Late Cretaceous polar wander of the pacific plate: Evidence of a rapid true polar wander event

Science

Volumn 287, Issue 5452, 2000, Pages 455-459

  Sager, William W ^a   Koppers, Anthony A P ^b  

^a TEXAS A AND M UNIVERSITY   (United States)

^b UNIVERSITY OF CALIFORNIA   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

ARGON;

APPARENT POLAR WANDER PATH; CRETACEOUS; PALEOMAGNETISM; PLATE MOTION;

ARTICLE; CHRONOLOGY; GEOGRAPHY; MAGNETIC FIELD; MOTION; PRIORITY JOURNAL; SEA;

PACIFIC OCEAN;

EID: 0034695677     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.287.5452.455     Document Type: Article

References (71)

1. P. Goldreich and A. Toomre, J. Geophys. Res. 74, 1555 (1969).
2. C. G. A. Harrison and T. Lindh, Nature 300, 251 (1982).
3. R. A. Livermore, F. J. Vine, A. G. Smith, Geophys. J. R. Astron. Soc. 79, 939 (1984).
4. J. Besse and V. Courtillot, J. Geophys. Res. 96, 4029 (1991).
5. J. L. Kirschvink, R. L. Ripperdan, D. A. Evans, Science 277, 541 (1997).
6. P. Wessel and L. W. Kroenke, Nature 387, 365 (1997).
7. A. A. P. Koppers, thesis, Free University, Amsterdam (1998).
8. R. G. Gordon, Annu. Rev Earth Planet. Sci. 15, 567 (1987).
9. _, Geophys. Res. Lett. 10, 709 (1983).
10. W. Sager and M. S. Pringle, J. Geophys. Res. 93, 11753 (1988).
11. K. E. Petronotis and R. G. Gordon, Geophys. J. Int. 139, 227 (1999).
12. Pacific seamount paleopoles have been derived using two similar inversion methods that produce comparable pole estimates. Both solve for a magnetization that approximates the magnetic anomaly shape and amplitude given the measured seamount topography. The least squares method assumes a uniform magnetization and simply minimizes residuals between observed and calculated anomalies [for example, M. L. Richards, V. Vacquier, G. D. Van Voorhis, Geophysics 32, 678 (1967)]. The seminorm method calculates a maximally uniform solution consistent with a magnetization containing random inhomogeneities [R. L. Parker, L. Shure, J. A. Hildebrand, Rev. Geophys. 25, 17 (1987)]. The least squares method makes no error estimate, but a 95% confidence ellipse for the paleopole can be calculated with the seminorm method (41). However, our calculations indicate that these error bounds may underestimate the true error [see (15)].
13. Pacific seamount paleopoles have been derived using two similar inversion methods that produce comparable pole estimates. Both solve for a magnetization that approximates the magnetic anomaly shape and amplitude given the measured seamount topography. The least squares method assumes a uniform magnetization and simply minimizes residuals between observed and calculated anomalies [for example, M. L. Richards, V. Vacquier, G. D. Van Voorhis, Geophysics 32, 678 (1967)]. The seminorm method calculates a maximally uniform solution consistent with a magnetization containing random inhomogeneities [R. L. Parker, L. Shure, J. A. Hildebrand, Rev. Geophys. 25, 17 (1987)]. The least squares method makes no error estimate, but a 95% confidence ellipse for the paleopole can be calculated with the seminorm method (41). However, our calculations indicate that these error bounds may underestimate the true error [see (15)].
14. Most prior analyses used the goodness-of-fit ratio (GFR), the mean of observed anomaly values divided by the mean of residuals (observed minus calculated anomalies), as a reliability criterion, rejecting results with GFR < 2.0. The GFR is an inadequate criterion for seminorm models because this method matches the observed anomaly to arbitrarily high precision owing to its inclusion of inhomogeneities in the magnetization model. Consequently, there is no accepted reliability criterion for such data. We rejected data with GFR < 2.5 and 95% confidence ellipse major semiaxes > 14°. Both criteria are arbitrary, but their effect is to remove from consideration seamounts with complex magnetic anomalies that are poorly modeled by these methods and therefore may violate model assumptions.
15. 39Ar incremental heating techniques, the imprint of alteration can be monitored and high-quality ages can be calculated on the basis of undisturbed sections within age spectra (42, 43) [A. A. P. Koppers, H. Staudigel, J. R. Wijbrans, Chem. Geol., in press].
