Paleosecular variation and GAD studies of 0–2 Ma flow sequences from the Aleutian Islands, Alaska



[1] A study was undertaken in the late 1960s to investigate paleosecular variations of the geomagnetic field as recorded in volcanic rocks from the Aleutian Islands. The early results were internally consistent but to be wholly credible by today's standards needed more detailed demagnetization. Complete thermal demagnetization protocols have been applied to the unmeasured archived samples from the six flow sequences used in the initial study, and 40Ar/39Ar techniques have been used to improve the time resolution. The flow sequences have ages ranging from about 50 ka to 2 Ma, and the number of sequential flows in the individual sequences varies from 8 to 21. After strict selection criteria were applied (MAD < 5°, α95 < 5°) for both demagnetization data from samples and samples within one flow, the number of acceptable flows per flow sequence dropped to between 5 and 15. With the exception of a sequence showing transitional field behavior, the between-flow dispersion and the α95 values for the other sequences were notably low with respect to secular variation models, and their mean directions were very close to the GAD field. Since the time represented by the individual sequences is not well determined, the low dispersion could represent very short eruption times. In contrast, the lack of dispersion with respect to the GAD field can be taken to indicate good time averaging. Since the locations of the sampled flows are at roughly the same latitude (about 50°N) but are spread over about 10° of longitude, the dispersion was calculated for both the locality-means and the flow-means. These data represent the whole 2 Myr and give a dispersion which is lower than current secular variation models predict. A similar data set published for locations in western Canada that are at roughly the same latitude and overlap in age with the Aleutian sites gives dispersions that are close to the model predictions. At face value this can be interpreted as indicating low secular variation for the Aleutian sites.

1. Introduction

[2] This study was initiated in 1968 as part of an attempt to investigate whether the low values of the secular variations of the geomagnetic field over the Pacific region today persist back through time. That the Pacific region (approaching a hemisphere) has lower secular variation than the other parts of the Earth has been recognized ever since global measurements of geomagnetic field variations were available [see, e.g., Fisk, 1931; Chapman and Bartels, 1940; Vestine et al., 1947]; thus it is of obvious importance to determine whether this is a transient, semipermanent or permanent feature of the field. If it were permanent, this would point to a significant asymmetry within the Earth, if semipermanent, the timescales involved could put severe constraints on models for the properties of the core and mantle, and if transient, the most probable cause could be statistical fluctuations in the operation of the geodynamo. There is still an ongoing debate as to the significance of the secular variation low in the Pacific hemisphere [e.g., Merrill et al., 1996; Johnson and Constable, 1997; Gubbins and Gibbons, 2004].

[3] At the time we carried out the original study [Bingham, 1971; Bingham and Stone, 1972], knowledge of the volcanic history of the Aleutian volcanic arc was somewhat rudimentary, and age control on the sequences sampled was poor. The paleomagnetic techniques used involved alternating field demagnetization of pilot samples followed by a single level “blanket” demagnetization of the remainder [Bingham and Stone, 1972]. The resulting data sets were internally consistent, and the final conclusion was that the data could not clearly distinguish between transient and nontransient models for Pacific paleosecular variation. This original study left us with a well-distributed set of undemagnetized oriented core samples to measure with present-day techniques for the present study. The site locations are shown in Figure 1.

Figure 1.

The locations of the sampled areas in the Aleutian Islands (CCR, RDH, KAN, DFB, ASH, NJC) and locations in Canada sampled and reported on by Mejia et al. [2002].

2. Field Methods

[4] All samples were collected using a gasoline powered core drill, with non magnetic coring tubes made of stainless steel tipped with diamonds mounted in phosphor bronze. The cores were orientated with respect to present-day north and horizontal using a moveable platform mounted on a slotted tube that could slide over the core. The moveable platform was used as a sun compass and as a mount for a magnetic compass. This compass was used to measure both the bearing of fiducial mark on the sample, and for shooting sight lines to prominent map features. These measurements enabled us to directly determine the local declination of the geomagnetic field. In cases where neither the sun compass nor sight lines were available, it was possible to extrapolate the magnetic declination. In cases where bad weather prohibited the use of the sun compass or sighting, we commonly returned to the location at a later date and obtained the orientations of the holes that had previously been marked with the orientation of the fiducial line.

[5] All samples were taken from sequences of flows, with a target of 6 to 8 cores recovered from each flow. In practice there were occasional flows where as few as 3 cores were recovered, and occasionally as many as 11. As far as possible, the individual samples from a given flow unit were demonstrably from the same flow outcrop, but sufficiently far apart (up to 50 m) that they were clearly giving independent magnetic records.

3. Laboratory Methods

3.1. Paleomagnetism

[6] All the cores were 2.54 cm in diameter, and were sliced into either two 2.1 cm or three 1 cm tall cylinders starting from the bottom of the core. In general, only one of these cylinders was used in the original study.

[7] Some of the original measurements were made using a two probe fluxgate magnetometer set up in the same way as an astatic magnetometer, but the bulk of them were made using a 5 Hz spinner magnetometer using fluxgate probes as sensors [Foster, 1966]. Alternating magnetic field (60 Hz) demagnetization was done using a 2-axis sample tumbler inside the demagnetizer coil [Hutchings, 1967] the whole system being located in a near zero magnetic field produced by three pairs of Helmholtz coils. The demagnetization protocol called for one sample from each flow to be demagnetized in a series of 8 steps up to a peak field of 200 mT. The demagnetizing field that minimized the between-flow scatter was then chosen as the level at which all remaining samples would treated. The values selected for the various flow sequences ranged between zero and 60 mT.

