Seismic attenuation tomography of the North-Western Himalaya using Coda waves

Abstract

<p>We use 4695 local waveforms from 1206 earthquakes (epicentral distance < 350 km and 2.0 ≤ Mw ≤ 5.5) recorded by IISER Kolkata network (IK) at 22 stations (32°N to 35°N latitude and 74°E to 77°E longitude), located within the North-Western Himalaya (28°N to 39°N latitude and 68°E to 81°E longitude). We study the coda waves which are generally the tail of a seismogram and arrive after the main seismic waves. We use the temporal decay of coda amplitude to calculate the coda quality factor (Q<sub>c</sub>) from which we estimate the attenuation (Q<sub>c</sub><sup>-1</sup>). We consider the single back-scattering model (Aki & Chouet, 1975) where both the scattering (Q<sub>sc</sub><sup>-1</sup>) and intrinsic (Q<sub>i</sub><sup>-1</sup>) component of the attenuation are included in the measurement. We use a lapse time of 2t<sub>s</sub> (t<sub>s</sub> is the S-wave arrival time) as the starting point of the coda window. Then, we consider multiple forward-scattering model, where the attenuation (Q<sub>c</sub><sup>-1</sup>) is dominantly dependent on the intrinsic (Q<sub>i</sub><sup>-1</sup>) component. In this model we use lapse time greater than 2t<sub>s</sub> so that the coda waves encounter multiple scatterers and enter the diffusive regime. We calculate the frequency dependent quality factor for each earthquake-receiver path at frequencies 1 to 14 Hz using the linear least squares approach on temporal decay of coda amplitude. We calculate Q<sub>0</sub> (quality factor at a reference frequency f<sub>0</sub> which is chosen to be 1 Hz for the analysis) and its frequency dependence (η) using weighted least squares approach on the power law dependence of Q<sub>c </sub>on frequency. To see the lateral variation of Q in our study area, we have produced 2-D maps by combining the Q<sub>c </sub>measurements together in a tomography. For single back-scattering model we use the back-projection algorithm which is based on the calculation of area overlap of ellipses with the gridded region. For multiple forward-scattering model, the same back-projection algorithm is modified to calculate the length overlap of traces with the gridded region. To understand the spatial resolution of the 2-D Q<sub>c </sub>maps, we use the point spreading function test which quantifies the recovery of Q<sub>c </sub>perturbation. In addition to this, we also perform a standard checkerboard resolution test to ensure simultaneous recovery of Q<sub>c </sub>perturbation. We observe low Q in the Kashmir basin and Lesser Himalaya and high Q in surrounding northeastern Higher Himalaya which clearly correspond to the coda wave attenuation signatures in the older Tethyan sedimentary rocks and crystalline igneous rocks in these regions respectively.</p>