16. 39Ar incremental heating techniques, the imprint of alteration can be monitored and high-quality ages can be calculated on the basis of undisturbed sections within age spectra (42, 43) [A. A. P. Koppers, H. Staudigel, J. R. Wijbrans, Chem. Geol., in press].
17. We used simple Fisher statistics to calculate mean paleomagnetic poles and 95% confidence circles. The poles were divided into groups on the basis of clustering of dates and pole positions (Web fig. 1 at Science Online, www.sciencemag.org/feature/data/ 1042962.shl). We tried using a moving window in time, as is done in some APWP calculations, but most windows were empty when small window widths were used. Only the 82-to 86-Ma group had to be separated by pole position because it was obvious that the poles of that age group contained far more scatter than any other. More elaborate schemes for pole and error calculations were rejected because seamount paleomagnetic pole errors are poorly quantified. Because some seamount poles may be more accurate than others, it may be desirable to weight poles by their error estimates [for example, R. G. Gordon and A. Cox, Geophys. J. R. Astron. Soc. 63, 619 (1980)]. However, confidence limits are not routinely calculated with the least squares inversion technique. Additionally, although a routine for calculating errors in seminorm inversion data has been developed (47), its reliability has been questioned (25). Our calculations using seminorm poles and confidence regions with the error-propagation method of Gordon and Cox indicate that estimated seminorm method-derived confidence ellipses are inconsistent with observed misfits between individual and mean paleomagnetic poles. Moreover, the implication is that the error estimates are too small. We tried to find a relation between pole misfit and GFR or the size of the seminorm 95% confidence ellipse major semiaxis, but the correlation was poor. Given the unknowns in pole uncertainties and the small amount of data used to calculate each mean pole (Table 2), the Fisher approximation seems adequate.
18. We used simple Fisher statistics to calculate mean paleomagnetic poles and 95% confidence circles. The poles were divided into groups on the basis of clustering of dates and pole positions (Web fig. 1 at Science Online, www.sciencemag.org/feature/data/ 1042962.shl). We tried using a moving window in time, as is done in some APWP calculations, but most windows were empty when small window widths were used. Only the 82-to 86-Ma group had to be separated by pole position because it was obvious that the poles of that age group contained far more scatter than any other. More elaborate schemes for pole and error calculations were rejected because seamount paleomagnetic pole errors are poorly quantified. Because some seamount poles may be more accurate than others, it may be desirable to weight poles by their error estimates [for example, R. G. Gordon and A. Cox, Geophys. J. R. Astron. Soc. 63, 619 (1980)]. However, confidence limits are not routinely calculated with the least squares inversion technique. Additionally, although a routine for calculating errors in seminorm inversion data has been developed (47), its reliability has been questioned (25). Our calculations using seminorm poles and confidence regions with the error-propagation method of Gordon and Cox indicate that estimated seminorm method-derived confidence ellipses are inconsistent with observed misfits between individual and mean paleomagnetic poles. Moreover, the implication is that the error estimates are too small. We tried to find a relation between pole misfit and GFR or the size of the seminorm 95% confidence ellipse major semiaxis, but the correlation was poor. Given the unknowns in pole uncertainties and the small amount of data used to calculate each mean pole (Table 2), the Fisher approximation seems adequate.
19. Mean pole ages can be calculated from multiple seamount ages using averages inversely weighted by analytical errors. However, this produces unrealistically small estimates of the standard deviation. This behavior reflects the fact that analytical errors for individual basalt ages are significantly smaller than the age range for typical seamount shield-building volcanism, which may last for as long as 5 to 10 My (42) [A. Abdel-Monem and P. W. Gast, Earth Planet. Sci. Lett. 2, 415 (1967); D. A. Clague et al., in The Geology of North America, Vol. N, The Eastern Pacific and Hawaii, E. L. Winterer, D. M. Hussong, R. W. Decker, Eds. (Geological Society of America, Boulder, CO, 1989), pp. 187-287; M. S. Pringle, H. Staudigel, J. Gee, Geology 19, 363 (1991)]. The processes causing such prolonged seamount volcanism are still poorly understood, but they seem to be a function of tectonic setting [D. Epp, J. Geophys. Res. 89, 11273 (1984); J. P. Morgan, W. J. Morgan, E. Price, J. Geophys. Res. 100, 8045 (1995); P. Wessel, Science 277, 802 (1997)]. For these reasons, calculating weighted averages based on analytical errors may significantly underestimate the actual geological error on average seamount ages and, by inference, on the mean magnetic pole ages.