[8] The more recent (2001–2003) measurements were all made on a 2-G, 2-axis cryogenic magnetometer, with thermal and alternating field (AF) demagnetizations carried out using a Schonstedt furnace and AF coil system. This whole system is housed in a magnetically shielded room. The demagnetization protocol called for an initial alternating field of 5 mT to remove any components acquired during storage, followed by thermal demagnetization at 150°, 200° and 250°C followed by 25° increments to 600°C. The results from this demagnetization procedure were then displayed as orthogonal Zijderveld plots (Z plots) [Zijderveld, 1967], outlier points removed, and a line fit calculated for those results that showed a systematic decrease in intensity while following a path toward the origin. Line-fit data from individual samples within a single flow unit were removed if they were displaced more than 30° from the mean of the remainder and/or had Maximum Angular Deviations (MAD) [Kirshvink, 1980] greater than 5°. The flow means were then calculated together with statistical analyses that included the two-tier analysis of Watson and Irving [1957]. Following the guidelines of Tauxe et al. [2003] the final selection was restricted to flow means with an alpha-95 < 5°, and in this case the statistics were calculated for both the directional data and the Virtual Geomagnetic Poles (VGPs) using the selected line-fit data as the starting point for both sets. These data are listed for each locality and include (1) data from the original study [Bingham and Stone, 1972]; (2) all the data that passed the selection criteria that required the line-fits to have MAD values of less than 5° and no individual sample line-fit directions that were displaced from the mean direction of the flow in question by more than 30°; and (3) data selected under (2) above, with the additional constraint that the alpha-95 confidence limits of the declination and inclination measurements should be less than 5° for the flow in question. Each of these sections of the tables show the mean values and the statistics associated with them. The final part of the table gives the results from each of the flows that survived the initial selection. Those flows that passed the final selection (3) are printed in bold, and all of them have the VGPs calculated from the flow mean Declinations and Inclinations. The number of samples and flows that were accepted and rejected is shown in Table 1.

Table 1. Samples and Flows That Were Accepted and Rejecteda
Flow SequenceTotal SamplesSamples MAD < 5Samples displ < 30°Total FlowsFlows α95 < 5
  • a

    The three letter code for each sampling locality is shown with the total number of samples collected, the number left after the MAD < 5° test, and the number left after removing those displaced by >30° from the flow mean. The last column gives the number of flow-means left after removing those with α95 < 5°.


3.2. Statistics

[9] There are many different statistical approaches that can be applied to data sets of this sort. In the interests of continuity we have used the same approach as was used in the original paper [Bingham and Stone, 1972] plus dispersion calculations following the suggestions of Johnson and Constable [1996] and Tauxe et al. [2003]. We have thus recorded the angular standard deviation, delta, defined by cos−1(R/N) where R is the sum of the unit vectors describing the measured values and N the number of measurements involved. The calculations of kappa (κ) and alpha-95 (α95) follow Fisher [1953], using the probability function for α95, and not the approximation used in many early publications. To determine whether the dispersion of the data points from within the individual flow units (κw) allowed a meaningful determination of the between-flow scatter (κb) the total scatter (st) as determined using the methods of Johnson and Constable [1996] was compared with the within-flow scatter (sw) using the method of Cox [1970] described by Johnson and Constable [1996]. The Watson and Irving [1957] two-tier test was also applied for comparison with the original data [Bingham and Stone, 1972] and showed that the between-flow scatter was statistically significant for all the flow sequences studied.

[10] In the data tables the total dispersion for each locality (combined within-flow and between-flow) is calculated about the vector mean and the mean VGP for the locality (stv and Stv, respectively) and about the GAD field (stp and Stp). The final estimate of the between-flow dispersion (sbp, Sbp) presumed to represent the secular variation, is calculated from

equation image


equation image

where sbp and Sbp are measures of the between-flow scatter about the GAD poles and GAD field directions taking into account the within-flow scatter.

3.3. Geochronology

[11] Volcanic rock samples from Round Head and Kanaton were dated by the 40Ar/39Ar method at the University of Wisconsin at Madison. Results and methodology are described by Jicha et al. [2004]. Samples from Crater Creek, Driftwood Bay, New Jersey Creek and Ashishik were dated by the 40Ar/39Ar method, at the Geochronology Laboratory of the Geophysical Institute at the University of Alaska Fairbanks. These samples were obtained from the unweathered core-ends or unused specimen cores of samples collected for paleomagnetic analysis, or from specimens previously demagnetized using alternating magnetic fields. For these analyses, the monitor mineral Bern-4B with an age of 17.25 Ma was used to monitor neutron flux (and calculate the irradiation parameter, J). The samples and standards were wrapped in aluminum foil and loaded into aluminum cans of 2.5 cm diameter and 6 cm height. The samples were irradiated in position 5c of the uranium enriched research reactor of McMaster University in Hamilton, Ontario, Canada for 0.5 megawatt-hours. Upon their return from the reactor, the sample and monitors were loaded into 2 mm diameter holes in a copper tray that was then loaded in a ultra-high vacuum extraction line. The monitors were fused, and samples heated, using a 6-watt argon-ion laser following the technique described by York et al. [1981], Layer et al. [1987], and Layer [2000]. Argon purification was achieved using a liquid nitrogen cold trap and a SAES Zr-Al getter at 400C. The samples were analyzed in a VG-3600 mass spectrometer at the Geophysical Institute, University of Alaska Fairbanks. The argon isotopes measured were corrected for system blank and mass discrimination, as well as calcium and potassium interference reactions following procedures outlined by McDougall and Harrison [1999]. Auxiliary Material Tables S1–S4 report the details from all analyses. Ages are quoted to the ±1 sigma level and calculated using the constants of Steiger and Jaeger [1977].

4. Data Sets

4.1. Geologic Setting

[12] The six localities sampled are all from exposures of extrusive rocks associated with the modern volcanoes of the Aleutian Island Arc. Six sequences of flow units were sampled. The smallest sequence contained 8 flows (Kanaton Ridge, Kanaga Island) and the largest 21 flows (Driftwood Bay, Unalaska Island). The relative stratigraphy within the six flow sequences is unambiguous; however, the overall time represented is more difficult to estimate. The absolute ages of each sequence were determined using 40Ar/39Ar mass spectrometry.

[13] The following descriptions are arranged from youngest to oldest.

4.2. Crater Creek (CRC) (53°23′N, 191°55′E)

[14] The Crater Creek site is located in the gorge cut by Crater Creek where it drains Okmok caldera located on Umnak Island. The gorge walls expose a fresh section of the precaldera volcano. Byers [1959] noted 15 flows at the location sampled, which probably correspond one for one with the 15 flows we sampled. Chatterjee [1971] describes these flows as ranging from tholeiites to olivine tholeiites. Between 5 and 11 cores per flow were taken.

4.2.1. Geochronology

[15] Three samples from each of three flows from the Crater Creek section were step-heated in two steps. Ages from all fractions were within 2-sigma of zero and there is no discernable difference in ages from the top to the bottom of the section. A composite isochron from the 18 steps gives an age of 20 ± 144 ka, with an initial 40Ar/39Ar ratio of 288 ± 4 (MSWD = 0.6). On the basis of these data it appears that the Crater Creek section was emplaced rapidly within the last ∼150 kyr, and based on the morphology of the crater, perhaps within the last 50 kyr.

