Historical Interstation Pattern Referencing (HIPR): An Application to PcP Waves Recorded in the Antarctic for ULVZ Imaging

Much of our knowledge on deep Earth structure is based on detailed analyses of seismic waveforms that often have small amplitude arrivals on seismograms; therefore, stacking is essential to obtain reliable signals above the noise level. We present a new iterative stacking scheme that incorporates Historical Interstation Pattern Referencing (HIPR) to improve data quality assessment. HIPR involves comparing travel‐time and data quality measurements between every station for every recorded event to establish historical patterns, which are then compared to individual measurements. Weights are determined based on the individual interstation measurement differences and their similarity to historical averages, and these weights are then used in our stacking algorithm. This approach not only refines the stacks made from high‐quality data but also allows some lower‐quality events that may have been dismissed with more traditional stacking approaches to contribute to our study. Our HIPR‐based stacking routine is illustrated through an application to core‐reflected PcP phases recorded by the Transantarctic Mountains Northern Network to investigate ultra‐low velocity zones (ULVZs). We focus on ULVZ structure to the east of New Zealand because this region is well‐sampled by our data set and also coincides with the boundary of the Pacific Large Low Shear Velocity Province (LLSVP), thereby allowing us to further assess possible ULVZ‐LLSVP relationships. The HIPR‐refined stacks display strong ULVZ evidence, and associated synthetic modeling suggests that the ULVZs in this region are likely associated with compositionally distinct material that has perhaps been swept by mantle convection currents to accumulate along the LLSVP boundary. Plain Language Summary Our understanding of deep Earth structure is largely based on studies that use small amplitude, earthquake‐generated signals that have interacted with structures near the Earth's core‐mantle boundary (CMB). Multiple seismic records are typically averaged (stacked) to improve the signal‐to‐noise ratio of these signals; however, the stacked signals can still be challenging to interpret if they are based on low quality data. Here, we develop a new iterative stacking approach that compares travel‐time and data quality measurements between all stations and all recorded events to develop historical patterns, which are then compared to individual measurements. Event and station weights based on the interstation measurement di