Mapping Upper Mantle Seismic Discontinuities Using Singular Spectrum Analysis

Abstract:

Seismic discontinuities are fundamental to the understanding of mantle composition and dynamics. Their depth and impedance are generally determined using secondary seismic phases, most commonly SS precursors and P-to-S converted waves. However, the analysis and interpretation using these approaches often suffer from incomplete data coverage, high noise levels and interfering seismic phases, especially near tectonically complex regions such as subduction zones and continental margins. To overcome these pitfalls, we apply Singular Spectrum Analysis (SSA) to remove random noise, reconstruct missing traces and enhance the robustness of SS precursors and P-to-S conversions from seismic discontinuities. Our method takes advantage of the predictability of time series in frequency-space domain and performs a rank reduction using a singular value decomposition of the trajectory matrix. We apply SSA to synthetic record sections as well as observations of 1) SS precursors beneath the northwestern Pacific subduction zones, and 2) P-to-S converted waves from the Western Canada Sedimentary Basin (WCSB). In comparison with raw or interpolated data, the SSA enhanced reflectivity maps show a greater resolution and a stronger negative correlation between the depths of the 410 and 660 km discontinuities. These effects can be attributed to the suppression of incoherent noise, which tends to reduce the signal amplitude during normal averaging procedures, through rank reduction and the emphasis of principle singular values. Our new results suggest a more laterally coherent 520 km reflection in the western Pacific regions. Similar improvements in data imaging are achieved in western Canada, where strong lateral variations in discontinuity topography are observed in the craton-Cordillera boundary zone. Improvements from SSA relative to conventional approaches are most notable in under-sampled regions.