4.2.2. Paleomagnetism

[16] The original study used a single alternating field demagnetization level of 60 mT [Bingham and Stone, 1972]. The distribution of the flow means was markedly elongate in a NW-SE direction, but the overall mean direction is not significantly different from the mean obtained using thermal demagnetization (this study).

[17] Thermal demagnetization indicates a range of blocking temperatures from flow to flow, with magnetite dominating in some, but more commonly showing that a range of magnetic carriers are present. The Z plots commonly give a clear straight-line path to the origin similar to that shown in (Figure 2a).

Figure 2.

The four panels show examples of Zijderveld diagrams (orthogonal plots) that are fairly representative of those found in the flow sequences studied. The left part of each figure shows the results of thermal demagnetization of selected samples. These plots show the declination (diamonds) plotted relative to N, S, E, W and inclination (squares) shown as true inclinations (i.e., projected onto the plane which includes the vector). The numbers adjacent to selected points indicate the demagnetization temperatures in °C. The right diagram shows the changes in total intensity with temperature during the demagnetization process.

[18] Using the selection criteria already described, two samples with MAD angles greater than 5° were excluded, seven more were excluded as too divergent (>30°) from the flow mean. Of the resultant flow means four out of 15 flows had alpha-95 values greater than 5° and were thus excluded from the final analysis. The mean MAD angle for all samples included in the calculations is 1.3° and the mean number of measurements per line fit is 7.6.

[19] Comparison of the flow means shows no significant serial correlation, although some pairs of flows show similar directions to each other (flows 2–3 and 6–7). The overall distribution of both VGPs and directions is elongate in a NW SE direction (Tables 2a2c, Figure 3a).

Figure 3.

The left part of the diagrams representing each of the six localities is an equal angle (stereographic) projection of the vector directions of the flow means. The center part shows the same data in an expanded format. The dots represent the data which passed all the selection criteria (highlighted data from the data tables), and the triangles represent the data that failed the final selection (flow means with an α95 > 5). The right-hand plot represents the same data displayed as VGPs. The circles represent the alpha-95 confidence limits about the means.

Table 2a. Crater Creek (53.47°N, 191.92°E), ∼50 kaa
  • a

    This and the following sets of data tables give the locality by name and by longitude and latitude followed by means of flow means and means of cores (samples) (N, number of measurements; R, resultant of N unit vectors; D&I, declination and inclination of the resultant vector; κ (kappa) and α95 (alpha 95) following Fisher [1953]; and delta the dispersion of the unit vectors). These results are shown for the original measurements [Bingham and Stone, 1972], and for the new results where the maximum angular deviation of the line-fit was <5° and individual samples deviated by <30° from the flow mean (Table 2a) and for the latter data with flows showing an α95 > 5° removed (Table 2b). This was done for the means of flows, and means of samples in each case. The total dispersion (st) was calculated following Tauxe et al. [2003]; stv represents the dispersion about the vector mean and stp about the GAD field. The within-flow dispersion (sw) and between-flow dispersion (sbp) were calculated following Cox [1970]. The same calculations are recorded for the VGPs with Longitude and Latitude replacing Declination and Inclination for the total (Stv and Stp), within-flow (Sw), and between-flow (Sbp) dispersion. Table 2c shows the data for the individual flows with the VGPs derived from the sample Declination and Inclination data. The individual flows in the final selection are in bold type. The subscript “v” indicates vector means and subscript “p” means of poles calculated from the individual sample measurements.

Original (1972) Results
Mean of flows15 35867.
All New Data With MAD < 5° and Deviation From Flow Mean < 30°
Mean of flows1514.77354.3069.9060.525.0010.07
Mean of all cores9996.62355.3069.3041.162.2012.59
Table 2b. New Data as in Table 2a With alpha95 < 5° for Flow Directionsa
Directional Data
Flow mean, D,I1110.86356.2068.7071.495.409.159.589.715.229.51
Flow mean VGPs1110.66131.8087.8029.598.5014.2415.1215.147.6314.86
Table 2c. Data for Individual Flowsa

4.3. Round Head (RDH) (51°54′N, 182°57′E)

[20] Round Head forms a prominent sea cliff near the easternmost extremity of Kanaga Island. The cliff is composed of thick (2–14 m) olivine basalt flows that appear to have been extruded from a vent on the flanks of Ancient Mount Kanaton, the ancestral volcano that later underwent a major caldera forming eruption. The flows appear to be very fresh, contain conspicuous phenocrysts of dark augite [Coats, 1956] and are separated by rubbly layers. The individual flows have variable apparent dips, and some lens out and disappear horizontally. However, thin silts and shale beds were noted about 6m below the lowermost columnar jointed flow. As far as could be determined, these beds were horizontal, thus indicating that the dips seen in the flow units are all primary. The flows were sampled between sea level and 250 m with between 5 and 9 cores per flow.

4.3.1. Geochronology

[21] The age of the flows sampled at Round Head is based on 40Ar/39Ar results from five incremental heatings from two flow units from near the bottom and the top of the section [Jicha et al., 2004]. The ages, 134.0 ± 3.1 ka and 130.5 ± 7.1 ka, allow the flows to have been erupted essentially contemporaneously, which is in accord with the tight clustering of the paleomagnetic directions described below.

4.3.2. Paleomagnetism

[22] In the original study [Bingham and Stone, 1972] all samples were demagnetized in steps up to a peak alternating field of 30 mT. The results of the thermal demagnetization carried out for the present study are similar to those obtained earlier. The Z plots give generally good straight line tracks toward the origin and the magnetic intensity versus temperature plots indicate a range of blocking temperatures up to and including magnetite. All samples have MAD angles less than 5° and none were significantly divergent (>30°) from the flow mean. Six out of 14 flows were rejected because of high (>5°) alpha-95 values. The remaining individual flow means are very tightly grouped, making it difficult to determine whether or not there are sequential relationships between them (Figure 3b). A two-tier analysis on the original data indicated that it was not possible to separate the between-flow and within-flow scatter; however, in this study both the two-tier and the Johnson and Constable analyses indicate that the between-flow dispersion value is significant (Tables 3a3c).

