Mantle Attenuation Estimated from Regional and Teleseismic P-waves of Deep Earthquakes and Surface Explosions

We estimated the network-averaged mantle attenuation t*(total) of 0.5 s beneath the North Korea test site (NKTS) by use of P-wave spectra and normalized spectral stacks from the 25 May 2009 declared nuclear test (mb 4.5; IDC). This value was checked using P-waves from seven deep (580–600 km) earthquakes (4.8 < M w < 5.5) in the Jilin-Heilongjiang, China region that borders with Russia and North Korea. These earthquakes are 200–300 km from the NKTS, within 200 km of the Global Seismic Network seismic station in Mudanjiang, China (MDJ) and the International Monitoring System primary arrays at Ussuriysk, Russia (USRK) and Wonju, Republic of Korea (KSRS). With the deep earthquakes, we split the t*(total) ray path into two segments: a t*(u), that represents the attenuation of the up-going ray from the deep hypocenters to the local-regional receivers, and t*(d), that represents the attenuation along the down-going ray to teleseismic receivers. The sum of t*(u) and t*(d) should be equal to t*(total), because they both share coincident ray paths. We estimated the upper-mantle attenuation t*(u) of 0.1 s at stations MDJ, USRK, and KSRS from individual and stacks of normalized P-wave spectra. We then estimated the average lower-mantle attenuation t*(d) of 0.4 s using stacked teleseismic P-wave spectra. We finally estimated a network average t*(total) of 0.5 s from the stacked teleseismic P-wave spectra from the 2009 nuclear test, which confirms the equality with the sum of t*(u) and t*(d). We included constraints on seismic moment, depth, and radiation pattern by using results from a moment tensor analysis and corner frequencies from modeling of P-wave spectra recorded at local distances. We also avoided finite-faulting effects by excluding earthquakes with complex source time functions. We assumed ω2 source models for earthquakes and explosions. The mantle attenuation beneath the NKTS is clearly different when compared with the network-averaged t* of 0.75 s for the western US and is similar to values of approximately 0.5 s for the Semipalatinsk test site within the 0.5–2 Hz range.