letters to nature 310 NATURE | VOL 416 | 21 MARCH 2002 | www.nature.com ¯ow line, reaching a minimum of 0.7 cm yr -1 at Vostok station. Along the eastern shoreline a signi®cant increase in accretion rate to 2.9 cm yr -1 is necessary to produce the grounded accreted ice. It is possible that some of the accreted ice thickening along the eastern shoreline is the result of compressive ¯ow and not accretion. The lower layers in the ice sheet (3,300±2,800 m) undergo a 7% thickening, in contrast to the shallower layers of the ice sheet (2,400±1,900 m) which thin by 5±13% when grounding. The accretion process is most evident along the lake shorelines. In the grounded ice adjacent to the southeastern Lake Vostok shoreline, we observe an average of 295 m of accreted ice in 21 radar pro®les (Fig. 1, inset). No accretion ice is imaged adjacent to the eastern grounding line in the northern 15 radar pro®les. We estimate the ice ¯ux out of the lake, projected onto a 165-km line downslope of the lake, parallel to local contours. Assuming an average thickness of 295 m of accretion ice and a mean velocity of 3myr -1 , we estimate an annual accreted ice volume ¯ux of 0.146 km 3 yr -1 along this line. For a lake volume 14 of 1,800 km 3 , the residence time is 13,300 years. Previous estimates of residence time range from 4,500 years (ref. 6) to 125,000 years (ref. 1). The lower estimate (4,500 years), derived from the He 4 /He 3 ratio 6 , may be related to the formation of the accreted ice samples along the western shoreline in regions isolated from the main lake circula- tionÐthat is, the shallow embayment or the western grounding line. Deeper samples of accreted ice should be more representative of the open lake. In the southern portion of Lake Vostok, the interaction between the lake and the overlying ice sheet is dominated by accretion. Earlier evidence for melting in this region 4 was based on radar data oblique to the ice ¯ow ®eld, and is, we believe, incorrect. Ice ¯ow over southern Lake Vostok has a strong along-lake component, and has had a consistent orientation for the past 16,000 yearsÐassum- ing that the velocity has been constant over this time period. The accretion process dominates at the lake shorelines, although some accretion continues over the middle of the lake. The accretion ice is generally transported out of the lake along the southeastern lake margin. Most samples of accretion ice analysed to date are derived either from the shallow embayment or from the western grounding line, environments that are not representative of the open lake system. M Received 29 August 2001; accepted 11 February 2002. 1. Kapitsa, A. P.,Ridley, J. K., Robin, G. de Q., Siegert, M. J. & Zotikov, I. A. A large freshwater lake beneath the ice of central East Antarctica. Nature 381, 684±686 (1996). 2. Karl, D. M. et al. Microorganisms in the accreted ice of Lake Vostok, Antarctica. Science 286, 2144± 2147 (1999). 3. Priscu, J. C. et al. Geomicrobiology of subglacial ice above Lake Vostok, Antarctica. Science 286, 2141± 2144 (1999). 4. Siegert, M. J., Kwok, K., Mayer, C. & Hubbard, B. Water exchange between the subglacial Lake Vostok and the overlying ice sheet. Nature 403, 643±646 (2000). 5. Jouzel, J. et al. More than 200 meters of lake ice above subglacial Lake Vostok, Antarctica. Science 286, 2138±2141 (1999). 6. Jean-Baptiste, P., Petit, J.-R., Lipenkov, V. Y., Raynaud, D. & Barkov, N. I. Constraints on hydrothermal processes and water exchange in Lake Vostok from helium isotopes. Nature 411, 460±462 (2001). 7. Siegert, M. J. et al. Physical, chemical and biological processes in Lake Vostok and other Antarctic subglacial lakes. Nature 414, 603±609 (2001). 8. Popov, S. V.,Mironov, A. V. & Sheremetiev, A. N. Average of the electromagnetic wave propagation velocity in ice measurements in Vostok station vicinity. Materialy Glyatsiologicheskikh Issledovanii (Data of Glaciological Studies) 90, 206±208 (2001). 9. Popkov, A. M., Kudryavtsev, G. A., Verkulich, S. R., Masolov, V. N. & Lukin,V. V. in International Workshop on Lake Vostok Study: Scienti®c Objectives and Technological Requirements 26 (Arctic and Antarctic Research Institute, St Petersburg, Russia, 1998). 