BACKGROUND AND RECENT RESULTS

Paleomagnetic data from sediments are available from numerous Pacific Deep Sea Drilling Project (DSDP) and ODP sites. These data can be examined to test whether they define an internally consistent picture of plate motion (Fig. 3). Unfortunately, the data are systematically shallower than coeval inclination values derived from igneous sources (e.g., Gordon, 1990). The differences are consistent with compaction-induced inclination shallowing (Tarduno, 1990; Butler, 1992). Sediments can acquire a detrital magnetization when magnetic grains orient in the presence of the Earth's geomagnetic field. A flattening of inclination, however, is well documented for certain sediment types, especially glacial varves. The error is described as

It was once thought that the deep-sea sediments might escape the effects of inclination error because their magnetization was thought to be a post-depositional remanent magnetization (pDRM), which was locked in not at the sediment-water interface, but slightly deeper, within an interval where magnetic grains were still free to rotate in pore spaces (Verosub, 1977; deMenocal et al., 1990). Nevertheless, observations from all the world's oceans appeared to define shallow inclinations in some deep-sea sediments (e.g., Celaya and Clement, 1988; Arason and Levi, 1990; Larson et al., 1992) and subsequent experimental results on synthetic samples have confirmed that a pDRM can be shallowed by compaction (Kodama and Sun, 1992; Sun and Kodama, 1992).

The Pacific sedimentary inclination data studies by Tarduno (1990), which are mainly Cretaceous in age, define a f = 0.52. Considerable scatter in this value is not surprising: sediments ranging from carbonates to volcaniclastics were combined and sediment type clearly must play a role in compaction. This calculation also relies on expected inclinations calculated from the sparse, ca. 1990 igneous data set (e.g., basalt colatitudes). Results from subsequent drilling have tended to confirm previous notions of mid-Cretaceous pole positions (Tarduno and Sager, 1995; Tarduno and Gee, 1995). In addition, whereas errors in the reference data (poles) can affect the degree of flattening, the overall distribution remains internally inconsistent and requires inclination shallowing. Any remaining doubt about the mechanism was put to rest by the magnetic anisotropy work by Hodych and Bijaksana (1993) on the same sites studied in Tarduno (1990). This work documented that the development of a magnetic fabric in Pacific deep-sea sediments was proportional to the flattening values predicted in Tarduno (1990).

The potential problems caused by inclination shallowing are particularly severe in our proposed study for several reasons. The amount of error varies with expected latitude. Because the expected latitude values in the moving hotspot hypothesis range between 40° and 60°, the errors will be at their maximum potential values for flattening values similar to those reported in natural sediments (Fig. 4A). These errors are of the same magnitude as the total inclination offset we seek to test (Fig. 4B). Whereas substantial advances have been made in using magnetic anisotropy to correct for inclination shallowing (e.g., Jackson et al., 1991), the potential errors are, in our opinion, too large for this to yield an unambiguous, high-resolution test of the fixed vs. moving hotspot models.

We note, however, that a few sites in the Pacific, particularly shallow-water sediments (Tarduno and Gee, 1995) have yielded data that do not appear to have suffered significant inclination shallowing. This may be due to relatively early cementation, and similar sediments could be encountered during drilling. However, because it can be difficult to obtain high-resolution age data on some shallow water sediments and recovery is problematic, they do not provide a viable alternative to basalt drilling.

New Paleolatitude Data for the Emperor Seamounts
During Leg 145, 87 m of lava flows were penetrated on Detroit Seamount (Fig. 1) (Rea, Basov, Scholl, and Allan, 1995). ⁴⁰Ar/³⁹Ar radiometric analyses yield an age (81.2 ± 1.3 Ma [Keller et al., 1995]) older than that assumed in hotspot-based plate-motion models (~75 Ma) (Duncan and Clague, 1985). Characteristic magnetizations derived from basalt samples have mainly negative inclinations indicating reversed polarity (Fig. 1). This polarity assignment is consistent with the radiometric age data, suggesting eruption of the basalts during Chron 33R (79-83 Ma) (Tarduno and Cottrell, 1997).