20. Mean pole ages can be calculated from multiple seamount ages using averages inversely weighted by analytical errors. However, this produces unrealistically small estimates of the standard deviation. This behavior reflects the fact that analytical errors for individual basalt ages are significantly smaller than the age range for typical seamount shield-building volcanism, which may last for as long as 5 to 10 My (42) [A. Abdel-Monem and P. W. Gast, Earth Planet. Sci. Lett. 2, 415 (1967); D. A. Clague et al., in The Geology of North America, Vol. N, The Eastern Pacific and Hawaii, E. L. Winterer, D. M. Hussong, R. W. Decker, Eds. (Geological Society of America, Boulder, CO, 1989), pp. 187-287; M. S. Pringle, H. Staudigel, J. Gee, Geology 19, 363 (1991)]. The processes causing such prolonged seamount volcanism are still poorly understood, but they seem to be a function of tectonic setting [D. Epp, J. Geophys. Res. 89, 11273 (1984); J. P. Morgan, W. J. Morgan, E. Price, J. Geophys. Res. 100, 8045 (1995); P. Wessel, Science 277, 802 (1997)]. For these reasons, calculating weighted averages based on analytical errors may significantly underestimate the actual geological error on average seamount ages and, by inference, on the mean magnetic pole ages.
21. Mean pole ages can be calculated from multiple seamount ages using averages inversely weighted by analytical errors. However, this produces unrealistically small estimates of the standard deviation. This behavior reflects the fact that analytical errors for individual basalt ages are significantly smaller than the age range for typical seamount shield-building volcanism, which may last for as long as 5 to 10 My (42) [A. Abdel-Monem and P. W. Gast, Earth Planet. Sci. Lett. 2, 415 (1967); D. A. Clague et al., in The Geology of North America, Vol. N, The Eastern Pacific and Hawaii, E. L. Winterer, D. M. Hussong, R. W. Decker, Eds. (Geological Society of America, Boulder, CO, 1989), pp. 187-287; M. S. Pringle, H. Staudigel, J. Gee, Geology 19, 363 (1991)]. The processes causing such prolonged seamount volcanism are still poorly understood, but they seem to be a function of tectonic setting [D. Epp, J. Geophys. Res. 89, 11273 (1984); J. P. Morgan, W. J. Morgan, E. Price, J. Geophys. Res. 100, 8045 (1995); P. Wessel, Science 277, 802 (1997)]. For these reasons, calculating weighted averages based on analytical errors may significantly underestimate the actual geological error on average seamount ages and, by inference, on the mean magnetic pole ages.
22. Mean pole ages can be calculated from multiple seamount ages using averages inversely weighted by analytical errors. However, this produces unrealistically small estimates of the standard deviation. This behavior reflects the fact that analytical errors for individual basalt ages are significantly smaller than the age range for typical seamount shield-building volcanism, which may last for as long as 5 to 10 My (42) [A. Abdel-Monem and P. W. Gast, Earth Planet. Sci. Lett. 2, 415 (1967); D. A. Clague et al., in The Geology of North America, Vol. N, The Eastern Pacific and Hawaii, E. L. Winterer, D. M. Hussong, R. W. Decker, Eds. (Geological Society of America, Boulder, CO, 1989), pp. 187-287; M. S. Pringle, H. Staudigel, J. Gee, Geology 19, 363 (1991)]. The processes causing such prolonged seamount volcanism are still poorly understood, but they seem to be a function of tectonic setting [D. Epp, J. Geophys. Res. 89, 11273 (1984); J. P. Morgan, W. J. Morgan, E. Price, J. Geophys. Res. 100, 8045 (1995); P. Wessel, Science 277, 802 (1997)]. For these reasons, calculating weighted averages based on analytical errors may significantly underestimate the actual geological error on average seamount ages and, by inference, on the mean magnetic pole ages.
