A total of 56 samples from Cores 165-998A-1H through 11H (0-5 Ma), 18 samples from Cores 165-998A-31X through 62X (22-34 Ma), and 18 samples from Cores 165-998B-3R through 23R (32-44 Ma) were stepwise AF demagnetized. Only 24 of these samples gave linear demagnetization paths and ChRM inclinations, with the rest either giving scattered directions with no ChRM or steep directions indicative of the drill-string overprint (Table 11, Table 16; Fig. 2).
These data were combined into two groups, one spanning 0-100 mbsf (2.1 Ma) and the other spanning 582-740 mbsf (38 Ma), with no data from the intervening intervals (Table 20; Fig. 9). In addition, the results from stepwise thermal demagnetization of 33 samples from Cores 165-998B-32R through 37R (848-902 mbsf; 49-52 Ma) were provided by V. Louvel and B. Galbrun (unpubl. data) (Table 16). The mean paleolatitude when corrected for the POAM bias is 7.9° (with asymmetrical 95% confidence limits of +5.1° or -7.9°) for the 33 samples. Site 998 would therefore have moved at least 6.5° northward over the past 50 m.y. or at an average rate of at least 14 km/m.y.
Site 998 only provides a proxy for the northward motion of the Caribbean plate because it has not been a part of that plate since the Eocene, though it has maintained a position to the north of the Caribbean plate (e.g., Pindell et al., 1988). The amount of northward motion estimated from this sparse data set is greater than that found in the larger data sets from Sites 999 and 1001, but is consistent within the relatively large uncertainties.
A total of 152 samples from Hole 999A and 146 samples from Hole 999B were stepwise AF and thermal demagnetized (Table 12). In general, AF demagnetization was more successful at removing the drill-string overprint than was thermal demagnetization. Typically, AF demagnetization between 25 and 70 mT was best at resolving the ChRM, whereas thermal demagnetization above 500°C was needed to remove the drill-string overprint. Of the 298 samples, 113 gave ChRMs that pass the discrete sample rejection criteria (Table 17). These span from 2 to 1065 mbsf, nearly the entire cored interval, though data are absent from 300 to 550 mbsf.
The paleolatitudes from the 113 samples and mean paleolatitudes for 10 intervals (0-50, 50-100, 100-200, 200-300, 500-600, 600-700, 700-800, 800-900, 900-1000, and 1000-1065 mbsf; Table 21) are plotted against depth in Figure 10. Lines fit through these indicate the Caribbean plate has moved northward at ~14 km/m.y. Over the past 65 m.y. this gives a total of 8° of northward translation, which places Site 999 very near the equator at the time of the K/T boundary impact event.
Split-core results obtained during Leg 165 had indicated that the intensity of magnetization was extremely low below 22.5 mbsf at Site 1000. The abrupt decrease in magnetization was attributed to reduction diagenesis. In addition, coring only reached limestones of early Miocene age. Therefore, the amount of latitudinal motion recorded by these sedimentary units would likely be small at best.
Nonetheless, we progressively AF demagnetized an additional 40 discrete samples to assess further the magnetization (Table 13). None of the samples from depths >22.5 mbsf gave stable ChRM directions and all displayed weak natural remanent magnetizations (NRM), typically <8 x 10-4 A/m. Data from this site do not provide any paleolatitude constraints and are not considered further.
As discussed above, progressive demagnetization of discrete paleomagnetic samples from basalt cores indicates that the drill-string overprint is removed by ~15-20 mT (Fig. 5). Because of this and because the split-core data outnumber the discrete by about an order of magnitude, our primary data for the basalt cores are the split-core inclinations after 25 mT demagnetization (or 20 mT when the 25 mT step is not available). We have carefully edited the data to avoid ends of core sections or gaps that occur within the core (see core photos on pp. 740-763 of Sigurdsson, Leckie, Acton, et al. [1997]). This process left us with 230 inclination estimates. This data set (Table 15) agrees well with mean inclinations determined from PCA of progressively demagnetized discrete samples (Fig. 6; Table 18). Only in the interval from 504.5 to 508 mbsf is there significant disagreement between split-core and discrete results, with both data sets displaying large variations in inclination over this short interval.
The data are divided into groups based on which "flow unit" they occur within, using the flow-unit divisions in Table 1 (Fig. 6). In some cases, two or more flow units may belong to the same flow or may belong to two or more flows that were extruded within a short interval of time (less than a few hundred years) relative to geomagnetic SV. In these cases, the variation in inclination from unit to unit can be used to group the flow units into units that represent independent samples of the geomagnetic field, which we refer to as "SV units." The goal of this procedure is to determine how many independent samples of the geomagnetic field are present within the 32-m-long cored interval in order to assess how well geomagnetic SV has been averaged. The more SV units present, the more time likely sampled by the volcanic rocks and the greater the likelihood that the mean paleomagnetic inclination represents that of a time-averaged, geocentric axial dipole field.