10. Whillans, I. M. Radio-echo layers and the recent stability of the West Antarctic ice sheet. Nature 264, 152±155 (1976). 11. Fujita, S. et al. Nature of radio echo layering in the Antarctic ice sheet detected by a two-frequency experiment. J. Geophys. Res. B 104, 13013±13024 (1999). 12. Kwok, R., Siegert, M. J. & Carsey, F. D. Ice motion over Lake Vostok, Antarctica: constraints on inferences regarding the accreted ice. J. Glaciol. 46, 689±694 (2000). 13. Carslaw, H. S. & Jaeger, J. C. Conduction of Heat in Solids 285 (Oxford Univ. Press, Oxford, 1959). 14. WuÈest, A. & Carmack, E. A priori estimates of mixing and circulation in the hard-to-reach water body of Lake Vostok. Ocean Model. 2, 29±43 (2000). 15. Jezek, K., Noltimier, K. & The RAMP Product Team. RAMP AMM-1 SAR Image Mosaic of Antarctica [digital media] (Alaska SAR Facility, Fairbanks; and the National Snow and Ice Data Center, Boulder, Colorado, 2001). Acknowledgements The ice-penetrating radar data were acquired by the US National Science Foundation's Support Of®ce for Aerogeophysical Research (SOAR) located at the University of Texas. We acknowledge the contributions of the Vostok ®eld team, including the SOAR team, the Kenn Borek ¯ight crews and the Raytheon East Camp crew. Comments from C. Bentley, R. Alley, E. Waddington and M. Siegert were appreciated. RADARSATsatellite data was provided by the Canadian Space Agency. This work was supported by the US National Science Foundation. Competing interests statement The authors declare that they have no competing ®nancial interests. Correspondence and requests for materials should be addressed to R.E.B. (e-mail: robinb@ldeo.columbia.edu). ................................................................. Development of anisotropic structure in the Earth's lower mantle by solid-state convection Allen K. McNamara*, Peter E. van Keken* & Shun-Ichiro Karato² * Department of Geological Sciences, University of Michigan, Ann Arbor, Michigan 48109-1063, USA ² Departmentof Geology and Geophysics, Yale University, New Haven, Connecticut 06520-8109, USA .............................................................................................................................................. Seismological observations reveal highly anisotropic patches at the bottom of the Earth's lower mantle, whereas the bulk of the mantle has been observed to be largely isotropic 1±4 . These patches have been interpreted to correspond to areas where subduction has taken place in the past or to areas where mantle plumes are upwelling, but the underlying cause for the anisotropy is unknownÐboth shape-preferred orientation of elastically hetero- genous materials 5 and lattice-preferred orientation of a homo- geneous material 6±8 have been proposed. Both of these mechanisms imply that large-strain deformation occurs within the anisotropic regions, but the geodynamic implications of the mechanisms differ. Shape-preferred orientation would imply the presence of large elastic (and hence chemical) heterogeneity whereas lattice-preferred orientation requires deformation at high stresses. Here we show, on the basis of numerical modelling incorporating mineral physics of elasticity and development of lattice-preferred orientation, that slab deformation in the deep lower mantle can account for the presence of strong anisotropy in the circum-Paci®c region. In this modelÐwhere development of the mineral fabric (the alignment of mineral grains) is caused solely by solid-state deformation of chemically homogeneous mantle materialÐanisotropy is caused by large-strain deforma- tion at high stresses, due to the collision of subducted slabs with the core±mantle boundary. Lattice-preferred orientation (LPO) leads to the development of a mineral fabric in material that deforms primarily by dislocation creep. Previous work 9 has shown that slabs cause high-stress regions in the lower mantle, which leads to localized regions of dislocation creep within a lower mantle dominated by diffusion creep under a wide range of rheological parameters. It has also been shown that hot, upwelling regions are low stress, and are therefore dominated by diffusion creep. The presence of dislocation creep is not in itself a © 2002 Macmillan Magazines Ltd