23. Mean pole ages can be calculated from multiple seamount ages using averages inversely weighted by analytical errors. However, this produces unrealistically small estimates of the standard deviation. This behavior reflects the fact that analytical errors for individual basalt ages are significantly smaller than the age range for typical seamount shield-building volcanism, which may last for as long as 5 to 10 My (42) [A. Abdel-Monem and P. W. Gast, Earth Planet. Sci. Lett. 2, 415 (1967); D. A. Clague et al., in The Geology of North America, Vol. N, The Eastern Pacific and Hawaii, E. L. Winterer, D. M. Hussong, R. W. Decker, Eds. (Geological Society of America, Boulder, CO, 1989), pp. 187-287; M. S. Pringle, H. Staudigel, J. Gee, Geology 19, 363 (1991)]. The processes causing such prolonged seamount volcanism are still poorly understood, but they seem to be a function of tectonic setting [D. Epp, J. Geophys. Res. 89, 11273 (1984); J. P. Morgan, W. J. Morgan, E. Price, J. Geophys. Res. 100, 8045 (1995); P. Wessel, Science 277, 802 (1997)]. For these reasons, calculating weighted averages based on analytical errors may significantly underestimate the actual geological error on average seamount ages and, by inference, on the mean magnetic pole ages.
24. Mean pole ages can be calculated from multiple seamount ages using averages inversely weighted by analytical errors. However, this produces unrealistically small estimates of the standard deviation. This behavior reflects the fact that analytical errors for individual basalt ages are significantly smaller than the age range for typical seamount shield-building volcanism, which may last for as long as 5 to 10 My (42) [A. Abdel-Monem and P. W. Gast, Earth Planet. Sci. Lett. 2, 415 (1967); D. A. Clague et al., in The Geology of North America, Vol. N, The Eastern Pacific and Hawaii, E. L. Winterer, D. M. Hussong, R. W. Decker, Eds. (Geological Society of America, Boulder, CO, 1989), pp. 187-287; M. S. Pringle, H. Staudigel, J. Gee, Geology 19, 363 (1991)]. The processes causing such prolonged seamount volcanism are still poorly understood, but they seem to be a function of tectonic setting [D. Epp, J. Geophys. Res. 89, 11273 (1984); J. P. Morgan, W. J. Morgan, E. Price, J. Geophys. Res. 100, 8045 (1995); P. Wessel, Science 277, 802 (1997)]. For these reasons, calculating weighted averages based on analytical errors may significantly underestimate the actual geological error on average seamount ages and, by inference, on the mean magnetic pole ages.
25. W. Sager and M. S. Pringle, in Seamounts, Islands, and Atolls, B. Keating, P. Fryer, R. Batiza, G. Boehlert, Eds. (American Geophysical Union, Washington, DC, 1987), pp. 133-162.
26. J. A. Tarduno and R. D. Cottrell, Earth Planet. Sci. Lett. 153, 171 (1998).
27. J. A. Tarduno and J. Gee, Nature 378, 477 (1995).
28. S. M. Vasas, K. E. Petronotis, R. G. Gordon, EOS Trans. Am. Geophys. Union 75, 203 (1994).
29. Four seamount poles come from seamounts that have reversed magnetic polarity that has been used to infer formation during Chron 33r. The initial argument was applied to three seamounts (Kona 55, Show, Chatauqua) by Gordon (9). His logic was that the seamounts are reversed in polarity and have paleomagnetic poles well to the south of latest Cretaceous poles. Because Chron 33r at 79 to 83 Ma [S. C. Cande and D. V. Kent, J. Geophys. Res. 97, 13917 (1992)] was the only long reversed-polarity period before the latest Cretaceous (the next significant reversed period-is Chron 31r at 70 to 72 Ma), these seamounts may have formed during Chron 33r. Sager and Pringle (10) added seamount C6 and noted that this seamount was normal with a reversed-polarity top, so it probably formed at the beginning of Chron 33r; they also noted that Chatauqua is reversed-polarity with a normal top, so it may have formed at the end of Chron 33r. Although these inferences were helpful at a time when few reliable radiometric dates were available, we decided it was inappropriate to insert these data into an otherwise well-dated data set. They are mentioned here because the location of three of these seamount poles is in the gap between older and younger Cretaceous poles and suggests, albeit weakly, that the polar shift occurred coincident with Chron 33r.