Table 3a. Round Head (51.9°N, 182.95°E), ∼120 kaa
Original (1972) Results
Mean of flows14 1.172.2401.423.9
All New Data With MAD < 5° and Deviation From Flow Mean < 30°
Mean of flows1413.968.5072.20370.102.104.06
Mean of all cores8584.258.5071.90112.001.507.62
Table 3b. New Data as in Table 3a With alpha95 < 5° for Flow Directionsa
Directional Data
Flow mean, D,I87.987.4871.71379.822.853.894.165.974.395.72
Flow mean VGPs87.96237.3084.4174.024.205.756.579.616.979.22
Table 3c. Data for Individual Flowsa

[23] The final values of alpha-95 (2.8°) and the dispersion (5.7°) indicate a much tighter grouping than would be expected if the total time represented by the sequence of flows was comparable to the timescales associated with the secular variation of the nondipole field (of the order of 1000 years) (Figure 3b). It should also be noted that the mean field and VGP directions do not coincide with, but are close to the GAD field, the VGP being within 2° of the GAD pole (Figure 3b).

[24] The mean MAD angle for all samples included in the calculations is 0.9° and the mean number of measurements per line fit is 8.8.

4.4. Kanaton (KAN) (51°54′N, 182°54′E)

[25] Ancient Mount Kanaton is the name given to the volcano that gave rise to the caldera rim that surrounds the currently active Kanaga Volcano, located on Kanaga Island [Coats, 1956]. The site sampled was part of the caldera rim known as Kanaton Ridge. It consists of nearly horizontal flows of basalt and andesite [Coats, 1947, 1956]. The morphology of the caldera rim suggests that there has been no significant tilting, leading to the assumption that any apparent dips to the flows are primary. Eight flows were sampled, with between 6 and 11 cores per flow except for the poorly exposed lowermost flow, from which 3 cores were taken.

4.4.1. Geochronology

[26] A single K/Ar age was of 184 ± 180 ka was obtained for a sample collected during the initial study. More recently, Jicha et al. [2004] published new 40Ar/39Ar ages determined using incremental heating experiments on samples from the same section of flow units. These new experiments give ages of 199.1 ± 2.5 ka and 198.1 ± 2.1 ka.

4.4.2. Paleomagnetism

[27] The original study [Bingham and Stone, 1972] was based on blanket demagnetization at 29 mT. Flow 8, from the highest part of the section, showed two distinct directions of magnetization. Both had shallow dips and showed opposite polarities. In retrospect we assume that there were in fact two flows. The results from this study are shown in Tables 4a4c.

Table 4a. Kanaton (51.9°N, 182.9°E), ∼200 kaa
Original (1972) Results
Mean of flows7 355.965.3136.35026.4
All New Data With MAD < 5° and Deviation From Flow Mean < 30°
Mean of flows76.97356.4067.60222.684.105.02
Mean of all cores5251.55356.6067.20113.271.907.54
Table 4b. New Data as in Table 4a With alpha95 < 5° for Flow Directionsa
Directional Data
Flow mean, D,I54.99355.5065.70298.924.404.204.696.005.785.66
Flow mean VGPs54.9744.3085.40124.246.906.517.298.8810.528.38
Table 4c. Data for Individual Flowsa

[28] The Z plots show good straight line tracks toward the origin above about 350°C and the magnetic intensity versus temperature shows a range of blocking temperatures up to about 600°C.

[29] Following thermal demagnetization, Flow 8 gave the same divergent directions as those seen in the original study, and was therefore excluded from the final data set. No samples were rejected as being divergent from the flow mean and the mean MAD angle for all samples included in the calculations is 1.8° and the mean number of measurements per line fit is 6.2.

[30] Flows 1 and 7 had alpha-95 > 5° and thus were excluded. The remaining five flows show no obvious serial correlation or grouping. However, the overall spread of the vectors is small. Thus they may represent a short time interval with respect to secular variation of the nondipole field (Figure 3c).

4.5. Driftwood Bay (DFB) (53°58′N, 193°06′E)

[31] Driftwood Bay is located on the north side of Unalaska Island, to the west of Dutch Harbor on the flanks of Makushin volcano. The flow sequence sampled is part of the Makushin series of Drewes et al. [1961] and consists primarily of basalt and andesite flows that range in thickness between 3 m and 17 m. Twenty-one flows were cored from a 300m high sea cliff 200m west of, and below the now abandoned White Alice communication site. The shallow dip away from the Makushin volcanic center is considered primary. In the field, the flows are commonly separated by grass-covered rubble up to 30 m thick, making it difficult to find evidence of weathering or other indicators of the time represented. However, the presence of the rubble units is a strong indication of between-flow weathering and/or deposition of sediments. The lowermost flow is an exception with about 20m of exposed beds of volcanogenic sediments between it and the flow above. Six or more cores were drilled from each flow unit.

4.5.1. Geochronology

[32] Two or three samples from each of three flows from the Driftwood Bay section were step-heated in 2–8 steps. Ages from the three flows were significantly different and so are considered separately. For Flow 21 (sample DFB1077), at the base of the section, a 3 sample, 8 fraction isochron age of 822 ± 150 ka (initial 40Ar/36Ar = 314 ± 19, MSWD = 1.2) reflects the beginning of this eruption sequence (Figure 4). A three sample, 15 fraction isochron age from a sample from flow 2 (DFB961) at the top of the section has an age of 349 ± 67 ka (Figure 4). A 2 sample, 5 fraction isochron age from flow 6 (DFB989) is poorly constrained at 449 ± 651 ka, falling between the ages from flow 21 and flow 2. Thus the Driftwood Bay sequence was emplaced over a period of approximately 470 ± 160 thousand years.

Figure 4.

Argon isotopic correlation diagrams for the uppermost (DFB961) and lowermost (DFB1077) dated samples from Driftwood Bay. All errors are reported at ±1 sigma. 40Ar/36Ari, the initial 40Ar/36Ar ratio determined from the isochron y-intercept; MSWD, Mean Square Weighted Deviation; n, number of fractions used in calculating the isochron age.

4.5.2. Paleomagnetism

[33] For the original study stepwise alternating demagnetization was carried out on selected pilot samples. These pilot demagnetizations showed very high magnetic stability with no discernable secondary components, so blanket demagnetizations were not made and the results are based on the initial magnetization (NRM) [Bingham and Stone, 1972].

[34] For this study all samples were thermally demagnetized. The Z plots show scattered directions below 300°C and good straight line tracks toward the origin at higher temperatures (Figure 2b). The magnetic intensity versus temperature plots show a spread of blocking temperatures between about 400 and 600°C. Five samples were rejected as being too divergent from the flow mean. The mean MAD angle for all samples included in the calculations is 1.05° and the mean number of measurements per line fit is 7.6. Six of the 21 flows sampled had alpha-95 values in excess of 5°, although only 2 exceeded 6° (Tables 5a5c).

