011 M.7 Hawkesbury, Ontario 2011 M.2 Les Escoumins, QC 2012 M.0 Port-Joli, Quebec 2012 M.1 New Brunswick 2012 M2.7 Petit-Rocher, NB 2

2 . u(r, f ) = F F s S( f ) exp( πfr/βq ) M o( f ) g( r,r o ) : 4πρ o β o FF s - free surface amplification and average radiation pattern S(f) - site amplification computed using average impedance from Boore and Joyner (1997) # r r r o g(r,r o ) = $ %(r o r) 1/ 2 r > r o - geometrical spreading from Street et al. (1975)with r o = 50 km Q = Q o f q - anelastic attenuation form of Aki and Chouet (1975) ρ o β o - density and S-wave velocity at the source depth Frequency (Hz) 4

3 Frequency (Hz) Frequency (Hz) 6

4 Frequency (Hz)

5 Frequency (Hz)

6 Frequency (Hz) Frequency (Hz) 12

7 Magnitude 5 There are few recordings of large earthquakes in NENA 4 Use small & moderate earthquakes to determine ground motion attenuation Distance (km) 1000 Estimate suitable Δσ from largest earthquakes Atkinson (2004) analyzed vertical and horizontal CNSN recordings down to M > 2 and out to r < 2000 km We analyzed the four best recorded moderate (4 M 5) earthquakes using stations out to r < 600 km 12 How Noise Affects Attenuation Estimates If we revise the usual decomposition to include noise ln v(r, f)+n I (f) = lnω(f) + lns I (f) - lng(r,r o ) - πfr/βq Where v(r, f) and n I (f) are the instrument-corrected signal and noise spectra, Ω(f) is the isotropic source term, S I (f) is the site amplification, g(r, r o ) is the geometrical spreading, and πfr/βq is the anelastic attenuation Noise is almost always additive in the power spectrum ln v(r, f)+n I (f) and affects recordings of smaller earthquakes and distant stations more strongly When n I (f)/v(r, f) increases, πfr/βq decreases (that is, Q increases). Worse, the distant stations dominate the attenuation estimates, as a result of the large r s in the linear term πfr/βq, so researchers must be careful to consider both noise and spatial homogeneity when they add these stations 14

8 5 Magnitude 4 Boatwright and Seekins (2011) 2011 analyses 2012 Atkinson (2004) Hypocentral Distance (km) Frequency (Hz) 16

9 5 Magnitude 4 Erickson et al. (2004) Atkinson (2004) Hypocentral Distance (km) Atkinson (2004) Magnitude 4 Benz et al. (1997) Hypocentral Distance (km)

10 Frequency (Hz) 19 Boore et al. (2010) re-evaluated stress parameters for AB95 and AB06 Stronger attenuation produces better fits from 0 to 200 km but cannot be extended past 600 km 20

11 . Two characteristics argue that the attenuation Final Speculation about Attenuation : Q = 410 f 0.5 for 0.2 f 20 Hz we have obtained for the Appalachian and southeastern Grenville Provinces is reasonable. We have restricted the recordings analyzed to the largest earthquakes at smaller distances (r < 600 km). This strategy optimizes the signal to noise in the data and enables us to resolve stronger attenuation. This strategy also serves to limit the area that the raypaths sample. The estimates of Q(f) are remarkably well fit by the Aki and Chouet (1975) function Q o f q on this broad frequency band 21 Conclusions The crustal waveguide in northeastern North America can be modeled very simply by revising r o = 50 km and using Q = 410 f 0.5 for the Appalachian and southeastern Grenville Provinces and Q = 52 f 0.5 for western Ontario These estimates of regional attenuation are lower than most previous estimates, particularly Benz et al. (1997), Erickson et al. (2004), Atkinson (2004), and Atkinson (this workshop) Because Q estimates are biased high by additive noise, it is important to analyze large earthquakes at regional distances ( r < 600 km). The available recordings of small and moderate earthquakes in NENA generally do not require the analysis of more distant stations. 21