30. J. Gee, H. Staudigel, L. Tauxe, T. Pick, Y. Gallet, J. Geophys. Res. 98, 11743 (1993).
31. J. Gee and M. Nakanishi, Proc. Ocean Drill. Program Sci. Results 144, 615 (1995).
32. The magnetization calculated using seamount anomaly inversion is a combination of the original magnetization acquired during cooling of the seamount basalts along with any secondary magnetization superimposed by other geologic factors. Of particular importance are induced magnetization, caused by the present-day geomagnetic field, and viscous magnetization, resulting from reorientation of the magnetization with time. Both of these are usually removed from standard paleomagnetic samples by alternating field or thermal demagnetization and measurement in a field-free space. Although many seamount basalts have stable magnetic directions and remanent magnetizations that are much larger than their induced counterparts, one might expect both induced and viscous magnetization bias to affect seamount paleomagnetic poles (22, 23). We cannot completely investigate this effect with seamount samples because dredged and drilled rocks typically represent only a small fraction of the seamount exterior, and for most seamounts even these samples do not exist. Studies of seamount rocks (22, 23) indicate that the induced magnetization represents 10 to 25% of the total magnetization. Because we know that most Pacific seamounts formed 20° to 30° south of their present positions, we can calculate that this degree of bias would cause a 2° to 6° shift in pole position for normally magnetized seamounts. This shift is small because of the small angle between the present-day and paleofield directions. The shift can be larger for reversely polarized seamounts because the induced and viscous magnetizations are in the opposite direction from the original magnetization. We know in what direction induced and viscous bias will move a seamount pole. With this bias, the pole of a normally polarized seamount will move from the zero-bias position toward the geomagnetic pole (in northern Greenland). The pole of a reversely magnetized seamount will move in the opposite direction until the bias is much greater than the original magnetization. Our argument that the polar shift is not caused by induced or viscous bias is based on the fact that the shift is nearly perpendicular to the direction of pole shift expected from this bias.
33. R. L. Parker, J. Geophys. Res. 96, 16101 (1990).
34. 63 angle of 5.6° (this is the radius of the cone within which 63% of poles fall relative to the mean). The scatter in an east-west direction seems somewhat larger than that in the north-south direction (Web fig. 2 at Science Online, www. sciencemag.org/feature/data/1042962.shl) because a given error in inclination produces only about half the error in pole space, as does an equivalent error in declination. A better calculation of mean pole locations and errors would take this anisotropy into account; however, all the mean poles were determined with small numbers of poles with uncertainties that are poorly quantified, so more sophisticated analysis may not be entirely appropriate and would not change the results significantly.
35. P. L. McFadden and F. J. Lowes, Geophys. J. R. Astron. Soc. 67, 19 (1981).
36. To compare the directions of APWP segments and to identify similarities, we needed to remove the effects of plate motions. One way to do this is to remove motions relative to the hot spots Although there is still considerable debate about the fixity of hot spots in the mantle and whether they constitute a reliable and accurate reference frame, many authors have used and continue to use motions of plates relative to the hot spots for tectonic studies. Of all the plates, the motion of the Pacific plate relative to the hot spots is probably the best known (6, 7), particularly for the period represented by the Hawaiian-Emperor chain (81 Ma to present). There is a discrepancy when trying to compare results from Pacific hot spots and those in the Indian and Atlantic oceans [G. D. Acton and R. G. Gordon, Science 263, 1246 (1994)], so we reconstructed Pacific data to the Pacific hot spots and other APWPs using Atlantic and Indian hot spots, with the assumption that the Pacific hot spots are fixed relative to those elsewhere. Pacific data were backtracked using the model of Koppers (7). Because this model has taken into account numerous Cretaceous seamount dates and several western Pacific seamount chains, it represents a significant improvement on prior models of Early Cretaceous Pacific plate motion. The two global synthetic APWPs (3, 4) were backtracked by their authors relative to the Atlantic and Indian hot spots. Although the details of the reconstructions were somewhat different, both are based on Morgan's models [W. J. Morgan, Tectonophysics 94, 123 (1983)] and so the results are comparable. Given that our comparison was only for major trends, we did not redo these reconstructions. For the North America data, we backtracked using a more recent plate-hot spot model (44). The difference caused by using this updated plate motion model should be small.
