11:00 AM - 11:15 AM
[SIT16-02] Adaptive grid waveform inversion for the 3D S-wave velocity structure in D″ beneath the Southern Atlantic
Keywords:D″ region, waveform inversion, African LLSVP, lowermost mantle
The D″ region is the thermal boundary layer at the lowermost several hundred kilometers of the mantle and plays an essential role in mantle dynamics. In D″, two large low-shear-velocity provinces (LLSVPs) are known to exist: one beneath the Pacific and another beneath Africa. Since their discovery in the 1980s, there has been debate on whether thelow shear wave velocity is mainly due to heterogeneity in temperature or chemical composition. Mantle circulation simulations have found that if mantle convection is driven solely by the effect of temperature, gatherings of small thermal plumes called “plume clusters” will be created, while large masses of chemical heterogeneity referred to as “thermochemical piles” will be formed if chemical heterogeneity plays a role in driving the circulation. Although many previous studies have carried out whole-mantle inversions and envisaged the seismic structure of the lowermost mantle, their horizontal resolution of around 1000 km is insufficient to distinguish between thermochemical piles and plume clusters.
In this study, we utilize a large amount of broadband body-wave seismograms from arrays in Africa that have recently become available, as well as those from surrounding regions, and conduct waveform inversion for the 3D S-wave velocity structure of the western boundary region of the African LLSVP. The seismograms are obtained from the IRIS datacenter and include data of seismic waves from deep- and intermediate-focus earthquakes recorded at epicentral distances of 70-100 degrees. We use the radial and vertical components in addition to the transverse component, incorporating the SKS phase into the inversion along with the S and ScS phases. ~3600 waveforms for each component are used. We develop methods for adaptive grid inversion to stabilize the results at a resolution as high as possible, given the limited amount and raypath coverage of our data. This enables us to achieve a resolution of ~250 km horizontally and ~50 km vertically, allowing insights into the internal structure of the African LLSVP.
Our inferred model images high-velocity zones of laterally ~500 km scale below Brazil, as well as several small-scale low-velocity anomalies ~500 km in width just above the core-mantle boundary (CMB). The high-velocity structures are thought to be remnants of the Farallon slab, but the slab seems to be split into several parts by the time it reaches D″. We also find that the low-velocity zones composing the African LLSVP may be stretched further to the west in a 100-km-thick layer just above the CMB. Our model implies that the African LLSVP seems closer to the picture of plume clusters than that of thermochemical piles.
In this study, we utilize a large amount of broadband body-wave seismograms from arrays in Africa that have recently become available, as well as those from surrounding regions, and conduct waveform inversion for the 3D S-wave velocity structure of the western boundary region of the African LLSVP. The seismograms are obtained from the IRIS datacenter and include data of seismic waves from deep- and intermediate-focus earthquakes recorded at epicentral distances of 70-100 degrees. We use the radial and vertical components in addition to the transverse component, incorporating the SKS phase into the inversion along with the S and ScS phases. ~3600 waveforms for each component are used. We develop methods for adaptive grid inversion to stabilize the results at a resolution as high as possible, given the limited amount and raypath coverage of our data. This enables us to achieve a resolution of ~250 km horizontally and ~50 km vertically, allowing insights into the internal structure of the African LLSVP.
Our inferred model images high-velocity zones of laterally ~500 km scale below Brazil, as well as several small-scale low-velocity anomalies ~500 km in width just above the core-mantle boundary (CMB). The high-velocity structures are thought to be remnants of the Farallon slab, but the slab seems to be split into several parts by the time it reaches D″. We also find that the low-velocity zones composing the African LLSVP may be stretched further to the west in a 100-km-thick layer just above the CMB. Our model implies that the African LLSVP seems closer to the picture of plume clusters than that of thermochemical piles.