Table 5a. Driftwood Bay (53.97°N, 193.27°E), 400–800 kaa
Original (1972) Results
Mean of flows21 0.969.
All New Data With MAD < 5° and Deviation From Flow Mean <30°
Mean of flows2120.661.4067.4059.324.2010.28
Mean of all cores112109.821.7067.7050.851.9011.33
Table 5b. New Data as in Table 5a With alpha95 < 5° for Flow Directionsa
Directional Data
Flow mean, D,I1514.851.4068.4095.783.908.018.298.463.268.35
Flow mean VGPs1514.65340.5088.2040.476.1012.3312.7712.906.2512.62
Table 5c. Data for Individual Flowsa

[35] The remaining 15 flows comprise the final data set and show two tight groups of sequential flows (2 through 4 and 5 through 11) and an apparently sequentially correlated sequence (15 through 20) (Figure 3d). This leaves flow 13 as a lone flow. The very tight grouping for flows 5 through 11 presumably indicates that they were erupted in a short time relative to the secular variation of the nondipole field, and probably represents a timescale of a few centuries or less. This apparently short timescale is hard to reconcile with the between-flow rubble described above; however, the source of the rubble, and its relationship to the flows, could not be determined in the field. The time represented by the whole sequence of flows is harder to assess, but the overall scatter of the flow directions is appropriate for time-averaged secular variation. At face value, the two tight groupings of directions represent two pulses of volcanic activity at 349 ± 67 ka and 470 ± 160 ka, respectively.

4.6. Ashishik (ASH) (53°32′N, 191°54′E)

[36] The Ashishik flow sequence sampled is located near to the New Jersey Creek locality on the NE end of Umnak Island. Both are part of the Ashishik Basalts of Byers [1959] and appear to be distal flows from the volcano that created Okmok caldera. The Ashishik flows, as sampled at this locality consists of aphyric and feldspathic basalt flows that range in thickness 3 m to 23 m, and are separated by well-weathered layers. From a distance the whole sequence appears to be horizontal; however, at close quarters the rubble interbeds make it impossible to determine the bedding attitude. Fourteen flows were sampled at 5 cores per flow.

4.6.1. Geochronology

[37] Three samples from each of three flows from the Ashishik section were step heated in two steps. Ages from all samples were within 2-sigma of one another, implying that there is no discernable difference in ages from the top to the bottom of the section. The first steps exhibited low Ca/K ratios and somewhat lower and more imprecise ages. The second fractions had Ca/K ratios >10 and greater precision. On the basis of our experience with the New Jersey Creek samples, it is probable that the heating schedule was not sufficient to detect the presence of excess argon. There was not sufficient isotopic variation to allow for determination of a precise isochron age (1568 ± 420 ka; Figure 5). The weighted average of the fusion steps from the 9 samples gives an age of 1869 ± 116 ka (Figure 5). However, as indicated by the isochron, there might be problems with excess argon with the samples; thus this age may be too old. On the basis of the inferred stratigraphic position and magnetic polarity (reversed), it is probable that the Ashishik section was deposited immediately after the Olduvai normal polarity interval, which spanned 1.77 Ma to 1.95 Ma [Cande and Kent, 1995].

Figure 5.

Age probability plot (with weighted mean age) and argon isotopic correlation diagram (with isochron age) for the fusion steps of nine 2-step analyses from Ashishik. Abbreviations as in Figure 4.

4.6.2. Paleomagnetism

[38] The results of the original study show that the uppermost flow was weakly magnetized and the individual samples gave scattered directions. This flow was not included in the original study. The remaining flows gave reversed directions after alternating field demagnetization. Progressive thermal demagnetization gave Z plots that show a very clear straight-line track toward the origin and generally show relatively sharp decreases in intensity between 400 and 600°C (Figure 2c). Line-fit analyses gave mean directions for each flow almost identical to those from the original study, but usually with larger alpha-95 confidence limits (Tables 6a6c, Figure 3e).

Table 6a. Ashishik (53.53°N, 191.9°E), ∼1.9 Maa
Original (1972) Results
Mean of flows13 198.1−81.520.86.317.1
All New Data With MAD < 5° and Deviation From Flow Mean < 35°
Mean of flows1312.31206.19−82.7717.4510.2118.72
Mean of all cores6561.31206.17−82.7817.364.3519.39
Table 6b. New Data as in Table 6a With alpha95 < 5° for Flow Directionsa
Directional Data
Flow mean, D,I65.65196.29−81.3114.4918.2119.5321.4225.214.2925.13
Flow mean VGPs65.00205.4066.504.9933.3033.5836.9344.894.2944.85
Table 6c. Data for Individual Flowsa

[39] For this data set, the divergence of individual samples from the flow mean was relaxed from 30° to 35°. The mean MAD angle for all samples included in the calculations is 0.99° and the mean number of measurements per line fit is 7.0. Seven out of the twelve flows were excluded because the within-flow alpha-95 was greater than 5° leaving five flows in the final data set. The flow means are distributed along a NNE-SSW line, and the mean of the final reversed directions, converted to its lower hemisphere equivalent, is displaced from the overall mean of all the flow sequences by about 30°.

[40] There is no indication in the demagnetization data that a significant secondary magnetization was present. This result has been included in the tabulated calculations, but not in the final analysis of the expected dispersion due to secular variation.

4.7. New Jersey Creek (NJC) (53°32′N, 191°55′E)

[41] This sampling site is located on the NE end of Umnak Island, on the flanks of Okmok caldera. These flows, together with those from the Ashishik locality, are part of the precaldera volcanic edifice and are collectively labeled the Ashishik Basalts by Byers [1959]. The New Jersey Creek flows range from 7m to 20m in thickness, display rubbly and apparently weathered layers between them, and are described as mafic phenocryst basalt [Byers, 1959]. In general the mafic phenocryst flows underlie, and may interfinger with, the aphyric and feldspathic flows typical of the Ashishik section; thus the two localities should be of about the same age, but their stratigraphic relationship is not clear. Nineteen flows were sampled over a stratigraphic thickness of about 150 meters with between 4 and 7 cores per flow. The whole sequence has a shallow dip to the north that has been interpreted as primary on the basis of its location on the north flank of the original volcanic edifice.