Flow-unit mean inclinations from stratigraphically adjacent units were compared. If a mean inclination was within three standard errors of the mean inclination of stratigraphically adjacent units (i.e., the inclinations differ insignificantly at greater than the 99% confidence level), then the inclinations for the units were combined and a SV-unit mean inclination was computed. Similarly, we converted each inclination into a paleolatitude and computed a SV-unit mean paleolatitude. Overall, the grouping process resulted in 27 flow units giving 12 SV units. Note that the units have both positive and negative inclinations, which could indicate field reversals or could simply reflect geomagnetic SV during an interval of constant polarity.
The 12 SV-unit mean paleolatitudes, which are used in subsequent calculations of the overall mean paleolatitude, are shown in Figure 11. To calculate the overall mean paleolatitude, we use two different approaches. First, using the age of the overlying limestone and radiometric ages from the basalt (76-80 Ma and 81 ±1 Ma, respectively), we might suspect that the basalts were all deposited in an interval of constant polarity prior to the deposition of the overlying sediment. This would most likely place the basalts entirely within the older portion of Chron 33n (73.619-79.075 Ma) or entirely within Chron 33r (79.075-83.00 Ma). Assuming all the basalts were extruded within an interval of constant polarity gives a mean paleolatitude of 4.3°(N or S) ±6.3°. Alternatively, both polarities could be present, with the basalts possibly being extruded over the interval from Chron 34n to 33n. Making no assumption about the polarity gives an unbiased mean paleolatitude of ±4.5°(+9.7°/-4.5°).
These estimates do not require that the Caribbean plate reside in the Northern Hemisphere, though the polarity through the K/T boundary section at this site clearly indicates a Northern Hemisphere position. Hence, the bulk of the Caribbean plate has likely been in the northern hemisphere since at least 80 Ma. Both basalt estimates place the Caribbean plate ~10° south of its current position. Within the uncertainties, Site 1001 could have been on or very near the equator in the Late Cretaceous.
A total of 76 samples (30 from basalt cores) from Hole 1001A and 13 samples from Hole 1001B were stepwise AF and thermally demagnetized. Both demagnetization methods typically revealed shallow inclinations after removal of a steep downward overprint direction (Fig. 4). Of the 59 samples from sedimentary units, 28 gave ChRM inclinations. The overprint was difficult to remove in the younger sediments (down to the base of Core 165-1001A-17R at 160 mbsf, middle Miocene age), with no ChRM inclination obtained for cores above this.
In addition, the results from stepwise demagnetization of 138 samples from sedimentary units from Hole 1001A and 141 samples from Hole 1001B were provided by V. Louvel and B. Galbrun (unpubl. data). The combined data set of 308 sample paleolatitudes is subdivided by depth into four intervals down to basaltic basement (160-211, 217-331, 331-396, and 396-484 mbsf).
The magnetostratigraphy (Fig. 12) is well resolved for the interval from 217 to 331 mbsf, corresponding to Chrons 24r-27n, and for the interval from 331 to 396 mbsf, corresponding to Chrons 27r-31r (pp. 314-315 of Sigurdsson, Leckie, Acton, et al., 1997; V. Louvel and B. Galbrun, unpubl. data; King et al., Chap. 8, this volume). Because the polarity of the samples can be ascertained, computation of paleolatitudes is straightforward and gives estimates with smaller uncertainty. To ensure that the mean paleolatitudes for these two intervals did not include samples with uncertain polarities, we excluded samples collected near polarity transitions or within zones where the polarity was uncertain (Fig. 12). Data from within these two intervals are important for several reasons. First, based on the magnetostratigraphy, particularly the sign of the inclinations within the chronozones, Site 1001 was in the Northern Hemisphere from Chron 31r to Chron 24r (from 71 to 54 Ma). Second, the smaller 95% confidence limits indicate that Site 1001 was at least 6° south of its current position in the Late Cretaceous and possibly within 1° of the equator.
The mean paleolatitudes for all four intervals indicate that the Caribbean plate has moved progressively northward by ~10° since the Late Cretaceous (Fig. 13). The limestones from 396 to 484 mbsf give a mean inclination of 14.4° and a mean unbiased paleolatitude of 2.9°, which are consistent with the mean inclination of 14.9° and mean unbiased paleolatitude of 4.5° estimated from the 12 SV units from the underlying basalt. The agreement of sedimentary and basalt estimates indicates that biases from inclination shallowing in the sediments are likely small and perhaps negligible.