Improving Receiver Function Imaging With High-resolution Radon Transform

Seismology is one of the most effective ways to study the Earth’s internal structures.With the increasing number of dense seismic arrays around the globe,the array method has gradually become an important tool for retrieving high-quality seismic signals and studying subtle crustal and mantle structures.The development of data processing and imaging methods suitable for seismic arrays has also become a hot topic of seismological research.In this thesis,we study the data processing strategy of receiver function(RF)for seismic arrays.As one of the important passive source seismic imaging methods,RF is widely applied to study the Earth’s interior in various scales.The RF method usually utilizes the deconvolution operation to remove the source-time function in the teleseismic wavefield and isolate the structural response beneath the stations.However,the seismic recordings of the temporary network are often contaminated by strong random and erratic noises,which seriously degrade the stability and accuracy of the RF calculation,thereby reducing the imaging quality.This thesis focuses on the key issues that affect the imaging quality of RFs and aims to develop an effective RF processing workflow based on the seismic arrays,and improve the stability of the RF calculation and the imaging resolution by using the Radon transform(RT)in wavefield regularization.The research content and results of this thesis are summarized as follows:Our study explores the application of the high-resolution RT to improve the traditional RF imaging procedure.Different from the traditional strategy of noise suppression after deconvolution,we introduce RT into the data preprocessing step(i.e.,before deconvolution)to denoise the original vertical-and radial-component teleseismic wavefields.This method seeks a sparse solution to the teleseismic wavefield in the τ-p domain by iteratively solving a least-squares minimization problem through the conjugate gradient(CG)algorithm.The synthetic data test of the 2D Moho-step model shows that the RT can effectively remove the nonlinear seismic phase including incoherent noise and scattered energy in RFs,improving their signal-to-noise ratio(SNR)by about 8 d B.Synthetic tests with varying level of Gaussian noise show that the new processing workflow can achieve robust performance for SNR as low as 0.1.By introducing missing traces to the synthetic data,we futher explore the influence of spatial sampling on the denosing performance of the RT.The result shows that RT is not sensitive to the spatially irregular sampling of the seismic array recordings if local linearity is present in the data.Finally,we simulate the amplitude anomalies that may occur during the real data acquisition and demonstrate that the new workflow can effectively suppress noise through wavefield regularization and improve the stability of the RT calculation.This thesis then uses the real data recorded by the Hi-CLIMB network on the Tibetan Plateau to further explore the ability of the processing workflow to suppress noise and improve the accuracy of RF imaging.After the teleseismic wavefield preprocessing,the RFs of a single seismic event are migrated using the common-conversion point(CCP)stacking method.The result shows a clear Moho interface structure along the entire cross-section profile at the depth of about 60 km with some moderate fluctuations.The stacking profile obtained from more(11)seismic events clearly delineates the morphology of the Moho and intra-crustal interfaces.A pronounced double-Moho structure can be observed south of 30.5°and multiple dipping structures are identified in depth range of 30-40 km beneath the central part of the cross-section,which may indicate low-angle thrusting faults resulting from the collision of different tectonic blocks.The denoising workflow successfully recovers the weak converted phase from the upper mantle,delineating an apparent north-dipping structure in the depth range of 100–150 km.This may reveal the location of the mid-lithospheric discontinuity(MLD)in the underthrusting Indian plate.In conclusion,this study proposes a new strategy for RF processing of array data.By using linear RT to denoise the teleseismic wavefield,the random noise and amplitude anomalies are effectively suppressed and the stability of RF deconvolution is improved.This study emphasizes the necessity of wavefield regularization in RF processing and provide new insight into developing array processing and imaging methods in the future.