4.7.1. Geochronology

[42] A single paleomagnetic core end from each of 19 flows from New Jersey Creek was selected for dating, although one flow (flow 18) had two different samples selected to check for inner-flow variability. The samples were collected with flow 1 being at the base of the section and flow 19 being at the top. For most flows, multistep step-heating was performed, with 5 – 12 duplicate “runs” on each sample (see Auxiliary Material). From these runs, plateau ages were calculated for 18 of the 20 samples; the other two samples having insufficient radiogenic argon. The plateau ages from the individual runs were then averaged together for an overall flow mean age. For most step-heats, the first fraction showed high amounts of nonradiogenic argon and variable ages, Ca/K ratios, and Cl/K ratios and were not included in plateau calculations. Isochrons were constructed for the flows to look for evidence of excess argon that might bias the age to be old. Figure 6 illustrates this process for Flow 11. Six step-heat runs were performed and plateau ages were calculated for each run (Figure 6a). These were averaged together to get an overall flow age of 1822 ± 41 ka. The composite isochron (Figure 6b) shows no evidence of excess argon in this sample. A similar process was followed for the other flows, and the results are summarized in Table 7.

Figure 6.

(a) Age spectra from 6 runs of sample NJC903 from flow 11 from New Jersey Creek. Plateau ages (and 1 sigma error) are shown for each run. (b) Composite argon isotopic correlation diagram for all 6 runs from flow 11. The calculated isochron does not include first step fractions from the runs. Abbreviations as in Figure 4.

Table 7. Summary of 40Ar/39Ar Ages From New Jersey Creeka
FlowSampleNumber of Step-Heat RunsWeighted Average Plateau AgeIsochron Age, kaInitial 40Ar/36ArComments
  • a

    All ages reported at ±1 sigma. Number of step-heat runs is the analyses used in calculation of the weighted average plateau age (see text). Bold: Samples used in calculation of overall average age.

Flow 1NJC84281991 ± 241832 ± 90317 ± 9Two-step step heats, isochron of high temperature steps shows excess. Not used in age calculation
Flow 2NJC852121800 ± 471873 ± 63284 ± 4Concordance between plateau and isochron ages
Flow 3NJC857101948 ± 351998 ± 62293 ± 2Two-step step heats. Not used in age calculation
Flow 4NJC86061832 ± 139--No well constrained isochron
Flow 5NJC866none-- Low K, no significant ages
Flow 6NJC873111815 ± 551626 ± 67301 ± 2Individual runs do not show significant excess argon
Flow 7NJC877111752 ± 451617 ± 60301 ± 1Individual runs do not show significant excess argon
Flow 8NJC88191972 ± 701761 ± 93302 ± 1Individual runs do not show significant excess argon
Flow 9NJC88982058 ± 512046 ± 65303 ± 2Variable excess argon (see text) not used in age calculation
Flow 10NJC89351864 ± 331793 ± 40302 ± 2Individual runs do not show significant excess argon
Flow 11NJC90361822 ± 411855 ± 194295 ± 8Concordance between plateau and isochron ages
Flow 12NJC90761717 ± 861642 ± 42299 ± 3Concordance between plateau and isochron ages
Flow 13NJC91371785 ± 611715 ± 89298 ± 3Concordance between plateau and isochron ages
Flow 14NJC91761974 ± 511975 ± 67297 ± 3Concordance between plateau and isochron ages
Flow 15NJC925101837 ± 541732 ± 94303 ± 4Two-step step heats. Not used in age calculation
Flow 16NJC93351905 ± 361970 ± 37277 ± 3Concordance between plateau and isochron ages
Flow 17NJC937none   Low K, no significant ages
Flow 18NJC94061807 ± 441654 ± 143298 ± 7Poorly constrained isochron, but concordance between plateau and isochron ages
Flow 18NJC94892023 ± 222104 ± 53277 ± 5Two-step step heats. Not used in age calculation
Flow 19NJC95161933 ± 1032053 ± 175293 ± 4Concordance between plateau and isochron ages
Average  1846 ± 19  13 plateau ages, MSWD = 1.9

[43] For Flow 9, plateau ages are significantly older than those from other samples (Table 7). Isochron analyses of several runs indicate the presence of excess argon, biasing the age of this sample. For example, Figure 7a shows that run 5 has an apparent plateau age of 2194 ± 78 ka, which is significantly older than the isochron age of 1865 ± 83 ka from the same sample. This isochron age is more similar to the majority of flow ages from flows that do not indicate the presence of excess argon. On the basis of this observation, the plateau ages from Flow 9 are not considered reliable and not included in the calculation of an overall mean age.

Figure 7.

(a) Age spectrum and argon isotopic correlation diagram from run 5 from sample NJC889 (flow 9) from New Jersey Creek. The apparent plateau age and isochron age are significantly different due to the presence of excess argon. (b) Composite argon isotopic correlation diagram for eight 2-step runs from sample NJC842 (flow 1). Isochron is constructed using only the second (fusion) step in each run. Abbreviations as in Figure 4.

[44] For Flows 1, 3, 15 and one sample from Flow 18, we performed 2-step runs with the assumption that there was not significant excess argon in the samples, and that the fusion step would produce reliable ages akin to the plateaus seen in more detailed runs. However, our experience with Flow 9 shows that this might not be the case, and in fact for flow 1, the fusion steps have a weighted mean age of 1991 ± 21 ka, but an isochron age of 1832 ± 90 ka (Figure 6b), again indicative of excess argon in this sample. Because a 2-step run is, in general, unable to conclusively rule out excess argon contamination, we do not include the results of these flows in the overall mean age calculation.

[45] Although there was significant variation in the ages, Ca/K and Cl/K from the flows, there was no systematic decrease in the determined ages from the bottom to the top of the section. Thus the 19 flows were probably emplaced over a short period of time. We feel that the weighted mean of the plateau ages from 13 of the 19 flows of 1846 ± 19 ka reflects the age of magnetization (Table 7). This age is consistent with the section being deposited during the Olduvai normal polarity interval.

4.7.2. Paleomagnetism

[46] The original study [Bingham and Stone, 1972] was based on blanket demagnetization with an alternating field of 47 mT. The distribution of flow directions was roughly circular except for the top 5 flows that gave sequential directions more or less following a path around the main grouping. For this study thermal demagnetization was used with the Z plots showing variable behavior, but commonly giving a straight-line track toward the origin between 525° and 600°C, plus a partial blocking temperature at about 350°C (Figure 2d).

[47] Following progressive thermal demagnetization five individual samples were rejected because they were too divergent from their flow means. The mean MAD angle for all samples included in the calculations is 1.1° and the mean number of measurements per line fit is 6.9. The whole data set (Tables 8a8c) shows a similar distribution to that seen in the original data set, but after the final selection only the top two flows (15 and 16) diverge from the group (Figure 3f).