37. To compare the directions of APWP segments and to identify similarities, we needed to remove the effects of plate motions. One way to do this is to remove motions relative to the hot spots Although there is still considerable debate about the fixity of hot spots in the mantle and whether they constitute a reliable and accurate reference frame, many authors have used and continue to use motions of plates relative to the hot spots for tectonic studies. Of all the plates, the motion of the Pacific plate relative to the hot spots is probably the best known (6, 7), particularly for the period represented by the Hawaiian-Emperor chain (81 Ma to present). There is a discrepancy when trying to compare results from Pacific hot spots and those in the Indian and Atlantic oceans [G. D. Acton and R. G. Gordon, Science 263, 1246 (1994)], so we reconstructed Pacific data to the Pacific hot spots and other APWPs using Atlantic and Indian hot spots, with the assumption that the Pacific hot spots are fixed relative to those elsewhere. Pacific data were backtracked using the model of Koppers (7). Because this model has taken into account numerous Cretaceous seamount dates and several western Pacific seamount chains, it represents a significant improvement on prior models of Early Cretaceous Pacific plate motion. The two global synthetic APWPs (3, 4) were backtracked by their authors relative to the Atlantic and Indian hot spots. Although the details of the reconstructions were somewhat different, both are based on Morgan's models [W. J. Morgan, Tectonophysics 94, 123 (1983)] and so the results are comparable. Given that our comparison was only for major trends, we did not redo these reconstructions. For the North America data, we backtracked using a more recent plate-hot spot model (44). The difference caused by using this updated plate motion model should be small.
38. J. F. Diehl, J. Geophys. Res. 96, 9887 (1991).
39. Figure 3 shows the sense of motion needed to explain the pole shift at 84 Ma. In the central Pacific, this appears as a counterclockwise rotation. A microplate rotation (17) was a possible explanation as long as the poles that did not seem to fit the APWPs (i.e., those of the 84E group) were all in the Musicians and South Hawaiian seamounts (the latter group is near the present-day Hawaiian Islands but not formed by the Hawaiian hot spot). However, even this initial explanation was based on the presumption of two separate microplates rotating the same amount (a fortuitous occurrence). The 84E group contains two seamounts (H11 and Nagata) that are located more than 1500 km from the purported microplates. Thus, the circumstances that would cause all of these seamounts to rotate the same amount are even more unlikely. Could the shift be attributed to a rapid rotation of the Pacific plate as a whole? This seems unlikely for two reasons. First, by Late Cretaceous time, the Pacific plate was large and had slabs engaged along most of its northern and western margins [T. W. C. Hilde, S. Uyeda, L. Kroenke, Tectonophysics 39, 145 (1977)]. A large rotation would have required those subduction zones to adopt strike-slip motions and drag slabs laterally through the mantle - an implausible explanation. Second, this rotation would have created highly arcuate seamount chains or chains with a younging-westward trend (as opposed to the observed younging-eastward trends), but no chains have been observed to fit this description. Consequently, there is no evidence suggesting a large rotation of the Pacific plate in its entirety.
40. T. Atwater, in The Geology of North America, Vol. N, The Eastern Pacific and Hawaii, E. L. Winterer, D. M. Hussong, R. W. Decker, Eds. (Geological Society of America, Boulder, CO, 1989), pp. 21-72.
41. K. D. Klitgord and H. Schouten, in The Geology of North America, Vol. M, the Western North Atlantic Region, P. R. Vogt and R. E. Tucholke, Eds. (Geological Society of America, Boulder, CO, 1986), pp. 351-378.
42. J.-Y. Royer, P. Patriat, H. W. Bergh, C. R. Scotese, Tectonophysics 155, 235 (1988).
43. M. S. Pringle, EOS Trans. Am. Geophys. Union 78 (fall suppl.), F728 (1997).
44. M. L. G. Tejada, J. J. Mahoney, R. A. Duncan, M. P. Hawkins, J. Petrol. 37, 361 (1996).
45. A. C. Kerr, J. Tarney, G. F. Mariner, A. Nivia, A. D. Saunders, in Large Igneous Provinces: Continental, Oceanic, and Planetary Flood Volcanism, J. J. Mahoney and M. F. Coffin, Eds. (American Geophysical Union, Washington, DC, 1997), pp. 123-144.
46. M. Stein and A. W. Hoffman, Nature 372, 63 (1994).
47. Figure 3 indicates that the rotation implied by the 84-Ma polar shift would cause the maximum shift in pateoposition in the Atlantic and Indian oceans. The motions of Atlantic-bordering continents are well known since Cretaceous time, so we calculated the expected latitude shift for two sites in the Atlantic realm, one on North America (Washington, D.C.) and one on Africa (Dakar). First, we backtracked the inferred TPW rotation pole in Pacific coordinates (24.1°N, 195.8°E) into the hot spot reference frame (-6.1°N, 228.2°E) using the plate-hot spot motion model of Koppers (7). Washington and Dakar were backtracked into the hot spot reference frame using a model for motions of the Atlantic-bordering plates relative to the mantle (44). Finally, Washington and Dakar were rotated using the TPW motion pole in the hot spot reference frame. Because these two sites were located nearly 90° from the rotation pole, they moved nearly the full amount of rotation, mostly in paleolatitude.
