Layer Stripping Forward Tomography with S, ScS, and Sdiff phases to sharpen images of deep mantle shear velocity heterogeneity
Abstract:The Earth’s deep mantle has been shown to be heterogeneous at a wide spectrum of scales, from fine scale (e.g.,1-10 km) to degree 2 low velocity provinces (1000’s of km lateral scale). Principle structures include large low shear velocity provinces (LLSVPs) beneath the Pacific Ocean and Africa and southern Atlanic Ocean, and a circum-Pacific higher velocity band. At intermediate scales, structures include D” discontinuities and details associated with LLSVPs, such as sharp sides. Shorter scales include ultra-low velocity zones (ULVZs) and regions of scattering of high frequency waves.
In this study we seek to improve upon our knowledge of global intermediate and small-scale shear wave structure in the lower mantle with an iterative Layer-Stripping Forward Tomography approach. This approach iteratively updates a starting tomographic model by mapping travel time residuals (between observations and predictions) from the surface (using shallow S wave phase) down to lowermost mantle depths (using deep S waves, ScS waves, and Sdiff waves). Modifications to the uppermost mantle are primarily aimed at properly correcting deep phases for shallow heterogeneity. Relatively uniform coverage of the deepest mantle is achieved with Sdiff. The core-reflected ScS contributes most strongly to our knowledge of shorter scale heterogeneity, but coverage is more limited that with Sdiff. Deep mantle S waves provide information on the vertical extent of lowermost mantle heterogeneity (e.g., LLSVPs), but similar to ScS, coverage is limited to source-receiver pairs with the proper distance range. We collected near 1200 global earthquakes global earthquakes from the last two decades, with depths greater than 30 km, and focused on data with clear and impulsive phases. In a semi-automatic scheme, we measured all S, ScS, and Sdiff times and document waveshape stability by constructing a mean shape of each phase for each event. Cross-correlation coefficients between individual records and the mean shape along with signal-to-noise ratios of each phase are used as weights during the iterative modeling approach. We will give attention to the strongest deep mantle heterogeneity, in both scale and amplitude, particularly LLSVPs whose general shapes are fairly well known, but their detailed structure remains enigmatic in many areas.