We study the directional variation in travel-time residuals using 13,820 P-wave arrivals from 1,998 teleseismic events (15° ≤ Δ ≤ 98°, 4.1 ≤ mb ≤ 7.3) recorded in 1991-1997 by the Kyrgyz Digital Seismic Network (KNET). Based on a modified version of the iasp91 model that accounts for the Kyrgyz crustal thickness beneath KNET, we convert P-wave travel times to travel-time residuals δt. The dependence of δt on backazimuth is modeled as one-, two-, and four-lobed variations in a horizontal plane (Backus, 1965).
A least-squares fit of the azimuthal variation of δ>t indicates that the crust in the northern Tien Shan is about 11-15 km thicker than it is in the Kazakh Shield and the Chu Depression. From nine KNET stations, the one-lobe model estimates that the slowest P-wave travel-time direction is -5.0° ± 4.8° (almost directly north) and the magnitude of variation is 1.71 ± 0.13 sec. This result is consistent with an upwelling lower mantle plume. For the two-lobe model, the slowest P-wave travel-time directions (anisotropy term) are 89.7° and 269.7° ± 4.7 (i.e., trending east-west). We find P-wave velocity anisotropy of 2.0%-2.9% associated with a layer with a thickness of 440 km at the top of the lower mantle. The fast direction of the P-wave travel-time (north-south) azimuthal anisotropy at the top of the lower mantle is (1) parallel to the absolute motion of the India plate and (2) close to the direction of the upwelling hot mantle flow. The last result suggests that the azimuthal anisotropy of the travel-time residuals is due to the shape-preferred orientation of middle-mantle material that results from plume intrusion. Shear-wave splitting studies (Makeyeva et al., 1992; Wolfe and Vernon, 1998) estimated the fast polarization direction to be parallel to the strike of the geological structures of the northern Tien Shan (71° ± 29°). Thus, the fast polarization direction determined from these shear-wave splitting studies using KNET data contradicts (differs by > 90°) the fast travel-time direction (-0.3° and 179.7° ± 4.7°) we determine here using P-wave travel-time residuals using KNET data. This suggests that the azimuthal anisotropy determined from P-wave travel-time variations and from shear-wave splitting in SKS and SKKS have different sources.