A potential problem in obtaining reliable paleomagnetic data from any basalt drill hole is the uncertain timescale between eruptions. If most flows reflect rapid eruptions, one could easily obtain a biased paleolatitude estimate by giving equal weight to each flow unit. To address this concern, the inclination-only averages derived from each flow unit (McFadden and Reid, 1982) must be checked for serial correlation (Cox, 1970; Kono, 1980; Tarduno and Sager, 1995). These analyses lead to inclination-group models (Fig. 5). The directional angular dispersion, estimated from the inclination-model data and transformed into pole space (Cox, 1970; Tarduno and Sager, 1995), is indistinguishable from the predicted virtual geomagnetic pole scatter from global data sets (McFadden et al., 1991) (Fig. 5). As discussed below, there is only one other paleomagnetic data set from the Emperor trend that satisfies these geomagnetic-sampling requirements.

The preferred inclination group model, where groups are distinct at >95% confidence (N = 10) (Kono, 1980), suggests a paleolatitude of 36.2° (+6.9°/-7.2°), clearly discordant from the present day latitude of Hawaii (~19°) (Fig. 5). This discrepancy is too large to be explained by tectonic tilt. Tilts of 1°-3° have been reported previously for some of the northern Emperor seamounts (Lonsdale et al., 1993). Because these tilts are small and the angle between the remanent magnetization vector and down-dip azimuth of tilt is large (>60°), the effect on the paleolatitude is negligible. Measurements made at unit contacts also fail to indicate significant dips (Tarduno and Cottrell, 1997).

The new paleomagnetic result directly questions the validity of the Late Cretaceous Pacific apparent polar wander path (Fig. 5). But how could these prior results be so errant? Previous Late Cretaceous poles are heavily or solely based on the inversion of magnetic surveys over seamounts (Gordon, 1983; Sager and Pringle, 1988). Reviews of the methods used to fit these poles suggest they are far more uncertain than commonly supposed (Parker, 1991). Viscous and induced magnetizations can also bias the resulting pole positions (Gee et al., 1989). Interestingly, high- latitude poles similar to the new colatitude result (Fig. 5) have been reported from preliminary analyses of marine magnetic anomaly skewness data of comparable age (Vasas et al., 1994).

Hotspot Motion and True Polar Wander
The other paleolatitude value from the Emperor trend that adequately averages secular variation was derived from Suiko Seamount (65 Ma) (Kono, 1980) (Fig. 1). The 8° discrepancy between the Suiko Seamount paleolatitude and the present-day latitude of the Hawaiian islands has been attributed previously to early Cenozoic true polar wander (Gordon and Cape, 1981; Sager and Bleil, 1987), which is defined as a rotation of the entire solid Earth in response to shifting mass heterogeneities in the mantle (Goldreich and Toomre, 1969). True polar wander predictions based on global paleomagnetic data from the continents (Besse and Courtillot, 1991), however, do not agree with the new Detroit Seamount data (Tarduno and Gee, 1995; Tarduno and Cottrell, 1997). Furthermore, renewed tests of Cretaceous true polar wander models show that the solid Earth rotations proposed are not seen in paleomagnetic data from regions where large changes in latitude should be observed (Cottrell and Tarduno, 2000b; Tarduno and Smirnov, 2001). Therefore, the true polar wander rotations proposed appear to be artifacts related to the fixed hotspot reference frame employed.

Because Late Cretaceous true polar wander predictions are inconsistent with the Pacific observations, we must now consider hotspot motion as an explanation for the difference between the paleomagnetic paleolatitude derived for Detroit Seamount (Tarduno and Cottrell, 1997) and that predicted by a fixed hotspot reference frame. We can isolate the latitudinal history of the Emperor seamounts from that of the Hawaiian chain by subtracting the difference between the present-day latitudes of the 43-Ma bend and Hawaii from the present-day latitudes of each of the Emperor seamounts. In effect, we slide the Emperor trend down the Hawaiian chain to the present-day latitude of Hawaii (Fig. 6). In so doing, we produce a plot predicting the paleolatitude of Emperor seamounts if they were formed by a hotspot moving at constant velocity beneath a stationary plate. The new Detroit Seamount result together with the Suiko Seamount data parallel this predicted trend and provide support for the hotspot motion hypothesis. Differences between the data and predicted values also allow for some northward plate motion. It is difficult to place error bounds on the rate of motion, because there are only two estimates of paleolatitude available. Nevertheless, the data suggest that the Hawaiian hotspot could have moved southward from 81 to 43 Ma (Norton, 1995) at a constant rate of 30-50 mm/yr, while the Pacific plate moved slowly northward in a paleomagnetic (spin axis) frame of reference (Fig. 5).