48. W. Sager and U. Bleil, Nature 326, 488 (1987).
49. R. A. Duncan and M. Storey, in Synthesis of Results from Scientific Drilling in the Indian Ocean, R. A. Duncan, D. K. Rea, R. B. Kidd, U. von Rad, J. K. Weissel, Eds. (American Geophysical Union, Washington, DC, 1992), pp. 91-103.
50. R. L. Parker, J. Geophys. Res. 93, 3105 (1988).
51. A. A. P. Koppers, H. Staudigel, J. R. Wijbrans, M. S. Pringle, Earth Planet. Sci. Lett. 163, 53 (1998).
52. M. S. Pringle, in The Mesozoic Pacific: Geology, Tectonics, and Volcanism, M. S. Pringle, W. W. Sager, W. V. Sliter, S. Stein, Eds. (American Geophysical Union, Washington, DC, 1993), pp. 187-215.
53. R. D. Müller, J.-Y. Royer, L. A. Lawver, Geology 21, 275 (1993).
54. W. W. Sager, J. Geophys. Res. 89, 6271 (1984).
55. _ and B. H. Keating, J. Geophys. Res. 89, 11135 (1984).
56. W. W. Sager, in Geology and Offshore Mineral Resources of the Central Pacific Basin, B. Keating and B. Bolton, Eds. (Springer-Verlag, New York, 1992), pp. 21-37.
57. C. G. A. Harrison, R. D. Jarrard, V. Vacquier, R. L. Larson, Geophys. J. R. Astron. Soc. 42, 859 (1975).
58. J. A. Hildebrand and R. L. Parker, J. Geophys. Res. 92, 12695 (1987).
59. W. W. Sager, R. A. Duncan, D. W. Handschumacher, in The Mesozoic Pacific: Geology, Tectonics, and Volcanism, M. S. Pringle, W. W. Sager, W. V. Sliter, S. Stein, Eds. (American Geophysical Union, Washington, DC, 1993), pp. 401-435.
60. W. W. Sager, thesis, University of Hawaii (1983).
61. P. Bryan, T. Shoberg, R. G. Gordon, K. E. Petronotis, D. D. Bergersen, in The Mesozoic Pacific: Geology, Tectonics, and Volcanism, M. S. Pringle, W. W. Sager, W. V. Sliter, S. Stein, Eds. (American Geophysical Union, Washington, DC, 1993). pp. 387-400.
62. Y. Ueda, J. Geomagn. Geoelectr. 49, 1481 (1985).
63. J. Francheteau, C. G. A. Harrison, J. G. Sclater, M. L. Richards, J. Geophys. Res. 75, 2035 (1970).
64. R. A. Duncan and D. A. Clague, EOS Trans. Am. Geophys. Union 65, 1076 (1984).
65. S. O. Schlanger et al., J. Geophys. Res. 89, 11261 (1984).
66. K. Saito and M. Ozima, Earth Planet. Sci. Lett. 33, 353 (1977).
67. M. S. Pringle and G. B. Dalrymple, in The Mesozoic Pacific: Geology, Tectonics, and Volcanism, M. S. Pringle, W. W. Sager, W. V. Sliter, S. Stein, Eds. (American Geophysical Union, Washington, DC, 1993), pp. 263-277.
68. E. L. Winterer et al., in The Mesozoic Pacific: Geology, Tectonics, and Volcanism, M. S. Pringle, W. W. Sager, W. V. Sliter, S. Stein, Eds. (American Geophysical Union, Washington, DC, 1993), pp. 307-334.
69. M. S. Pringle and R. A. Duncan, Proc. Ocean Drill. Program Sci. Results 144, 547 (1995).
70. Y. Takigami, I. Kaneoka, T. Ishii, Y. Nakamura, Palaeogeogr. Palaeoclimatol. Palaeoecol. 71, 71 (1989).
71. We thank R. Gordon and an anonymous reviewer for constructive comments. Supported by NSF grants OCE98-11326 ( W.W.S.) and OCE97-30394 (A.A.P.K.) and a TALENT-NATO grant by the Netherlands Science Foundation (NWO) (A.A.P.K.).