Table 8a. New Jersey Creek (53.53°N, 191.92°N), ∼1.9 Maa
Original (1972) Results
Mean of flows19 350.564.71223.17.1
All New Data With MAD < 5° and Deviation From Flow Mean < 30°
Mean of flows1817.91345.1865.63195.712.485.63
Mean of all cores9896.73344.5865.4176.301.659.24
Table 8b. New Data as in Table 8a With alpha95 < 5° for Flow Directionsa
Directional Data
Flow mean, D,I109.96342.5066.90246.983.104.895.
Flow mean VGPs109.9293.8078.60110.654.607.317.7214.306.9513.98
Table 8c. Data for Individual Flowsa

[48] The remainder of the flows have directions that are tightly grouped, and show no clear sequential directions. The mean VGPs and directions are close to, but do not coincide with the GAD field.

5. Paleointensity

[49] Samples from two flows from each of the localities were submitted to University of Florida for paleointensity determinations. All but three of these samples failed to give reliable intensity estimates. Two of the three measurements were from the 2 Ma reversed/transitional sequence at the ASH locality and the third one from the youngest flow sequence (about 50 kyr) at CRC. These results are given in Table 9 for completeness.

Table 9. Successful Paleointensity Results With Quality Parameters as Defined by Coe et al. [1978]a
Location FlowNT1T2σbσb/bfgqF, μTVADM Am2
CRC-393005750.0210.0170.9560.77344.54363.99.73E + 22
ASH-1074755750.0180.0430.8190.80215.43721.63.29E + 22
ASH-984005750.0210.0390.7930.79116.29227.04.11E + 22

6. Analysis

[50] For the six Aleutian localities sampled, the age range is from about 50 ka to 2 Ma, the latitude spread is between 51.9°N and 53.97°N. The between-flow dispersion (sbp) of the locality-mean magnetization directions varies between 5.66° and 9.51°, excluding Ashishik (ASH), which has a dispersion of 25.13°, and probably includes part of a polarity transition. The dispersion for the equivalent VGPs (Sbp) varies between 8.38° and 14.86° (ASH 44.85) (Table 10).

Table 10. Summary of the Aleutian Flow Sequencesa
  • a

    The top part of the table shows results for the magnetic directions (D&I), and the bottom part shows results for Virtual Geomagnetic Poles (VGP). The symbols are the same as those described for Table 2.

Crater Creek1110.86356.268.771.495.409.715.229.51
Round Head87.987.571.7379.822.855.974.395.72
Driftwood Bay1514.851.468.495.783.908.463.268.35
Ashishik (REV)65.65196.3−81.314.4918.2125.214.2925.13
New Jersey creek109.96342.566.9246.983.
Crater Creek1110.66131.8087.8029.598.5015.147.6314.86
Round Head87.96237.384.4174.024.209.616.979.22
Driftwood Bay1514.65340.588.240.476.1012.906.2512.62
Ashishik (REV)65.00205.466.54.9933.3044.894.2944.85
New Jersey creek109.9293.878.6110.654.6014.306.9513.98

[51] The Ashishik data have not been included in the final calculations of dispersion. None of the localities have definitive indications of the time span represented, and only the Driftwood Bay flow sequence gives a significant spread of 40Ar/39Ar ages from within the flow sequence. The face-value time span represented by these flows is about 480 kyr. The dispersion of the observed magnetic directions for Driftwood Bay (8.35) is compatible with the expected dispersion predicted by secular variation models [Johnson and Constable, 1996].

[52] In trying to assess the time represented by the flow sequences it should be noted that the geocentric axial dipole (GAD) field direction and the axial field pole both lie within the alpha-95 confidence limits for three of the localities (CRC, KAN, DFB). For one of the other two localities (RDH) the alpha-95 circle of confidence for the mean field direction is within 2° of the GAD direction, and the other (NJC) is within 5° (Table 10, Figure 3).

[53] The closeness of all the locality means to the GAD field can be interpreted to mean that the time averaging was sufficient to remove the majority of the short period components of the secular variation field. Counter to this argument is the observation that the dispersion of the magnetic field directions about their locality means (stv) for three of the five localities (RDH, KAN, NJC) is less than 5.2° for the directions and less than 7.8° for the VGPs (Tables 3a3c, 4a4c, and 7). These values are considerably lower than secular variation models predict. The dispersion for the other two localities is still low, but closer to predicted values (Table 9, Figures 8 and 9). If the low values are because the flow sequences represent too short a time interval, then it is curious that the mean field is so close to the GAD field, especially since the localities are separated in time by at least tens of thousands of years and have a total spread of about two million years.

Figure 8.

The top panel shows the dispersion versus latitude from the Johnson and Constable [1996] model for vector directions, combined with data from the last 5 Myr. The bottom panels show the dispersion of the vector directions from Aleutian sites (circles) and from Canadian sites (squares). The circle and square show the combined data set. The error bars represent the standard deviations of the data sets.

Figure 9.

This figure is the same as Figure 8, but for VGPs instead of vector directions.

[54] All of the localities used in this study are within about 2° of latitude of each other, but spread over about 10° of longitude. The similarity of the latitudes allows the data to be combined to obtain a more robust estimate of the paleosecular variation.

[55] If the low dispersions within localities are interpreted as representing short time spans for the sampled sections, then the total dispersion of the five locality means might be expected to approximate the secular variation. However, the dispersion (st) of the five locality means is 4.05° for the declination and inclination and 6.42° for the VGPs (Table 11). These values are very low with respect to the Johnson and Constable [1996] model. An alternative approach is to consider each flow-mean direction as representing a snap shot of the geomagnetic field. Since the number of flows is relatively large (N = 49), this should smooth out the effect of occasional flows being too close to one another in time to record changes in the field. The results of both analyses are shown in Table 11.

Table 11. Mean Magnetic Directions and VGPs for the Aleutian, Canadian, and Combined Aleutian and Canadian Data Setsa
  • a

    The top part of the table shows results for the magnetic directions (D&I), and the bottom part shows results for Virtual Geomagnetic Poles (VGP). Each part is divided into data for combined localities (flow sequences) and combined individual flows. The symbols are the same as those described for Table 2.

