Frequency dependence of attenuation components including mantle leakage in Garhwal Himalaya based on a modified MLTWA method

Abstract

SUMMARY Attenuation mechanism of seismic waves derived with an appropriate scattering model is a mandate for high-frequency earthquake ground-motion modelling. The assumption of uniform half-space is not always realistic and can have significant implications on the results obtained. We tested this assumption for the case of Garhwal Himalaya using a two-layered half-space model. To this end, we analysed the 1999 Chamoli earthquake aftershock sequence in Garhwal Himalaya based on the MLTWA method assuming both the uniform and layered models. The data set consists of 384 seismograms from 72 aftershocks recorded by a temporary seismic network of twelve stations. The uniform model cannot fully describe the energy loss by intrinsic and scattering attenuation, expecting bias in the corresponding attenuation estimates which are constrained by comparing results of numerical simulations with analytical solutions. By comparing the coda envelopes of both the analytical and numerical experiments for a suite of models that include the varying degree of scattering and intrinsic attenuation of crust and mantle (half-space), we could obtain improved attenuation estimates. Monte Carlo method was used to numerically solve the radiative transfer equation to deal with multiple isotropic scattering of seismic waves in 3-D heterogeneous acoustic media. The coda envelopes for the uniform model could reasonably fit those for the layered ones. The envelope fits suggest that the intrinsic and scattering coefficients for the uniform model are overestimated, by a factor that is frequency-dependent, as compared to those for the layered earth model. The factor lies in the range 4.0–7.5 (5.75 ± 1.75) below ∼3 Hz and 2.8–4.0 (3.4 ± 0.6) above 3 Hz for intrinsic, and the range 1.1–2.5 (1.8 ± 0.7) for scattering coefficient at 1–12 Hz. Under the assumption of a layered model, we could further reinterpret the apparent frequency-dependent energy loss as a sum of frequency-dependent mantle leakage and frequency-independent intrinsic attenuation. The results in the assumption of a layered model suggest strong scattering in the upper crust and weak mantle leakage in the frequency range 1.5–24 Hz. The small-scale random heterogeneities responsible for the observed scattering likely represent the fluid-filled fractured-crust, as evidenced from seismic tomography in the source area of the 1999 event.