Reconciling PP and P'P' precursor observations of a complex 660 km seismic discontinuity

High frequency precursors to P'P' almost invariably observe a narrow 660 km discontinuity, whereas PP precursor studies at long periods struggle to detect a reflection from the '660' despite its apparent sharpness to P'P'. To investigate these contradictory observations we compare PP and P'P' precursors in the same region. Using short period P'P' precursors we observe a sharp 660 km discontinuity, which appears to vary in depth substantially. The apparent topography on the '660' is too large to originate solely from thermal variations, regardless of its cause, therefore indicating chemical variations at the base of the mantle transition zone. Long period P'P' precursors show no '660' as they are sensitive to a larger area and thus average out the apparent topography, in agreement with long period PP precursors. Instead, we see some evidence in both long period data types for a reflection from 720 km depth, which is likely to correspond to a phase change in the garnet system.