Aleutian localities54.99356.1068.50398.463.804.375.124.05
Aleutian flows4948.56356.3268.65108.231.977.924.597.70
Canadian localities98.90358.4071.2079.345.809.727.629.02
Canadian flows4342.29355.8270.6259.492.8510.834.9210.57
Combined localities1413.89357.5070.20113.373.707.566.477.10
Combined flows9290.84356.1069.5678.241.689.344.749.17
Aleutian localities54.9791.3087.80157.796.106.907.846.43
Aleutian flows4947.89104.2387.8943.433.1212.528.2712.07
Canadian localities98.76233.7084.2032.989.1015.4112.3614.25
Canadian flows4341.20212.7884.8123.334.6117.694.8617.50
Combined localities1413.70224.8086.9043.916.1012.689.9112.07
Combined flows9288.97185.0187.6929.982.7415.078.0114.78

[56] It is of interest to note that there is no evidence of far-sidedness or right-handedness in the data.

[57] We have expanded on this approach by incorporating the results from the paleosecular variation study published by Mejia et al. [2002]. This study is based on flow sequences and individual flows from localities in western Canada. These localities are all east of those used in the present study, but at similar latitudes. The latitude range for the Canadian data is 49.83°N to 51.95°N which overlaps with the Aleutian latitudes of 51.90°N to 53.97°N giving an overall latitude spread of 4.14°. The two data sets combined thus represent a band of latitude at 52°N ± 2.1° and just over 57° of longitude in length (182.90°E to 240.25°E). Table 12 shows the dispersion of both directions and VGPs for the nine groups of flows. For these analyses flows with α-95 values less than 5° were selected and the dispersion statistics calculated using the same methods as were used for the Aleutian data. The Mejia et al. [2002] data set covers 50 ka to 760 ka, which overlaps well with the bulk of the Aleutian data.

Table 12. Results Recalculated From Selected Data Published by Mejia et al. [2002]a
  • a

    The top part of the table shows results for the magnetic directions (D&I), and the bottom part shows results for Virtual Geomagnetic Poles (VGP). The first column shows their 2-letter code plus the assigned age in ka. The following columns give the following: N, the number of flows sampled; R, resultant of N unit vectors; D and I, declination and inclination of the resultant vector (top part); VGP latitude and longitude (bottom part); κ (kappa) and α95 (alpha 95) following Fisher [1953]; circular standard deviation delta; stp and Stp, the total dispersion; swp and Swp, the within-flow dispersion; sbp and Sbp, the between-flow dispersion.

CL 5032.97351.4076.4057.5016.4014.674.0114.62
ST 15098.9115.1066.7093.725.3010.554.2810.42
CW 24043.9116.9075.8032.9016.3017.385.7217.28
CW 30022.0014.1070.60317.0614.208.775.248.55
CW 35076.95333.9068.30132.485.3012.475.7812.30
ST 40021.99353.1054.90176.4119.0020.826.4120.69
CW 547109.97342.3067.80259.273.008.594.718.46
CW 56133.0018.9074.90961.454.0010.794.7210.66
KA 76033.00349.0078.70541.555.3013.393.4313.35
GL 5032.89223.6074.7018.0229.926.696.9526.60
ST 15098.83328.6080.6046.247.7015.656.4715.45
CW 24043.75266.3073.9012.0027.7029.889.6629.71
CW 30021.99304.5081.00112.1223.8014.848.3514.50
CW 35076.88160.9073.8051.878.520.828.9420.59
ST 40021.9973.9073.50128.5822.2024.327.8124.15
CW 547109.91153.7078.90101.064.8014.227.1014.04
CW 56132.99278.6075.80311.197.0017.998.1017.76
KA 76032.99228.1070.60160.559.8024.556.4424.46

[58] The Aleutian and Canadian flow-mean data have also been combined (Table 11) and give a between-flow dispersion for these 92 magnetic directions of 9.17°, and 14.78° for the VGPs. Both of these values fall within the error bars of most paleosecular variation models. The flow-mean data from the Canadian and Aleutian data sets separately show that the Aleutians have recorded lower values and the Canadian data values close to the mean values of the secular variation model. The difference between the two data sets is clearly seen in the distribution of the VGPs (Figure 10). The Aleutian VGPs are markedly closer to the GAD pole and have a much tighter distribution. Comparing the Aleutian and Canadian data sets shows that for individual localities with 5 or more flows (3 from Canada, 5 from the Aleutians) the dispersion for the Canadian data was generally large with respect to the Aleutian data, which could reflect differences in the length of time recorded by the flow sequences (Figures 8 and 9). However, comparing the dispersion of all the locality means also shows a larger dispersion for the Canadian data, as does a comparison of the individual flow means (Table 11).

Figure 10.

Aleutian and Canadian Virtual Geomagnetic Poles and the circles representing dispersion (Sbp) about their means.

7. Conclusions

[59] The remeasurement of samples collected in 1968 and succeeding years gives essentially the same results as the original study, but now more believable. The samples come from six sequences of lava flows from localities along the Aleutian volcanic arc spanning about 2 Myr. The data used for analysis of the magnetic field changes through time resulted from multilevel thermal demagnetization, and the results selected using strict criteria which rejected samples with a MAD angle greater than 5°, and rejected flows with alpha-95 confidence circles greater than 5°. The data showed that for three localities the dispersion was lower than would be expected using the Johnson and Constable [1996] model, the other two localities gave dispersions that are within the error bars of the model. The sixth site is interpreted as including a partial polarity transition and was thus rejected. Because all the localities are at about the same latitude (within 2°) it is reasonable to combine them. The combined flow-mean data from the Aleutian localities gives a dispersion which is very low with respect to the Johnson and Constable [1996] model. This can be interpreted in terms of anomalous secular variation in the North Pacific, or possibly by having sampled intervals of time that were too short. The latter explanation would have to address the probability of sampling a near GAD field in at least three locations well separated in time. By combining the flow-mean data from the Aleutian and Canadian localities we have obtained estimates of the dispersion of both the direction of the recorded geomagnetic field and the equivalent VGPs representing latitudes within 2° of 50°N and stretching from 182°E to 241°N. This composite data set gives dispersion values that are very close to those predicted by the Johnson and Constable model, but the disparity in the number of localities involved may be masking the differences between the two data sets (Table 11, Figures 8 and 9).


[60] The authors would like to acknowledge D. K. Bingham, B. N. Chaterjee, and H. S. Nielson, who braved the Aleutian weather to collect samples, Reeve-Aleutian Airlines for their generous help in getting us from island to island, K. M. Creer for suggesting the initial paleosecular variation study, Neil Opdyke and Victoria Mejia for the paleointensity studies, and the State of Alaska and the National Science Foundation (grants EAR-9870309 and EAR-9943700) for financial support.