Studying Seismic Anisotropy In The Crust,Upper Mantle,and Lowermost Mantle Using Shear Wave Splitting

Seismic anisotropy is an important tool to study the internal structure,dynamic activities and stress state of the Earth.Shear wave splitting can be used to calculate the numerical characteristics of seismic anisotropy,including the fast direction and the delay time of slow wave.The characteristics of shear wave splitting in different layers of the Earth can be studied by using different seismic phases.For example,direct S wave can be used to study crustal anisotropy,which can reflect crustal structure and stress state.SKS and SKKS seismic phases can be used to study the anisotropy of the upper mantle,which can reflect structures,plate movement and other dynamic characteristics of the upper mantle.The anisotropy of core-mantle boundary can be predicted by the comparison of SKS and SKKS splitting results,which provides an important basis for the study of structures,composition,dynamic process of the lowermost mantle and the interaction with the outer core.In this paper,seismic anisotropy of crust(Alaska),upper mantle(continental United States)and core-mantle boundary(continental North America and the northeast Pacific Ocean)are studied by using shear wave splitting of the seismic phases mentioned above.Seismic anisotropy of core-mantle boundary is the key point of this paper,we creatively propose group splitting analysis method,the results of which show that D”anisotropy is characterized by small-scale variation and random distribution,and ascribe these observations to partial melting driven by chemical anomaly.In the study of crust anisotropy,direct S wave data recorded by Alaska seismic stations are used to calculate shear wave splitting parameters.Note that different from SKS(SKKS)shear wave splitting analysis,direct S waves have different focal depths,so their splitting results reflect anisotropy of different depth.Because of this,we normalize delay times to the value of per km.We obtain 27,751 splitting results,most of which are concentrated in the Alaskan subduction zone.Then the geographical distribution of fast direction and normalized delay time are analyzed,and several stations with relatively consistent fast direction are selected,which reflect the average direction of stress nearby them.The normalized delay time can reflect the strength of crustal stress to some extent.Then we analyze temporal changes of the normalized delay times in several specific areas and find that there are obviously regular variations before and after large earthquakes.Based on the seismic data of EarthScope stations,shear wave splitting database of SKS and SKKS phases are established,including 12160(10056 SKS,2104 SKKS)results with splitting data and 35836(29288 SKS,6548 SKKS)with no splitting data.Then the distribution characteristics of upper mantle anisotropy in the US continent are analyzed,and some stations conforming to one-layered anisotropy model are selected,which will be used for the subsequent upper mantle anisotropy correction.We constrain D”anisotropy beneath the North American continent and northeastern Pacific using two approaches:1)joint splitting analysis of SKS and SKKS phase pair for a common event,in which we obtain 158 pairs exhibiting discrepant splitting results and 791 pairs non-discrepant splitting results;and 2)group splitting analysis of SKS(or SKKS)phase from neighboring events recorded at a common station,in which we observe 109 2°x2° grids with consistent splitting parameters,and 164 grids with abrupt changes from splitting to no splitting within 30-100 km.The seismic data from both analyses indicate that small-scale variations of D”anisotropy are widespread beneath the studied regions,with a lateral scale up to tens of kilometers.For portion of the data recorded at the stations of simple upper mantle anisotropy,we correct for the effects of upper mantle anisotropy and obtain the splitting parameters of D”anisotropy.The inferred D”anisotropy exhibits a changing geographic pattern and lateral transition of anisotropy to a lateral scale of tens of kilometers.Such a length scale of changing anisotropy is also confirmed by synthetical modeling of the seismic data.We suggest that the inferred small-scale anisotropies could be best explained by the shape preferred orientation of widespread small-scale partial melt pockets derived by a composition change produced early in the Earth's history,a similar compositional origin that was invoked to explain the African Anomaly in the lower mantle.